From pgsql-performance-owner@postgresql.org Mon Aug 1 01:05:06 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3260352B7F for ; Mon, 1 Aug 2005 01:05:04 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 63398-06 for ; Mon, 1 Aug 2005 04:04:55 +0000 (GMT) Received: from outbound.mailhop.org (outbound.mailhop.org [63.208.196.171]) by svr1.postgresql.org (Postfix) with ESMTP id 812C052AC2 for ; Mon, 1 Aug 2005 01:04:53 -0300 (ADT) Received: from ppp-66-143-176-222.dsl.austtx.swbell.net ([66.143.176.222] helo=mail.herk.net) by outbound.mailhop.org with esmtpsa (TLSv1:DES-CBC3-SHA:168) (Exim 4.51) id 1DzRXn-000Bll-SZ for pgsql-performance@postgresql.org; Mon, 01 Aug 2005 00:04:52 -0400 Received: from localhost (localhost [127.0.0.1]) (uid 500) by mail.herk.net with local; Sun, 31 Jul 2005 23:04:49 -0500 id 00114186.42ED9F61.00001AB0 X-Mail-Handler: MailHop Outbound by DynDNS.org X-Originating-IP: 66.143.176.222 X-Report-Abuse-To: abuse@dyndns.org (see http://www.mailhop.org/outbound/abuse.html for abuse reporting information) X-MHO-User: parkerm Date: Sun, 31 Jul 2005 23:04:49 -0500 From: Michael Parker To: pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 Message-ID: <20050801040443.GA6527@mail.herk.net> References: <42E80C28.2040608@aptalaska.net> <1122532023.25574.1.camel@localhost.localdomain> <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="=_mail-6832-1122869089-0001-2" Content-Disposition: inline In-Reply-To: <27897.1122748133@sss.pgh.pa.us> User-Agent: Mutt/1.5.1i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 tagged_above=0 required=5 tests=FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/1 X-Sequence-Number: 13748 This is a MIME-formatted message. If you see this text it means that your E-mail software does not support MIME-formatted messages. --=_mail-6832-1122869089-0001-2 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable Hi All, As a SpamAssassin developer, who by my own admission has real problem getting PostgreSQL to work well, I must thank everyone for their feedback on this issue. Believe me when I say what is in the tree now is a far cry from what used to be there, orders of magnitude faster for sure. I think there are several good ideas that have come out of this thread and I've set about attempting to implement them. Here is a version of the stored procedure, based in large part by the one written by Tom Lane, that accepts and array of tokens and loops over them to either update or insert them into the database (I'm not including the greatest_int/least_int procedures but you've all seen them before): CREATE OR REPLACE FUNCTION put_tokens(inuserid INTEGER, intokenary BYTEA[], inspam_count INTEGER, inham_count INTEGER, inatime INTEGER) RETURNS VOID AS '=20 DECLARE _token BYTEA; new_tokens INTEGER :=3D 0; BEGIN for i in array_lower(intokenary, 1) .. array_upper(intokenary, 1) LOOP _token :=3D intokenary[i]; UPDATE bayes_token SET spam_count =3D greatest_int(spam_count + inspam_count, 0), ham_count =3D greatest_int(ham_count + inham_count, 0), atime =3D greatest_int(atime, inatime) WHERE id =3D inuserid=20 AND token =3D _token;=20 IF NOT FOUND THEN=20 -- we do not insert negative counts, just return true IF NOT (inspam_count < 0 OR inham_count < 0) THEN INSERT INTO bayes_token (id, token, spam_count, ham_count, atime)=20 VALUES (inuserid, _token, inspam_count, inham_count, inatime);=20 IF FOUND THEN new_tokens :=3D new_tokens + 1; END IF; END IF; END IF; END LOOP; UPDATE bayes_vars SET token_count =3D token_count + new_tokens, newest_token_age =3D greatest_int(newest_token_age, inatime), oldest_token_age =3D least_int(oldest_token_age, inatime) WHERE id =3D inuserid; RETURN; END;=20 ' LANGUAGE 'plpgsql';=20 This version is about 32x faster than the old version, with the default fsync value and autovacuum running in the background. The next hurdle, and I've just posted to the DBD::Pg list, is escaping/quoting the token strings. They are true binary strings, substrings of SHA1 hashes, I don't think the original data set provided puts them in the right context. They have proved to be tricky. I'm unable to call the stored procedure from perl because I keep getting a malformed array litteral error. Here is some example code that shows the issue: #!/usr/bin/perl -w # from a new db, do this first # INSERT INTO bayes_vars VALUES (1,'nobody',0,0,0,0,0,0,2147483647,0); use strict; use DBI; use DBD::Pg qw(:pg_types); use Digest::SHA1 qw(sha1); my $dbh =3D DBI->connect("DBI:Pg:dbname=3Dspamassassin","postgres") || die; my @dataary; # Input is just a list of words (ie /usr/share/dict/words) stop after 150 while(<>) { chomp; push(@dataary, substr(sha1($_), -5)); # to see it work with normal string comment out above and uncomment below # push(@dataary, $_); last if scalar(@dataary) >=3D 150; } my $datastring =3D join(",", map { '"' . bytea_esc($_) . '"' } @dataary); my $sql =3D "select put_tokens(1, '{$datastring}', 1, 1, 10000)"; my $sth =3D $dbh->prepare($sql); my $rc =3D $sth->execute(); unless ($rc) { print "Error: " . $dbh->errstr() . "\n"; } $sth->finish(); sub bytea_esc { my ($str) =3D @_; my $buf =3D ""; foreach my $char (split(//,$str)) { if (ord($char) =3D=3D 0) { $buf .=3D "\\\\000"; } elsif (ord($char) =3D=3D 39) { $buf .=3D "\\\\047"; } elsif (ord($char) =3D=3D 92) { $buf .=3D "\\\\134"; } else { $buf .=3D $char; } } return $buf; } Any ideas? or thoughts on the revised procedure? I'd greatly appriciate them. Sorry for the length, but hopefully it give a good enough example. Thanks Michael Parker --=_mail-6832-1122869089-0001-2 Content-Type: application/pgp-signature Content-Transfer-Encoding: 7bit Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.0.7 (GNU/Linux) iD8DBQFC7Z9bG4km+uS4gOIRAklxAJ918GyBQKOnrblMY24M3GIsiVXjWQCfafn5 K2WAqkOXTbmL9oLtY5UkchE= =FGrI -----END PGP SIGNATURE----- --=_mail-6832-1122869089-0001-2-- From pgsql-performance-owner@postgresql.org Mon Aug 1 01:42:39 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 315BC52A92 for ; Mon, 1 Aug 2005 01:42:36 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 96541-07 for ; Mon, 1 Aug 2005 04:42:34 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id A6EE052896 for ; Mon, 1 Aug 2005 01:42:33 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j714gUpc024579; Mon, 1 Aug 2005 00:42:30 -0400 (EDT) To: Michael Parker Cc: pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 In-reply-to: <20050801040443.GA6527@mail.herk.net> References: <42E80C28.2040608@aptalaska.net> <1122532023.25574.1.camel@localhost.localdomain> <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <20050801040443.GA6527@mail.herk.net> Comments: In-reply-to Michael Parker message dated "Sun, 31 Jul 2005 23:04:49 -0500" Date: Mon, 01 Aug 2005 00:42:30 -0400 Message-ID: <24578.1122871350@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/2 X-Sequence-Number: 13749 Michael Parker writes: > The next hurdle, and I've just posted to the DBD::Pg list, is > escaping/quoting the token strings. If you're trying to write a bytea[] literal, I think the most reliable way to write the individual bytes is \\\\nnn where nnn is *octal*. The idea here is: * string literal parser takes off one level of backslashing, leaving \\nnn * array input parser takes off another level, leaving \nnn * bytea input parser knows about backslashed octal values Note it has to be 3 octal digits every time, no abbreviations. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 1 01:54:48 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7186752812 for ; Mon, 1 Aug 2005 01:54:47 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 98576-09 for ; Mon, 1 Aug 2005 04:54:38 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id A937D52AFC for ; Mon, 1 Aug 2005 01:54:37 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j714rMLp024658; Mon, 1 Aug 2005 00:53:22 -0400 (EDT) To: Michael Parker Cc: pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 In-reply-to: <20050801040443.GA6527@mail.herk.net> References: <42E80C28.2040608@aptalaska.net> <1122532023.25574.1.camel@localhost.localdomain> <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <20050801040443.GA6527@mail.herk.net> Comments: In-reply-to Michael Parker message dated "Sun, 31 Jul 2005 23:04:49 -0500" Date: Mon, 01 Aug 2005 00:53:22 -0400 Message-ID: <24657.1122872002@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/3 X-Sequence-Number: 13750 Michael Parker writes: > sub bytea_esc { > my ($str) = @_; > my $buf = ""; > foreach my $char (split(//,$str)) { > if (ord($char) == 0) { $buf .= "\\\\000"; } > elsif (ord($char) == 39) { $buf .= "\\\\047"; } > elsif (ord($char) == 92) { $buf .= "\\\\134"; } > else { $buf .= $char; } > } > return $buf; > } Oh, I see the problem: you forgot to convert " to a backslash sequence. It would probably also be wise to convert anything >= 128 to a backslash sequence, so as to avoid any possible problems with multibyte character encodings. You wouldn't see this issue in a SQL_ASCII database, but I suspect it would rise up to bite you with other encoding settings. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 1 13:34:19 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C2A2D52B53 for ; Mon, 1 Aug 2005 13:34:18 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 47285-10 for ; Mon, 1 Aug 2005 16:34:07 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 47D8C52B3D for ; Mon, 1 Aug 2005 13:34:06 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j71GWYsQ010469; Mon, 1 Aug 2005 12:32:34 -0400 (EDT) To: "Luke Lonergan" Cc: "Bruce Momjian" , "Mark Wong" , "Joshua D. Drake" , "Patrick Welche" , pgsql-performance@postgresql.org, maryedie@osdl.org Subject: Re: [PATCHES] COPY FROM performance improvements In-reply-to: References: Comments: In-reply-to "Luke Lonergan" message dated "Fri, 29 Jul 2005 10:54:08 -0700" Date: Mon, 01 Aug 2005 12:32:34 -0400 Message-ID: <10468.1122913954@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/4 X-Sequence-Number: 13751 "Luke Lonergan" writes: > On 7/29/05 5:37 AM, "Bruce Momjian" wrote: >> Where is the most recent version of the COPY patch? > I've attached it here, sorry to the list owner for the patch inclusion / > off-topic. This patch appears to reverse out the most recent committed changes in copy.c. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 1 14:09:23 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id F0CC052B6C for ; Mon, 1 Aug 2005 14:09:21 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 57555-10 for ; Mon, 1 Aug 2005 17:09:19 +0000 (GMT) Received: from larry.aptalaska.net (larry.aptalaska.net [64.186.96.3]) by svr1.postgresql.org (Postfix) with ESMTP id 61CF752B53 for ; Mon, 1 Aug 2005 14:09:18 -0300 (ADT) Received: from [192.168.98.9] (rdbck-static-287.palmer.mtaonline.net [64.4.232.33]) (authenticated bits=0) by larry.aptalaska.net (8.13.4/8.13.4) with ESMTP id j71HMx7V014924 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Mon, 1 Aug 2005 09:23:00 -0800 Message-ID: <42EE5720.2020306@aptalaska.net> Date: Mon, 01 Aug 2005 09:08:48 -0800 From: Matthew Schumacher User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: Michael Parker , pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 References: <42E80C28.2040608@aptalaska.net> <1122532023.25574.1.camel@localhost.localdomain> <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <20050801040443.GA6527@mail.herk.net> <24657.1122872002@sss.pgh.pa.us> In-Reply-To: <24657.1122872002@sss.pgh.pa.us> X-Enigmail-Version: 0.91.0.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit X-Spam-Bayes: Learn: unavailable X-Scanned-By: MIMEDefang 2.52 on 64.186.96.3 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.008 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/5 X-Sequence-Number: 13752 Tom Lane wrote: > Michael Parker writes: > >>sub bytea_esc { >> my ($str) = @_; >> my $buf = ""; >> foreach my $char (split(//,$str)) { >> if (ord($char) == 0) { $buf .= "\\\\000"; } >> elsif (ord($char) == 39) { $buf .= "\\\\047"; } >> elsif (ord($char) == 92) { $buf .= "\\\\134"; } >> else { $buf .= $char; } >> } >> return $buf; >>} > > > Oh, I see the problem: you forgot to convert " to a backslash sequence. > > It would probably also be wise to convert anything >= 128 to a backslash > sequence, so as to avoid any possible problems with multibyte character > encodings. You wouldn't see this issue in a SQL_ASCII database, but I > suspect it would rise up to bite you with other encoding settings. > > regards, tom lane Here is some code that applies Toms Suggestions: 38c39,41 < if (ord($char) == 0) { $buf .= "\\\\000"; } --- > if (ord($char) >= 128) { $buf .= "\\\\" . sprintf ("%lo", ord($char)); } > elsif (ord($char) == 0) { $buf .= "\\\\000"; } > elsif (ord($char) == 34) { $buf .= "\\\\042"; } But this begs the question, why not escape everything? schu From pgsql-performance-owner@postgresql.org Mon Aug 1 16:48:40 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6D9AD52B81 for ; Mon, 1 Aug 2005 16:48:39 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 00978-05 for ; Mon, 1 Aug 2005 19:48:34 +0000 (GMT) Received: from mail.Mi8.com (nycgw05.mi8.com [63.240.6.46]) by svr1.postgresql.org (Postfix) with ESMTP id 667F8529B6 for ; Mon, 1 Aug 2005 16:48:33 -0300 (ADT) Received: from 172.16.1.112 by mail.Mi8.com with ESMTP (- GW05 Welcome to Mi8 Corporation www.Mi8.com); Mon, 01 Aug 2005 15:48:27 -0400 X-Server-Uuid: E0C866E6-C6CD-48B5-AE61-E57E73CF3CC7 Received: from MI8NYCMAIL06.Mi8.com ([172.16.1.175]) by D01SMTP02.Mi8.com with Microsoft SMTPSVC(6.0.3790.211); Mon, 1 Aug 2005 15:48:26 -0400 Received: from 141.156.39.156 ([141.156.39.156]) by MI8NYCMAIL06.Mi8.com ([172.16.1.219]) via Exchange Front-End Server mi8owa.mi8.com ( [172.16.1.104]) with Microsoft Exchange Server HTTP-DAV ; Mon, 1 Aug 2005 15:48:26 -0500 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Mon, 01 Aug 2005 15:48:35 -0400 Subject: Re: [PATCHES] COPY FROM performance improvements From: "Alon Goldshuv" To: pgsql-performance@postgresql.org Cc: "Luke Lonergan" Message-ID: In-Reply-To: MIME-Version: 1.0 X-OriginalArrivalTime: 01 Aug 2005 19:48:26.0998 (UTC) FILETIME=[FBFDE960:01C596D1] X-WSS-ID: 6EF0A30126S57177-01-01 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.581 tagged_above=0 required=5 tests=FORGED_RCVD_HELO, RCVD_NUMERIC_HELO X-Spam-Level: * X-Archive-Number: 200508/6 X-Sequence-Number: 13753 Tom, >> I've attached it here, sorry to the list owner for the patch inclusion / >> off-topic. > > This patch appears to reverse out the most recent committed changes in > copy.c. Which changes do you refer to? I thought I accommodated all the recent changes (I recall some changes to the tupletable/tupleslot interface, HEADER in cvs, and hex escapes and maybe one or 2 more). What did I miss? Thanks. Alon. From pgsql-performance-owner@postgresql.org Mon Aug 1 17:04:29 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6F09952C26 for ; Mon, 1 Aug 2005 17:04:27 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 17736-01 for ; Mon, 1 Aug 2005 20:04:22 +0000 (GMT) Received: from larry.aptalaska.net (larry.aptalaska.net [64.186.96.3]) by svr1.postgresql.org (Postfix) with ESMTP id ECD1252C3C for ; Mon, 1 Aug 2005 17:04:20 -0300 (ADT) Received: from [192.168.98.9] (rdbck-static-287.palmer.mtaonline.net [64.4.232.33]) (authenticated bits=0) by larry.aptalaska.net (8.13.4/8.13.4) with ESMTP id j71KIWF6025299 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Mon, 1 Aug 2005 12:18:32 -0800 Message-ID: <42EE8043.1050207@aptalaska.net> Date: Mon, 01 Aug 2005 12:04:19 -0800 From: Matthew Schumacher User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 References: <42E80C28.2040608@aptalaska.net> <1122532023.25574.1.camel@localhost.localdomain> <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> In-Reply-To: <27897.1122748133@sss.pgh.pa.us> X-Enigmail-Version: 0.91.0.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit X-Spam-Bayes: Learn: unavailable X-Scanned-By: MIMEDefang 2.52 on 64.186.96.3 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.007 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/7 X-Sequence-Number: 13754 Tom Lane wrote: > > Revised insertion procedure: > > > CREATE or replace FUNCTION put_tokens (_id INTEGER, > _tokens BYTEA[], > _spam_count INTEGER, > _ham_count INTEGER, > _atime INTEGER) > RETURNS VOID AS > $$ > declare _token bytea; > new_tokens integer := 0; > BEGIN > for i in array_lower(_tokens,1) .. array_upper(_tokens,1) > LOOP > _token := _tokens[i]; > UPDATE bayes_token > SET spam_count = spam_count + _spam_count, > ham_count = ham_count + _ham_count, > atime = _atime > WHERE id = _id > AND token = _token; > > IF not found THEN > INSERT INTO bayes_token VALUES (_id, _token, _spam_count, > _ham_count, _atime); > new_tokens := new_tokens + 1; > END IF; > END LOOP; > if new_tokens > 0 THEN > UPDATE bayes_vars SET token_count = token_count + new_tokens > WHERE id = _id; > IF NOT FOUND THEN > RAISE EXCEPTION 'unable to update token_count in bayes_vars'; > END IF; > END IF; > RETURN; > END; > $$ > LANGUAGE plpgsql; > Tom, thanks for all your help on this, I think we are fairly close to having this done in a proc. The biggest problem we are running into now is that the data gets inserted as an int. Even though your proc defines _token as byeta, I get numbers in the table. For example: select put_tokens2(1, '{"\\246\\323\\061\\332\\277"}', 1, 1, 10000); produces this: id | token | spam_count | ham_count | atime ----+-----------------+------------+-----------+------- 1 | 246323061332277 | 1 | 1 | 10000 I'm not sure why this is happening, perhaps the problem is obvious to you? Thanks, schu From pgsql-performance-owner@postgresql.org Mon Aug 1 18:18:31 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0E46352C00 for ; Mon, 1 Aug 2005 18:18:30 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 68863-01 for ; Mon, 1 Aug 2005 21:18:25 +0000 (GMT) Received: from boutiquenumerique.com (boutiquenumerique.com [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 271DF52B0E for ; Mon, 1 Aug 2005 18:18:23 -0300 (ADT) Received: (qmail 26635 invoked from network); 1 Aug 2005 23:18:32 +0200 Received: from unknown (HELO localhost) (boutiquenumerique-lists@192.168.0.4) by boutiquenumerique.com with SMTP; 1 Aug 2005 23:18:32 +0200 To: "Matthew Schumacher" , "Tom Lane" Cc: pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 References: <42E80C28.2040608@aptalaska.net> <1122532023.25574.1.camel@localhost.localdomain> <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> Message-ID: Date: Mon, 01 Aug 2005 23:18:23 +0200 From: PFC Organization: =?iso-8859-15?Q?La_Boutique_Num=E9rique?= Content-Type: text/plain; format=flowed; delsp=yes; charset=iso-8859-15 MIME-Version: 1.0 Content-Transfer-Encoding: 8bit In-Reply-To: <42EE8043.1050207@aptalaska.net> User-Agent: Opera M2/8.0 (Linux, build 1095) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/8 X-Sequence-Number: 13755 > select put_tokens2(1, '{"\\246\\323\\061\\332\\277"}', 1, 1, 10000); Try adding more backslashes until it works (seems that you need \\\\ or something). Don't DBI convert the language types to postgres quoted forms on its own ? From pgsql-performance-owner@postgresql.org Mon Aug 1 18:28:35 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3454752E31 for ; Mon, 1 Aug 2005 18:28:33 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 63606-07 for ; Mon, 1 Aug 2005 21:28:30 +0000 (GMT) Received: from larry.aptalaska.net (larry.aptalaska.net [64.186.96.3]) by svr1.postgresql.org (Postfix) with ESMTP id 3EAC252DF6 for ; Mon, 1 Aug 2005 18:28:29 -0300 (ADT) Received: from [192.168.98.9] (rdbck-static-287.palmer.mtaonline.net [64.4.232.33]) (authenticated bits=0) by larry.aptalaska.net (8.13.4/8.13.4) with ESMTP id j71LgbQr011048 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Mon, 1 Aug 2005 13:42:38 -0800 Message-ID: <42EE93F8.5010003@aptalaska.net> Date: Mon, 01 Aug 2005 13:28:24 -0800 From: Matthew Schumacher User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: PFC Cc: Tom Lane , pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 References: <42E80C28.2040608@aptalaska.net> <1122532023.25574.1.camel@localhost.localdomain> <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> In-Reply-To: X-Enigmail-Version: 0.91.0.0 Content-Type: text/plain; charset=ISO-8859-15 Content-Transfer-Encoding: 7bit X-Spam-Bayes: Learn: ham X-Scanned-By: MIMEDefang 2.52 on 64.186.96.3 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.007 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/9 X-Sequence-Number: 13756 PFC wrote: > > >> select put_tokens2(1, '{"\\246\\323\\061\\332\\277"}', 1, 1, 10000); > > > Try adding more backslashes until it works (seems that you need \\\\ > or something). > Don't DBI convert the language types to postgres quoted forms on its > own ? > Your right.... I am finding that the proc is not the problem as I suspected, it works correctly when I am not calling it from perl, something isn't escaped correctly.... schu From pgsql-performance-owner@postgresql.org Mon Aug 1 18:50:02 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2589952BA2 for ; Mon, 1 Aug 2005 18:49:36 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 79715-06 for ; Mon, 1 Aug 2005 21:49:30 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [67.100.216.10]) by svr1.postgresql.org (Postfix) with ESMTP id 4DC0452BE0 for ; Mon, 1 Aug 2005 18:49:29 -0300 (ADT) Received: by flake.decibel.org (Postfix, from userid 1001) id C41D015233; Mon, 1 Aug 2005 16:49:31 -0500 (CDT) Date: Mon, 1 Aug 2005 16:49:31 -0500 From: "Jim C. Nasby" To: Matthew Schumacher Cc: PFC , Tom Lane , pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 Message-ID: <20050801214931.GF60019@decibel.org> References: <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> <42EE93F8.5010003@aptalaska.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <42EE93F8.5010003@aptalaska.net> X-Operating-System: FreeBSD 4.11-RELEASE-p10 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.003 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/10 X-Sequence-Number: 13757 On Mon, Aug 01, 2005 at 01:28:24PM -0800, Matthew Schumacher wrote: > PFC wrote: > > > > > >> select put_tokens2(1, '{"\\246\\323\\061\\332\\277"}', 1, 1, 10000); > > > > > > Try adding more backslashes until it works (seems that you need \\\\ > > or something). > > Don't DBI convert the language types to postgres quoted forms on its > > own ? > > > > Your right.... I am finding that the proc is not the problem as I > suspected, it works correctly when I am not calling it from perl, > something isn't escaped correctly.... I'm not sure who's responsible for DBI::Pg (Josh?), but would it make sense to add better support for bytea to DBI::Pg? ISTM there should be a better way of doing this than adding gobs of \'s. -- Jim C. Nasby, Database Consultant decibel@decibel.org Give your computer some brain candy! www.distributed.net Team #1828 Windows: "Where do you want to go today?" Linux: "Where do you want to go tomorrow?" FreeBSD: "Are you guys coming, or what?" From pgsql-performance-owner@postgresql.org Mon Aug 1 19:09:28 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6337452C2F for ; Mon, 1 Aug 2005 19:09:26 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 85765-03 for ; Mon, 1 Aug 2005 22:09:21 +0000 (GMT) Received: from outbound.mailhop.org (outbound.mailhop.org [63.208.196.171]) by svr1.postgresql.org (Postfix) with ESMTP id 7C5E952BFA for ; Mon, 1 Aug 2005 19:09:20 -0300 (ADT) Received: from ppp-66-143-176-222.dsl.austtx.swbell.net ([66.143.176.222] helo=mail.herk.net) by outbound.mailhop.org with esmtpsa (TLSv1:DES-CBC3-SHA:168) (Exim 4.51) id 1DziTK-000A8w-7n for pgsql-performance@postgresql.org; Mon, 01 Aug 2005 18:09:22 -0400 Received: from [192.168.1.100] ([192.168.1.1]) (AUTH: LOGIN parker@herk.net) by mail.herk.net with esmtp; Mon, 01 Aug 2005 17:09:20 -0500 id 0011A3A1.42EE9D90.0000707B X-Mail-Handler: MailHop Outbound by DynDNS.org X-Originating-IP: 66.143.176.222 X-Report-Abuse-To: abuse@dyndns.org (see http://www.mailhop.org/outbound/abuse.html for abuse reporting information) X-MHO-User: parkerm Message-ID: <42EE9D83.8000405@pobox.com> Date: Mon, 01 Aug 2005 17:09:07 -0500 From: Michael Parker User-Agent: Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US; rv:1.7.6) Gecko/20050317 Thunderbird/1.0.2 Mnenhy/0.7.2.0 X-Accept-Language: en-us, en Mime-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="=_mail-28795-1122934160-0001-2" To: pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 References: <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> <42EE93F8.5010003@aptalaska.net> <20050801214931.GF60019@decibel.org> In-Reply-To: <20050801214931.GF60019@decibel.org> X-Enigmail-Version: 0.91.0.0 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 tagged_above=0 required=5 tests=FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/11 X-Sequence-Number: 13758 This is a MIME-formatted message. If you see this text it means that your E-mail software does not support MIME-formatted messages. --=_mail-28795-1122934160-0001-2 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Jim C. Nasby wrote: >I'm not sure who's responsible for DBI::Pg (Josh?), but would it make >sense to add better support for bytea to DBI::Pg? ISTM there should be a >better way of doing this than adding gobs of \'s. > > It has support for binding a bytea parameter, but in this case we're trying to build up an array and pass that into a stored procedure. The $dbh->quote() method for DBD::Pg lacks the ability to quote bytea types. There is actually a TODO note in the code about adding support for quoting Pg specific types. Presumabliy the difficulties we are having with this would be solved by that, once it has been implemented. In the meantime, I believe it's just a matter of getting the right escapes happen so that the procedure is inserting values that we can later get via a select and using bind_param() with the PG_BYTEA type. Michael --=_mail-28795-1122934160-0001-2 Content-Type: application/pgp-signature; name="signature.asc" Content-Transfer-Encoding: 7bit Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (Darwin) iD8DBQFC7p2HG4km+uS4gOIRAn9gAJwL+6XcJLjYloAjO3hOFKvlcTbzVwCeLwsy RJPPQZDWRYPIer0eOcixutE= =xv0t -----END PGP SIGNATURE----- --=_mail-28795-1122934160-0001-2-- From pgsql-performance-owner@postgresql.org Mon Aug 1 19:19:34 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A24ED52E45 for ; Mon, 1 Aug 2005 19:19:31 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 85345-06 for ; Mon, 1 Aug 2005 22:19:26 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.193]) by svr1.postgresql.org (Postfix) with ESMTP id 1246952931 for ; Mon, 1 Aug 2005 19:19:25 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i4so1100741wra for ; Mon, 01 Aug 2005 15:19:28 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:mime-version:content-type; b=S1ulFd5kkedaClcKR43dRSSo6OxcSgQUPkE3lhBTmCCkvYm9UBUcQSYLW2NuGwvveXY3W9EPNabq1FS0/oniPjGkLUyjwoZ1MXi8wJzd+aZ437sznaga3sxqsLrtNad8cw22RDtk31ffq9JojMptw5B5d0a3Z3yJP8MeMczFpbg= Received: by 10.54.8.60 with SMTP id 60mr3266886wrh; Mon, 01 Aug 2005 15:19:28 -0700 (PDT) Received: by 10.54.42.25 with HTTP; Mon, 1 Aug 2005 15:19:27 -0700 (PDT) Message-ID: Date: Tue, 2 Aug 2005 00:19:27 +0200 From: Meetesh Karia Reply-To: meetesh.karia@alumni.duke.edu To: pgsql-performance@postgresql.org Subject: Planner incorrectly choosing seq scan over index scan Mime-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_Part_3517_6467213.1122934767947" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.231 tagged_above=0 required=5 tests=AWL, HTML_40_50, HTML_MESSAGE, RCVD_BY_IP X-Spam-Level: X-Archive-Number: 200508/12 X-Sequence-Number: 13759 ------=_Part_3517_6467213.1122934767947 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Hi all, We're using 8.0.3 and we're seeing a problem where the planner is choosing = a=20 seq scan and hash join over an index scan. If I set enable_hashjoin to off,= =20 then I get the plan I'm expecting and the query runs a lot faster. I've als= o=20 tried lowering the random page cost (even to 1) but the planner still=20 chooses to use the hash join. Does anyone have any thoughts/suggestions? I saw that there was a thread=20 recently in which the planner wasn't correctly estimating the cost for=20 queries using LIMIT. Is it possible that something similar is happening her= e=20 (perhaps because of the sort) and that the patch Tom proposed would fix it? Thanks. Here are the various queries and plans: Normal settings ------------------------ explain analyze select c.sourceId, c.targetId, abs(c.tr - c.sr ) as xmy, (c.sr - s.ar ) * (c.tr - t.ar) as xy, (c.sr - s.ar ) * (c.sr - s.ar) as x2, (c.tr - t.ar ) * (c.tr - t.ar) as y2 from candidates617004 c, lte_user s, lte_user t where c.sourceId =3D s.user_id and c.targetId =3D t.user_id order by c.sourceId, c.targetId; QUERY PLAN Sort (cost=3D13430.57..13439.24 rows=3D3467 width=3D48) (actual time=3D 1390.000..1390.000 rows=3D3467 loops=3D1) Sort Key: c.sourceid, c.targetid -> Merge Join (cost=3D9912.07..13226.72 rows=3D3467 width=3D48) (actual tim= e=3D 1344.000..1375.000 rows=3D3467 loops=3D1) Merge Cond: ("outer".user_id =3D "inner".sourceid) -> Index Scan using lte_user_pkey on lte_user s (cost=3D0.00..16837.71rows=3D279395 width=3D16) (actual time=3D 0.000..95.000 rows=3D50034 loops=3D1) -> Sort (cost=3D9912.07..9920.73 rows=3D3467 width=3D40) (actual time=3D 1156.000..1156.000 rows=3D3467 loops=3D1) Sort Key: c.sourceid -> Hash Join (cost=3D8710.44..9708.21 rows=3D3467 width=3D40) (actual time= =3D 1125.000..1156.000 rows=3D3467 loops=3D1) Hash Cond: ("outer".targetid =3D "inner".user_id) -> Seq Scan on candidates617004 c (cost=3D0.00..67.67 rows=3D3467 width=3D3= 2)=20 (actual time=3D0.000..0.000 rows=3D3467 loops=3D1) -> Hash (cost=3D8011.95..8011.95 rows=3D279395 width=3D16) (actual time=3D 1125.000..1125.000 rows=3D0 loops=3D1) -> Seq Scan on lte_user t (cost=3D0.00..8011.95 rows=3D279395 width=3D16) (= actual=20 time=3D0.000..670.000 rows=3D279395 loops=3D1) Total runtime: 1406.000 ms enable_hashjoin disabled ---------------------------------------- QUERY PLAN Sort (cost=3D14355.37..14364.03 rows=3D3467 width=3D48) (actual time=3D 391.000..391.000 rows=3D3467 loops=3D1) Sort Key: c.sourceid, c.targetid -> Nested Loop (cost=3D271.52..14151.51 rows=3D3467 width=3D48) (actual tim= e=3D 203.000..359.000 rows=3D3467 loops=3D1) -> Merge Join (cost=3D271.52..3490.83 rows=3D3467 width=3D40) (actual time= =3D 203.000..218.000 rows=3D3467 loops=3D1) Merge Cond: ("outer".user_id =3D "inner".sourceid) -> Index Scan using lte_user_pkey on lte_user s (cost=3D0.00..16837.71rows=3D279395 width=3D16) (actual time=3D 0.000..126.000 rows=3D50034 loops=3D1) -> Sort (cost=3D271.52..280.19 rows=3D3467 width=3D32) (actual time=3D15.000..30.000rows=3D3467 loops=3D1) Sort Key: c.sourceid -> Seq Scan on candidates617004 c (cost=3D0.00..67.67 rows=3D3467 width=3D3= 2)=20 (actual time=3D0.000..0.000 rows=3D3467 loops=3D1) -> Index Scan using lte_user_pkey on lte_user t (cost=3D0.00..3.03 rows=3D1= =20 width=3D16) (actual time=3D0.031..0.036 rows=3D1 loops=3D3467) Index Cond: ("outer".targetid =3D t.user_id) Total runtime: 406.000 ms random_page_cost set to 1.5 ---------------------------------------------- QUERY PLAN Sort (cost=3D12702.62..12711.29 rows=3D3467 width=3D48) (actual time=3D 1407.000..1407.000 rows=3D3467 loops=3D1) Sort Key: c.sourceid, c.targetid -> Merge Join (cost=3D9912.07..12498.77 rows=3D3467 width=3D48) (actual tim= e=3D 1391.000..1407.000 rows=3D3467 loops=3D1) Merge Cond: ("outer".user_id =3D "inner".sourceid) -> Index Scan using lte_user_pkey on lte_user s (cost=3D0.00..12807.34rows=3D279395 width=3D16) (actual time=3D 0.000..46.000 rows=3D50034 loops=3D1) -> Sort (cost=3D9912.07..9920.73 rows=3D3467 width=3D40) (actual time=3D 1188.000..1188.000 rows=3D3467 loops=3D1) Sort Key: c.sourceid -> Hash Join (cost=3D8710.44..9708.21 rows=3D3467 width=3D40) (actual time= =3D 1157.000..1188.000 rows=3D3467 loops=3D1) Hash Cond: ("outer".targetid =3D "inner".user_id) -> Seq Scan on candidates617004 c (cost=3D0.00..67.67 rows=3D3467 width=3D3= 2)=20 (actual time=3D0.000..15.000 rows=3D3467 loops=3D1) -> Hash (cost=3D8011.95..8011.95 rows=3D279395 width=3D16) (actual time=3D 1157.000..1157.000 rows=3D0 loops=3D1) -> Seq Scan on lte_user t (cost=3D0.00..8011.95 rows=3D279395 width=3D16) (= actual=20 time=3D0.000..750.000 rows=3D279395 loops=3D1) Total runtime: 1422.000 ms random_page_cost set to 1.5 and enable_hashjoin set to false ---------------------------------------------------------------------------= ----------------------- QUERY PLAN Sort (cost=3D13565.58..13574.25 rows=3D3467 width=3D48) (actual time=3D 390.000..390.000 rows=3D3467 loops=3D1) Sort Key: c.sourceid, c.targetid -> Nested Loop (cost=3D271.52..13361.73 rows=3D3467 width=3D48) (actual tim= e=3D 203.000..360.000 rows=3D3467 loops=3D1) -> Merge Join (cost=3D271.52..2762.88 rows=3D3467 width=3D40) (actual time= =3D 203.000..250.000 rows=3D3467 loops=3D1) Merge Cond: ("outer".user_id =3D "inner".sourceid) -> Index Scan using lte_user_pkey on lte_user s (cost=3D0.00..12807.34rows=3D279395 width=3D16) (actual time=3D 0.000..48.000 rows=3D50034 loops=3D1) -> Sort (cost=3D271.52..280.19 rows=3D3467 width=3D32) (actual time=3D15.000..31.000rows=3D3467 loops=3D1) Sort Key: c.sourceid -> Seq Scan on candidates617004 c (cost=3D0.00..67.67 rows=3D3467 width=3D3= 2)=20 (actual time=3D0.000..15.000 rows=3D3467 loops=3D1) -> Index Scan using lte_user_pkey on lte_user t (cost=3D0.00..3.02 rows=3D1= =20 width=3D16) (actual time=3D0.023..0.023 rows=3D1 loops=3D3467) Index Cond: ("outer".targetid =3D t.user_id) Total runtime: 406.000 ms Thanks, Meetesh ------=_Part_3517_6467213.1122934767947 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Hi all,

We're using 8.0.3 and we're seeing a problem where the planner is choosing a seq scan and hash join over an index scan.  If I set enable_hashjoin to off, then I get the plan I'm expecting and the query runs a lot faster.  I've also tried lowering the random page cost (even to 1) but the planner still chooses to use the hash join.

Does anyone have any thoughts/suggestions?  I saw that there was a thread recently in which the planner wasn't correctly estimating the cost for queries using LIMIT.  Is it possible that something similar is happening here (perhaps because of the sort) and that the patch Tom proposed would fix it?

Thanks.  Here are the various queries and plans:

Normal settings
------------------------
explain analyze
    select
        c.sourceId,
        c.targetId,
        abs(c.tr= - c.sr) as xmy,
        (c.sr - s.ar) * (c.tr -= t.ar) as xy,
        (c.sr - s.ar) * (c.sr -= s.ar) as x2,
        (c.tr - t.ar) * (c.tr -= t.ar) as y2
    from
        candidates617004 c,
        lte_user s,
        lte_user t
    where
        c.sourceId =3D s.user_id
        and c.targetId =3D t.user_id
    order by
        c.sourceId,
        c.targetId;

QUERY PLAN
Sort  (cost=3D13430.57..13439.24 rows=3D3467 width=3D48) (actual time= =3D1390.000..1390.000 rows=3D3467 loops=3D1)
  Sort Key: c.sourceid, c.targetid
  ->  Merge Join  (cost=3D9912.07..13226.72 rows=3D3467 width=3D48) (actual time=3D1344.000..1375.000 rows=3D3467 loops=3D1)
        Merge Cond: ("outer".u= ser_id =3D "inner".sourceid)
        ->  Index Scan using lte_user_pkey on lte_user s  (cost=3D0.00..16837.71 rows=3D279395 width=3D16) (actual time=3D0.000..95.000 rows=3D50034 loops=3D1)
        ->  Sort  (cost=3D9912.07..9920.73 rows=3D3467 width=3D40) (actual time=3D1156.000..1156.000 rows=3D3467 loops=3D1)
            &nb= sp; Sort Key: c.sourceid
            &nb= sp; ->  Hash Join  (cost=3D8710.44..9708.21 rows=3D3467 width=3D40= ) (actual time=3D1125.000..1156.000 rows=3D3467 loops=3D1)
            &nb= sp;       Hash Cond: ("outer".targetid =3D "inner".user_id)
            &nb= sp;       ->  Seq Scan on candidates617004 c  (cost=3D0.00..67.67 rows=3D3467 width=3D32) (actual time=3D0.000..0.000 rows=3D3467 loops=3D1)<= br>             &nb= sp;       ->  Hash  (cost=3D8011.95..8011.95 rows=3D279395 width=3D16) (actual time=3D1125.000..1125.000 rows=3D0 loops=3D1)
            &nb= sp;             ->  Seq Scan on lte_user t  (cost=3D0.00..8011.95 rows=3D279395 width=3D16) (actual time=3D0.000..670.000 rows=3D279395 loops= =3D1)
Total runtime: 1406.000 ms

enable_hashjoin disabled
----------------------------------------
QUERY PLAN
Sort  (cost=3D14355.37..14364.03 rows=3D3467 width=3D48) (actual time= =3D391.000..391.000 rows=3D3467 loops=3D1)
  Sort Key: c.sourceid, c.targetid
  ->  Nested Loop  (cost=3D271.52..14151.51 rows=3D3467 w= idth=3D48) (actual time=3D203.000..359.000 rows=3D3467 loops=3D1)
        ->  Merge Join  (cost=3D271.52..3490.83 rows=3D3467 width=3D40) (actual time=3D203.000..218= .000 rows=3D3467 loops=3D1)
            &nb= sp; Merge Cond: ("outer".user_id =3D "inner".sourceid)<= br>             &nb= sp; ->  Index Scan using lte_user_pkey on lte_user s  (cost=3D0.00..16837.71 rows=3D279395 width=3D16) (actual time=3D0.000..126.= 000 rows=3D50034 loops=3D1)
            &nb= sp; ->  Sort  (cost=3D271.52..280.19 rows=3D3467 width=3D32) (actu= al time=3D15.000..30.000 rows=3D3467 loops=3D1)
            &nb= sp;       Sort Key: c.sourceid
            &nb= sp;       ->  Seq Scan on candidates617004 c  (cost=3D0.00..67.67 rows=3D3467 width=3D32) (actual time=3D0.000..0.000 rows=3D3467 loops=3D1)<= br>         ->  Index Scan using lte_user_pkey on lte_user t  (cost=3D0.00..3.03 rows=3D1 width=3D16) (actual time=3D0.031..0.036 rows=3D1 loops=3D3467)
            &nb= sp; Index Cond: ("outer".targetid =3D t.user_id)
Total runtime: 406.000 ms

random_page_cost set to 1.5
----------------------------------------------
QUERY PLAN
Sort  (cost=3D12702.62..12711.29 rows=3D3467 width=3D48) (actual time= =3D1407.000..1407.000 rows=3D3467 loops=3D1)
  Sort Key: c.sourceid, c.targetid
  ->  Merge Join  (cost=3D9912.07..12498.77 rows=3D3467 width=3D48) (actual time=3D1391.000..1407.000 rows=3D3467 loops=3D1)
        Merge Cond: ("outer".u= ser_id =3D "inner".sourceid)
        ->  Index Scan using lte_user_pkey on lte_user s  (cost=3D0.00..12807.34 rows=3D279395 width=3D16) (actual time=3D0.000..46.000 rows=3D50034 loops=3D1)
        ->  Sort  (cost=3D9912.07..9920.73 rows=3D3467 width=3D40) (actual time=3D1188.000..1188.000 rows=3D3467 loops=3D1)
            &nb= sp; Sort Key: c.sourceid
            &nb= sp; ->  Hash Join  (cost=3D8710.44..9708.21 rows=3D3467 width=3D40= ) (actual time=3D1157.000..1188.000 rows=3D3467 loops=3D1)
            &nb= sp;       Hash Cond: ("outer".targetid =3D "inner".user_id)
            &nb= sp;       ->  Seq Scan on candidates617004 c  (cost=3D0.00..67.67 rows=3D3467 width=3D32) (actual time=3D0.000..15.000 rows=3D3467 loops=3D1)=
            &nb= sp;       ->  Hash  (cost=3D8011.95..8011.95 rows=3D279395 width=3D16) (actual time=3D1157.000..1157.000 rows=3D0 loops=3D1)
            &nb= sp;             ->  Seq Scan on lte_user t  (cost=3D0.00..8011.95 rows=3D279395 width=3D16) (actual time=3D0.000..750.000 rows=3D279395 loops= =3D1)
Total runtime: 1422.000 ms

random_page_cost set to 1.5 and enable_hashjoin set to false
---------------------------------------------------------------------------= -----------------------
QUERY PLAN
Sort  (cost=3D13565.58..13574.25 rows=3D3467 width=3D48) (actual time= =3D390.000..390.000 rows=3D3467 loops=3D1)
  Sort Key: c.sourceid, c.targetid
  ->  Nested Loop  (cost=3D271.52..13361.73 rows=3D3467 w= idth=3D48) (actual time=3D203.000..360.000 rows=3D3467 loops=3D1)
        ->  Merge Join  (cost=3D271.52..2762.88 rows=3D3467 width=3D40) (actual time=3D203.000..250= .000 rows=3D3467 loops=3D1)
            &nb= sp; Merge Cond: ("outer".user_id =3D "inner".sourceid)<= br>             &nb= sp; ->  Index Scan using lte_user_pkey on lte_user s  (cost=3D0.00..12807.34 rows=3D279395 width=3D16) (actual time=3D0.000..48.0= 00 rows=3D50034 loops=3D1)
            &nb= sp; ->  Sort  (cost=3D271.52..280.19 rows=3D3467 width=3D32) (actu= al time=3D15.000..31.000 rows=3D3467 loops=3D1)
            &nb= sp;       Sort Key: c.sourceid
            &nb= sp;       ->  Seq Scan on candidates617004 c  (cost=3D0.00..67.67 rows=3D3467 width=3D32) (actual time=3D0.000..15.000 rows=3D3467 loops=3D1)=
        ->  Index Scan using lte_user_pkey on lte_user t  (cost=3D0.00..3.02 rows=3D1 width=3D16) (actual time=3D0.023..0.023 rows=3D1 loops=3D3467)
            &nb= sp; Index Cond: ("outer".targetid =3D t.user_id)
Total runtime: 406.000 ms

Thanks,
Meetesh
------=_Part_3517_6467213.1122934767947-- From pgsql-performance-owner@postgresql.org Mon Aug 1 19:37:30 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BC0FF52894 for ; Mon, 1 Aug 2005 19:37:27 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91710-06 for ; Mon, 1 Aug 2005 22:37:21 +0000 (GMT) Received: from mail.nordicbet.com (mail.nordicbet.com [193.69.167.130]) by svr1.postgresql.org (Postfix) with ESMTP id BE84652B4E for ; Mon, 1 Aug 2005 19:37:16 -0300 (ADT) Received: from [84.48.51.94] (helo=tobias.nordicbet.invalid) (Authenticated Sender=tobias@nordicbet.com) by mail.nordicbet.com with esmtpa (Exim 4.50 #1 (Debian)) id 1DzisS-0003iy-QW; Tue, 02 Aug 2005 00:35:29 +0200 Received: by tobias.nordicbet.invalid (Postfix, from userid 500) id AE30FE003B; Tue, 2 Aug 2005 00:37:08 +0200 (CEST) Date: Tue, 2 Aug 2005 00:37:08 +0200 From: Tobias Brox To: meetesh.karia@alumni.duke.edu Cc: pgsql-performance@postgresql.org Subject: Re: Planner incorrectly choosing seq scan over index scan Message-ID: <20050801223708.GE17649@tobias.lan> References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: Organization: Group Nordicbet User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/13 X-Sequence-Number: 13760 [Meetesh Karia - Tue at 12:19:27AM +0200] > We're using 8.0.3 and we're seeing a problem where the planner is choosing a > seq scan and hash join over an index scan. If I set enable_hashjoin to off, > then I get the plan I'm expecting and the query runs a lot faster. I've also > tried lowering the random page cost (even to 1) but the planner still > chooses to use the hash join. Have you tried increasing the statistics collection? -- Tobias Brox, +47-91700050 Nordicbet, IT dept From pgsql-performance-owner@postgresql.org Mon Aug 1 20:16:44 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3877152C2A for ; Mon, 1 Aug 2005 20:16:43 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 99393-07 for ; Mon, 1 Aug 2005 23:16:38 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id D701B52C15 for ; Mon, 1 Aug 2005 20:16:34 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050801231636m9100g65e3e>; Mon, 1 Aug 2005 23:16:36 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id C3F565604F; Mon, 1 Aug 2005 18:16:35 -0500 (CDT) Received: from [192.168.1.11] (liliana.arbash-meinel.com [192.168.1.11]) by juju.arbash-meinel.com (Postfix) with ESMTP id 6FDF355FD0; Mon, 1 Aug 2005 18:16:30 -0500 (CDT) Message-ID: <42EEAD4B.4010905@arbash-meinel.com> Date: Mon, 01 Aug 2005 18:16:27 -0500 From: John Arbash Meinel User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050727) X-Accept-Language: en-us, en MIME-Version: 1.0 To: meetesh.karia@alumni.duke.edu Cc: pgsql-performance@postgresql.org Subject: Re: Planner incorrectly choosing seq scan over index scan References: In-Reply-To: X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig3CCABE5B81B1FDFC3904F657" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.046 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/14 X-Sequence-Number: 13761 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig3CCABE5B81B1FDFC3904F657 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Meetesh Karia wrote: > Hi all, > > We're using 8.0.3 and we're seeing a problem where the planner is > choosing a seq scan and hash join over an index scan. If I set > enable_hashjoin to off, then I get the plan I'm expecting and the query > runs a lot faster. I've also tried lowering the random page cost (even > to 1) but the planner still chooses to use the hash join. > > Does anyone have any thoughts/suggestions? I saw that there was a > thread recently in which the planner wasn't correctly estimating the > cost for queries using LIMIT. Is it possible that something similar is > happening here (perhaps because of the sort) and that the patch Tom > proposed would fix it? > > Thanks. Here are the various queries and plans: > > Normal settings ... > QUERY PLAN > Sort (cost=13430.57..13439.24 rows=3467 width=48) (actual > time=1390.000..1390.000 rows=3467 loops=1) > Sort Key: c.sourceid, c.targetid > -> Merge Join (cost=9912.07..13226.72 rows=3467 width=48) (actual > time=1344.000..1375.000 rows=3467 loops=1) > Merge Cond: ("outer".user_id = "inner".sourceid) > -> Index Scan using lte_user_pkey on lte_user s > (cost=0.00..16837.71 rows=279395 width=16) (actual time=0.000..95.000 > rows=50034 loops=1) This is where the planner is messing up, and mis-estimating the selectivity. It is expecting to get 280k rows, but only needs to get 50k. I assume lte_user is the bigger table, and that candidates617004 has some subset. Has lte_user and candidates617004 been recently ANALYZEd? All estimates, except for the expected number of rows from lte_user seem to be okay. Is user_id the primary key for lte_user? I'm trying to figure out how you can get 50k rows, by searching a primary key, against a 3.5k rows. Is user_id only part of the primary key for lte_user? Can you give us the output of: \d lte_user \d candidates617004 So that we have the description of the tables, and what indexes you have defined? Also, if you could describe the table layouts, that would help. John =:-> > -> Sort (cost=9912.07..9920.73 rows=3467 width=40) (actual > time=1156.000..1156.000 rows=3467 loops=1) > Sort Key: c.sourceid > -> Hash Join (cost=8710.44..9708.21 rows=3467 width=40) > (actual time=1125.000..1156.000 rows=3467 loops=1) > Hash Cond: ("outer".targetid = "inner".user_id) > -> Seq Scan on candidates617004 c > (cost=0.00..67.67 rows=3467 width=32) (actual time=0.000..0.000 > rows=3467 loops=1) > -> Hash (cost=8011.95..8011.95 rows=279395 > width=16) (actual time=1125.000..1125.000 rows=0 loops=1) > -> Seq Scan on lte_user t > (cost=0.00..8011.95 rows=279395 width=16) (actual time=0.000..670.000 > rows=279395 loops=1) > Total runtime: 1406.000 ms > > enable_hashjoin disabled > ---------------------------------------- > QUERY PLAN > Sort (cost=14355.37..14364.03 rows=3467 width=48) (actual > time=391.000..391.000 rows=3467 loops=1) > Sort Key: c.sourceid, c.targetid > -> Nested Loop (cost=271.52..14151.51 rows=3467 width=48) (actual > time=203.000..359.000 rows=3467 loops=1) > -> Merge Join (cost=271.52..3490.83 rows=3467 width=40) > (actual time=203.000..218.000 rows=3467 loops=1) > Merge Cond: ("outer".user_id = "inner".sourceid) > -> Index Scan using lte_user_pkey on lte_user s > (cost=0.00..16837.71 rows=279395 width=16) (actual time=0.000..126.000 > rows=50034 loops=1) > -> Sort (cost=271.52..280.19 rows=3467 width=32) (actual > time=15.000..30.000 rows=3467 loops=1) > Sort Key: c.sourceid > -> Seq Scan on candidates617004 c > (cost=0.00..67.67 rows=3467 width=32) (actual time=0.000..0.000 > rows=3467 loops=1) > -> Index Scan using lte_user_pkey on lte_user t > (cost=0.00..3.03 rows=1 width=16) (actual time=0.031..0.036 rows=1 > loops=3467) > Index Cond: ("outer".targetid = t.user_id) > Total runtime: 406.000 ms > > random_page_cost set to 1.5 > ---------------------------------------------- > QUERY PLAN > Sort (cost=12702.62..12711.29 rows=3467 width=48) (actual > time=1407.000..1407.000 rows=3467 loops=1) > Sort Key: c.sourceid, c.targetid > -> Merge Join (cost=9912.07..12498.77 rows=3467 width=48) (actual > time=1391.000..1407.000 rows=3467 loops=1) > Merge Cond: ("outer".user_id = "inner".sourceid) > -> Index Scan using lte_user_pkey on lte_user s > (cost=0.00..12807.34 rows=279395 width=16) (actual time=0.000..46.000 > rows=50034 loops=1) > -> Sort (cost=9912.07..9920.73 rows=3467 width=40) (actual > time=1188.000..1188.000 rows=3467 loops=1) > Sort Key: c.sourceid > -> Hash Join (cost=8710.44..9708.21 rows=3467 width=40) > (actual time=1157.000..1188.000 rows=3467 loops=1) > Hash Cond: ("outer".targetid = "inner".user_id) > -> Seq Scan on candidates617004 c > (cost=0.00..67.67 rows=3467 width=32) (actual time=0.000..15.000 > rows=3467 loops=1) > -> Hash (cost=8011.95..8011.95 rows=279395 > width=16) (actual time=1157.000..1157.000 rows=0 loops=1) > -> Seq Scan on lte_user t > (cost=0.00..8011.95 rows=279395 width=16) (actual time=0.000..750.000 > rows=279395 loops=1) > Total runtime: 1422.000 ms > > random_page_cost set to 1.5 and enable_hashjoin set to false > -------------------------------------------------------------------------------------------------- > QUERY PLAN > Sort (cost=13565.58..13574.25 rows=3467 width=48) (actual > time=390.000..390.000 rows=3467 loops=1) > Sort Key: c.sourceid, c.targetid > -> Nested Loop (cost=271.52..13361.73 rows=3467 width=48) (actual > time=203.000..360.000 rows=3467 loops=1) > -> Merge Join (cost=271.52..2762.88 rows=3467 width=40) > (actual time=203.000..250.000 rows=3467 loops=1) > Merge Cond: ("outer".user_id = "inner".sourceid) > -> Index Scan using lte_user_pkey on lte_user s > (cost=0.00..12807.34 rows=279395 width=16) (actual time=0.000..48.000 > rows=50034 loops=1) > -> Sort (cost=271.52..280.19 rows=3467 width=32) (actual > time=15.000..31.000 rows=3467 loops=1) > Sort Key: c.sourceid > -> Seq Scan on candidates617004 c > (cost=0.00..67.67 rows=3467 width=32) (actual time=0.000..15.000 > rows=3467 loops=1) > -> Index Scan using lte_user_pkey on lte_user t > (cost=0.00..3.02 rows=1 width=16) (actual time=0.023..0.023 rows=1 > loops=3467) > Index Cond: ("outer".targetid = t.user_id) > Total runtime: 406.000 ms > > Thanks, > Meetesh --------------enig3CCABE5B81B1FDFC3904F657 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC7q1OJdeBCYSNAAMRAkenAJ94Lfxat19qbFY9gaS+cXNfINK3RQCeP2eK jEJ19FqiH59B9F75X/RRTAE= =5VvY -----END PGP SIGNATURE----- --------------enig3CCABE5B81B1FDFC3904F657-- From pgsql-performance-owner@postgresql.org Mon Aug 1 20:30:30 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D7AC152C2A for ; Mon, 1 Aug 2005 20:30:29 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 00446-07 for ; Mon, 1 Aug 2005 23:30:24 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.205]) by svr1.postgresql.org (Postfix) with ESMTP id 1F80B529E1 for ; Mon, 1 Aug 2005 20:30:23 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i4so1110508wra for ; Mon, 01 Aug 2005 16:30:27 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:cc:in-reply-to:mime-version:content-type:references; b=m4BqNLZ3iSOshYiEm5uKTke3ymmdGumSz2y8iZGcPJibEXrkHHNBwcaAN0qpJEYkIwpM77MHnfZ/IE+EX+f4wo8s8Uo1HO/qpEO44an0Bi+6tYVEDf2QMy3FVoN8PNc2H6eoiuZyG0RakgnHwOBVVyfjY2dU/322g43BmELMa9E= Received: by 10.54.129.7 with SMTP id b7mr3320848wrd; Mon, 01 Aug 2005 16:30:26 -0700 (PDT) Received: by 10.54.42.25 with HTTP; Mon, 1 Aug 2005 16:30:26 -0700 (PDT) Message-ID: Date: Tue, 2 Aug 2005 01:30:26 +0200 From: Meetesh Karia Reply-To: meetesh.karia@alumni.duke.edu To: Tobias Brox Subject: Re: Planner incorrectly choosing seq scan over index scan Cc: pgsql-performance@postgresql.org In-Reply-To: <20050801223708.GE17649@tobias.lan> Mime-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_Part_3553_30794248.1122939026870" References: <20050801223708.GE17649@tobias.lan> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.196 tagged_above=0 required=5 tests=AWL, HTML_30_40, HTML_MESSAGE, RCVD_BY_IP X-Spam-Level: X-Archive-Number: 200508/15 X-Sequence-Number: 13762 ------=_Part_3553_30794248.1122939026870 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Are you referring to the statistics gathering target for ANALYZE? Based on= =20 your email, I just tried the following and then re-ran the explain analyze= =20 but got the same "incorrect" plan: alter table candidates617004 alter column sourceId set statistics 1000, alter column targetId set statistics 1000; analyze candidates617004; alter table lte_user alter column user_id set statistics 1000; analyze lte_user; Thanks for your suggestion, Meetesh On 8/2/05, Tobias Brox wrote: >=20 > [Meetesh Karia - Tue at 12:19:27AM +0200] > > We're using 8.0.3 and we're seeing a problem where the planner is=20 > choosing a > > seq scan and hash join over an index scan. If I set enable_hashjoin to= =20 > off, > > then I get the plan I'm expecting and the query runs a lot faster. I've= =20 > also > > tried lowering the random page cost (even to 1) but the planner still > > chooses to use the hash join. >=20 > Have you tried increasing the statistics collection? >=20 > -- > Tobias Brox, +47-91700050 > Nordicbet, IT dept > ------=_Part_3553_30794248.1122939026870 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Are you referring to the statistics gathering target for ANALYZE?  Based on your email, I just tried the following and then re-ran the explain analyze but got the same "incorrect" plan:

alter table candidates617004
    alter column sourceId set statistics 1000,
    alter column targetId set statistics 1000;
analyze candidates617004;

alter table lte_user
    alter column user_id set statistics 1000;
analyze lte_user;

Thanks for your suggestion,
Meetesh

On 8/2/05, Tobias Brox <tobias@nordi= cbet.com> wrote:
[Meetesh Karia - Tue at 12:19:27AM +0200]
> We're using 8.0.3 and we'= re seeing a problem where the planner is choosing a
> seq scan and ha= sh join over an index scan. If I set enable_hashjoin to off,
> then I= get the plan I'm expecting and the query runs a lot faster. I've also
> tried lowering the random page cost (even to 1) but the planner st= ill
> chooses to use the hash join.

Have you tried increasing = the statistics collection?

--
Tobias Brox, +47-91700050
Nordic= bet, IT dept

------=_Part_3553_30794248.1122939026870-- From pgsql-patches-owner@postgresql.org Mon Aug 1 20:51:18 2005 X-Original-To: pgsql-patches-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5859252BAB for ; Mon, 1 Aug 2005 20:51:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 09155-04 for ; Mon, 1 Aug 2005 23:51:12 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 1A19652BA9 for ; Mon, 1 Aug 2005 20:51:11 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j71NpBIY026132; Mon, 1 Aug 2005 19:51:11 -0400 (EDT) To: "Alon Goldshuv" Cc: pgsql-patches@postgresql.org, "Luke Lonergan" Subject: Re: [PERFORM] COPY FROM performance improvements In-reply-to: References: Comments: In-reply-to "Alon Goldshuv" message dated "Mon, 01 Aug 2005 15:48:35 -0400" Date: Mon, 01 Aug 2005 19:51:10 -0400 Message-ID: <26131.1122940270@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/25 X-Sequence-Number: 16972 "Alon Goldshuv" writes: >> This patch appears to reverse out the most recent committed changes in >> copy.c. > Which changes do you refer to? I thought I accommodated all the recent > changes (I recall some changes to the tupletable/tupleslot interface, HEADER > in cvs, and hex escapes and maybe one or 2 more). What did I miss? The latest touch of copy.c, namely this patch: 2005-07-10 17:13 tgl * doc/src/sgml/ref/create_type.sgml, src/backend/commands/copy.c, src/backend/commands/typecmds.c, src/backend/tcop/fastpath.c, src/backend/tcop/postgres.c, src/backend/utils/adt/arrayfuncs.c, src/backend/utils/adt/date.c, src/backend/utils/adt/numeric.c, src/backend/utils/adt/rowtypes.c, src/backend/utils/adt/timestamp.c, src/backend/utils/adt/varbit.c, src/backend/utils/adt/varchar.c, src/backend/utils/adt/varlena.c, src/backend/utils/mb/mbutils.c, src/include/catalog/catversion.h, src/include/catalog/pg_proc.h, src/test/regress/expected/type_sanity.out, src/test/regress/sql/type_sanity.sql: Change typreceive function API so that receive functions get the same optional arguments as text input functions, ie, typioparam OID and atttypmod. Make all the datatypes that use typmod enforce it the same way in typreceive as they do in typinput. This fixes a problem with failure to enforce length restrictions during COPY FROM BINARY. It was rather obvious, given that the first chunk of the patch backed up the file's CVS version stamp from 1.247 to 1.246 :-( regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 1 20:56:19 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6000252C36 for ; Mon, 1 Aug 2005 20:56:18 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 07610-10 for ; Mon, 1 Aug 2005 23:56:10 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.207]) by svr1.postgresql.org (Postfix) with ESMTP id 04E3652975 for ; Mon, 1 Aug 2005 20:56:09 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i4so1113717wra for ; Mon, 01 Aug 2005 16:56:13 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:cc:in-reply-to:mime-version:content-type:references; b=PeOiPrwWnSbHxFREi+8ocmiW59eoHLVtllcMi9DCleHsrDVXW2/+rF6sP/v1cZAjaRFmU4kqQAmx7Rmo7rexL6fLQn8tNbdhFJT9symG+tPAQf6Yyg0YSreHbajvUkdBJ3g3JUrCnskoBT+8hGMUNqs53eTxjvUJ6Ysj8wEbQNY= Received: by 10.54.57.21 with SMTP id f21mr3130742wra; Mon, 01 Aug 2005 16:56:13 -0700 (PDT) Received: by 10.54.42.25 with HTTP; Mon, 1 Aug 2005 16:56:12 -0700 (PDT) Message-ID: Date: Tue, 2 Aug 2005 01:56:13 +0200 From: Meetesh Karia Reply-To: meetesh.karia@alumni.duke.edu To: John Arbash Meinel Subject: Re: Planner incorrectly choosing seq scan over index scan Cc: pgsql-performance@postgresql.org In-Reply-To: <42EEAD4B.4010905@arbash-meinel.com> Mime-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_Part_3573_25248839.1122940573051" References: <42EEAD4B.4010905@arbash-meinel.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.199 tagged_above=0 required=5 tests=AWL, HTML_50_60, HTML_MESSAGE, RCVD_BY_IP X-Spam-Level: X-Archive-Number: 200508/16 X-Sequence-Number: 13763 ------=_Part_3573_25248839.1122940573051 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Thanks John. I've answered your questions below: Has lte_user and candidates617004 been recently ANALYZEd? All estimates, > except for the expected number of rows from lte_user seem to be okay. I ANALYZEd both tables just before putting together my first email. And,=20 unfortunately, modifying the statistics target didn't help either. Is user_id the primary key for lte_user? Yes=20 I'm trying to figure out how you can get 50k rows, by searching a > primary key, against a 3.5k rows. Is user_id only part of the primary > key for lte_user? Hmmm ... I missed that before. But, that surprises me too. Especially since= =20 sourceId in the candidates table has only 1 value. Also, user_id is the=20 complete primary key for lte_user. Can you give us the output of: > \d lte_user > \d candidates617004 Sure, here they are: lte=3D# \d lte_user Table "public.lte_user" Column | Type | Modifiers ---------------+-----------------------------+----------- user_id | bigint | not null firstname | character varying(255) | lastname | character varying(255) | address1 | character varying(255) | address2 | character varying(255) | city | character varying(255) | state | character varying(255) | zip | character varying(255) | phone1 | character varying(255) | phone2 | character varying(255) | username | character varying(255) | password | character varying(255) | deleted | boolean | not null ext_cust_id | character varying(255) | aboutme | character varying(255) | birthday | timestamp without time zone | fm_id | bigint | ar | double precision | Indexes: "lte_user_pkey" PRIMARY KEY, btree (user_id) "idx_user_extid" btree (ext_cust_id) "idx_user_username" btree (username) Foreign-key constraints: "fk_user_fm" FOREIGN KEY (fm_id) REFERENCES fm(fm_id) lte=3D# \d candidates617004 Table "public.candidates617004" Column | Type | Modifiers --------------+------------------+----------- fmid | bigint | sourceid | bigint | sr | double precision | targetid | bigint | tr | double precision |=20 Also, if you could describe the table layouts, that would help. Sure. The lte_user table is just a collection of users. user_id is assigned= =20 uniquely using a sequence. During some processing, we create a candidates= =20 table (candidates617004 in our case). This table is usually a temp table.= =20 sourceid is a user_id (in this case it is always 617004) and targetid is=20 also a user_id (2860 distinct values out of 3467). The rest of the=20 information is either only used in the select clause or not used at all=20 during this processing. Did I miss something in the table layout description that would be helpful? Thanks for your help! Meetesh > -> Sort (cost=3D9912.07..9920.73 rows=3D3467 width=3D40) (actual > > time=3D1156.000..1156.000 rows=3D3467 loops=3D1) > > Sort Key: c.sourceid > > -> Hash Join (cost=3D8710.44..9708.21 rows=3D3467 width=3D40) > > (actual time=3D1125.000..1156.000 rows=3D3467 loops=3D1) > > Hash Cond: ("outer".targetid =3D "inner".user_id) > > -> Seq Scan on candidates617004 c > > (cost=3D0.00..67.67 rows=3D3467 width=3D32) (actual time=3D0.000..0.000 > > rows=3D3467 loops=3D1) > > -> Hash (cost=3D8011.95..8011.95 rows=3D279395 > > width=3D16) (actual time=3D1125.000..1125.000 rows=3D0 loops=3D1) > > -> Seq Scan on lte_user t > > (cost=3D0.00..8011.95 rows=3D279395 width=3D16) (actual time=3D0.000..6= 70.000 > > rows=3D279395 loops=3D1) > > Total runtime: 1406.000 ms > > > > enable_hashjoin disabled > > ---------------------------------------- > > QUERY PLAN > > Sort (cost=3D14355.37..14364.03 rows=3D3467 width=3D48) (actual > > time=3D391.000..391.000 rows=3D3467 loops=3D1) > > Sort Key: c.sourceid, c.targetid > > -> Nested Loop (cost=3D271.52..14151.51 rows=3D3467 width=3D48) (actual > > time=3D203.000..359.000 rows=3D3467 loops=3D1) > > -> Merge Join (cost=3D271.52..3490.83 rows=3D3467 width=3D40) > > (actual time=3D203.000..218.000 rows=3D3467 loops=3D1) > > Merge Cond: ("outer".user_id =3D "inner".sourceid) > > -> Index Scan using lte_user_pkey on lte_user s > > (cost=3D0.00..16837.71 rows=3D279395 width=3D16) (actual time=3D0.000..= 126.000 > > rows=3D50034 loops=3D1) > > -> Sort (cost=3D271.52..280.19 rows=3D3467 width=3D32) (actual > > time=3D15.000..30.000 rows=3D3467 loops=3D1) > > Sort Key: c.sourceid > > -> Seq Scan on candidates617004 c > > (cost=3D0.00..67.67 rows=3D3467 width=3D32) (actual time=3D0.000..0.000 > > rows=3D3467 loops=3D1) > > -> Index Scan using lte_user_pkey on lte_user t > > (cost=3D0.00..3.03 rows=3D1 width=3D16) (actual time=3D0.031..0.036 row= s=3D1 > > loops=3D3467) > > Index Cond: ("outer".targetid =3D t.user_id) > > Total runtime: 406.000 ms > > > > random_page_cost set to 1.5 > > ---------------------------------------------- > > QUERY PLAN > > Sort (cost=3D12702.62..12711.29 rows=3D3467 width=3D48) (actual > > time=3D1407.000..1407.000 rows=3D3467 loops=3D1) > > Sort Key: c.sourceid, c.targetid > > -> Merge Join (cost=3D9912.07..12498.77 rows=3D3467 width=3D48) (actual > > time=3D1391.000..1407.000 rows=3D3467 loops=3D1) > > Merge Cond: ("outer".user_id =3D "inner".sourceid) > > -> Index Scan using lte_user_pkey on lte_user s > > (cost=3D0.00..12807.34 rows=3D279395 width=3D16) (actual time=3D0.000..= 46.000 > > rows=3D50034 loops=3D1) > > -> Sort (cost=3D9912.07..9920.73 rows=3D3467 width=3D40) (actual > > time=3D1188.000..1188.000 rows=3D3467 loops=3D1) > > Sort Key: c.sourceid > > -> Hash Join (cost=3D8710.44..9708.21 rows=3D3467 width=3D40) > > (actual time=3D1157.000..1188.000 rows=3D3467 loops=3D1) > > Hash Cond: ("outer".targetid =3D "inner".user_id) > > -> Seq Scan on candidates617004 c > > (cost=3D0.00..67.67 rows=3D3467 width=3D32) (actual time=3D0.000..15.00= 0 > > rows=3D3467 loops=3D1) > > -> Hash (cost=3D8011.95..8011.95 rows=3D279395 > > width=3D16) (actual time=3D1157.000..1157.000 rows=3D0 loops=3D1) > > -> Seq Scan on lte_user t > > (cost=3D0.00..8011.95 rows=3D279395 width=3D16) (actual time=3D0.000..7= 50.000 > > rows=3D279395 loops=3D1) > > Total runtime: 1422.000 ms > > > > random_page_cost set to 1.5 and enable_hashjoin set to false > >=20 > -------------------------------------------------------------------------= ------------------------- > > QUERY PLAN > > Sort (cost=3D13565.58..13574.25 rows=3D3467 width=3D48) (actual > > time=3D390.000..390.000 rows=3D3467 loops=3D1) > > Sort Key: c.sourceid, c.targetid > > -> Nested Loop (cost=3D271.52..13361.73 rows=3D3467 width=3D48) (actual > > time=3D203.000..360.000 rows=3D3467 loops=3D1) > > -> Merge Join (cost=3D271.52..2762.88 rows=3D3467 width=3D40) > > (actual time=3D203.000..250.000 rows=3D3467 loops=3D1) > > Merge Cond: ("outer".user_id =3D "inner".sourceid) > > -> Index Scan using lte_user_pkey on lte_user s > > (cost=3D0.00..12807.34 rows=3D279395 width=3D16) (actual time=3D0.000..= 48.000 > > rows=3D50034 loops=3D1) > > -> Sort (cost=3D271.52..280.19 rows=3D3467 width=3D32) (actual > > time=3D15.000..31.000 rows=3D3467 loops=3D1) > > Sort Key: c.sourceid > > -> Seq Scan on candidates617004 c > > (cost=3D0.00..67.67 rows=3D3467 width=3D32) (actual time=3D0.000..15.00= 0 > > rows=3D3467 loops=3D1) > > -> Index Scan using lte_user_pkey on lte_user t > > (cost=3D0.00..3.02 rows=3D1 width=3D16) (actual time=3D0.023..0.023 row= s=3D1 > > loops=3D3467) > > Index Cond: ("outer".targetid =3D t.user_id) > > Total runtime: 406.000 ms > > > > Thanks, > > Meetesh >=20 >=20 >=20 > ------=_Part_3573_25248839.1122940573051 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Thanks John.  I've answered your questions below:

Has lte_user and candidates617004 been recently ANALYZEd? All estimates,except for the expected number of rows from lte_user seem to be okay.

I ANALYZEd both tables just before putting together my first email.  And, unfortunately, modifying the statistics target didn't help either.

Is user= _id the primary key for lte_user?

Yes

I'm try= ing to figure out how you can get 50k rows, by searching a
primary key, = against a=20 3.5k rows. Is user_id only part of the primary
key for lte_user?

Hmmm ... I missed that before.  But, that surprises me too.  Especially since sourceId in the candidates table has only 1 value.  Also, user_id is the complete primary key for lte_user.

Can you= give us the output of:
\d lte_user
\d candidates617004

Sure, here they are:

lte=3D# \d lte_user
            &nb= sp;    Table "public.lte_user"
    Column     |            Type            = ; | Modifiers
---------------+-----------------------------+-----------
 user_id       | bigint           &nb= sp;          | not null
 firstname     | character varying(255)  = ;    |
 lastname      | character varying(255) =      |
 address1      | character varying(255) =      |
 address2      | character varying(255) =      |
 city          | characte= r varying(255)      |
 state         | character var= ying(255)      |
 zip           | cha= racter varying(255)      |
 phone1        | character varying(= 255)      |
 phone2        | character varying(= 255)      |
 username      | character varying(255) =      |
 password      | character varying(255) =      |
 deleted       | boolean           &n= bsp;         | not null
 ext_cust_id   | character varying(255)   &nb= sp;  |
 aboutme       | character varying(255)&= nbsp;     |
 birthday      | timestamp without time zone = |
 fm_id       | bigint           &nb= sp;          |
 ar            | double precision            |
Indexes:
    "lte_user_pkey" PRIMARY KEY, btree (user_id)     "idx_user_extid" btree (ext_cust_id)
    "idx_user_username" btree (username)
Foreign-key constraints:
    "fk_user_fm" FOREIGN KEY (fm_id) REFERENCES fm= (fm_id)

lte=3D# \d candidates617004
       Table "public.candidates617004&qu= ot;
    Column    |       Type       | Modifiers
--------------+------------------+-----------
 fmid       | bigint   &n= bsp;       |
 sourceid     | bigint    &nbs= p;      |
 sr            | double precision |=
 targetid     | bigint    &nbs= p;      |
 tr           | double precision |

Also, i= f you could describe the table layouts, that would help.

Sure.  The lte_user table is just a collection of users.  user_id is assigned uniquely using a sequence.  During some processing, we create a candidates table (candidates617004 in our case).  This table is usually a temp table.  sourceid is a user_id (in this case it is always 617004) and targetid is also a user_id (2860 distinct values out of 3467).  The rest of the information is either only used in the select clause or not used at all during this processing.

Did I miss something in the table layout description that would be helpful?=

Thanks for your help!
Meetesh

> &nb= sp;       ->  Sort  (cost=3D9912.07..9920.73 rows=3D3467 width=3D40) (actual
> time=3D1156.000..1156.000 rows=3D3467 loops=3D1= )
>           =     Sort Key: c.sourceid
>    &nbs= p;          ->  Hash Join  (cost=3D8710.44..9708.21 rows=3D3467 width=3D40)
> (actual time=3D1125.000..1156.000 rows=3D3467 loops=3D1= )
>           =           Hash Cond: ("outer".targetid =3D "inner".user_id)
&g= t;            &= nbsp;        ->  Seq Scan on candidates617004 c
> (cost=3D0.00..67.67= rows=3D3467 width=3D32) (actual time=3D0.000..0.000
> rows=3D3467 lo= ops=3D1)
>          = ;           ->  Hash  (cost=3D8011.95..8011.95 rows=3D279395
= > width=3D16) (actual time=3D1125.000..1125.000 rows=3D0 loops=3D1)
&= gt;            =             &nb= sp;  ->  Seq Scan on lte_user t
> (cost=3D0.00..8011.95 rows= =3D279395 width=3D16) (actual time=3D0.000..670.000
> rows=3D279395 l= oops=3D1)
> Total runtime: 1406.000 ms
>
> enable_hashjoi= n disabled
> ----------------------------------------
> QUERY PLAN
> Sort  (cost=3D14355.37..14364.03 rows= =3D3467 width=3D48) (actual
> time=3D391.000..391.000 rows=3D3467 loo= ps=3D1)
>   Sort Key: c.sourceid, c.targetid
> &= nbsp; ->  Nested Loop  (cost=3D271.52..14151.51 rows=3D3467 width=3D48) (actual
> time=3D203.000..359.000 rows=3D346= 7 loops=3D1)
>         ->  Merge Join  (cost=3D271.52..3490.83 rows=3D3467 width=3D40)
> (actual time=3D203.000..218.000 rows=3D3467 loops=3D1)<= br>>           &n= bsp;   Merge Cond: ("outer".user_id =3D "inner".sourceid)
&= gt;            =    ->  Index Scan using lte_user_pkey on lte_user s
> (cost= =3D0.00..16837.71 rows=3D279395 width=3D16) (actual time=3D0.000..126.000> rows=3D50034 loops=3D1)
>      &= nbsp;        ->  Sort  (cost=3D271.52..280.19 rows=3D3467 width=3D32) (actual
> time=3D15.000..30.000 rows=3D3467 loops=3D1)>           &nbs= p;         Sort Key: c.sourceid
>        = ;             ->  Seq Scan on candidates617004 c
> (cost=3D0.00..67.67= rows=3D3467 width=3D32) (actual time=3D0.000..0.000
> rows=3D3467 lo= ops=3D1)
>         -> = ; Index Scan using lte_user_pkey on lte_user t
> (cost=3D0.00..3= .03 rows=3D1 width=3D16) (actual time=3D0.031..0.036 rows=3D1
> loops=3D= 3467)
>          &n= bsp;    Index Cond: ("outer".targetid =3D t.user_id)
> Total runtim= e: 406.000 ms
>
> random_page_cost set to 1.5
> ---------= -------------------------------------
> QUERY PLAN
> Sort =  (cost=3D 12702.62..12711.29 rows=3D3467 width=3D48) (actual
> time=3D1407.000.= .1407.000 rows=3D3467 loops=3D1)
>   Sort Key: c.sourceid, = c.targetid
>   ->  Merge Join  (cost= =3D9912.07..12498.77 rows=3D3467 width=3D48) (actual
> time=3D1391.000..1407.000 rows=3D3467 loops=3D1)
> &nbs= p;       Merge Cond: ("outer".user_= id =3D "inner".sourceid)
>     &nb= sp;   ->  Index Scan using lte_user_pkey on lte_user= s
> (cost=3D0.00..12807.34 rows=3D279395 width=3D16) (actual time=3D0.000..46.000
> rows=3D5003= 4 loops=3D1)
>         ->  Sort  (cost=3D9912.07..9920.73 rows=3D3467 width=3D40) (actual
> time=3D1188.000..1188.000 rows=3D3467 loops=3D1= )
>           =     Sort Key: c.sourceid
>    &nbs= p;          ->  Hash Join  (cost=3D8710.44..9708.21 rows=3D3467 width=3D40)
> (actual time=3D1157.000..1188.000 rows=3D3467 loops=3D1= )
>           =           Hash Cond: ("outer".targetid =3D "inner".user_id)
&g= t;            &= nbsp;        ->  Seq Scan on candidates617004 c
> (cost=3D0.00..67.67= rows=3D3467 width=3D32) (actual time=3D0.000..15.000
> rows=3D3467 l= oops=3D1)
>         &nbs= p;           ->  Hash  (cost=3D8011.95..8011.95 rows=3D279395
= > width=3D16) (actual time=3D1157.000..1157.000 rows=3D0 loops=3D1)
&= gt;            =             &nb= sp;  ->  Seq Scan on lte_user t
> (cost=3D0.00..8011.95 rows= =3D279395 width=3D16) (actual time=3D0.000..750.000
> rows=3D279395 l= oops=3D1)
> Total runtime: 1422.000 ms
>
> random_page_co= st set to 1.5 and enable_hashjoin set to false
> ------------------------------------------------------------------= --------------------------------
> QUERY PLAN
> Sort  = ;(cost=3D13565.58..13574.25 rows=3D3467 width=3D48) (actual
> time=3D= 390.000..390.000 rows=3D3467 loops=3D1)
>   Sort Key: c.sourceid, c.targeti= d
>   ->  Nested Loop  (cost=3D271.5= 2..13361.73 rows=3D3467 width=3D48) (actual
> time=3D203.000..360.000= rows=3D3467 loops=3D1)
>       &n= bsp; ->  Merge Join  (cost=3D271.52..2762.88 rows=3D3467 width=3D40)
> (actual time=3D203.000..250.000 rows=3D3467 loops=3D1)<= br>>           &n= bsp;   Merge Cond: ("outer".user_id =3D "inner".sourceid)
&= gt;            =    ->  Index Scan using lte_user_pkey on lte_user s
> (cost= =3D0.00..12807.34 rows=3D279395 width=3D16) (actual time=3D0.000..48.000> rows=3D50034 loops=3D1)
>      &n= bsp;        ->  Sort  (cost=3D271.52..280.19 rows=3D3467 width=3D32) (actual
> time=3D15.000..31.000 rows=3D3467 loops=3D1)>           &nbs= p;         Sort Key: c.sourceid
>        = ;             ->  Seq Scan on candidates617004 c
> (cost=3D0.00..67.67= rows=3D3467 width=3D32) (actual time=3D0.000..15.000
> rows=3D3467 l= oops=3D1)
>         ->&nbs= p; Index Scan using lte_user_pkey on lte_user t
> (cost=3D 0.00..3.02 rows=3D1 width=3D16) (actual time=3D0.023..0.023 rows=3D1
>= ; loops=3D3467)
>        &nbs= p;      Index Cond: ("outer".targetid =3D t.user_id)
> Total runtim= e: 406.000 ms
>
> Thanks,
> Meetesh




------=_Part_3573_25248839.1122940573051-- From pgsql-performance-owner@postgresql.org Mon Aug 1 21:15:33 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 885C152BC3 for ; Mon, 1 Aug 2005 21:15:31 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 12235-05 for ; Tue, 2 Aug 2005 00:15:26 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id AB6F4529E1 for ; Mon, 1 Aug 2005 21:15:25 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j720FQaO026293; Mon, 1 Aug 2005 20:15:26 -0400 (EDT) To: meetesh.karia@alumni.duke.edu Cc: John Arbash Meinel , pgsql-performance@postgresql.org Subject: Re: Planner incorrectly choosing seq scan over index scan In-reply-to: References: <42EEAD4B.4010905@arbash-meinel.com> Comments: In-reply-to Meetesh Karia message dated "Tue, 02 Aug 2005 01:56:13 +0200" Date: Mon, 01 Aug 2005 20:15:26 -0400 Message-ID: <26292.1122941726@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/17 X-Sequence-Number: 13764 Meetesh Karia writes: > Sure. The lte_user table is just a collection of users. user_id is assigned= > uniquely using a sequence. During some processing, we create a candidates= > table (candidates617004 in our case). This table is usually a temp table.= > sourceid is a user_id (in this case it is always 617004) and targetid is=20 > also a user_id (2860 distinct values out of 3467). The rest of the=20 > information is either only used in the select clause or not used at all=20 > during this processing. If you know that sourceid has only a single value, it'd probably be helpful to call out that value in the query, ie, where ... AND c.sourceId = 617004 ... regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 2 04:06:02 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B7C1752E97 for ; Tue, 2 Aug 2005 04:05:59 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 12448-05 for ; Tue, 2 Aug 2005 07:05:50 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.198]) by svr1.postgresql.org (Postfix) with ESMTP id 032AA52E96 for ; Tue, 2 Aug 2005 04:05:48 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i4so1162935wra for ; Tue, 02 Aug 2005 00:05:50 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:cc:in-reply-to:mime-version:content-type:references; b=A6mAIeGuY5dNZIzyOJIpQxh/LfPDIZDEvq3UmrT3GMMr6qM4IF8pb6vdpcVcaxN4xeo5TNqb6r0XhfxO0u0LAErHlyjbORUtilr+Rkp1u3CVoiM5x10SqNczvE+yX9CWhtuBaivsZRkqlR+NUvbJwMARog3+wCHcblu9b1Z2hlg= Received: by 10.54.2.57 with SMTP id 57mr3555676wrb; Tue, 02 Aug 2005 00:05:50 -0700 (PDT) Received: by 10.54.42.25 with HTTP; Tue, 2 Aug 2005 00:05:50 -0700 (PDT) Message-ID: Date: Tue, 2 Aug 2005 09:05:50 +0200 From: Meetesh Karia Reply-To: meetesh.karia@alumni.duke.edu To: Tom Lane Subject: Re: Planner incorrectly choosing seq scan over index scan Cc: John Arbash Meinel , pgsql-performance@postgresql.org In-Reply-To: <26292.1122941726@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_Part_3703_21659964.1122966350462" References: <42EEAD4B.4010905@arbash-meinel.com> <26292.1122941726@sss.pgh.pa.us> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.185 tagged_above=0 required=5 tests=AWL, HTML_40_50, HTML_MESSAGE, RCVD_BY_IP X-Spam-Level: X-Archive-Number: 200508/18 X-Sequence-Number: 13765 ------=_Part_3703_21659964.1122966350462 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Thanks Tom, That modifies the query plan slightly, but the planner still decides to do = a=20 hash join for the lte_user table aliased 't'. Though, if I make this change= =20 and set enable_hashjoin to off, the query plan (and execution time) gets=20 even better. enable_hashjoin =3D on ---------------------------------- QUERY PLAN Sort (cost=3D10113.35..10122.02 rows=3D3467 width=3D48) (actual time=3D 1203.000..1203.000 rows=3D3467 loops=3D1) Sort Key: c.sourceid, c.targetid -> Nested Loop (cost=3D8711.19..9909.50 rows=3D3467 width=3D48) (actual tim= e=3D 1156.000..1203.000 rows=3D3467 loops=3D1) -> Index Scan using lte_user_pkey on lte_user s (cost=3D0.00..3.02 rows=3D1= =20 width=3D16) (actual time=3D0.000..0.000 rows=3D1 loops=3D1) Index Cond: (617004 =3D user_id) -> Hash Join (cost=3D8711.19..9776.46 rows=3D3467 width=3D40) (actual time= =3D 1156.000..1187.000 rows=3D3467 loops=3D1) Hash Cond: ("outer".targetid =3D "inner".user_id) -> Seq Scan on candidates617004 c (cost=3D0.00..76.34 rows=3D3467 width=3D3= 2)=20 (actual time=3D0.000..16.000 rows=3D3467 loops=3D1) Filter: (sourceid =3D 617004) -> Hash (cost=3D8012.55..8012.55 rows=3D279455 width=3D16) (actual time=3D 1141.000..1141.000 rows=3D0 loops=3D1) -> Seq Scan on lte_user t (cost=3D0.00..8012.55 rows=3D279455 width=3D16) (= actual=20 time=3D0.000..720.000 rows=3D279395 loops=3D1) Total runtime: 1218.000 ms enable_hashjoin =3D off ----------------------------------- QUERY PLAN Sort (cost=3D10942.56..10951.22 rows=3D3467 width=3D48) (actual time=3D 188.000..188.000 rows=3D3467 loops=3D1) Sort Key: c.sourceid, c.targetid -> Nested Loop (cost=3D0.00..10738.71 rows=3D3467 width=3D48) (actual time= =3D 0.000..188.000 rows=3D3467 loops=3D1) -> Index Scan using lte_user_pkey on lte_user s (cost=3D0.00..3.02 rows=3D1= =20 width=3D16) (actual time=3D0.000..0.000 rows=3D1 loops=3D1) Index Cond: (617004 =3D user_id) -> Nested Loop (cost=3D0.00..10605.67 rows=3D3467 width=3D40) (actual time= =3D 0.000..157.000 rows=3D3467 loops=3D1) -> Seq Scan on candidates617004 c (cost=3D0.00..76.34 rows=3D3467 width=3D3= 2)=20 (actual time=3D0.000..15.000 rows=3D3467 loops=3D1) Filter: (sourceid =3D 617004) -> Index Scan using lte_user_pkey on lte_user t (cost=3D0.00..3.02 rows=3D1= =20 width=3D16) (actual time=3D0.028..0.037 rows=3D1 loops=3D3467) Index Cond: ("outer".targetid =3D t.user_id) Total runtime: 188.000 ms Thanks, Meetesh On 8/2/05, Tom Lane wrote: >=20 > Meetesh Karia writes: > > Sure. The lte_user table is just a collection of users. user_id is=20 > assigned=3D > > uniquely using a sequence. During some processing, we create a=20 > candidates=3D > > table (candidates617004 in our case). This table is usually a temp=20 > table.=3D > > sourceid is a user_id (in this case it is always 617004) and targetid= =20 > is=3D20 > > also a user_id (2860 distinct values out of 3467). The rest of the=3D20 > > information is either only used in the select clause or not used at=20 > all=3D20 > > during this processing. >=20 > If you know that sourceid has only a single value, it'd probably be > helpful to call out that value in the query, ie, > where ... AND c.sourceId =3D 617004 ... >=20 > regards, tom lane > ------=_Part_3703_21659964.1122966350462 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Thanks Tom,

That modifies the query plan slightly, but the planner still decides to do a hash join for the lte_user table aliased 't'.  Though, if I make this change and set enable_hashjoin to off, the query plan (and execution time) gets even better.

enable_hashjoin =3D on
----------------------------------
QUERY PLAN
Sort  (cost=3D10113.35..10122.02 rows=3D3467 width=3D48) (actual time= =3D1203.000..1203.000 rows=3D3467 loops=3D1)
  Sort Key: c.sourceid, c.targetid
  ->  Nested Loop  (cost=3D8711.19..9909.50 rows=3D3467 width=3D48) (actual time=3D1156.000..1203.000 rows=3D3467 loops=3D1)
        ->  Index Scan using lte_user_pkey on lte_user s  (cost=3D0.00..3.02 rows=3D1 width=3D16) (actual time=3D0.000..0.000 rows=3D1 loops=3D1)
            &nb= sp; Index Cond: (617004 =3D user_id)
        ->  Hash Join  (cost=3D8711.19..9776.46 rows=3D3467 width=3D40) (actual time=3D1156.000..1187.000 rows=3D3467 loops=3D1)
            &nb= sp; Hash Cond: ("outer".targetid =3D "inner".user_id)             &nb= sp; ->  Seq Scan on candidates617004 c  (cost=3D0.00..76.34 rows=3D3467 width=3D32) (actual time=3D0.000..16.000 rows=3D3467 loops=3D1)=
            &nb= sp;       Filter: (sourceid =3D 617004)
            &nb= sp; ->  Hash  (cost=3D8012.55..8012.55 rows=3D279455 width=3D16) (actual time=3D1141.000..1141.000 rows=3D0 loops=3D1)
            &nb= sp;       ->  Seq Scan on lte_user t  (cost=3D0.00..8012.55 rows=3D279455 width=3D16) (actual time=3D0.000..720.000 rows=3D279395 loops= =3D1)
Total runtime: 1218.000 ms

enable_hashjoin =3D off
-----------------------------------
QUERY PLAN
Sort  (cost=3D10942.56..10951.22 rows=3D3467 width=3D48) (actual time= =3D188.000..188.000 rows=3D3467 loops=3D1)
  Sort Key: c.sourceid, c.targetid
  ->  Nested Loop  (cost=3D0.00..10738.71 rows=3D3467 wid= th=3D48) (actual time=3D0.000..188.000 rows=3D3467 loops=3D1)
        ->  Index Scan using lte_user_pkey on lte_user s  (cost=3D0.00..3.02 rows=3D1 width=3D16) (actual time=3D0.000..0.000 rows=3D1 loops=3D1)
            &nb= sp; Index Cond: (617004 =3D user_id)
        ->  Nested Loop  (cost=3D0.00..10605.67 rows=3D3467 width=3D40) (actual time=3D0.000..157.000 rows=3D3467 loops=3D1)
            &nb= sp; ->  Seq Scan on candidates617004 c  (cost=3D0.00..76.34 rows=3D3467 width=3D32) (actual time=3D0.000..15.000 rows=3D3467 loops=3D1)=
            &nb= sp;       Filter: (sourceid =3D 617004)
            &nb= sp; ->  Index Scan using lte_user_pkey on lte_user t  (cost=3D0.00..3.02 rows=3D1 width=3D16) (actual time=3D0.028..0.037 rows=3D= 1 loops=3D3467)
            &nb= sp;       Index Cond: ("outer".targetid =3D t.user_id)
Total runtime: 188.000 ms

Thanks,
Meetesh

On 8/2/05, Tom Lane <tgl@sss= .pgh.pa.us> wrote:
Meetesh Karia <meetesh.karia@= gmail.com> writes:
> Sure. The lte_user table is just a collec= tion of users. user_id is assigned=3D
> uniquely using a sequence. Du= ring some processing, we create a candidates=3D
> table (candidates617004 in our case). This table is usually a temp= table.=3D
> sourceid is a user_id (in this case it is always 617004)= and targetid is=3D20
> also a user_id (2860 distinct values out of 3= 467). The rest of the=3D20
> information is either only used in the select clause or not used a= t all=3D20
> during this processing.

If you know that sourceid= has only a single value, it'd probably be
helpful to call out that valu= e in the query, ie,
        where ... AND c.sourceI= d =3D 617004 ...

        &nb= sp;            =    regards, tom lane

------=_Part_3703_21659964.1122966350462-- From pgsql-patches-owner@postgresql.org Tue Aug 2 12:00:17 2005 X-Original-To: pgsql-patches-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0908D52F18 for ; Tue, 2 Aug 2005 12:00:14 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 25433-01 for ; Tue, 2 Aug 2005 15:00:02 +0000 (GMT) Received: from mail.Mi8.com (nycgw01.mi8.com [63.240.6.42]) by svr1.postgresql.org (Postfix) with ESMTP id 7F9DA52F14 for ; Tue, 2 Aug 2005 11:59:56 -0300 (ADT) Received: from 172.16.1.118 by mail.Mi8.com with ESMTP (- GW01 Welcome to Mi8 Corporation www.Mi8.com); Tue, 02 Aug 2005 10:59:42 -0400 X-Server-Uuid: F1A2E19A-84E4-48DD-8F48-B475613F58B2 Received: from MI8NYCMAIL06.Mi8.com ([172.16.1.175]) by d01smtp03.Mi8.com with Microsoft SMTPSVC(6.0.3790.1830); Tue, 2 Aug 2005 10:59:42 -0400 Received: from 141.156.39.156 ([141.156.39.156]) by MI8NYCMAIL06.Mi8.com ([172.16.1.219]) via Exchange Front-End Server mi8owa.mi8.com ( [172.16.1.104]) with Microsoft Exchange Server HTTP-DAV ; Tue, 2 Aug 2005 10:59:39 -0500 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Tue, 02 Aug 2005 10:59:43 -0400 Subject: Re: COPY FROM performance improvements From: "Alon Goldshuv" To: pgsql-patches@postgresql.org Message-ID: In-Reply-To: MIME-Version: 1.0 X-OriginalArrivalTime: 02 Aug 2005 14:59:42.0607 (UTC) FILETIME=[D04361F0:01C59772] X-WSS-ID: 6EF155D427K725273-01-01 Content-Type: multipart/mixed; boundary=B_3205825185_103175599 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.569 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO, RCVD_NUMERIC_HELO X-Spam-Level: * X-Archive-Number: 200508/28 X-Sequence-Number: 16975 > This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. --B_3205825185_103175599 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit New patch attached. It includes very minor changes. These are changes that were committed to CVS 3 weeks ago (copy.c 1.247) which I missed in the previous patch. Alon. --B_3205825185_103175599 Content-Type: application/octet-stream; name=copy_parse_improvements_V16.patch Content-Disposition: attachment; filename=copy_parse_improvements_V16.patch Content-Transfer-Encoding: base64 SW5kZXg6IHNyYy9iYWNrZW5kL2NvbW1hbmRzL2NvcHkuYwo9PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09ClJDUyBm aWxlOiAvcHJvamVjdHMvY3Zzcm9vdC9wZ3NxbC9zcmMvYmFja2VuZC9jb21tYW5kcy9jb3B5 LmMsdgpyZXRyaWV2aW5nIHJldmlzaW9uIDEuMjQ3CmRpZmYgLWMgLXIxLjI0NyBjb3B5LmMK KioqIHNyYy9iYWNrZW5kL2NvbW1hbmRzL2NvcHkuYwkxMCBKdWwgMjAwNSAyMToxMzo1OCAt MDAwMAkxLjI0NwotLS0gc3JjL2JhY2tlbmQvY29tbWFuZHMvY29weS5jCTIgQXVnIDIwMDUg MTQ6Mzg6MzcgLTAwMDAKKioqKioqKioqKioqKioqCioqKiA1MCw1NSAqKioqCi0tLSA1MCw1 NiAtLS0tCiAgCiAgI2RlZmluZSBJU09DVEFMKGMpICgoKGMpID49ICcwJykgJiYgKChjKSA8 PSAnNycpKQogICNkZWZpbmUgT0NUVkFMVUUoYykgKChjKSAtICcwJykKKyAjZGVmaW5lIENP UFlfQlVGX1NJWkUgNjU1MzYKICAKICAvKgogICAqIFJlcHJlc2VudHMgdGhlIGRpZmZlcmVu dCBzb3VyY2UvZGVzdCBjYXNlcyB3ZSBuZWVkIHRvIHdvcnJ5IGFib3V0IGF0CioqKioqKioq KioqKioqKgoqKiogNjMsNzggKioqKgogIH0gQ29weURlc3Q7CiAgCiAgLyoKLSAgKiBTdGF0 ZSBpbmRpY2F0b3Igc2hvd2luZyB3aGF0IHN0b3BwZWQgQ29weVJlYWRBdHRyaWJ1dGUoKQot ICAqLwotIHR5cGVkZWYgZW51bSBDb3B5UmVhZFJlc3VsdAotIHsKLSAJTk9STUFMX0FUVFIs Ci0gCUVORF9PRl9MSU5FLAotIAlVTlRFUk1JTkFURURfRklFTEQKLSB9IENvcHlSZWFkUmVz dWx0OwotIAotIC8qCiAgICoJUmVwcmVzZW50cyB0aGUgZW5kLW9mLWxpbmUgdGVybWluYXRv ciB0eXBlIG9mIHRoZSBpbnB1dAogICAqLwogIHR5cGVkZWYgZW51bSBFb2xUeXBlCi0tLSA2 NCw2OSAtLS0tCioqKioqKioqKioqKioqKgoqKiogOTcsMTAyICoqKioKLS0tIDg4LDk2IC0t LS0KICBzdGF0aWMgRW9sVHlwZSBlb2xfdHlwZTsJCS8qIEVPTCB0eXBlIG9mIGlucHV0ICov CiAgc3RhdGljIGludAljbGllbnRfZW5jb2Rpbmc7CS8qIHJlbW90ZSBzaWRlJ3MgY2hhcmFj dGVyIGVuY29kaW5nICovCiAgc3RhdGljIGludAlzZXJ2ZXJfZW5jb2Rpbmc7CS8qIGxvY2Fs IGVuY29kaW5nICovCisgc3RhdGljIGNoYXIJZW9sX2NoWzJdOwkJCS8qIFRoZSBieXRlIHZh bHVlcyBvZiB0aGUgMSBvciAyIGVvbCBieXRlcyAqLworIHN0YXRpYyBib29sIGNsaWVudF9l bmNvZGluZ19vbmx5OyAvKiB0cnVlIGlmIGNsaWVudCBlbmNvZGluZyBpcyBhIG5vbgorICAg ICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgKiBzdXBwb3J0ZWQgc2VydmVyIGVu Y29kaW5nICovCiAgCiAgLyogdGhlc2UgYXJlIGp1c3QgZm9yIGVycm9yIG1lc3NhZ2VzLCBz ZWUgY29weV9pbl9lcnJvcl9jYWxsYmFjayAqLwogIHN0YXRpYyBib29sIGNvcHlfYmluYXJ5 OwkJLyogaXMgaXQgYSBiaW5hcnkgY29weT8gKi8KKioqKioqKioqKioqKioqCioqKiAxMDQs MTI2ICoqKioKICBzdGF0aWMgaW50CWNvcHlfbGluZW5vOwkJLyogbGluZSBudW1iZXIgZm9y IGVycm9yIG1lc3NhZ2VzICovCiAgc3RhdGljIGNvbnN0IGNoYXIgKmNvcHlfYXR0bmFtZTsJ LyogY3VycmVudCBhdHQgZm9yIGVycm9yIG1lc3NhZ2VzICovCiAgCiAgCiAgLyoKICAgKiBU aGVzZSBzdGF0aWMgdmFyaWFibGVzIGFyZSB1c2VkIHRvIGF2b2lkIGluY3VycmluZyBvdmVy aGVhZCBmb3IgZWFjaAohICAqIGF0dHJpYnV0ZSBwcm9jZXNzZWQuICBhdHRyaWJ1dGVfYnVm IGlzIHJldXNlZCBvbiBlYWNoIENvcHlSZWFkQXR0cmlidXRlCiAgICogY2FsbCB0byBob2xk IHRoZSBzdHJpbmcgYmVpbmcgcmVhZCBpbi4gIFVuZGVyIG5vcm1hbCB1c2UgaXQgd2lsbCBz b29uCiAgICogZ3JvdyB0byBhIHN1aXRhYmxlIHNpemUsIGFuZCB0aGVuIHdlIHdpbGwgYXZv aWQgcGFsbG9jL3BmcmVlIG92ZXJoZWFkCiAgICogZm9yIHN1YnNlcXVlbnQgYXR0cmlidXRl cy4gIE5vdGUgdGhhdCBDb3B5UmVhZEF0dHJpYnV0ZSByZXR1cm5zIGEgcG9pbnRlcgohICAq IHRvIGF0dHJpYnV0ZV9idWYncyBkYXRhIGJ1ZmZlciEKICAgKi8KISBzdGF0aWMgU3RyaW5n SW5mb0RhdGEgYXR0cmlidXRlX2J1ZjsKICAKICAvKgogICAqIFNpbWlsYXJseSwgbGluZV9i dWYgaG9sZHMgdGhlIHdob2xlIGlucHV0IGxpbmUgYmVpbmcgcHJvY2Vzc2VkIChpdHMKICAg KiBjdXJzb3IgZmllbGQgcG9pbnRzIHRvIHRoZSBuZXh0IGNoYXJhY3RlciB0byBiZSByZWFk IGJ5IENvcHlSZWFkQXR0cmlidXRlKS4KICAgKiBUaGUgaW5wdXQgY3ljbGUgaXMgZmlyc3Qg dG8gcmVhZCB0aGUgd2hvbGUgbGluZSBpbnRvIGxpbmVfYnVmLCBjb252ZXJ0IGl0CiAgICog dG8gc2VydmVyIGVuY29kaW5nLCBhbmQgdGhlbiBleHRyYWN0IGluZGl2aWR1YWwgYXR0cmli dXRlIGZpZWxkcyBpbnRvCiEgICogYXR0cmlidXRlX2J1Zi4gIChXZSB1c2VkIHRvIGhhdmUg Q29weVJlYWRBdHRyaWJ1dGUgcmVhZCB0aGUgaW5wdXQgc291cmNlCiAgICogZGlyZWN0bHks IGJ1dCB0aGF0IGNhdXNlZCBhIGxvdCBvZiBlbmNvZGluZyBpc3N1ZXMgYW5kIHVubmVjZXNz YXJ5IGxvZ2ljCiAgICogY29tcGxleGl0eS4pCiAgICovCi0tLSA5OCwxNDEgLS0tLQogIHN0 YXRpYyBpbnQJY29weV9saW5lbm87CQkvKiBsaW5lIG51bWJlciBmb3IgZXJyb3IgbWVzc2Fn ZXMgKi8KICBzdGF0aWMgY29uc3QgY2hhciAqY29weV9hdHRuYW1lOwkvKiBjdXJyZW50IGF0 dCBmb3IgZXJyb3IgbWVzc2FnZXMgKi8KICAKKyAvKgorICAqIFN0YXRpYyB2YXJpYWJsZXMg Zm9yIGJ1ZmZlcmVkIGlucHV0IHBhcnNpbmcgCisgICovCisgc3RhdGljIGNoYXIgaW5wdXRf YnVmW0NPUFlfQlVGX1NJWkUgKyAxXTsgLyogZXh0cmEgYnl0ZSBmb3IgJ1wwJyAqLworIHN0 YXRpYyBpbnQJYnVmZmVyX2luZGV4OwkJLyogaW5wdXQgYnVmZmVyIGluZGV4ICovCisgc3Rh dGljIGJvb2wgZW5kX21hcmtlcjsKKyBzdGF0aWMgY2hhciAqYmVnbG9jOworIHN0YXRpYyBj aGFyICplbmRsb2M7Cisgc3RhdGljIGJvb2wgYnVmX2RvbmU7CQkJLyogZmluaXNoZWQgcHJv Y2Vzc2luZyB0aGUgY3VycmVudCBidWZmZXIgKi8KKyBzdGF0aWMgYm9vbCBsaW5lX2RvbmU7 CQkJLyogZmluaXNoZWQgcHJvY2Vzc2luZyB0aGUgd2hvbGUgbGluZSBvcgorICAgICAgICAg ICAgICAgICAgICAgICAgICAgICAgICAgICogc3RvcHBlZCBpbiB0aGUgbWlkZGxlICovCisg LyogdGhlc2UgYXJlIGZvciBDU1YgZm9ybWF0ICovCisgc3RhdGljIGJvb2wJaW5fcXVvdGU7 Cisgc3RhdGljIGJvb2wgbGFzdF93YXNfZXNjOworIAorIC8qIHRoZXNlIGFyZSBmb3IgVEVY VCBmb3JtYXQgKi8KKyBzdGF0aWMgYm9vbCBlc2NfaW5fcHJldmJ1ZjsJCS8qIGVzY2FwZSB3 YXMgbGFzdCBjaGFyYWN0ZXIgb2YgdGhlIGRhdGEKKyAgICAgICAgICAgICAgICAgICAgICAg ICAgICAgICAgICAqIGlucHV0IGJ1ZmZlciAqLworIHN0YXRpYyBib29sIGNyX2luX3ByZXZi dWY7CQkvKiBDUiB3YXMgbGFzdCBjaGFyYWN0ZXIgb2YgdGhlIGRhdGEgaW5wdXQKKyAgICAg ICAgICAgICAgICAgICAgICAgICAgICAgICAgICAqIGJ1ZmZlciAqLworIAogIAogIC8qCiAg ICogVGhlc2Ugc3RhdGljIHZhcmlhYmxlcyBhcmUgdXNlZCB0byBhdm9pZCBpbmN1cnJpbmcg b3ZlcmhlYWQgZm9yIGVhY2gKISAgKiBhdHRyaWJ1dGUgcHJvY2Vzc2VkLiAgYXR0cl9idWYg aXMgcmV1c2VkIG9uIGVhY2ggQ29weVJlYWRBdHRyaWJ1dGUKICAgKiBjYWxsIHRvIGhvbGQg dGhlIHN0cmluZyBiZWluZyByZWFkIGluLiAgVW5kZXIgbm9ybWFsIHVzZSBpdCB3aWxsIHNv b24KICAgKiBncm93IHRvIGEgc3VpdGFibGUgc2l6ZSwgYW5kIHRoZW4gd2Ugd2lsbCBhdm9p ZCBwYWxsb2MvcGZyZWUgb3ZlcmhlYWQKICAgKiBmb3Igc3Vic2VxdWVudCBhdHRyaWJ1dGVz LiAgTm90ZSB0aGF0IENvcHlSZWFkQXR0cmlidXRlIHJldHVybnMgYSBwb2ludGVyCiEgICog dG8gYXR0cl9idWYncyBkYXRhIGJ1ZmZlciEKICAgKi8KISBzdGF0aWMgU3RyaW5nSW5mb0Rh dGEgYXR0cl9idWY7IAogIAogIC8qCiAgICogU2ltaWxhcmx5LCBsaW5lX2J1ZiBob2xkcyB0 aGUgd2hvbGUgaW5wdXQgbGluZSBiZWluZyBwcm9jZXNzZWQgKGl0cwogICAqIGN1cnNvciBm aWVsZCBwb2ludHMgdG8gdGhlIG5leHQgY2hhcmFjdGVyIHRvIGJlIHJlYWQgYnkgQ29weVJl YWRBdHRyaWJ1dGUpLgogICAqIFRoZSBpbnB1dCBjeWNsZSBpcyBmaXJzdCB0byByZWFkIHRo ZSB3aG9sZSBsaW5lIGludG8gbGluZV9idWYsIGNvbnZlcnQgaXQKICAgKiB0byBzZXJ2ZXIg ZW5jb2RpbmcsIGFuZCB0aGVuIGV4dHJhY3QgaW5kaXZpZHVhbCBhdHRyaWJ1dGUgZmllbGRz IGludG8KISAgKiBhdHRyX2J1Zi4gIChXZSB1c2VkIHRvIGhhdmUgQ29weVJlYWRBdHRyaWJ1 dGUgcmVhZCB0aGUgaW5wdXQgc291cmNlCiAgICogZGlyZWN0bHksIGJ1dCB0aGF0IGNhdXNl ZCBhIGxvdCBvZiBlbmNvZGluZyBpc3N1ZXMgYW5kIHVubmVjZXNzYXJ5IGxvZ2ljCiAgICog Y29tcGxleGl0eS4pCiAgICovCioqKioqKioqKioqKioqKgoqKiogMTM3LDE1NSAqKioqCiAg c3RhdGljIHZvaWQgQ29weUZyb20oUmVsYXRpb24gcmVsLCBMaXN0ICphdHRudW1saXN0LCBi b29sIGJpbmFyeSwgYm9vbCBvaWRzLAogICBjaGFyICpkZWxpbSwgY2hhciAqbnVsbF9wcmlu dCwgYm9vbCBjc3ZfbW9kZSwgY2hhciAqcXVvdGUsIGNoYXIgKmVzY2FwZSwKICAJCSBMaXN0 ICpmb3JjZV9ub3RudWxsX2F0dHMsIGJvb2wgaGVhZGVyX2xpbmUpOwohIHN0YXRpYyBib29s IENvcHlSZWFkTGluZShjaGFyICogcXVvdGUsIGNoYXIgKiBlc2NhcGUpOwohIHN0YXRpYyBj aGFyICpDb3B5UmVhZEF0dHJpYnV0ZShjb25zdCBjaGFyICpkZWxpbSwgY29uc3QgY2hhciAq bnVsbF9wcmludCwKISAJCQkJICBDb3B5UmVhZFJlc3VsdCAqcmVzdWx0LCBib29sICppc251 bGwpOwohIHN0YXRpYyBjaGFyICpDb3B5UmVhZEF0dHJpYnV0ZUNTVihjb25zdCBjaGFyICpk ZWxpbSwgY29uc3QgY2hhciAqbnVsbF9wcmludCwKISAJCQkJCSBjaGFyICpxdW90ZSwgY2hh ciAqZXNjYXBlLAohIAkJCQkJIENvcHlSZWFkUmVzdWx0ICpyZXN1bHQsIGJvb2wgKmlzbnVs bCk7CiAgc3RhdGljIERhdHVtIENvcHlSZWFkQmluYXJ5QXR0cmlidXRlKGludCBjb2x1bW5f bm8sIEZtZ3JJbmZvICpmbGluZm8sCiAgCQkJCQkJT2lkIHR5cGlvcGFyYW0sIGludDMyIHR5 cG1vZCwgYm9vbCAqaXNudWxsKTsKICBzdGF0aWMgdm9pZCBDb3B5QXR0cmlidXRlT3V0KGNo YXIgKnN0cmluZywgY2hhciAqZGVsaW0pOwogIHN0YXRpYyB2b2lkIENvcHlBdHRyaWJ1dGVP dXRDU1YoY2hhciAqc3RyaW5nLCBjaGFyICpkZWxpbSwgY2hhciAqcXVvdGUsCiAgCQkJCQlj aGFyICplc2NhcGUsIGJvb2wgZm9yY2VfcXVvdGUpOwogIHN0YXRpYyBMaXN0ICpDb3B5R2V0 QXR0bnVtcyhSZWxhdGlvbiByZWwsIExpc3QgKmF0dG5hbWVsaXN0KTsKICBzdGF0aWMgdm9p ZCBsaW1pdF9wcmludG91dF9sZW5ndGgoU3RyaW5nSW5mbyBidWYpOwogIAogIC8qIEludGVy bmFsIGNvbW11bmljYXRpb25zIGZ1bmN0aW9ucyAqLwogIHN0YXRpYyB2b2lkIFNlbmRDb3B5 QmVnaW4oYm9vbCBiaW5hcnksIGludCBuYXR0cyk7Ci0tLSAxNTIsMTc2IC0tLS0KICBzdGF0 aWMgdm9pZCBDb3B5RnJvbShSZWxhdGlvbiByZWwsIExpc3QgKmF0dG51bWxpc3QsIGJvb2wg YmluYXJ5LCBib29sIG9pZHMsCiAgIGNoYXIgKmRlbGltLCBjaGFyICpudWxsX3ByaW50LCBi b29sIGNzdl9tb2RlLCBjaGFyICpxdW90ZSwgY2hhciAqZXNjYXBlLAogIAkJIExpc3QgKmZv cmNlX25vdG51bGxfYXR0cywgYm9vbCBoZWFkZXJfbGluZSk7CiEgc3RhdGljIGJvb2wgQ29w eVJlYWRMaW5lVGV4dChzaXplX3QgYnl0ZXNyZWFkLCBjaGFyICplc2NhcGUpOwohIHN0YXRp YyBib29sIENvcHlSZWFkTGluZUNTVihzaXplX3QgYnl0ZXNyZWFkLCBjaGFyICpxdW90ZSwg Y2hhciAqZXNjYXBlKTsKISBzdGF0aWMgdm9pZCBDb3B5UmVhZEF0dHJpYnV0ZXNUZXh0KGNv bnN0IGNoYXIgKmRlbGltLCBjb25zdCBjaGFyICplc2NhcGUsIGNvbnN0IGNoYXIgKm51bGxf cHJpbnQsCiEgCQkJCQkJCQkgICBpbnQgbnVsbF9wcmludF9sZW4sIGNoYXIgKm51bGxzLCBM aXN0ICphdHRudW1saXN0LCAKISAJCQkJCQkJCSAgIGludCAqYXR0cl9vZmZzZXRzLCBpbnQg bnVtX3BoeXNfYXR0cnMsIEZvcm1fcGdfYXR0cmlidXRlICphdHRyKTsKISBzdGF0aWMgdm9p ZCBDb3B5UmVhZEF0dHJpYnV0ZXNDU1YoY29uc3QgY2hhciAqZGVsaW0sIGNvbnN0IGNoYXIg Km51bGxfcHJpbnQsIGNoYXIgKnF1b3RlLAohIAkJCQkJCQkJICBjaGFyICplc2NhcGUsIGlu dCBudWxsX3ByaW50X2xlbiwgY2hhciAqbnVsbHMsIExpc3QgKmF0dG51bWxpc3QsIAohIAkJ CQkJCQkJICBpbnQgKmF0dHJfb2Zmc2V0cywgaW50IG51bV9waHlzX2F0dHJzLCBGb3JtX3Bn X2F0dHJpYnV0ZSAqYXR0cik7CiAgc3RhdGljIERhdHVtIENvcHlSZWFkQmluYXJ5QXR0cmli dXRlKGludCBjb2x1bW5fbm8sIEZtZ3JJbmZvICpmbGluZm8sCiAgCQkJCQkJT2lkIHR5cGlv cGFyYW0sIGludDMyIHR5cG1vZCwgYm9vbCAqaXNudWxsKTsKKyBzdGF0aWMgY2hhciAqQ29w eVJlYWRPaWRBdHRyKGNvbnN0IGNoYXIgKmRlbGltLCBjb25zdCBjaGFyICpudWxsX3ByaW50 LCBpbnQgbnVsbF9wcmludF9sZW4sCisgCQkJCSAgICAgICAgICAgICBib29sICppc251bGwp OwogIHN0YXRpYyB2b2lkIENvcHlBdHRyaWJ1dGVPdXQoY2hhciAqc3RyaW5nLCBjaGFyICpk ZWxpbSk7CiAgc3RhdGljIHZvaWQgQ29weUF0dHJpYnV0ZU91dENTVihjaGFyICpzdHJpbmcs IGNoYXIgKmRlbGltLCBjaGFyICpxdW90ZSwKICAJCQkJCWNoYXIgKmVzY2FwZSwgYm9vbCBm b3JjZV9xdW90ZSk7CiAgc3RhdGljIExpc3QgKkNvcHlHZXRBdHRudW1zKFJlbGF0aW9uIHJl bCwgTGlzdCAqYXR0bmFtZWxpc3QpOwogIHN0YXRpYyB2b2lkIGxpbWl0X3ByaW50b3V0X2xl bmd0aChTdHJpbmdJbmZvIGJ1Zik7Cisgc3RhdGljIGJvb2wgRGV0ZWN0TGluZUVuZChzaXpl X3QgYnl0ZXNyZWFkLCBjaGFyICpxdW90ZSwgY2hhciAqZXNjYXBlKTsKKyAKICAKICAvKiBJ bnRlcm5hbCBjb21tdW5pY2F0aW9ucyBmdW5jdGlvbnMgKi8KICBzdGF0aWMgdm9pZCBTZW5k Q29weUJlZ2luKGJvb2wgYmluYXJ5LCBpbnQgbmF0dHMpOwoqKioqKioqKioqKioqKioKKioq IDE1OSwxNzUgKioqKgogIHN0YXRpYyB2b2lkIENvcHlTZW5kU3RyaW5nKGNvbnN0IGNoYXIg KnN0cik7CiAgc3RhdGljIHZvaWQgQ29weVNlbmRDaGFyKGNoYXIgYyk7CiAgc3RhdGljIHZv aWQgQ29weVNlbmRFbmRPZlJvdyhib29sIGJpbmFyeSk7CiEgc3RhdGljIHZvaWQgQ29weUdl dERhdGEodm9pZCAqZGF0YWJ1ZiwgaW50IGRhdGFzaXplKTsKISBzdGF0aWMgaW50CUNvcHlH ZXRDaGFyKHZvaWQpOwogIAogICNkZWZpbmUgQ29weUdldEVvZigpICAoZmVfZW9mKQotIHN0 YXRpYyBpbnQJQ29weVBlZWtDaGFyKHZvaWQpOwotIHN0YXRpYyB2b2lkIENvcHlEb25lUGVl ayhpbnQgYywgYm9vbCBwaWNrdXApOwogIHN0YXRpYyB2b2lkIENvcHlTZW5kSW50MzIoaW50 MzIgdmFsKTsKICBzdGF0aWMgaW50MzIgQ29weUdldEludDMyKHZvaWQpOwogIHN0YXRpYyB2 b2lkIENvcHlTZW5kSW50MTYoaW50MTYgdmFsKTsKICBzdGF0aWMgaW50MTYgQ29weUdldElu dDE2KHZvaWQpOwogIAogIAogIC8qCiAgICogU2VuZCBjb3B5IHN0YXJ0L3N0b3AgbWVzc2Fn ZXMgZm9yIGZyb250ZW5kIGNvcGllcy4gIFRoZXNlIGhhdmUgY2hhbmdlZAotLS0gMTgwLDE5 NyAtLS0tCiAgc3RhdGljIHZvaWQgQ29weVNlbmRTdHJpbmcoY29uc3QgY2hhciAqc3RyKTsK ICBzdGF0aWMgdm9pZCBDb3B5U2VuZENoYXIoY2hhciBjKTsKICBzdGF0aWMgdm9pZCBDb3B5 U2VuZEVuZE9mUm93KGJvb2wgYmluYXJ5KTsKISBzdGF0aWMgaW50CUNvcHlHZXREYXRhKHZv aWQgKmRhdGFidWYsIGludCBkYXRhc2l6ZSk7CiAgCiAgI2RlZmluZSBDb3B5R2V0RW9mKCkg IChmZV9lb2YpCiAgc3RhdGljIHZvaWQgQ29weVNlbmRJbnQzMihpbnQzMiB2YWwpOwogIHN0 YXRpYyBpbnQzMiBDb3B5R2V0SW50MzIodm9pZCk7CiAgc3RhdGljIHZvaWQgQ29weVNlbmRJ bnQxNihpbnQxNiB2YWwpOwogIHN0YXRpYyBpbnQxNiBDb3B5R2V0SW50MTYodm9pZCk7CiAg CisgLyogYnl0ZSBzY2FuaW5nIHV0aWxzICovCisgc3RhdGljIGNoYXIgKnNjYW5UZXh0TGlu ZShjb25zdCBjaGFyICpzLCBjaGFyIGMsIHNpemVfdCBsZW4pOworIHN0YXRpYyBjaGFyICpz Y2FuQ1NWTGluZShjb25zdCBjaGFyICpzLCBjaGFyIGMxLCBjaGFyIGMyLCBjaGFyIGMzLCBz aXplX3QgbGVuKTsKKyBzdGF0aWMgY2hhciAqc2NhblRleHRBdHRyKGNvbnN0IGNoYXIgKnMs IGNoYXIgYzEsIGNoYXIgYzIsIHNpemVfdCBsZW4pOwogIAogIC8qCiAgICogU2VuZCBjb3B5 IHN0YXJ0L3N0b3AgbWVzc2FnZXMgZm9yIGZyb250ZW5kIGNvcGllcy4gIFRoZXNlIGhhdmUg Y2hhbmdlZAoqKioqKioqKioqKioqKioKKioqIDM4MiwzOTUgKioqKgogICAqIEl0IHNlZW1z IHVud2lzZSB0byBhbGxvdyB0aGUgQ09QWSBJTiB0byBjb21wbGV0ZSBub3JtYWxseSBpbiB0 aGF0IGNhc2UuCiAgICoKICAgKiBOQjogbm8gZGF0YSBjb252ZXJzaW9uIGlzIGFwcGxpZWQg YnkgdGhlc2UgZnVuY3Rpb25zCiAgICovCiEgc3RhdGljIHZvaWQKICBDb3B5R2V0RGF0YSh2 b2lkICpkYXRhYnVmLCBpbnQgZGF0YXNpemUpCiAgewogIAlzd2l0Y2ggKGNvcHlfZGVzdCkK ICAJewogIAkJY2FzZSBDT1BZX0ZJTEU6CiEgCQkJZnJlYWQoZGF0YWJ1ZiwgZGF0YXNpemUs IDEsIGNvcHlfZmlsZSk7CiAgCQkJaWYgKGZlb2YoY29weV9maWxlKSkKICAJCQkJZmVfZW9m ID0gdHJ1ZTsKICAJCQlicmVhazsKLS0tIDQwNCw0MjIgLS0tLQogICAqIEl0IHNlZW1zIHVu d2lzZSB0byBhbGxvdyB0aGUgQ09QWSBJTiB0byBjb21wbGV0ZSBub3JtYWxseSBpbiB0aGF0 IGNhc2UuCiAgICoKICAgKiBOQjogbm8gZGF0YSBjb252ZXJzaW9uIGlzIGFwcGxpZWQgYnkg dGhlc2UgZnVuY3Rpb25zCisgICoKKyAgKiBSZXR1cm5zOiB0aGUgbnVtYmVyIG9mIGJ5dGVz IHRoYXQgd2VyZSBzdWNjZXNzZnVsbHkgcmVhZAorICAqIGludG8gdGhlIGRhdGEgYnVmZmVy LgogICAqLwohIHN0YXRpYyBpbnQKICBDb3B5R2V0RGF0YSh2b2lkICpkYXRhYnVmLCBpbnQg ZGF0YXNpemUpCiAgeworIAlzaXplX3QJCWJ5dGVzcmVhZCA9IDA7CisgCiAgCXN3aXRjaCAo Y29weV9kZXN0KQogIAl7CiAgCQljYXNlIENPUFlfRklMRToKISAJCQlieXRlc3JlYWQgPSBm cmVhZChkYXRhYnVmLCAxLCBkYXRhc2l6ZSwgY29weV9maWxlKTsKICAJCQlpZiAoZmVvZihj b3B5X2ZpbGUpKQogIAkJCQlmZV9lb2YgPSB0cnVlOwogIAkJCWJyZWFrOwoqKioqKioqKioq KioqKioKKioqIDQwMSw0MDYgKioqKgotLS0gNDI4LDQzNSAtLS0tCiAgCQkJCQkJKGVycmNv ZGUoRVJSQ09ERV9DT05ORUNUSU9OX0ZBSUxVUkUpLAogIAkJCQkJCSBlcnJtc2coInVuZXhw ZWN0ZWQgRU9GIG9uIGNsaWVudCBjb25uZWN0aW9uIikpKTsKICAJCQl9CisgCQkJYnl0ZXNy ZWFkICs9IGRhdGFzaXplOwkJLyogdXBkYXRlIHRoZSBjb3VudCBvZiBieXRlcyB0aGF0IHdl cmUKKyAJCQkJCQkJCQkJICogcmVhZCBzbyBmYXIgKi8KICAJCQlicmVhazsKICAJCWNhc2Ug Q09QWV9ORVdfRkU6CiAgCQkJd2hpbGUgKGRhdGFzaXplID4gMCAmJiAhZmVfZW9mKQoqKioq KioqKioqKioqKioKKioqIDQyOSw0MzUgKioqKgogIAkJCQkJCWNhc2UgJ2MnOgkJLyogQ29w eURvbmUgKi8KICAJCQkJCQkJLyogQ09QWSBJTiBjb3JyZWN0bHkgdGVybWluYXRlZCBieSBm cm9udGVuZCAqLwogIAkJCQkJCQlmZV9lb2YgPSB0cnVlOwohIAkJCQkJCQlyZXR1cm47CiAg CQkJCQkJY2FzZSAnZic6CQkvKiBDb3B5RmFpbCAqLwogIAkJCQkJCQllcmVwb3J0KEVSUk9S LAogIAkJCQkJCQkJCShlcnJjb2RlKEVSUkNPREVfUVVFUllfQ0FOQ0VMRUQpLAotLS0gNDU4 LDQ2NCAtLS0tCiAgCQkJCQkJY2FzZSAnYyc6CQkvKiBDb3B5RG9uZSAqLwogIAkJCQkJCQkv KiBDT1BZIElOIGNvcnJlY3RseSB0ZXJtaW5hdGVkIGJ5IGZyb250ZW5kICovCiAgCQkJCQkJ CWZlX2VvZiA9IHRydWU7CiEgCQkJCQkJCXJldHVybiBieXRlc3JlYWQ7CiAgCQkJCQkJY2Fz ZSAnZic6CQkvKiBDb3B5RmFpbCAqLwogIAkJCQkJCQllcmVwb3J0KEVSUk9SLAogIAkJCQkJ CQkJCShlcnJjb2RlKEVSUkNPREVfUVVFUllfQ0FOQ0VMRUQpLAoqKioqKioqKioqKioqKioK KioqIDQ1OSw1OTggKioqKgogIAkJCQkJYXZhaWwgPSBkYXRhc2l6ZTsKICAJCQkJcHFfY29w eW1zZ2J5dGVzKGNvcHlfbXNnYnVmLCBkYXRhYnVmLCBhdmFpbCk7CiAgCQkJCWRhdGFidWYg PSAodm9pZCAqKSAoKGNoYXIgKikgZGF0YWJ1ZiArIGF2YWlsKTsKICAJCQkJZGF0YXNpemUg LT0gYXZhaWw7CiAgCQkJfQogIAkJCWJyZWFrOwogIAl9Ci0gfQotIAotIHN0YXRpYyBpbnQK LSBDb3B5R2V0Q2hhcih2b2lkKQotIHsKLSAJaW50CQkJY2g7Ci0gCi0gCXN3aXRjaCAoY29w eV9kZXN0KQotIAl7Ci0gCQljYXNlIENPUFlfRklMRToKLSAJCQljaCA9IGdldGMoY29weV9m aWxlKTsKLSAJCQlicmVhazsKLSAJCWNhc2UgQ09QWV9PTERfRkU6Ci0gCQkJY2ggPSBwcV9n ZXRieXRlKCk7Ci0gCQkJaWYgKGNoID09IEVPRikKLSAJCQl7Ci0gCQkJCS8qIE9ubHkgYSBc LiB0ZXJtaW5hdG9yIGlzIGxlZ2FsIEVPRiBpbiBvbGQgcHJvdG9jb2wgKi8KLSAJCQkJZXJl cG9ydChFUlJPUiwKLSAJCQkJCQkoZXJyY29kZShFUlJDT0RFX0NPTk5FQ1RJT05fRkFJTFVS RSksCi0gCQkJCQkJIGVycm1zZygidW5leHBlY3RlZCBFT0Ygb24gY2xpZW50IGNvbm5lY3Rp b24iKSkpOwotIAkJCX0KLSAJCQlicmVhazsKLSAJCWNhc2UgQ09QWV9ORVdfRkU6Ci0gCQkJ ewotIAkJCQl1bnNpZ25lZCBjaGFyIGNjOwotIAotIAkJCQlDb3B5R2V0RGF0YSgmY2MsIDEp OwotIAkJCQlpZiAoZmVfZW9mKQotIAkJCQkJY2ggPSBFT0Y7Ci0gCQkJCWVsc2UKLSAJCQkJ CWNoID0gY2M7Ci0gCQkJCWJyZWFrOwotIAkJCX0KLSAJCWRlZmF1bHQ6Ci0gCQkJY2ggPSBF T0Y7Ci0gCQkJYnJlYWs7Ci0gCX0KLSAJaWYgKGNoID09IEVPRikKLSAJCWZlX2VvZiA9IHRy dWU7Ci0gCXJldHVybiBjaDsKLSB9Ci0gCi0gLyoKLSAgKiBDb3B5UGVla0NoYXIgcmVhZHMg YSBieXRlIGluICJwZWVrYWJsZSIgbW9kZS4KLSAgKgotICAqIGFmdGVyIGVhY2ggY2FsbCB0 byBDb3B5UGVla0NoYXIsIGEgY2FsbCB0byBDb3B5RG9uZVBlZWsgX211c3RfCi0gICogZm9s bG93LCB1bmxlc3MgRU9GIHdhcyByZXR1cm5lZC4KLSAgKgotICAqIENvcHlEb25lUGVlayB3 aWxsIGVpdGhlciB0YWtlIHRoZSBwZWVrZWQgY2hhciBvZmYgdGhlIHN0cmVhbQotICAqIChp ZiBwaWNrdXAgaXMgdHJ1ZSkgb3IgbGVhdmUgaXQgb24gdGhlIHN0cmVhbSAoaWYgcGlja3Vw IGlzIGZhbHNlKS4KLSAgKi8KLSBzdGF0aWMgaW50Ci0gQ29weVBlZWtDaGFyKHZvaWQpCi0g ewotIAlpbnQJCQljaDsKLSAKLSAJc3dpdGNoIChjb3B5X2Rlc3QpCi0gCXsKLSAJCWNhc2Ug Q09QWV9GSUxFOgotIAkJCWNoID0gZ2V0Yyhjb3B5X2ZpbGUpOwotIAkJCWJyZWFrOwotIAkJ Y2FzZSBDT1BZX09MRF9GRToKLSAJCQljaCA9IHBxX3BlZWtieXRlKCk7Ci0gCQkJaWYgKGNo ID09IEVPRikKLSAJCQl7Ci0gCQkJCS8qIE9ubHkgYSBcLiB0ZXJtaW5hdG9yIGlzIGxlZ2Fs IEVPRiBpbiBvbGQgcHJvdG9jb2wgKi8KLSAJCQkJZXJlcG9ydChFUlJPUiwKLSAJCQkJCQko ZXJyY29kZShFUlJDT0RFX0NPTk5FQ1RJT05fRkFJTFVSRSksCi0gCQkJCQkJIGVycm1zZygi dW5leHBlY3RlZCBFT0Ygb24gY2xpZW50IGNvbm5lY3Rpb24iKSkpOwotIAkJCX0KLSAJCQli cmVhazsKLSAJCWNhc2UgQ09QWV9ORVdfRkU6Ci0gCQkJewotIAkJCQl1bnNpZ25lZCBjaGFy IGNjOwotIAotIAkJCQlDb3B5R2V0RGF0YSgmY2MsIDEpOwotIAkJCQlpZiAoZmVfZW9mKQot IAkJCQkJY2ggPSBFT0Y7Ci0gCQkJCWVsc2UKLSAJCQkJCWNoID0gY2M7Ci0gCQkJCWJyZWFr OwotIAkJCX0KLSAJCWRlZmF1bHQ6Ci0gCQkJY2ggPSBFT0Y7Ci0gCQkJYnJlYWs7Ci0gCX0K LSAJaWYgKGNoID09IEVPRikKLSAJCWZlX2VvZiA9IHRydWU7Ci0gCXJldHVybiBjaDsKLSB9 Ci0gCi0gc3RhdGljIHZvaWQKLSBDb3B5RG9uZVBlZWsoaW50IGMsIGJvb2wgcGlja3VwKQot IHsKLSAJaWYgKGZlX2VvZikKLSAJCXJldHVybjsJCQkJCS8qIGNhbid0IHVuZ2V0IGFuIEVP RiAqLwotIAlzd2l0Y2ggKGNvcHlfZGVzdCkKLSAJewotIAkJY2FzZSBDT1BZX0ZJTEU6Ci0g CQkJaWYgKCFwaWNrdXApCi0gCQkJewotIAkJCQkvKiBXZSBkb24ndCB3YW50IHRvIHBpY2sg aXQgdXAgLSBzbyBwdXQgaXQgYmFjayBpbiB0aGVyZSAqLwotIAkJCQl1bmdldGMoYywgY29w eV9maWxlKTsKLSAJCQl9Ci0gCQkJLyogSWYgd2Ugd2FudGVkIHRvIHBpY2sgaXQgdXAsIGl0 J3MgYWxyZWFkeSBkb25lICovCi0gCQkJYnJlYWs7Ci0gCQljYXNlIENPUFlfT0xEX0ZFOgot IAkJCWlmIChwaWNrdXApCi0gCQkJewotIAkJCQkvKiBXZSB3YW50IHRvIHBpY2sgaXQgdXAg Ki8KLSAJCQkJKHZvaWQpIHBxX2dldGJ5dGUoKTsKLSAJCQl9CiAgCiEgCQkJLyoKISAJCQkg KiBJZiB3ZSBkaWRuJ3Qgd2FudCB0byBwaWNrIGl0IHVwLCBqdXN0IGxlYXZlIGl0IHdoZXJl IGl0CiEgCQkJICogc2l0cwohIAkJCSAqLwohIAkJCWJyZWFrOwohIAkJY2FzZSBDT1BZX05F V19GRToKISAJCQlpZiAoIXBpY2t1cCkKISAJCQl7CiEgCQkJCS8qIFdlIGRvbid0IHdhbnQg dG8gcGljayBpdCB1cCAtIHNvIHB1dCBpdCBiYWNrIGluIHRoZXJlICovCiEgCQkJCWNvcHlf bXNnYnVmLT5jdXJzb3ItLTsKISAJCQl9CiEgCQkJLyogSWYgd2Ugd2FudGVkIHRvIHBpY2sg aXQgdXAsIGl0J3MgYWxyZWFkeSBkb25lICovCiEgCQkJYnJlYWs7CiEgCX0KICB9CiAgCi0g CiAgLyoKICAgKiBUaGVzZSBmdW5jdGlvbnMgZG8gYXBwbHkgc29tZSBkYXRhIGNvbnZlcnNp b24KICAgKi8KLS0tIDQ4OCw1MDMgLS0tLQogIAkJCQkJYXZhaWwgPSBkYXRhc2l6ZTsKICAJ CQkJcHFfY29weW1zZ2J5dGVzKGNvcHlfbXNnYnVmLCBkYXRhYnVmLCBhdmFpbCk7CiAgCQkJ CWRhdGFidWYgPSAodm9pZCAqKSAoKGNoYXIgKikgZGF0YWJ1ZiArIGF2YWlsKTsKKyAJCQkJ Ynl0ZXNyZWFkICs9IGF2YWlsOwkJLyogdXBkYXRlIHRoZSBjb3VudCBvZiBieXRlcyB0aGF0 IHdlcmUKKyAJCQkJCQkJCQkJICogcmVhZCBzbyBmYXIgKi8KICAJCQkJZGF0YXNpemUgLT0g YXZhaWw7CiAgCQkJfQogIAkJCWJyZWFrOwogIAl9CiAgCiEgCXJldHVybiBieXRlc3JlYWQ7 CiAgfQogIAogIC8qCiAgICogVGhlc2UgZnVuY3Rpb25zIGRvIGFwcGx5IHNvbWUgZGF0YSBj b252ZXJzaW9uCiAgICovCioqKioqKioqKioqKioqKgoqKiogOTY4LDk4MCAqKioqCiAgCX0K ICAKICAJLyogU2V0IHVwIHZhcmlhYmxlcyB0byBhdm9pZCBwZXItYXR0cmlidXRlIG92ZXJo ZWFkLiAqLwohIAlpbml0U3RyaW5nSW5mbygmYXR0cmlidXRlX2J1Zik7CiAgCWluaXRTdHJp bmdJbmZvKCZsaW5lX2J1Zik7CiAgCWxpbmVfYnVmX2NvbnZlcnRlZCA9IGZhbHNlOwogIAog IAljbGllbnRfZW5jb2RpbmcgPSBwZ19nZXRfY2xpZW50X2VuY29kaW5nKCk7CiAgCXNlcnZl cl9lbmNvZGluZyA9IEdldERhdGFiYXNlRW5jb2RpbmcoKTsKICAKICAJY29weV9kZXN0ID0g Q09QWV9GSUxFOwkJLyogZGVmYXVsdCAqLwogIAljb3B5X2ZpbGUgPSBOVUxMOwogIAljb3B5 X21zZ2J1ZiA9IE5VTEw7Ci0tLSA4NzMsOTA1IC0tLS0KICAJfQogIAogIAkvKiBTZXQgdXAg dmFyaWFibGVzIHRvIGF2b2lkIHBlci1hdHRyaWJ1dGUgb3ZlcmhlYWQuICovCiEgCWluaXRT dHJpbmdJbmZvKCZhdHRyX2J1Zik7CiAgCWluaXRTdHJpbmdJbmZvKCZsaW5lX2J1Zik7CiAg CWxpbmVfYnVmX2NvbnZlcnRlZCA9IGZhbHNlOwogIAogIAljbGllbnRfZW5jb2RpbmcgPSBw Z19nZXRfY2xpZW50X2VuY29kaW5nKCk7CiAgCXNlcnZlcl9lbmNvZGluZyA9IEdldERhdGFi YXNlRW5jb2RpbmcoKTsKICAKKyAJLyoKKyAJICogY2hlY2sgaWYgdGhlIGNsaWVudCBlbmNv ZGluZyBpcyBvbmUgb2YgdGhlIDUgZW5jb2RpbmdzCisgCSAqIHRoYXQgYXJlIG5vdCBzdXBw b3J0ZWQgYXMgYSBzZXJ2ZXIgZW5jb2RpbmdzLgorIAkgKi8KKyAJc3dpdGNoIChjbGllbnRf ZW5jb2RpbmcpCisgCXsKKyAJCWNhc2UgUEdfU0pJUzoKKyAJCWNhc2UgUEdfQklHNToKKyAJ CWNhc2UgUEdfR0JLOgorIAkJY2FzZSBQR19VSEM6CisgCQljYXNlIFBHX0dCMTgwMzA6Cisg CQkJY2xpZW50X2VuY29kaW5nX29ubHkgPSB0cnVlOworIAkJCWJyZWFrOworIAkJZGVmYXVs dDoKKyAJCQljbGllbnRfZW5jb2Rpbmdfb25seSA9IGZhbHNlOworIAl9CisgCisgCWlmKCFj c3ZfbW9kZSkKKyAJICAgIGVzY2FwZSA9ICJcXCI7CQkvKiBkZWZhdWx0IGZvciB0ZXh0IGZv cm1hdCAqLworIAogIAljb3B5X2Rlc3QgPSBDT1BZX0ZJTEU7CQkvKiBkZWZhdWx0ICovCiAg CWNvcHlfZmlsZSA9IE5VTEw7CiAgCWNvcHlfbXNnYnVmID0gTlVMTDsKKioqKioqKioqKioq KioqCioqKiAxMTA1LDExMTMgKioqKgogIAkJCQkJIGVycm1zZygiY291bGQgbm90IHdyaXRl IHRvIGZpbGUgXCIlc1wiOiAlbSIsCiAgCQkJCQkJCWZpbGVuYW1lKSkpOwogIAl9CiEgCXBm cmVlKGF0dHJpYnV0ZV9idWYuZGF0YSk7CiAgCXBmcmVlKGxpbmVfYnVmLmRhdGEpOwogIAog IAkvKgogIAkgKiBDbG9zZSB0aGUgcmVsYXRpb24uCUlmIHJlYWRpbmcsIHdlIGNhbiByZWxl YXNlIHRoZSBBY2Nlc3NTaGFyZUxvY2sKICAJICogd2UgZ290OyBpZiB3cml0aW5nLCB3ZSBz aG91bGQgaG9sZCB0aGUgbG9jayB1bnRpbCBlbmQgb2YKLS0tIDEwMzAsMTAzOSAtLS0tCiAg CQkJCQkgZXJybXNnKCJjb3VsZCBub3Qgd3JpdGUgdG8gZmlsZSBcIiVzXCI6ICVtIiwKICAJ CQkJCQkJZmlsZW5hbWUpKSk7CiAgCX0KISAJcGZyZWUoYXR0cl9idWYuZGF0YSk7CiAgCXBm cmVlKGxpbmVfYnVmLmRhdGEpOwogIAorIAogIAkvKgogIAkgKiBDbG9zZSB0aGUgcmVsYXRp b24uCUlmIHJlYWRpbmcsIHdlIGNhbiByZWxlYXNlIHRoZSBBY2Nlc3NTaGFyZUxvY2sKICAJ ICogd2UgZ290OyBpZiB3cml0aW5nLCB3ZSBzaG91bGQgaG9sZCB0aGUgbG9jayB1bnRpbCBl bmQgb2YKKioqKioqKioqKioqKioqCioqKiAxMzQ1LDEzNTEgKioqKgogIAkJCQl7CiAgCQkJ CQlieXRlYQkgICAqb3V0cHV0Ynl0ZXM7CiAgCiEgCQkJCQlvdXRwdXRieXRlcyA9IERhdHVt R2V0Qnl0ZWFQKEZ1bmN0aW9uQ2FsbDEoJm91dF9mdW5jdGlvbnNbYXR0bnVtIC0gMV0sCiAg CQkJCQkJCQkJCQkJCQkJICAgdmFsdWUpKTsKICAJCQkJCS8qIFdlIGFzc3VtZSB0aGUgcmVz dWx0IHdpbGwgbm90IGhhdmUgYmVlbiB0b2FzdGVkICovCiAgCQkJCQlDb3B5U2VuZEludDMy KFZBUlNJWkUob3V0cHV0Ynl0ZXMpIC0gVkFSSERSU1opOwotLS0gMTI3MSwxMjc4IC0tLS0K ICAJCQkJewogIAkJCQkJYnl0ZWEJICAgKm91dHB1dGJ5dGVzOwogIAohIAkJCQkJb3V0cHV0 Ynl0ZXMgPQohIAkJCQkJCURhdHVtR2V0Qnl0ZWFQKEZ1bmN0aW9uQ2FsbDEoJm91dF9mdW5j dGlvbnNbYXR0bnVtIC0gMV0sCiAgCQkJCQkJCQkJCQkJCQkJICAgdmFsdWUpKTsKICAJCQkJ CS8qIFdlIGFzc3VtZSB0aGUgcmVzdWx0IHdpbGwgbm90IGhhdmUgYmVlbiB0b2FzdGVkICov CiAgCQkJCQlDb3B5U2VuZEludDMyKFZBUlNJWkUob3V0cHV0Ynl0ZXMpIC0gVkFSSERSU1op OwoqKioqKioqKioqKioqKioKKioqIDEzOTUsMTQwNCAqKioqCiAgCQlpZiAoY29weV9hdHRu YW1lKQogIAkJewogIAkJCS8qIGVycm9yIGlzIHJlbGV2YW50IHRvIGEgcGFydGljdWxhciBj b2x1bW4gKi8KISAJCQlsaW1pdF9wcmludG91dF9sZW5ndGgoJmF0dHJpYnV0ZV9idWYpOwog IAkJCWVycmNvbnRleHQoIkNPUFkgJXMsIGxpbmUgJWQsIGNvbHVtbiAlczogXCIlc1wiIiwK ICAJCQkJCSAgIGNvcHlfcmVsbmFtZSwgY29weV9saW5lbm8sIGNvcHlfYXR0bmFtZSwKISAJ CQkJCSAgIGF0dHJpYnV0ZV9idWYuZGF0YSk7CiAgCQl9CiAgCQllbHNlCiAgCQl7Ci0tLSAx MzIyLDEzMzEgLS0tLQogIAkJaWYgKGNvcHlfYXR0bmFtZSkKICAJCXsKICAJCQkvKiBlcnJv ciBpcyByZWxldmFudCB0byBhIHBhcnRpY3VsYXIgY29sdW1uICovCiEgCQkJbGltaXRfcHJp bnRvdXRfbGVuZ3RoKCZhdHRyX2J1Zik7CiAgCQkJZXJyY29udGV4dCgiQ09QWSAlcywgbGlu ZSAlZCwgY29sdW1uICVzOiBcIiVzXCIiLAogIAkJCQkJICAgY29weV9yZWxuYW1lLCBjb3B5 X2xpbmVubywgY29weV9hdHRuYW1lLAohIAkJCQkJICAgYXR0cl9idWYuZGF0YSk7CiAgCQl9 CiAgCQllbHNlCiAgCQl7CioqKioqKioqKioqKioqKgoqKiogMTQwNiwxNDExICoqKioKLS0t IDEzMzMsMTM0MCAtLS0tCiAgCQkJaWYgKGxpbmVfYnVmX2NvbnZlcnRlZCB8fAogIAkJCQlj bGllbnRfZW5jb2RpbmcgPT0gc2VydmVyX2VuY29kaW5nKQogIAkJCXsKKyAJCQkJLyogU3Ry aXAgb2ZmIHRoZSBuZXdsaW5lICovCisgCQkJCSoobGluZV9idWYuZGF0YSArIGxpbmVfYnVm LmxlbiAtIDEpID0gJ1wwJzsKICAJCQkJbGltaXRfcHJpbnRvdXRfbGVuZ3RoKCZsaW5lX2J1 Zik7CiAgCQkJCWVycmNvbnRleHQoIkNPUFkgJXMsIGxpbmUgJWQ6IFwiJXNcIiIsCiAgCQkJ CQkJICAgY29weV9yZWxuYW1lLCBjb3B5X2xpbmVubywKKioqKioqKioqKioqKioqCioqKiAx NDg1LDE0OTEgKioqKgogIAlPaWQJCQlpbl9mdW5jX29pZDsKICAJRGF0dW0JICAgKnZhbHVl czsKICAJY2hhcgkgICAqbnVsbHM7Ci0gCWJvb2wJCWRvbmUgPSBmYWxzZTsKICAJYm9vbAkJ aXNudWxsOwogIAlSZXN1bHRSZWxJbmZvICpyZXN1bHRSZWxJbmZvOwogIAlFU3RhdGUJICAg KmVzdGF0ZSA9IENyZWF0ZUV4ZWN1dG9yU3RhdGUoKTsgLyogZm9yIEV4ZWNDb25zdHJhaW50 cygpICovCi0tLSAxNDE0LDE0MTkgLS0tLQoqKioqKioqKioqKioqKioKKioqIDE0OTYsMTUw MSAqKioqCi0tLSAxNDI0LDE0MzQgLS0tLQogIAlFeHByQ29udGV4dCAqZWNvbnRleHQ7CQkv KiB1c2VkIGZvciBFeGVjRXZhbEV4cHIgZm9yIGRlZmF1bHQgYXR0cyAqLwogIAlNZW1vcnlD b250ZXh0IG9sZGNvbnRleHQgPSBDdXJyZW50TWVtb3J5Q29udGV4dDsKICAJRXJyb3JDb250 ZXh0Q2FsbGJhY2sgZXJyY29udGV4dDsKKyAgICAgaW50CQkgICAqYXR0cl9vZmZzZXRzOwor IAlpbnQJCQludWxsX3ByaW50X2xlbjsgLyogbGVuZ3RoIG9mIG51bGwgcHJpbnQgKi8KKyAJ Ym9vbAkJbm9fbW9yZV9kYXRhOworIAlMaXN0Q2VsbCAgICpjdXI7CisgCiAgCiAgCXR1cERl c2MgPSBSZWxhdGlvbkdldERlc2NyKHJlbCk7CiAgCWF0dHIgPSB0dXBEZXNjLT5hdHRyczsK KioqKioqKioqKioqKioqCioqKiAxNjA3LDE2MTMgKioqKgogIAkJfQogIAl9CiAgCiEgCS8q IFByZXBhcmUgdG8gY2F0Y2ggQUZURVIgdHJpZ2dlcnMuICovCiAgCUFmdGVyVHJpZ2dlckJl Z2luUXVlcnkoKTsKICAKICAJLyoKLS0tIDE1NDAsMTU0OCAtLS0tCiAgCQl9CiAgCX0KICAK ISAJLyoKISAJICogUHJlcGFyZSB0byBjYXRjaCBBRlRFUiB0cmlnZ2Vycy4KISAJICovCiAg CUFmdGVyVHJpZ2dlckJlZ2luUXVlcnkoKTsKICAKICAJLyoKKioqKioqKioqKioqKioqCioq KiAxNjcxLDE2NzYgKioqKgotLS0gMTYwNiwxNjEyIC0tLS0KICAKICAJdmFsdWVzID0gKERh dHVtICopIHBhbGxvYyhudW1fcGh5c19hdHRycyAqIHNpemVvZihEYXR1bSkpOwogIAludWxs cyA9IChjaGFyICopIHBhbGxvYyhudW1fcGh5c19hdHRycyAqIHNpemVvZihjaGFyKSk7Cisg CWF0dHJfb2Zmc2V0cyA9IChpbnQgKikgcGFsbG9jKG51bV9waHlzX2F0dHJzICogc2l6ZW9m KGludCkpOwogIAogIAkvKiBNYWtlIHJvb20gZm9yIGEgUEFSQU1fRVhFQyB2YWx1ZSBmb3Ig ZG9tYWluIGNvbnN0cmFpbnQgY2hlY2tzICovCiAgCWlmIChoYXNDb25zdHJhaW50cykKKioq KioqKioqKioqKioqCioqKiAxNjkxLDE3MTIgKioqKgogIAllcnJjb250ZXh0LnByZXZpb3Vz ID0gZXJyb3JfY29udGV4dF9zdGFjazsKICAJZXJyb3JfY29udGV4dF9zdGFjayA9ICZlcnJj b250ZXh0OwogIAohIAkvKiBvbiBpbnB1dCBqdXN0IHRocm93IHRoZSBoZWFkZXIgbGluZSBh d2F5ICovCiEgCWlmIChoZWFkZXJfbGluZSkKICAJewohIAkJY29weV9saW5lbm8rKzsKISAJ CWRvbmUgPSBDb3B5UmVhZExpbmUocXVvdGUsIGVzY2FwZSkgOwogIAl9CiAgCiEgCXdoaWxl ICghZG9uZSkKICAJewogIAkJYm9vbAkJc2tpcF90dXBsZTsKICAJCU9pZAkJCWxvYWRlZF9v aWQgPSBJbnZhbGlkT2lkOwogIAogIAkJQ0hFQ0tfRk9SX0lOVEVSUlVQVFMoKTsKICAKLSAJ CWNvcHlfbGluZW5vKys7Ci0gCiAgCQkvKiBSZXNldCB0aGUgcGVyLXR1cGxlIGV4cHJjb250 ZXh0ICovCiAgCQlSZXNldFBlclR1cGxlRXhwckNvbnRleHQoZXN0YXRlKTsKICAKLS0tIDE2 MjcsMTY4MSAtLS0tCiAgCWVycmNvbnRleHQucHJldmlvdXMgPSBlcnJvcl9jb250ZXh0X3N0 YWNrOwogIAllcnJvcl9jb250ZXh0X3N0YWNrID0gJmVycmNvbnRleHQ7CiAgCiEgCS8qCiEg CSAqIGluaXRpYWxpemUgYnVmZmVyZWQgc2NhbiB2YXJpYWJsZXMuCiEgCSAqLwohIAlpZihj c3ZfbW9kZSkKISAJewohIAkgICAgaW5fcXVvdGUgPSBmYWxzZTsKISAJCWxhc3Rfd2FzX2Vz YyA9IGZhbHNlOwohIAl9CiEgCiEgCW51bGxfcHJpbnRfbGVuID0gc3RybGVuKG51bGxfcHJp bnQpOwohIAkKISAJLyogU2V0IHVwIGRhdGEgYnVmZmVyIHRvIGhvbGQgYSBjaHVuayBvZiBk YXRhICovCiEgCU1lbVNldChpbnB1dF9idWYsICcgJywgQ09QWV9CVUZfU0laRSAqIHNpemVv ZihjaGFyKSk7CiEgCWlucHV0X2J1ZltDT1BZX0JVRl9TSVpFXSA9ICdcMCc7CiEgCiEgCW5v X21vcmVfZGF0YSA9IGZhbHNlOwkJLyogbm8gbW9yZSBpbnB1dCBkYXRhIHRvIHJlYWQgZnJv bSBmaWxlIG9yIEZFICovCiEgCWxpbmVfZG9uZSA9IHRydWU7CiEgCWJ1Zl9kb25lID0gZmFs c2U7CiEgCiEgCWRvCiAgCXsKISAJCXNpemVfdAkgICBieXRlc3JlYWQgPSAwOwohIAkJCiEg CQkvKiByZWFkIGEgY2h1bmsgb2YgZGF0YSBpbnRvIHRoZSBidWZmZXIgKi8KISAJCWlmICgh YmluYXJ5KQohIAkJewohIAkJICAgYnl0ZXNyZWFkID0gQ29weUdldERhdGEoaW5wdXRfYnVm LCBDT1BZX0JVRl9TSVpFKTsKISAJCSAgIGJ1Zl9kb25lID0gZmFsc2U7CiEgCiEgCQkgICAv KiBzZXQgYnVmZmVyIHBvaW50ZXJzIHRvIGJlZ2lubmluZyBvZiB0aGUgYnVmZmVyICovCiEg CQkgICBiZWdsb2MgPSBpbnB1dF9idWY7CiEgCQkgICBidWZmZXJfaW5kZXggPSAwOwogIAl9 CiAgCiEgCQkvKgohIAkJICogY29udGludWUgaWYgc29tZSBieXRlcyB3ZXJlIHJlYWQgb3Ig aWYgd2UgZGlkbid0IHJlYWNoIEVPRi4gaWYgd2UKISAJCSAqIGJvdGggcmVhY2hlZCBFT0Yg X2FuZF8gbm8gYnl0ZXMgd2VyZSByZWFkLCBxdWl0IHRoZSBsb29wIHdlIGFyZQohIAkJICog ZG9uZQohIAkJICovCiEgCQlpZiAoYnl0ZXNyZWFkID4gMCB8fCAhZmVfZW9mIHx8IGJpbmFy eSkKISAJCXsKISAKISAJCQl3aGlsZSAoIWJ1Zl9kb25lKQogIAl7CiAgCQlib29sCQlza2lw X3R1cGxlOwogIAkJT2lkCQkJbG9hZGVkX29pZCA9IEludmFsaWRPaWQ7CiAgCiAgCQlDSEVD S19GT1JfSU5URVJSVVBUUygpOwogIAogIAkJLyogUmVzZXQgdGhlIHBlci10dXBsZSBleHBy Y29udGV4dCAqLwogIAkJUmVzZXRQZXJUdXBsZUV4cHJDb250ZXh0KGVzdGF0ZSk7CiAgCioq KioqKioqKioqKioqKgoqKiogMTcxNiwxNzQ2ICoqKioKICAJCS8qIEluaXRpYWxpemUgYWxs IHZhbHVlcyBmb3Igcm93IHRvIE5VTEwgKi8KICAJCU1lbVNldCh2YWx1ZXMsIDAsIG51bV9w aHlzX2F0dHJzICogc2l6ZW9mKERhdHVtKSk7CiAgCQlNZW1TZXQobnVsbHMsICduJywgbnVt X3BoeXNfYXR0cnMgKiBzaXplb2YoY2hhcikpOwogIAogIAkJaWYgKCFiaW5hcnkpCiAgCQl7 Ci0gCQkJQ29weVJlYWRSZXN1bHQgcmVzdWx0ID0gTk9STUFMX0FUVFI7Ci0gCQkJY2hhcgkg ICAqc3RyaW5nOwotIAkJCUxpc3RDZWxsICAgKmN1cjsKICAKICAJCQkvKiBBY3R1YWxseSBy ZWFkIHRoZSBsaW5lIGludG8gbWVtb3J5IGhlcmUgKi8KISAJCQlkb25lID0gY3N2X21vZGUg PyAKISAJCQkJQ29weVJlYWRMaW5lKHF1b3RlLCBlc2NhcGUpIDogQ29weVJlYWRMaW5lKE5V TEwsIE5VTEwpOwogIAogIAkJCS8qCiEgCQkJICogRU9GIGF0IHN0YXJ0IG9mIGxpbmUgbWVh bnMgd2UncmUgZG9uZS4gIElmIHdlIHNlZSBFT0YgYWZ0ZXIKISAJCQkgKiBzb21lIGNoYXJh Y3RlcnMsIHdlIGFjdCBhcyB0aG91Z2ggaXQgd2FzIG5ld2xpbmUgZm9sbG93ZWQKISAJCQkg KiBieSBFT0YsIGllLCBwcm9jZXNzIHRoZSBsaW5lIGFuZCB0aGVuIGV4aXQgbG9vcCBvbiBu ZXh0CiEgCQkJICogaXRlcmF0aW9uLgogIAkJCSAqLwohIAkJCWlmIChkb25lICYmIGxpbmVf YnVmLmxlbiA9PSAwKQogIAkJCQlicmVhazsKICAKICAJCQlpZiAoZmlsZV9oYXNfb2lkcykK ICAJCQl7CiAgCQkJCS8qIGNhbid0IGJlIGluIENTViBtb2RlIGhlcmUgKi8KISAJCQkJc3Ry aW5nID0gQ29weVJlYWRBdHRyaWJ1dGUoZGVsaW0sIG51bGxfcHJpbnQsCiEgCQkJCQkJCQkJ CSAgICZyZXN1bHQsICZpc251bGwpOwogIAogIAkJCQlpZiAoaXNudWxsKQogIAkJCQkJZXJl cG9ydChFUlJPUiwKLS0tIDE2ODUsMTcyOSAtLS0tCiAgCQkvKiBJbml0aWFsaXplIGFsbCB2 YWx1ZXMgZm9yIHJvdyB0byBOVUxMICovCiAgCQlNZW1TZXQodmFsdWVzLCAwLCBudW1fcGh5 c19hdHRycyAqIHNpemVvZihEYXR1bSkpOwogIAkJTWVtU2V0KG51bGxzLCAnbicsIG51bV9w aHlzX2F0dHJzICogc2l6ZW9mKGNoYXIpKTsKKyAJCQkJLyogcmVzZXQgYXR0cmlidXRlIHBv aW50ZXJzICovCisgCQkJCU1lbVNldChhdHRyX29mZnNldHMsIDAsIG51bV9waHlzX2F0dHJz ICogc2l6ZW9mKGludCkpOwogIAogIAkJaWYgKCFiaW5hcnkpCiAgCQl7CiAgCiAgCQkJLyog QWN0dWFsbHkgcmVhZCB0aGUgbGluZSBpbnRvIG1lbW9yeSBoZXJlICovCiEgCQkJCQlsaW5l X2RvbmUgPSBjc3ZfbW9kZSA/CiEgCQkJCQkJCQlDb3B5UmVhZExpbmVDU1YoYnl0ZXNyZWFk LCBxdW90ZSwgZXNjYXBlKSA6IAohIAkJCQkJCQkJQ29weVJlYWRMaW5lVGV4dChieXRlc3Jl YWQsIGVzY2FwZSk7CiAgCiAgCQkJLyoKISAJCQkJCSAqIGlmIGZpbmlzaGVkIHByb2Nlc3Np bmcgZGF0YSBsaW5lIC0gaW5jcmVtZW50IGxpbmUgY291bnQuCiEgCQkJCQkgKiBPdGhlcndp c2UsIGlmIGVvZiBpcyBub3QgeWV0IHJlYWNoZWQsIHdlIHNraXAgYXR0IHBhcnNpbmcgCiEg CQkJCQkgKiBhbmQgcmVhZCBtb3JlIGRhdGEuIEJ1dCBpZiBlb2YgX3dhc18gcmVhY2hlZCBp dCBtZWFucyAKISAJCQkJCSAqIHRoYXQgdGhlIG9yaWdpbmFsIGxhc3QgZGF0YSBsaW5lIGlz IGRlZmVjdGl2ZSBhbmQgCiEgCQkJCQkgKiB3ZSB3YW50IHRvIGNhdGNoIHRoYXQgZXJyb3Ig bGF0ZXIgb24uCiAgCQkJICovCiEgCQkJCQlpZiAobGluZV9kb25lKQohIAkJCQkJCWNvcHlf bGluZW5vKys7CiEgCQkJCQllbHNlIGlmICghZmVfZW9mIHx8IGVuZF9tYXJrZXIgKQogIAkJ CQlicmVhazsKICAKKyAJCQkJCWlmIChoZWFkZXJfbGluZSkKKyAJCQkJCXsKKyAJCQkJCQls aW5lX2J1Zi5sZW4gPSAwOwkJLyogd2UgY2FuIHJlc2V0IGxpbmUgYnVmZmVyIG5vdy4gKi8K KyAJCQkJCQlsaW5lX2J1Zi5kYXRhWzBdID0gJ1wwJzsKKyAJCQkJCQlsaW5lX2J1Zi5jdXJz b3IgPSAwOworIAkJCQkJCWhlYWRlcl9saW5lID0gZmFsc2U7CisgCQkJCQkJY29udGludWU7 CisgCQkJCQl9CisgCQkJCQkKICAJCQlpZiAoZmlsZV9oYXNfb2lkcykKICAJCQl7CisgCQkJ CQkJY2hhcgkgICAqb2lkX3N0cmluZzsKKyAKICAJCQkJLyogY2FuJ3QgYmUgaW4gQ1NWIG1v ZGUgaGVyZSAqLwohIAkJCQkJCW9pZF9zdHJpbmcgPSBDb3B5UmVhZE9pZEF0dHIoZGVsaW0s IG51bGxfcHJpbnQsIG51bGxfcHJpbnRfbGVuLAohIAkJCQkJCQkJCQkJCQkgJmlzbnVsbCk7 CiAgCiAgCQkJCWlmIChpc251bGwpCiAgCQkJCQllcmVwb3J0KEVSUk9SLAoqKioqKioqKioq KioqKioKKioqIDE3NTAsMTc1NiAqKioqCiAgCQkJCXsKICAJCQkJCWNvcHlfYXR0bmFtZSA9 ICJvaWQiOwogIAkJCQkJbG9hZGVkX29pZCA9IERhdHVtR2V0T2JqZWN0SWQoRGlyZWN0RnVu Y3Rpb25DYWxsMShvaWRpbiwKISAJCQkJCQkJCQkJCSAgIENTdHJpbmdHZXREYXR1bShzdHJp bmcpKSk7CiAgCQkJCQlpZiAobG9hZGVkX29pZCA9PSBJbnZhbGlkT2lkKQogIAkJCQkJCWVy ZXBvcnQoRVJST1IsCiAgCQkJCQkJCQkoZXJyY29kZShFUlJDT0RFX0JBRF9DT1BZX0ZJTEVf Rk9STUFUKSwKLS0tIDE3MzMsMTczOSAtLS0tCiAgCQkJCXsKICAJCQkJCWNvcHlfYXR0bmFt ZSA9ICJvaWQiOwogIAkJCQkJbG9hZGVkX29pZCA9IERhdHVtR2V0T2JqZWN0SWQoRGlyZWN0 RnVuY3Rpb25DYWxsMShvaWRpbiwKISAJCQkJCQkJCQkJCQkgICBDU3RyaW5nR2V0RGF0dW0o b2lkX3N0cmluZykpKTsKICAJCQkJCWlmIChsb2FkZWRfb2lkID09IEludmFsaWRPaWQpCiAg CQkJCQkJZXJlcG9ydChFUlJPUiwKICAJCQkJCQkJCShlcnJjb2RlKEVSUkNPREVfQkFEX0NP UFlfRklMRV9GT1JNQVQpLAoqKioqKioqKioqKioqKioKKioqIDE3NTksMTc5MiAqKioqCiAg CQkJCX0KICAJCQl9CiAgCiEgCQkJLyogTG9vcCB0byByZWFkIHRoZSB1c2VyIGF0dHJpYnV0 ZXMgb24gdGhlIGxpbmUuICovCiAgCQkJZm9yZWFjaChjdXIsIGF0dG51bWxpc3QpCiAgCQkJ ewogIAkJCQlpbnQJCQlhdHRudW0gPSBsZmlyc3RfaW50KGN1cik7CiAgCQkJCWludAkJCW0g PSBhdHRudW0gLSAxOwogIAohIAkJCQkvKgohIAkJCQkgKiBJZiBwcmlvciBhdHRyIG9uIHRo aXMgbGluZSB3YXMgZW5kZWQgYnkgbmV3bGluZSwKISAJCQkJICogY29tcGxhaW4uCiEgCQkJ CSAqLwohIAkJCQlpZiAocmVzdWx0ICE9IE5PUk1BTF9BVFRSKQohIAkJCQkJZXJlcG9ydChF UlJPUiwKISAJCQkJCQkJKGVycmNvZGUoRVJSQ09ERV9CQURfQ09QWV9GSUxFX0ZPUk1BVCks CiEgCQkJCQkJCSBlcnJtc2coIm1pc3NpbmcgZGF0YSBmb3IgY29sdW1uIFwiJXNcIiIsCiEg CQkJCQkJCQkJTmFtZVN0cihhdHRyW21dLT5hdHRuYW1lKSkpKTsKICAKISAJCQkJaWYgKGNz dl9tb2RlKQohIAkJCQl7CiEgCQkJCQlzdHJpbmcgPSBDb3B5UmVhZEF0dHJpYnV0ZUNTVihk ZWxpbSwgbnVsbF9wcmludCwgcXVvdGUsCiEgCQkJCQkJCQkJCQkgICBlc2NhcGUsICZyZXN1 bHQsICZpc251bGwpOwohIAkJCQkJaWYgKHJlc3VsdCA9PSBVTlRFUk1JTkFURURfRklFTEQp CiEgCQkJCQkJZXJlcG9ydChFUlJPUiwKISAJCQkJCQkJCShlcnJjb2RlKEVSUkNPREVfQkFE X0NPUFlfRklMRV9GT1JNQVQpLAohIAkJCQkJCQkgICBlcnJtc2coInVudGVybWluYXRlZCBD U1YgcXVvdGVkIGZpZWxkIikpKTsKISAJCQkJfQogIAkJCQllbHNlCiEgCQkJCQlzdHJpbmcg PSBDb3B5UmVhZEF0dHJpYnV0ZShkZWxpbSwgbnVsbF9wcmludCwKISAJCQkJCQkJCQkJCSAg ICZyZXN1bHQsICZpc251bGwpOwogIAogIAkJCQlpZiAoY3N2X21vZGUgJiYgaXNudWxsICYm IGZvcmNlX25vdG51bGxbbV0pCiAgCQkJCXsKLS0tIDE3NDIsMTc3MCAtLS0tCiAgCQkJCX0K ICAJCQl9CiAgCiEgCQkJCQkvKiBwYXJzZSBhbGwgdGhlIGF0dHJpYnV0ZSBpbiB0aGUgZGF0 YSBsaW5lICovCiEgCQkJCQlpZihjc3ZfbW9kZSkKISAJCQkJCQlDb3B5UmVhZEF0dHJpYnV0 ZXNDU1YoZGVsaW0sIG51bGxfcHJpbnQsIHF1b3RlLCBlc2NhcGUsIG51bGxfcHJpbnRfbGVu LCAKISAJCQkJCQkJCQkJCW51bGxzLCBhdHRudW1saXN0LCBhdHRyX29mZnNldHMsIG51bV9w aHlzX2F0dHJzLCBhdHRyKTsKISAJCQkJCWVsc2UKISAJCQkJCQlDb3B5UmVhZEF0dHJpYnV0 ZXNUZXh0KGRlbGltLCBlc2NhcGUsIG51bGxfcHJpbnQsIG51bGxfcHJpbnRfbGVuLAohIAkJ CQkJCQkJCQkJIG51bGxzLCBhdHRudW1saXN0LCBhdHRyX29mZnNldHMsIG51bV9waHlzX2F0 dHJzLCBhdHRyKTsKISAKISAJCQkJCS8qCiEgCQkJCQkgKiBMb29wIHRvIHJlYWQgdGhlIHVz ZXIgYXR0cmlidXRlcyBvbiB0aGUgbGluZS4KISAJCQkJCSAqLwogIAkJCWZvcmVhY2goY3Vy LCBhdHRudW1saXN0KQogIAkJCXsKICAJCQkJaW50CQkJYXR0bnVtID0gbGZpcnN0X2ludChj dXIpOwogIAkJCQlpbnQJCQltID0gYXR0bnVtIC0gMTsKKyAJCQkJCQljaGFyICAgICAgICpz dHJpbmc7CiAgCiEgCQkJCQkJc3RyaW5nID0gYXR0cl9idWYuZGF0YSArIGF0dHJfb2Zmc2V0 c1ttXTsKICAKISAJCQkJCQlpZiAobnVsbHNbbV0gPT0gJyAnKQohIAkJCQkJCQlpc251bGwg PSBmYWxzZTsKICAJCQkJZWxzZQohIAkJCQkJCQlpc251bGwgPSB0cnVlOwogIAogIAkJCQlp ZiAoY3N2X21vZGUgJiYgaXNudWxsICYmIGZvcmNlX25vdG51bGxbbV0pCiAgCQkJCXsKKioq KioqKioqKioqKioqCioqKiAxODA2LDE4MjMgKioqKgogIAkJCQkJY29weV9hdHRuYW1lID0g TlVMTDsKICAJCQkJfQogIAkJCX0KLSAKLSAJCQkvKgotIAkJCSAqIENvbXBsYWluIGlmIHRo ZXJlIGFyZSBtb3JlIGZpZWxkcyBvbiB0aGUgaW5wdXQgbGluZS4KLSAJCQkgKgotIAkJCSAq IFNwZWNpYWwgY2FzZTogaWYgd2UncmUgcmVhZGluZyBhIHplcm8tY29sdW1uIHRhYmxlLCB3 ZSB3b24ndAotIAkJCSAqIHlldCBoYXZlIGNhbGxlZCBDb3B5UmVhZEF0dHJpYnV0ZSgpIGF0 IGFsbDsgc28gbm8gZXJyb3IgaWYKLSAJCQkgKiBsaW5lIGlzIGVtcHR5LgotIAkJCSAqLwot IAkJCWlmIChyZXN1bHQgPT0gTk9STUFMX0FUVFIgJiYgbGluZV9idWYubGVuICE9IDApCi0g CQkJCWVyZXBvcnQoRVJST1IsCi0gCQkJCQkJKGVycmNvZGUoRVJSQ09ERV9CQURfQ09QWV9G SUxFX0ZPUk1BVCksCi0gCQkJCQkgICBlcnJtc2coImV4dHJhIGRhdGEgYWZ0ZXIgbGFzdCBl eHBlY3RlZCBjb2x1bW4iKSkpOwogIAkJfQogIAkJZWxzZQogIAkJewotLS0gMTc4NCwxNzg5 IC0tLS0KKioqKioqKioqKioqKioqCioqKiAxODI4LDE4MzQgKioqKgogIAkJCWZsZF9jb3Vu dCA9IENvcHlHZXRJbnQxNigpOwogIAkJCWlmIChDb3B5R2V0RW9mKCkgfHwgZmxkX2NvdW50 ID09IC0xKQogIAkJCXsKISAJCQkJZG9uZSA9IHRydWU7CiAgCQkJCWJyZWFrOwogIAkJCX0K ICAKLS0tIDE3OTQsMTgwMSAtLS0tCiAgCQkJZmxkX2NvdW50ID0gQ29weUdldEludDE2KCk7 CiAgCQkJaWYgKENvcHlHZXRFb2YoKSB8fCBmbGRfY291bnQgPT0gLTEpCiAgCQkJewohIAkJ CQkJCWJ1Zl9kb25lID0gdHJ1ZTsKISAJCQkJCQlub19tb3JlX2RhdGEgPSB0cnVlOwogIAkJ CQlicmVhazsKICAJCQl9CiAgCioqKioqKioqKioqKioqKgoqKiogMTg4NSwxODkxICoqKioK ICAJCQkJbnVsbHNbZGVmbWFwW2ldXSA9ICcgJzsKICAJCX0KICAKISAJCS8qIE5leHQgYXBw bHkgYW55IGRvbWFpbiBjb25zdHJhaW50cyAqLwogIAkJaWYgKGhhc0NvbnN0cmFpbnRzKQog IAkJewogIAkJCVBhcmFtRXhlY0RhdGEgKnBybWRhdGEgPSAmZWNvbnRleHQtPmVjeHRfcGFy YW1fZXhlY192YWxzWzBdOwotLS0gMTg1MiwxODYwIC0tLS0KICAJCQkJbnVsbHNbZGVmbWFw W2ldXSA9ICcgJzsKICAJCX0KICAKISAJCQkJLyoKISAJCQkJICogTmV4dCBhcHBseSBhbnkg ZG9tYWluIGNvbnN0cmFpbnRzCiEgCQkJCSAqLwogIAkJaWYgKGhhc0NvbnN0cmFpbnRzKQog IAkJewogIAkJCVBhcmFtRXhlY0RhdGEgKnBybWRhdGEgPSAmZWNvbnRleHQtPmVjeHRfcGFy YW1fZXhlY192YWxzWzBdOwoqKioqKioqKioqKioqKioKKioqIDE5MTIsMTkyNCAqKioqCiAg CQkJfQogIAkJfQogIAohIAkJLyogQW5kIG5vdyB3ZSBjYW4gZm9ybSB0aGUgaW5wdXQgdHVw bGUuICovCiAgCQl0dXBsZSA9IGhlYXBfZm9ybXR1cGxlKHR1cERlc2MsIHZhbHVlcywgbnVs bHMpOwogIAogIAkJaWYgKG9pZHMgJiYgZmlsZV9oYXNfb2lkcykKICAJCQlIZWFwVHVwbGVT ZXRPaWQodHVwbGUsIGxvYWRlZF9vaWQpOwogIAohIAkJLyogVHJpZ2dlcnMgYW5kIHN0dWZm IG5lZWQgdG8gYmUgaW52b2tlZCBpbiBxdWVyeSBjb250ZXh0LiAqLwogIAkJTWVtb3J5Q29u dGV4dFN3aXRjaFRvKG9sZGNvbnRleHQpOwogIAogIAkJc2tpcF90dXBsZSA9IGZhbHNlOwot LS0gMTg4MSwxODk3IC0tLS0KICAJCQl9CiAgCQl9CiAgCiEgCQkJCS8qCiEgCQkJCSAqIEFu ZCBub3cgd2UgY2FuIGZvcm0gdGhlIGlucHV0IHR1cGxlLgohIAkJCQkgKi8KICAJCXR1cGxl ID0gaGVhcF9mb3JtdHVwbGUodHVwRGVzYywgdmFsdWVzLCBudWxscyk7CiAgCiAgCQlpZiAo b2lkcyAmJiBmaWxlX2hhc19vaWRzKQogIAkJCUhlYXBUdXBsZVNldE9pZCh0dXBsZSwgbG9h ZGVkX29pZCk7CiAgCiEgCQkJCS8qCiEgCQkJCSAqIFRyaWdnZXJzIGFuZCBzdHVmZiBuZWVk IHRvIGJlIGludm9rZWQgaW4gcXVlcnkgY29udGV4dC4KISAJCQkJICovCiAgCQlNZW1vcnlD b250ZXh0U3dpdGNoVG8ob2xkY29udGV4dCk7CiAgCiAgCQlza2lwX3R1cGxlID0gZmFsc2U7 CioqKioqKioqKioqKioqKgoqKiogMTk0NSwxOTU1ICoqKioKICAJCQkvKiBQbGFjZSB0dXBs ZSBpbiB0dXBsZSBzbG90ICovCiAgCQkJRXhlY1N0b3JlVHVwbGUodHVwbGUsIHNsb3QsIElu dmFsaWRCdWZmZXIsIGZhbHNlKTsKICAKISAJCQkvKiBDaGVjayB0aGUgY29uc3RyYWludHMg b2YgdGhlIHR1cGxlICovCiAgCQkJaWYgKHJlbC0+cmRfYXR0LT5jb25zdHIpCiAgCQkJCUV4 ZWNDb25zdHJhaW50cyhyZXN1bHRSZWxJbmZvLCBzbG90LCBlc3RhdGUpOwogIAohIAkJCS8q IE9LLCBzdG9yZSB0aGUgdHVwbGUgYW5kIGNyZWF0ZSBpbmRleCBlbnRyaWVzIGZvciBpdCAq LwogIAkJCXNpbXBsZV9oZWFwX2luc2VydChyZWwsIHR1cGxlKTsKICAKICAJCQlpZiAocmVz dWx0UmVsSW5mby0+cmlfTnVtSW5kaWNlcyA+IDApCi0tLSAxOTE4LDE5MzIgLS0tLQogIAkJ CS8qIFBsYWNlIHR1cGxlIGluIHR1cGxlIHNsb3QgKi8KICAJCQlFeGVjU3RvcmVUdXBsZSh0 dXBsZSwgc2xvdCwgSW52YWxpZEJ1ZmZlciwgZmFsc2UpOwogIAohIAkJCQkJLyoKISAJCQkJ CSAqIENoZWNrIHRoZSBjb25zdHJhaW50cyBvZiB0aGUgdHVwbGUKISAJCQkJCSAqLwogIAkJ CWlmIChyZWwtPnJkX2F0dC0+Y29uc3RyKQogIAkJCQlFeGVjQ29uc3RyYWludHMocmVzdWx0 UmVsSW5mbywgc2xvdCwgZXN0YXRlKTsKICAKISAJCQkJCS8qCiEgCQkJCQkgKiBPSywgc3Rv cmUgdGhlIHR1cGxlIGFuZCBjcmVhdGUgaW5kZXggZW50cmllcyBmb3IgaXQKISAJCQkJCSAq LwogIAkJCXNpbXBsZV9oZWFwX2luc2VydChyZWwsIHR1cGxlKTsKICAKICAJCQlpZiAocmVz dWx0UmVsSW5mby0+cmlfTnVtSW5kaWNlcyA+IDApCioqKioqKioqKioqKioqKgoqKiogMTk1 OCwxOTc2ICoqKioKICAJCQkvKiBBRlRFUiBST1cgSU5TRVJUIFRyaWdnZXJzICovCiAgCQkJ RXhlY0FSSW5zZXJ0VHJpZ2dlcnMoZXN0YXRlLCByZXN1bHRSZWxJbmZvLCB0dXBsZSk7CiAg CQl9CiAgCX0KICAKISAJLyogRG9uZSwgY2xlYW4gdXAgKi8KICAJZXJyb3JfY29udGV4dF9z dGFjayA9IGVycmNvbnRleHQucHJldmlvdXM7CiAgCiAgCU1lbW9yeUNvbnRleHRTd2l0Y2hU byhvbGRjb250ZXh0KTsKICAKISAJLyogRXhlY3V0ZSBBRlRFUiBTVEFURU1FTlQgaW5zZXJ0 aW9uIHRyaWdnZXJzICovCiAgCUV4ZWNBU0luc2VydFRyaWdnZXJzKGVzdGF0ZSwgcmVzdWx0 UmVsSW5mbyk7CiAgCiEgCS8qIEhhbmRsZSBxdWV1ZWQgQUZURVIgdHJpZ2dlcnMgKi8KICAJ QWZ0ZXJUcmlnZ2VyRW5kUXVlcnkoZXN0YXRlKTsKICAKICAJcGZyZWUodmFsdWVzKTsKICAJ cGZyZWUobnVsbHMpOwogIAotLS0gMTkzNSwxOTcyIC0tLS0KICAJCQkvKiBBRlRFUiBST1cg SU5TRVJUIFRyaWdnZXJzICovCiAgCQkJRXhlY0FSSW5zZXJ0VHJpZ2dlcnMoZXN0YXRlLCBy ZXN1bHRSZWxJbmZvLCB0dXBsZSk7CiAgCQl9CisgCisgCQkJCWxpbmVfYnVmLmxlbiA9IDA7 CQkvKiB3ZSBjYW4gcmVzZXQgbGluZSBidWZmZXIgbm93LiAqLworIAkJCQlsaW5lX2J1Zi5k YXRhWzBdID0gJ1wwJzsKKyAJCQkJbGluZV9idWYuY3Vyc29yID0gMDsKKyAJCQl9CQkJCQkv KiBlbmQgd2hpbGUoIWJ1Zl9kb25lKSAqLworIAkJfQkJCQkJCS8qIGVuZCBpZiAoYnl0ZXNy ZWFkID4gMCB8fCAhZmVfZW9mKSAqLworIAkJZWxzZQorIAkJCS8qIG5vIGJ5dGVzIHJlYWQs IGVuZCBvZiBkYXRhICovCisgCQl7CisgCQkJbm9fbW9yZV9kYXRhID0gdHJ1ZTsKICAJfQor IAl9IHdoaWxlICghbm9fbW9yZV9kYXRhKTsKICAKISAJLyoKISAJICogRG9uZSwgY2xlYW4g dXAKISAJICovCiAgCWVycm9yX2NvbnRleHRfc3RhY2sgPSBlcnJjb250ZXh0LnByZXZpb3Vz OwogIAogIAlNZW1vcnlDb250ZXh0U3dpdGNoVG8ob2xkY29udGV4dCk7CiAgCiEgCS8qIAkg CiEgICAgICAqIEV4ZWN1dGUgQUZURVIgU1RBVEVNRU5UIGluc2VydGlvbiB0cmlnZ2Vycwoh IAkgKi8KICAJRXhlY0FTSW5zZXJ0VHJpZ2dlcnMoZXN0YXRlLCByZXN1bHRSZWxJbmZvKTsK ICAKISAJLyoKISAJICogSGFuZGxlIHF1ZXVlZCBBRlRFUiB0cmlnZ2VycwohIAkgKi8KICAJ QWZ0ZXJUcmlnZ2VyRW5kUXVlcnkoZXN0YXRlKTsKICAKKyAJcGZyZWUoYXR0cl9vZmZzZXRz KTsKKyAKICAJcGZyZWUodmFsdWVzKTsKICAJcGZyZWUobnVsbHMpOwogIAoqKioqKioqKioq KioqKioKKioqIDE5OTAsMjI2MiAqKioqCiAgCiAgCiAgLyoKISAgKiBSZWFkIHRoZSBuZXh0 IGlucHV0IGxpbmUgYW5kIHN0YXNoIGl0IGluIGxpbmVfYnVmLCB3aXRoIGNvbnZlcnNpb24g dG8KISAgKiBzZXJ2ZXIgZW5jb2RpbmcuCiEgICoKISAgKiBSZXN1bHQgaXMgdHJ1ZSBpZiBy ZWFkIHdhcyB0ZXJtaW5hdGVkIGJ5IEVPRiwgZmFsc2UgaWYgdGVybWluYXRlZAohICAqIGJ5 IG5ld2xpbmUuCiAgICovCiAgc3RhdGljIGJvb2wKISBDb3B5UmVhZExpbmUoY2hhciAqIHF1 b3RlLCBjaGFyICogZXNjYXBlKQogIHsKISAJYm9vbAkJcmVzdWx0OwohIAlib29sCQljaGFu Z2VfZW5jb2RpbmcgPSAoY2xpZW50X2VuY29kaW5nICE9IHNlcnZlcl9lbmNvZGluZyk7CiEg CWludAkJCWM7CiEgCWludAkJCW1ibGVuOwohIAlpbnQJCQlqOwohIAl1bnNpZ25lZCBjaGFy IHNbMl07CiAgCWNoYXIJICAgKmN2dDsKISAJYm9vbCAgICAgICAgaW5fcXVvdGUgPSBmYWxz ZSwgbGFzdF93YXNfZXNjID0gZmFsc2UsIGNzdl9tb2RlID0gZmFsc2U7CiEgCWNoYXIgICAg ICAgIHF1b3RlYyA9ICdcMCcsIGVzY2FwZWMgPSAnXDAnOwohIAohIAlpZiAocXVvdGUpCiEg CXsKISAJCWNzdl9tb2RlID0gdHJ1ZTsKISAJCXF1b3RlYyA9IHF1b3RlWzBdOwohIAkJZXNj YXBlYyA9IGVzY2FwZVswXTsKISAJCS8qIGlnbm9yZSBzcGVjaWFsIGVzY2FwZSBwcm9jZXNz aW5nIGlmIGl0J3MgdGhlIHNhbWUgYXMgcXVvdGVjICovCiEgCQlpZiAocXVvdGVjID09IGVz Y2FwZWMpCiEgCQkJZXNjYXBlYyA9ICdcMCc7CiEgCX0KISAKISAKISAJc1sxXSA9IDA7CiEg CiEgCS8qIHJlc2V0IGxpbmVfYnVmIHRvIGVtcHR5ICovCiEgCWxpbmVfYnVmLmxlbiA9IDA7 CiEgCWxpbmVfYnVmLmRhdGFbMF0gPSAnXDAnOwohIAlsaW5lX2J1Zi5jdXJzb3IgPSAwOwog IAogIAkvKiBtYXJrIHRoYXQgZW5jb2RpbmcgY29udmVyc2lvbiBoYXNuJ3Qgb2NjdXJyZWQg eWV0ICovCiAgCWxpbmVfYnVmX2NvbnZlcnRlZCA9IGZhbHNlOwogIAohIAkvKiBzZXQgZGVm YXVsdCBzdGF0dXMgKi8KISAJcmVzdWx0ID0gZmFsc2U7CiAgCiAgCS8qCiEgCSAqIEluIHRo aXMgbG9vcCB3ZSBvbmx5IGNhcmUgZm9yIGRldGVjdGluZyBuZXdsaW5lcyAoXHIgYW5kL29y IFxuKSBhbmQKISAJICogdGhlIGVuZC1vZi1jb3B5IG1hcmtlciAoXC4pLiAgCiEgCSAqCiEg CSAqIEluIFRleHQgbW9kZSwgZm9yIGJhY2t3YXJkcyBjb21wYXRpYmlsaXR5IHdlIGFsbG93 CiEgCSAqIGJhY2tzbGFzaGVzIHRvIGVzY2FwZSBuZXdsaW5lIGNoYXJhY3RlcnMuICBCYWNr c2xhc2hlcyBvdGhlciB0aGFuCiEgCSAqIHRoZSBlbmQgbWFya2VyIGdldCBwdXQgaW50byB0 aGUgbGluZV9idWYsIHNpbmNlIENvcHlSZWFkQXR0cmlidXRlCiEgCSAqIGRvZXMgaXRzIG93 biBlc2NhcGUgcHJvY2Vzc2luZy4JCiEgCSAqCiEgCSAqIEluIENTViBtb2RlLCBDUiBhbmQg TkwgaW5zaWRlIHEgcXVvdGVkIGZpZWxkIGFyZSBqdXN0IHBhcnQgb2YgdGhlCiEgCSAqIGRh dGEgdmFsdWUgYW5kIGFyZSBwdXQgaW4gbGluZV9idWYuIFdlIGtlZXAganVzdCBlbm91Z2gg c3RhdGUKISAJICogdG8ga25vdyBpZiB3ZSBhcmUgY3VycmVudGx5IGluIGEgcXVvdGVkIGZp ZWxkIG9yIG5vdC4KISAJICoKISAJICogVGhlc2UgZm91ciBjaGFyYWN0ZXJzLCBhbmQgb25s eSB0aGVzZSBmb3VyLCBhcmUgYXNzdW1lZCB0aGUgc2FtZSBpbiAKISAJICogZnJvbnRlbmQg YW5kIGJhY2tlbmQgZW5jb2RpbmdzLgohIAkgKgohIAkgKiBXZSBkbyBub3QgYXNzdW1lIHRo YXQgc2Vjb25kIGFuZCBsYXRlciBieXRlcyBvZiBhIGZyb250ZW5kCiEgCSAqIG11bHRpYnl0 ZSBjaGFyYWN0ZXIgY291bGRuJ3QgbG9vayBsaWtlIEFTQ0lJIGNoYXJhY3RlcnMuCiAgCSAq LwohIAlmb3IgKDs7KQogIAl7CiEgCQljID0gQ29weUdldENoYXIoKTsKISAJCWlmIChjID09 IEVPRikKICAJCXsKISAJCQlyZXN1bHQgPSB0cnVlOwohIAkJCWJyZWFrOwogIAkJfQogIAoh IAkJaWYgKGNzdl9tb2RlKQogIAkJewogIAkJCS8qICAKISAJCQkgKiBEZWFsaW5nIHdpdGgg cXVvdGVzIGFuZCBlc2NhcGVzIGhlcmUgaXMgbWlsZGx5IHRyaWNreS4gSWYgdGhlCiEgCQkJ ICogcXVvdGUgY2hhciBpcyBhbHNvIHRoZSBlc2NhcGUgY2hhciwgdGhlcmUncyBubyBwcm9i bGVtIC0gd2UgIAohIAkJCSAqIGp1c3QgdXNlIHRoZSBjaGFyIGFzIGEgdG9nZ2xlLiBJZiB0 aGV5IGFyZSBkaWZmZXJlbnQsIHdlIG5lZWQKISAJCQkgKiB0byBlbnN1cmUgdGhhdCB3ZSBv bmx5IHRha2UgYWNjb3VudCBvZiBhbiBlc2NhcGUgaW5zaWRlIGEgcXVvdGVkCiEgCQkJICog ZmllbGQgYW5kIGltbWVkaWF0ZWx5IHByZWNlZGluZyBhIHF1b3RlIGNoYXIsIGFuZCBub3Qg dGhlCiEgCQkJICogc2Vjb25kIGluIGEgZXNjYXBlLWVzY2FwZSBzZXF1ZW5jZS4KICAJCQkg Ki8gCiAgCiEgCQkJaWYgKGluX3F1b3RlICYmIGMgPT0gZXNjYXBlYykKISAJCQkJbGFzdF93 YXNfZXNjID0gISBsYXN0X3dhc19lc2M7CiEgCQkJaWYgKGMgPT0gcXVvdGVjICYmICEgbGFz dF93YXNfZXNjKQohIAkJCQlpbl9xdW90ZSA9ICEgaW5fcXVvdGU7CiEgCQkJaWYgKGMgIT0g ZXNjYXBlYykKISAJCQkJbGFzdF93YXNfZXNjID0gZmFsc2U7CiAgCiAgCQkJLyoKISAJCQkg KiB1cGRhdGluZyB0aGUgbGluZSBjb3VudCBmb3IgZW1iZWRkZWQgQ1IgYW5kL29yIExGIGNo YXJzIGlzIAohIAkJCSAqIG5lY2Vzc2FyaWx5IGEgbGl0dGxlIGZyYWdpbGUgLSB0aGlzIHRl c3QgaXMgcHJvYmFibHkgYWJvdXQgCiEgCQkJICogdGhlIGJlc3Qgd2UgY2FuIGRvLgogIAkJ CSAqLyAKISAJCQlpZiAoaW5fcXVvdGUgJiYgYyA9PSAoZW9sX3R5cGUgPT0gRU9MX0NSID8g J1xyJyA6ICdcbicpKSAKISAJCQkJY29weV9saW5lbm8rKzsgCiEgCQl9CiAgCiEgCQlpZiAo IWluX3F1b3RlICYmIGMgPT0gJ1xyJykKICAJCXsKISAJCQlpZiAoZW9sX3R5cGUgPT0gRU9M X05MKQogIAkJCXsKISAJCQkJaWYgKCEgY3N2X21vZGUpCiEgCQkJCQllcmVwb3J0KEVSUk9S LAohIAkJCQkJCQkoZXJyY29kZShFUlJDT0RFX0JBRF9DT1BZX0ZJTEVfRk9STUFUKSwKISAJ CQkJCQkJIGVycm1zZygibGl0ZXJhbCBjYXJyaWFnZSByZXR1cm4gZm91bmQgaW4gZGF0YSIp LAohIAkJCQkJCQkgZXJyaGludCgiVXNlIFwiXFxyXCIgdG8gcmVwcmVzZW50IGNhcnJpYWdl IHJldHVybi4iKSkpOwohIAkJCQllbHNlCiEgCQkJCQllcmVwb3J0KEVSUk9SLAohIAkJCQkJ CQkoZXJyY29kZShFUlJDT0RFX0JBRF9DT1BZX0ZJTEVfRk9STUFUKSwKISAJCQkJCQkJIGVy cm1zZygidW5xdW90ZWQgY2FycmlhZ2UgcmV0dXJuIGZvdW5kIGluIENTViBkYXRhIiksCiEg CQkJCQkJCSBlcnJoaW50KCJVc2UgcXVvdGVkIENTViBmaWVsZCB0byByZXByZXNlbnQgY2Fy cmlhZ2UgcmV0dXJuLiIpKSk7CiAgCQkJfQohIAkJCS8qIENoZWNrIGZvciBcclxuIG9uIGZp cnN0IGxpbmUsIF9hbmRfIGhhbmRsZSBcclxuLiAqLwohIAkJCWlmIChlb2xfdHlwZSA9PSBF T0xfVU5LTk9XTiB8fCBlb2xfdHlwZSA9PSBFT0xfQ1JOTCkKICAJCQl7CiEgCQkJCWludAkJ CWMyID0gQ29weVBlZWtDaGFyKCk7CiAgCiEgCQkJCWlmIChjMiA9PSAnXG4nKQohIAkJCQl7 CiEgCQkJCQlDb3B5RG9uZVBlZWsoYzIsIHRydWUpOwkJLyogZWF0IG5ld2xpbmUgKi8KISAJ CQkJCWVvbF90eXBlID0gRU9MX0NSTkw7CiAgCQkJCX0KICAJCQkJZWxzZQogIAkJCQl7CiEg CQkJCQkvKiBmb3VuZCBcciwgYnV0IG5vIFxuICovCiAgCQkJCQlpZiAoZW9sX3R5cGUgPT0g RU9MX0NSTkwpCiAgCQkJCQl7CiEgCQkJCQkJaWYgKCFjc3ZfbW9kZSkKISAJCQkJCQkJZXJl cG9ydChFUlJPUiwKISAJCQkJCQkJCQkoZXJyY29kZShFUlJDT0RFX0JBRF9DT1BZX0ZJTEVf Rk9STUFUKSwKISAJCQkJCQkJCQkgZXJybXNnKCJsaXRlcmFsIGNhcnJpYWdlIHJldHVybiBm b3VuZCBpbiBkYXRhIiksCiEgCQkJCQkJCQkJIGVycmhpbnQoIlVzZSBcIlxcclwiIHRvIHJl cHJlc2VudCBjYXJyaWFnZSByZXR1cm4uIikpKTsKICAJCQkJCQllbHNlCiEgCQkJCQkJCWVy ZXBvcnQoRVJST1IsCiEgCQkJCQkJCQkJKGVycmNvZGUoRVJSQ09ERV9CQURfQ09QWV9GSUxF X0ZPUk1BVCksCiEgCQkJCQkJCQkJIGVycm1zZygidW5xdW90ZWQgY2FycmlhZ2UgcmV0dXJu IGZvdW5kIGluIGRhdGEiKSwKISAJCQkJCQkJCQkgZXJyaGludCgiVXNlIHF1b3RlZCBDU1Yg ZmllbGQgdG8gcmVwcmVzZW50IGNhcnJpYWdlIHJldHVybi4iKSkpOwohIAogIAkJCQkJfQog IAogIAkJCQkJLyoKISAJCQkJCSAqIGlmIHdlIGdvdCBoZXJlLCBpdCBpcyB0aGUgZmlyc3Qg bGluZSBhbmQgd2UgZGlkbid0CiEgCQkJCQkgKiBnZXQgXG4sIHNvIHB1dCBpdCBiYWNrCiAg CQkJCQkgKi8KISAJCQkJCUNvcHlEb25lUGVlayhjMiwgZmFsc2UpOwohIAkJCQkJZW9sX3R5 cGUgPSBFT0xfQ1I7CiAgCQkJCX0KICAJCQl9Ci0gCQkJYnJlYWs7CiAgCQl9CiEgCQlpZiAo IWluX3F1b3RlICYmIGMgPT0gJ1xuJykKICAJCXsKISAJCQlpZiAoZW9sX3R5cGUgPT0gRU9M X0NSIHx8IGVvbF90eXBlID09IEVPTF9DUk5MKQohIAkJCXsKISAJCQkJaWYgKCFjc3ZfbW9k ZSkKISAJCQkJCWVyZXBvcnQoRVJST1IsCiEgCQkJCQkJCShlcnJjb2RlKEVSUkNPREVfQkFE X0NPUFlfRklMRV9GT1JNQVQpLAohIAkJCQkJCQkgZXJybXNnKCJsaXRlcmFsIG5ld2xpbmUg Zm91bmQgaW4gZGF0YSIpLAohIAkJCQkJCQkgZXJyaGludCgiVXNlIFwiXFxuXCIgdG8gcmVw cmVzZW50IG5ld2xpbmUuIikpKTsKISAJCQkJZWxzZQohIAkJCQkJZXJlcG9ydChFUlJPUiwK ISAJCQkJCQkJKGVycmNvZGUoRVJSQ09ERV9CQURfQ09QWV9GSUxFX0ZPUk1BVCksCiEgCQkJ CQkJCSBlcnJtc2coInVucXVvdGVkIG5ld2xpbmUgZm91bmQgaW4gZGF0YSIpLAohIAkJCQkJ CQkgZXJyaGludCgiVXNlIHF1b3RlZCBDU1YgZmllbGQgdG8gcmVwcmVzZW50IG5ld2xpbmUu IikpKTsKISAJCQkJCQogIAkJCX0KLSAJCQllb2xfdHlwZSA9IEVPTF9OTDsKLSAJCQlicmVh azsKLSAJCX0KICAKISAJCWlmICgobGluZV9idWYubGVuID09IDAgfHwgIWNzdl9tb2RlKSAm JiBjID09ICdcXCcpCiEgCQl7CiEgCQkJaW50IGMyOwohIAkJCQohIAkJCWlmIChjc3ZfbW9k ZSkKISAJCQkJYzIgPSBDb3B5UGVla0NoYXIoKTsKISAJCQllbHNlCiEgCQkJCWMyID0gYyA9 IENvcHlHZXRDaGFyKCk7CiAgCiEgCQkJaWYgKGMyID09IEVPRikKICAJCQl7CiEgCQkJCXJl c3VsdCA9IHRydWU7CiEgCQkJCWlmIChjc3ZfbW9kZSkKISAJCQkJCUNvcHlEb25lUGVlayhj MiwgdHJ1ZSk7CiAgCQkJCWJyZWFrOwogIAkJCX0KISAJCQlpZiAoYzIgPT0gJy4nKQohIAkJ CXsKISAJCQkJaWYgKGNzdl9tb2RlKQohIAkJCQkJQ29weURvbmVQZWVrKGMyLCB0cnVlKTsg LyogYWxsb3cga2VlcCBjYWxsaW5nIEdldENoYXIoKSAqLwogIAohIAkJCQlpZiAoZW9sX3R5 cGUgPT0gRU9MX0NSTkwpCiEgCQkJCXsKISAJCQkJCWMgPSBDb3B5R2V0Q2hhcigpOwohIAkJ CQkJaWYgKGMgPT0gJ1xuJykKISAJCQkJCQllcmVwb3J0KEVSUk9SLAohIAkJCQkJCQkJKGVy cmNvZGUoRVJSQ09ERV9CQURfQ09QWV9GSUxFX0ZPUk1BVCksCiEgCQkJCQkJCQkgZXJybXNn KCJlbmQtb2YtY29weSBtYXJrZXIgZG9lcyBub3QgbWF0Y2ggcHJldmlvdXMgbmV3bGluZSBz dHlsZSIpKSk7CiEgCQkJCQlpZiAoYyAhPSAnXHInKQohIAkJCQkJCWVyZXBvcnQoRVJST1Is CiEgCQkJCQkJCQkoZXJyY29kZShFUlJDT0RFX0JBRF9DT1BZX0ZJTEVfRk9STUFUKSwKISAJ CQkJCQkJCSBlcnJtc2coImVuZC1vZi1jb3B5IG1hcmtlciBjb3JydXB0IikpKTsKICAJCQkJ fQotIAkJCQljID0gQ29weUdldENoYXIoKTsKLSAJCQkJaWYgKGMgIT0gJ1xyJyAmJiBjICE9 ICdcbicpCi0gCQkJCQllcmVwb3J0KEVSUk9SLAotIAkJCQkJCQkoZXJyY29kZShFUlJDT0RF X0JBRF9DT1BZX0ZJTEVfRk9STUFUKSwKLSAJCQkJCQkJIGVycm1zZygiZW5kLW9mLWNvcHkg bWFya2VyIGNvcnJ1cHQiKSkpOwotIAkJCQlpZiAoKGVvbF90eXBlID09IEVPTF9OTCAmJiBj ICE9ICdcbicpIHx8Ci0gCQkJCQkoZW9sX3R5cGUgPT0gRU9MX0NSTkwgJiYgYyAhPSAnXG4n KSB8fAotIAkJCQkJKGVvbF90eXBlID09IEVPTF9DUiAmJiBjICE9ICdccicpKQotIAkJCQkJ ZXJlcG9ydChFUlJPUiwKLSAJCQkJCQkJKGVycmNvZGUoRVJSQ09ERV9CQURfQ09QWV9GSUxF X0ZPUk1BVCksCi0gCQkJCQkJCSBlcnJtc2coImVuZC1vZi1jb3B5IG1hcmtlciBkb2VzIG5v dCBtYXRjaCBwcmV2aW91cyBuZXdsaW5lIHN0eWxlIikpKTsKICAKICAJCQkJLyoKISAJCQkJ ICogSW4gcHJvdG9jb2wgdmVyc2lvbiAzLCB3ZSBzaG91bGQgaWdub3JlIGFueXRoaW5nIGFm dGVyCiEgCQkJCSAqIFwuIHVwIHRvIHRoZSBwcm90b2NvbCBlbmQgb2YgY29weSBkYXRhLgko WFhYIG1heWJlCiEgCQkJCSAqIGJldHRlciBub3QgdG8gdHJlYXQgXC4gYXMgc3BlY2lhbD8p CiAgCQkJCSAqLwohIAkJCQlpZiAoY29weV9kZXN0ID09IENPUFlfTkVXX0ZFKQogIAkJCQl7 CiEgCQkJCQl3aGlsZSAoYyAhPSBFT0YpCiEgCQkJCQkJYyA9IENvcHlHZXRDaGFyKCk7CiAg CQkJCX0KISAJCQkJcmVzdWx0ID0gdHJ1ZTsJLyogcmVwb3J0IEVPRiAqLwogIAkJCQlicmVh azsKICAJCQl9CiAgCQkJCiEgCQkJaWYgKGNzdl9tb2RlKQohIAkJCQlDb3B5RG9uZVBlZWso YzIsIGZhbHNlKTsgLyogbm90IGEgZG90LCBzbyBwdXQgaXQgYmFjayAqLyAKICAJCQllbHNl CiEgCQkJCS8qIG5vdCBFT0YgbWFyaywgc28gZW1pdCBcIGFuZCBmb2xsb3dpbmcgY2hhciBs aXRlcmFsbHkgKi8KISAJCQkJYXBwZW5kU3RyaW5nSW5mb0NoYXJNYWNybygmbGluZV9idWYs ICdcXCcpOwogIAkJfQogIAohIAkJYXBwZW5kU3RyaW5nSW5mb0NoYXJNYWNybygmbGluZV9i dWYsIGMpOwogIAogIAkJLyoKISAJCSAqIFdoZW4gY2xpZW50IGVuY29kaW5nICE9IHNlcnZl ciwgbXVzdCBiZSBjYXJlZnVsIHRvIHJlYWQgdGhlCiEgCQkgKiBleHRyYSBieXRlcyBvZiBh IG11bHRpYnl0ZSBjaGFyYWN0ZXIgZXhhY3RseSwgc2luY2UgdGhlCiEgCQkgKiBlbmNvZGlu ZyBtaWdodCBub3QgZW5zdXJlIHRoZXkgZG9uJ3QgbG9vayBsaWtlIEFTQ0lJLiAgV2hlbiB0 aGUKISAJCSAqIGVuY29kaW5ncyBhcmUgdGhlIHNhbWUsIHdlIG5lZWQgbm90IGRvIHRoaXMs IHNpbmNlIG5vIHNlcnZlcgohIAkJICogZW5jb2Rpbmcgd2UgdXNlIGhhcyBBU0NJSS1saWtl IGZvbGxvd2luZyBieXRlcy4KICAJCSAqLwohIAkJaWYgKGNoYW5nZV9lbmNvZGluZykKISAJ CXsKISAJCQlzWzBdID0gYzsKISAJCQltYmxlbiA9IHBnX2VuY29kaW5nX21ibGVuKGNsaWVu dF9lbmNvZGluZywgcyk7CiEgCQkJZm9yIChqID0gMTsgaiA8IG1ibGVuOyBqKyspCiEgCQkJ ewohIAkJCQljID0gQ29weUdldENoYXIoKTsKISAJCQkJaWYgKGMgPT0gRU9GKQogIAkJCQl7 CiEgCQkJCQlyZXN1bHQgPSB0cnVlOwogIAkJCQkJYnJlYWs7CiAgCQkJCX0KLSAJCQkJYXBw ZW5kU3RyaW5nSW5mb0NoYXJNYWNybygmbGluZV9idWYsIGMpOwogIAkJCX0KISAJCQlpZiAo cmVzdWx0KQohIAkJCQlicmVhazsJCQkvKiBvdXQgb2Ygb3V0ZXIgbG9vcCAqLwogIAkJfQot IAl9CQkJCQkJCS8qIGVuZCBvZiBvdXRlciBsb29wICovCiAgCiEgCS8qIERvbmUgcmVhZGlu ZyB0aGUgbGluZS4gIENvbnZlcnQgaXQgdG8gc2VydmVyIGVuY29kaW5nLiAqLwohIAlpZiAo Y2hhbmdlX2VuY29kaW5nKQogIAl7CiAgCQljdnQgPSAoY2hhciAqKSBwZ19jbGllbnRfdG9f c2VydmVyKCh1bnNpZ25lZCBjaGFyICopIGxpbmVfYnVmLmRhdGEsCiAgCQkJCQkJCQkJCSAg IGxpbmVfYnVmLmxlbik7Ci0tLSAxOTg2LDI0MjQgLS0tLQogIAogIAogIC8qCiEgICogRmlu ZHMgdGhlIG5leHQgVEVYVCBsaW5lIHRoYXQgaXMgaW4gdGhlIGlucHV0IGJ1ZmZlciBhbmQg bG9hZHMgCiEgICogaXQgaW50byBsaW5lX2J1Zi4gUmV0dXJucyBhbiBpbmRpY2F0aW9uIGlm IHRoZSBsaW5lIHRoYXQgd2FzIHJlYWQgCiEgICogaXMgY29tcGxldGUgKGlmIGFuIHVuZXNj YXBlZCBsaW5lLWVuZCB3YXMgZW5jb3VudGVyZWQpLiBJZiB3ZSAKISAgKiByZWFjaGVkIHRo ZSBlbmQgb2YgYnVmZmVyIGJlZm9yZSB0aGUgd2hvbGUgbGluZSB3YXMgd3JpdHRlbiBpbnRv IHRoZQohICAqIGxpbmUgYnVmZmVyIHRoZW4gcmV0dXJucyBmYWxzZS4KICAgKi8KICBzdGF0 aWMgYm9vbAohIENvcHlSZWFkTGluZVRleHQoc2l6ZV90IGJ5dGVzcmVhZCwgY2hhciAqZXNj YXBlKQogIHsKISAJaW50CQkJbGluZXNpemU7CiEgCWJvb2wJCXRyYW5zY29kZSA9IChjbGll bnRfZW5jb2RpbmcgIT0gc2VydmVyX2VuY29kaW5nKTsKICAJY2hhcgkgICAqY3Z0OwohIAlj aGFyICAgICAgICBlc2NhcGVjID0gJ1wwJzsKICAKICAJLyogbWFyayB0aGF0IGVuY29kaW5n IGNvbnZlcnNpb24gaGFzbid0IG9jY3VycmVkIHlldCAqLwogIAlsaW5lX2J1Zl9jb252ZXJ0 ZWQgPSBmYWxzZTsKICAKISAJLyoKISAJICogc2V0IHRoZSBlc2NhcGUgY2hhciBmb3IgdGV4 dCBmb3JtYXQgKCdcXCcgYnkgZGVmYXVsdCkuCiEgCSAqLwohIAllc2NhcGVjID0gZXNjYXBl WzBdOwogIAogIAkvKgohIAkgKiBEZXRlY3QgZW5kIG9mIGxpbmUgdHlwZSBpZiBub3QgYWxy ZWFkeSBkZXRlY3RlZC4KICAJICovCiEgCWlmIChlb2xfdHlwZSA9PSBFT0xfVU5LTk9XTikK ICAJewohIAkJYm9vbAkJZW9sX2RldGVjdGVkID0gRGV0ZWN0TGluZUVuZChieXRlc3JlYWQs IE5VTEwsIGVzY2FwZSk7CiEgCiEgCQlpZiAoIWVvbF9kZXRlY3RlZCkKICAJCXsKISAJCQkv KiBsb2FkIGVudGlyZSBpbnB1dCBidWZmZXIgaW50byBsaW5lIGJ1ZiwgYW5kIHF1aXQgKi8K ISAJCQlhcHBlbmRCaW5hcnlTdHJpbmdJbmZvKCZsaW5lX2J1ZiwgaW5wdXRfYnVmLCBDT1BZ X0JVRl9TSVpFKTsKISAJCQlidWZfZG9uZSA9IHRydWU7CiEgCQkJcmV0dXJuIGZhbHNlOwoh IAkJfQogIAkJfQogIAohIAkvKgohIAkgKiBTcGVjaWFsIGNhc2U6IGVvbCBpcyBDUk5MLCBs YXN0IGJ5dGUgb2YgcHJldmlvdXMgYnVmZmVyIHdhcyBhbgohIAkgKiB1bmVzY2FwZWQgQ1Ig YW5kIDFzdCBieXRlIG9mIGN1cnJlbnQgYnVmZmVyIGlzIE5MLiBXZSBjaGVjayBmb3IKISAJ ICogdGhhdCBoZXJlLgohIAkgKi8KISAJaWYgKGVvbF90eXBlID09IEVPTF9DUk5MKQohIAl7 CiEgCQkvKiBpZiB3ZSBzdGFydGVkIHNjYW5uaW5nIGZyb20gdGhlIDFzdCBieXRlIG9mIHRo ZSBidWZmZXIgKi8KISAJCWlmIChiZWdsb2MgPT0gaW5wdXRfYnVmKQohIAkJewohIAkJCS8q IGFuZCBoYWQgYSBDUiBpbiBsYXN0IGJ5dGUgb2YgcHJldiBidWYgKi8KISAJCQlpZiAoY3Jf aW5fcHJldmJ1ZikKISAJCQl7CiEgCQkJCS8qCiEgCQkJCSAqIGlmIHRoaXMgMXN0IGJ5dGUg aW4gYnVmZmVyIGlzIDJuZCBieXRlIG9mIGxpbmUgZW5kIHNlcXVlbmNlCiEgCQkJCSAqIChs aW5lZmVlZCkKISAJCQkJICovCiEgCQkJCWlmICgqYmVnbG9jID09IGVvbF9jaFsxXSkKICAJ CXsKICAJCQkvKiAgCiEgCQkJCQkgKiBsb2FkIHRoYXQgb25lIGxpbmVmZWVkIGJ5dGUgYW5k IGluZGljYXRlIHdlIGFyZSBkb25lCiEgCQkJCQkgKiB3aXRoIHRoZSBkYXRhIGxpbmUKICAJ CQkgKi8gCisgCQkJCQlhcHBlbmRCaW5hcnlTdHJpbmdJbmZvKCZsaW5lX2J1ZiwgYmVnbG9j LCAxKTsKKyAJCQkJCWJ1ZmZlcl9pbmRleCsrOworIAkJCQkJYmVnbG9jKys7CisgCQkJCQll c2NfaW5fcHJldmJ1ZiA9IGZhbHNlOworIAkJCQkJY3JfaW5fcHJldmJ1ZiA9IGZhbHNlOwor IAkJCQkJcmV0dXJuIHRydWU7CisgCQkJCX0KKyAJCQl9CiAgCiEgCQkJY3JfaW5fcHJldmJ1 ZiA9IGZhbHNlOwohIAkJfQohIAl9CiAgCiAgCQkJLyoKISAJICogKHdlIG5lZWQgYSBsb29w IHNvIHRoYXQgaWYgZW9sX2NoIGlzIGZvdW5kLCBidXQgcHJldiBjaCBpcyBiYWNrc2xhc2gs CiEgCSAqIHdlIGNhbiBzZWFyY2ggZm9yIHRoZSBuZXh0IGVvbF9jaCkKICAJCQkgKi8gCiEg CXdoaWxlICh0cnVlKQohIAl7CiEgCQkvKiByZWFjaGVkIGVuZCBvZiBidWZmZXIgKi8KISAJ CWlmICgoZW5kbG9jID0gc2NhblRleHRMaW5lKGJlZ2xvYywgZW9sX2NoWzBdLCBieXRlc3Jl YWQgLSBidWZmZXJfaW5kZXgpKSA9PSBOVUxMKQohIAkJewohIAkJCWxpbmVzaXplID0gQ09Q WV9CVUZfU0laRSAtIChiZWdsb2MgLSBpbnB1dF9idWYpOwohIAkJCWFwcGVuZEJpbmFyeVN0 cmluZ0luZm8oJmxpbmVfYnVmLCBiZWdsb2MsIGxpbmVzaXplKTsKICAKISAJCQlpZiAobGlu ZV9idWYubGVuID4gMSkKICAJCXsKISAJCQkJY2hhcgkgICAqbGFzdF9jaCA9IGxpbmVfYnVm LmRhdGEgKyBsaW5lX2J1Zi5sZW4gLSAxOyAvKiBiZWZvcmUgdGVybWluYXRpbmcgXDAgKi8K ISAKISAJCQkJaWYgKCpsYXN0X2NoID09IGVzY2FwZWMpCiAgCQkJewohIAkJCQkJZXNjX2lu X3ByZXZidWYgPSB0cnVlOwohIAohIAkJCQkJaWYgKGxpbmVfYnVmLmxlbiA+IDIpCiEgCQkJ CQl7CiEgCQkJCQkJbGFzdF9jaC0tOwohIAkJCQkJCWlmICgqbGFzdF9jaCA9PSBlc2NhcGVj KQohIAkJCQkJCQllc2NfaW5fcHJldmJ1ZiA9IGZhbHNlOwogIAkJCX0KISAJCQkJfQohIAkJ CQllbHNlIGlmICgqbGFzdF9jaCA9PSAnXHInKQogIAkJCXsKISAJCQkJCWlmIChlb2xfdHlw ZSA9PSBFT0xfQ1JOTCkKISAJCQkJCQljcl9pbl9wcmV2YnVmID0gdHJ1ZTsKISAJCQkJfQoh IAkJCX0KICAKISAJCQlsaW5lX2RvbmUgPSBmYWxzZTsKISAJCQlidWZfZG9uZSA9IHRydWU7 CiEgCQkJYnJlYWs7CiAgCQkJCX0KICAJCQkJZWxzZQorIAkJCS8qIGZvdW5kIHRoZSAxc3Qg ZW9sIGNoIGluIGlucHV0X2J1Zi4gKi8KICAJCQkJewohIAkJCWJvb2wJCWVvbF9mb3VuZCA9 IHRydWU7CiEgCQkJYm9vbAkJZW9sX2VzY2FwZWQgPSB0cnVlOwohIAohIAkJCS8qCiEgCQkJ ICogTG9hZCB0aGF0IHBpZWNlIG9mIGRhdGEgKHBvdGVudGlhbGx5IGEgZGF0YSBsaW5lKSBp bnRvIHRoZSBsaW5lIGJ1ZmZlciwKISAJCQkgKiBhbmQgdXBkYXRlIHRoZSBwb2ludGVycyBm b3IgdGhlIG5leHQgc2Nhbi4KISAJCQkgKi8KISAJCQlsaW5lc2l6ZSA9IGVuZGxvYyAtIGJl Z2xvYyArIDE7CiEgCQkJYXBwZW5kQmluYXJ5U3RyaW5nSW5mbygmbGluZV9idWYsIGJlZ2xv YywgbGluZXNpemUpOwohIAkJCWJ1ZmZlcl9pbmRleCArPSBsaW5lc2l6ZTsKISAJCQliZWds b2MgPSBlbmRsb2MgKyAxOwohIAogIAkJCQkJaWYgKGVvbF90eXBlID09IEVPTF9DUk5MKQog IAkJCQkJewohIAkJCQkvKiBjaGVjayBpZiB0aGVyZSBpcyBhICdcbicgYWZ0ZXIgdGhlICdc cicgKi8KISAJCQkJaWYgKCooZW5kbG9jICsgMSkgPT0gJ1xuJykKISAJCQkJewohIAkJCQkJ LyogdGhpcyBpcyBhIGxpbmUgZW5kICovCiEgCQkJCQlhcHBlbmRCaW5hcnlTdHJpbmdJbmZv KCZsaW5lX2J1ZiwgYmVnbG9jLCAxKTsJCS8qIGxvYWQgdGhhdCAnXG4nICovCiEgCQkJCQli dWZmZXJfaW5kZXgrKzsKISAJCQkJCWJlZ2xvYysrOwohIAkJCQl9CiAgCQkJCQkJZWxzZQoh IAkJCQkJLyoganVzdCBhIENSLCBub3QgYSBsaW5lIGVuZCAqLwohIAkJCQkJZW9sX2ZvdW5k ID0gZmFsc2U7CiAgCQkJCQl9CiAgCiAgCQkJCQkvKgohIAkJCSAqIGluIHNvbWUgY2FzZXMs IHRoaXMgZW5kIG9mIGxpbmUgY2hhciBoYXBwZW5zIHRvIGJlIHRoZQohIAkJCSAqIGxhc3Qg Y2hhcmFjdGVyIGluIHRoZSBidWZmZXIuIHdlIG5lZWQgdG8gY2F0Y2ggdGhhdC4KICAJCQkJ CSAqLwohIAkJCWlmIChidWZmZXJfaW5kZXggPj0gYnl0ZXNyZWFkKQohIAkJCQlidWZfZG9u ZSA9IHRydWU7CiEgCiEgCQkJLyoKISAJCQkgKiBDaGVjayBpZiB0aGUgMXN0IGVuZCBvZiBs aW5lIGNoIGlzIGVzY2FwZWQuCiEgCQkJICovCiEgCQkJaWYgKGVuZGxvYyAhPSBpbnB1dF9i dWYpCS8qIGNhbiB3ZSBsb29rIDEgY2hhciBiYWNrPyAqLwohIAkJCXsKISAJCQkJaWYgKCoo ZW5kbG9jIC0gMSkgIT0gZXNjYXBlYykJLyogcHJldiBjaGFyIGlzIG5vdCBhbiBlc2NhcGUg Ki8KISAJCQkJCWVvbF9lc2NhcGVkID0gZmFsc2U7CiEgCQkJCWVsc2UKISAJCQkJCS8qIHBy ZXYgY2hhciBpcyBhbiBlc2NhcGUgKi8KISAJCQkJewohIAkJCQkJaWYgKGVuZGxvYyAhPSAo aW5wdXRfYnVmICsgMSkpCQkvKiBjYW4gd2UgbG9vayBhbm90aGVyCiEgCQkJCQkJCQkJCQkJ CQkgKiBjaGFyIGJhY2s/ICovCiEgCQkJCQl7CiEgCQkJCQkJLyogaXQncyBhIGRvdWJsZSBl c2NhcGUgY2hhciwgc28gaXQncyBub3QgYW4gZXNjYXBlICovCiEgCQkJCQkJaWYgKCooZW5k bG9jIC0gMikgPT0gZXNjYXBlYykKISAJCQkJCQkJZW9sX2VzY2FwZWQgPSBmYWxzZTsKISAJ CQkJCQkvKiBlbHNlIGl0J3MgYSBzaW5nbGUgZXNjYXBlIGNoYXIsIHNvIEVPTCBpcyBhc2Nh cGVkICovCiEgCQkJCQl9CiEgCQkJCQllbHNlCiEgCQkJCQl7CiEgCQkJCQkJLyogd2UgbmVl ZCB0byBjaGVjayBpbiB0aGUgbGFzdCBidWZmZXIgKi8KISAJCQkJCQlpZiAoZXNjX2luX3By ZXZidWYpCQkKISAJCQkJCQkJZW9sX2VzY2FwZWQgPSBmYWxzZTsKICAJCQkJfQogIAkJCX0K ICAJCX0KISAJCQllbHNlCiEgCQkJCS8qCiEgCQkJCSAqIHRoaXMgZW9sIGNoIGlzIGZpcnN0 IGNoIGluIGJ1ZmZlciwgY2hlY2sgZm9yIGVzY2FwZSBpbiBwcmV2IGJ1ZgohIAkJCQkgKi8K ICAJCXsKISAJCQkJaWYgKCFlc2NfaW5fcHJldmJ1ZikKISAJCQkJCWVvbF9lc2NhcGVkID0g ZmFsc2U7CiAgCQkJfQogIAohIAkJCWVzY19pbl9wcmV2YnVmID0gZmFsc2U7CQkvKiByZXNl dCB2YXJpYWJsZSAqLwogIAohIAkJCS8qCiEgCQkJICogaWYgZW9sIHdhcyBmb3VuZCwgYW5k IGl0IGlzbid0IGVzY2FwZWQsIGxpbmUgaXMgZG9uZQohIAkJCSAqLwohIAkJCWlmICgoZW9s X2VzY2FwZWQgPT0gZmFsc2UpICYmIGVvbF9mb3VuZCkKICAJCQl7CiEgCQkJCWxpbmVfZG9u ZSA9IHRydWU7CiAgCQkJCWJyZWFrOwogIAkJCX0KISAJCQkJZWxzZQohIAkJCQkvKiBzdGF5 IGluIHRoZSBsb29wIGFuZCBwcm9jZXNzIHNvbWUgbW9yZSBkYXRhLiAqLwohIAkJCQlsaW5l X2RvbmUgPSBmYWxzZTsKISAJCQkJCQohIAkJfQkJCQkJCS8qIGVuZCBvZiBmb3VuZCBlb2xf Y2ggKi8KISAJCX0KICAKISAJLyoKISAJICogRG9uZSByZWFkaW5nIHRoZSBsaW5lLiBDb252 ZXJ0IGl0IHRvIHNlcnZlciBlbmNvZGluZy4KISAJICovCiEgCWlmICh0cmFuc2NvZGUpCiEg CXsKISAJCWN2dCA9IChjaGFyICopIHBnX2NsaWVudF90b19zZXJ2ZXIoKHVuc2lnbmVkIGNo YXIgKikgbGluZV9idWYuZGF0YSwKISAJCQkJCQkJCQkJICAgbGluZV9idWYubGVuKTsKISAJ CWlmIChjdnQgIT0gbGluZV9idWYuZGF0YSkKISAJCXsKISAJCQkvKiB0cmFuc2ZlciBjb252 ZXJ0ZWQgZGF0YSBiYWNrIHRvIGxpbmVfYnVmICovCiEgCQkJbGluZV9idWYubGVuID0gMDsK ISAJCQlsaW5lX2J1Zi5kYXRhWzBdID0gJ1wwJzsKISAJCQlsaW5lX2J1Zi5jdXJzb3IgPSAw OwohIAkJCWFwcGVuZEJpbmFyeVN0cmluZ0luZm8oJmxpbmVfYnVmLCBjdnQsIHN0cmxlbihj dnQpKTsKISAJCX0KISAJfQohIAkJCQohIAkvKiBpbmRpY2F0ZSB0aGF0IGNvbnZlcnNpb24g aGFkIG9jY3VyZWQgKi8KISAJbGluZV9idWZfY29udmVydGVkID0gdHJ1ZTsKISAKISAJLyoK ISAJICogY2hlY2sgaWYgdGhpcyBsaW5lIGlzIGFuIGVuZCBtYXJrZXIgLS0gIlwuIgohIAkg Ki8KISAJZW5kX21hcmtlciA9IGZhbHNlOwohIAohIAlzd2l0Y2ggKGVvbF90eXBlKQohIAkJ CXsKISAJCWNhc2UgRU9MX05MOgohIAkJCWlmICghc3RyY21wKGxpbmVfYnVmLmRhdGEsICJc XC5cbiIpKQohIAkJCQllbmRfbWFya2VyID0gdHJ1ZTsKISAJCQlicmVhazsKISAJCWNhc2Ug RU9MX0NSOgohIAkJCWlmICghc3RyY21wKGxpbmVfYnVmLmRhdGEsICJcXC5cciIpKQohIAkJ CQllbmRfbWFya2VyID0gdHJ1ZTsKISAJCQlicmVhazsKISAJCWNhc2UgRU9MX0NSTkw6CiEg CQkJaWYgKCFzdHJjbXAobGluZV9idWYuZGF0YSwgIlxcLlxyXG4iKSkKISAJCQkJZW5kX21h cmtlciA9IHRydWU7CiEgCQkJYnJlYWs7CiEgCQljYXNlIEVPTF9VTktOT1dOOgohIAkJCQli cmVhazsKISAJCQl9CiEgCiEgCWlmIChlbmRfbWFya2VyKQohIAkJCXsKISAJCWZlX2VvZiA9 IHRydWU7CiEgCQkvKiB3ZSBkb24ndCB3YW50IHRvIHByb2Nlc3MgYSBcLiBhcyBkYXRhIGxp bmUsIHdhbnQgdG8gcXVpdC4gKi8KISAJCWxpbmVfZG9uZSA9IGZhbHNlOwohIAkJYnVmX2Rv bmUgPSB0cnVlOwohIAl9CiEgCiEgCXJldHVybiBsaW5lX2RvbmU7CiEgfQohIAohIC8qCiEg ICogRmluZHMgdGhlIG5leHQgQ1NWIGxpbmUgdGhhdCBpcyBpbiB0aGUgaW5wdXQgYnVmZmVy IGFuZCBsb2FkcyAKISAgKiBpdCBpbnRvIGxpbmVfYnVmLiBSZXR1cm5zIGFuIGluZGljYXRp b24gaWYgdGhlIGxpbmUgdGhhdCB3YXMgcmVhZCAKISAgKiBpcyBjb21wbGV0ZSAoaWYgYW4g dW5lc2NhcGVkIGxpbmUtZW5kIHdhcyBlbmNvdW50ZXJlZCkuIElmIHdlIAohICAqIHJlYWNo ZWQgdGhlIGVuZCBvZiBidWZmZXIgYmVmb3JlIHRoZSB3aG9sZSBsaW5lIHdhcyB3cml0dGVu IGludG8gdGhlCiEgICogbGluZSBidWZmZXIgdGhlbiByZXR1cm5zIGZhbHNlLgohICAqLwoh IHN0YXRpYyBib29sCiEgQ29weVJlYWRMaW5lQ1NWKHNpemVfdCBieXRlc3JlYWQsIGNoYXIg KnF1b3RlLCBjaGFyICplc2NhcGUpCiEgewohIAlpbnQJCQlsaW5lc2l6ZTsKISAJY2hhcgkg ICAqY3Z0OwohIAljaGFyICAgICAgICBxdW90ZWMgPSAnXDAnLAohICAgICAgICAgICAgICAg ICBlc2NhcGVjID0gJ1wwJzsKISAJYm9vbAkJdHJhbnNjb2RlID0gKGNsaWVudF9lbmNvZGlu ZyAhPSBzZXJ2ZXJfZW5jb2RpbmcpOwohIAkKISAJLyogbWFyayB0aGF0IGVuY29kaW5nIGNv bnZlcnNpb24gaGFzbid0IG9jY3VycmVkIHlldCAqLwohIAlsaW5lX2J1Zl9jb252ZXJ0ZWQg PSBmYWxzZTsKISAJCiEgCWVzY2FwZWMgPSBlc2NhcGVbMF07CiEgCXF1b3RlYyA9IHF1b3Rl WzBdOwohIAkKISAgICAgLyogaWdub3JlIHNwZWNpYWwgZXNjYXBlIHByb2Nlc3NpbmcgaWYg aXQncyB0aGUgc2FtZSBhcyBxdW90ZWMgKi8KISAJaWYgKHF1b3RlYyA9PSBlc2NhcGVjKQoh IAkJZXNjYXBlYyA9ICdcMCc7CiEgCiEgCS8qCiEgCSAqIERldGVjdCBlbmQgb2YgbGluZSB0 eXBlIGlmIG5vdCBhbHJlYWR5IGRldGVjdGVkLgohIAkgKi8KISAJaWYgKGVvbF90eXBlID09 IEVPTF9VTktOT1dOKQohIAl7CiEgCQlib29sCQllb2xfZGV0ZWN0ZWQgPSBEZXRlY3RMaW5l RW5kKGJ5dGVzcmVhZCwgcXVvdGUsIGVzY2FwZSk7CiEgCQkKISAJCWlmICghZW9sX2RldGVj dGVkKQohIAkJewohIAkJCS8qIGxvYWQgZW50aXJlIGlucHV0IGJ1ZmZlciBpbnRvIGxpbmUg YnVmLCBhbmQgcXVpdCAqLwohIAkJCWFwcGVuZEJpbmFyeVN0cmluZ0luZm8oJmxpbmVfYnVm LCBpbnB1dF9idWYsIENPUFlfQlVGX1NJWkUpOwohIAkJCWxpbmVfZG9uZSA9IGZhbHNlOwoh IAkJCWJ1Zl9kb25lID0gdHJ1ZTsKISAJCQkKISAJCQlyZXR1cm4gbGluZV9kb25lOwohIAkJ fQohIAl9CiEgCQohIAkvKgohIAkgKiBTcGVjaWFsIGNhc2U6IGVvbCBpcyBDUk5MLCBsYXN0 IGJ5dGUgb2YgcHJldmlvdXMgYnVmZmVyIHdhcyBhbgohIAkgKiB1bmVzY2FwZWQgQ1IgYW5k IDFzdCBieXRlIG9mIGN1cnJlbnQgYnVmZmVyIGlzIE5MLiBXZSBjaGVjayBmb3IKISAJICog dGhhdCBoZXJlLgohIAkgKi8KISAJCQkJaWYgKGVvbF90eXBlID09IEVPTF9DUk5MKQohIAkJ CQl7CiEgCQkvKiBpZiB3ZSBzdGFydGVkIHNjYW5uaW5nIGZyb20gdGhlIDFzdCBieXRlIG9m IHRoZSBidWZmZXIgKi8KISAJCWlmIChiZWdsb2MgPT0gaW5wdXRfYnVmKQohIAkJewohIAkJ CS8qIGFuZCBoYWQgYSBDUiBpbiBsYXN0IGJ5dGUgb2YgcHJldiBidWYgKi8KISAJCQlpZiAo Y3JfaW5fcHJldmJ1ZikKISAJCQl7CiEgCQkJCS8qCiEgCQkJCSAqIGlmIHRoaXMgMXN0IGJ5 dGUgaW4gYnVmZmVyIGlzIDJuZCBieXRlIG9mIGxpbmUgZW5kIHNlcXVlbmNlCiEgCQkJCSAq IChsaW5lZmVlZCkKISAJCQkJICovCiEgCQkJCWlmICgqYmVnbG9jID09IGVvbF9jaFsxXSkK ISAJCQkJewohIAkJCQkJLyoKISAJCQkJCSAqIGxvYWQgdGhhdCBvbmUgbGluZWZlZWQgYnl0 ZSBhbmQgaW5kaWNhdGUgd2UgYXJlIGRvbmUKISAJCQkJCSAqIHdpdGggdGhlIGRhdGEgbGlu ZQohIAkJCQkJICovCiEgCQkJCQlhcHBlbmRCaW5hcnlTdHJpbmdJbmZvKCZsaW5lX2J1Ziwg YmVnbG9jLCAxKTsKISAJCQkJCWJ1ZmZlcl9pbmRleCsrOwohIAkJCQkJYmVnbG9jKys7CiEg CQkJCQkKISAJCQkJCWxpbmVfZG9uZSA9IHRydWU7CiEgCQkJCQllc2NfaW5fcHJldmJ1ZiA9 IGZhbHNlOwohIAkJCQkJY3JfaW5fcHJldmJ1ZiA9IGZhbHNlOwohIAkJCQkJCiEgCQkJCQly ZXR1cm4gbGluZV9kb25lOwohIAkJCQl9CiEgCQkJfQohIAkJCQohIAkJCWNyX2luX3ByZXZi dWYgPSBmYWxzZTsKISAJCX0KICAJCQkJfQogIAogIAkJCQkvKgohIAkgKiAod2UgbmVlZCBh IGxvb3Agc28gdGhhdCBpZiBlb2xfY2ggaXMgZm91bmQsIGJ1dCB3ZSBhcmUgaW4gcXVvdGVz LAohIAkgKiB3ZSBjYW4gc2VhcmNoIGZvciB0aGUgbmV4dCBlb2xfY2gpCiAgCQkJCSAqLwoh IAl3aGlsZSAodHJ1ZSkKISAJewohIAkJLyogcmVhY2hlZCBlbmQgb2YgYnVmZmVyICovCiEg CQlpZiAoKGVuZGxvYyA9IHNjYW5DU1ZMaW5lKGJlZ2xvYywgZW9sX2NoWzBdLCBlc2NhcGVj LCBxdW90ZWMsIGJ5dGVzcmVhZCAtIGJ1ZmZlcl9pbmRleCkpID09IE5VTEwpCiEgCQkJCXsK ISAJCQlsaW5lc2l6ZSA9IENPUFlfQlVGX1NJWkUgLSAoYmVnbG9jIC0gaW5wdXRfYnVmKTsK ISAJCQlhcHBlbmRCaW5hcnlTdHJpbmdJbmZvKCZsaW5lX2J1ZiwgYmVnbG9jLCBsaW5lc2l6 ZSk7CiEgCQkJCiEgCQkJaWYgKGxhc3Rfd2FzX2VzYykKISAJCQkJZXNjX2luX3ByZXZidWYg PSB0cnVlOwohIAkJCQohIAkJCWlmIChsaW5lX2J1Zi5sZW4gPiAxKQohIAkJCXsKISAJCQkJ Y2hhcgkgICAqbGFzdF9jaCA9IGxpbmVfYnVmLmRhdGEgKyBsaW5lX2J1Zi5sZW4gLSAxOyAv KiBiZWZvcmUgdGVybWluYXRpbmcgXDAgKi8KISAJCQkJCiEgCQkJCWlmICgqbGFzdF9jaCA9 PSAnXHInKQogIAkJCQl7CiEgCQkJCQlpZiAoZW9sX3R5cGUgPT0gRU9MX0NSTkwpCiEgCQkJ CQkJY3JfaW5fcHJldmJ1ZiA9IHRydWU7CiAgCQkJCX0KISAJCQl9CiEgCQkJCiEgCQkJbGlu ZV9kb25lID0gZmFsc2U7CiEgCQkJYnVmX2RvbmUgPSB0cnVlOwogIAkJCQlicmVhazsKICAJ CQl9CisgCQllbHNlCisgCQkJLyogZm91bmQgMXN0IGVvbCBjaGFyIGluIGlucHV0X2J1Zi4g Ki8KKyAJCXsKKyAJCQlib29sCQllb2xfZm91bmQgPSB0cnVlOwogIAkJCQohIAkJCS8qCiEg CQkJICogTG9hZCB0aGF0IHBpZWNlIG9mIGRhdGEgKHBvdGVudGlhbGx5IGEgZGF0YSBsaW5l KSBpbnRvIHRoZSBsaW5lIGJ1ZmZlciwKISAJCQkgKiBhbmQgdXBkYXRlIHRoZSBwb2ludGVy cyBmb3IgdGhlIG5leHQgc2Nhbi4KISAJCQkgKi8KISAJCQlsaW5lc2l6ZSA9IGVuZGxvYyAt IGJlZ2xvYyArIDE7CiEgCQkJYXBwZW5kQmluYXJ5U3RyaW5nSW5mbygmbGluZV9idWYsIGJl Z2xvYywgbGluZXNpemUpOwohIAkJCWJ1ZmZlcl9pbmRleCArPSBsaW5lc2l6ZTsKISAJCQli ZWdsb2MgPSBlbmRsb2MgKyAxOwohIAkJCQohIAkJCS8qIGVuZCBvZiBsaW5lIG9ubHkgaWYg bm90IGluIHF1b3RlcyAqLwohIAkJCWlmKGluX3F1b3RlKQohIAkJCXsKISAJCQkJbGluZV9k b25lID0gZmFsc2U7CiEgCQkJCS8qIGJ1ZiBkb25lLCBidXQgc3RpbGwgaW4gcXVvdGUgKi8K ISAJCQkJaWYgKGJ1ZmZlcl9pbmRleCA+PSBieXRlc3JlYWQpCiEgCQkJCXsKISAJCQkJCWJ1 Zl9kb25lID0gdHJ1ZTsKISAJCQkJCWJyZWFrOwohIAkJCQl9CiEgCQkJfQohIAkJCWVsc2UK ISAJCQl7CiEgCQkJCS8qIGlmIGRvcyBlb2wsIGNoZWNrIGZvciAnXG4nIGFmdGVyIHRoZSAn XHInICovCiEgCQkJCWlmIChlb2xfdHlwZSA9PSBFT0xfQ1JOTCkKISAJCQkJewohIAkJCQkJ aWYgKCooZW5kbG9jICsgMSkgPT0gJ1xuJykKISAJCQkJCXsKISAJCQkJCQkvKiB0aGlzIGlz IGEgbGluZSBlbmQgKi8KISAJCQkJCQlhcHBlbmRCaW5hcnlTdHJpbmdJbmZvKCZsaW5lX2J1 ZiwgYmVnbG9jLCAxKTsJCS8qIGxvYWQgdGhhdCAnXG4nICovCiEgCQkJCQkJYnVmZmVyX2lu ZGV4Kys7CiEgCQkJCQkJYmVnbG9jKys7CiEgCQkJCQl9CiAgCQkJZWxzZQohIAkJCQkJCS8q IGp1c3QgYSBDUiwgbm90IGEgbGluZSBlbmQgKi8KISAJCQkJCQllb2xfZm91bmQgPSBmYWxz ZTsKICAJCX0KICAKISAJCQkJLyoKISAJCQkJICogaW4gc29tZSBjYXNlcywgdGhpcyBlbmQg b2YgbGluZSBjaGFyIGhhcHBlbnMgdG8gYmUgdGhlCiEgCQkJCSAqIGxhc3QgY2hhcmFjdGVy IGluIHRoZSBidWZmZXIuIHdlIG5lZWQgdG8gY2F0Y2ggdGhhdC4KISAJCQkJICovCiEgCQkJ CWlmIChidWZmZXJfaW5kZXggPj0gYnl0ZXNyZWFkKQohIAkJCQkJYnVmX2RvbmUgPSB0cnVl OwogIAogIAkJLyoKISAJCQkJICogaWYgZW9sIHdhcyBmb3VuZCBsaW5lIGlzIGRvbmUKICAJ CSAqLwohIAkJCQlpZiAoZW9sX2ZvdW5kKQogIAkJCQl7CiEgCQkJCQlsaW5lX2RvbmUgPSB0 cnVlOwogIAkJCQkJYnJlYWs7CiAgCQkJCX0KICAJCQl9CiEgCQl9CQkJCQkJLyogZW5kIG9m IGZvdW5kIGVvbF9jaCAqLwogIAkJfQogIAohIAkvKgohIAkgKiBEb25lIHJlYWRpbmcgdGhl IGxpbmUuIENvbnZlcnQgaXQgdG8gc2VydmVyIGVuY29kaW5nLgohIAkgKi8KISAJaWYgKHRy YW5zY29kZSkKICAJewogIAkJY3Z0ID0gKGNoYXIgKikgcGdfY2xpZW50X3RvX3NlcnZlcigo dW5zaWduZWQgY2hhciAqKSBsaW5lX2J1Zi5kYXRhLAogIAkJCQkJCQkJCQkgICBsaW5lX2J1 Zi5sZW4pOwoqKioqKioqKioqKioqKioKKioqIDIyNjUsMjI4NSAqKioqCiAgCQkJLyogdHJh bnNmZXIgY29udmVydGVkIGRhdGEgYmFjayB0byBsaW5lX2J1ZiAqLwogIAkJCWxpbmVfYnVm LmxlbiA9IDA7CiAgCQkJbGluZV9idWYuZGF0YVswXSA9ICdcMCc7CiAgCQkJYXBwZW5kQmlu YXJ5U3RyaW5nSW5mbygmbGluZV9idWYsIGN2dCwgc3RybGVuKGN2dCkpOwogIAkJfQogIAl9 CiAgCiEgCS8qIE5vdyBpdCdzIHNhZmUgdG8gdXNlIHRoZSBidWZmZXIgaW4gZXJyb3IgbWVz c2FnZXMgKi8KICAJbGluZV9idWZfY29udmVydGVkID0gdHJ1ZTsKICAKISAJcmV0dXJuIHJl c3VsdDsKICB9CiAgCiAgLyoKICAgKglSZXR1cm4gZGVjaW1hbCB2YWx1ZSBmb3IgYSBoZXhh ZGVjaW1hbCBkaWdpdAogICAqLwogIHN0YXRpYwohIGludCBHZXREZWNpbWFsRnJvbUhleChj aGFyIGhleCkKICB7CiAgCWlmIChpc2RpZ2l0KGhleCkpCiAgCQlyZXR1cm4gaGV4IC0gJzAn OwotLS0gMjQyNywyNDgwIC0tLS0KICAJCQkvKiB0cmFuc2ZlciBjb252ZXJ0ZWQgZGF0YSBi YWNrIHRvIGxpbmVfYnVmICovCiAgCQkJbGluZV9idWYubGVuID0gMDsKICAJCQlsaW5lX2J1 Zi5kYXRhWzBdID0gJ1wwJzsKKyAJCQlsaW5lX2J1Zi5jdXJzb3IgPSAwOwogIAkJCWFwcGVu ZEJpbmFyeVN0cmluZ0luZm8oJmxpbmVfYnVmLCBjdnQsIHN0cmxlbihjdnQpKTsKICAJCX0K ICAJfQogIAohIAkvKiBpbmRpY2F0ZSB0aGF0IGNvbnZlcnNpb24gaGFkIG9jY3VyZWQgKi8K ICAJbGluZV9idWZfY29udmVydGVkID0gdHJ1ZTsKICAKISAJLyoKISAJICogY2hlY2sgaWYg dGhpcyBsaW5lIGlzIGFuIGVuZCBtYXJrZXIgLS0gIlwuIgohIAkgKi8KISAJZW5kX21hcmtl ciA9IGZhbHNlOwohIAohIAlzd2l0Y2ggKGVvbF90eXBlKQohIAl7CiEgCQljYXNlIEVPTF9O TDoKISAJCQlpZiAoIXN0cmNtcChsaW5lX2J1Zi5kYXRhLCAiXFwuXG4iKSkKISAJCQkJZW5k X21hcmtlciA9IHRydWU7CiEgCQkJYnJlYWs7CiEgCQljYXNlIEVPTF9DUjoKISAJCQlpZiAo IXN0cmNtcChsaW5lX2J1Zi5kYXRhLCAiXFwuXHIiKSkKISAJCQkJZW5kX21hcmtlciA9IHRy dWU7CiEgCQkJYnJlYWs7CiEgCQljYXNlIEVPTF9DUk5MOgohIAkJCWlmICghc3RyY21wKGxp bmVfYnVmLmRhdGEsICJcXC5cclxuIikpCiEgCQkJCWVuZF9tYXJrZXIgPSB0cnVlOwohIAkJ CWJyZWFrOwohIAkJY2FzZSBFT0xfVU5LTk9XTjoKISAJCQlicmVhazsKISAJfQohIAohIAlp ZiAoZW5kX21hcmtlcikKISAJewohIAkJZmVfZW9mID0gdHJ1ZTsKISAJCS8qIHdlIGRvbid0 IHdhbnQgdG8gcHJvY2VzcyBhIFwuIGFzIGRhdGEgbGluZSwgd2FudCB0byBxdWl0LiAqLwoh IAkJbGluZV9kb25lID0gZmFsc2U7CiEgCQlidWZfZG9uZSA9IHRydWU7CiEgCX0KISAKISAJ cmV0dXJuIGxpbmVfZG9uZTsKICB9CiAgCiAgLyoKICAgKglSZXR1cm4gZGVjaW1hbCB2YWx1 ZSBmb3IgYSBoZXhhZGVjaW1hbCBkaWdpdAogICAqLwogIHN0YXRpYwohIGludAohIEdldERl Y2ltYWxGcm9tSGV4KGNoYXIgaGV4KQogIHsKICAJaWYgKGlzZGlnaXQoaGV4KSkKICAJCXJl dHVybiBoZXggLSAnMCc7CioqKioqKioqKioqKioqKgoqKiogMjI4NywyMzQ0ICoqKioKICAJ CXJldHVybiB0b2xvd2VyKGhleCkgLSAnYScgKyAxMDsKICB9CiAgCiEgLyotLS0tLS0tLS0t CiEgICogUmVhZCB0aGUgdmFsdWUgb2YgYSBzaW5nbGUgYXR0cmlidXRlLCBwZXJmb3JtaW5n IGRlLWVzY2FwaW5nIGFzIG5lZWRlZC4KISAgKgohICAqIGRlbGltIGlzIHRoZSBjb2x1bW4g ZGVsaW1pdGVyIHN0cmluZyAobXVzdCBiZSBqdXN0IG9uZSBieXRlIGZvciBub3cpLgohICAq IG51bGxfcHJpbnQgaXMgdGhlIG51bGwgbWFya2VyIHN0cmluZy4gIE5vdGUgdGhhdCB0aGlz IGlzIGNvbXBhcmVkIHRvCiEgICogdGhlIHByZS1kZS1lc2NhcGVkIGlucHV0IHN0cmluZy4K ISAgKgohICAqICpyZXN1bHQgaXMgc2V0IHRvIGluZGljYXRlIHdoYXQgdGVybWluYXRlZCB0 aGUgcmVhZDoKISAgKgkJTk9STUFMX0FUVFI6CWNvbHVtbiBkZWxpbWl0ZXIKISAgKgkJRU5E X09GX0xJTkU6CWVuZCBvZiBsaW5lCiEgICogSW4gZWl0aGVyIGNhc2UsIHRoZSBzdHJpbmcg cmVhZCB1cCB0byB0aGUgdGVybWluYXRvciBpcyByZXR1cm5lZC4KISAgKgohICAqICppc251 bGwgaXMgc2V0IHRydWUgb3IgZmFsc2UgZGVwZW5kaW5nIG9uIHdoZXRoZXIgdGhlIGlucHV0 IG1hdGNoZWQKISAgKiB0aGUgbnVsbCBtYXJrZXIuICBOb3RlIHRoYXQgdGhlIGNhbGxlciBj YW5ub3QgY2hlY2sgdGhpcyBzaW5jZSB0aGUKISAgKiByZXR1cm5lZCBzdHJpbmcgd2lsbCBi ZSB0aGUgcG9zdC1kZS1lc2NhcGluZyBlcXVpdmFsZW50LCB3aGljaCBtYXkKISAgKiBsb29r IHRoZSBzYW1lIGFzIHNvbWUgdmFsaWQgZGF0YSBzdHJpbmcuCiEgICotLS0tLS0tLS0tCiAg ICovCiEgc3RhdGljIGNoYXIgKgohIENvcHlSZWFkQXR0cmlidXRlKGNvbnN0IGNoYXIgKmRl bGltLCBjb25zdCBjaGFyICpudWxsX3ByaW50LAohIAkJCQkgIENvcHlSZWFkUmVzdWx0ICpy ZXN1bHQsIGJvb2wgKmlzbnVsbCkKICB7CiAgCWNoYXIJCWM7CiEgCWNoYXIJCWRlbGltYyA9 IGRlbGltWzBdOwohIAlpbnQJCQlzdGFydF9jdXJzb3IgPSBsaW5lX2J1Zi5jdXJzb3I7CiEg CWludAkJCWVuZF9jdXJzb3I7CiEgCWludAkJCWlucHV0X2xlbjsKICAKISAJLyogcmVzZXQg YXR0cmlidXRlX2J1ZiB0byBlbXB0eSAqLwohIAlhdHRyaWJ1dGVfYnVmLmxlbiA9IDA7CiEg CWF0dHJpYnV0ZV9idWYuZGF0YVswXSA9ICdcMCc7CiAgCiEgCS8qIHNldCBkZWZhdWx0IHN0 YXR1cyAqLwohIAkqcmVzdWx0ID0gRU5EX09GX0xJTkU7CiAgCiEgCWZvciAoOzspCiAgCXsK ISAJCWVuZF9jdXJzb3IgPSBsaW5lX2J1Zi5jdXJzb3I7CiEgCQlpZiAobGluZV9idWYuY3Vy c29yID49IGxpbmVfYnVmLmxlbikKISAJCQlicmVhazsKISAJCWMgPSBsaW5lX2J1Zi5kYXRh W2xpbmVfYnVmLmN1cnNvcisrXTsKISAJCWlmIChjID09IGRlbGltYykKICAJCXsKISAJCQkq cmVzdWx0ID0gTk9STUFMX0FUVFI7CiEgCQkJYnJlYWs7CiAgCQl9CiEgCQlpZiAoYyA9PSAn XFwnKQogIAkJewohIAkJCWlmIChsaW5lX2J1Zi5jdXJzb3IgPj0gbGluZV9idWYubGVuKQoh IAkJCQlicmVhazsKISAJCQljID0gbGluZV9idWYuZGF0YVtsaW5lX2J1Zi5jdXJzb3IrK107 CiEgCQkJc3dpdGNoIChjKQogIAkJCXsKICAJCQkJY2FzZSAnMCc6CiAgCQkJCWNhc2UgJzEn OgotLS0gMjQ4MiwyNzcxIC0tLS0KICAJCXJldHVybiB0b2xvd2VyKGhleCkgLSAnYScgKyAx MDsKICB9CiAgCiEgLyoKISAgKiBEZXRlY3RlZCB0aGUgZW9sIHR5cGUgYnkgbG9va2luZyBh dCB0aGUgZmlyc3QgZGF0YSByb3cuCiEgICogUG9zc2libGUgZW9sIHR5cGVzIGFyZSBOTCwg Q1IsIG9yIENSTkwuIElmIGVvbCB0eXBlIHdhcwohICAqIGRldGVjdGVkLCBpdCBpcyBzZXQg YW5kIGEgYm9vbGVhbiB0cnVlIGlzIHJldHVybmVkIHRvCiEgICogaW5kaWNhdGVkIGRldGVj dGlvbiB3YXMgc3VjY2Vzc2Z1bC4gSWYgdGhlIGZpcnN0IGRhdGEgcm93CiEgICogaXMgbG9u Z2VyIHRoYW4gdGhlIGlucHV0IGJ1ZmZlciwgd2UgcmV0dXJuIGZhbHNlIGFuZCB3aWxsCiEg ICogdHJ5IGFnYWluIGluIHRoZSBuZXh0IGJ1ZmZlci4KICAgKi8KISBzdGF0aWMgYm9vbAoh IERldGVjdExpbmVFbmQoc2l6ZV90IGJ5dGVzcmVhZCwgY2hhciAqcXVvdGUsIGNoYXIgKmVz Y2FwZSkKICB7CisgCWludCAgICAgICAgIGluZGV4ID0gMDsKICAJY2hhcgkJYzsKISAJY2hh cgkJcXVvdGVjID0gJ1wwJywKISAJCWVzY2FwZWMgPSAnXDAnOwohIAlib29sICAgICAgICBj c3YgPSBmYWxzZTsKICAKISAJaWYgKHF1b3RlKSAvKiBDU1YgZm9ybWF0ICovCiEgCXsKISAJ CWNzdiA9IHRydWU7CiEgCQlxdW90ZWMgPSBxdW90ZVswXTsKISAJCWVzY2FwZWMgPSBlc2Nh cGVbMF07CiEgCQkvKiBpZ25vcmUgc3BlY2lhbCBlc2NhcGUgcHJvY2Vzc2luZyBpZiBpdCdz IHRoZSBzYW1lIGFzIHF1b3RlYyAqLwohIAkJaWYgKHF1b3RlYyA9PSBlc2NhcGVjKQohIAkJ CWVzY2FwZWMgPSAnXDAnOwohIAl9CiAgCiEgCXdoaWxlIChpbmRleCA8IENPUFlfQlVGX1NJ WkUpCiEgCXsKISAJCWMgPSBpbnB1dF9idWZbaW5kZXhdOwogIAohIAkJaWYoY3N2KQogIAl7 CiEgCQkJaWYgKGluX3F1b3RlICYmIGMgPT0gZXNjYXBlYykKISAJCQkJbGFzdF93YXNfZXNj ID0gIWxhc3Rfd2FzX2VzYzsKISAJCQlpZiAoYyA9PSBxdW90ZWMgJiYgIWxhc3Rfd2FzX2Vz YykKISAJCQkJaW5fcXVvdGUgPSAhaW5fcXVvdGU7CiEgCQkJaWYgKGMgIT0gZXNjYXBlYykK ISAJCQkJbGFzdF93YXNfZXNjID0gZmFsc2U7CiEgCQl9CiEgCQkKISAJCWlmIChjID09ICdc bicpCiAgCQl7CiEgCQkJaWYoIWNzdiB8fCAoY3N2ICYmICFpbl9xdW90ZSkpCiEgCQkJewoh IAkJCQllb2xfdHlwZSA9IEVPTF9OTDsKISAJCQkJZW9sX2NoWzBdID0gJ1xuJzsKISAJCQkJ ZW9sX2NoWzFdID0gJ1wwJzsKISAJCQkJCiEgCQkJCWluX3F1b3RlID0gZmFsc2U7CiEgCQkJ CWxhc3Rfd2FzX2VzYyA9IGZhbHNlOwohIAkJCQlyZXR1cm4gdHJ1ZTsKISAJCQl9CiAgCQl9 CiEgCQlpZiAoYyA9PSAnXHInKQogIAkJewohIAkJCWlmKCFjc3YgfHwgKGNzdiAmJiAhaW5f cXVvdGUpKQohICAgICAgICAgICAgIHsKISAJCQkJaWYgKGlucHV0X2J1ZltpbmRleCArIDFd ID09ICdcbicpIC8qIGFsd2F5cyBzYWZlICovCiEgCQkJCQl7CiEgCQkJCQllb2xfdHlwZSA9 IEVPTF9DUk5MOwohIAkJCQkJZW9sX2NoWzBdID0gJ1xyJzsKISAJCQkJCWVvbF9jaFsxXSA9 ICdcbic7CiEgCQkJCQl9CiEgCQkJCQllbHNlCiEgCQkJCQl7CiEgCQkJCQkJZW9sX3R5cGUg PSBFT0xfQ1I7CiEgCQkJCQkJZW9sX2NoWzBdID0gJ1xyJzsKISAJCQkJCQllb2xfY2hbMV0g PSAnXDAnOwohIAkJCQkJfQohIAkJCQkJCiEgCQkJCQlpbl9xdW90ZSA9IGZhbHNlOwohIAkJ CQkJbGFzdF93YXNfZXNjID0gZmFsc2U7CiEgCQkJCQlyZXR1cm4gdHJ1ZTsKISAJCQl9CiEg CQl9CiEgCiEgCQlpbmRleCsrOwohIAl9CiEgCiEgcmV0dXJuIGZhbHNlOwohIH0KISAKISAv KgohICAqIFJlYWQgYWxsIFRFWFQgYXR0cmlidXRlcy4gQXR0cmlidXRlcyBhcmUgcGFyc2Vk IGZyb20gbGluZV9idWYgYW5kCiEgICogaW5zZXJ0ZWQgKGFsbCBhdCBvbmNlKSB0byBhdHRy X2J1Ziwgd2hpbGUgc2F2aW5nIHBvaW50ZXJzIHRvCiEgICogZWFjaCBhdHRyaWJ1dGUncyBz dGFydGluZyBwb3NpdGlvbi4KISAgKgohICAqIFdoZW4gdGhpcyByb3V0aW5lIGZpbmlzaGVz IGV4ZWN1dGlvbiBib3RoIHRoZSBudWxscyBhcnJheSBhbmQKISAgKiB0aGUgYXR0cl9vZmZz ZXRzIGFycmF5IGFyZSB1cGRhdGVkLiBUaGUgYXR0cl9vZmZzZXRzIHdpbGwgaW5jbHVkZQoh ICAqIHRoZSBvZmZzZXQgZnJvbSB0aGUgYmVnaW5uaW5nIG9mIHRoZSBhdHRyaWJ1dGUgYXJy YXkgb2Ygd2hpY2gKISAgKiBlYWNoIGF0dHJpYnV0ZSBiZWdpbnMuIElmIGEgc3BlY2lmaWMg YXR0cmlidXRlIGlzIG5vdCB1c2VkIGZvciB0aGlzCiEgICogQ09QWSBjb21tYW5kIChvbW1p dHRlZCBmcm9tIHRoZSBjb2x1bW4gbGlzdCksIGEgdmFsdWUgb2YgMCB3aWxsIGJlIGFzc2ln bmVkLgohICAqIEZvciBleGFtcGxlOiBmb3IgdGFibGUgZm9vKGEsYixjLGQsZSkgYW5kIENP UFkgZm9vKGEsYixlKQohICAqIGF0dHJfb2Zmc2V0cyBtYXkgbG9vayBzb21ldGhpbmcgbGlr ZSB0aGlzIGFmdGVyIHRoaXMgcm91dGluZQohICAqIHJldHVybnM6IFswLDIwLDAsMCw1NV0u IFRoYXQgbWVhbnMgdGhhdCBjb2x1bW4gImEiIHZhbHVlIHN0YXJ0cwohICAqIGF0IGJ5dGUg b2Zmc2V0IDAsICJiIiBpbiAyMCBhbmQgImUiIGluIDU1LCBpbiBhdHRyX2J1Zi4KISAgKgoh ICAqIEluIHRoZSBhdHRyaWJ1dGUgYnVmZmVyIChhdHRyX2J1ZikgZWFjaCBhdHRyaWJ1dGUK ISAgKiBpcyB0ZXJtaW5hdGVkIHdpdGggYSAnXDAnLCBhbmQgdGhlcmVmb3JlIGJ5IHVzaW5n IHRoZSBhdHRyX29mZnNldHMKISAgKiBhcnJheSB3ZSBjb3VsZCBwb2ludCB0byBhIGJlZ2lu bmluZyBvZiBhbiBhdHRyaWJ1dGUgYW5kIGhhdmUgaXQKISAgKiBiZWhhdmUgYXMgYSBDIHN0 cmluZywgbXVjaCBsaWtlIHByZXZpb3VzbHkgZG9uZSBpbiBDT1BZLgohICAqCiEgICogQW5v dGhlciBhc3BlY3QgdG8gaW1wcm92aW5nIHBlcmZvcm1hbmNlIGlzIHJlZHVjaW5nIHRoZSBm cmVxdWVuY3kKISAgKiBvZiBkYXRhIGxvYWQgaW50byBidWZmZXJzLiBUaGUgb3JpZ2luYWwg Q09QWSByZWFkIGF0dHJpYnV0ZSBjb2RlCiEgICogbG9hZGVkIGEgY2hhcmFjdGVyIGF0IGEg dGltZS4gSW4gaGVyZSB3ZSB0cnkgdG8gbG9hZCBhIGNodW5rIG9mIGRhdGEKISAgKiBhdCBh IHRpbWUuIFVzdWFsbHkgYSBjaHVuayB3aWxsIGluY2x1ZGUgYSBmdWxsIGRhdGEgcm93CiEg ICogKHVubGVzcyB3ZSBoYXZlIGFuIGVzY2FwZWQgZGVsaW0pLiBUaGF0IGVmZmVjdGl2ZWx5 IHJlZHVjZXMgdGhlIG51bWJlciBvZgohICAqIGxvYWRzIGJ5IGEgZmFjdG9yIG9mIG51bWJl ciBvZiBieXRlcyBwZXIgcm93LiBUaGlzIGltcHJvdmVzIHBlcmZvcm1hbmNlCiEgICogZ3Jl YXRseSwgdW5mb3J0dW5hdGVseSBpdCBhZGQgbW9yZSBjb21wbGV4aXR5IHRvIHRoZSBjb2Rl LgohICAqCiEgICogR2xvYmFsIHBhcnRpY2lwYW50cyBpbiBwYXJzaW5nIGxvZ2ljOgohICAq CiEgICogbGluZV9idWYuY3Vyc29yIC0tIGFuIG9mZnNldCBmcm9tIGJlZ2lubmluZyBvZiB0 aGUgbGluZSBidWZmZXIKISAgKiB0aGF0IGluZGljYXRlcyB3aGVyZSB3ZSBhcmUgYWJvdXQg dG8gYmVnaW4gdGhlIG5leHQgc2Nhbi4gTm90ZSB0aGF0CiEgICogaWYgd2UgaGF2ZSBXSVRI IE9JRFMgdGhpcyBjdXJzb3IgaXMgYWxyZWFkeSBzaGlmdGVkIHBhc3QgdGhlIGZpcnN0CiEg ICogT0lEIGF0dHJpYnV0ZS4KISAgKgohICAqIGF0dHJfYnVmLmN1cnNvciAtLSBhbiBvZmZz ZXQgZnJvbSB0aGUgYmVnaW5uaW5nIG9mIHRoZQohICAqIGF0dHJpYnV0ZSBidWZmZXIgdGhh dCBpbmRpY2F0ZXMgd2hlcmUgdGhlIGN1cnJlbnQgYXR0cmlidXRlIGJlZ2lucy4KISAgKi8K ISBzdGF0aWMgdm9pZCAKISBDb3B5UmVhZEF0dHJpYnV0ZXNUZXh0KGNvbnN0IGNoYXIgKmRl bGltLCBjb25zdCBjaGFyICplc2NhcGUsIGNvbnN0IGNoYXIgKm51bGxfcHJpbnQsCiEgCQkJ CQkgICBpbnQgbnVsbF9wcmludF9sZW4sIGNoYXIgKm51bGxzLCBMaXN0ICphdHRudW1saXN0 LCAKISAJCQkJCSAgIGludCAqYXR0cl9vZmZzZXRzLCBpbnQgbnVtX3BoeXNfYXR0cnMsIEZv cm1fcGdfYXR0cmlidXRlICphdHRyKQohIHsKISAJY2hhcgkJZGVsaW1jID0gZGVsaW1bMF07 CQkvKiBkZWxpbWl0ZXIgY2hhcmFjdGVyICovCiEgCWNoYXIgICAgICAgIGVzY2FwZWMgPSBl c2NhcGVbMF07ICAgIC8qIGVzY2FwZSBjaGFyYWN0ZXIgICAgKi8KISAJY2hhcgkgICAqc2Nh bl9zdGFydDsJCS8qIHBvaW50ZXIgdG8gbGluZSBidWZmZXIgZm9yIHNjYW4gc3RhcnQuICov CiEgCWNoYXIJICAgKnNjYW5fZW5kOwkJLyogcG9pbnRlciB0byBsaW5lIGJ1ZmZlciB3aGVy ZSBjaGFyIHdhcyBmb3VuZCAqLwohIAlpbnQJCQlhdHRyX3ByZV9sZW47CS8qIGF0dHIgcmF3 IGxlbiwgYmVmb3JlIHByb2Nlc3NpbmcgZXNjYXBlcyAqLwohIAlpbnQJCQlhdHRyX3Bvc3Rf bGVuOwkvKiBjdXJyZW50IGF0dHIgbGVuIGFmdGVyIGVzY2FwaW5nICovCiEgCWludAkJCW07 CQkJCS8qIGF0dHJpYnV0ZSBpbmRleCBiZWluZyBwYXJzZWQgKi8KISAJaW50CQkJYnl0ZXNf cmVtYWluaW5nOy8qIG51bSBieXRlcyByZW1haW5pbmcgdG8gYmUgc2Nhbm5lZCBpbiBsaW5l IGJ1ZiAqLwohIAlpbnQJCQljaHVua19zdGFydDsJLyogb2Zmc2V0IHRvIGJlZ2lubmluZyBv ZiBsaW5lIGNodW5rIHRvIGxvYWQgKi8KISAJaW50CQkJY2h1bmtfbGVuOwkJLyogbGVuZ3Ro IG9mIGNodW5rIG9mIGRhdGEgdG8gbG9hZCB0byBhdHRyIGJ1ZiAqLwohIAlpbnQJCQlvY3Rf dmFsOwkJLyogYnl0ZSB2YWx1ZSBmb3Igb2N0YWwgZXNjYXBlcyAqLwohIAlpbnQgICAgICAg ICBoZXhfdmFsOyAgICAgICAgLyogYnl0ZSB2YWx1ZSBmb3IgaGV4YWRlY2ltYWwgZXNjYXBl cyAqLwohIAljaGFyICAgICAgICBoZXhjaGFyOyAgICAgICAgLyogY2hhciB0aGF0IGFwcGVh cnMgYWZ0ZXIgXHggZm9yIGhleCB2YWx1ZXMgKi8KISAJaW50CQkJYXR0bnVtOwkJCS8qIGF0 dHJpYnV0ZSBudW1iZXIgYmVpbmcgcGFyc2VkICovCiEgCUxpc3RDZWxsICAgKmN1cjsJCQkv KiBjdXJzb3IgdG8gYXR0cmlidXRlIGxpc3QgdXNlZCBmb3IgdGhpcyBDT1BZICovCiEgCWlu dAkJCWF0dHJpYnV0ZTsKISAJCQohIAkJLyoKISAJCSAqIGluaXQgdmFyaWFibGVzIGZvciBh dHRyaWJ1dGUgc2NhbgohIAkJICovCiEgCQlhdHRyX2J1Zi5sZW4gPSAwOwohIAkJYXR0cl9i dWYuZGF0YVswXSA9ICdcMCc7CiEgCQlhdHRyX2J1Zi5jdXJzb3IgPSAwOwohIAkJLyogY3Vy c29yIGlzIG5vdyA+IDAgaWYgd2UgY29weSBXSVRIIE9JRFMgKi8KISAJCXNjYW5fc3RhcnQg PSBsaW5lX2J1Zi5kYXRhICsgbGluZV9idWYuY3Vyc29yOwohIAkJY3VyID0gbGlzdF9oZWFk KGF0dG51bWxpc3QpOwohIAkJYXR0bnVtID0gbGZpcnN0X2ludChjdXIpOwohIAkJbSA9IGF0 dG51bSAtIDE7CiEgCQljaHVua19zdGFydCA9IGxpbmVfYnVmLmN1cnNvcjsKISAJCWNodW5r X2xlbiA9IDA7CiEgCQlhdHRyX3ByZV9sZW4gPSAwOwohIAkJYXR0cl9wb3N0X2xlbiA9IDA7 CiEgCQkKISAJCS8qCiEgCQkgKiBTY2FuIHRocm91Z2ggdGhlIGxpbmUgYnVmZmVyIHRvIHJl YWQgYWxsIGF0dHJpYnV0ZXMgZGF0YQohIAkJICovCiEgCQl3aGlsZSAobGluZV9idWYuY3Vy c29yIDwgbGluZV9idWYubGVuKQohIAkJewohIAkJCWJ5dGVzX3JlbWFpbmluZyA9IGxpbmVf YnVmLmxlbiAtIGxpbmVfYnVmLmN1cnNvcjsKISAJCQkKISAJCQlpZiAoKHNjYW5fZW5kID0g c2NhblRleHRBdHRyKHNjYW5fc3RhcnQsIGRlbGltYywgZXNjYXBlYywgYnl0ZXNfcmVtYWlu aW5nKSkKISAJCQkJPT0gTlVMTCkKISAJCQl7CiEgCQkJCS8qIEdPVCBUTyBFTkQgT0YgTElO RSBCVUZGRVIgKi8KISAJCQkJCiEgCQkJCWlmIChjdXIgPT0gTlVMTCkKISAJCQkJCWVyZXBv cnQoRVJST1IsCiEgCQkJCQkJCShlcnJjb2RlKEVSUkNPREVfQkFEX0NPUFlfRklMRV9GT1JN QVQpLAohIAkJCQkJCQkgZXJybXNnKCJleHRyYSBkYXRhIGFmdGVyIGxhc3QgZXhwZWN0ZWQg Y29sdW1uIikpKTsKISAJCQkJCiEgCQkJCWF0dG51bSA9IGxmaXJzdF9pbnQoY3VyKTsKISAJ CQkJbSA9IGF0dG51bSAtIDE7CiEgCQkJCQohIAkJCQkvKiBkb24ndCBjb3VudCBlb2wgY2hh cihzKSBpbiBhdHRyIGFuZCBjaHVuayBsZW4gY2FsY3VsYXRpb24gKi8KISAJCQkJaWYgKGVv bF90eXBlID09IEVPTF9DUk5MKQohIAkJCQl7CiEgCQkJCQlhdHRyX3ByZV9sZW4gKz0gYnl0 ZXNfcmVtYWluaW5nIC0gMjsKISAJCQkJCWNodW5rX2xlbiA9IGxpbmVfYnVmLmxlbiAtIGNo dW5rX3N0YXJ0IC0gMjsKISAJCQkJfQohIAkJCQllbHNlCiEgCQkJCXsKISAJCQkJCWF0dHJf cHJlX2xlbiArPSBieXRlc19yZW1haW5pbmcgLSAxOwohIAkJCQkJY2h1bmtfbGVuID0gbGlu ZV9idWYubGVuIC0gY2h1bmtfc3RhcnQgLSAxOwohIAkJCQl9CiEgCQkJCQohIAkJCQkvKiBj aGVjayBpZiB0aGlzIGlzIGEgTlVMTCB2YWx1ZSBvciBkYXRhIHZhbHVlIChhc3N1bWVkIE5V TEwpICovCiEgCQkJCWlmIChhdHRyX3ByZV9sZW4gPT0gbnVsbF9wcmludF9sZW4KISAJCQkJ CSYmCiEgCQkJCQlzdHJuY21wKGxpbmVfYnVmLmRhdGEgKyBsaW5lX2J1Zi5sZW4gLSBhdHRy X3ByZV9sZW4gLSAxLCBudWxsX3ByaW50LCBhdHRyX3ByZV9sZW4pCiEgCQkJCQk9PSAwKQoh IAkJCQkJbnVsbHNbbV0gPSAnbic7CiEgCQkJCWVsc2UKISAJCQkJCW51bGxzW21dID0gJyAn OwohIAkJCQkKISAJCQkJYXR0cl9vZmZzZXRzW21dID0gYXR0cl9idWYuY3Vyc29yOwohIAkJ CQkKISAJCQkJCiEgCQkJCS8qIGxvYWQgdGhlIGxhc3QgY2h1bmssIHRoZSB3aG9sZSBidWZm ZXIgaW4gbW9zdCBjYXNlcyAqLwohIAkJCQlhcHBlbmRCaW5hcnlTdHJpbmdJbmZvKCZhdHRy X2J1ZiwgbGluZV9idWYuZGF0YSArIGNodW5rX3N0YXJ0LCBjaHVua19sZW4pOwohIAkJCQkK ISAJCQkJbGluZV9idWYuY3Vyc29yICs9IGF0dHJfcHJlX2xlbiArIDI7CQkvKiBza2lwIGVv bCBjaGFyIGFuZAohIAkJCQkJKiAnXDAnIHRvIGV4aXQgbG9vcCAqLwohIAkJCQkKISAJCQkJ aWYgKGxuZXh0KGN1cikgIT0gTlVMTCkKISAJCQkJewohIAkJCQkJLyoKISAJCQkJCSAqIEZv ciBhbiBlbXB0eSBkYXRhIGxpbmUsIHRoZSBwcmV2aW91cyBDT1BZIGNvZGUgd2lsbAohIAkJ CQkJICogZmFpbCBpdCBkdXJpbmcgdGhlIGNvbnZlcnNpb24gc3RhZ2UuIFdlIGNhbiBmYWls IGl0IGhlcmUKISAJCQkJCSAqIGFscmVhZHksIGJ1dCB0aGVuIHdlIHdpbGwgZmFpbCB0aGUg cmVncmVzc2lvbiB0ZXN0cyBiL2MKISAJCQkJCSAqIG9mIGEgZGlmZmVyZW50IGVycm9yIG1l c3NhZ2UuIHRoYXQncyB3aHkgd2UgcmV0dXJuIHNvIHdlCiEgCQkJCQkgKiBjYW4gZ2V0IHRo ZSBzYW1lIGVycm9yIG1lc3NhZ2UgdGhhdCByZWdyZXNzIGV4cGVjdHMuIGFoaC4uLgohIAkJ CQkJICogdGhpcyBjb25kaXRpb25hbCBpcyB1bm5lY2Vzc2FyeSBhbmQgc2hvdWxkIGJlIHJl bW92ZWQgc29vbi4KISAJCQkJCSAqLwohIAkJCQkJaWYgKGxpbmVfYnVmLmxlbiA+IDEpCiEg CQkJCQkJZXJlcG9ydChFUlJPUiwKISAJCQkJCQkJCShlcnJjb2RlKEVSUkNPREVfQkFEX0NP UFlfRklMRV9GT1JNQVQpLAohIAkJCQkJCQkJIGVycm1zZygibWlzc2luZyBkYXRhIGZvciBj b2x1bW4gXCIlc1wiIiwKISAJCQkJCQkJCQkJTmFtZVN0cihhdHRyW20gKyAxXS0+YXR0bmFt ZSkpKSk7CiEgCQkJCQllbHNlCiEgCQkJCQkJcmV0dXJuOwohIAkJCQl9CiEgCQkJfQohIAkJ CWVsc2UKISAJCQkJLyogRk9VTkQgQSBERUxJTUlURVIgT1IgRVNDQVBFICovCiEgCQkJewoh IAkJCQlpZiAoY3VyID09IE5VTEwpCiEgCQkJCQllcmVwb3J0KEVSUk9SLAohIAkJCQkJCQko ZXJyY29kZShFUlJDT0RFX0JBRF9DT1BZX0ZJTEVfRk9STUFUKSwKISAJCQkJCQkJIGVycm1z ZygiZXh0cmEgZGF0YSBhZnRlciBsYXN0IGV4cGVjdGVkIGNvbHVtbiIpKSk7CiEgCQkJCQoh IAkJCQlpZiAoKnNjYW5fZW5kID09IGRlbGltYykJCS8qIGZvdW5kIGEgZGVsaW1pdGVyICov CiEgCQkJCXsKISAJCQkJCWF0dG51bSA9IGxmaXJzdF9pbnQoY3VyKTsKISAJCQkJCW0gPSBh dHRudW0gLSAxOwohIAkJCQkJCiEgCQkJCQkvKiAod2UgZG9uJ3QgaW5jbHVkZSB0aGUgZGVs aW1pdGVyIGNoIGluIGxlbmd0aCkgKi8KISAJCQkJCWF0dHJfcHJlX2xlbiArPSBzY2FuX2Vu ZCAtIHNjYW5fc3RhcnQ7CiEgCQkJCQkvKiAod2UgZG9uJ3QgaW5jbHVkZSB0aGUgZGVsaW1p dGVyIGNoIGluIGxlbmd0aCkgKi8KISAJCQkJCWF0dHJfcG9zdF9sZW4gKz0gc2Nhbl9lbmQg LSBzY2FuX3N0YXJ0OwohIAkJCQkJCiEgCQkJCQkvKiBjaGVjayBpZiB0aGlzIGlzIGEgbnVs bCBwcmludCBvciBkYXRhIChhc3N1bWVkIE5VTEwpICovCiEgCQkJCQlpZiAoYXR0cl9wcmVf bGVuID09IG51bGxfcHJpbnRfbGVuCiEgCQkJCQkJJiYKISAJCQkJCQlzdHJuY21wKHNjYW5f ZW5kIC0gYXR0cl9wcmVfbGVuLCBudWxsX3ByaW50LCBhdHRyX3ByZV9sZW4pCiEgCQkJCQkJ PT0gMCkKISAJCQkJCQludWxsc1ttXSA9ICduJzsKISAJCQkJCWVsc2UKISAJCQkJCQludWxs c1ttXSA9ICcgJzsKISAJCQkJCQohIAkJCQkJLyogc2V0IHRoZSBwb2ludGVyIHRvIG5leHQg YXR0cmlidXRlIHBvc2l0aW9uICovCiEgCQkJCQlhdHRyX29mZnNldHNbbV0gPSBhdHRyX2J1 Zi5jdXJzb3I7CiEgCQkJCQkKISAJCQkJCS8qCiEgCQkJCQkgKiB1cGRhdGUgYnVmZmVyIGN1 cnNvcnMgdG8gb3VyIGN1cnJlbnQgbG9jYXRpb24sICsxIHRvIHNraXAKISAJCQkJCSAqIHRo ZSBkZWxpbWMKISAJCQkJCSAqLwohIAkJCQkJbGluZV9idWYuY3Vyc29yID0gc2Nhbl9lbmQg LSBsaW5lX2J1Zi5kYXRhICsgMTsKISAJCQkJCWF0dHJfYnVmLmN1cnNvciArPSBhdHRyX3Bv c3RfbGVuICsgMTsKISAJCQkJCQohIAkJCQkJLyogcHJlcGFyZSBzY2FuIGZvciBuZXh0IGF0 dHIgKi8KISAJCQkJCXNjYW5fc3RhcnQgPSBsaW5lX2J1Zi5kYXRhICsgbGluZV9idWYuY3Vy c29yOwohIAkJCQkJY3VyID0gbG5leHQoY3VyKTsKISAJCQkJCWF0dHJfcHJlX2xlbiA9IDA7 CiEgCQkJCQlhdHRyX3Bvc3RfbGVuID0gMDsKISAJCQkJfQohIAkJCQllbHNlCiEgCQkJCQkv KiBmb3VuZCBhbiBlc2NhcGUgY2hhcmFjdGVyICovCiEgCQkJCXsKISAJCQkJCWNoYXIJCW5l eHRjID0gKihzY2FuX2VuZCArIDEpOwohIAkJCQkJY2hhcgkJbmV3YzsKISAJCQkJCWludAkJ CXNraXAgPSAyOwohIAkJCQkJCiEgCQkJCQljaHVua19sZW4gPSAoc2Nhbl9lbmQgLSBsaW5l X2J1Zi5kYXRhKSAtIGNodW5rX3N0YXJ0ICsgMTsKISAJCQkJCQohIAkJCQkJLyogbG9hZCBh IGNodW5rIG9mIGRhdGEgKi8KISAJCQkJCWFwcGVuZEJpbmFyeVN0cmluZ0luZm8oJmF0dHJf YnVmLCBsaW5lX2J1Zi5kYXRhICsgY2h1bmtfc3RhcnQsIGNodW5rX2xlbik7CiEgCQkJCQkK ISAJCQkJCXN3aXRjaCAobmV4dGMpCiAgCQkJewogIAkJCQljYXNlICcwJzoKICAJCQkJY2Fz ZSAnMSc6CioqKioqKioqKioqKioqKgoqKiogMjM0OCwyNDQ1ICoqKioKICAJCQkJY2FzZSAn NSc6CiAgCQkJCWNhc2UgJzYnOgogIAkJCQljYXNlICc3JzoKISAJCQkJCS8qIGhhbmRsZSBc MDEzICovCiEgCQkJCQl7CiEgCQkJCQkJaW50CQkJdmFsOwogIAohIAkJCQkJCXZhbCA9IE9D VFZBTFVFKGMpOwohIAkJCQkJCWlmIChsaW5lX2J1Zi5jdXJzb3IgPCBsaW5lX2J1Zi5sZW4p CiEgCQkJCQkJewohIAkJCQkJCQljID0gbGluZV9idWYuZGF0YVtsaW5lX2J1Zi5jdXJzb3Jd OwohIAkJCQkJCQlpZiAoSVNPQ1RBTChjKSkKISAJCQkJCQkJewohIAkJCQkJCQkJbGluZV9i dWYuY3Vyc29yKys7CiEgCQkJCQkJCQl2YWwgPSAodmFsIDw8IDMpICsgT0NUVkFMVUUoYyk7 CiEgCQkJCQkJCQlpZiAobGluZV9idWYuY3Vyc29yIDwgbGluZV9idWYubGVuKQogIAkJCQkJ CQkJewohIAkJCQkJCQkJCWMgPSBsaW5lX2J1Zi5kYXRhW2xpbmVfYnVmLmN1cnNvcl07CiEg CQkJCQkJCQkJaWYgKElTT0NUQUwoYykpCiAgCQkJCQkJCQkJewohIAkJCQkJCQkJCQlsaW5l X2J1Zi5jdXJzb3IrKzsKISAJCQkJCQkJCQkJdmFsID0gKHZhbCA8PCAzKSArIE9DVFZBTFVF KGMpOwogIAkJCQkJCQkJCX0KICAJCQkJCQkJCX0KISAJCQkJCQkJfQohIAkJCQkJCX0KISAJ CQkJCQljID0gdmFsICYgMDM3NzsKISAJCQkJCX0KICAJCQkJCWJyZWFrOwogIAkJCQljYXNl ICd4JzoKICAJCQkJCS8qIEhhbmRsZSBceDNGICovCiEgCQkJCQlpZiAobGluZV9idWYuY3Vy c29yIDwgbGluZV9idWYubGVuKQohIAkJCQkJewohIAkJCQkJCWNoYXIgaGV4Y2hhciA9IGxp bmVfYnVmLmRhdGFbbGluZV9idWYuY3Vyc29yXTsKICAKICAJCQkJCQlpZiAoaXN4ZGlnaXQo aGV4Y2hhcikpCiAgCQkJCQkJewohIAkJCQkJCQlpbnQgdmFsID0gR2V0RGVjaW1hbEZyb21I ZXgoaGV4Y2hhcik7CiAgCiEgCQkJCQkJCWxpbmVfYnVmLmN1cnNvcisrOwohIAkJCQkJCQlp ZiAobGluZV9idWYuY3Vyc29yIDwgbGluZV9idWYubGVuKQohIAkJCQkJCQl7CiEgCQkJCQkJ CQloZXhjaGFyID0gbGluZV9idWYuZGF0YVtsaW5lX2J1Zi5jdXJzb3JdOwogIAkJCQkJCQkJ aWYgKGlzeGRpZ2l0KGhleGNoYXIpKQogIAkJCQkJCQkJewohIAkJCQkJCQkJCWxpbmVfYnVm LmN1cnNvcisrOwohIAkJCQkJCQkJCXZhbCA9ICh2YWwgPDwgNCkgKyBHZXREZWNpbWFsRnJv bUhleChoZXhjaGFyKTsKISAJCQkJCQkJCX0KISAJCQkJCQkJfQohIAkJCQkJCQljID0gdmFs ICYgMHhmZjsKICAJCQkJCQl9CiAgCQkJCQl9CiAgCQkJCQlicmVhazsKICAJCQkJY2FzZSAn Yic6CiEgCQkJCQljID0gJ1xiJzsKICAJCQkJCWJyZWFrOwogIAkJCQljYXNlICdmJzoKISAJ CQkJCWMgPSAnXGYnOwogIAkJCQkJYnJlYWs7CiAgCQkJCWNhc2UgJ24nOgohIAkJCQkJYyA9 ICdcbic7CiAgCQkJCQlicmVhazsKICAJCQkJY2FzZSAncic6CiEgCQkJCQljID0gJ1xyJzsK ICAJCQkJCWJyZWFrOwogIAkJCQljYXNlICd0JzoKISAJCQkJCWMgPSAnXHQnOwogIAkJCQkJ YnJlYWs7CiAgCQkJCWNhc2UgJ3YnOgohIAkJCQkJYyA9ICdcdic7CiAgCQkJCQlicmVhazsK ICAKICAJCQkJCS8qCiEgCQkJCQkgKiBpbiBhbGwgb3RoZXIgY2FzZXMsIHRha2UgdGhlIGNo YXIgYWZ0ZXIgJ1wnCiEgCQkJCQkgKiBsaXRlcmFsbHkKICAJCQkJCSAqLwogIAkJCX0KISAJ CX0KISAJCWFwcGVuZFN0cmluZ0luZm9DaGFyTWFjcm8oJmF0dHJpYnV0ZV9idWYsIGMpOwog IAl9CiAgCiEgCS8qIGNoZWNrIHdoZXRoZXIgcmF3IGlucHV0IG1hdGNoZWQgbnVsbCBtYXJr ZXIgKi8KISAJaW5wdXRfbGVuID0gZW5kX2N1cnNvciAtIHN0YXJ0X2N1cnNvcjsKISAJaWYg KGlucHV0X2xlbiA9PSBzdHJsZW4obnVsbF9wcmludCkgJiYKISAJCXN0cm5jbXAoJmxpbmVf YnVmLmRhdGFbc3RhcnRfY3Vyc29yXSwgbnVsbF9wcmludCwgaW5wdXRfbGVuKSA9PSAwKQog IAkJKmlzbnVsbCA9IHRydWU7CiEgCWVsc2UKISAJCSppc251bGwgPSBmYWxzZTsKICAKISAJ cmV0dXJuIGF0dHJpYnV0ZV9idWYuZGF0YTsKISB9CiAgCiAgCiAgLyoKISAgKiBSZWFkIHRo ZSB2YWx1ZSBvZiBhIHNpbmdsZSBhdHRyaWJ1dGUgaW4gQ1NWIG1vZGUsCiAgICogcGVyZm9y bWluZyBkZS1lc2NhcGluZyBhcyBuZWVkZWQuIEVzY2FwaW5nIGRvZXMgbm90IGZvbGxvdyB0 aGUgbm9ybWFsCiAgICogUG9zdGdyZVNRTCB0ZXh0IG1vZGUsIGJ1dCBpbnN0ZWFkICJzdGFu ZGFyZCIgKGkuZS4gY29tbW9uKSBDU1YgdXNhZ2UuCiAgICoKLS0tIDI3NzUsMjk0OCAtLS0t CiAgCQkJCWNhc2UgJzUnOgogIAkJCQljYXNlICc2JzoKICAJCQkJY2FzZSAnNyc6CiEgCQkJ CQkJCW9jdF92YWwgPSBPQ1RWQUxVRShuZXh0Yyk7CiEgCQkJCQkJCW5leHRjID0gKihzY2Fu X2VuZCArIDIpOwogIAohIAkJCQkJCQkvKgohIAkJCQkJCQkgKiAobm8gbmVlZCBmb3Igb3V0 IGJhZCBhY2Nlc3MgY2hlY2sgc2luY2UgbGluZSBpZgohIAkJCQkJCQkJKiBidWZmZXJlZCkK ISAJCQkJCQkJICovCiEgCQkJCQkJCWlmIChJU09DVEFMKG5leHRjKSkKICAJCQkJCQkJCXsK ISAJCQkJCQkJCXNraXArKzsKISAJCQkJCQkJCW9jdF92YWwgPSAob2N0X3ZhbCA8PCAzKSAr IE9DVFZBTFVFKG5leHRjKTsKISAJCQkJCQkJCW5leHRjID0gKihzY2FuX2VuZCArIDMpOwoh IAkJCQkJCQkJaWYgKElTT0NUQUwobmV4dGMpKQogIAkJCQkJCQkJCXsKISAJCQkJCQkJCQlz a2lwKys7CiEgCQkJCQkJCQkJb2N0X3ZhbCA9IChvY3RfdmFsIDw8IDMpICsgT0NUVkFMVUUo bmV4dGMpOwogIAkJCQkJCQkJCX0KICAJCQkJCQkJCX0KISAJCQkJCQkJCW5ld2MgPSBvY3Rf dmFsICYgMDM3NzsJLyogdGhlIGVzY2FwZWQgYnl0ZSB2YWx1ZSAqLwogIAkJCQkJYnJlYWs7 CiAgCQkJCWNhc2UgJ3gnOgogIAkJCQkJLyogSGFuZGxlIFx4M0YgKi8KISAJCQkJCQkJaGV4 Y2hhciA9ICooc2Nhbl9lbmQgKyAyKTsKICAKICAJCQkJCQlpZiAoaXN4ZGlnaXQoaGV4Y2hh cikpCiAgCQkJCQkJewohIAkJCQkJCQkJc2tpcCsrOwohIAkJCQkJCQkJaGV4X3ZhbCA9IEdl dERlY2ltYWxGcm9tSGV4KGhleGNoYXIpOwogIAohIAkJCQkJCQkJaGV4Y2hhciA9ICooc2Nh bl9lbmQgKyAzKTsKICAJCQkJCQkJCWlmIChpc3hkaWdpdChoZXhjaGFyKSkKICAJCQkJCQkJ CXsKISAJCQkJCQkJCQlza2lwKys7CiEgCQkJCQkJCQkJaGV4X3ZhbCA9IChoZXhfdmFsIDw8 IDQpICsgR2V0RGVjaW1hbEZyb21IZXgoaGV4Y2hhcik7CiAgCQkJCQkJfQorIAkJCQkJCQkJ bmV3YyA9IGhleF92YWwgJiAweGZmOwogIAkJCQkJfQorIAkJCQkJCQkJZWxzZSAvKiAiXHgi IHdpdGggbm8gaGV4IHZhbHVlICovCisgCQkJCQkJCQkJbmV3YyA9IG5leHRjOwogIAkJCQkJ YnJlYWs7CiAgCQkJCWNhc2UgJ2InOgohIAkJCQkJCQluZXdjID0gJ1xiJzsKICAJCQkJCWJy ZWFrOwogIAkJCQljYXNlICdmJzoKISAJCQkJCQkJbmV3YyA9ICdcZic7CiAgCQkJCQlicmVh azsKICAJCQkJY2FzZSAnbic6CiEgCQkJCQkJCW5ld2MgPSAnXG4nOwogIAkJCQkJYnJlYWs7 CiAgCQkJCWNhc2UgJ3InOgohIAkJCQkJCQluZXdjID0gJ1xyJzsKICAJCQkJCWJyZWFrOwog IAkJCQljYXNlICd0JzoKISAJCQkJCQkJbmV3YyA9ICdcdCc7CiAgCQkJCQlicmVhazsKICAJ CQkJY2FzZSAndic6CiEgCQkJCQkJCW5ld2MgPSAnXHYnOwohIAkJCQkJCQlicmVhazsKISAJ CQkJCQlkZWZhdWx0OgohIAkJCQkJCQlpZiAobmV4dGMgPT0gZGVsaW1jKQohIAkJCQkJCQkJ bmV3YyA9IGRlbGltYzsKISAJCQkJCQkJZWxzZSBpZiAobmV4dGMgPT0gZXNjYXBlYykKISAJ CQkJCQkJCW5ld2MgPSBlc2NhcGVjOwohIAkJCQkJCQllbHNlCiEgCQkJCQkJCQkvKiBubyBl c2NhcGUgc2VxdWVuY2UsIHRha2UgbmV4dCBjaGFyIGxpdGVyYWx5ICovCiEgCQkJCQkJCQlu ZXdjID0gbmV4dGM7CiAgCQkJCQlicmVhazsKKyAJCQkJCX0KKyAJCQkJCQorIAkJCQkJLyog dXBkYXRlIHRvIGN1cnJlbnQgbGVuZ3RoLCBhZGQgZXNjYXBlIGFuZCBlc2NhcGVkIGNoYXJz ICAqLworIAkJCQkJYXR0cl9wcmVfbGVuICs9IHNjYW5fZW5kIC0gc2Nhbl9zdGFydCArIDI7 CisgCQkJCQkvKiB1cGRhdGUgdG8gY3VycmVudCBsZW5ndGgsIGVzY2FwZWQgY2hhciAqLwor IAkJCQkJYXR0cl9wb3N0X2xlbiArPSBzY2FuX2VuZCAtIHNjYW5fc3RhcnQgKyAxOwogIAog IAkJCQkJLyoKISAJCQkJCSAqIE5lZWQgdG8gZ2V0IHJpZCBvZiB0aGUgZXNjYXBlIGNoYXJh Y3Rlci4gVGhpcyBpcyBkb25lIGJ5CiEgCQkJCQkgKiBsb2FkaW5nIHRoZSBjaHVuayB1cCB0 byBpbmNsdWRpbmcgdGhlIGVzY2FwZSBjaGFyYWN0ZXIKISAJCQkJCSAqIGludG8gdGhlIGF0 dHJpYnV0ZSBidWZmZXIuIFRoZW4gb3ZlcndyaXR0aW5nIHRoZSBiYWNrc2xhc2gKISAJCQkJ CSAqIHdpdGggdGhlIGVzY2FwZWQgc2VxdWVuY2Ugb3IgY2hhciwgYW5kIGNvbnRpbnVpbmcg dG8gc2NhbgohIAkJCQkJICogZnJvbSAqYWZ0ZXIqIHRoZSBjaGFyIHRoYW4gaXMgYWZ0ZXIg dGhlIGVzY2FwZSBpbiBsaW5lIGJ1Zi4KICAJCQkJCSAqLworIAkJCQkJKihhdHRyX2J1Zi5k YXRhICsgYXR0cl9idWYubGVuIC0gMSkgPSBuZXdjOworIAkJCQkJbGluZV9idWYuY3Vyc29y ID0gc2Nhbl9lbmQgLSBsaW5lX2J1Zi5kYXRhICsgc2tpcDsKKyAJCQkJCXNjYW5fc3RhcnQg PSBzY2FuX2VuZCArIHNraXA7CisgCQkJCQljaHVua19zdGFydCA9IGxpbmVfYnVmLmN1cnNv cjsKKyAJCQkJCWNodW5rX2xlbiA9IDA7CiAgCQkJfQohIAkJCQkKISAJCQl9CQkJCQkJLyog ZW5kIGRlbGltaXRlci9iYWNrc2xhc2ggKi8KISAKISAJCX0JCQkJCQkJLyogZW5kIGxpbmUg YnVmZmVyIHNjYW4uICovCiEgCiEgCS8qCiEgCSAqIFJlcGxhY2UgYWxsIGRlbGltaXRlcnMg d2l0aCBOVUxMIGZvciBzdHJpbmcgdGVybWluYXRpb24uCiEgCSAqIE5PVEU6IG9ubHkgZGVs aW1pdGVycyAoTk9UIG5lY2Vzc2FyaWx5IGFsbCBkZWxpbWMpIGFyZSByZXBsYWNlZC4KISAJ ICogRXhhbXBsZSAoZGVsaW1jID0gJ3wnKToKISAJICogLSBCZWZvcmU6ICBmICAxCXwgIGYg IFx8ICAyICB8CWYgIDMKISAJICogLSBBZnRlciA6ICBmICAxIFwwICBmICAgfCAgMiBcMAlm ICAzCiEgCSAqLwohIAlmb3IgKGF0dHJpYnV0ZSA9IDE7IGF0dHJpYnV0ZSA8IG51bV9waHlz X2F0dHJzOyBhdHRyaWJ1dGUrKykKISAJewohIAkJaWYgKGF0dHJfb2Zmc2V0c1thdHRyaWJ1 dGVdICE9IDApCiEgCQkJKihhdHRyX2J1Zi5kYXRhICsgYXR0cl9vZmZzZXRzW2F0dHJpYnV0 ZV0gLSAxKSA9ICdcMCc7CiAgCX0KICAKISB9CiEgCiEgCiEgLyoKISAgKiBSZWFkIGEgYmlu YXJ5IGF0dHJpYnV0ZQohICAqLwohIHN0YXRpYyBEYXR1bQohIENvcHlSZWFkQmluYXJ5QXR0 cmlidXRlKGludCBjb2x1bW5fbm8sIEZtZ3JJbmZvICpmbGluZm8sIAohIAkJCQkJCU9pZCB0 eXBpb3BhcmFtLCBpbnQzMiB0eXBtb2QsCiEgCQkJCQkJYm9vbCAqaXNudWxsKQohIHsKISAJ aW50MzIJCWZsZF9zaXplOwohIAlEYXR1bQkJcmVzdWx0OwohIAkKISAJZmxkX3NpemUgPSBD b3B5R2V0SW50MzIoKTsKISAJaWYgKENvcHlHZXRFb2YoKSkKISAJCWVyZXBvcnQoRVJST1Is CiEgCQkJCShlcnJjb2RlKEVSUkNPREVfQkFEX0NPUFlfRklMRV9GT1JNQVQpLAohIAkJCQkg ZXJybXNnKCJ1bmV4cGVjdGVkIEVPRiBpbiBDT1BZIGRhdGEiKSkpOwohIAlpZiAoZmxkX3Np emUgPT0gLTEpCiEgCXsKICAJCSppc251bGwgPSB0cnVlOwohIAkJcmV0dXJuIChEYXR1bSkg MDsKISAJfQohIAlpZiAoZmxkX3NpemUgPCAwKQohIAkJZXJlcG9ydChFUlJPUiwKISAJCQkJ KGVycmNvZGUoRVJSQ09ERV9CQURfQ09QWV9GSUxFX0ZPUk1BVCksCiEgCQkJCSBlcnJtc2co ImludmFsaWQgZmllbGQgc2l6ZSIpKSk7CiAgCiEgCS8qIHJlc2V0IGF0dHJfYnVmIHRvIGVt cHR5LCBhbmQgbG9hZCByYXcgZGF0YSBpbiBpdCAqLwohIAlhdHRyX2J1Zi5sZW4gPSAwOwoh IAlhdHRyX2J1Zi5kYXRhWzBdID0gJ1wwJzsKISAJYXR0cl9idWYuY3Vyc29yID0gMDsKISAJ CiEgCWVubGFyZ2VTdHJpbmdJbmZvKCZhdHRyX2J1ZiwgZmxkX3NpemUpOwohIAkKISAJQ29w eUdldERhdGEoYXR0cl9idWYuZGF0YSwgZmxkX3NpemUpOwohIAlpZiAoQ29weUdldEVvZigp KQohIAkJZXJlcG9ydChFUlJPUiwKISAJCQkJKGVycmNvZGUoRVJSQ09ERV9CQURfQ09QWV9G SUxFX0ZPUk1BVCksCiEgCQkJCSBlcnJtc2coInVuZXhwZWN0ZWQgRU9GIGluIENPUFkgZGF0 YSIpKSk7CiAgCisgCWF0dHJfYnVmLmxlbiA9IGZsZF9zaXplOworIAlhdHRyX2J1Zi5kYXRh W2ZsZF9zaXplXSA9ICdcMCc7CisgCQorIAkvKiBDYWxsIHRoZSBjb2x1bW4gdHlwZSdzIGJp bmFyeSBpbnB1dCBjb252ZXJ0ZXIgKi8KKyAJcmVzdWx0ID0gRnVuY3Rpb25DYWxsMyhmbGlu Zm8sCisgCQkJCQkJICAgUG9pbnRlckdldERhdHVtKCZhdHRyX2J1ZiksCisgCQkJCQkJICAg T2JqZWN0SWRHZXREYXR1bSh0eXBpb3BhcmFtKSwKKyAJCQkJCQkgICBJbnQzMkdldERhdHVt KHR5cG1vZCkpOworIAkKKyAJLyogVHJvdWJsZSBpZiBpdCBkaWRuJ3QgZWF0IHRoZSB3aG9s ZSBidWZmZXIgKi8KKyAJaWYgKGF0dHJfYnVmLmN1cnNvciAhPSBhdHRyX2J1Zi5sZW4pCisg CQllcmVwb3J0KEVSUk9SLAorIAkJCQkoZXJyY29kZShFUlJDT0RFX0lOVkFMSURfQklOQVJZ X1JFUFJFU0VOVEFUSU9OKSwKKyAJCQkJIGVycm1zZygiaW5jb3JyZWN0IGJpbmFyeSBkYXRh IGZvcm1hdCIpKSk7CisgCQorIAkqaXNudWxsID0gZmFsc2U7CisgCXJldHVybiByZXN1bHQ7 CisgfQogIAogIC8qCiEgICogUmVhZCBhbGwgdGhlIGF0dHJpYnV0ZXMgb2YgdGhlIGRhdGEg bGluZSBpbiBDU1YgbW9kZSwKICAgKiBwZXJmb3JtaW5nIGRlLWVzY2FwaW5nIGFzIG5lZWRl ZC4gRXNjYXBpbmcgZG9lcyBub3QgZm9sbG93IHRoZSBub3JtYWwKICAgKiBQb3N0Z3JlU1FM IHRleHQgbW9kZSwgYnV0IGluc3RlYWQgInN0YW5kYXJkIiAoaS5lLiBjb21tb24pIENTViB1 c2FnZS4KICAgKgoqKioqKioqKioqKioqKioKKioqIDI0NDgsMjQ3MiAqKioqCiAgICoKICAg KiBudWxsX3ByaW50IGlzIHRoZSBudWxsIG1hcmtlciBzdHJpbmcuICBOb3RlIHRoYXQgdGhp cyBpcyBjb21wYXJlZCB0bwogICAqIHRoZSBwcmUtZGUtZXNjYXBlZCBpbnB1dCBzdHJpbmcg KHRodXMgaWYgaXQgaXMgcXVvdGVkIGl0IGlzIG5vdCBhIE5VTEwpLgotICAqCi0gICogKnJl c3VsdCBpcyBzZXQgdG8gaW5kaWNhdGUgd2hhdCB0ZXJtaW5hdGVkIHRoZSByZWFkOgotICAq CQlOT1JNQUxfQVRUUjoJY29sdW1uIGRlbGltaXRlcgotICAqCQlFTkRfT0ZfTElORToJZW5k IG9mIGxpbmUKLSAgKgkJVU5URVJNSU5BVEVEX0ZJRUxEIG5vIHF1b3RlIGRldGVjdGVkIGF0 IGVuZCBvZiBhIHF1b3RlZCBmaWVsZAotICAqCi0gICogSW4gYW55IGNhc2UsIHRoZSBzdHJp bmcgcmVhZCB1cCB0byB0aGUgdGVybWluYXRvciAob3IgZW5kIG9mIGZpbGUpCi0gICogaXMg cmV0dXJuZWQuCi0gICoKLSAgKiAqaXNudWxsIGlzIHNldCB0cnVlIG9yIGZhbHNlIGRlcGVu ZGluZyBvbiB3aGV0aGVyIHRoZSBpbnB1dCBtYXRjaGVkCi0gICogdGhlIG51bGwgbWFya2Vy LiAgTm90ZSB0aGF0IHRoZSBjYWxsZXIgY2Fubm90IGNoZWNrIHRoaXMgc2luY2UgdGhlCi0g ICogcmV0dXJuZWQgc3RyaW5nIHdpbGwgYmUgdGhlIHBvc3QtZGUtZXNjYXBpbmcgZXF1aXZh bGVudCwgd2hpY2ggbWF5Ci0gICogbG9vayB0aGUgc2FtZSBhcyBzb21lIHZhbGlkIGRhdGEg c3RyaW5nLgogICAqLS0tLS0tLS0tLQogICAqLwohIAohIHN0YXRpYyBjaGFyICoKISBDb3B5 UmVhZEF0dHJpYnV0ZUNTVihjb25zdCBjaGFyICpkZWxpbSwgY29uc3QgY2hhciAqbnVsbF9w cmludCwgY2hhciAqcXVvdGUsCiEgCQkJCQkgY2hhciAqZXNjYXBlLCBDb3B5UmVhZFJlc3Vs dCAqcmVzdWx0LCBib29sICppc251bGwpCiAgewogIAljaGFyCQlkZWxpbWMgPSBkZWxpbVsw XTsKICAJY2hhcgkJcXVvdGVjID0gcXVvdGVbMF07Ci0tLSAyOTUxLDI5NjIgLS0tLQogICAq CiAgICogbnVsbF9wcmludCBpcyB0aGUgbnVsbCBtYXJrZXIgc3RyaW5nLiAgTm90ZSB0aGF0 IHRoaXMgaXMgY29tcGFyZWQgdG8KICAgKiB0aGUgcHJlLWRlLWVzY2FwZWQgaW5wdXQgc3Ry aW5nICh0aHVzIGlmIGl0IGlzIHF1b3RlZCBpdCBpcyBub3QgYSBOVUxMKS4KICAgKi0tLS0t LS0tLS0KICAgKi8KISBzdGF0aWMgdm9pZAohIENvcHlSZWFkQXR0cmlidXRlc0NTVihjb25z dCBjaGFyICpkZWxpbSwgY29uc3QgY2hhciAqbnVsbF9wcmludCwgY2hhciAqcXVvdGUsCiEg CQkJCQkgIGNoYXIgKmVzY2FwZSwgaW50IG51bGxfcHJpbnRfbGVuLCBjaGFyICpudWxscywg TGlzdCAqYXR0bnVtbGlzdCwgCiEgCQkJCQkgIGludCAqYXR0cl9vZmZzZXRzLCBpbnQgbnVt X3BoeXNfYXR0cnMsIEZvcm1fcGdfYXR0cmlidXRlICphdHRyKQogIHsKICAJY2hhcgkJZGVs aW1jID0gZGVsaW1bMF07CiAgCWNoYXIJCXF1b3RlYyA9IHF1b3RlWzBdOwoqKioqKioqKioq KioqKioKKioqIDI0NzcsMjUwMiAqKioqCiAgCWludAkJCWlucHV0X2xlbjsKICAJYm9vbAkJ aW5fcXVvdGUgPSBmYWxzZTsKICAJYm9vbAkJc2F3X3F1b3RlID0gZmFsc2U7CiAgCiEgCS8q IHJlc2V0IGF0dHJpYnV0ZV9idWYgdG8gZW1wdHkgKi8KISAJYXR0cmlidXRlX2J1Zi5sZW4g PSAwOwohIAlhdHRyaWJ1dGVfYnVmLmRhdGFbMF0gPSAnXDAnOwogIAotIAkvKiBzZXQgZGVm YXVsdCBzdGF0dXMgKi8KLSAJKnJlc3VsdCA9IEVORF9PRl9MSU5FOwogIAogIAlmb3IgKDs7 KQogIAl7CiAgCQllbmRfY3Vyc29yID0gbGluZV9idWYuY3Vyc29yOwogIAkJaWYgKGxpbmVf YnVmLmN1cnNvciA+PSBsaW5lX2J1Zi5sZW4pCiAgCQkJYnJlYWs7CiAgCQljID0gbGluZV9i dWYuZGF0YVtsaW5lX2J1Zi5jdXJzb3IrK107CiAgCiAgCQkvKiB1bnF1b3RlZCBmaWVsZCBk ZWxpbWl0ZXIgICovCiAgCQlpZiAoIWluX3F1b3RlICYmIGMgPT0gZGVsaW1jKQogIAkJewoh IAkJCSpyZXN1bHQgPSBOT1JNQUxfQVRUUjsKISAJCQlicmVhazsKICAJCX0KICAKICAJCS8q IHN0YXJ0IG9mIHF1b3RlZCBmaWVsZCAob3IgcGFydCBvZiBmaWVsZCkgKi8KLS0tIDI5Njcs MzA0NyAtLS0tCiAgCWludAkJCWlucHV0X2xlbjsKICAJYm9vbAkJaW5fcXVvdGUgPSBmYWxz ZTsKICAJYm9vbAkJc2F3X3F1b3RlID0gZmFsc2U7CisgCWludAkJCWF0dG51bTsJCQkvKiBh dHRyaWJ1dGUgbnVtYmVyIGJlaW5nIHBhcnNlZCAqLworIAlpbnQJCQltOwkJCQkvKiBhdHRy aWJ1dGUgaW5kZXggYmVpbmcgcGFyc2VkICovCisgCUxpc3RDZWxsICAgKmN1cjsJCQkvKiBj dXJzb3IgdG8gYXR0cmlidXRlIGxpc3QgdXNlZCBmb3IgdGhpcyBDT1BZICovCiAgCiEgCS8q CiEgCSAqIGluaXQgdmFyaWFibGVzIGZvciBhdHRyaWJ1dGUgc2NhbgohIAkgKi8KISAJYXR0 cl9idWYubGVuID0gMDsKISAJYXR0cl9idWYuZGF0YVswXSA9ICdcMCc7CiEgCWF0dHJfYnVm LmN1cnNvciA9IDA7CiEgCQohIAljdXIgPSBsaXN0X2hlYWQoYXR0bnVtbGlzdCk7CiEgCWF0 dG51bSA9IGxmaXJzdF9pbnQoY3VyKTsKISAJbSA9IGF0dG51bSAtIDE7CiEgCWlucHV0X2xl biA9IDA7CiAgCiAgCiAgCWZvciAoOzspCiAgCXsKICAJCWVuZF9jdXJzb3IgPSBsaW5lX2J1 Zi5jdXJzb3I7CisgCQkKKyAJCS8qIGZpbmlzaGVkIHByb2Nlc3NpbmcgYXR0cmlidXRlcyBp biBsaW5lICovCiAgCQlpZiAobGluZV9idWYuY3Vyc29yID49IGxpbmVfYnVmLmxlbikKKyAJ CXsKKyAJCQkvKiBjaGVjayB3aGV0aGVyIHJhdyBpbnB1dCBtYXRjaGVkIG51bGwgbWFya2Vy ICovCisgCQkJaW5wdXRfbGVuID0gZW5kX2N1cnNvciAtIHN0YXJ0X2N1cnNvcjsKKyAJCQlp ZiAoIXNhd19xdW90ZSAmJiBpbnB1dF9sZW4gPT0gbnVsbF9wcmludF9sZW4gJiYKKyAJCQkJ c3RybmNtcCgmbGluZV9idWYuZGF0YVtzdGFydF9jdXJzb3JdLCBudWxsX3ByaW50LCBpbnB1 dF9sZW4pID09IDApCisgCQkJCW51bGxzW21dID0gJ24nOworIAkJCWVsc2UKKyAJCQkJbnVs bHNbbV0gPSAnICc7CisgCQkJCisgCQkJaWYgKGluX3F1b3RlKQorIAkJCQllcmVwb3J0KEVS Uk9SLAorIAkJCQkJCShlcnJjb2RlKEVSUkNPREVfQkFEX0NPUFlfRklMRV9GT1JNQVQpLAor IAkJCQkJCSBlcnJtc2coInVudGVybWluYXRlZCBDU1YgcXVvdGVkIGZpZWxkIikpKTsKKyAJ CQkKKyAJCQlpZiAobG5leHQoY3VyKSAhPSBOVUxMKQorIAkJCQllcmVwb3J0KEVSUk9SLAor IAkJCQkJCShlcnJjb2RlKEVSUkNPREVfQkFEX0NPUFlfRklMRV9GT1JNQVQpLAorIAkJCQkJ CSBlcnJtc2coIm1pc3NpbmcgZGF0YSBmb3IgY29sdW1uIFwiJXNcIiIsCisgCQkJCQkJCQlO YW1lU3RyKGF0dHJbbSArIDFdLT5hdHRuYW1lKSkpKTsKICAJCQlicmVhazsKKyAJCX0KKyAJ CQogIAkJYyA9IGxpbmVfYnVmLmRhdGFbbGluZV9idWYuY3Vyc29yKytdOwogIAogIAkJLyog dW5xdW90ZWQgZmllbGQgZGVsaW1pdGVyICAqLwogIAkJaWYgKCFpbl9xdW90ZSAmJiBjID09 IGRlbGltYykKICAJCXsKISAJCQkvKiBjaGVjayB3aGV0aGVyIHJhdyBpbnB1dCBtYXRjaGVk IG51bGwgbWFya2VyICovCiEgCQkJaW5wdXRfbGVuID0gZW5kX2N1cnNvciAtIHN0YXJ0X2N1 cnNvcjsKISAJCQlpZiAoIXNhd19xdW90ZSAmJiBpbnB1dF9sZW4gPT0gbnVsbF9wcmludF9s ZW4gJiYKISAJCQkJc3RybmNtcCgmbGluZV9idWYuZGF0YVtzdGFydF9jdXJzb3JdLCBudWxs X3ByaW50LCBpbnB1dF9sZW4pID09IDApCiEgCQkJCW51bGxzW21dID0gJ24nOwohIAkJCWVs c2UKISAJCQkJbnVsbHNbbV0gPSAnICc7CiEgICAgICAgICAgICAgCiEgCQkJLyogdGVybWlu YXRlIGF0dHIgc3RyaW5nIHdpdGggJ1wwJyAqLwohIAkJCWFwcGVuZFN0cmluZ0luZm9DaGFy TWFjcm8oJmF0dHJfYnVmLCAnXDAnKTsKISAJCQlhdHRyX2J1Zi5jdXJzb3IrKzsKISAJCQkK ISAJCQkvKiBzZXR1cCBuZXh0IGF0dHJpYnV0ZSBzY2FuICovCiEgCQkJY3VyID0gbG5leHQo Y3VyKTsKISAJCQkKISAJCQlpZiAoY3VyID09IE5VTEwpCiEgCQkJCWVyZXBvcnQoRVJST1Is CiEgCQkJCQkJKGVycmNvZGUoRVJSQ09ERV9CQURfQ09QWV9GSUxFX0ZPUk1BVCksCiEgCQkJ CQkJIGVycm1zZygiZXh0cmEgZGF0YSBhZnRlciBsYXN0IGV4cGVjdGVkIGNvbHVtbiIpKSk7 CiEgCQkJCiEgCQkJYXR0bnVtID0gbGZpcnN0X2ludChjdXIpOwohIAkJCW0gPSBhdHRudW0g LSAxOwohIAkJCWF0dHJfb2Zmc2V0c1ttXSA9IGF0dHJfYnVmLmN1cnNvcjsKISAJCQlzdGFy dF9jdXJzb3IgPSBsaW5lX2J1Zi5jdXJzb3I7CiEgCQkJY29udGludWU7CiAgCQl9CiAgCiAg CQkvKiBzdGFydCBvZiBxdW90ZWQgZmllbGQgKG9yIHBhcnQgb2YgZmllbGQpICovCioqKioq KioqKioqKioqKgoqKiogMjUyMCwyNTI3ICoqKioKICAKICAJCQkJaWYgKG5leHRjID09IGVz Y2FwZWMgfHwgbmV4dGMgPT0gcXVvdGVjKQogIAkJCQl7CiEgCQkJCQlhcHBlbmRTdHJpbmdJ bmZvQ2hhck1hY3JvKCZhdHRyaWJ1dGVfYnVmLCBuZXh0Yyk7CiAgCQkJCQlsaW5lX2J1Zi5j dXJzb3IrKzsKICAJCQkJCWNvbnRpbnVlOwogIAkJCQl9CiAgCQkJfQotLS0gMzA2NSwzMDcz IC0tLS0KICAKICAJCQkJaWYgKG5leHRjID09IGVzY2FwZWMgfHwgbmV4dGMgPT0gcXVvdGVj KQogIAkJCQl7CiEgCQkJCQlhcHBlbmRTdHJpbmdJbmZvQ2hhck1hY3JvKCZhdHRyX2J1Ziwg bmV4dGMpOwogIAkJCQkJbGluZV9idWYuY3Vyc29yKys7CisgCQkJCQlhdHRyX2J1Zi5jdXJz b3IrKzsKICAJCQkJCWNvbnRpbnVlOwogIAkJCQl9CiAgCQkJfQoqKioqKioqKioqKioqKioK KioqIDI1MzcsMjYxNiAqKioqCiAgCQkJaW5fcXVvdGUgPSBmYWxzZTsKICAJCQljb250aW51 ZTsKICAJCX0KISAJCWFwcGVuZFN0cmluZ0luZm9DaGFyTWFjcm8oJmF0dHJpYnV0ZV9idWYs IGMpOwogIAl9CiAgCi0gCWlmIChpbl9xdW90ZSkKLSAJCSpyZXN1bHQgPSBVTlRFUk1JTkFU RURfRklFTEQ7Ci0gCi0gCS8qIGNoZWNrIHdoZXRoZXIgcmF3IGlucHV0IG1hdGNoZWQgbnVs bCBtYXJrZXIgKi8KLSAJaW5wdXRfbGVuID0gZW5kX2N1cnNvciAtIHN0YXJ0X2N1cnNvcjsK LSAJaWYgKCFzYXdfcXVvdGUgJiYgaW5wdXRfbGVuID09IHN0cmxlbihudWxsX3ByaW50KSAm JgotIAkJc3RybmNtcCgmbGluZV9idWYuZGF0YVtzdGFydF9jdXJzb3JdLCBudWxsX3ByaW50 LCBpbnB1dF9sZW4pID09IDApCi0gCQkqaXNudWxsID0gdHJ1ZTsKLSAJZWxzZQotIAkJKmlz bnVsbCA9IGZhbHNlOwotIAotIAlyZXR1cm4gYXR0cmlidXRlX2J1Zi5kYXRhOwogIH0KICAK ICAvKgohICAqIFJlYWQgYSBiaW5hcnkgYXR0cmlidXRlCiAgICovCiEgc3RhdGljIERhdHVt CiEgQ29weVJlYWRCaW5hcnlBdHRyaWJ1dGUoaW50IGNvbHVtbl9ubywgRm1nckluZm8gKmZs aW5mbywKISAJCQkJCQlPaWQgdHlwaW9wYXJhbSwgaW50MzIgdHlwbW9kLAogIAkJCQkJCWJv b2wgKmlzbnVsbCkKICB7CiEgCWludDMyCQlmbGRfc2l6ZTsKISAJRGF0dW0JCXJlc3VsdDsK ISAKISAJZmxkX3NpemUgPSBDb3B5R2V0SW50MzIoKTsKISAJaWYgKENvcHlHZXRFb2YoKSkK ISAJCWVyZXBvcnQoRVJST1IsCiEgCQkJCShlcnJjb2RlKEVSUkNPREVfQkFEX0NPUFlfRklM RV9GT1JNQVQpLAohIAkJCQkgZXJybXNnKCJ1bmV4cGVjdGVkIEVPRiBpbiBDT1BZIGRhdGEi KSkpOwohIAlpZiAoZmxkX3NpemUgPT0gLTEpCiEgCXsKISAJCSppc251bGwgPSB0cnVlOwoh IAkJcmV0dXJuIChEYXR1bSkgMDsKICAJfQohIAlpZiAoZmxkX3NpemUgPCAwKQohIAkJZXJl cG9ydChFUlJPUiwKISAJCQkJKGVycmNvZGUoRVJSQ09ERV9CQURfQ09QWV9GSUxFX0ZPUk1B VCksCiEgCQkJCSBlcnJtc2coImludmFsaWQgZmllbGQgc2l6ZSIpKSk7CiAgCiEgCS8qIHJl c2V0IGF0dHJpYnV0ZV9idWYgdG8gZW1wdHksIGFuZCBsb2FkIHJhdyBkYXRhIGluIGl0ICov CiEgCWF0dHJpYnV0ZV9idWYubGVuID0gMDsKISAJYXR0cmlidXRlX2J1Zi5kYXRhWzBdID0g J1wwJzsKISAJYXR0cmlidXRlX2J1Zi5jdXJzb3IgPSAwOwogIAohIAllbmxhcmdlU3RyaW5n SW5mbygmYXR0cmlidXRlX2J1ZiwgZmxkX3NpemUpOwogIAotIAlDb3B5R2V0RGF0YShhdHRy aWJ1dGVfYnVmLmRhdGEsIGZsZF9zaXplKTsKLSAJaWYgKENvcHlHZXRFb2YoKSkKLSAJCWVy ZXBvcnQoRVJST1IsCi0gCQkJCShlcnJjb2RlKEVSUkNPREVfQkFEX0NPUFlfRklMRV9GT1JN QVQpLAotIAkJCQkgZXJybXNnKCJ1bmV4cGVjdGVkIEVPRiBpbiBDT1BZIGRhdGEiKSkpOwog IAohIAlhdHRyaWJ1dGVfYnVmLmxlbiA9IGZsZF9zaXplOwohIAlhdHRyaWJ1dGVfYnVmLmRh dGFbZmxkX3NpemVdID0gJ1wwJzsKICAKISAJLyogQ2FsbCB0aGUgY29sdW1uIHR5cGUncyBi aW5hcnkgaW5wdXQgY29udmVydGVyICovCiEgCXJlc3VsdCA9IEZ1bmN0aW9uQ2FsbDMoZmxp bmZvLAohIAkJCQkJCSAgIFBvaW50ZXJHZXREYXR1bSgmYXR0cmlidXRlX2J1ZiksCiEgCQkJ CQkJICAgT2JqZWN0SWRHZXREYXR1bSh0eXBpb3BhcmFtKSwKISAJCQkJCQkgICBJbnQzMkdl dERhdHVtKHR5cG1vZCkpOwogIAotIAkvKiBUcm91YmxlIGlmIGl0IGRpZG4ndCBlYXQgdGhl IHdob2xlIGJ1ZmZlciAqLwotIAlpZiAoYXR0cmlidXRlX2J1Zi5jdXJzb3IgIT0gYXR0cmli dXRlX2J1Zi5sZW4pCi0gCQllcmVwb3J0KEVSUk9SLAotIAkJCQkoZXJyY29kZShFUlJDT0RF X0lOVkFMSURfQklOQVJZX1JFUFJFU0VOVEFUSU9OKSwKLSAJCQkJIGVycm1zZygiaW5jb3Jy ZWN0IGJpbmFyeSBkYXRhIGZvcm1hdCIpKSk7CiAgCi0gCSppc251bGwgPSBmYWxzZTsKLSAJ cmV0dXJuIHJlc3VsdDsKLSB9CiAgCiAgLyoKICAgKiBTZW5kIHRleHQgcmVwcmVzZW50YXRp b24gb2Ygb25lIGF0dHJpYnV0ZSwgd2l0aCBjb252ZXJzaW9uIGFuZCBlc2NhcGluZwotLS0g MzA4MywzMTQ5IC0tLS0KICAJCQlpbl9xdW90ZSA9IGZhbHNlOwogIAkJCWNvbnRpbnVlOwog IAkJfQohIAkJYXBwZW5kU3RyaW5nSW5mb0NoYXJNYWNybygmYXR0cl9idWYsIGMpOwohIAkJ YXR0cl9idWYuY3Vyc29yKys7CiAgCX0KICAKICB9CiAgCiAgLyoKISAgKiBSZWFkIHRoZSBm aXJzdCBhdHRyaWJ1dGUuIFRoaXMgaXMgbWFpbmx5IHVzZWQgdG8gbWFpbnRhaW4gc3VwcG9y dAohICAqIGZvciBhbiBPSUQgY29sdW1uLiBBbGwgdGhlIHJlc3Qgb2YgdGhlIGNvbHVtbnMg d2lsbCBiZSByZWFkIGF0IG9uY2Ugd2l0aAohICAqIENvcHlSZWFkQXR0cmlidXRlc1RleHQu CiAgICovCiEgc3RhdGljIGNoYXIgKgohIENvcHlSZWFkT2lkQXR0cihjb25zdCBjaGFyICpk ZWxpbSwgY29uc3QgY2hhciAqbnVsbF9wcmludCwgaW50IG51bGxfcHJpbnRfbGVuLAogIAkJ CQkJCWJvb2wgKmlzbnVsbCkKICB7CiEgCWNoYXIJCWRlbGltYyA9IGRlbGltWzBdOwohIAlj aGFyCSAgICpzdGFydF9sb2MgPSBsaW5lX2J1Zi5kYXRhICsgbGluZV9idWYuY3Vyc29yOwoh IAljaGFyCSAgICplbmRfbG9jOwohIAlpbnQJCQlhdHRyX2xlbiA9IDA7CiEgCWludAkJCWJ5 dGVzX3JlbWFpbmluZzsKISAKISAJLyogcmVzZXQgYXR0cmlidXRlIGJ1ZiB0byBlbXB0eSAq LwohIAlhdHRyX2J1Zi5sZW4gPSAwOwohIAlhdHRyX2J1Zi5kYXRhWzBdID0gJ1wwJzsKISAJ YXR0cl9idWYuY3Vyc29yID0gMDsKISAKISAJLyogIyBvZiBieXRlcyB0aGF0IHdlcmUgbm90 IHlldCBwcm9jZXNzZWQgaW4gdGhpcyBsaW5lICovCiEgCWJ5dGVzX3JlbWFpbmluZyA9IGxp bmVfYnVmLmxlbiAtIGxpbmVfYnVmLmN1cnNvcjsKISAKISAJLyogZ290IHRvIGVuZCBvZiBs aW5lICovCiEgCWlmICgoZW5kX2xvYyA9IHNjYW5UZXh0TGluZShzdGFydF9sb2MsIGRlbGlt YywgYnl0ZXNfcmVtYWluaW5nKSkgPT0gTlVMTCkKISAJewohIAkJYXR0cl9sZW4gPSBieXRl c19yZW1haW5pbmcgLSAxOyAvKiBkb24ndCBjb3VudCAnXG4nIGluIGxlbiBjYWxjdWxhdGlv biAqLwohIAkJYXBwZW5kQmluYXJ5U3RyaW5nSW5mbygmYXR0cl9idWYsIHN0YXJ0X2xvYywg YXR0cl9sZW4pOwohIAkJbGluZV9idWYuY3Vyc29yICs9IGF0dHJfbGVuICsgMjsJCS8qIHNr aXAgJ1xuJyBhbmQgJ1wwJyAqLwogIAl9CiEgCWVsc2UKISAJCS8qIGZvdW5kIGEgZGVsaW1p dGVyICovCiEgCXsKISAJCS8qCiEgCQkgKiAod2UgZG9uJ3QgY2FyZSBpZiBkZWxpbSB3YXMg cHJlY2VkZWQgd2l0aCBhIGJhY2tzbGFzaCwgYmVjYXVzZSBpdCdzCiEgCQkgKiBhbiBpbnZh bGlkIE9JRCBhbnl3YXkpCiEgCQkgKi8KICAKISAJCWF0dHJfbGVuID0gZW5kX2xvYyAtIHN0 YXJ0X2xvYzsgLyogd2UgZG9uJ3QgaW5jbHVkZSB0aGUgZGVsaW1pdGVyIGNoICovCiAgCiEg CQlhcHBlbmRCaW5hcnlTdHJpbmdJbmZvKCZhdHRyX2J1Ziwgc3RhcnRfbG9jLCBhdHRyX2xl bik7CiEgCQlsaW5lX2J1Zi5jdXJzb3IgKz0gYXR0cl9sZW4gKyAxOwohIAl9CiAgCiAgCiEg CS8qIGNoZWNrIHdoZXRoZXIgcmF3IGlucHV0IG1hdGNoZWQgbnVsbCBtYXJrZXIgKi8KISAJ aWYgKGF0dHJfbGVuID09IG51bGxfcHJpbnRfbGVuICYmIHN0cm5jbXAoc3RhcnRfbG9jLCBu dWxsX3ByaW50LCBhdHRyX2xlbikgPT0gMCkKISAJCSppc251bGwgPSB0cnVlOwohIAllbHNl CiEgCQkqaXNudWxsID0gZmFsc2U7CiAgCiEgCXJldHVybiBhdHRyX2J1Zi5kYXRhOwohIH0K ICAKICAKICAKICAvKgogICAqIFNlbmQgdGV4dCByZXByZXNlbnRhdGlvbiBvZiBvbmUgYXR0 cmlidXRlLCB3aXRoIGNvbnZlcnNpb24gYW5kIGVzY2FwaW5nCioqKioqKioqKioqKioqKgoq KiogMjgwMCwyODAyICoqKioKLS0tIDMzMzMsMzQzNiAtLS0tCiAgCiAgCXJldHVybiBhdHRu dW1zOwogIH0KKyAKKyAvKgorICAqIFRoZSBmb2xsb3dpbmcgYXJlIGN1c3RvbSB2ZXJzaW9u cyBvZiB0aGUgc3RyaW5nIGZ1bmN0aW9uIHN0cmNocigpLgorICAqIEFzIG9wcG9zZWQgdG8g dGhlIG9yaWdpbmFsIHN0cmNociB3aGljaCBzZWFyY2hlcyB0aHJvdWdoCisgICogYSBzdHJp bmcgdW50aWwgdGhlIHRhcmdldCBjaGFyYWN0ZXIgaXMgZm91bmQsIG9yIGEgTlVMTCBpcwor ICAqIGZvdW5kLCB0aGlzIHZlcnNpb24gd2lsbCBub3QgcmV0dXJuIHdoZW4gYSBOVUxMIGlz IGZvdW5kLgorICAqIEluc3RlYWQgaXQgd2lsbCBzZWFyY2ggdGhyb3VnaCBhIHByZS1kZWZp bmVkIGxlbmd0aCBvZgorICAqIGJ5dGVzIGFuZCB3aWxsIHJldHVybiBvbmx5IGlmIHRoZSB0 YXJnZXQgY2hhcmFjdGVyKHMpIGlzIHJlYWNoZWQuCisgICoKKyAgKiBJZiBvdXIgY2xpZW50 IGVuY29kaW5nIGlzIG5vdCBhIHN1cHBvcnRlZCBzZXJ2ZXIgZW5jb2RpbmcsIHdlCisgICog a25vdyB0aGF0IGl0IGlzIG5vdCBzYWZlIHRvIGxvb2sgYXQgZWFjaCBjaGFyYWN0ZXIgYXMg dHJhaWxpbmcKKyAgKiBieXRlIGluIGEgbXVsdGlieXRlIGNoYXJhY3RlciBtYXkgYmUgYSA3 LWJpdCBBU0NJSSBlcXVpdmFsZW50LgorICAqIFRoZXJlZm9yZSB3ZSB1c2UgcGdfZW5jb2Rp bmdfbWJsZW4gdG8gc2tpcCB0byB0aGUgZW5kIG9mIHRoZQorICAqIGNoYXJhY3Rlci4KKyAg KgorICAqIHJldHVybnM6CisgICoJIHBvaW50ZXIgdG8gYyAtIGlmIGMgaXMgbG9jYXRlZCB3 aXRoaW4gdGhlIHN0cmluZy4KKyAgKgkgTlVMTCAtIGlmIGMgd2FzIG5vdCBmb3VuZCBpbiBz cGVjaWZpZWQgbGVuZ3RoIG9mIHNlYXJjaC4gTm90ZToKKyAgKgkJCXRoaXMgRE9FU04nVCBt ZWFuIHRoYXQgYSAnXDAnIHdhcyByZWFjaGVkLgorICAqLworIGNoYXIgKgorIHNjYW5UZXh0 TGluZShjb25zdCBjaGFyICpzLCBjaGFyIGVvbCwgc2l6ZV90IGxlbikKKyB7CisgCWNvbnN0 IGNoYXIgKnN0YXJ0OworIAorIAlpZiAoY2xpZW50X2VuY29kaW5nX29ubHkgJiYgIWxpbmVf YnVmX2NvbnZlcnRlZCkKKyAJeworIAkJaW50CQkJbWJsZW4gPSBwZ19lbmNvZGluZ19tYmxl bihjbGllbnRfZW5jb2RpbmcsICh1bnNpZ25lZCBjaGFyKilzKTsKKyAKKyAJCWZvciAoc3Rh cnQgPSBzOyAqcyAhPSBlb2wgJiYgcyA8IHN0YXJ0ICsgbGVuOyBzICs9IG1ibGVuKQorIAkJ CW1ibGVuID0gcGdfZW5jb2RpbmdfbWJsZW4oY2xpZW50X2VuY29kaW5nLCAodW5zaWduZWQg Y2hhciopcyk7CisgCQkKKyAJCXJldHVybiAoKCpzID09IGVvbCkgPyAoY2hhciAqKSBzIDog TlVMTCk7CisgCX0KKyAJZWxzZQorIAkJcmV0dXJuIG1lbWNocihzLGVvbCxsZW4pOworIH0K KyAKKyAKKyBjaGFyICoKKyBzY2FuQ1NWTGluZShjb25zdCBjaGFyICpzLCBjaGFyIGVvbCwg Y2hhciBlc2NhcGVjLCBjaGFyIHF1b3RlYywgc2l6ZV90IGxlbikKKyB7CisgCWNvbnN0IGNo YXIgKnN0YXJ0OworIAkKKyAJaWYgKGNsaWVudF9lbmNvZGluZ19vbmx5ICYmICFsaW5lX2J1 Zl9jb252ZXJ0ZWQpCisgCXsKKyAJCWludAkJCW1ibGVuID0gcGdfZW5jb2RpbmdfbWJsZW4o Y2xpZW50X2VuY29kaW5nLCAodW5zaWduZWQgY2hhciopcyk7CisgCQkKKyAJCWZvciAoc3Rh cnQgPSBzOyAqcyAhPSBlb2wgJiYgIHMgPCBzdGFydCArIGxlbjsgcyArPSBtYmxlbikKKyAJ CXsKKyAJCQlpZiAoaW5fcXVvdGUgJiYgKnMgPT0gZXNjYXBlYykKKyAJCQkJbGFzdF93YXNf ZXNjID0gIWxhc3Rfd2FzX2VzYzsKKyAJCQlpZiAoKnMgPT0gcXVvdGVjICYmICFsYXN0X3dh c19lc2MpCisgCQkJCWluX3F1b3RlID0gIWluX3F1b3RlOworIAkJCWlmICgqcyAhPSBlc2Nh cGVjKQorIAkJCQlsYXN0X3dhc19lc2MgPSBmYWxzZTsKKyAJCQkJCQorIAkJCW1ibGVuID0g cGdfZW5jb2RpbmdfbWJsZW4oY2xpZW50X2VuY29kaW5nLCAodW5zaWduZWQgY2hhciopcyk7 CisgICAgICAgICB9CisgCX0KKyAJZWxzZQorIAkJLyogc2FmZSB0byBzY3JvbGwgYnl0ZSBi eSBieXRlICovCisgCXsKKyAJCWZvciAoc3RhcnQgPSBzOyAqcyAhPSBlb2wgICYmIHMgPCBz dGFydCArIGxlbjsgcysrKQorIAkJeworIAkJCWlmIChpbl9xdW90ZSAmJiAqcyA9PSBlc2Nh cGVjKQorIAkJCQlsYXN0X3dhc19lc2MgPSAhbGFzdF93YXNfZXNjOworIAkJCWlmICgqcyA9 PSBxdW90ZWMgJiYgIWxhc3Rfd2FzX2VzYykKKyAJCQkJaW5fcXVvdGUgPSAhaW5fcXVvdGU7 CisgCQkJaWYgKCpzICE9IGVzY2FwZWMpCisgCQkJCWxhc3Rfd2FzX2VzYyA9IGZhbHNlOwor IAkJfQorIAkJCQorIAl9CisgCQorIAlpZigqcyAhPSBlc2NhcGVjKSAvKiBmb3VuZCBlb2xf Y2ggKi8KKyAJCWxhc3Rfd2FzX2VzYyA9IGZhbHNlOworIAkKKyAJcmV0dXJuICgoKnMgPT0g ZW9sKSA/IChjaGFyICopIHMgOiBOVUxMKTsKKyB9CisgCisgLyoKKyAgKiBTY2FuIGZvciAx IG9mIDIgY2hhcmFjdGVycyBuZWNlc3NhcnkgZm9yIGF0dHJpYnV0ZSBwYXJzaW5nLiBObyBu ZWVkCisgICogZm9yIGNoZWNraW5nIGZvciBtdWx0aS1ieXRlIGNoYXJhY3RlcnMgc2luY2Ug Y29udmVyc2lvbiBhbHJlYWR5IAorICAqIGhhcHBlbmVkLgorICAqLworIGNoYXIgKgorIHNj YW5UZXh0QXR0cihjb25zdCBjaGFyICpzLCBjaGFyIGMxLCBjaGFyIGMyLCBzaXplX3QgbGVu KQorIHsKKyAJY29uc3QgY2hhciAqc3RhcnQ7CisgCQorIAlmb3IgKHN0YXJ0ID0gczsgKnMg IT0gYzEgJiYgKnMgIT0gYzIgJiYgcyA8IHN0YXJ0ICsgbGVuOyBzKyspCisgCQk7CisgCQor IAlyZXR1cm4gKCpzICE9ICdcMCcgPyAoY2hhciAqKSBzIDogTlVMTCk7CisgfQorIAorIAor IAorIAorIAo= --B_3205825185_103175599-- From pgsql-patches-owner@postgresql.org Tue Aug 2 12:03:57 2005 X-Original-To: pgsql-patches-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 760BB52C6D; Tue, 2 Aug 2005 12:03:54 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 20548-10; Tue, 2 Aug 2005 15:03:43 +0000 (GMT) Received: from mail.Mi8.com (nycgw05.mi8.com [63.240.6.46]) by svr1.postgresql.org (Postfix) with ESMTP id ADEBE52F23; Tue, 2 Aug 2005 12:03:40 -0300 (ADT) Received: from 172.16.1.110 by mail.Mi8.com with ESMTP (- GW05 Welcome to Mi8 Corporation www.Mi8.com); Tue, 02 Aug 2005 11:03:35 -0400 X-Server-Uuid: E0C866E6-C6CD-48B5-AE61-E57E73CF3CC7 Received: from MI8NYCMAIL06.Mi8.com ([172.16.1.175]) by D01SMTP01.Mi8.com with Microsoft SMTPSVC(6.0.3790.211); Tue, 2 Aug 2005 11:03:35 -0400 Received: from 141.156.39.156 ([141.156.39.156]) by MI8NYCMAIL06.Mi8.com ([172.16.1.219]) via Exchange Front-End Server mi8owa.mi8.com ( [172.16.1.104]) with Microsoft Exchange Server HTTP-DAV ; Tue, 2 Aug 2005 11:03:35 -0500 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Tue, 02 Aug 2005 11:03:44 -0400 Subject: Re: [PERFORM] COPY FROM performance improvements From: "Alon Goldshuv" To: pgsql-patches@postgresql.org, "pgsql-performance@postgresql.org" Message-ID: In-Reply-To: <26131.1122940270@sss.pgh.pa.us> MIME-Version: 1.0 X-OriginalArrivalTime: 02 Aug 2005 15:03:35.0747 (UTC) FILETIME=[5B39B930:01C59773] X-WSS-ID: 6EF154CD26S618902-01-01 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.569 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO, RCVD_NUMERIC_HELO X-Spam-Level: * X-Archive-Number: 200508/29 X-Sequence-Number: 16976 Tom, Thanks for pointing it out. I made the small required modifications to match copy.c version 1.247 and sent it to -patches list. New patch is V16. Alon. On 8/1/05 7:51 PM, "Tom Lane" wrote: > "Alon Goldshuv" writes: >>> This patch appears to reverse out the most recent committed changes in >>> copy.c. > >> Which changes do you refer to? I thought I accommodated all the recent >> changes (I recall some changes to the tupletable/tupleslot interface, HEADER >> in cvs, and hex escapes and maybe one or 2 more). What did I miss? > > The latest touch of copy.c, namely this patch: > > 2005-07-10 17:13 tgl > > * doc/src/sgml/ref/create_type.sgml, src/backend/commands/copy.c, > src/backend/commands/typecmds.c, src/backend/tcop/fastpath.c, > src/backend/tcop/postgres.c, src/backend/utils/adt/arrayfuncs.c, > src/backend/utils/adt/date.c, src/backend/utils/adt/numeric.c, > src/backend/utils/adt/rowtypes.c, > src/backend/utils/adt/timestamp.c, src/backend/utils/adt/varbit.c, > src/backend/utils/adt/varchar.c, src/backend/utils/adt/varlena.c, > src/backend/utils/mb/mbutils.c, src/include/catalog/catversion.h, > src/include/catalog/pg_proc.h, > src/test/regress/expected/type_sanity.out, > src/test/regress/sql/type_sanity.sql: Change typreceive function > API so that receive functions get the same optional arguments as > text input functions, ie, typioparam OID and atttypmod. Make all > the datatypes that use typmod enforce it the same way in typreceive > as they do in typinput. This fixes a problem with failure to > enforce length restrictions during COPY FROM BINARY. > > It was rather obvious, given that the first chunk of the patch backed up > the file's CVS version stamp from 1.247 to 1.246 :-( > > regards, tom lane > From pgsql-performance-owner@postgresql.org Tue Aug 2 13:05:00 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1939D52C8A for ; Tue, 2 Aug 2005 13:04:55 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 40098-02 for ; Tue, 2 Aug 2005 16:04:50 +0000 (GMT) Received: from mail.nordicbet.com (mail.nordicbet.com [193.69.167.130]) by svr1.postgresql.org (Postfix) with ESMTP id 3B81552BE2 for ; Tue, 2 Aug 2005 13:04:48 -0300 (ADT) Received: from [84.48.51.94] (helo=tobias.nordicbet.invalid) (Authenticated Sender=tobias@nordicbet.com) by mail.nordicbet.com with esmtpa (Exim 4.50 #1 (Debian)) id 1DzzE4-0004oe-Rn for ; Tue, 02 Aug 2005 18:02:56 +0200 Received: by tobias.nordicbet.invalid (Postfix, from userid 500) id 06DA6E012E; Tue, 2 Aug 2005 18:04:34 +0200 (CEST) Date: Tue, 2 Aug 2005 18:04:34 +0200 From: Tobias Brox To: pgsql-performance@postgresql.org Subject: "nice"/low priority Query Message-ID: <20050802160434.GD9278@tobias.lan> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Organization: Group Nordicbet User-Agent: Mutt/1.5.8i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/20 X-Sequence-Number: 13767 The short question: Is there any ways to give postgresql a hint that a particular SQL call should be run at lower priority? Since every db connection has a pid, I can manually run "renice" to scheduele it by the OS - but of course I can't do it manually all the time. The long story: We have a constantly growing database, and hence also a constantly growing load on the database server. A hardware upgrade has for different reasons been postponed, and it's still beeing postponed. We were hitting the first capacity problems in June, though so far I've managed to keep the situation in check by tuning the configuration, adding indices, optimizing queries, doing cacheing in the application, and at one point in the code I'm even asking the database for "explain plan", grepping out the estimated cost number, and referring the user to take contact with the IT-dept if he really needs the report. But I digress. Still there are lots of CPU power available - normally the server runs with 50-80% of the CPUs idle, it's just the spikes that kills us. We basically have two kind of queries that are significant - an ever-ongoing "critical" rush of simple queries, both reading and writing to the database, plus some few heavy "non-critical" read-only queries that may cause significant iowait. The problem comes when we are so unlucky that two or three heavy queries are run simultaneously; we get congestion problems - instead of the applications just running a bit slower, they run _much_ slower. Ideally, if it was trivial to give priorities, it should be possible to keep the CPUs running at 100% for hours without causing critical problems...? -- Tobias Brox, +47-91700050 Tromso, Norway From pgsql-performance-owner@postgresql.org Tue Aug 2 13:19:49 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id EF7CE52C9F for ; Tue, 2 Aug 2005 13:19:47 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 44138-01 for ; Tue, 2 Aug 2005 16:19:45 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 56F8E529D2 for ; Tue, 2 Aug 2005 13:19:44 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j72GJUYJ011485; Tue, 2 Aug 2005 12:19:30 -0400 (EDT) To: Tobias Brox Cc: pgsql-performance@postgresql.org Subject: Re: "nice"/low priority Query In-reply-to: <20050802160434.GD9278@tobias.lan> References: <20050802160434.GD9278@tobias.lan> Comments: In-reply-to Tobias Brox message dated "Tue, 02 Aug 2005 18:04:34 +0200" Date: Tue, 02 Aug 2005 12:19:30 -0400 Message-ID: <11484.1122999570@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/21 X-Sequence-Number: 13768 Tobias Brox writes: > Is there any ways to give postgresql a hint that a > particular SQL call should be run at lower priority? Since every db > connection has a pid, I can manually run "renice" to scheduele it by the OS > - but of course I can't do it manually all the time. And it won't help you anyway, because renice only affects CPU priority not I/O scheduling ... which, by your description, is the real problem. I think the only thing that's likely to help much is trying to arrange that the "simple" queries only need to touch pages that are already in memory. Some playing around with shared_buffer sizing might help. Also, if you're not on PG 8.0.*, an update might help. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 2 14:00:48 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 57AC9529D2 for ; Tue, 2 Aug 2005 14:00:46 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 53174-02 for ; Tue, 2 Aug 2005 17:00:34 +0000 (GMT) Received: from mail.metronet.co.uk (mail.metronet.co.uk [213.162.97.75]) by svr1.postgresql.org (Postfix) with ESMTP id F032452E8E for ; Tue, 2 Aug 2005 14:00:30 -0300 (ADT) Received: from mainbox.archonet.com (84-51-143-99.archon037.adsl.metronet.co.uk [84.51.143.99]) by smtp.metronet.co.uk (MetroNet Mail) with ESMTP id DC2B540C3C5; Tue, 2 Aug 2005 18:00:17 +0100 (BST) Received: from localhost (localhost.localdomain [127.0.0.1]) by mainbox.archonet.com (Postfix) with ESMTP id 304D815EDA; Tue, 2 Aug 2005 17:54:56 +0100 (BST) Received: from mainbox.archonet.com ([127.0.0.1]) by localhost (mainbox [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 06583-10; Tue, 2 Aug 2005 17:54:53 +0100 (BST) Received: from [192.168.1.17] (client17.office.archonet.com [192.168.1.17]) by mainbox.archonet.com (Postfix) with ESMTP id 1D39C15ED9; Tue, 2 Aug 2005 17:54:53 +0100 (BST) Message-ID: <42EFA55C.3080903@archonet.com> Date: Tue, 02 Aug 2005 17:54:52 +0100 From: Richard Huxton User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: Tobias Brox , pgsql-performance@postgresql.org Subject: Re: "nice"/low priority Query References: <20050802160434.GD9278@tobias.lan> <11484.1122999570@sss.pgh.pa.us> In-Reply-To: <11484.1122999570@sss.pgh.pa.us> Content-Type: text/plain; charset=ISO-8859-15; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/22 X-Sequence-Number: 13769 Tom Lane wrote: > Tobias Brox writes: > >>Is there any ways to give postgresql a hint that a >>particular SQL call should be run at lower priority? Since every db >>connection has a pid, I can manually run "renice" to scheduele it by the OS >>- but of course I can't do it manually all the time. > > And it won't help you anyway, because renice only affects CPU priority > not I/O scheduling ... which, by your description, is the real problem. > > I think the only thing that's likely to help much is trying to arrange > that the "simple" queries only need to touch pages that are already in > memory. Some playing around with shared_buffer sizing might help. > Also, if you're not on PG 8.0.*, an update might help. Would it be useful to be able to re-use the vacuum_cost_xxx settings in 8.0 for this sort of thing? I'm thinking a long-running report query isn't that different from a vacuum. -- Richard Huxton Archonet Ltd From pgsql-performance-owner@postgresql.org Tue Aug 2 14:25:55 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0893A52E74 for ; Tue, 2 Aug 2005 14:25:54 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 58127-06 for ; Tue, 2 Aug 2005 17:25:51 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [67.100.216.10]) by svr1.postgresql.org (Postfix) with ESMTP id 5280452CA7 for ; Tue, 2 Aug 2005 14:25:50 -0300 (ADT) Received: by flake.decibel.org (Postfix, from userid 1001) id 86CA715232; Tue, 2 Aug 2005 12:25:50 -0500 (CDT) Date: Tue, 2 Aug 2005 12:25:50 -0500 From: "Jim C. Nasby" To: Tom Lane Cc: Tobias Brox , pgsql-performance@postgresql.org Subject: Re: "nice"/low priority Query Message-ID: <20050802172550.GZ60019@decibel.org> References: <20050802160434.GD9278@tobias.lan> <11484.1122999570@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <11484.1122999570@sss.pgh.pa.us> X-Operating-System: FreeBSD 4.11-RELEASE-p10 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.003 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/23 X-Sequence-Number: 13770 On Tue, Aug 02, 2005 at 12:19:30PM -0400, Tom Lane wrote: > Tobias Brox writes: > > Is there any ways to give postgresql a hint that a > > particular SQL call should be run at lower priority? Since every db > > connection has a pid, I can manually run "renice" to scheduele it by the OS > > - but of course I can't do it manually all the time. > > And it won't help you anyway, because renice only affects CPU priority > not I/O scheduling ... which, by your description, is the real problem. Actually, from what I've read 4.2BSD actually took priority into account when scheduling I/O. I don't know if this behavior is still present in FreeBSD or the like, though. So depending on the OS, priority could play a role in determining I/O scheduling. -- Jim C. Nasby, Database Consultant decibel@decibel.org Give your computer some brain candy! www.distributed.net Team #1828 Windows: "Where do you want to go today?" Linux: "Where do you want to go tomorrow?" FreeBSD: "Are you guys coming, or what?" From pgsql-performance-owner@postgresql.org Tue Aug 2 15:57:55 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id EFBF45298A for ; Tue, 2 Aug 2005 15:57:53 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 81274-02 for ; Tue, 2 Aug 2005 18:57:51 +0000 (GMT) Received: from cdrsmtp01.yellowbook.com (unknown [64.199.226.83]) by svr1.postgresql.org (Postfix) with ESMTP id 545C0528FC for ; Tue, 2 Aug 2005 15:57:49 -0300 (ADT) Received: from ybcdrmta01.corp.ybusa.net ([10.5.17.170]) by cdrsmtp01.yellowbook.com with Microsoft SMTPSVC(6.0.3790.0); Tue, 2 Aug 2005 14:01:14 -0500 Received: from YBCDREX01.corp.ybusa.net ([10.5.17.162]) by ybcdrmta01.corp.ybusa.net with Microsoft SMTPSVC(6.0.3790.211); Tue, 2 Aug 2005 13:57:49 -0500 X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----_=_NextPart_001_01C59794.14BE874F" Subject: pg_dump for table with bytea takes a long time Date: Tue, 2 Aug 2005 13:57:50 -0500 Message-ID: <3FB3AAE149F4AD4D8499E15EE8CF41A3460487@YBCDREX01.corp.ybusa.net> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: pg_dump for table with bytea takes a long time Thread-Index: AcWXlBPdbGKJiWXZTU2vktSIqJrfAg== From: "Sailer, Denis (YBUSA-CDR)" To: X-OriginalArrivalTime: 02 Aug 2005 18:57:49.0402 (UTC) FILETIME=[13DB4BA0:01C59794] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.374 tagged_above=0 required=5 tests=AWL, HTML_MESSAGE, HTML_TEXT_AFTER_BODY, HTML_TEXT_AFTER_HTML X-Spam-Level: X-Archive-Number: 200508/24 X-Sequence-Number: 13771 This is a multi-part message in MIME format. ------_=_NextPart_001_01C59794.14BE874F Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Dumping a database which contains a table with a bytea column takes approximately 25 hours and 45 minutes. The database has 26 tables in it. The other 25 tables take less than 5 minutes to dump so almost all time is spent dumping the bytea table. =20 prd1=3D# \d ybnet.ebook_master; Table "ybnet.ebook_master" Column | Type | Modifiers --------------+---------+----------- region_key | integer | not null book_key | integer | not null pub_sequence | integer | not null section_code | integer | not null pagenbr | integer | not null pdffile | bytea | Indexes: "ebook_master_pkey" PRIMARY KEY, btree (book_key, pub_sequence, section_code, pagenbr, region_key) Foreign-key constraints: "FK1_book_year" FOREIGN KEY (book_key, pub_sequence, region_key) REFERENCES ybnet.book_year(book_key, pub_sequence, region_key) "FK1_ebook_section" FOREIGN KEY (section_code) REFERENCES ybnet.ebook_section(sectioncode) Tablespace: "ebook" =20 The tablespace ebook is 65504295 bytes in size and the ebook_master table has 61-1GB files associated to it. =20 The command to dump the database is: =20 pg_dump --file=3D$DUMP_FILE --format=3Dc --data-only --verbose --host=3Dybcdrdbp01 $DATABASE =20 I also perform a hot backup of this database using pg_start_backup(), tar, and pg_stop_backup(). It takes only 20 minutes to create a tar ball of the entire 62GB. I like the speed of this method, but it does not allow me to restore 1 table at a time.=20 =20 The version of postgres is PostgreSQL 8.0.0 on i686-pc-linux-gnu, compiled by GCC gcc (GCC) 3.2.2 =20 The machine has 4 Xeon 3.00 GHz processors with hyper-threading on and 4GB of memory. Postgres is supported by two file systems connected to an EMC SAN disk array. One 2 GB one for the log files and a second 500 GB one for the data and indexes. All output files for the backup files are placed onto the 500 GB volume group and then backed up to an external storage manager. =20 Portions of the config file are: =20 shared_buffers =3D 16384 work_mem =3D 8192 maintenance_work_mem =3D 16384 =20 max_fsm_pages =3D 512000 max_fsm_relations =3D 1000 fsync =3D true =20 # - Checkpoints - checkpoint_segments =3D 20 =20 # - Planner Cost Constants - effective_cache_size =3D 262144 random_page_cost =3D 3 =20 =20 I am looking for ideas for making the backup of the above table much faster. ------_=_NextPart_001_01C59794.14BE874F Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable

Dumping a database which contains a table with a = bytea column takes approximately 25 hours and 45 minutes.  The database = has 26 tables in it. The other 25 tables take less than 5 minutes to dump so = almost all time is spent dumping the bytea table.

 

prd1=3D# \d = ybnet.ebook_master;

     Table "ybnet.ebook_master"

    Column    |  Type   | = Modifiers

--------------+---------+-----------

 region_key   | integer | not null

 book_key     | integer | not null

 pub_sequence = | integer | not null

 section_code = | integer | not null

 pagenbr      | integer | not null

 pdffile      | bytea   |

Indexes:

    = "ebook_master_pkey" PRIMARY KEY, btree (book_key, pub_sequence, section_code, pagenbr, = region_key)

Foreign-key = constraints:

    "FK1_book_year" FOREIGN KEY (book_key, pub_sequence, = region_key) REFERENCES ybnet.book_year(book_key, pub_sequence, = region_key)

    "FK1_ebook_section" FOREIGN KEY (section_code) REFERENCES = ybnet.ebook_section(sectioncode)

Tablespace: = "ebook"

 

The tablespace ebook is 65504295 bytes in size and = the ebook_master table has 61-1GB files associated to it.

 

The command to dump the database = is:

 

pg_dump --file=3D$DUMP_FILE --format=3Dc = --data-only --verbose -–host=3Dybcdrdbp01 $DATABASE

 

I also perform a hot backup of this database using = pg_start_backup(), tar, and pg_stop_backup().  It takes only 20 minutes to create a = tar ball of the entire 62GB.  I like the speed of this method, but it does = not allow me to restore 1 table at a time.

 

The version of postgres is PostgreSQL 8.0.0 on i686-pc-linux-gnu, compiled by GCC gcc (GCC) = 3.2.2

 

The machine has 4 Xeon 3.00 GHz processors = with hyper-threading on and 4GB of memory.  Postgres is supported by two = file systems connected to an EMC SAN disk array.  One 2 GB one for the = log files and a second 500 GB one for the data and indexes.  All output = files for the backup files are placed onto the 500 GB volume group and then = backed up to an external storage manager.

 

Portions of the config file = are:

 

shared_buffers =3D = 16384

work_mem =3D = 8192

maintenance_work_mem =3D = 16384

 

max_fsm_pages =3D = 512000

max_fsm_relations =3D 1000

fsync =3D = true

 

# - Checkpoints = -

checkpoint_segments =3D 20

 

# - Planner Cost = Constants -

effective_cache_size =3D 262144

random_page_cost =3D 3

 

 

I am looking for ideas for making the backup of the = above table much faster.

=00 ------_=_NextPart_001_01C59794.14BE874F-- From pgsql-performance-owner@postgresql.org Tue Aug 2 16:59:43 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 67F1E528CC for ; Tue, 2 Aug 2005 16:59:41 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91341-07 for ; Tue, 2 Aug 2005 19:59:30 +0000 (GMT) Received: from mail.nordicbet.com (mail.nordicbet.com [193.69.167.130]) by svr1.postgresql.org (Postfix) with ESMTP id 94A8B52A9E for ; Tue, 2 Aug 2005 16:59:28 -0300 (ADT) Received: from [84.48.51.94] (helo=tobias.nordicbet.invalid) (Authenticated Sender=tobias@nordicbet.com) by mail.nordicbet.com with esmtpa (Exim 4.50 #1 (Debian)) id 1E02tD-0004eD-N5; Tue, 02 Aug 2005 21:57:38 +0200 Received: by tobias.nordicbet.invalid (Postfix, from userid 500) id 5085FE004E; Tue, 2 Aug 2005 21:59:16 +0200 (CEST) Date: Tue, 2 Aug 2005 21:59:15 +0200 From: Tobias Brox To: Tobias Brox Cc: pgsql-performance@postgresql.org Subject: Re: "nice"/low priority Query Message-ID: <20050802195915.GA9691@tobias.lan> References: <20050802160434.GD9278@tobias.lan> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20050802160434.GD9278@tobias.lan> Organization: Group Nordicbet User-Agent: Mutt/1.5.8i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/25 X-Sequence-Number: 13772 [Tobias Brox - Tue at 06:04:34PM +0200] > (...) and at one > point in the code I'm even asking the database for "explain plan", grepping > out the estimated cost number, and referring the user to take contact with > the IT-dept if he really needs the report. But I digress. I just came to think about some more "dirty" tricks I can do. I have turned on stats collection in the configuration; now, if I do: select count(*) from pg_stat_activity where not current_query like '%'; or, eventually: select count(*) from pg_stat_activity where not current_query like '%' and query_start+'1 second'; Wed, 3 Aug 2005 06:02:43 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 76751-05 for ; Wed, 3 Aug 2005 09:02:33 +0000 (GMT) Received: from mail.metronet.co.uk (mail.metronet.co.uk [213.162.97.75]) by svr1.postgresql.org (Postfix) with ESMTP id 6B01052ABA for ; Wed, 3 Aug 2005 06:02:32 -0300 (ADT) Received: from mainbox.archonet.com (84-51-143-99.archon037.adsl.metronet.co.uk [84.51.143.99]) by smtp.metronet.co.uk (MetroNet Mail) with ESMTP id 5A2E140E099; Wed, 3 Aug 2005 10:02:29 +0100 (BST) Received: from localhost (localhost.localdomain [127.0.0.1]) by mainbox.archonet.com (Postfix) with ESMTP id 3028515EDA; Wed, 3 Aug 2005 09:53:33 +0100 (BST) Received: from mainbox.archonet.com ([127.0.0.1]) by localhost (mainbox [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 10207-08; Wed, 3 Aug 2005 09:53:29 +0100 (BST) Received: from [192.168.1.17] (client17.office.archonet.com [192.168.1.17]) by mainbox.archonet.com (Postfix) with ESMTP id A281715ED9; Wed, 3 Aug 2005 09:53:29 +0100 (BST) Message-ID: <42F08609.5060709@archonet.com> Date: Wed, 03 Aug 2005 09:53:29 +0100 From: Richard Huxton User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tobias Brox Cc: pgsql-performance@postgresql.org Subject: Re: "nice"/low priority Query References: <20050802160434.GD9278@tobias.lan> <20050802195915.GA9691@tobias.lan> In-Reply-To: <20050802195915.GA9691@tobias.lan> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/26 X-Sequence-Number: 13773 Tobias Brox wrote: > [Tobias Brox - Tue at 06:04:34PM +0200] > >>(...) and at one >>point in the code I'm even asking the database for "explain plan", grepping >>out the estimated cost number, and referring the user to take contact with >>the IT-dept if he really needs the report. But I digress. > > > I just came to think about some more "dirty" tricks I can do. I have turned > on stats collection in the configuration; now, if I do: > > select count(*) from pg_stat_activity where not current_query like '%'; > > or, eventually: > > select count(*) from pg_stat_activity > where not current_query like '%' and query_start+'1 second' > it will give a hint about how busy the database server is, thus I can > eventually let the application sleep and retry if there are any other heavy > queries in progress. Or - create a table with an estimated_cost column, when you start a new "heavy" query, insert that query's cost, then sleep SUM(estimated_cost)/100 secs or something. When the query ends, delete the cost-row. Hmm - actually rather than dividing by 100, perhaps make it a tunable value. -- Richard Huxton Archonet Ltd From pgsql-performance-owner@postgresql.org Wed Aug 3 09:48:53 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 78C2B52972 for ; Wed, 3 Aug 2005 09:48:46 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 30383-05 for ; Wed, 3 Aug 2005 12:48:35 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.198]) by svr1.postgresql.org (Postfix) with ESMTP id 2796F52808 for ; Wed, 3 Aug 2005 09:48:34 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i8so119919wra for ; Wed, 03 Aug 2005 05:48:38 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:cc:in-reply-to:mime-version:content-type:references; b=UXXQJcEQJk0qKskeAO5hn53xadj7ZwXK1ofiFZXH/xu+0XM4qtEDL7Aca1Gd+hEd+BMqyvbKt0CYQJuHZJE8EteOuvyxYW/5TuSYmgAIj7RCgDaDEM6Q4EK6gVbrgXezwMyxJXUhD9Ti7+2vVsS4eStEw/vMPn+rALnTk3uEfPU= Received: by 10.54.14.29 with SMTP id 29mr579846wrn; Wed, 03 Aug 2005 05:48:38 -0700 (PDT) Received: by 10.54.42.25 with HTTP; Wed, 3 Aug 2005 05:48:38 -0700 (PDT) Message-ID: Date: Wed, 3 Aug 2005 14:48:38 +0200 From: Meetesh Karia Reply-To: meetesh.karia@alumni.duke.edu To: Tom Lane Subject: Re: Planner incorrectly choosing seq scan over index scan Cc: John Arbash Meinel , pgsql-performance@postgresql.org In-Reply-To: Mime-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_Part_247_16072047.1123073318453" References: <42EEAD4B.4010905@arbash-meinel.com> <26292.1122941726@sss.pgh.pa.us> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.177 tagged_above=0 required=5 tests=AWL, HTML_40_50, HTML_MESSAGE, RCVD_BY_IP X-Spam-Level: X-Archive-Number: 200508/27 X-Sequence-Number: 13774 ------=_Part_247_16072047.1123073318453 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Btw - I tried playing around with some of the other planner cost constants= =20 but I wasn't able to get the planner to choose the index scan. It seems lik= e=20 the issue is that the estimated cost for fetching one row from the index ( 3.02) is a little high in my case. Is there any way that I can adjust that= =20 cost estimate? Are there any side effects of doing that? Or is my best=20 solution to simple set enable_hashjoin to off for this query? Thanks, Meetesh On 8/2/05, Meetesh Karia wrote: >=20 > Thanks Tom, >=20 > That modifies the query plan slightly, but the planner still decides to d= o=20 > a hash join for the lte_user table aliased 't'. Though, if I make this=20 > change and set enable_hashjoin to off, the query plan (and execution time= )=20 > gets even better. >=20 > enable_hashjoin =3D on > ---------------------------------- > QUERY PLAN > Sort (cost=3D10113.35..10122.02 rows=3D3467 width=3D48) (actual time=3D > 1203.000..1203.000 rows=3D3467 loops=3D1) > Sort Key: c.sourceid, c.targetid > -> Nested Loop (cost=3D8711.19..9909.50 rows=3D3467 width=3D48) (actual t= ime=3D > 1156.000..1203.000 rows=3D3467 loops=3D1) > -> Index Scan using lte_user_pkey on lte_user s (cost=3D0.00..3.02 rows= =3D1=20 > width=3D16) (actual time=3D0.000..0.000 rows=3D1 loops=3D1) > Index Cond: (617004 =3D user_id) > -> Hash Join (cost=3D8711.19..9776.46 rows=3D3467 width=3D40) (actual tim= e=3D > 1156.000..1187.000 rows=3D3467 loops=3D1) > Hash Cond: ("outer".targetid =3D "inner".user_id) > -> Seq Scan on candidates617004 c (cost=3D0.00..76.34 rows=3D3467 width= =3D32)=20 > (actual time=3D0.000..16.000 rows=3D3467 loops=3D1) > Filter: (sourceid =3D 617004) > -> Hash (cost=3D8012.55..8012.55 rows=3D279455 width=3D16) (actual time= =3D > 1141.000..1141.000 rows=3D0 loops=3D1) > -> Seq Scan on lte_user t (cost=3D0.00..8012.55 rows=3D279455 width=3D16)= =20 > (actual time=3D0.000..720.000 rows=3D279395 loops=3D1) > Total runtime: 1218.000 ms >=20 > enable_hashjoin =3D off > ----------------------------------- > QUERY PLAN > Sort (cost=3D10942.56..10951.22 rows=3D3467 width=3D48) (actual time=3D > 188.000..188.000 rows=3D3467 loops=3D1) > Sort Key: c.sourceid, c.targetid > -> Nested Loop (cost=3D0.00..10738.71 rows=3D3467 width=3D48) (actual tim= e=3D > 0.000..188.000 rows=3D3467 loops=3D1) > -> Index Scan using lte_user_pkey on lte_user s (cost=3D0.00..3.02 rows= =3D1=20 > width=3D16) (actual time=3D0.000..0.000 rows=3D1 loops=3D1) > Index Cond: (617004 =3D user_id) > -> Nested Loop (cost=3D0.00..10605.67 rows=3D3467 width=3D40) (actual tim= e=3D > 0.000..157.000 rows=3D3467 loops=3D1) > -> Seq Scan on candidates617004 c (cost=3D0.00..76.34 rows=3D3467 width= =3D32)=20 > (actual time=3D0.000..15.000 rows=3D3467 loops=3D1) > Filter: (sourceid =3D 617004) > -> Index Scan using lte_user_pkey on lte_user t (cost=3D0.00..3.02 rows= =3D1=20 > width=3D16) (actual time=3D0.028..0.037 rows=3D1 loops=3D3467) > Index Cond: ("outer".targetid =3D t.user_id) > Total runtime: 188.000 ms >=20 > Thanks, > Meetesh >=20 > On 8/2/05, Tom Lane wrote: > >=20 > > Meetesh Karia writes: > > > Sure. The lte_user table is just a collection of users. user_id is=20 > > assigned=3D > > > uniquely using a sequence. During some processing, we create a=20 > > candidates=3D=20 > > > table (candidates617004 in our case). This table is usually a temp=20 > > table.=3D > > > sourceid is a user_id (in this case it is always 617004) and targetid= =20 > > is=3D20 > > > also a user_id (2860 distinct values out of 3467). The rest of the=3D= 20=20 > > > information is either only used in the select clause or not used at= =20 > > all=3D20 > > > during this processing. > >=20 > > If you know that sourceid has only a single value, it'd probably be > > helpful to call out that value in the query, ie,=20 > > where ... AND c.sourceId =3D 617004 ... > >=20 > > regards, tom lane > >=20 >=20 > ------=_Part_247_16072047.1123073318453 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Btw - I tried playing around with some of the other planner cost constants but I wasn't able to get the planner to choose the index scan.  It seems like the issue is that the estimated cost for fetching one row from the index (3.02) is a little high in my case.  Is there any way that I can adjust that cost estimate?  Are there any side effects of doing that?  Or is my best solution to simple set enable_hashjoin to off for this query?

Thanks,
Meetesh

On 8/2/05, Meetesh Karia <meetesh.karia@gmail.com> wrote:
Thanks Tom,

That modifies the query plan slightly, but the planner still decides to do a hash join for the lte_user table aliased 't'.  Though, if I make this change and set enable_hashjoin to off, the query plan (and execution time) gets even better.

enable_hashjoin =3D on
----------------------------------
QUERY PLAN
Sort  (cost=3D10113.35..10122.02 rows=3D3467 width=3D48) (actual time= =3D1203.000..1203.000 rows=3D3467 loops=3D1)
  Sort Key: c.sourceid, c.targetid
   ->  Nested Loop  (cost=3D8711.19..9909.50 rows=3D3467 width=3D48) (actual time=3D1156.000..1203.000 rows=3D3467 loops=3D1)
        ->  Index Scan using lte_user_pkey on lte_user s  (cost=3D0.00..3.02 rows=3D1 width=3D16) (actual time=3D0.000..0.000 rows=3D1 loops=3D1)
            &nb= sp; Index Cond: (617004 =3D user_id)
        ->  Hash Join  (cost=3D8711.19..9776.46 rows=3D3467 width=3D40) (actual time=3D1156.000..1187.000 rows=3D3467 loops=3D1)
            &nb= sp; Hash Cond: ("outer".targetid =3D "inner".user_id)             &nb= sp; ->  Seq Scan on candidates617004 c  (cost=3D0.00..76.34 rows=3D3467 width=3D32) (actual time=3D0.000..16.000 rows=3D3467 loops=3D1)=
            &nb= sp;       Filter: (sourceid =3D 617004)
            &nb= sp; ->  Hash  (cost=3D8012.55..8012.55 rows=3D279455 width=3D16) (actual time=3D1141.000..1141.000 rows=3D0 loops=3D1)
            &nb= sp;       ->  Seq Scan on lte_user t  (cost=3D0.00..8012.55 rows=3D279455 width=3D16) (actual time=3D0.000..720.000 rows=3D279395 loops= =3D1)
Total runtime: 1218.000 ms

enable_hashjoin =3D off
-----------------------------------
QUERY PLAN
Sort  (cost=3D10942.56..10951.22 rows=3D3467 width=3D48) (actual time= =3D188.000..188.000 rows=3D3467 loops=3D1)
  Sort Key: c.sourceid, c.targetid
   ->  Nested Loop  (cost=3D0.00..10738.71 rows=3D3467 wid= th=3D48) (actual time=3D0.000..188.000 rows=3D3467 loops=3D1)
        ->  Index Scan using lte_user_pkey on lte_user s  (cost=3D0.00..3.02 rows=3D1 width=3D16) (actual time=3D0.000..0.000 rows=3D1 loops=3D1)
            &nb= sp; Index Cond: (617004 =3D user_id)
        ->  Nested Loop  (cost=3D0.00..10605.67 rows=3D3467 width=3D40) (actual time=3D0.000..157.000 rows=3D3467 loops=3D1)
            &nb= sp; ->  Seq Scan on candidates617004 c  (cost=3D0.00..76.34 rows=3D3467 width=3D32) (actual time=3D0.000..15.000 rows=3D3467 loops=3D1)=
            &nb= sp;       Filter: (sourceid =3D 617004)
            &nb= sp; ->  Index Scan using lte_user_pkey on lte_user t  (cost=3D0.00..3.02 rows=3D1 width=3D16) (actual time=3D0.028..0.037 rows=3D= 1 loops=3D3467)
            &nb= sp;       Index Cond: ("outer".targetid =3D t.user_id)
 Total runtime: 188.000 ms

Thanks,
Meetesh


<= div>On 8/2/05, To= m Lane < tgl@sss.pgh.pa.us> wrote:
Meetesh Karia <meetesh.karia@= gmail.com> writes:
> Sure. The lte_user table is just a collec= tion of users. user_id is assigned=3D
> uniquely using a sequence. During some processing, we create a can= didates=3D
> table (candidates617004 in our case). This table is usually a temp= table.=3D
> sourceid is a user_id (in this case it is always 617004)= and targetid is=3D20
> also a user_id (2860 distinct values out of 3= 467). The rest of the=3D20
> information is either only used in the select clause or not used a= t all=3D20
> during this processing.

If you know that sourceid= has only a single value, it'd probably be
helpful to call out that valu= e in the query, ie,
        where ... AND c.sourceI= d =3D 617004 ...

        &nb= sp;            =    regards, tom lane


------=_Part_247_16072047.1123073318453-- From pgsql-performance-owner@postgresql.org Wed Aug 3 10:12:16 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 745E652851 for ; Wed, 3 Aug 2005 10:12:15 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 37238-04 for ; Wed, 3 Aug 2005 13:12:04 +0000 (GMT) Received: from TWMAIL.ESNCC.COM (static-66-173-159-28.t1.cavtel.net [66.173.159.28]) by svr1.postgresql.org (Postfix) with ESMTP id 9C91352824 for ; Wed, 3 Aug 2005 10:12:02 -0300 (ADT) Received: from spawar2i8uvlb9 ([150.125.117.63]) by TWMAIL.ESNCC.COM with Microsoft SMTPSVC(5.0.2195.6713); Wed, 3 Aug 2005 09:12:05 -0400 From: "Lane Van Ingen" To: Subject: Is There A Windows Version of Performance Tuning Documents? Date: Wed, 3 Aug 2005 09:15:34 -0400 Message-ID: MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook IMO, Build 9.0.2416 (9.0.2911.0) X-MimeOLE: Produced By Microsoft MimeOLE V6.00.3790.181 Importance: Normal X-OriginalArrivalTime: 03 Aug 2005 13:12:05.0409 (UTC) FILETIME=[F1E2E910:01C5982C] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 tagged_above=0 required=5 tests=FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/28 X-Sequence-Number: 13775 I have in my possession some performance tuning documents authored by Bruce Momjian, Josh Berkus, and others. They give good information on utilities to use (like ipcs, sar, vmstat, etc) to evaluate disk, memory, etc. performance on Unix-based systems. Problem is, I have applications running on Windows 2003, and have worked mostly on Unix before. Was wondering if anyone knows where there might be a Windows performance document that tells what to use / where to look in Windows for some of this data. I am thinking that I may not seeing what I need in perfmon or the Windows task manager. Want to answer questions like: How much memory is being used for disk buffer cache? How to I lock shared memory for PostgreSQL (if possible at all)? How to determine if SWAP (esp. page-in) activity is hurting me? Does Windows use a 'unified buffer cache' or not? How do I determine how much space is required to do most of my sorts in RAM? From pgsql-performance-owner@postgresql.org Wed Aug 3 12:08:09 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 29E1B528C2 for ; Wed, 3 Aug 2005 12:08:08 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83451-10 for ; Wed, 3 Aug 2005 15:08:05 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id 308FB528B0 for ; Wed, 3 Aug 2005 12:08:03 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050803150800m9100g65v6e>; Wed, 3 Aug 2005 15:08:00 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 382C556051; Wed, 3 Aug 2005 10:07:58 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id 8B65956016; Wed, 3 Aug 2005 10:07:47 -0500 (CDT) Message-ID: <42F0DDC3.9080707@arbash-meinel.com> Date: Wed, 03 Aug 2005 10:07:47 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Lane Van Ingen , Postgresql Performance Subject: Re: Is There A Windows Version of Performance Tuning Documents? References: In-Reply-To: X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig6F8271126F508DC29657ACEF" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.045 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/29 X-Sequence-Number: 13776 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig6F8271126F508DC29657ACEF Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Lane Van Ingen wrote: > I have in my possession some performance tuning documents authored by Bruce > Momjian, Josh Berkus, and others. They give good information on utilities to > use (like ipcs, sar, vmstat, etc) to evaluate disk, memory, etc. performance > on Unix-based systems. > > Problem is, I have applications running on Windows 2003, and have worked > mostly on Unix before. Was wondering if anyone knows where there might be a > Windows performance document that tells what to use / where to look in > Windows for some of this data. I am thinking that I may not seeing what I > need > in perfmon or the Windows task manager. > > Want to answer questions like: > How much memory is being used for disk buffer cache? > How to I lock shared memory for PostgreSQL (if possible at all)? > How to determine if SWAP (esp. page-in) activity is hurting me? > Does Windows use a 'unified buffer cache' or not? > How do I determine how much space is required to do most of my sorts in > RAM? > I don't know of any specific documentation. I would mention the TaskManager as the first place I would look (Ctrl+Shift+Esc, or right click on the task bar). You can customize the columns that it shows in the process view, so you can get an idea if something is paging, how much I/O it is using, etc. I'm sure there are other better tools, but this one is pretty easy to get to, and shows quite a bit. John =:-> --------------enig6F8271126F508DC29657ACEF Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC8N3DJdeBCYSNAAMRAiVuAJ47FwYtwDCfkfVPiqkUf7YLH56zLwCePHMz HYvS+fdAtzmqNRkCSjMtx4U= =ffBB -----END PGP SIGNATURE----- --------------enig6F8271126F508DC29657ACEF-- From pgsql-performance-owner@postgresql.org Thu Aug 4 05:14:17 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0534852840 for ; Thu, 4 Aug 2005 05:14:16 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 63444-01 for ; Thu, 4 Aug 2005 08:14:04 +0000 (GMT) Received: from larry.aptalaska.net (larry.aptalaska.net [64.186.96.3]) by svr1.postgresql.org (Postfix) with ESMTP id 7638252BB0 for ; Thu, 4 Aug 2005 05:14:01 -0300 (ADT) Received: from [192.168.98.9] (rdbck-static-287.palmer.mtaonline.net [64.4.232.33]) (authenticated bits=0) by larry.aptalaska.net (8.13.4/8.13.4) with ESMTP id j748DwPo010731 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO) for ; Thu, 4 Aug 2005 00:13:59 -0800 Message-ID: <42F1CE43.8040801@aptalaska.net> Date: Thu, 04 Aug 2005 00:13:55 -0800 From: Matthew Schumacher User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 References: <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> <42EE93F8.5010003@aptalaska.net> <20050801214931.GF60019@decibel.org> <42EE9D83.8000405@pobox.com> In-Reply-To: <42EE9D83.8000405@pobox.com> X-Enigmail-Version: 0.91.0.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit X-Spam-Bayes: Learn: ham X-Scanned-By: MIMEDefang 2.52 on 64.186.96.3 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/30 X-Sequence-Number: 13777 Okay, Here is the status of the SA updates and a question: Michael got SA changed to pass an array of tokens to the proc so right there we gained a ton of performance due to connections and transactions being grouped into one per email instead of one per token. Now I am working on making the proc even faster. Since we have all of the tokens coming in as an array, it should be possible to get this down to just a couple of queries. I have the proc using IN and NOT IN statements to update everything at once from a temp table, but it progressively gets slower because the temp table is growing between vacuums. At this point it's slightly slower than the old update or else insert on every token. What I really want to do is have the token array available as a record so that I can query against it, but not have it take up the resources of a real table. If I could copy from an array into a record then I can even get rid of the loop. Anyone have any thoughts on how to do this? CREATE OR REPLACE FUNCTION put_tokens(inuserid INTEGER, intokenary BYTEA[], inspam_count INTEGER, inham_count INTEGER, inatime INTEGER) RETURNS VOID AS ' DECLARE _token BYTEA; BEGIN for i in array_lower(intokenary, 1) .. array_upper(intokenary, 1) LOOP _token := intokenary[i]; INSERT INTO bayes_token_tmp VALUES (_token); END LOOP; UPDATE bayes_token SET spam_count = greatest_int(spam_count + inspam_count, 0), ham_count = greatest_int(ham_count + inham_count , 0), atime = greatest_int(atime, 1000) WHERE id = inuserid AND (token) IN (SELECT intoken FROM bayes_token_tmp); UPDATE bayes_vars SET token_count = token_count + (SELECT count(intoken) FROM bayes_token_tmp WHERE intoken NOT IN (SELECT token FROM bayes_token)), newest_token_age = greatest_int(newest_token_age, inatime), oldest_token_age = least_int(oldest_token_age, inatime) WHERE id = inuserid; INSERT INTO bayes_token SELECT inuserid, intoken, inspam_count, inham_count, inatime FROM bayes_token_tmp WHERE (inspam_count > 0 OR inham_count > 0) AND (intoken) NOT IN (SELECT token FROM bayes_token); delete from bayes_token_tmp; RETURN; END; ' LANGUAGE 'plpgsql'; CREATE OR REPLACE FUNCTION greatest_int (integer, integer) RETURNS INTEGER IMMUTABLE STRICT AS 'SELECT CASE WHEN $1 < $2 THEN $2 ELSE $1 END;' LANGUAGE SQL; CREATE OR REPLACE FUNCTION least_int (integer, integer) RETURNS INTEGER IMMUTABLE STRICT AS 'SELECT CASE WHEN $1 < $2 THEN $1 ELSE $2 END;' LANGUAGE SQL; From pgsql-performance-owner@postgresql.org Thu Aug 4 05:24:43 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6F8A552BE8 for ; Thu, 4 Aug 2005 05:24:42 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 62034-06 for ; Thu, 4 Aug 2005 08:24:33 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.196]) by svr1.postgresql.org (Postfix) with ESMTP id A49AC528D7 for ; Thu, 4 Aug 2005 05:24:30 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id 70so79467wra for ; Thu, 04 Aug 2005 01:24:32 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:mime-version:content-type; b=W7LoDuRvU8OEs7sYgpSMx0el88e9u/sOKm6XoBHa6Y100MNBiGMYWRWVmtCyJJ4hZXpXFy7B68pWub+po8+Go4adx8Q5bbcJTLwLQ5mDtR0BWntd5P7864J6lbzPJqKLC1s4sPLmXjmZIzjkuJs9GkyMJ6TEVwvFV9DfakjWK/o= Received: by 10.54.7.21 with SMTP id 21mr1301386wrg; Thu, 04 Aug 2005 01:24:32 -0700 (PDT) Received: by 10.54.111.19 with HTTP; Thu, 4 Aug 2005 01:24:32 -0700 (PDT) Message-ID: Date: Thu, 4 Aug 2005 13:54:32 +0530 From: prasanna s Reply-To: prasanna s To: postgres Subject: Indexed views. Mime-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_Part_3202_21493592.1123143872388" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.033 tagged_above=0 required=5 tests=HTML_10_20, HTML_MESSAGE, HTML_SHORT_LENGTH, RCVD_BY_IP X-Spam-Level: * X-Archive-Number: 200508/31 X-Sequence-Number: 13778 ------=_Part_3202_21493592.1123143872388 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Does postgres support indexed views/materialised views that some of the=20 other databases support? Thanks Prasanna S ------=_Part_3202_21493592.1123143872388 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Does postgres support indexed views/materialised views that some of the oth= er databases support?

Thanks
Prasanna S
------=_Part_3202_21493592.1123143872388-- From pgsql-performance-owner@postgresql.org Thu Aug 4 05:34:32 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B164252C8A for ; Thu, 4 Aug 2005 05:34:31 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 67774-01 for ; Thu, 4 Aug 2005 08:34:30 +0000 (GMT) Received: from houston.familyhealth.com.au (houston.au.fhnetwork.com [203.22.197.21]) by svr1.postgresql.org (Postfix) with ESMTP id CA21B52C86 for ; Thu, 4 Aug 2005 05:34:28 -0300 (ADT) Received: from houston.familyhealth.com.au (localhost [127.0.0.1]) by houston.familyhealth.com.au (Postfix) with ESMTP id 2978224FF1; Thu, 4 Aug 2005 16:34:26 +0800 (WST) Received: from [127.0.0.1] (work-40.internal [192.168.0.40]) by houston.familyhealth.com.au (Postfix) with ESMTP id E81F024FE5; Thu, 4 Aug 2005 16:34:22 +0800 (WST) Message-ID: <42F1D3BA.1070107@familyhealth.com.au> Date: Thu, 04 Aug 2005 16:37:14 +0800 From: Christopher Kings-Lynne User-Agent: Mozilla Thunderbird 1.0.6 (Windows/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: prasanna s Cc: postgres Subject: Re: Indexed views. References: In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-familyhealth-MailScanner-Information: Please contact the ISP for more information X-familyhealth-MailScanner: Found to be clean X-familyhealth-MailScanner-From: chriskl@familyhealth.com.au X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.062 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/32 X-Sequence-Number: 13779 No, unless you use some custom triggers. prasanna s wrote: > Does postgres support indexed views/materialised views that some of the > other databases support? > > Thanks > Prasanna S From pgsql-performance-owner@postgresql.org Thu Aug 4 05:45:50 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6DC4852BB0 for ; Thu, 4 Aug 2005 05:45:48 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 67952-06 for ; Thu, 4 Aug 2005 08:45:42 +0000 (GMT) Received: from boutiquenumerique.com (boutiquenumerique.com [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id BC5A052B01 for ; Thu, 4 Aug 2005 05:45:40 -0300 (ADT) Received: (qmail 11009 invoked from network); 4 Aug 2005 10:45:51 +0200 Received: from unknown (HELO localhost) (boutiquenumerique-lists@192.168.0.4) by boutiquenumerique.com with SMTP; 4 Aug 2005 10:45:51 +0200 To: "Matthew Schumacher" , pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 References: <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> <42EE93F8.5010003@aptalaska.net> <20050801214931.GF60019@decibel.org> <42EE9D83.8000405@pobox.com> <42F1CE43.8040801@aptalaska.net> Message-ID: Date: Thu, 04 Aug 2005 10:45:40 +0200 From: PFC Organization: =?iso-8859-15?Q?La_Boutique_Num=E9rique?= Content-Type: text/plain; format=flowed; delsp=yes; charset=iso-8859-15 MIME-Version: 1.0 Content-Transfer-Encoding: 8bit In-Reply-To: <42F1CE43.8040801@aptalaska.net> User-Agent: Opera M2/8.0 (Linux, build 1095) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/33 X-Sequence-Number: 13780 > What I really want to do is have the token array available as a record > so that I can query against it, but not have it take up the resources of > a real table. If I could copy from an array into a record then I can > even get rid of the loop. Anyone have any thoughts on how to do this? You could make a set-returning-function (about 3 lines) which RETURNs NEXT every element in the array ; then you can use this SRF just like a table and SELECT from it, join it with your other tables, etc. From pgsql-performance-owner@postgresql.org Thu Aug 4 05:52:48 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B4BBE52B04 for ; Thu, 4 Aug 2005 05:52:46 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 68535-09 for ; Thu, 4 Aug 2005 08:52:35 +0000 (GMT) Received: from epbsa003.epam.com (epbsa003.epam.com [195.56.119.212]) by svr1.postgresql.org (Postfix) with ESMTP id D5A4552901 for ; Thu, 4 Aug 2005 05:52:33 -0300 (ADT) Received: from [127.0.0.1] ([10.0.11.147]) by epbsa003.epam.com with Microsoft SMTPSVC(6.0.3790.1830); Thu, 4 Aug 2005 10:52:33 +0200 Message-ID: <42F1D739.2090309@forgeahead.hu> Date: Thu, 04 Aug 2005 10:52:09 +0200 From: Laszlo Hornyak User-Agent: Mozilla Thunderbird 1.0.2 (Windows/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Re: Indexed views. References: In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-OriginalArrivalTime: 04 Aug 2005 08:52:33.0899 (UTC) FILETIME=[DAF4CBB0:01C598D1] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/34 X-Sequence-Number: 13781 prasanna s wrote: > Does postgres support indexed views/materialised views that some of > the other databases support? > > Thanks > Prasanna S Hi! It is not supported, but perhaps this will help you: http://www.varlena.com/varlena/GeneralBits/Tidbits/matviews.html From pgsql-performance-owner@postgresql.org Thu Aug 4 11:10:33 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2D61952C7D for ; Thu, 4 Aug 2005 11:10:23 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 45969-03 for ; Thu, 4 Aug 2005 14:10:21 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id E0F0F52C92 for ; Thu, 4 Aug 2005 11:10:15 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050804141015m9100g5scfe>; Thu, 4 Aug 2005 14:10:19 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id C58CE56051; Thu, 4 Aug 2005 09:10:14 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id 6BE7856016; Thu, 4 Aug 2005 09:10:07 -0500 (CDT) Message-ID: <42F221BB.5000702@arbash-meinel.com> Date: Thu, 04 Aug 2005 09:10:03 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Matthew Schumacher Cc: pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 References: <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> <42EE93F8.5010003@aptalaska.net> <20050801214931.GF60019@decibel.org> <42EE9D83.8000405@pobox.com> <42F1CE43.8040801@aptalaska.net> In-Reply-To: <42F1CE43.8040801@aptalaska.net> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig2E83272BA840438B90498DA8" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.045 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/35 X-Sequence-Number: 13782 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig2E83272BA840438B90498DA8 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Matthew Schumacher wrote: > Okay, > > Here is the status of the SA updates and a question: > > Michael got SA changed to pass an array of tokens to the proc so right > there we gained a ton of performance due to connections and transactions > being grouped into one per email instead of one per token. > > Now I am working on making the proc even faster. Since we have all of > the tokens coming in as an array, it should be possible to get this down > to just a couple of queries. > > I have the proc using IN and NOT IN statements to update everything at > once from a temp table, but it progressively gets slower because the > temp table is growing between vacuums. At this point it's slightly > slower than the old update or else insert on every token. I recommend that you drop and re-create the temp table. There is no reason to have it around, considering you delete and re-add everything. That means you never have to vacuum it, since it always only contains the latest rows. > > What I really want to do is have the token array available as a record > so that I can query against it, but not have it take up the resources of > a real table. If I could copy from an array into a record then I can > even get rid of the loop. Anyone have any thoughts on how to do this? > My one question here, is the inspam_count and inham_count *always* the same for all tokens? I would have thought each token has it's own count. Anyway, there are a few lines I would change: > > CREATE OR REPLACE FUNCTION put_tokens(inuserid INTEGER, > intokenary BYTEA[], > inspam_count INTEGER, > inham_count INTEGER, > inatime INTEGER) > RETURNS VOID AS ' > DECLARE > _token BYTEA; > BEGIN > -- create the table at the start of the procedure CREATE TEMP TABLE bayes_token_tmp (intoken bytea); -- You might also add primary key if you are going to be adding -- *lots* of entries, but it sounds like you are going to have -- less than 1 page, so it doesn't matter > for i in array_lower(intokenary, 1) .. array_upper(intokenary, 1) > LOOP > _token := intokenary[i]; > INSERT INTO bayes_token_tmp VALUES (_token); > END LOOP; > > UPDATE > bayes_token > SET > spam_count = greatest_int(spam_count + inspam_count, 0), > ham_count = greatest_int(ham_count + inham_count , 0), > atime = greatest_int(atime, 1000) > WHERE > id = inuserid > AND -- (token) IN (SELECT intoken FROM bayes_token_tmp); EXISTS (SELECT token FROM bayes_token_tmp WHERE intoken=token LIMIT 1); -- I would also avoid your intoken (NOT) IN (SELECT token FROM -- bayes_token) There are a few possibilities, but to me -- as your bayes_token table becomes big, this will start -- to be the slow point -- Rather than doing 2 NOT IN queries, it *might* be faster to do DELETE FROM bayes_token_tmp WHERE NOT EXISTS (SELECT token FROM bayes_token WHERE token=intoken); > > UPDATE > bayes_vars > SET -- token_count = token_count + (SELECT count(intoken) FROM -- bayes_token_tmp WHERE intoken NOT IN (SELECT token FROM bayes_token)), token_count = token_count + (SELECT count(intoken) FROM bayes_token_tmp) -- You don't need the where NOT IN, since we already removed those rows > newest_token_age = greatest_int(newest_token_age, inatime), > oldest_token_age = least_int(oldest_token_age, inatime) > WHERE > id = inuserid; > > INSERT INTO > bayes_token > SELECT > inuserid, > intoken, > inspam_count, > inham_count, > inatime > FROM > bayes_token_tmp > WHERE > (inspam_count > 0 OR inham_count > 0) -- AND -- (intoken) NOT IN (SELECT token FROM bayes_token); -- You don't need either of those lines, again because we already -- filtered -- delete from bayes_token_tmp; -- And rather than deleting all of the entries just DROP TABLE bayes_token_tmp; > > RETURN; > END; > ' LANGUAGE 'plpgsql'; > > CREATE OR REPLACE FUNCTION greatest_int (integer, integer) > RETURNS INTEGER > IMMUTABLE STRICT > AS 'SELECT CASE WHEN $1 < $2 THEN $2 ELSE $1 END;' > LANGUAGE SQL; > > CREATE OR REPLACE FUNCTION least_int (integer, integer) > RETURNS INTEGER > IMMUTABLE STRICT > AS 'SELECT CASE WHEN $1 < $2 THEN $1 ELSE $2 END;' > LANGUAGE SQL; > So to clarify, here is my finished function: ------------------------------------ CREATE OR REPLACE FUNCTION put_tokens(inuserid INTEGER, intokenary BYTEA[], inspam_count INTEGER, inham_count INTEGER, inatime INTEGER) RETURNS VOID AS ' DECLARE _token BYTEA; BEGIN CREATE TEMP TABLE bayes_token_tmp (intoken bytea); for i in array_lower(intokenary, 1) .. array_upper(intokenary, 1) LOOP _token := intokenary[i]; INSERT INTO bayes_token_tmp VALUES (_token); END LOOP; UPDATE bayes_token SET spam_count = greatest_int(spam_count + inspam_count, 0), ham_count = greatest_int(ham_count + inham_count , 0), atime = greatest_int(atime, 1000) WHERE id = inuserid AND EXISTS (SELECT token FROM bayes_token_tmp WHERE intoken=token LIMIT 1); DELETE FROM bayes_token_tmp WHERE NOT EXISTS (SELECT token FROM bayes_token WHERE token=intoken); UPDATE bayes_vars SET token_count = token_count + (SELECT count(intoken) FROM bayes_token_tmp), newest_token_age = greatest_int(newest_token_age, inatime), oldest_token_age = least_int(oldest_token_age, inatime) WHERE id = inuserid; INSERT INTO bayes_token SELECT inuserid, intoken, inspam_count, inham_count, inatime FROM bayes_token_tmp WHERE (inspam_count > 0 OR inham_count > 0) DROP TABLE bayes_token_tmp; RETURN; END; ' LANGUAGE 'plpgsql'; --------------enig2E83272BA840438B90498DA8 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC8iG+JdeBCYSNAAMRAuQpAJ9tr8Idt+akJTjmymh9QKw6ZRzr5QCglUC/ Mm4izKm+0Gcl10WMNDY8Dmw= =YH+y -----END PGP SIGNATURE----- --------------enig2E83272BA840438B90498DA8-- From pgsql-performance-owner@postgresql.org Thu Aug 4 11:37:12 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D754552B17 for ; Thu, 4 Aug 2005 11:37:09 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 51941-07 for ; Thu, 4 Aug 2005 14:37:02 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id AC8765295A for ; Thu, 4 Aug 2005 11:37:01 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j74Eb362015611; Thu, 4 Aug 2005 10:37:03 -0400 (EDT) To: Matthew Schumacher Cc: pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 In-reply-to: <42F1CE43.8040801@aptalaska.net> References: <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> <42EE93F8.5010003@aptalaska.net> <20050801214931.GF60019@decibel.org> <42EE9D83.8000405@pobox.com> <42F1CE43.8040801@aptalaska.net> Comments: In-reply-to Matthew Schumacher message dated "Thu, 04 Aug 2005 00:13:55 -0800" Date: Thu, 04 Aug 2005 10:37:02 -0400 Message-ID: <15610.1123166222@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/36 X-Sequence-Number: 13783 Matthew Schumacher writes: > for i in array_lower(intokenary, 1) .. array_upper(intokenary, 1) > LOOP > _token := intokenary[i]; > INSERT INTO bayes_token_tmp VALUES (_token); > END LOOP; > UPDATE > bayes_token > SET > spam_count = greatest_int(spam_count + inspam_count, 0), > ham_count = greatest_int(ham_count + inham_count , 0), > atime = greatest_int(atime, 1000) > WHERE > id = inuserid > AND > (token) IN (SELECT intoken FROM bayes_token_tmp); I don't really see why you think that this path is going to lead to better performance than where you were before. Manipulation of the temp table is never going to be free, and IN (sub-select) is always inherently not fast, and NOT IN (sub-select) is always inherently awful. Throwing a pile of simple queries at the problem is not necessarily the wrong way ... especially when you are doing it in plpgsql, because you've already eliminated the overhead of network round trips and repeated planning of the queries. regards, tom lane From pgsql-performance-owner@postgresql.org Thu Aug 4 12:08:38 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E46C0528C2 for ; Thu, 4 Aug 2005 12:08:35 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 61577-10 for ; Thu, 4 Aug 2005 15:08:24 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id A04AA5295A for ; Thu, 4 Aug 2005 12:08:22 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050804150821m9200an2ave>; Thu, 4 Aug 2005 15:08:21 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 09A2C56051; Thu, 4 Aug 2005 10:08:21 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id 9A76956016; Thu, 4 Aug 2005 10:08:13 -0500 (CDT) Message-ID: <42F22F59.7090901@arbash-meinel.com> Date: Thu, 04 Aug 2005 10:08:09 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: Matthew Schumacher , pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 References: <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> <42EE93F8.5010003@aptalaska.net> <20050801214931.GF60019@decibel.org> <42EE9D83.8000405@pobox.com> <42F1CE43.8040801@aptalaska.net> <15610.1123166222@sss.pgh.pa.us> In-Reply-To: <15610.1123166222@sss.pgh.pa.us> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig7354D6CDCD06455821ED204B" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.045 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/37 X-Sequence-Number: 13784 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig7354D6CDCD06455821ED204B Content-Type: multipart/mixed; boundary="------------030100080708090409060609" This is a multi-part message in MIME format. --------------030100080708090409060609 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Tom Lane wrote: > Matthew Schumacher writes: > >> for i in array_lower(intokenary, 1) .. array_upper(intokenary, 1) >> LOOP >> _token := intokenary[i]; >> INSERT INTO bayes_token_tmp VALUES (_token); >> END LOOP; > > >> UPDATE >> bayes_token >> SET >> spam_count = greatest_int(spam_count + inspam_count, 0), >> ham_count = greatest_int(ham_count + inham_count , 0), >> atime = greatest_int(atime, 1000) >> WHERE >> id = inuserid >> AND >> (token) IN (SELECT intoken FROM bayes_token_tmp); > > > I don't really see why you think that this path is going to lead to > better performance than where you were before. Manipulation of the > temp table is never going to be free, and IN (sub-select) is always > inherently not fast, and NOT IN (sub-select) is always inherently > awful. Throwing a pile of simple queries at the problem is not > necessarily the wrong way ... especially when you are doing it in > plpgsql, because you've already eliminated the overhead of network > round trips and repeated planning of the queries. So for an IN (sub-select), does it actually pull all of the rows from the other table, or is the planner smart enough to stop once it finds something? Is IN (sub-select) about the same as EXISTS (sub-select WHERE x=y)? What about NOT IN (sub-select) versus NOT EXISTS (sub-select WHERE x=y) I would guess that the EXISTS/NOT EXISTS would be faster, though it probably would necessitate using a nested loop (at least that seems to be the way the query is written). I did some tests on a database with 800k rows, versus a temp table with 2k rows. I did one sequential test (1-2000, with 66 rows missing), and one sparse test (1-200, 100000-100200, 200000-200200, ... with 658 rows missing). If found that NOT IN did indeed have to load the whole table. IN was smart enough to do a nested loop. EXISTS and NOT EXISTS did a sequential scan on my temp table, with a SubPlan filter (which looks a whole lot like a Nested Loop). What I found was that IN performed about the same as EXISTS (since they are both effectively doing a nested loop), but that NOT IN took 4,000ms while NOT EXISTS was the same speed as EXISTS at around 166ms. Anyway, so it does seem like NOT IN is not a good choice, but IN seems to be equivalent to EXISTS, and NOT EXISTS is also very fast. Is this generally true, or did I just get lucky on my data? John =:-> > > regards, tom lane > > ---------------------------(end of broadcast)--------------------------- > TIP 3: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faq > --------------030100080708090409060609 Content-Type: text/plain; x-mac-type="0"; x-mac-creator="0"; name="random_exists.txt" Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="random_exists.txt" mifar07=# EXPLAIN ANALYZE SELECT id FROM ids WHERE id IN (SELECT id FROM object_t); QUERY PLAN --------------------------------------------------------------------------------------------------------------------------------- Nested Loop IN Join (cost=0.00..9851.68 rows=2140 width=4) (actual time=0.085..183.889 rows=1351 loops=1) -> Seq Scan on ids (cost=0.00..31.40 rows=2140 width=4) (actual time=0.014..24.032 rows=2009 loops=1) -> Index Scan using object_t_pkey on object_t (cost=0.00..4.58 rows=1 width=4) (actual time=0.071..0.071 rows=1 loops=2009) Index Cond: ("outer".id = object_t.id) Total runtime: 184.823 ms (5 rows) Time: 186.931 ms mifar07=# EXPLAIN ANALYZE SELECT id FROM ids WHERE EXISTS (SELECT id FROM object_t o WHERE o.id = ids.id); QUERY PLAN ------------------------------------------------------------------------------------------------------------------------------------- Seq Scan on ids (cost=0.00..9824.93 rows=1070 width=4) (actual time=0.086..165.053 rows=1351 loops=1) Filter: (subplan) SubPlan -> Index Scan using object_t_pkey on object_t o (cost=0.00..4.58 rows=1 width=4) (actual time=0.025..0.025 rows=1 loops=2009) Index Cond: (id = $0) Total runtime: 165.995 ms (6 rows) Time: 167.795 ms mifar07=# EXPLAIN ANALYZE SELECT id FROM ids WHERE id NOT IN (SELECT id FROM object_t); QUERY PLAN --------------------------------------------------------------------------------------------------------------------------- Seq Scan on ids (cost=36410.51..36447.26 rows=1070 width=4) (actual time=4168.247..4172.080 rows=658 loops=1) Filter: (NOT (hashed subplan)) SubPlan -> Seq Scan on object_t (cost=0.00..34381.81 rows=811481 width=4) (actual time=0.044..2464.296 rows=811481 loops=1) Total runtime: 4210.784 ms (5 rows) Time: 4212.276 ms mifar07=# EXPLAIN ANALYZE SELECT id FROM ids WHERE NOT EXISTS (SELECT id FROM object_t o WHERE o.id = ids.id); QUERY PLAN ------------------------------------------------------------------------------------------------------------------------------------- Seq Scan on ids (cost=0.00..9824.93 rows=1070 width=4) (actual time=0.372..164.510 rows=658 loops=1) Filter: (NOT (subplan)) SubPlan -> Index Scan using object_t_pkey on object_t o (cost=0.00..4.58 rows=1 width=4) (actual time=0.064..0.064 rows=1 loops=2009) Index Cond: (id = $0) Total runtime: 165.016 ms (6 rows) Time: 166.786 ms --------------030100080708090409060609 Content-Type: text/plain; x-mac-type="0"; x-mac-creator="0"; name="sequential_exists_test.txt" Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="sequential_exists_test.txt" mifar07=# EXPLAIN ANALYZE SELECT id FROM ids WHERE id IN (SELECT id FROM object_t); QUERY PLAN --------------------------------------------------------------------------------------------------------------------------------- Nested Loop IN Join (cost=0.00..9851.68 rows=2140 width=4) (actual time=0.069..97.567 rows=1934 loops=1) -> Seq Scan on ids (cost=0.00..31.40 rows=2140 width=4) (actual time=0.012..1.868 rows=2000 loops=1) -> Index Scan using object_t_pkey on object_t (cost=0.00..4.58 rows=1 width=4) (actual time=0.045..0.045 rows=1 loops=2000) Index Cond: ("outer".id = object_t.id) Total runtime: 98.236 ms (5 rows) Time: 99.921 ms mifar07=# EXPLAIN ANALYZE SELECT id FROM ids WHERE EXISTS (SELECT id FROM object_t o WHERE o.id = ids.id); QUERY PLAN ------------------------------------------------------------------------------------------------------------------------------------- Seq Scan on ids (cost=0.00..9824.93 rows=1070 width=4) (actual time=0.071..74.158 rows=1934 loops=1) Filter: (subplan) SubPlan -> Index Scan using object_t_pkey on object_t o (cost=0.00..4.58 rows=1 width=4) (actual time=0.013..0.013 rows=1 loops=2000) Index Cond: (id = $0) Total runtime: 74.798 ms (6 rows) Time: 86.287 ms mifar07=# EXPLAIN ANALYZE SELECT id FROM ids WHERE id NOT IN (SELECT id FROM object_t); QUERY PLAN --------------------------------------------------------------------------------------------------------------------------- Seq Scan on ids (cost=36410.51..36447.26 rows=1070 width=4) (actual time=4024.613..4028.774 rows=66 loops=1) Filter: (NOT (hashed subplan)) SubPlan -> Seq Scan on object_t (cost=0.00..34381.81 rows=811481 width=4) (actual time=0.040..2503.374 rows=811481 loops=1) Total runtime: 4122.327 ms (5 rows) Time: 4134.659 ms mifar07=# EXPLAIN ANALYZE SELECT id FROM ids WHERE NOT EXISTS (SELECT id FROM object_t o WHERE o.id = ids.id); QUERY PLAN ------------------------------------------------------------------------------------------------------------------------------------- Seq Scan on ids (cost=0.00..9824.93 rows=1070 width=4) (actual time=0.220..92.611 rows=66 loops=1) Filter: (NOT (subplan)) SubPlan -> Index Scan using object_t_pkey on object_t o (cost=0.00..4.58 rows=1 width=4) (actual time=0.043..0.043 rows=1 loops=2000) Index Cond: (id = $0) Total runtime: 92.743 ms (6 rows) Time: 94.190 ms mifar07=# --------------030100080708090409060609-- --------------enig7354D6CDCD06455821ED204B Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC8i9cJdeBCYSNAAMRAr5bAKCXMOelVjmGxp0O0x6BK9Orn0tYYwCeLGt6 Q+6Co3FT9oCLnVyrtGxt/ec= =CC8z -----END PGP SIGNATURE----- --------------enig7354D6CDCD06455821ED204B-- From pgsql-performance-owner@postgresql.org Thu Aug 4 13:16:00 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 43E04529F6 for ; Thu, 4 Aug 2005 13:14:18 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78148-09 for ; Thu, 4 Aug 2005 16:14:14 +0000 (GMT) Received: from larry.aptalaska.net (larry.aptalaska.net [64.186.96.3]) by svr1.postgresql.org (Postfix) with ESMTP id 6A7FB529D3 for ; Thu, 4 Aug 2005 13:14:13 -0300 (ADT) Received: from [192.168.98.9] (rdbck-static-287.palmer.mtaonline.net [64.4.232.33]) (authenticated bits=0) by larry.aptalaska.net (8.13.4/8.13.4) with ESMTP id j74GDct8026452 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Thu, 4 Aug 2005 08:13:38 -0800 Message-ID: <42F23EAF.2070603@aptalaska.net> Date: Thu, 04 Aug 2005 08:13:35 -0800 From: Matthew Schumacher User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: John A Meinel Cc: pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 References: <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> <42EE93F8.5010003@aptalaska.net> <20050801214931.GF60019@decibel.org> <42EE9D83.8000405@pobox.com> <42F1CE43.8040801@aptalaska.net> <42F221BB.5000702@arbash-meinel.com> In-Reply-To: <42F221BB.5000702@arbash-meinel.com> X-Enigmail-Version: 0.91.0.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit X-Spam-Bayes: Learn: ham X-Scanned-By: MIMEDefang 2.52 on 64.186.96.3 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/38 X-Sequence-Number: 13785 John A Meinel wrote: > Matthew Schumacher wrote: > > I recommend that you drop and re-create the temp table. There is no > reason to have it around, considering you delete and re-add everything. > That means you never have to vacuum it, since it always only contains > the latest rows. Whenever I have a create temp and drop statement I get these errors: select put_tokens(1, '{"\\\\000"}', 1, 1, 1000); ERROR: relation with OID 582248 does not exist CONTEXT: SQL statement "INSERT INTO bayes_token_tmp VALUES ( $1 )" PL/pgSQL function "put_tokens" line 12 at SQL statement > > > > My one question here, is the inspam_count and inham_count *always* the > same for all tokens? I would have thought each token has it's own count. > Anyway, there are a few lines I would change: No, we get the userid, inspam, inham, and atime, and they are the same for each token. If we have a different user we call the proc again. > -- create the table at the start of the procedure > CREATE TEMP TABLE bayes_token_tmp (intoken bytea); > -- You might also add primary key if you are going to be adding > -- *lots* of entries, but it sounds like you are going to have > -- less than 1 page, so it doesn't matter This causes errors, see above.... > -- (token) IN (SELECT intoken FROM bayes_token_tmp); > EXISTS (SELECT token FROM bayes_token_tmp > WHERE intoken=token LIMIT 1); > > -- I would also avoid your intoken (NOT) IN (SELECT token FROM > -- bayes_token) There are a few possibilities, but to me > -- as your bayes_token table becomes big, this will start > -- to be the slow point > > -- Rather than doing 2 NOT IN queries, it *might* be faster to do > DELETE FROM bayes_token_tmp > WHERE NOT EXISTS (SELECT token FROM bayes_token > WHERE token=intoken); > > I'll look into this. thanks, schu From pgsql-performance-owner@postgresql.org Thu Aug 4 13:16:24 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 37DAB529D3 for ; Thu, 4 Aug 2005 13:16:22 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 79071-05 for ; Thu, 4 Aug 2005 16:16:17 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 2CC9452896 for ; Thu, 4 Aug 2005 13:16:16 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j74GGAX9016346; Thu, 4 Aug 2005 12:16:10 -0400 (EDT) To: John A Meinel Cc: Matthew Schumacher , pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 In-reply-to: <42F22F59.7090901@arbash-meinel.com> References: <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> <42EE93F8.5010003@aptalaska.net> <20050801214931.GF60019@decibel.org> <42EE9D83.8000405@pobox.com> <42F1CE43.8040801@aptalaska.net> <15610.1123166222@sss.pgh.pa.us> <42F22F59.7090901@arbash-meinel.com> Comments: In-reply-to John A Meinel message dated "Thu, 04 Aug 2005 10:08:09 -0500" Date: Thu, 04 Aug 2005 12:16:10 -0400 Message-ID: <16345.1123172170@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/39 X-Sequence-Number: 13786 John A Meinel writes: > Tom Lane wrote: >> I don't really see why you think that this path is going to lead to >> better performance than where you were before. > So for an IN (sub-select), does it actually pull all of the rows from > the other table, or is the planner smart enough to stop once it finds > something? It stops when it finds something --- but it's still a join operation in essence. I don't see that putting the values one by one into a table and then joining is going to be a win compared to just processing the values one at a time against the main table. > Is IN (sub-select) about the same as EXISTS (sub-select WHERE x=y)? > What about NOT IN (sub-select) versus NOT EXISTS (sub-select WHERE x=y) The EXISTS variants are actually worse, because we've not spent as much time teaching the planner how to optimize them. There's effectively only one decent plan for an EXISTS, which is that the subselect's "x" is indexed and we do an indexscan probe using the outer "y" for each outer row. IN and NOT IN can do that, or several alternative plans that might be better depending on data statistics. However, that's cold comfort for Matthew's application -- the only way he'd get any benefit from all those planner smarts is if he ANALYZEs the temp table after loading it and then EXECUTEs the main query (so that it gets re-planned every time). Plus, at least some of those alternative plans would require an index on the temp table, which is unlikely to be worth the cost of setting up. And finally, this formulation requires separate IN and NOT IN tests that are necessarily going to do a lot of redundant work. There's enough overhead here that I find it highly doubtful that it'll be a win compared to the original approach of retail queries against the main table. regards, tom lane From pgsql-performance-owner@postgresql.org Thu Aug 4 13:32:08 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8AC0852A76 for ; Thu, 4 Aug 2005 13:18:06 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78178-10 for ; Thu, 4 Aug 2005 16:18:05 +0000 (GMT) Received: from larry.aptalaska.net (larry.aptalaska.net [64.186.96.3]) by svr1.postgresql.org (Postfix) with ESMTP id E1BDF52A56 for ; Thu, 4 Aug 2005 13:18:04 -0300 (ADT) Received: from [192.168.98.9] (rdbck-static-287.palmer.mtaonline.net [64.4.232.33]) (authenticated bits=0) by larry.aptalaska.net (8.13.4/8.13.4) with ESMTP id j74GI1Be027516 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Thu, 4 Aug 2005 08:18:02 -0800 Message-ID: <42F23FB6.5010405@aptalaska.net> Date: Thu, 04 Aug 2005 08:17:58 -0800 From: Matthew Schumacher User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 References: <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> <42EE93F8.5010003@aptalaska.net> <20050801214931.GF60019@decibel.org> <42EE9D83.8000405@pobox.com> <42F1CE43.8040801@aptalaska.net> <15610.1123166222@sss.pgh.pa.us> In-Reply-To: <15610.1123166222@sss.pgh.pa.us> X-Enigmail-Version: 0.91.0.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit X-Spam-Bayes: Learn: unavailable X-Scanned-By: MIMEDefang 2.52 on 64.186.96.3 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/40 X-Sequence-Number: 13787 Tom Lane wrote: > I don't really see why you think that this path is going to lead to > better performance than where you were before. Manipulation of the > temp table is never going to be free, and IN (sub-select) is always > inherently not fast, and NOT IN (sub-select) is always inherently > awful. Throwing a pile of simple queries at the problem is not > necessarily the wrong way ... especially when you are doing it in > plpgsql, because you've already eliminated the overhead of network > round trips and repeated planning of the queries. > > regards, tom lane The reason why I think this may be faster is because I would avoid running an update on data that needs to be inserted which saves searching though the table for a matching token. Perhaps I should do the insert first, then drop those tokens from the temp table, then do my updates in a loop. I'll have to do some benchmarking... schu From pgsql-performance-owner@postgresql.org Thu Aug 4 14:34:03 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 84E8B52A56 for ; Thu, 4 Aug 2005 14:34:01 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 97153-06 for ; Thu, 4 Aug 2005 17:33:58 +0000 (GMT) Received: from mailbox.samurai.com (mailbox.samurai.com [205.207.28.82]) by svr1.postgresql.org (Postfix) with ESMTP id BD0AE5282F for ; Thu, 4 Aug 2005 14:33:55 -0300 (ADT) Received: from localhost (mailbox.samurai.com [205.207.28.82]) by mailbox.samurai.com (Postfix) with ESMTP id DB697239BEA; Thu, 4 Aug 2005 13:33:55 -0400 (EDT) Received: from mailbox.samurai.com ([205.207.28.82]) by localhost (mailbox.samurai.com [205.207.28.82]) (amavisd-new, port 10024) with LMTP id 09013-01-2; Thu, 4 Aug 2005 13:33:25 -0400 (EDT) Received: from [192.168.0.105] (CPE3285f62f0d42-CM0011aec5ebbc.cpe.net.cable.rogers.com [69.199.116.160]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (No client certificate requested) by mailbox.samurai.com (Postfix) with ESMTP id 38626239C3C; Thu, 4 Aug 2005 13:32:40 -0400 (EDT) Message-ID: <42F25137.4030107@samurai.com> Date: Thu, 04 Aug 2005 13:32:39 -0400 From: Neil Conway User-Agent: Mozilla Thunderbird 1.0.2 (Macintosh/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: "Jim C. Nasby" Cc: Tom Lane , Tobias Brox , pgsql-performance@postgresql.org Subject: Re: "nice"/low priority Query References: <20050802160434.GD9278@tobias.lan> <11484.1122999570@sss.pgh.pa.us> <20050802172550.GZ60019@decibel.org> In-Reply-To: <20050802172550.GZ60019@decibel.org> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at mailbox.samurai.com X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/41 X-Sequence-Number: 13788 Jim C. Nasby wrote: > Actually, from what I've read 4.2BSD actually took priority into account > when scheduling I/O. FWIW, you can set I/O priority in recent versions of the Linux kernel using ionice, which is part of RML's schedutils package (which was recently merged into util-linux). -Neil From pgsql-performance-owner@postgresql.org Thu Aug 4 14:36:01 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E115A52AB8 for ; Thu, 4 Aug 2005 14:35:53 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 97627-09 for ; Thu, 4 Aug 2005 17:35:50 +0000 (GMT) Received: from larry.aptalaska.net (larry.aptalaska.net [64.186.96.3]) by svr1.postgresql.org (Postfix) with ESMTP id A52A352A97 for ; Thu, 4 Aug 2005 14:35:48 -0300 (ADT) Received: from [192.168.98.9] (rdbck-static-287.palmer.mtaonline.net [64.4.232.33]) (authenticated bits=0) by larry.aptalaska.net (8.13.4/8.13.4) with ESMTP id j74HZiHp013869 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Thu, 4 Aug 2005 09:35:45 -0800 Message-ID: <42F251F0.1070407@aptalaska.net> Date: Thu, 04 Aug 2005 09:35:44 -0800 From: Matthew Schumacher User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 References: <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> <42EE93F8.5010003@aptalaska.net> <20050801214931.GF60019@decibel.org> <42EE9D83.8000405@pobox.com> <42F1CE43.8040801@aptalaska.net> <15610.1123166222@sss.pgh.pa.us> <42F23FB6.5010405@aptalaska.net> In-Reply-To: <42F23FB6.5010405@aptalaska.net> X-Enigmail-Version: 0.91.0.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit X-Spam-Bayes: Learn: ham X-Scanned-By: MIMEDefang 2.52 on 64.186.96.3 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/42 X-Sequence-Number: 13789 Matthew Schumacher wrote: > Tom Lane wrote: > > >>I don't really see why you think that this path is going to lead to >>better performance than where you were before. Manipulation of the >>temp table is never going to be free, and IN (sub-select) is always >>inherently not fast, and NOT IN (sub-select) is always inherently >>awful. Throwing a pile of simple queries at the problem is not >>necessarily the wrong way ... especially when you are doing it in >>plpgsql, because you've already eliminated the overhead of network >>round trips and repeated planning of the queries. >> >> regards, tom lane > > > The reason why I think this may be faster is because I would avoid > running an update on data that needs to be inserted which saves > searching though the table for a matching token. > > Perhaps I should do the insert first, then drop those tokens from the > temp table, then do my updates in a loop. > > I'll have to do some benchmarking... > > schu Tom, I think your right, whenever I do a NOT IN it does a full table scan against bayes_token and since that table is going to get very big doing the simple query in a loop that uses an index seems a bit faster. John, thanks for your help, it was worth a try, but it looks like the looping is just faster. Here is what I have so far in case anyone else has ideas before I abandon it: CREATE OR REPLACE FUNCTION put_tokens(inuserid INTEGER, intokenary BYTEA[], inspam_count INTEGER, inham_count INTEGER, inatime INTEGER) RETURNS VOID AS ' DECLARE _token BYTEA; BEGIN UPDATE bayes_token SET spam_count = greatest_int(spam_count + inspam_count, 0), ham_count = greatest_int(ham_count + inham_count , 0), atime = greatest_int(atime, inatime) WHERE id = inuserid AND (token) IN (SELECT bayes_token_tmp FROM bayes_token_tmp(intokenary)); UPDATE bayes_vars SET token_count = token_count + ( SELECT count(bayes_token_tmp) FROM bayes_token_tmp(intokenary) WHERE bayes_token_tmp NOT IN (SELECT token FROM bayes_token)), newest_token_age = greatest_int(newest_token_age, inatime), oldest_token_age = least_int(oldest_token_age, inatime) WHERE id = inuserid; INSERT INTO bayes_token SELECT inuserid, bayes_token_tmp, inspam_count, inham_count, inatime FROM bayes_token_tmp(intokenary) WHERE (inspam_count > 0 OR inham_count > 0) AND (bayes_token_tmp) NOT IN (SELECT token FROM bayes_token); RETURN; END; ' LANGUAGE 'plpgsql'; CREATE OR REPLACE FUNCTION bayes_token_tmp(intokenary BYTEA[]) RETURNS SETOF bytea AS ' BEGIN for i in array_lower(intokenary, 1) .. array_upper(intokenary, 1) LOOP return next intokenary[i]; END LOOP; RETURN; end ' language 'plpgsql'; CREATE OR REPLACE FUNCTION greatest_int (integer, integer) RETURNS INTEGER IMMUTABLE STRICT AS 'SELECT CASE WHEN $1 < $2 THEN $2 ELSE $1 END;' LANGUAGE SQL; CREATE OR REPLACE FUNCTION least_int (integer, integer) RETURNS INTEGER IMMUTABLE STRICT AS 'SELECT CASE WHEN $1 < $2 THEN $1 ELSE $2 END;' LANGUAGE SQL; From pgsql-performance-owner@postgresql.org Thu Aug 4 16:36:35 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4FBD052A28 for ; Thu, 4 Aug 2005 16:36:32 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 38890-09 for ; Thu, 4 Aug 2005 19:36:23 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id D129552A34 for ; Thu, 4 Aug 2005 16:36:21 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050804193622m92009hu98e>; Thu, 4 Aug 2005 19:36:22 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 98B2256051; Thu, 4 Aug 2005 14:36:21 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id 8B73056016; Thu, 4 Aug 2005 14:36:15 -0500 (CDT) Message-ID: <42F26E2C.8010607@arbash-meinel.com> Date: Thu, 04 Aug 2005 14:36:12 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Matthew Schumacher Cc: Tom Lane , pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 References: <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> <42EE93F8.5010003@aptalaska.net> <20050801214931.GF60019@decibel.org> <42EE9D83.8000405@pobox.com> <42F1CE43.8040801@aptalaska.net> <15610.1123166222@sss.pgh.pa.us> <42F23FB6.5010405@aptalaska.net> <42F251F0.1070407@aptalaska.net> In-Reply-To: <42F251F0.1070407@aptalaska.net> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigE5BF5323050368A690114C1F" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.045 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/43 X-Sequence-Number: 13790 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigE5BF5323050368A690114C1F Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Matthew Schumacher wrote: > Matthew Schumacher wrote: > >>Tom Lane wrote: >> >> >> >>>I don't really see why you think that this path is going to lead to >>>better performance than where you were before. Manipulation of the >>>temp table is never going to be free, and IN (sub-select) is always >>>inherently not fast, and NOT IN (sub-select) is always inherently >>>awful. Throwing a pile of simple queries at the problem is not >>>necessarily the wrong way ... especially when you are doing it in >>>plpgsql, because you've already eliminated the overhead of network >>>round trips and repeated planning of the queries. >>> >>> regards, tom lane >> >> >>The reason why I think this may be faster is because I would avoid >>running an update on data that needs to be inserted which saves >>searching though the table for a matching token. >> >>Perhaps I should do the insert first, then drop those tokens from the >>temp table, then do my updates in a loop. >> >>I'll have to do some benchmarking... >> >>schu > > > Tom, I think your right, whenever I do a NOT IN it does a full table > scan against bayes_token and since that table is going to get very big > doing the simple query in a loop that uses an index seems a bit faster. > > John, thanks for your help, it was worth a try, but it looks like the > looping is just faster. > > Here is what I have so far in case anyone else has ideas before I > abandon it: Surely this isn't what you have. You have *no* loop here, and you have stuff like: AND (bayes_token_tmp) NOT IN (SELECT token FROM bayes_token); I'm guessing this isn't your last version of the function. As far as putting the CREATE TEMP TABLE inside the function, I think the problem is that the first time it runs, it compiles the function, and when it gets to the UPDATE/INSERT with the temporary table name, at compile time it hard-codes that table id. I tried getting around it by using "EXECUTE" which worked, but it made the function horribly slow. So I don't recommend it. Anyway, if you want us to evaluate it, you really need to send us the real final function. John =:-> > > CREATE OR REPLACE FUNCTION put_tokens(inuserid INTEGER, > intokenary BYTEA[], > inspam_count INTEGER, > inham_count INTEGER, > inatime INTEGER) > RETURNS VOID AS ' > DECLARE > _token BYTEA; > BEGIN > > UPDATE > bayes_token > SET > spam_count = greatest_int(spam_count + inspam_count, 0), > ham_count = greatest_int(ham_count + inham_count , 0), > atime = greatest_int(atime, inatime) > WHERE > id = inuserid > AND > (token) IN (SELECT bayes_token_tmp FROM bayes_token_tmp(intokenary)); > > UPDATE > bayes_vars > SET > token_count = token_count + ( > SELECT > count(bayes_token_tmp) > FROM > bayes_token_tmp(intokenary) > WHERE > bayes_token_tmp NOT IN (SELECT token FROM bayes_token)), > newest_token_age = greatest_int(newest_token_age, inatime), > oldest_token_age = least_int(oldest_token_age, inatime) > WHERE > id = inuserid; > > INSERT INTO > bayes_token > SELECT > inuserid, > bayes_token_tmp, > inspam_count, > inham_count, > inatime > FROM > bayes_token_tmp(intokenary) > WHERE > (inspam_count > 0 OR inham_count > 0) > AND > (bayes_token_tmp) NOT IN (SELECT token FROM bayes_token); > > RETURN; > END; > ' LANGUAGE 'plpgsql'; > > CREATE OR REPLACE FUNCTION bayes_token_tmp(intokenary BYTEA[]) RETURNS > SETOF bytea AS > ' > BEGIN > for i in array_lower(intokenary, 1) .. array_upper(intokenary, 1) > LOOP > return next intokenary[i]; > END LOOP; > RETURN; > end > ' > language 'plpgsql'; > > CREATE OR REPLACE FUNCTION greatest_int (integer, integer) > RETURNS INTEGER > IMMUTABLE STRICT > AS 'SELECT CASE WHEN $1 < $2 THEN $2 ELSE $1 END;' > LANGUAGE SQL; > > CREATE OR REPLACE FUNCTION least_int (integer, integer) > RETURNS INTEGER > IMMUTABLE STRICT > AS 'SELECT CASE WHEN $1 < $2 THEN $1 ELSE $2 END;' > LANGUAGE SQL; > > ---------------------------(end of broadcast)--------------------------- > TIP 4: Have you searched our list archives? > > http://archives.postgresql.org > --------------enigE5BF5323050368A690114C1F Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC8m4vJdeBCYSNAAMRAvw8AJ0dAsgk5dgc0itJcCipWyX9Hs2gUACfa3gZ 8Z7cmebXPoBQBQqjTDaF3nU= =ahb7 -----END PGP SIGNATURE----- --------------enigE5BF5323050368A690114C1F-- From pgsql-performance-owner@postgresql.org Thu Aug 4 19:38:23 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 825F55282F for ; Thu, 4 Aug 2005 19:38:22 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 81498-09 for ; Thu, 4 Aug 2005 22:38:13 +0000 (GMT) Received: from larry.aptalaska.net (larry.aptalaska.net [64.186.96.3]) by svr1.postgresql.org (Postfix) with ESMTP id 4E09452800 for ; Thu, 4 Aug 2005 19:38:12 -0300 (ADT) Received: from [192.168.98.9] (rdbck-static-287.palmer.mtaonline.net [64.4.232.33]) (authenticated bits=0) by larry.aptalaska.net (8.13.4/8.13.4) with ESMTP id j74MbfVE015904 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Thu, 4 Aug 2005 14:37:42 -0800 Message-ID: <42F298B5.5010405@aptalaska.net> Date: Thu, 04 Aug 2005 14:37:41 -0800 From: Matthew Schumacher User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: John A Meinel Cc: pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 References: <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> <42EE93F8.5010003@aptalaska.net> <20050801214931.GF60019@decibel.org> <42EE9D83.8000405@pobox.com> <42F1CE43.8040801@aptalaska.net> <15610.1123166222@sss.pgh.pa.us> <42F23FB6.5010405@aptalaska.net> <42F251F0.1070407@aptalaska.net> <42F26E2C.8010607@arbash-meinel.com> In-Reply-To: <42F26E2C.8010607@arbash-meinel.com> X-Enigmail-Version: 0.91.0.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit X-Spam-Bayes: Learn: unavailable X-Scanned-By: MIMEDefang 2.52 on 64.186.96.3 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/44 X-Sequence-Number: 13791 John A Meinel wrote: > Surely this isn't what you have. You have *no* loop here, and you have > stuff like: > AND > (bayes_token_tmp) NOT IN (SELECT token FROM bayes_token); > > I'm guessing this isn't your last version of the function. > > As far as putting the CREATE TEMP TABLE inside the function, I think the > problem is that the first time it runs, it compiles the function, and > when it gets to the UPDATE/INSERT with the temporary table name, at > compile time it hard-codes that table id. > > I tried getting around it by using "EXECUTE" which worked, but it made > the function horribly slow. So I don't recommend it. > > Anyway, if you want us to evaluate it, you really need to send us the > real final function. > > John > =:-> It is the final function. It doesn't need a loop because of the bayes_token_tmp function I added. The array is passed to it and it returns a record set so I can work off of it like it's a table. So the function works the same way it before, but instead of using SELECT intoken from TEMPTABLE, you use SELECT bayes_token_tmp from bayes_token_tmp(intokenary). I think this is more efficient than the create table overhead, especially because the incoming record set won't be to big. Thanks, schu From pgsql-performance-owner@postgresql.org Thu Aug 4 19:43:09 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B2362528B3 for ; Thu, 4 Aug 2005 19:43:08 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83408-08 for ; Thu, 4 Aug 2005 22:42:59 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id 713F25282F for ; Thu, 4 Aug 2005 19:42:58 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050804224259m92009ipade>; Thu, 4 Aug 2005 22:42:59 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 26B8F56051; Thu, 4 Aug 2005 17:42:59 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id 81F4F56016; Thu, 4 Aug 2005 17:42:53 -0500 (CDT) Message-ID: <42F299EC.8080809@arbash-meinel.com> Date: Thu, 04 Aug 2005 17:42:52 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Matthew Schumacher Cc: pgsql-performance@postgresql.org Subject: Re: Performance problems testing with Spamassassin 3.1.0 References: <42E9749F.6000709@aptalaska.net> <1122616211.5691.340.camel@lamb.mcmillan.net.nz> <42EA6966.8090402@aptalaska.net> <1122668371.5691.352.camel@lamb.mcmillan.net.nz> <19405.1122670917@sss.pgh.pa.us> <42EAA410.9010700@aptalaska.net> <27897.1122748133@sss.pgh.pa.us> <42EE8043.1050207@aptalaska.net> <42EE93F8.5010003@aptalaska.net> <20050801214931.GF60019@decibel.org> <42EE9D83.8000405@pobox.com> <42F1CE43.8040801@aptalaska.net> <15610.1123166222@sss.pgh.pa.us> <42F23FB6.5010405@aptalaska.net> <42F251F0.1070407@aptalaska.net> <42F26E2C.8010607@arbash-meinel.com> <42F298B5.5010405@aptalaska.net> In-Reply-To: <42F298B5.5010405@aptalaska.net> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig30078318CC01CA21BF5E0E03" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.045 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/45 X-Sequence-Number: 13792 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig30078318CC01CA21BF5E0E03 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Matthew Schumacher wrote: > John A Meinel wrote: > > >>Surely this isn't what you have. You have *no* loop here, and you have >>stuff like: >> AND >> (bayes_token_tmp) NOT IN (SELECT token FROM bayes_token); >> >>I'm guessing this isn't your last version of the function. >> >>As far as putting the CREATE TEMP TABLE inside the function, I think the >>problem is that the first time it runs, it compiles the function, and >>when it gets to the UPDATE/INSERT with the temporary table name, at >>compile time it hard-codes that table id. >> >>I tried getting around it by using "EXECUTE" which worked, but it made >>the function horribly slow. So I don't recommend it. >> >>Anyway, if you want us to evaluate it, you really need to send us the >>real final function. >> >>John >>=:-> > > > It is the final function. It doesn't need a loop because of the > bayes_token_tmp function I added. The array is passed to it and it > returns a record set so I can work off of it like it's a table. So the > function works the same way it before, but instead of using SELECT > intoken from TEMPTABLE, you use SELECT bayes_token_tmp from > bayes_token_tmp(intokenary). > > I think this is more efficient than the create table overhead, > especially because the incoming record set won't be to big. > > Thanks, > > schu > > Well, I would at least recommend that you change the "WHERE bayes_token_tmp NOT IN (SELECT token FROM bayes_token)" with a "WHERE NOT EXISTS (SELECT toke FROM bayes_token WHERE token=bayes_token_tmp)" You might try experimenting with the differences, but on my system the NOT IN has to do a full sequential scan on bayes_token and load all entries into a list, while NOT EXISTS can do effectively a nested loop. The nested loop requires that there is an index on bayes_token(token), but I'm pretty sure there is anyway. Again, in my testing, it was a difference of 4200ms versus 180ms. (800k rows in my big table, 2k in the temp one) John =:-> --------------enig30078318CC01CA21BF5E0E03 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC8pnsJdeBCYSNAAMRApkHAKCXHa4ddIXDqGJRk4y9CRLeeEVjOACfcIqh FmYng0yOyadGlXkBQRAgH/E= =mZVF -----END PGP SIGNATURE----- --------------enig30078318CC01CA21BF5E0E03-- From pgsql-performance-owner@postgresql.org Fri Aug 5 08:11:36 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3658E52986 for ; Fri, 5 Aug 2005 08:11:35 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 77920-07 for ; Fri, 5 Aug 2005 11:11:32 +0000 (GMT) Received: from mail2.aeccom.com (port-212-202-101-158.static.qsc.de [212.202.101.158]) by svr1.postgresql.org (Postfix) with ESMTP id 67AF8529BD for ; Fri, 5 Aug 2005 08:11:29 -0300 (ADT) Received: from [192.168.2.12] (port-83-236-156-26.static.qsc.de [83.236.156.26]) by mail2.aeccom.com (Postfix) with ESMTP id E88E0108; Fri, 5 Aug 2005 13:11:31 +0200 (CEST) Message-ID: <42F34963.5080403@aeccom.com> Date: Fri, 05 Aug 2005 13:11:31 +0200 From: =?ISO-8859-1?Q?Dirk_Lutzeb=E4ck?= Organization: AEC/communications GmbH User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Cc: Sven Geisler Subject: Performance problems on 4-way AMD Opteron 875 (dual core) Content-Type: multipart/alternative; boundary="------------050707070801080103090606" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.232 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO, HTML_10_20, HTML_MESSAGE, HTML_TITLE_EMPTY X-Spam-Level: X-Archive-Number: 200508/46 X-Sequence-Number: 13793 This is a multi-part message in MIME format. --------------050707070801080103090606 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit [[I'm posting this on behalf of my co-worker who cannot post to this list at the moment]] Hi, I had installed PostgreSQL on a 4-way AMD Opteron 875 (dual core) and the performance isn't on the expected level. Details: The "old" server is a 4-way XEON MP 3.0 GHz with 4MB L3 cache, 32 GB RAM (PC1600) and local FC-RAID 10. Hyper-Threading is off. (DL580) The "old" server is using Red Hat Enterprise Linux 3 Update 5. The "new" server is a 4-way Opteron 875 with 1 MB L2 cache, 32 GB RAM (PC3200) and the same local FC-RAID 10. (HP DL585) The "new" server is using Red Hat Enterprise Linux 4 (with the latest x86_64 kernel from Red Hat - 2.6.9-11.ELsmp #1 SMP Fri May 20 18:25:30 EDT 2005 x86_64) I use PostgreSQL version 8.0.3. The issue is that the Opteron is slower as the XEON MP under high load. I have created a test with parallel queries which are typical for my application. The queries are in a range of small queries (0.1 seconds) and larger queries using join (15 seconds). The test starts parallel clients. Each clients runs the queries in a random order. The test takes care that a client use always the same random order to get valid results. Here are the number of queries which the server has finished in a fix period of time. I used PostgreSQL 8.1 snapshot from last week compiled as 64bit binary for DL585-64bit. I used PostgreSQL 8.0.3 compiled as 32bit binary for DL585-32bit and DL580. During the tests everything which is needed is in the file cache. I didn't have read activity. Context switch spikes are over 50000 during the test on both server. My feeling is that the XEON has a tick more context switches. PostgreSQL params: max_locks_per_transaction = 256 shared_buffers = 40000 effective_cache_size = 3840000 work_mem = 300000 maintenance_work_mem = 512000 wal_buffers = 32 checkpoint_segments = 24 I was expecting two times more queries on the DL585. The DL585 with PostgreSQL 8.0.3 32bit does meltdown earlier as the XEON in production use. Please compare 4 clients and 8 clients. With 4 clients the Opteron is in front and with 8 clients the XEON doesn't meltdown that much as the Opteron. I don't have any idea what cause this. Benchmarks like SAP's SD 2-tier showing that the DL585 can handle nearly three times more load as the DL580 with XEON 3.0. We choose the 4-way Opteron 875 based on such benchmark to replace the 4-way XEON MP. Does anyone have comments or ideas on which I have to focus my work? I guess, the shared buffer cause the meltdown when to many clients are accessing the same data. I didn't understand why the 4-way XEON MP 3.0 can deal with this better as the 4-way Opteron 875. The system load on the Opteron is never over 3.0. The XEON MP has a load up to 4.0. Should I try other settings for PostgreSQL in postgresql.conf? Should I try other setting for the compilation? I will compile the latest PostgreSQL 8.1 snapshot for 32bit to evaluate the new shared buffer code from Tom. I think, the 64bit is slow because my queries are CPU intensive. Can someone provide a commercial support contact for this issue? Sven. --------------050707070801080103090606 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit
[[I'm posting this on behalf of my co-worker who cannot post to this list at the moment]]

Hi,

I had installed PostgreSQL on a 4-way AMD Opteron 875 (dual core) and the performance isn't on the expected level.

Details:
The "old" server is a 4-way XEON MP 3.0 GHz with 4MB L3 cache, 32 GB RAM (PC1600)  and local FC-RAID 10. Hyper-Threading is off. (DL580)
The "old" server is using Red Hat Enterprise Linux 3 Update 5.
The "new" server is a 4-way Opteron 875 with 1 MB L2 cache, 32 GB RAM (PC3200) and the same local FC-RAID 10. (HP DL585)
The "new" server is using Red Hat Enterprise Linux 4 (with the latest x86_64 kernel from Red Hat - 2.6.9-11.ELsmp #1 SMP Fri May 20 18:25:30 EDT 2005 x86_64)
I use PostgreSQL version 8.0.3.

The issue is that the Opteron is slower as the XEON MP under high load. I have created a test with parallel queries which are typical for my application. The queries are in a range of small queries (0.1 seconds) and larger queries using join (15 seconds).
The test starts parallel clients. Each clients runs the queries in a random order. The test takes care that a client use always the same random order to get valid results.

Here are the number of queries which the server has finished in a fix period of time.
I used PostgreSQL 8.1 snapshot from last week compiled as 64bit binary for DL585-64bit.
I used PostgreSQL 8.0.3 compiled as 32bit binary for DL585-32bit and DL580.
During the tests everything which is needed is in the file cache. I didn't have read activity.
Context switch  spikes are over 50000 during the test on both server. My feeling is that the XEON has a tick more context switches.



PostgreSQL params:
max_locks_per_transaction = 256
shared_buffers = 40000
effective_cache_size = 3840000
work_mem = 300000
maintenance_work_mem = 512000
wal_buffers = 32
checkpoint_segments = 24


I was expecting two times more queries on the DL585. The DL585 with PostgreSQL 8.0.3 32bit does meltdown earlier as the XEON in production use. Please compare 4 clients and 8 clients. With 4 clients the Opteron is in front and with 8 clients the XEON doesn't meltdown that much as the Opteron.

I don't have any idea what cause this. Benchmarks like SAP's SD 2-tier showing that the DL585 can handle nearly three times more load as the DL580 with XEON 3.0. We choose the 4-way Opteron 875 based on such benchmark to replace the 4-way XEON MP.

Does anyone have comments or ideas on which I have to focus my work?

I guess, the shared buffer cause the meltdown when to many clients are accessing the same data.
I didn't understand why the 4-way XEON MP 3.0 can deal with this better as the 4-way Opteron 875.
The system load on the Opteron is never over 3.0. The XEON MP has a load up to 4.0.

Should I try other settings for PostgreSQL in postgresql.conf?
Should I try other setting for the compilation?

I will compile the latest PostgreSQL 8.1 snapshot for 32bit to evaluate the new shared buffer code from Tom.
I think, the 64bit is slow because my queries are CPU intensive.

Can someone provide a commercial support contact for this issue?

Sven.

--------------050707070801080103090606-- From pgsql-performance-owner@postgresql.org Fri Aug 5 09:27:25 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 20DE552BB7 for ; Fri, 5 Aug 2005 09:27:24 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 96596-04 for ; Fri, 5 Aug 2005 12:27:18 +0000 (GMT) Received: from vms042pub.verizon.net (vms042pub.verizon.net [206.46.252.42]) by svr1.postgresql.org (Postfix) with ESMTP id 58BFE52BC6 for ; Fri, 5 Aug 2005 09:27:17 -0300 (ADT) Received: from osgiliath.mathom.us ([70.108.53.154]) by vms042.mailsrvcs.net (Sun Java System Messaging Server 6.2 HotFix 0.04 (built Dec 24 2004)) with ESMTPA id <0IKR00GYS19KGZG0@vms042.mailsrvcs.net> for pgsql-performance@postgresql.org; Fri, 05 Aug 2005 07:27:21 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by osgiliath.mathom.us (Postfix) with ESMTP id 11A8C600498; Fri, 05 Aug 2005 08:27:20 -0400 (EDT) Received: from osgiliath.mathom.us ([127.0.0.1]) by localhost (osgiliath [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 18342-02-6; Fri, 05 Aug 2005 08:27:19 -0400 (EDT) Received: by osgiliath.mathom.us (Postfix, from userid 1000) id D291C60048A; Fri, 05 Aug 2005 08:27:19 -0400 (EDT) Date: Fri, 05 Aug 2005 08:27:19 -0400 From: Michael Stone Subject: Re: Performance problems on 4-way AMD Opteron 875 (dual core) In-reply-to: <42F34963.5080403@aeccom.com> To: Dirk =?iso-8859-1?Q?Lutzeb=E4ck?= Cc: pgsql-performance@postgresql.org, Sven Geisler Mail-Followup-To: Dirk =?iso-8859-1?Q?Lutzeb=E4ck?= , pgsql-performance@postgresql.org, Sven Geisler Message-id: <20050805122719.GW19080@mathom.us> MIME-version: 1.0 Content-type: text/plain; charset=iso-8859-1; format=flowed Content-transfer-encoding: 8BIT Content-disposition: inline X-Pgp-Fingerprint: 53 FF 38 00 E7 DD 0A 9C 84 52 84 C5 EE DF 7C 88 X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mathom.us References: <42F34963.5080403@aeccom.com> User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.007 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/47 X-Sequence-Number: 13794 On Fri, Aug 05, 2005 at 01:11:31PM +0200, Dirk Lutzeb�ck wrote: >I will compile the latest PostgreSQL 8.1 snapshot for 32bit to evaluate >the new shared buffer code from Tom. >I think, the 64bit is slow because my queries are CPU intensive. Have you actually tried it or are you guessing? If you're guessing, then compile it as a 64 bit binary and benchmark that. Mike Stone From pgsql-performance-owner@postgresql.org Fri Aug 5 09:56:25 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 84F4752836 for ; Fri, 5 Aug 2005 09:56:24 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 01448-07 for ; Fri, 5 Aug 2005 12:56:20 +0000 (GMT) Received: from mail2.aeccom.com (port-212-202-101-158.static.qsc.de [212.202.101.158]) by svr1.postgresql.org (Postfix) with ESMTP id 98A1C5293F for ; Fri, 5 Aug 2005 09:56:19 -0300 (ADT) Received: from [192.168.2.12] (port-83-236-156-26.static.qsc.de [83.236.156.26]) by mail2.aeccom.com (Postfix) with ESMTP id BC565108; Fri, 5 Aug 2005 14:56:23 +0200 (CEST) Message-ID: <42F361F7.8070201@aeccom.com> Date: Fri, 05 Aug 2005 14:56:23 +0200 From: =?ISO-8859-1?Q?Dirk_Lutzeb=E4ck?= Organization: AEC/communications GmbH User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Michael Stone Cc: pgsql-performance@postgresql.org, Sven Geisler Subject: Re: Performance problems on 4-way AMD Opteron 875 (dual References: <42F34963.5080403@aeccom.com> <20050805122719.GW19080@mathom.us> In-Reply-To: <20050805122719.GW19080@mathom.us> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.077 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/48 X-Sequence-Number: 13795 Michael Stone wrote: > On Fri, Aug 05, 2005 at 01:11:31PM +0200, Dirk Lutzeb�ck wrote: > >> I will compile the latest PostgreSQL 8.1 snapshot for 32bit to >> evaluate the new shared buffer code from Tom. >> I think, the 64bit is slow because my queries are CPU intensive. > > > Have you actually tried it or are you guessing? If you're guessing, then > compile it as a 64 bit binary and benchmark that. > > Mike Stone We tried it. 64bit 8.1dev was slower than 32bit 8.0.3. Dirk From pgsql-performance-owner@postgresql.org Fri Aug 5 10:53:18 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 769BC529BA for ; Fri, 5 Aug 2005 10:53:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 16749-07 for ; Fri, 5 Aug 2005 13:53:13 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id DA67852909 for ; Fri, 5 Aug 2005 10:53:12 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j75DrEXI008259; Fri, 5 Aug 2005 09:53:14 -0400 (EDT) To: =?ISO-8859-1?Q?Dirk_Lutzeb=E4ck?= Cc: pgsql-performance@postgresql.org, Sven Geisler Subject: Re: Performance problems on 4-way AMD Opteron 875 (dual core) In-reply-to: <42F34963.5080403@aeccom.com> References: <42F34963.5080403@aeccom.com> Comments: In-reply-to =?ISO-8859-1?Q?Dirk_Lutzeb=E4ck?= message dated "Fri, 05 Aug 2005 13:11:31 +0200" Date: Fri, 05 Aug 2005 09:53:14 -0400 Message-ID: <8258.1123249994@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/49 X-Sequence-Number: 13796 =?ISO-8859-1?Q?Dirk_Lutzeb=E4ck?= writes: > Here are the number of queries which the server has finished in a fix > period of time. Uh, you never actually supplied any numbers (or much of any other specifics about what was tested, either). My first reaction is "don't vary more than one experimental parameter at a time". There is no way to tell whether the discrepancy is due to the different hardware, different Postgres version, or 32-bit vs 64-bit build. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Aug 5 15:35:12 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2630452848 for ; Fri, 5 Aug 2005 15:35:11 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 18968-06 for ; Fri, 5 Aug 2005 18:35:03 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.207]) by svr1.postgresql.org (Postfix) with ESMTP id 1F6335286E for ; Fri, 5 Aug 2005 15:35:02 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i11so433141wra for ; Fri, 05 Aug 2005 11:35:03 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:mime-version:content-type:content-transfer-encoding:content-disposition; b=R1QoxStqWrghhpk7TOc0/qoVi4LnLpOAqN2FwQ1OVbrYBD9b2o1CbmUi7zddcH3Y376XQY+yQOT3we4BGeYS52Qxbnth+2VK+Dk/UbhC903liHyF85MhyMDJ+edbUgTJzy7TZksuddd1m/g3Iucn4QoxDwhZKV3hZx8rQzlBoE8= Received: by 10.54.34.54 with SMTP id h54mr2799504wrh; Fri, 05 Aug 2005 11:35:03 -0700 (PDT) Received: by 10.54.119.7 with HTTP; Fri, 5 Aug 2005 11:35:03 -0700 (PDT) Message-ID: <41b0fe890508051135712a086b@mail.gmail.com> Date: Fri, 5 Aug 2005 11:35:03 -0700 From: Rhett Garber Reply-To: Rhett Garber To: pgsql-performance@postgresql.org Subject: Why hash join instead of nested loop? Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 tagged_above=0 required=5 tests=RCVD_BY_IP X-Spam-Level: X-Archive-Number: 200508/50 X-Sequence-Number: 13797 I've got similiar queries that I think should be evaluated (as displayed through 'explain') the same, but don't. Hopefully this is the rigth place to send such a question and one of you can help explain this to me. The Tables: Connection - 1.2 million entries, about 60 megs, 3 integer fields that link two tables together (i.e. an identifier and two foreign keys). has an index on the identifier and either of the foreign keys. rtmessagestate - very small, 5 entries rtmessage - pretty big, 80,000 entries The Queries: =20 select rtmessagestate.* from rtmessagestate, connection where connection_registry_id =3D 40105 and obj1 =3D 73582 and obj2 =3D rtmessagestate.id; returns 1 in 13.7 ms = =20 select rtmessage.id, subject from rtmessage, connection where connection_registry_id =3D 40003 and obj1 =3D 4666 and obj2 =3D rtmessage.id; returns 12 in 2 ms Some more possibly important details: entries in Connection with connection_registry_id =3D 40105: 30,000 entries with this id and obj1 =3D 73582: 1 entries in Connection with connection_registry_id =3D 40003: 6,000 entries with this id and obj1 =3D 4666: 20 =20 but as I said before, there is an btree index on (connection_registry_id, o= bj1) Explain: The first query, breaks down as: Hash Join (cost=3D5.96..7.04 rows=3D1 width=3D14) Hash Cond: ("outer".id =3D "inner".obj2) -> Seq Scan on rtmessagestate (cost=3D0.00..1.05 rows=3D5 width= =3D14) -> Hash (cost=3D5.96..5.96 rows=3D1 width=3D4) -> Index Scan using connection_regid_obj1_index on connection (cost=3D0.00..5.96 rows=3D1 width=3D4) Index Cond: ((connection_registry_id =3D 40105) AND (obj1 =3D 73582))(6 rows) While the second query is: Nested Loop (cost=3D0.00..11.62 rows=3D2 width=3D38) -> Index Scan using connection_regid_obj1_index on connection=20 (cost=3D0.00..5.96 rows=3D1 width=3D4) Index Cond: ((connection_registry_id =3D 40003) AND (obj1 =3D = 4666)) -> Index Scan using rtmessage_pkey on rtmessage=20 (cost=3D0.00..5.65 rows=3D1 width=3D38) Index Cond: ("outer".obj2 =3D rtmessage.id) (5 rows) Actually running these queries shows that the second one (nested loop) is much faster than the hash join, presumably because of hash startup costs. Any ideas how I can make them both use the nested loop. I assume that this would be the fastest for both. Oddly enough, running the 1st query (rtmessagestate) as two queries or with a sub query is way faster than doing the join. And yes, I realize this schema may not be the most efficient for these examples, but it seems to be the most flexible. I'm working on some schema improvements also but if I could understand why this is slow that woudl probably help also. Thanks for you help, Rhett From pgsql-performance-owner@postgresql.org Fri Aug 5 19:13:54 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 161D952836 for ; Fri, 5 Aug 2005 19:13:53 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 63894-07 for ; Fri, 5 Aug 2005 22:13:51 +0000 (GMT) Received: from viefep20-int.chello.at (viefep12-int.chello.at [213.46.255.25]) by svr1.postgresql.org (Postfix) with ESMTP id C28035286E for ; Fri, 5 Aug 2005 19:13:49 -0300 (ADT) Received: from OTTO ([213.222.172.216]) by viefep20-int.chello.at (InterMail vM.6.01.04.04 201-2131-118-104-20050224) with SMTP id <20050805221350.KZHC25440.viefep20-int.chello.at@OTTO>; Sat, 6 Aug 2005 00:13:50 +0200 Message-ID: <001801c59a0a$f4be7480$9a00a8c0@OTTO> From: =?ISO-8859-1?Q?Havasv=F6lgyi_Ott=F3?= To: "Rhett Garber" , References: <41b0fe890508051135712a086b@mail.gmail.com> Subject: Re: Why hash join instead of nested loop? Date: Sat, 6 Aug 2005 00:13:49 +0200 MIME-Version: 1.0 Content-Type: text/plain; format=flowed; charset="ISO-8859-1"; reply-type=original Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 6.00.2900.2527 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2527 X-Virus-Scanned: by amavisd-new at hub.org X-Archive-Number: 200508/51 X-Sequence-Number: 13798 Rhett, Please post the explain analyze for both queries. From that we can see the predicted and the actual costs of them. Regards, Otto ----- Original Message ----- From: "Rhett Garber" To: Sent: Friday, August 05, 2005 8:35 PM Subject: [PERFORM] Why hash join instead of nested loop? I've got similiar queries that I think should be evaluated (as displayed through 'explain') the same, but don't. Hopefully this is the rigth place to send such a question and one of you can help explain this to me. The Tables: Connection - 1.2 million entries, about 60 megs, 3 integer fields that link two tables together (i.e. an identifier and two foreign keys). has an index on the identifier and either of the foreign keys. rtmessagestate - very small, 5 entries rtmessage - pretty big, 80,000 entries The Queries: select rtmessagestate.* from rtmessagestate, connection where connection_registry_id = 40105 and obj1 = 73582 and obj2 = rtmessagestate.id; returns 1 in 13.7 ms select rtmessage.id, subject from rtmessage, connection where connection_registry_id = 40003 and obj1 = 4666 and obj2 = rtmessage.id; returns 12 in 2 ms Some more possibly important details: entries in Connection with connection_registry_id = 40105: 30,000 entries with this id and obj1 = 73582: 1 entries in Connection with connection_registry_id = 40003: 6,000 entries with this id and obj1 = 4666: 20 but as I said before, there is an btree index on (connection_registry_id, obj1) Explain: The first query, breaks down as: Hash Join (cost=5.96..7.04 rows=1 width=14) Hash Cond: ("outer".id = "inner".obj2) -> Seq Scan on rtmessagestate (cost=0.00..1.05 rows=5 width=14) -> Hash (cost=5.96..5.96 rows=1 width=4) -> Index Scan using connection_regid_obj1_index on connection (cost=0.00..5.96 rows=1 width=4) Index Cond: ((connection_registry_id = 40105) AND (obj1 = 73582))(6 rows) While the second query is: Nested Loop (cost=0.00..11.62 rows=2 width=38) -> Index Scan using connection_regid_obj1_index on connection (cost=0.00..5.96 rows=1 width=4) Index Cond: ((connection_registry_id = 40003) AND (obj1 = 4666)) -> Index Scan using rtmessage_pkey on rtmessage (cost=0.00..5.65 rows=1 width=38) Index Cond: ("outer".obj2 = rtmessage.id) (5 rows) Actually running these queries shows that the second one (nested loop) is much faster than the hash join, presumably because of hash startup costs. Any ideas how I can make them both use the nested loop. I assume that this would be the fastest for both. Oddly enough, running the 1st query (rtmessagestate) as two queries or with a sub query is way faster than doing the join. And yes, I realize this schema may not be the most efficient for these examples, but it seems to be the most flexible. I'm working on some schema improvements also but if I could understand why this is slow that woudl probably help also. Thanks for you help, Rhett ---------------------------(end of broadcast)--------------------------- TIP 3: Have you checked our extensive FAQ? http://www.postgresql.org/docs/faq From pgsql-performance-owner@postgresql.org Fri Aug 5 20:17:03 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 296D65286E for ; Fri, 5 Aug 2005 20:17:02 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 77157-07 for ; Fri, 5 Aug 2005 23:16:51 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.195]) by svr1.postgresql.org (Postfix) with ESMTP id 78B9952A07 for ; Fri, 5 Aug 2005 20:16:50 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i8so670635wra for ; Fri, 05 Aug 2005 16:16:51 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=gaNGSJjfzitffKDFJqemxGJ7iBw1JZLXyiTg9M6rf7HYm+xa2Vm/Su+eQq2Uv0Psa4ceFUYNii91D94Ut4hdniuvuBX1eCYja/mZEVnh1JCyftRmp1aSto7kOYi8Hib7R7npjJz5NvQdJvzvuUG2W0QHEIXzsWh0Oct6S2AkxyQ= Received: by 10.54.27.46 with SMTP id a46mr3029134wra; Fri, 05 Aug 2005 16:16:51 -0700 (PDT) Received: by 10.54.119.7 with HTTP; Fri, 5 Aug 2005 16:16:51 -0700 (PDT) Message-ID: <41b0fe8905080516164464e04f@mail.gmail.com> Date: Fri, 5 Aug 2005 16:16:51 -0700 From: Rhett Garber Reply-To: Rhett Garber To: =?ISO-8859-1?Q?Havasv=F6lgyi_Ott=F3?= Subject: Re: Why hash join instead of nested loop? Cc: pgsql-performance@postgresql.org In-Reply-To: <001801c59a0a$f4be7480$9a00a8c0@OTTO> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <41b0fe890508051135712a086b@mail.gmail.com> <001801c59a0a$f4be7480$9a00a8c0@OTTO> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 tagged_above=0 required=5 tests=RCVD_BY_IP X-Spam-Level: X-Archive-Number: 200508/52 X-Sequence-Number: 13799 On 8/5/05, Havasv=F6lgyi Ott=F3 wrote: > Please post the explain analyze for both queries. From that we can see th= e > predicted and the actual costs of them. > select rtmessagestate.* from rtmessagestate, connection where > connection_registry_id =3D 40105 and obj1 =3D 73582 and obj2 =3D > rtmessagestate.id; Hash Join (cost=3D5.96..7.04 rows=3D1 width=3D14) (actual time=3D10.591..10.609 rows=3D1 loops=3D1) Hash Cond: ("outer".id =3D "inner".obj2) -> Seq Scan on rtmessagestate (cost=3D0.00..1.05 rows=3D5 width=3D14) (actual time=3D0.011..0.022 rows=3D5 loops=3D1) -> Hash (cost=3D5.96..5.96 rows=3D1 width=3D4) (actual time=3D0.109..0.109 rows=3D0 loops=3D1) -> Index Scan using connection_regid_obj1_index on connection (cost=3D0.00..5.96 rows=3D1 width=3D4) (actual time=3D0.070..0.= 076 rows=3D1 loops=3D1) Index Cond: ((connection_registry_id =3D 40105) AND (obj1 =3D 73582)) Total runtime: 11.536 ms (7 rows) > select rtmessage.id, subject from rtmessage, connection where > connection_registry_id =3D 40003 and obj1 =3D 4666 and obj2 =3D > rtmessage.id; Nested Loop (cost=3D0.00..11.62 rows=3D2 width=3D38) (actual time=3D0.186..0.970 rows=3D12 loops=3D1) -> Index Scan using connection_regid_obj1_index on connection=20 (cost=3D0.00..5.96 rows=3D1 width=3D4) (actual time=3D0.109..0.308 rows=3D1= 2 loops=3D1) Index Cond: ((connection_registry_id =3D 40003) AND (obj1 =3D 4666= )) -> Index Scan using rtmessage_pkey on rtmessage (cost=3D0.00..5.65 rows=3D1 width=3D38) (actual time=3D0.032..0.039 rows=3D1 loops=3D12) Index Cond: ("outer".obj2 =3D rtmessage.id) Total runtime: 1.183 ms (6 rows) Rhett From pgsql-performance-owner@postgresql.org Sat Aug 6 02:04:22 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AB1CE528B8 for ; Sat, 6 Aug 2005 02:04:20 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 57096-05 for ; Sat, 6 Aug 2005 05:04:10 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 18A1E52A38 for ; Sat, 6 Aug 2005 02:04:09 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7653pPB015426; Sat, 6 Aug 2005 01:03:51 -0400 (EDT) To: Rhett Garber Cc: =?ISO-8859-1?Q?Havasv=F6lgyi_Ott=F3?= , pgsql-performance@postgresql.org Subject: Re: Why hash join instead of nested loop? In-reply-to: <41b0fe8905080516164464e04f@mail.gmail.com> References: <41b0fe890508051135712a086b@mail.gmail.com> <001801c59a0a$f4be7480$9a00a8c0@OTTO> <41b0fe8905080516164464e04f@mail.gmail.com> Comments: In-reply-to Rhett Garber message dated "Fri, 05 Aug 2005 16:16:51 -0700" Date: Sat, 06 Aug 2005 01:03:51 -0400 Message-ID: <15425.1123304631@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/53 X-Sequence-Number: 13800 Rhett Garber writes: > Hash Join (cost=5.96..7.04 rows=1 width=14) (actual > time=10.591..10.609 rows=1 loops=1) > Hash Cond: ("outer".id = "inner".obj2) > -> Seq Scan on rtmessagestate (cost=0.00..1.05 rows=5 width=14) > (actual time=0.011..0.022 rows=5 loops=1) > -> Hash (cost=5.96..5.96 rows=1 width=4) (actual > time=0.109..0.109 rows=0 loops=1) > -> Index Scan using connection_regid_obj1_index on > connection (cost=0.00..5.96 rows=1 width=4) (actual time=0.070..0.076 > rows=1 loops=1) > Index Cond: ((connection_registry_id = 40105) AND (obj1 > = 73582)) Total runtime: 11.536 ms > (7 rows) [ scratches head... ] If the hash table build takes only 0.109 msec and loads only one row into the hash table, and the scan of rtmessagestate takes only 0.022 msec and produces only 5 rows, it is real hard to see how the join takes 10.609 msec overall. Unless the id and obj2 columns are of a datatype with an incredibly slow equality function. What is the datatype involved here, anyway? And what PG version are we speaking of? regards, tom lane From pgsql-performance-owner@postgresql.org Sat Aug 6 10:16:14 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id F0E9B52A81 for ; Sat, 6 Aug 2005 10:16:12 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 59717-09 for ; Sat, 6 Aug 2005 13:16:03 +0000 (GMT) Received: from sccrmhc13.comcast.net (sccrmhc13.comcast.net [204.127.202.64]) by svr1.postgresql.org (Postfix) with ESMTP id 5C64F52A73 for ; Sat, 6 Aug 2005 10:16:02 -0300 (ADT) Received: from [127.0.0.1] (c-24-4-66-2.hsd1.ca.comcast.net[24.4.66.2]) by comcast.net (sccrmhc13) with ESMTP id <2005080613160501300dkqnbe>; Sat, 6 Aug 2005 13:16:05 +0000 Message-ID: <42F4B812.7070501@comcast.net> Date: Sat, 06 Aug 2005 06:16:02 -0700 From: Patrick Hatcher User-Agent: Mozilla Thunderbird 1.0.2 (Windows/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: "pgsql-performance@postgresql.org" Subject: Slow update statement Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=2.22 tagged_above=0 required=5 tests=DNS_FROM_RFC_ABUSE, DNS_FROM_RFC_POST, TO_ADDRESS_EQ_REAL X-Spam-Level: ** X-Archive-Number: 200508/54 X-Sequence-Number: 13801 [Reposted from General section with updated information] Pg 7.4.5 I'm running an update statement on about 12 million records using the following query: Update table_A set F1 = b.new_data from table_B b where b.keyfield = table_A.keyfield both keyfields are indexed, all other keys in table_A were dropped, yet this job has been running over 15 hours. Is this normal? I stopped the process the first time after 3 hours of running due to excessive log rotation and reset the conf file to these settings: wal_buffers = 64 # min 4, 8KB each # - Checkpoints - checkpoint_segments = 128 # in logfile segments, min 1, 16MB each checkpoint_timeout = 1800 # range 30-3600, in seconds #checkpoint_warning = 30 # 0 is off, in seconds #commit_delay = 0 # range 0-100000, in microseconds #commit_siblings = 5 # range 1-1000 Would it just be quicker to run a JOIN statement to a temp file and then reinsert? TIA Patrick From pgsql-performance-owner@postgresql.org Sat Aug 6 10:34:37 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 354A152A73 for ; Sat, 6 Aug 2005 10:34:36 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 71348-03 for ; Sat, 6 Aug 2005 13:34:34 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id 8369C52A71 for ; Sat, 6 Aug 2005 10:34:33 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050806133436m9100g6kqae>; Sat, 6 Aug 2005 13:34:36 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id D437656053; Sat, 6 Aug 2005 08:34:35 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id BEC3956016; Sat, 6 Aug 2005 08:34:31 -0500 (CDT) Message-ID: <42F4BC61.20102@arbash-meinel.com> Date: Sat, 06 Aug 2005 08:34:25 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Patrick Hatcher Cc: "pgsql-performance@postgresql.org" Subject: Re: Slow update statement References: <42F4B812.7070501@comcast.net> In-Reply-To: <42F4B812.7070501@comcast.net> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigB102F44065A4841D42023663" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.045 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/55 X-Sequence-Number: 13802 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigB102F44065A4841D42023663 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Patrick Hatcher wrote: > [Reposted from General section with updated information] > Pg 7.4.5 > > I'm running an update statement on about 12 million records using the > following query: > > Update table_A > set F1 = b.new_data > from table_B b > where b.keyfield = table_A.keyfield > > both keyfields are indexed, all other keys in table_A were dropped, yet > this job has been running over 15 hours. Is > this normal? Can you do an EXPLAIN UPDATE so that we can have an idea what the planner is trying to do? My personal concern is if it doing something like pulling in all rows from b, and then one by one updating table_A, but as it is going, it can't retire any dead rows, because you are still in a transaction. So you are getting a lot of old rows, which it has to pull in to realize it was old. How many rows are in table_B? I can see that possibly doing it in smaller chunks might be faster, as would inserting into another table. But I would do more of a test and see what happens. John =:-> > > I stopped the process the first time after 3 hours of running due to > excessive log rotation and reset the conf file to these settings: > > > wal_buffers = 64 # min 4, 8KB each > > # - Checkpoints - > > checkpoint_segments = 128 # in logfile segments, min 1, 16MB each > checkpoint_timeout = 1800 # range 30-3600, in seconds > #checkpoint_warning = 30 # 0 is off, in seconds > #commit_delay = 0 # range 0-100000, in microseconds > #commit_siblings = 5 # range 1-1000 > > > Would it just be quicker to run a JOIN statement to a temp file and then > reinsert? > TIA Patrick > > > ---------------------------(end of broadcast)--------------------------- > TIP 9: In versions below 8.0, the planner will ignore your desire to > choose an index scan if your joining column's datatypes do not > match > --------------enigB102F44065A4841D42023663 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC9LxlJdeBCYSNAAMRAtL9AJ98wz0ruVi1wics+ko476ylNgbT3QCffWN8 arXwBSyM/BCE48I0BmunjUA= =KQQn -----END PGP SIGNATURE----- --------------enigB102F44065A4841D42023663-- From pgsql-performance-owner@postgresql.org Sat Aug 6 11:12:51 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E1B075292D for ; Sat, 6 Aug 2005 11:12:49 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78521-02 for ; Sat, 6 Aug 2005 14:12:39 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 26C4C52AB1 for ; Sat, 6 Aug 2005 11:12:38 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j76EChYV017917; Sat, 6 Aug 2005 10:12:43 -0400 (EDT) To: Patrick Hatcher Cc: "pgsql-performance@postgresql.org" Subject: Re: Slow update statement In-reply-to: <42F4B812.7070501@comcast.net> References: <42F4B812.7070501@comcast.net> Comments: In-reply-to Patrick Hatcher message dated "Sat, 06 Aug 2005 06:16:02 -0700" Date: Sat, 06 Aug 2005 10:12:43 -0400 Message-ID: <17916.1123337563@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/56 X-Sequence-Number: 13803 Patrick Hatcher writes: > I'm running an update statement on about 12 million records using the > following query: > Update table_A > set F1 = b.new_data > from table_B b > where b.keyfield = table_A.keyfield What does EXPLAIN show for this? Do you have any foreign key references to table_A from elsewhere? regards, tom lane From pgsql-patches-owner@postgresql.org Sat Aug 6 18:05:05 2005 X-Original-To: pgsql-patches-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B585052B73 for ; Sat, 6 Aug 2005 18:05:03 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 81449-03 for ; Sat, 6 Aug 2005 21:04:53 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 4679D52B71 for ; Sat, 6 Aug 2005 18:04:52 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j76L4qVg026468; Sat, 6 Aug 2005 17:04:53 -0400 (EDT) To: "Alon Goldshuv" Cc: pgsql-patches@postgresql.org Subject: Re: COPY FROM performance improvements In-reply-to: References: Comments: In-reply-to "Alon Goldshuv" message dated "Tue, 02 Aug 2005 10:59:43 -0400" Date: Sat, 06 Aug 2005 17:04:52 -0400 Message-ID: <26467.1123362292@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/46 X-Sequence-Number: 16993 "Alon Goldshuv" writes: > New patch attached. It includes very minor changes. These are changes that > were committed to CVS 3 weeks ago (copy.c 1.247) which I missed in the > previous patch. I've applied this with (rather extensive) revisions. I didn't like what you had done with the control structure --- loading the input buffer only at the outermost loop level was a bad design choice IMHO. You had sprinkled the code with an unreasonable number of special cases in order to try to cope with the effects of that mistake, but there were lots of problems still left. Some of the bugs I noticed: * Broke old-protocol COPY, since that has no provision for stopping at the EOF marker except by parsing the data carefully to start with. The backend would just hang up unless the total data size chanced to be a multiple of 64K. * Subtle change in interpretation of \. EOF marker (the existing code will recognize it even when not at start of line). * Seems to have thrown away detection of newline format discrepancies. * Fails for zero-column tables. * Broke display of column values during error context callback (would always show the first column contents no matter which one is being complained of). * DetectLineEnd mistakenly assumes CR mode if very last character of first bufferload is CR; need to reserve judgment until next char is available. * DetectLineEnd fails to account for backslashed control characters, so it will e.g. accept \ followed by \n as determining the newline style. * Fails to apply encoding conversion if first line exceeds copy buf size, because when DetectLineEnd fails the quick-exit path doesn't do it. * There seem to be several bugs associated with the fact that input_buf[] always has 64K of data in it even when EOF has been reached on the input. One example: echo -n 123 >zzz1 psql> create temp table t1(f1 text); psql> copy t1 from '/home/tgl/zzz1'; psql> select * from t1; hmm ... where'd that 64K of whitespace come from? I rewrote the patch in a way that retained most of the speedups without changing the basic control structure (except for replacing multiple CopyReadAttribute calls with one CopyReadAttributes call per line). I had some difficulty in generating test cases that weren't largely I/O-bound, but AFAICT the patch as applied is about the same speed as what you submitted. regards, tom lane From pgsql-patches-owner@postgresql.org Sat Aug 6 23:33:20 2005 X-Original-To: pgsql-patches-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 465D652B48 for ; Sat, 6 Aug 2005 23:33:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 50369-10 for ; Sun, 7 Aug 2005 02:33:14 +0000 (GMT) Received: from mail.Mi8.com (nycgw05.mi8.com [63.240.6.46]) by svr1.postgresql.org (Postfix) with ESMTP id AA98252B2E for ; Sat, 6 Aug 2005 23:33:12 -0300 (ADT) Received: from 172.16.1.118 by mail.Mi8.com with ESMTP (- GW05 Welcome to Mi8 Corporation www.Mi8.com); Sat, 06 Aug 2005 22:33:08 -0400 X-Server-Uuid: E0C866E6-C6CD-48B5-AE61-E57E73CF3CC7 Received: from MI8NYCMAIL06.Mi8.com ([172.16.1.175]) by d01smtp03.Mi8.com with Microsoft SMTPSVC(6.0.3790.1830); Sat, 6 Aug 2005 22:33:08 -0400 Received: from 24.5.173.15 ([24.5.173.15]) by MI8NYCMAIL06.Mi8.com ( [172.16.1.219]) via Exchange Front-End Server mi8owa.mi8.com ( [172.16.1.106]) with Microsoft Exchange Server HTTP-DAV ; Sat, 6 Aug 2005 22:33:07 -0500 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Sat, 06 Aug 2005 19:33:07 -0700 Subject: Re: COPY FROM performance improvements From: "Luke Lonergan" To: "Tom Lane" , "Alon Goldshuv" Cc: pgsql-patches@postgresql.org Message-ID: In-Reply-To: <26467.1123362292@sss.pgh.pa.us> MIME-Version: 1.0 X-OriginalArrivalTime: 07 Aug 2005 02:33:08.0216 (UTC) FILETIME=[58CA7F80:01C59AF8] X-WSS-ID: 6EEBAD6E26S3784053-01-01 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.594 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO, RCVD_NUMERIC_HELO X-Spam-Level: * X-Archive-Number: 200508/47 X-Sequence-Number: 16994 Tom, Thanks for finding the bugs and reworking things. > I had some difficulty in generating test cases that weren't largely > I/O-bound, but AFAICT the patch as applied is about the same speed > as what you submitted. You achieve the important objective of knocking the parsing stage down a lot, but your parsing code is actually about 20% slower than Alon's. Before your patch: Time: 14205.606 ms With your patch: Time: 10565.374 ms With Alon's patch: Time: 10289.845 ms The parsing part of the code in your version is slower, but as a percentage of the total it's hidden. The loss of 0.3 seconds on 143MB means: - If parsing takes a total of 0.9 seconds, the parsing rate is 160MB/s (143/0.9) - If we add another 0.3 seconds to parsing to bring it to 1.2, then the parsing rate becomes 120MB/s When we improve the next stages of the processing (attribute conversion, write-to disk), this will stand out a lot more. Our objective is to get the COPY rate *much* faster than the current poky rate of 14MB/s (after this patch). - Luke On 8/6/05 2:04 PM, "Tom Lane" wrote: > "Alon Goldshuv" writes: >> New patch attached. It includes very minor changes. These are changes that >> were committed to CVS 3 weeks ago (copy.c 1.247) which I missed in the >> previous patch. > > I've applied this with (rather extensive) revisions. I didn't like what > you had done with the control structure --- loading the input buffer > only at the outermost loop level was a bad design choice IMHO. You had > sprinkled the code with an unreasonable number of special cases in order > to try to cope with the effects of that mistake, but there were lots > of problems still left. Some of the bugs I noticed: > > * Broke old-protocol COPY, since that has no provision for stopping at > the EOF marker except by parsing the data carefully to start with. The > backend would just hang up unless the total data size chanced to be a > multiple of 64K. > > * Subtle change in interpretation of \. EOF marker (the existing code > will recognize it even when not at start of line). > > * Seems to have thrown away detection of newline format discrepancies. > > * Fails for zero-column tables. > > * Broke display of column values during error context callback (would > always show the first column contents no matter which one is being > complained of). > > * DetectLineEnd mistakenly assumes CR mode if very last character of first > bufferload is CR; need to reserve judgment until next char is available. > > * DetectLineEnd fails to account for backslashed control characters, > so it will e.g. accept \ followed by \n as determining the newline > style. > > * Fails to apply encoding conversion if first line exceeds copy buf > size, because when DetectLineEnd fails the quick-exit path doesn't do > it. > > * There seem to be several bugs associated with the fact that input_buf[] > always has 64K of data in it even when EOF has been reached on the > input. One example: > echo -n 123 >zzz1 > psql> create temp table t1(f1 text); > psql> copy t1 from '/home/tgl/zzz1'; > psql> select * from t1; > hmm ... where'd that 64K of whitespace come from? > > I rewrote the patch in a way that retained most of the speedups without > changing the basic control structure (except for replacing multiple > CopyReadAttribute calls with one CopyReadAttributes call per line). > > I had some difficulty in generating test cases that weren't largely > I/O-bound, but AFAICT the patch as applied is about the same speed > as what you submitted. > > regards, tom lane > > ---------------------------(end of broadcast)--------------------------- > TIP 2: Don't 'kill -9' the postmaster > From pgsql-patches-owner@postgresql.org Sat Aug 6 23:54:26 2005 X-Original-To: pgsql-patches-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2717C52BBC for ; Sat, 6 Aug 2005 23:54:14 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49581-06 for ; Sun, 7 Aug 2005 02:54:10 +0000 (GMT) Received: from mail.mi8.com (d01gw04.mi8.com [63.240.6.44]) by svr1.postgresql.org (Postfix) with ESMTP id A144E52BC1 for ; Sat, 6 Aug 2005 23:54:08 -0300 (ADT) Received: from 172.16.1.110 by mail.mi8.com with ESMTP (- Welcome to Mi8 Corporation www.Mi8.com (D4)); Sat, 06 Aug 2005 22:54:05 -0400 X-Server-Uuid: C8FB4D43-1108-484A-A898-3CBCC7906230 Received: from MI8NYCMAIL06.Mi8.com ([172.16.1.175]) by D01SMTP01.Mi8.com with Microsoft SMTPSVC(6.0.3790.211); Sat, 6 Aug 2005 22:54:05 -0400 Received: from 24.5.173.15 ([24.5.173.15]) by MI8NYCMAIL06.Mi8.com ( [172.16.1.219]) via Exchange Front-End Server mi8owa.mi8.com ( [172.16.1.106]) with Microsoft Exchange Server HTTP-DAV ; Sat, 6 Aug 2005 22:54:04 -0500 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Sat, 06 Aug 2005 19:54:04 -0700 Subject: Re: COPY FROM performance improvements From: "Luke Lonergan" To: "Tom Lane" , "Alon Goldshuv" Cc: pgsql-patches@postgresql.org Message-ID: In-Reply-To: <26467.1123362292@sss.pgh.pa.us> MIME-Version: 1.0 X-OriginalArrivalTime: 07 Aug 2005 02:54:05.0519 (UTC) FILETIME=[4633B5F0:01C59AFB] X-WSS-ID: 6EEBA8472X4762688-01-01 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.593 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO, RCVD_NUMERIC_HELO X-Spam-Level: * X-Archive-Number: 200508/48 X-Sequence-Number: 16995 Tom, The previous timings were for a table with 15 columns of mixed type. We also test with 1 column to make the parsing overhead more apparent. In the case of 1 text column with 145MB of input data: Your patch: Time: 6612.599 ms Alon's patch: Time: 6119.244 ms Alon's patch is 7.5% faster here, where it was only 3% faster on the 15 column case. This is consistent with a large difference in parsing speed between your approach and Alon's. I'm pretty sure that the "mistake" you refer to is responsible for the speed improvement, and was deliberately chosen to minimize memory copies, etc. Given that we're looking ahead to getting much higher speeds, approaching current high performance disk speeds, we've been looking more closely at the parsing speed. It comes down to a tradeoff between elegant code and speed. We'll prove it in lab tests soon, where we measure the parsing rate directly, but these experiments show it clearly, though indirectly. - Luke On 8/6/05 2:04 PM, "Tom Lane" wrote: > "Alon Goldshuv" writes: >> New patch attached. It includes very minor changes. These are changes that >> were committed to CVS 3 weeks ago (copy.c 1.247) which I missed in the >> previous patch. > > I've applied this with (rather extensive) revisions. I didn't like what > you had done with the control structure --- loading the input buffer > only at the outermost loop level was a bad design choice IMHO. You had > sprinkled the code with an unreasonable number of special cases in order > to try to cope with the effects of that mistake, but there were lots > of problems still left. Some of the bugs I noticed: > > * Broke old-protocol COPY, since that has no provision for stopping at > the EOF marker except by parsing the data carefully to start with. The > backend would just hang up unless the total data size chanced to be a > multiple of 64K. > > * Subtle change in interpretation of \. EOF marker (the existing code > will recognize it even when not at start of line). > > * Seems to have thrown away detection of newline format discrepancies. > > * Fails for zero-column tables. > > * Broke display of column values during error context callback (would > always show the first column contents no matter which one is being > complained of). > > * DetectLineEnd mistakenly assumes CR mode if very last character of first > bufferload is CR; need to reserve judgment until next char is available. > > * DetectLineEnd fails to account for backslashed control characters, > so it will e.g. accept \ followed by \n as determining the newline > style. > > * Fails to apply encoding conversion if first line exceeds copy buf > size, because when DetectLineEnd fails the quick-exit path doesn't do > it. > > * There seem to be several bugs associated with the fact that input_buf[] > always has 64K of data in it even when EOF has been reached on the > input. One example: > echo -n 123 >zzz1 > psql> create temp table t1(f1 text); > psql> copy t1 from '/home/tgl/zzz1'; > psql> select * from t1; > hmm ... where'd that 64K of whitespace come from? > > I rewrote the patch in a way that retained most of the speedups without > changing the basic control structure (except for replacing multiple > CopyReadAttribute calls with one CopyReadAttributes call per line). > > I had some difficulty in generating test cases that weren't largely > I/O-bound, but AFAICT the patch as applied is about the same speed > as what you submitted. > > regards, tom lane > > ---------------------------(end of broadcast)--------------------------- > TIP 2: Don't 'kill -9' the postmaster > From pgsql-patches-owner@postgresql.org Sun Aug 7 00:25:48 2005 X-Original-To: pgsql-patches-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A0F2D5282E for ; Sun, 7 Aug 2005 00:25:43 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 79905-03 for ; Sun, 7 Aug 2005 03:25:31 +0000 (GMT) Received: from mail.Mi8.com (d01gw01.mi8.com [63.240.6.47]) by svr1.postgresql.org (Postfix) with ESMTP id 5676652AE0 for ; Sun, 7 Aug 2005 00:25:29 -0300 (ADT) Received: from 172.16.1.110 by mail.Mi8.com with ESMTP (- Welcome to Mi8 Corporation www.Mi8.com (D1)); Sat, 06 Aug 2005 23:25:24 -0400 X-Server-Uuid: 241911D6-425B-44B9-A073-E3FE0F8FC774 Received: from MI8NYCMAIL06.Mi8.com ([172.16.1.175]) by D01SMTP01.Mi8.com with Microsoft SMTPSVC(6.0.3790.211); Sat, 6 Aug 2005 23:25:24 -0400 Received: from 24.5.173.15 ([24.5.173.15]) by MI8NYCMAIL06.Mi8.com ( [172.16.1.219]) via Exchange Front-End Server mi8owa.mi8.com ( [172.16.1.106]) with Microsoft Exchange Server HTTP-DAV ; Sat, 6 Aug 2005 23:25:24 -0500 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Sat, 06 Aug 2005 20:25:24 -0700 Subject: Re: COPY FROM performance improvements From: "Luke Lonergan" To: "PostgreSQL-patches" Message-ID: In-Reply-To: <26467.1123362292@sss.pgh.pa.us> MIME-Version: 1.0 X-OriginalArrivalTime: 07 Aug 2005 03:25:24.0531 (UTC) FILETIME=[A62E3030:01C59AFF] X-WSS-ID: 6EEBA0AE2B414362915-01-01 Content-Type: multipart/mixed; boundary=B_3206204724_2604846 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.592 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO, RCVD_NUMERIC_HELO X-Spam-Level: * X-Archive-Number: 200508/49 X-Sequence-Number: 16996 > This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. --B_3206204724_2604846 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Tom, My direct e-mails to you are apparently blocked, so I'll send this to the list. I've attached the case we use for load performance testing, with the data generator modified to produce a single row version of the dataset. I do believe that you/we will need to invert the processing loop to get the maximum parsing speed. We will be implementing much higher loading speeds which require it to compete with Oracle, Netezza, Teradata, so we'll have to work this out for the best interests of our users. - Luke --B_3206204724_2604846 Content-Type: application/octet-stream; name=IVP.tgz Content-Disposition: attachment; filename=IVP.tgz Content-Transfer-Encoding: base64 H4sIAJx59UIAA+w8aXfbxrX5KvyKG8aOSJkiCXCRTUVtFUlx2Go7kuLk1dZRQWBIjgViGAwg imn939+9s4AgRcr2q+O+nhjHpoCZu83dZgV6r87rX/3OV6PRauy02/i30djptBb+musrt+Ht tL2dVtNDONfzOs2voP17C0ZXJlM/AfgqikTMkqEfr4F7X/1/6dVD+wcJ81N2E/ZrcvR78Gi4 jcfs32p5O2R/b6fTpkeEb7U9tH/j9xBm+fqD2/+br+t9Htf7vhw5DgtGYv4DVyMuAf/5MfRi VFMU+SkXMbzyIx7q2/NEDBN/DAORwCFjk5cJY3FNo/dSmPIoAq5xwQfpjycRmwPCoZ/6yJoB kkL1wtgPRjxmtYIQ+4OUJZCSKDJI+CSFJIvxPgsCJuUgi6JZFWYiAzkSWRTCyL9jyAqBYh4P V/GK/TELNfFSOEyZTMN+qYos/FQRCrC9gYhjFqSQCsgk0UlHDAvHYz8OuwYXYCJ/jSAnURD6 fwwZP5ICIuGHtumoNh8VkhZZP8ahVifsbULD4CyyKPUGC1oZ+DySJC+7Z0GWsiryFbfISxEe cGT+poTRXovE8E1JGYwqMPQlKn86mtUK1B0Zod7Ac5xkDNsDMHiOU4OJn45q5CzfXLIUsomi MhFJCnE27rNEKtIiS2yTgwhDjCVd5/Dlj0KmN+dnF1d7bUz0ji05Pnu5RxysRjSrb3oxooZM y5/FATmc1GaS2USxLCighrLxu8lNDlnTPn3w8+HePybT8B+LEitDhDxBI4uEs/dKfdi72HuC tOpWSMfhA3gN2yE8mYPA9S5Rjx0AZaTtGEoXbCzuyMCThN1xkUkbECqCqiBu/wzldDZhsDlj cpNMOElEwFhYgRISQguFGIISbxXHJ3gPe1BC4FKBH4BMxUTFaRywBaGs9QgoIWnYoWn4bA0c iyRT4NrVrlgy5jHKi60w7qZjJUF6v2aoj1qtVtII9zwFF28H3MF/OqcoPRxQL2Md3Wpx0Q4z Q0Z3SBSw4aOCmigJMXUg3pzVJab0nBW3KolEgFRDComcvWFIfUD6QHfkqQt8n8wdFsRg8F45 Fpts88+bPPQxDo0AmK4uf40OxuEi85LWxBx3njRShn6Kde5HSOH3VRbo86G6dZE/KqMoz4eJ I4MRG/vEeNeUaNIkSU4cypiGUz5GyhhSVejDnZ8EIz+pQjC/JXOkbMjwls1vB/Pb4fx2RCmG JTyoAp9TeDu/vZ3fRnO08fw2ngMIewuV0twfHtPl5va/d23+PyRjGnogYuzG00znJNmF/ULH mTvfyJfQpxKTwDCWit3w1wvdk+0CYzEt9qYLkfCw4yst9q3bE6Ce4tE+lhh88i72Pz0s+2wX jf+p6dtDhsNbH1Ptp58OftT8z6Xxf9ttd77M/z7HtcL+B68uP60PfJj9Gx20vdfo0Py/3Wm4 X+z/Oa419j+KUxoVfxoeNP/vtFpr7d90d5bsv+M2v8z/P8tVlzjRoNlhLb1P627Nq/+QcDjx Z+A9h4bXbb3othvgYQjX6049S3GSWQtWwbW6zZ0luNEDuGa30ei2OjnciX/LiDkCNuuXWQx/ xf9uEzy367aLBMc+jxXf9kMwutFgh3+cfvtTXWvi/4JNhOQ08foEPN4X/w0d/+1220VQl+K/ g2ngS/x/hmvR9l/i5492rYt/IdJPxuO98Y+d/WL8d2hL4Ev8f4ary+7TLh/GTGSyWx+JMasH dzJB83/JBX+Ea0X867HWJ+TxePzjZK/l2vm/12jtUPx77S/j/89yfcPNLst3Mg25qI3+5CwU Rby/XJbweEhl88KSGe+XsCzEAX3MoHd6dfTy6AIaecmr/YuDH/cvwM1LrnonR5dX+yfn4OVl pz+dHF30DqA5J3Wy/8tx7/Tosvf3I0A3aTkOj1MYsnTG/KRM9+jArLJri8ciTkcrykN/Vix1 7gQP9cL1DTp/Ge07BPUTJ2Iqq3CpWhpxmcKUfquEHGfjQERYSw9beJuN4xvavcGiUGS0Cm5L xzxeUejfM0ncCZ8iTYnkJ8OgCmo1emsLH+4q8E8HgLZ8sJZXIxYP0xFigQbSRnhd0Mx1FbYi EewiREFuSb9LZVNdhoWFFmdjarMqLbRy1zzivW7ja4zkayo1raIKaqYqr6onap8Bs9JOaEsu U0vLr59fwx78EzZhswqbNfqp08+bN/Rbpp8K/VRVYWMT3ikyMvHjsOx6FfX0Q+/4CLZ4TDNH LfIAymXS4bZbge/A09oDmGCj00G5lEl/yLrwVEKK3d0NTXlveDzJ0hsicUMbwmb38kYMbkgT N6m4MRmRvYlLVbLJ68Z1ZTff5Cq76uGd5Q9lLdDeQEzQmxSCe10tJaVKBfb20LGPj2FZMJYk tO2ICLQ8rTZpn+qtSNr1w7Kct7vA29O80UMMdykDPx5orh5yfRpFvFT91ti1Al/vgbuaOSEq 7loBIAZAKDBIxNgulkOEcUj6y8Xx1ogDUK9b+qeLBFMBfQZWqSGSi0KiZ2XcVciKXkPbWXlv jX72SHu7piCTaD7Jf2N7DVsUjLL4VhWh59lCH1N2YAGJ3nRECkZLDTAbyLIOomoxv1BYkxUq SmPElJSmGlqv/8BRE7SFMNcUubcfpLTrbTbaLVGTGqVGYFPSIPQzDC2RqrI3DUVWRzbGBCLi vRHKKLdev0r4WG01Ekqa+DxSptL0zK6vSgWazrZ7vUdhs2tlPoplljC9aY4kErZJRycGAx5w hoEtJ37AlL/5KBfWG3K2NaRHzWVR88+eaQG1isciYTcyzzDlbxV0dQmnkgu1H+mt2AJ7xUyr QgwKilQYW2VNaiJkZa+ss2RlrHiXibIYqMLKlqYwZ3QgJrMCMb0rtNiuNAkms3KRQzU3gaHS i4OEjUldBVITQZuKid3SijGvPK47JE1asymjXr9gkq0i6RQwyC/yINjVeL2Yp9yPsN0KeWr4 TJeDZfowWKargmW6Oljq9XM/kawooOo+lA6nIgnpQAKZrsyJNP9uydxY9OxZIXo+2hOJxwNd Tj/KD3W3PV3th4ihwy4Vt+W5ll/z66r1sULHZbEo2RJmIT0AfIRbU6sMxnLoI3blmbtrarfK 0490eo2Xe/QcvUqUbf206I1U8E43LE+OC3qhNq5UR2WVBj6Njdda+aPt/F9vGZsstnRAUsea y2r0VQzB6ZoQ1Fqw/bIeC3bhTQk79BJ1wcX0l3c9clmc9fjz3POahrMVGq2JMf4OWbmxbBcc o13nTD6IJr82Qy3UfT3XB2zBJUvNcS59/iTvjhGLB1IDbZjB7MbGXrHnVsNxDbLhJ4k/k1RM AtrRb1fj47VRHOcjGXVaKydSxXH8ZMJC6gvMSRODu7GxYeZB9tFMguyjme3Yx3xCZAoKEwlA tji2hjs/ytiDBmwUZxePQSr1mfahfyufno/uG9eAhbkUu/Pxva6hbRZaoTMVaqg/r3G95kp6 ps0PqbVW0OksUnAXKUBhxqFq3F0wV2HyoWp22o069W8L5Ly15Ly15ExNYwW5pqoyJl4gp2sa K8jZmgVCrbWEWmvl0jWtJUrttZTaa0VqrxKps5ZQZ61Iuqa1SGhHFRpPXyCkaxq1hre7RMjW PF8k9Xyt+Z6vlcnUtNF+zWX7vVhL78Vaei8eoec2rtd5q6pZ6a26ZjU993qNDXTNKmPqGm+J kLeWkLeWkKpZMqbbXNvC5toWNh9pYWstvdZaeq1FepTS5ss4ZjpZNWs2dr0mZ1m1HObLFZVd NY/Op57v/qNrQ3rIrNZ96IynXYPh1bdVzqsQq7G3LaUlMOx/aMULyfmzKoxERiU8VmfAJQtE HFpoLDUIPDY4eKfQ8G+OaZFJyEV8/97g+/cW37/X+P69wccq/17j+/cW3xCImD+Rr8l8x0f7 55eWbiziec3p2amp/D+tJqxZSygOkxTMwvxE1b6l2rff2bUveDsfPMkpT4MRlIs2ff32ej72 Deg4o4mvrmN9di59AfH107AQjCTz22e0lrRhYZ+GpeqDYVTBb5BxtegxJAjh9xPm3+7uFiUy gfVhEhnggkTa17DqY8VBXK1wrXFUuaZE91qrG6DdexXxVUPGXcLINSQXx4dE6FqD0PyMc5qc lDVPxK3kiFDKR9s0nlylsnz082FKy8ELajNhhrV2kXpRXZhvTCStACmo0EaphtLL2qsorQVa pIVxqoFoIXwVnZUAcxrOhhJ5Y4XFbGIxzVIMlVBrgHXysKITOPFeA0z5RUtHgGTg3iVliLJi ZZxJy37SOy3rTKIIb7vXVY2W21y92rCMkuefdVjOBiW3lQI2ql7TKBiT3hqQ9gsC0cnwMZBC Eqi1wu2nNU//AP10859SVWtbaZHUo1KvSduay5qEYIZhH+bbBnhViqp1BpikBpHw/6009S5f U3mrQ1bn3oWY/VfJrL28UysMIXUBGs7OaiwkClqxK2zv1uwSqYyd4LwxUQfbGQ3qaRAxh1/a PioilKlkW2FsWcoKqKLnBnMiC3tNH0SCyrZy9poqUfyd9/9W7P/aQ3mfjMf7zn+0m2r/13Wb rtvaofPfnfbOl/Mfn+U6ONgbBoFz9v1fL/fUxr8AvZkrHOfolyOszc+GOYf7V/u097YX9tXb Eul96lweHR2qsuI5Uuf04ujg7OLwcs/VFnacg7Pzq97Z6eXe9lnT+ab4BNvjTqtYRK8ZBjXR dTaelA8OKoC/pq4C2wE82aoFjuNHURdrSO6KBdwWWEJCL+IYqA1no1af11vB6d6KW4E/4ZNt ZsVxAuwWYhTkL/o1yJxf/qyoFZ7nuP9pw37gtSL+i5b8JDwej/9my2s2zPkvr9V0O3T+s7Xj fYn/z3H9FGN3HMEVDiBoGxy7LFrKg8OjkzO4YOj+9MK0SJmsOYfsjkWCFjr7s40uHE14AC8R R8IhH/IUqRzdp7RTJp1DTsvz/Sy1wPuoZO6cz+npTYIwY3YHbZKw7cQy9FPaRzAL85NEhFmg F76dkI0Fvfa1mdL8z7wHrLhoHpCyYBTzAKUxlTWnAZcTFnCf3gnEaclYv7ru90Wmif50pduM pJ1GzQWg08/6BdJ+NqT33+lFt+GQSfV2HMJ4AD8wJaS0UmrBxHjiJ3opWL+2HEVwxxKJeA4S vpxJnKsg/V8zrvcUpeMSxxMe83E2XgPgkUi0F8/iEImvAkKQc5ao1sUBczzdDImDJJIZnxHg 4PwnkBOMbnxsIk+UOJnhfQursF08ZGqbL5sME3X44M9Y1wZ4yVKtC/yLM/BJgQ1SuUrUIoYc CUFQTpM4HyS+HKEfNIntpchQgVzKTJUgzilLcW52C5MIh+B5hRYjTUQ0L0P2+l3l5iE+dZDd 6ZVXh7/NEgG9Xh1eCREKAc05xg7A2YTFL4/x/rnF/gVeVBt9LHkBymXpqwbSviScTWCKswwx RcWg9OgtYpBO0Yw4Zou1ThyU7ZCFXL9AbMW/ZAnZ1mlRk4/usTWxci0swGYf+1lMXzNAZKXt LEo5vZ8oFRap0X7uwMFWnsXqhMBl6qcSdeoHtyTatvFNScVOm/icIfIdZ1N8QiZXGLYEiP68 Re9jby1RcFBlP0mGfgg/sz4c8/hWOjuk6Mks4cNRCqdoNrTkPEZ6740R+10IVb46f6hoMI5f 2zdBgigmwqvOlG0m9J2GAZvClLFbjLIpOoM6dSJHfDKx72sOeIz0h0i7dkVbiigcokuBRhzY EERbUUJQoPw3FlZ1yBIMISo3S6c47SAB1ChHxfl7wvxnhpJFAVHBesQkaA2WUy3A59uK1PYa 2GQn0Ydo3xKlZWEfrYK5oetkKamJ2P6FYR4lerKGrFC3mBGZPzY5DplGM8omWE+HmOicScjv eJj56sMXfZIRQsw2OGgn7k7ORm3UByaulTlRKEpdH5+5rmwByQkhHwwKJ14GArvGKWlxSpto i4TQ5ES+61yxe83UYuXkELhK/usjN2xhFbxfYISuyegQVJ7AsLuILa5WMGhVLZJVbZkyFbvY aVDHwAI/k8zBxmKIx/pNYdrli/gYOy39qYilFuaHdwRFKg5NaIdPVh8oqb0g0aNEnjcKh4Wq SrOZhNZDAqoFIo5mtv+SUNKfMEjBV1+2KMFYhKinHp0WiVmqPPJ4/1Sl04IUxnr6IyToCKnP 4/w8EtAL0RHbJhwUL/LDUL2DH/t9dcSIPoTi0KcoVOyMMMkPRwpXtU0fFsA+wB8SsK9dNhH0 yYxCM0LBJHkxL3zJwyQMTKapOggQKtEwcWyrTwmQwKkQkXSQRzBSuQDTczpaSgZLotJycMaj UDUtooGKpGVxdNWZw+5pO1glSVqJVyz1VwtUWxBKPiKzH89gnKU0MkbkvkmHCwrOLe0oU3uN HIGObulOleS6w/GJdjjyd5tDajgwWDcueGxYgL1cor/roGu7P6tOTMKL5/UTVsf80qj/cu6c 07dEpBRJ9xzROBLr9XqAqt8/OcQh0wiHteDWGvDyx9/0d0cI1tFjg67rPYeT72eoqov9E+dH PwmxQ5XY89E5im4TA0PXDhLGnFeYaES36WEZWFlUxKScdjnuqBrTWhJuOWpI0F0BJNVYgYBq cPqqd9jbh/gHkQSMJHb0oZGx+tMXCCTrCoHgpXVx23EciggHhY4dm6pooQC6Y2qYQV3VURxg JOmPmHCKa0rhjka0eH5Gx6K3HDOQ6NoBhTX+i1qjT8LplGXBqHukk34LlOcfIZBmhOHoEQaK AM4JijLuNnGU8zfoI1coEw6F+riehzo5TMWxj1cH5/XeeSVPAGVbYPomMziuqSHk+0aQH+ZP 8IEO5a12KK/decSh2gsOBU7BwZC+DnsvL5NTf1IxXmd85SXTzuIpea56cOFjbQzESI3orr49 Vt6lUorTaZGvvkJR0FC5dox36mHr95FPSQ320Q2GMyT898sPdQVt0O8T7HH6ZKHciOZzGNSt fbQlt5CMHodSz3qHQzf74as+KwaSSplGwOqyhI6WsIazhuVJQ2HOcKU+D2UfqafSKZnybZLF hd6AvvKlvrulilRupZOCKAxO2xDVKcwZaLWXKJA1KELs6a8suJ2BeumUmNE2n/mCl6rVFLFD mthRixou40RlgjJJZVNZgwuDg1PLTqtx38JeF+m7HejzVIUSmQd7BxSp7Xr3zeet5XqjTHRc yleo4oASRVF+hU4iRpQ6nJFxrZqaY9EU63/bu/63to2k/zP7V2wpbTC1BcZ8aciROwo05QqB F0i/PHf35GRbBgXbciU7hPeu//s7n5ldaWXJVtKS9On7WNcjIO3O7s7Mzrfdnb1iF+vaYgcJ 1KCBiQYPsAYxwrfs9qb+GKkqnh4afHIfdrGwymqK/IPARZ5HRs8EC8wmC1rPj61bRj3qy57x RtC9CbKplxCq7gIyoKKEXWpMU8uOfY+cGNiozI08bXl7nr6Z0OgZxfTamRtaUYc7iccepHEg y12AJ0kOa4QFMB2ZNd1gBFBD5gxQ0h/QTBvDQMNEBG8wLxFhOMOUcZIEI1l5DMVgjUMJYaL8 t37Yx5q9p8+kUiLzgNzTxBZmZvU5aw2ghTRdIQ+6kELYWKh6nNJqOBa7m+1FBMjqhvvZmiTv kSYkYbujR/J36Ch0IZIyjG4S0qWqlPCL/tSygTnZ6wYw49NNlApfuhHMEYYi+m8C6QF3gEYz kvx7CfdTclvZcd4Sm6I78ALLSeSx7z/D9SdBSgSMOBdaWQmDPDMI4kk4Zzyr2e8iXzfRSPLT VZgrRMfgHfFDGIAZ7NSe4TzCZmVjKRg/EJmVbWOPLCVMq3bITleOv25I82P0M0DCi5MzDTRh 34rNCqc6DTL58diRJmR0Qjhe253dMmq7s5XYlzcgkIAjMTHTB1ZT/J+hjvmaZr+nWvpbmLRw 5frRmF0yTILboD8icfYjmNfQcxSTrBcjJxyy90ezvw8DmDht/CThPEiSzY+mwIP2e9R2l7dF IomhHUCkSfAZBYr9jqzqlbThcbxnTrjnG8zenj/pj2eNmXxZcrKtTIR064Q9miWmmuJNiYmx 5WOiIRKElWgUY/vznCAdhBaRto2lcQ6vxiCPjZxlFzSOogHNsZ4knILf76lMr4kcEb+RIJLJ 2Dcew3vrsLrOz+0e2QhMT30V8fsBh1bu4bPZLkj4MhHkMvuAqwPT4NswIZrqX+hHSEzvjnAM wuEICLtk5GaEUDYnxQ7klKG4Kr4mk46PdZCLzvom3zvFEu4+kMDFj1FM6DYoMT3NHFT3LaTp cOKpU7RrT7Ug3jHOo6ZNEsvBD6N+qreKzbFMBKbCjrrGqQ8RRKhhppRFGm2gkdAxgyORFdGo Af/Shyhi1fE2jMfAuHD+sjLxFRHGw0B8fLZCJyPWC46akF3kkdFPNDEHEMXdmNS6Wj3cqwFn NPfkdIhohJk6su/HN4HlP6CHpvmtSsMaHNahHvgQD8V+ez8GmXbWRrJlVjR0Ko9huu98wsyx ryT8gGhtLlhLWBWfxdetI+ObQXX2ZaUI2d8IqzKdS83lJODcmipz6+ppitUb3i0dnY/YI15H jWiZE5gOWWodH/yEqH5CrXWY3tRKasdTd6TBr1BuWUcjE8j52SGfI/QcI9LawoSb0j4zHclq Ui4VbxBmoZGPZKN5ig3XNlOvkjlKLeSksm9MdlJhyMCP+yE1YLpkI3LKduHs4GexNwgLbyHj XLlAuoe6MkgG/ptIJjjiWzYSpqTrBLC10dje+MKEbIxBIL2vZTY9bw5x7JDOLbZzCCBGrTah S8MuLM3TODmhQ8hn5VLMrsgEWIXhF0V3RBFs1SeppxHGJsVAnB0b/ZMzgQxfqGXDGcsisLF5 Ihq6DJBxDM8Sv/uWJBrNYR70zkajG9ywF2koJcccrRlH1vGdhH1gtcA+h/7Qid8LxBFIjUmm nEUPtBQHVhEW4tCAJh++z8sCliBsb+ubPrYMBsk62b3mFxgcpAuJWm2/m7eOe6aX1Alh/67H SyVlKyWwSgYmMMTug+NQivmeMWoucDBlkCua1dTG9xJjcIGMIyFiWhsrHXA1L8QtIs/xezI3 hjTTCt3bUz/C8vQ54mWii8GQj0j6aXlx5Tt9HAsiXUKVHevQV831rYbZLMXmtWFesB7JaXKO krtwKA6BmMSS8zG1gmmQik01ad9LM1m3YcAk4c3Qm+21zEOZ8WFUwYdJHTcwn5kAFp31VPoM Ow96dcSyjMyeWubSkFC+oxb7UZKYwLZ1HaW7Is/aMn+ZsJMRdDcCJAPClJblEiRbrevtL1x4 0ogQ2iExMw5NAMf/15gwPuKYtwFpErIhJwhO4bBRkva/XoCtsrFn8BOPF/Sm1vOu4BlcHOq7 4EHCd+kqRye6GSIo3wlixPAgPtrh0AZLaT5rjoWRe5YQ3/fHPics6cMsJsVN6CGQ3OZ2tl5o FmAC51CaDTyUxBaszMUEYgLmDRTScjMNDO6CaMHU8gXSLUiICyUmJdygsRwlJRU4iKjhTNDw KhDXvIdap9GxJ5QeWlJYhuqF7wLHwn5LfBjxYkwniuGTSoZjEdRhPDU6mgmRitqM4pKxjwRj ZIyG5pxZ0KZfg5Sp8zghrpn0yDwm9MbA/Q4WZ3W6OquLy7NHE0cbkz7B6SRDcSJgmuVbjGAW o9ZQhz4Bno0BrzqTdjDwR2kmcDRdty3X04br+sXWxkbd8yRCRLxsdg4A4ZhBObfDd4P7OYvc 0wfkwA1tL4JutvKEMO4AhhCbcrFQ1mSatYUjDkUEqWHbJbN1jE0KWOahfwJem1JsLvPilKyL 9IkXSdRkZxTd5a7UNukJsQZCdg6zW4dOBN8gQMpw8D4HP4a84ssii227eByCiqR5ElkCpk4Z /lAw84lrWOekFJtCvYDn+RUHvb4RAAbVnISeWBNzUiMIhXWk7oRjCekqKnXtCXWMpwOoTh1o G6M1RXeEkIvdBMIrkAEbXVjxNwv+6KT8qtO4uonDR3wIVfCd7hkhVrIxbvgwHRtGJcV4Q/xD RlAUiqFvmTSddjJj8mKKRASiiOxcqXGMIEWM9W36yevLcGDtNHw15JiUlUep9mPbjH5FmnjD garX9++TzMkdRjZpP9Z3jZ0mc2p68IJubg1zqajz6rOE2r3vqB4/ZtvkqCWBQxXJ+i/rvek4 DApMObPp6h0A2L556gBGl5E4d7htALErSbFg5uH0cJB/2qwMi90I2ytbwchmXq4hPYMt7tiC IcwpMAibJYZP+6bP2JK06+3u8EqCJR1vKzlv25WmdH+Jt9FW7h8y19rpOnDXmgqxeNKHgsx2 kC5Lw3BBgErwCof91eUp3OlwbNnY78MlfnAScFsvVBDiqdvxeLS3vn5/f+8Nwk4cYaUJGwvW ZYtLsi4Z9d+tGxve85PRO94YM3dfzEkvk5LpTRNGkVhtTqqzQ2zLgce+3fyCzjMmGKxyYHpk ackOgdS65ps3+iHEjOBKyjKiWBDog9PrxvHL6+NLyG5zKgcikCR20uFs5TBmvZn7eGzS8CzS gxEhaiWheahhG+JKYx0wCeyKnVYpMFiTU+MURk/LuulTkONnAj4VoawkDs1zAK27q4IZmxrf Wv626z3iwBDe1ER29piIS+qbGOEkOheGkxNCZk8+9eER0M+FjTyzM0U30iXKrYqdT+7Gp285 mw3ZoTT9c75BLnhmSS3Ioy510wY6yNhBls6crUVm/xSbRtbLtBa+BGRzKw/K3bRhoxnDaNjI GjUQSQfU5bz0hAzeOnZIwOhHjDJA9FZzry4OSXQNieNpGg54x1IAfQXkT0NMLBtbl+KZzgL6 CFIoE1kY4X4NsW0IgT1wBsJU8DFHPn0EHj3eVZZtKvt5NhLteB5IjYmVm23fiIObCRkqBrlq GrnQ93UTc8G9D3bbGrtxLCH4x5f6zB/V7QxhvPgjwv0IGxwCG40x/gN9FX55ghOV2LPEWo6N cmiwoHOnnmQ8ZspmTtu9sdxhItiQajfiWUnCn+YOFkzZdlb3rCSzwOgQ5nbs21UKIqkENY1G Y03D9/SQcUJzzVfTSISaAu18rNnZm4bmo523sI1vp2Z/+0FZp0NmmBe8C8zFJ/0HmdKTROL6 JOnCbrqnSXX2/im77xrYffdPs3af3oxS12bfmfmQTNrZrSk0lYC+gGeMuMrpNwwlUySpMAyH NQ/zOHjH9ymQ9xopM1pJmkGIODprXPpDcsKX6+D8PTXpdOz4028eDZT3PqZbH43YwKKzXaFT xz7CTzLtTXBpCqFAN//KV0Sx0z9k38h0Bma7zx61p74jzfIWG5ksnTAvB7Z9Ke5MTvfOKRI0 vJGOuK7OC2+DCcLUWKsN0EfTG5+4OvyF/BhZoXevPhJTw6oPfFDGdyMu3qWH25A1kQcChBJ1 PTFJdrJ9dFl6CWYgxdm3ZqKYwewT9K/V9oxtpNObSNM9pDNELF/0MQll0cWa+XZ7oxF4mduq EhMdcprx3D8I72OxdcfokfjGYWJNf+Qgi1g7c/ybB3ZjdD6HmxJnvyVCaOI6UQm//5CEjsue hm6UbKjUrJB8s3iBzYA4ewZ7K/VopYPi8bpd9sd71paS+SrDym/WxBbcbAdu6f7bF5E1zsNh g+fXlTUrrDJ9bjfXvuhHbRqXQOAFFoausYLq9m1s6UoSns0lE8rgRqX6MrTZODJBMkWeDFb/ U96VJl4lAZOdy8irCz9JOCmP49kYSWwslgROQQbpglXDyRHxzLAra42EB0+9jMbBHu5ogaZm nsxhNw4wh9cgdDnk6me+EFtRTBs1Mr3x1E5xF/MPTLM0l49EMiYjnJxLEKXioKvZR0ETifpL vgg2b4yTPddMnvAcGKdTgClb3AQt0pMPPTAD5GDk2WJW5eljEiWAulIkMCUY3KkMbRjYrHwY L3ekKycyJDdZ4A88/QJRJt61M4k75lYzZTYJDHjnKdnH2p+QZxgb5FDlW7KL/MTBi+Bkalic /w6j4eUwGZicubiQExr5ofj4xBV3i9vNv/cJQfriluYvWZ6nYTvGElpWjFilpQ/jELl8CIup JX+KTbvwHQQAYvaKF0gHfszBidl1WNJ2jU1Cs4bMfVIYZ9ZJ4m07JObTakcOcr8PSVD/YBaO IEHzPc2gkMg5v7nRP0Rxm/qWK9VsoGhd/xSObr3z+EZ/CxtE7Ga0/S5ffNMU/3vgDxtnftzh axeH66X1XyH+eEYiLjFrWmILeA7IQ918+lT6ALlxeXAkptx1xCkrToYdQsbF8QsyMWlCN1ps t/OiDxRAPwy6mSNfbEkxY5YBzcGUFRM3lsfHdaJ+dPOgDE34LM+3MVkct1GPSHVycsV8ev3d +dnV+UvR5gMyEnyPHO63P13WNO8BtNqC6+ODka+mH/rw/OLny5MX313r1cMaMLzByPDUMgrL 8g9ZxyHWAmEPuVxVAs5EyF5dHWTrEORrTOTiFE9dhXLXJZtbZ+iuYIFYT5gdO0IyhpToXjxI d+Wfvrg4VaeGTdmwwBtt3rDr0YbRMBmmkYXDI49qDCfvuEdkYLlrnYiAHJyyhP0IzX5zdRQm txm7dRjDm/rvEZa3DzpAqa//8saX3/5GXYgTT6RUD2U8UtvP9ZfqIkCQ/PzJC3bfdHTD/5aX LzdfWqxLs9Nphl6XJFWpZ3X90jv0QDYvjYfZVRjzJzFjJKsEJdzAiGLgyoV+cHqqmbmu9OXx 1fHlD8dHBP8AjhyLfqc+t8br0kOJ98eBPeZGrt7Y2sFhjFADjEXeLkg2CwcOZ5lsCNjJrnbD /1OqBq06OFHoNUE7y0L4XS6DPgYGCPQYFIcdJeEuuCedHt+5vykfLgd5LKHPDodTTR/DIEOQ SIKNV94BAclJYNFoappcf5bjs3/6p+T8r7nk5/HaqMj/vrGx3bTnf1s7fP/3ztb24vzvJ3ne M/97aa53xfk/tJv/Q94MsDpvfvXfmSRL8jdZsSi/r03yEK69anI0cQmbFQg5DBsAVdNrttq6 qXV58PLo9dnBT/orFODKJr8G0rnQh1VAQUaUGlLHAPyvymT1Uu2HcdCeYA3jdYwcvKvZC71G 3iZ1Sv1HLeG3xnNJwiu/o1zyzH5JsxNJ7rxfS8DDVZ4BnVOWO2CdfKpqib4vcd1cCSRP+bW0 HawemESkoiIlBZn8zn2s6/JurK+pJULvIYJwEg5bdWoRerN0qlhnlI10xvTJIJISABQa0zAi Me/bdKu8YL+ErJMy3jzoKTQi+YIZb5aMuGbRgfqfuflfs9a5Ut0dLdfiaks9kxyGWDmI4/ry txIIwJJ82kvZYAtXKQP6mZZsMgQizfa+ROin/9QS1UAm15RDbOP5TiCdz1pW6Cv3G/EM5+lG im9zubKbalkVmeyrfRf4M+Uy6PQ38DpypK7NQFiNJpZs3pczfJLd2q6+myMAJmM5KuwRKECz eXQTF2e5krxZivNLp9lUV9v0qfE8zTpd8ww0VKNP1JcT50iHgWPzayMEIPvayV7mbMTpilJK QB/Qlpaw4cj0iz+YnUwhOWUzW8jDxxIbtyEAYWghqW9N8yJVf5pxZI+TAcfOUWGsMgUEWhQN TCjFnMgchWbXiDkBmRrhUb9rsdNzc4zfs0kaDS0GR6aQcEjDnsnuTOJYDjDYrpojrwTe1Egh jDIqsAfk5DX3MzSao3XGB0wTNg/T+Y8wJY8SQLpvcHd6V3bwOgNIeK2tjw2oBgQS2SPenfUz A+oJwBO7/inEH2N1UV5w+vTILEPfhBJcd04HZWAAqUiaRooU0zyPFN/RPcSLUia2KZ0NLJqv FuVjdzOeAcc9EtJhlUfYi2W4IlHYJt9Yq/KJ6Qpp9Ldu0zYaqcGaA8kgn8m/Ua9HWEDG9iUj YmWUjgDVz/ddAOBqVpTjeBJA/Bz7yQNHdRrYoMJHpw2rg9T4wCtVS5mu8PsdXBQt8sq9C4Gn vYyYptxBt5suwHBFg1qT3h+4smdt5Qtic/0HzyoMnfV7vUA8kwhuNaXEF0UCs0rd4DGHyGHN QxBNZVeeE3eHXgn5MsbAUnsslVN2ldfYsok9XSbyEA1FM/Leb3OKDGKS7//g+mMj8HkiO9pR +i+6fn/ftQWcL7CM8inHw1LmXsvxLWlbUWBSjnTG0r52oMKyoLY4N16mb1fT0rXPpD9Gq2pt OA+9yQoRC+UGYTTmzAFY+ep8xc3u7zsczaDLO6qznqbfss5K976SMTh6fXryINNvaX+gftfX 08RzpyJt+PaELzgNaXEeUn97vf4kuYUpQtOKYdikcPmOWVNQ12aocif3/Xup8qx8hSbPCpYp cs4IX6bJH1ceOqNz0uyuyaUiHywQTaf/30nEAjHKRGKhUIlMdLGEn1Oyx/mSTd288Cm0kp+u NNNZFJhiInwcsDOET1Z8nvRxSkH8uOPIxM+MMVj543w28ud9RuSIn2JXHfmTfXT6yx0sCqAC s6YSqITcEEH5+xN4gFeOFCoyf4UYcruXF0N/dKBi8XyUZ2b87/bx2qjI/7nZ3LT3v25ub7d2 OP63s7j/8ZM8n4c9slx7+vXrzojEAv6PYJ+8hGpW2I2CP1jw4c0z9TlZu2EvKwfNmV7XiD+Q 1ZesMipcI0utUL5HDnBWgf9yKmykFeQfZCleY9/88Kuv+AqT9GbIk5erfr1d06u+/otu67+S CbOn2zX36si0wHO3QHZPpaSS9muQfqv+F1v7+6Qev/xSf0Z1viCuxN81/d//4s8t+VNnDXCm ev2fVrO++bROP1sbzs/cn79mV1yaBPdS7euqaikByC6adMYKyYSXmBhLa2v2fkc2fpaWUvXg vMuEv/MyuyPIhUY6CpFS53YD3K5Q0r7JKI1qmjxVcSHsa7TgPrlOlRfJ97G8TK7L+Q5wv391 onfU7fX1OWbxKnlm4wgDqxVjWBJfrY4SPysULMR7TRiipOhvCtk+E2P5N8UOCCXTlT/A0H72 3sH+ZwtL4c/0lOj/9LfHaoP1/86c+983Nuz9z1stvv99d2N7Y6H/P8UD+mP3rdcZ9z9WG1X0 b7Yc+vP6b2u3tVj//SQPdkVpYgKvHd7wkfGmhjH07eX5mX6y3m0jV8d6iZDIUsA/0T+eXH+n j45PT85Oro8v9ZP/PlkogT/LA9KGb0evrWmUeG0/eUTfD0+F/9faIJ/P3P/Q2t3YpPLbuzs7 i/n/KZ7PP1tvh8N1EF0hU+oYC+SrNbbyg85tpJf5vB+2z+IrGYnL5qo03URwuhudcHLaTrAq 95wgrtzXF+eX169fvjrbX2mm73A3wuujk8v9lc303en5C749YqWVvjo8+ubs/IhebaWvzl9d X7y63l/ZTt9IiR25to2PJ1y8QJv7K7Zl+jK6eU1aTTfudeNIr9jmdaOvV0y7uDJiuRHqxkiv CAAa7wqA683nXzZ18/lzvSKtK76n5h965a+6Efyim/pfz2Qx0iLtme6FwAgSDpXiJBt/szCy TeX01+0soBX7ws348fiT4p5eZdTWKaKdzqZoXTGAbIfN6U/qNvlSf6qOm/6i67K98ojUXnqG Kk/fOaQNh+G426YWqbHfzFkxObRvg6MPa9xMY9tI4510wrTBrjQXaAz18iUawCo6itR1dPdX vcqHkZ48BMkTs6LZCYJuTS9zTU7Q5cstTGYQyAWwr5epwnKuESo80I24NxMHGqM0P94TH5Ph 1S/9b0k6FdjIxYXwSsY+ROGrlwdnOf6x+NrieZj80heRAIGwYsvzDS/MJb+dgtzjw0F3bocP z44ep78dvbxCwJZ/c4f/aO20eD72A/sPsagGm/OPbfnJA/tvd3t7pv+3u7MJ+29zd2d7e2cL /v/W1vYi/v9Jnpz95+HY3K2HX6EaoVP+Pbrv/nvK+VOKhcdnEIg2duDIejEbT88PjgqOZb1d 79S79aDeq9/Ub+u40/au3q8P6sN6ZJ3OFdN0lYu5rJ+nrSuSVk6nSkA4/etIQt9sPM847e4z fPA8BvW5voBNybeX6aWl1caoRu9OI05HCaWHqxvodQevcU6ej1GyaL7FqXn6dItPMBaQBJc/ d29wjo+sgNVGFx/T45crr65oTFRngtc2Zxb0Wobe1UavpjhLSEz0gbTf3mptQsRv6cat03Sj mzXUmBjYDqGmZDrm/6Nf+Df1VN7/t9ti/6+1vU1iYovjP5vbi/n/KR6+R08s2wbxgk51gVr6 Gyay+UcvX0s2eJxsN4dQsWEWyZF/mXBm8zQLvbnzbzmFwDl+rZHCbJtxKJkokm1C89m51bWa gM9JjmVrfKOLe7jHT/583W2zsEo7zd9cdVa4wg9w+9FN7m/bm9yHdL4s/a0oMDhNN0Omz43R HbYjr2MZanyzZiCQpWvhZjC7bTOF0dIfTXp+MP+N0P9obVTO/+amif/ubO9u8Pzf3lrc//lJ ns9t+OTg+rv99UkSr7MqWe8GwQjpU4frkqVzXRgXxsJeabHRDU1w/rwCWGoW4Hb4v1QhaWBq zIedK/lbwDfgLnffvxFTfrqpGS1NF/ujKfnbHsz/y+ODo7Pjj9dG1f6f7dYu5n+rtbW7s8vr f63m1sL+/yRPY86j/ofTK51IjiD+FztA+HK8ufW+4Vv6OKMHn8yoO/daOPd72NRDf/ff+rrp Ib9VdnuOb/Mz8Vfk9g35UIlcG2AvB8HM83QuXZHfldPcKkuow5s7bFJjXuB4PYojmLx1+TPu cBal9WDcWTdf9hSS9Tuz/g31Y/3NZtK9466+3viaPaeikFCq6enjbji23hQHzvi0UZoKxg79 iOTnC85b17VXpsp1nYwxxWlUBA3YOBzHyIIrmerl3AGOh+Pkdk9SywadyTjN6Y6cyJnPZpKA IK3ZIEQGMFxDaDfSeNmiv3av/7W2HLlazp3Ptha6+I/mv/b02tqa/gcN6F/6mLu4s6U4Gae5 gyZNKi53VzApkglfwYKULtxRSGeb9kcMSjMaHJIw6Uc5O+hQLi405MY+7dFkbFvgbKOSN5bM 0725XPqhj7o+vz44vSK+aOhrWLTps6c3N7zWDlLTdWhm6EtkFL1EaFS+WiSuDpFHve8P72q2 1KnkfXRK0Zdv+JyE/URftjZ3NpvkH9E39kGRuc/A3vF2dp82t57qM96FtU5doFLmlOM3kgRn T2+o/cZ7/0+B7STJS8f4oiYTVWJI8UF+gFKPYP9r+awRx84oQhiwjWD/YFPH0X2Nuq8kNN0H ijYGLW9zY5dD1MgiIe82POQiwrvkIdG5d4/KM3+0eK98xP6Px2wAeUSjj9DG/P0fze0tsvlh /+O/Fu//3drdXOj/T/IcXZ5f6OuDb06P9biJTKwdkjrPlDoko/D6OPuyirxa+uTldV3k9fXx T9e1Yrk3pKZWD787Pvze1Hi+j/wk28QCTX3w8kjzy7+Yd61mrUYwvzu+PLm+Wh3jiNoUvF7Q Loe3WQJvsxregGhdCq9VAq9VDc8fzYC3VQJv633691AOb7sE3nY1vDeTGfTYKYG38z7w+uXw dkvg7b4H/iY35fC+LoH3dTW8JBiVw3taAu9pNbyIFFMZPJJoBXjNjWp4w+htObyS+dF8j/nR DTrl8ErmR7NsfqiTl1fHl9eY2edm/v5wcPrq+EqvyhxtbtSfcP455J5/QjUqK2x+aIXtD6yw ufWhLex+7DFsfv2hLTQ/tIWnH9rCdJemakC05mpsVpC6pMJ8NJVUmE/qYoUKUpe0MJ/UjzCG ClKXtDCf1CUtzCd1SQsVpIbWy9VoVZC6pMJ8NJVUmE/qYoUKUpe0MJ/UjzCGClKXtDCf1CUt zCd1SQsVpIZBkquxVUHqkgrz0VRSYT6pixUqSF3SwnxSP8IYKkhd0sJ8Upe0MJ/UJS1UzuqH fI3tylldqFA1IwoVqmb1dIXKWV1ooWpW/+4xVM7qQgtVs7rQQtWsLrRQQWqY8bkaO1VmWbFC hUlTrFBhlhUqVJllxRYqzLLfP4Yqs6zYQoVZVmyhwiwrtlBJ6n6+xm4lqQsVqtBUqFBF6ukK laQutFBF6t89hkpSF1qoInWhhSpSF1qo0tXk/OZqfF2lq4sVKvRcsUKFri5UqNLVxRYqdPXv H0OVri62UKGriy1U6OpiCxWkRlwiV+NpBalLKsxHU0mF+aQuVqggdUkL80n9CGOoIHVJC/NJ XdLCfFKXtFBBaoSM3BrNjQpSl1SYj6aSCvNJXaxQQeqSFuaT+hHGUEHqkhbmk7qkhfmkLmmh gtSI5uVqVEXLSirMR1NJhfmkLlaoIHVJC/NJ/QhjqCB1SQvzSV3SwnxSl7RQQWoEWnM1qqJl JRXmo6mkwnxSFytUkLqkhfmkfoQxVJC6pIX5pC5pYT6pS1qY7tIfvQa3eBbP4lk8i2fxLJ7F s3gWz+JZPItn8SyexbN4Fs/iWTyLZ/E89vN/0A4qNADwAAA= --B_3206204724_2604846-- From pgsql-patches-owner@postgresql.org Sun Aug 7 01:08:28 2005 X-Original-To: pgsql-patches-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6A34952A70 for ; Sun, 7 Aug 2005 01:08:27 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 86158-06 for ; Sun, 7 Aug 2005 04:08:19 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 572FA52883 for ; Sun, 7 Aug 2005 01:08:18 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7748G4M029008; Sun, 7 Aug 2005 00:08:16 -0400 (EDT) To: "Luke Lonergan" Cc: "Alon Goldshuv" , pgsql-patches@postgresql.org Subject: Re: COPY FROM performance improvements In-reply-to: References: Comments: In-reply-to "Luke Lonergan" message dated "Sat, 06 Aug 2005 19:33:07 -0700" Date: Sun, 07 Aug 2005 00:08:16 -0400 Message-ID: <29007.1123387696@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/50 X-Sequence-Number: 16997 "Luke Lonergan" writes: >> I had some difficulty in generating test cases that weren't largely >> I/O-bound, but AFAICT the patch as applied is about the same speed >> as what you submitted. > You achieve the important objective of knocking the parsing stage down a > lot, but your parsing code is actually about 20% slower than Alon's. I would like to see the exact test case you are using to make this claim; the tests I did suggested my code is the same speed or faster. The particular test case I was using was the "tenk1" data from the regression database, duplicated out to about 600K rows so as to run long enough to measure with some degree of repeatability. As best I can tell, my version of CopyReadAttributes is significantly quicker than Alon's, approximately balancing out the fact that my version of CopyReadLine is slower. I did the latter first, and would now be tempted to rewrite it in the same style as CopyReadAttributes, ie one pass of memory-to-memory copy using pointers rather than buffer indexes. BTW, late today I figured out a way to get fairly reproducible non-I/O-bound numbers about COPY FROM: use a trigger that suppresses the actual inserts, thus: create table foo ... create function noway() returns trigger as 'begin return null; end' language plpgsql; create trigger noway before insert on foo for each row execute procedure noway(); then repeat: copy foo from '/tmp/foo.data'; If the source file is not too large to fit in kernel disk cache, then after the first iteration there is no I/O at all. I got numbers that were reproducible within less than 1%, as opposed to 5% or more variation when the thing was partially I/O bound. Pretty useless in the real world, of course, but great for timing COPY's data-pushing. regards, tom lane From pgsql-patches-owner@postgresql.org Sun Aug 7 02:22:23 2005 X-Original-To: pgsql-patches-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9DEB652800 for ; Sun, 7 Aug 2005 02:22:22 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78644-03 for ; Sun, 7 Aug 2005 05:22:12 +0000 (GMT) Received: from mail.Mi8.com (d01gw01.mi8.com [63.240.6.47]) by svr1.postgresql.org (Postfix) with ESMTP id 47B0A52903 for ; Sun, 7 Aug 2005 02:22:11 -0300 (ADT) Received: from 172.16.1.112 by mail.Mi8.com with ESMTP (- Welcome to Mi8 Corporation www.Mi8.com (D1)); Sun, 07 Aug 2005 01:21:59 -0400 X-Server-Uuid: 241911D6-425B-44B9-A073-E3FE0F8FC774 Received: from MI8NYCMAIL06.Mi8.com ([172.16.1.175]) by D01SMTP02.Mi8.com with Microsoft SMTPSVC(6.0.3790.211); Sun, 7 Aug 2005 01:21:05 -0400 Received: from 24.5.173.15 ([24.5.173.15]) by MI8NYCMAIL06.Mi8.com ( [172.16.1.219]) via Exchange Front-End Server mi8owa.mi8.com ( [172.16.1.106]) with Microsoft Exchange Server HTTP-DAV ; Sun, 7 Aug 2005 01:21:04 -0500 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Sat, 06 Aug 2005 22:21:03 -0700 Subject: Re: COPY FROM performance improvements From: "Luke Lonergan" To: "Tom Lane" Cc: "Alon Goldshuv" , pgsql-patches@postgresql.org Message-ID: In-Reply-To: <29007.1123387696@sss.pgh.pa.us> MIME-Version: 1.0 X-OriginalArrivalTime: 07 Aug 2005 05:21:05.0321 (UTC) FILETIME=[CF36B190:01C59B0F] X-WSS-ID: 6EEB45CB2B414402914-01-01 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.592 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO, RCVD_NUMERIC_HELO X-Spam-Level: * X-Archive-Number: 200508/51 X-Sequence-Number: 16998 Tom, On 8/6/05 9:08 PM, "Tom Lane" wrote: > "Luke Lonergan" writes: >>> I had some difficulty in generating test cases that weren't largely >>> I/O-bound, but AFAICT the patch as applied is about the same speed >>> as what you submitted. > >> You achieve the important objective of knocking the parsing stage down a >> lot, but your parsing code is actually about 20% slower than Alon's. > > I would like to see the exact test case you are using to make this > claim; the tests I did suggested my code is the same speed or faster. I showed mine - you show yours :-) Apparently our e-mail crossed. > As best I can tell, my version of CopyReadAttributes is significantly > quicker than Alon's, approximately balancing out the fact that my > version of CopyReadLine is slower. I did the latter first, and would > now be tempted to rewrite it in the same style as CopyReadAttributes, > ie one pass of memory-to-memory copy using pointers rather than buffer > indexes. See previous timings - looks like Alon's parsing is substantially faster. However, I'd like him to confirm by running with the "shunt" placed at different stages, in this case between parse and attribute conversion (not attribute parse). > BTW, late today I figured out a way to get fairly reproducible > non-I/O-bound numbers about COPY FROM: use a trigger that suppresses > the actual inserts, thus: > > create table foo ... > create function noway() returns trigger as > 'begin return null; end' language plpgsql; > create trigger noway before insert on foo > for each row execute procedure noway(); > then repeat: > copy foo from '/tmp/foo.data'; Cool! That's a better way than hacking code and inserting shunts. Alon will likely hit this tomorrow. - Luke From pgsql-performance-owner@postgresql.org Sun Aug 7 23:01:12 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7330A52C0F for ; Sun, 7 Aug 2005 23:01:09 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 33156-07 for ; Mon, 8 Aug 2005 02:01:05 +0000 (GMT) Received: from sccrmhc13.comcast.net (sccrmhc13.comcast.net [63.240.76.28]) by svr1.postgresql.org (Postfix) with ESMTP id D7C1352BBA for ; Sun, 7 Aug 2005 23:01:03 -0300 (ADT) Received: from [127.0.0.1] (c-24-4-66-2.hsd1.ca.comcast.net[24.4.66.2]) by comcast.net (sccrmhc13) with ESMTP id <2005080802005501300dpesle>; Mon, 8 Aug 2005 02:00:55 +0000 Message-ID: <42F6BCD3.4070600@comcast.net> Date: Sun, 07 Aug 2005 19:00:51 -0700 From: Patrick Hatcher User-Agent: Mozilla Thunderbird 1.0.2 (Windows/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: John A Meinel Cc: "pgsql-performance@postgresql.org" Subject: Re: Slow update statement References: <42F4B812.7070501@comcast.net> <42F4BC61.20102@arbash-meinel.com> In-Reply-To: <42F4BC61.20102@arbash-meinel.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.793 tagged_above=0 required=5 tests=AWL, DNS_FROM_RFC_ABUSE, DNS_FROM_RFC_POST, DNS_FROM_RFC_WHOIS X-Spam-Level: * X-Archive-Number: 200508/57 X-Sequence-Number: 13804 Sorry went out of town for the weekend. The update did occur, but I have no idea when it finished. Here's the actual query and the explain Update: cdm.bcp_ddw_ck_cus = 12.7 M cdm.cdm_ddw_customer = 12.8M explain update cdm.cdm_ddw_customer set indiv_fkey = b.indiv_fkey from cdm.bcp_ddw_ck_cus b where cdm.cdm_ddw_customer.cus_nbr = b.cus_num; Hash Join (cost=1246688.42..4127248.31 rows=12702676 width=200) Hash Cond: ("outer".cus_num = "inner".cus_nbr) -> Seq Scan on bcp_ddw_ck_cus b (cost=0.00..195690.76 rows=12702676 width=16) -> Hash (cost=874854.34..874854.34 rows=12880834 width=192) -> Seq Scan on cdm_ddw_customer (cost=0.00..874854.34 rows=12880834 width=192) John A Meinel wrote: >Patrick Hatcher wrote: > > >>[Reposted from General section with updated information] >>Pg 7.4.5 >> >>I'm running an update statement on about 12 million records using the >>following query: >> >>Update table_A >>set F1 = b.new_data >>from table_B b >>where b.keyfield = table_A.keyfield >> >>both keyfields are indexed, all other keys in table_A were dropped, yet >>this job has been running over 15 hours. Is >>this normal? >> >> > >Can you do an EXPLAIN UPDATE so that we can have an idea what the >planner is trying to do? > >My personal concern is if it doing something like pulling in all rows >from b, and then one by one updating table_A, but as it is going, it >can't retire any dead rows, because you are still in a transaction. So >you are getting a lot of old rows, which it has to pull in to realize it >was old. > >How many rows are in table_B? > >I can see that possibly doing it in smaller chunks might be faster, as >would inserting into another table. But I would do more of a test and >see what happens. > >John >=:-> > > > >>I stopped the process the first time after 3 hours of running due to >>excessive log rotation and reset the conf file to these settings: >> >> >>wal_buffers = 64 # min 4, 8KB each >> >># - Checkpoints - >> >>checkpoint_segments = 128 # in logfile segments, min 1, 16MB each >>checkpoint_timeout = 1800 # range 30-3600, in seconds >>#checkpoint_warning = 30 # 0 is off, in seconds >>#commit_delay = 0 # range 0-100000, in microseconds >>#commit_siblings = 5 # range 1-1000 >> >> >>Would it just be quicker to run a JOIN statement to a temp file and then >>reinsert? >>TIA Patrick >> >> >>---------------------------(end of broadcast)--------------------------- >>TIP 9: In versions below 8.0, the planner will ignore your desire to >> choose an index scan if your joining column's datatypes do not >> match >> >> >> > > > From pgsql-performance-owner@postgresql.org Sun Aug 7 23:09:18 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BD76652C55 for ; Sun, 7 Aug 2005 23:09:12 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 87512-05 for ; Mon, 8 Aug 2005 02:09:07 +0000 (GMT) Received: from sccrmhc11.comcast.net (sccrmhc11.comcast.net [204.127.202.55]) by svr1.postgresql.org (Postfix) with ESMTP id 6052152C47 for ; Sun, 7 Aug 2005 23:09:06 -0300 (ADT) Received: from [127.0.0.1] (c-24-4-66-2.hsd1.ca.comcast.net[24.4.66.2]) by comcast.net (sccrmhc11) with ESMTP id <200508080209060110017uqse>; Mon, 8 Aug 2005 02:09:10 +0000 Message-ID: <42F6BEC0.9090504@comcast.net> Date: Sun, 07 Aug 2005 19:09:04 -0700 From: Patrick Hatcher User-Agent: Mozilla Thunderbird 1.0.2 (Windows/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: "pgsql-performance@postgresql.org" Subject: Re: Slow update statement References: <42F4B812.7070501@comcast.net> <17916.1123337563@sss.pgh.pa.us> In-Reply-To: <17916.1123337563@sss.pgh.pa.us> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.943 tagged_above=0 required=5 tests=AWL, DNS_FROM_RFC_ABUSE, DNS_FROM_RFC_POST, DNS_FROM_RFC_WHOIS X-Spam-Level: * X-Archive-Number: 200508/58 X-Sequence-Number: 13805 Sorry went out of town for the weekend. The update did occur, but I have no idea when it finished. Here's the actual query and the explain Update: cdm.bcp_ddw_ck_cus = 12.7 M cdm.cdm_ddw_customer = 12.8M explain update cdm.cdm_ddw_customer set indiv_fkey = b.indiv_fkey from cdm.bcp_ddw_ck_cus b where cdm.cdm_ddw_customer.cus_nbr = b.cus_num; Here's the table layout. It's the first time I noticed this, but there is a PK on the cus_nbr and an index. Does really need to be both and could this be causing the issue? I thought that if a primary key was designated, it was automatically indexed.: CREATE TABLE cdm.cdm_ddw_customer ( cus_nbr int8 NOT NULL, ph_home int8, ph_day int8, email_adr varchar(255), name_prefix varchar(5), name_first varchar(20), name_middle varchar(20), name_last varchar(30), name_suffix varchar(5), addr1 varchar(40), addr2 varchar(40), addr3 varchar(40), city varchar(25), state varchar(7), zip varchar(10), country varchar(16), gender varchar(1), lst_dte date, add_dte date, reg_id int4, indiv_fkey int8, CONSTRAINT ddwcus_pk PRIMARY KEY (cus_nbr) ) WITH OIDS; CREATE INDEX cdm_ddwcust_id_idx ON cdm.cdm_ddw_customer USING btree (cus_nbr); CREATE TABLE cdm.bcp_ddw_ck_cus ( cus_num int8, indiv_fkey int8 NOT NULL ) WITHOUT OIDS; Tom Lane wrote: >Patrick Hatcher writes: > > >>I'm running an update statement on about 12 million records using the >>following query: >> >> > > > >>Update table_A >>set F1 = b.new_data >>from table_B b >>where b.keyfield = table_A.keyfield >> >> > >What does EXPLAIN show for this? > >Do you have any foreign key references to table_A from elsewhere? > > regards, tom lane > > > From pgsql-performance-owner@postgresql.org Mon Aug 8 00:48:43 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E40725288B for ; Mon, 8 Aug 2005 00:48:41 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 59195-02 for ; Mon, 8 Aug 2005 03:48:31 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id EB77052967 for ; Mon, 8 Aug 2005 00:48:30 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j783mPn6020255; Sun, 7 Aug 2005 23:48:27 -0400 (EDT) To: Patrick Hatcher Cc: John A Meinel , "pgsql-performance@postgresql.org" Subject: Re: Slow update statement In-reply-to: <42F6BCD3.4070600@comcast.net> References: <42F4B812.7070501@comcast.net> <42F4BC61.20102@arbash-meinel.com> <42F6BCD3.4070600@comcast.net> Comments: In-reply-to Patrick Hatcher message dated "Sun, 07 Aug 2005 19:00:51 -0700" Date: Sun, 07 Aug 2005 23:48:25 -0400 Message-ID: <20254.1123472905@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/59 X-Sequence-Number: 13806 Patrick Hatcher writes: > Hash Join (cost=1246688.42..4127248.31 rows=12702676 width=200) > Hash Cond: ("outer".cus_num = "inner".cus_nbr) > -> Seq Scan on bcp_ddw_ck_cus b (cost=0.00..195690.76 rows=12702676 > width=16) > -> Hash (cost=874854.34..874854.34 rows=12880834 width=192) > -> Seq Scan on cdm_ddw_customer (cost=0.00..874854.34 > rows=12880834 width=192) Yipes, that's a bit of a large hash table, if the planner's estimates are on-target. What do you have work_mem (sort_mem if pre 8.0) set to, and how does that compare to actual available RAM? I'm thinking you might have set work_mem too large and the thing is now swap-thrashing. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 8 01:38:13 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CB36C52C28 for ; Mon, 8 Aug 2005 01:35:51 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 66084-02 for ; Mon, 8 Aug 2005 04:35:43 +0000 (GMT) Received: from sccrmhc13.comcast.net (sccrmhc13.comcast.net [204.127.202.64]) by svr1.postgresql.org (Postfix) with ESMTP id B7C7F52B73 for ; Mon, 8 Aug 2005 01:35:41 -0300 (ADT) Received: from [127.0.0.1] (c-24-4-66-2.hsd1.ca.comcast.net[24.4.66.2]) by comcast.net (sccrmhc13) with ESMTP id <2005080804353901300dl953e>; Mon, 8 Aug 2005 04:35:39 +0000 Message-ID: <42F6E118.20206@comcast.net> Date: Sun, 07 Aug 2005 21:35:36 -0700 From: Patrick Hatcher User-Agent: Mozilla Thunderbird 1.0.2 (Windows/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: John A Meinel , "pgsql-performance@postgresql.org" Subject: Re: Slow update statement References: <42F4B812.7070501@comcast.net> <42F4BC61.20102@arbash-meinel.com> <42F6BCD3.4070600@comcast.net> <20254.1123472905@sss.pgh.pa.us> In-Reply-To: <20254.1123472905@sss.pgh.pa.us> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.719 tagged_above=0 required=5 tests=AWL, DNS_FROM_RFC_ABUSE, DNS_FROM_RFC_POST, DNS_FROM_RFC_WHOIS X-Spam-Level: * X-Archive-Number: 200508/60 X-Sequence-Number: 13807 At the time this was the only process running on the box so I set sort_mem= 228000; It's a 12G box. Tom Lane wrote: >Patrick Hatcher writes: > > >> Hash Join (cost=1246688.42..4127248.31 rows=12702676 width=200) >> Hash Cond: ("outer".cus_num = "inner".cus_nbr) >> -> Seq Scan on bcp_ddw_ck_cus b (cost=0.00..195690.76 rows=12702676 >>width=16) >> -> Hash (cost=874854.34..874854.34 rows=12880834 width=192) >> -> Seq Scan on cdm_ddw_customer (cost=0.00..874854.34 >>rows=12880834 width=192) >> >> > >Yipes, that's a bit of a large hash table, if the planner's estimates >are on-target. What do you have work_mem (sort_mem if pre 8.0) set to, >and how does that compare to actual available RAM? I'm thinking you >might have set work_mem too large and the thing is now swap-thrashing. > > regards, tom lane > > > From pgsql-performance-owner@postgresql.org Mon Aug 8 09:05:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id F1DA752A7B for ; Mon, 8 Aug 2005 09:05:11 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 47716-08 for ; Mon, 8 Aug 2005 12:05:00 +0000 (GMT) Received: from purple.bdb.fi (purple.bdb.fi [195.197.212.62]) by svr1.postgresql.org (Postfix) with ESMTP id D8E7852C2D for ; Mon, 8 Aug 2005 09:04:58 -0300 (ADT) Received: by purple.bdb.fi (Postfix, from userid 101) id BE7FA3261; Mon, 8 Aug 2005 15:03:21 +0300 (EETDST) Received: from localhost (localhost [127.0.0.1]) by purple.bdb.fi (Postfix) with ESMTP id B50484972; Mon, 8 Aug 2005 15:03:21 +0300 (EETDST) Date: Mon, 8 Aug 2005 15:03:21 +0300 (EETDST) From: Kari Lavikka To: Luke Lonergan Cc: pgsql-performance@postgresql.org Subject: Re: Finding bottleneck In-Reply-To: Message-ID: References: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/61 X-Sequence-Number: 13808 Hi! Oprofile looks quite interesting. I'm not very familiar with postgresql internals, but here's some report output: CPU: AMD64 processors, speed 2190.23 MHz (estimated) Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit mask of 0x00 (No unit mask) count 100000 samples % symbol name 13513390 16.0074 AtEOXact_CatCache 4492257 5.3213 StrategyGetBuffer 2279285 2.6999 AllocSetAlloc 2121509 2.5130 LWLockAcquire 2023574 2.3970 hash_seq_search 1971358 2.3352 nocachegetattr 1837168 2.1762 GetSnapshotData 1793693 2.1247 SearchCatCache 1777385 2.1054 hash_search 1460804 1.7304 ExecMakeFunctionResultNoSets 1360930 1.6121 _bt_compare 1344604 1.5928 yyparse 1318407 1.5617 LWLockRelease 1290814 1.5290 FunctionCall2 1137544 1.3475 ExecEvalVar 1102236 1.3057 hash_any 912677 1.0811 OpernameGetCandidates 877993 1.0400 ReadBufferInternal 783908 0.9286 TransactionIdPrecedes 772886 0.9155 MemoryContextAllocZeroAligned 679768 0.8052 StrategyBufferLookup 609339 0.7218 equal 600584 0.7114 PGSemaphoreLock And btw, I tried to strace lingering queries under different loads. When number of concurrent queries increases, lseek and read syscalls stay within quite constant limits but number of semop calls quadruples. Are there some buffer locking issues? |\__/| ( oo ) Kari Lavikka - tuner@bdb.fi - (050) 380 3808 __ooO( )Ooo_______ _____ ___ _ _ _ _ _ _ _ "" On Thu, 28 Jul 2005, Luke Lonergan wrote: > On 7/28/05 2:21 AM, "Kari Lavikka" wrote: > > There's a new profiling tool called oprofile: > > http://oprofile.sourceforge.net/download/ > > that can be run without instrumenting the binaries beforehand. To actually > find out what the code is doing during these stalls, oprofile can show you > in which routines the CPU is spending time when you start/stop the > profiling. > > As an alternative to the "guess->change parameters->repeat" approach, this > is the most direct way to find the exact nature of the problem. > > - Luke > > > > ---------------------------(end of broadcast)--------------------------- > TIP 5: don't forget to increase your free space map settings > From pgsql-performance-owner@postgresql.org Mon Aug 8 11:39:27 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 404C052B49 for ; Mon, 8 Aug 2005 11:39:16 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 52922-08 for ; Mon, 8 Aug 2005 14:39:12 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 2976452B0B for ; Mon, 8 Aug 2005 11:39:11 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j78EdF7H028286; Mon, 8 Aug 2005 10:39:15 -0400 (EDT) To: Kari Lavikka Cc: pgsql-performance@postgresql.org Subject: Re: Finding bottleneck In-reply-to: References: Comments: In-reply-to Kari Lavikka message dated "Mon, 08 Aug 2005 15:03:21 +0300" Date: Mon, 08 Aug 2005 10:39:15 -0400 Message-ID: <28285.1123511955@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/62 X-Sequence-Number: 13809 Kari Lavikka writes: > samples % symbol name > 13513390 16.0074 AtEOXact_CatCache That seems quite odd --- I'm not used to seeing that function at the top of a profile. What is the workload being profiled, exactly? regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 8 12:19:13 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id DB03452DAD for ; Mon, 8 Aug 2005 12:19:05 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 17017-05 for ; Mon, 8 Aug 2005 15:18:56 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 43FF552B53 for ; Mon, 8 Aug 2005 12:18:55 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j78FIsZB028598; Mon, 8 Aug 2005 11:18:54 -0400 (EDT) To: Patrick Hatcher Cc: "pgsql-performance@postgresql.org" Subject: Re: Slow update statement In-reply-to: <42F6BEC0.9090504@comcast.net> References: <42F4B812.7070501@comcast.net> <17916.1123337563@sss.pgh.pa.us> <42F6BEC0.9090504@comcast.net> Comments: In-reply-to Patrick Hatcher message dated "Sun, 07 Aug 2005 19:09:04 -0700" Date: Mon, 08 Aug 2005 11:18:53 -0400 Message-ID: <28597.1123514333@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/63 X-Sequence-Number: 13810 Patrick Hatcher writes: > Here's the table layout. It's the first time I noticed this, but there > is a PK on the cus_nbr and an index. Does really need to be both and > could this be causing the issue? I thought that if a primary key was > designated, it was automatically indexed.: The duplicate index is certainly a waste, but it's no more expensive to maintain than any other index would be; it doesn't seem likely that that would account for any huge slowdown. A long-shot theory occurs to me upon noticing that your join keys are int8: 7.4 had a pretty bad hash function for int8, to wit it took the low order half of the integer and ignored the high order half. For ordinary distributions of key values this made no difference, but I recall seeing at least one real-world case where the information was all in the high half of the key, and so the hash join degenerated to a sequential search because all the entries went into the same hash bucket. Were you assigning cus_nbrs nonsequentially by any chance? regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 8 12:24:57 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 182DD528FA for ; Mon, 8 Aug 2005 12:24:55 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 54979-01 for ; Mon, 8 Aug 2005 15:24:45 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id 37FB1529BE for ; Mon, 8 Aug 2005 12:24:44 -0300 (ADT) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: Finding bottleneck Date: Mon, 8 Aug 2005 11:24:42 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3417DD026@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Finding bottleneck Thread-Index: AcWcJ0fWQTl93s16T6yCKgBHt7uf3QABbjaw From: "Merlin Moncure" To: "Tom Lane" Cc: , "Kari Lavikka" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.055 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/64 X-Sequence-Number: 13811 > Kari Lavikka writes: > > samples % symbol name > > 13513390 16.0074 AtEOXact_CatCache >=20 > That seems quite odd --- I'm not used to seeing that function at the top > of a profile. What is the workload being profiled, exactly? He is running a commit_delay of 80000. Could that be playing a role? Merlin From pgsql-performance-owner@postgresql.org Mon Aug 8 12:37:31 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 01B755293B for ; Mon, 8 Aug 2005 12:37:30 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 66243-10 for ; Mon, 8 Aug 2005 15:37:22 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id DBDC8528CF for ; Mon, 8 Aug 2005 12:37:21 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j78FbKcc028802; Mon, 8 Aug 2005 11:37:20 -0400 (EDT) To: "Merlin Moncure" Cc: pgsql-performance@postgresql.org, "Kari Lavikka" Subject: Re: Finding bottleneck In-reply-to: <6EE64EF3AB31D5448D0007DD34EEB3417DD026@Herge.rcsinc.local> References: <6EE64EF3AB31D5448D0007DD34EEB3417DD026@Herge.rcsinc.local> Comments: In-reply-to "Merlin Moncure" message dated "Mon, 08 Aug 2005 11:24:42 -0400" Date: Mon, 08 Aug 2005 11:37:19 -0400 Message-ID: <28801.1123515439@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/65 X-Sequence-Number: 13812 "Merlin Moncure" writes: >> Kari Lavikka writes: >>> samples % symbol name >>> 13513390 16.0074 AtEOXact_CatCache >> >> That seems quite odd --- I'm not used to seeing that function at the top >> of a profile. What is the workload being profiled, exactly? > He is running a commit_delay of 80000. Could that be playing a role? It wouldn't cause AtEOXact_CatCache to suddenly get expensive. (I have little or no faith in the value of nonzero commit_delay, though.) regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 8 13:21:04 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id F142D52967 for ; Mon, 8 Aug 2005 13:21:02 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 18584-01 for ; Mon, 8 Aug 2005 16:20:56 +0000 (GMT) Received: from purple.bdb.fi (purple.bdb.fi [195.197.212.62]) by svr1.postgresql.org (Postfix) with ESMTP id 9619B529BE for ; Mon, 8 Aug 2005 13:20:49 -0300 (ADT) Received: by purple.bdb.fi (Postfix, from userid 101) id 2915831AD; Mon, 8 Aug 2005 19:19:09 +0300 (EETDST) Received: from localhost (localhost [127.0.0.1]) by purple.bdb.fi (Postfix) with ESMTP id 1EB484972; Mon, 8 Aug 2005 19:19:09 +0300 (EETDST) Date: Mon, 8 Aug 2005 19:19:09 +0300 (EETDST) From: Kari Lavikka To: Merlin Moncure Cc: Tom Lane , pgsql-performance@postgresql.org Subject: Re: Finding bottleneck In-Reply-To: <6EE64EF3AB31D5448D0007DD34EEB3417DD026@Herge.rcsinc.local> Message-ID: References: <6EE64EF3AB31D5448D0007DD34EEB3417DD026@Herge.rcsinc.local> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/66 X-Sequence-Number: 13813 Actually I modified postgresql.conf a bit and there isn't commit delay any more. That didn't make noticeable difference though.. Workload is generated by a website with about 1000 dynamic page views a second. Finland's biggest site among youths btw. Anyway, there are about 70 tables and here's some of the most important: relname | reltuples ----------------------------------+------------- comment | 1.00723e+08 comment_archive | 9.12764e+07 channel_comment | 6.93912e+06 image | 5.80314e+06 admin_event | 5.1936e+06 user_channel | 3.36877e+06 users | 325929 channel | 252267 Queries to "comment" table are mostly IO-bound but are performing quite well. Here's an example: (SELECT u.nick, c.comment, c.private, c.admin, c.visible, c.parsable, c.uid_sender, to_char(c.stamp, 'DD.MM.YY HH24:MI') AS stamp, c.comment_id FROM comment c INNER JOIN users u ON u.uid = c.uid_sender WHERE u.status = 'a' AND c.image_id = 15500900 AND c.uid_target = 780345 ORDER BY uid_target DESC, image_id DESC, c.comment_id DESC) LIMIT 36 And explain analyze: Limit (cost=0.00..6.81 rows=1 width=103) (actual time=0.263..17.522 rows=12 loops=1) -> Nested Loop (cost=0.00..6.81 rows=1 width=103) (actual time=0.261..17.509 rows=12 loops=1) -> Index Scan Backward using comment_uid_target_image_id_comment_id_20050527 on "comment" c (cost=0.00..3.39 rows=1 width=92) (actual time=0.129..16.213 rows=12 loops=1) Index Cond: ((uid_target = 780345) AND (image_id = 15500900)) -> Index Scan using users_pkey on users u (cost=0.00..3.40 rows=1 width=15) (actual time=0.084..0.085 rows=1 loops=12) Index Cond: (u.uid = "outer".uid_sender) Filter: (status = 'a'::bpchar) Total runtime: 17.653 ms We are having performance problems with some smaller tables and very simple queries. For example: SELECT u.uid, u.nick, extract(epoch from uc.stamp) AS stamp FROM user_channel uc INNER JOIN users u USING (uid) WHERE channel_id = 281321 AND u.status = 'a' ORDER BY uc.channel_id, upper(uc.nick) And explain analyze: Nested Loop (cost=0.00..200.85 rows=35 width=48) (actual time=0.414..38.128 rows=656 loops=1) -> Index Scan using user_channel_channel_id_nick on user_channel uc (cost=0.00..40.18 rows=47 width=27) (actual time=0.090..0.866 rows=667 loops=1) Index Cond: (channel_id = 281321) -> Index Scan using users_pkey on users u (cost=0.00..3.40 rows=1 width=25) (actual time=0.048..0.051 rows=1 loops=667) Index Cond: ("outer".uid = u.uid) Filter: (status = 'a'::bpchar) Total runtime: 38.753 ms Under heavy load these queries tend to take several minutes to execute although there's plenty of free cpu available. There aren't any blocking locks in pg_locks. |\__/| ( oo ) Kari Lavikka - tuner@bdb.fi - (050) 380 3808 __ooO( )Ooo_______ _____ ___ _ _ _ _ _ _ _ "" On Mon, 8 Aug 2005, Merlin Moncure wrote: >> Kari Lavikka writes: >>> samples % symbol name >>> 13513390 16.0074 AtEOXact_CatCache >> >> That seems quite odd --- I'm not used to seeing that function at the > top >> of a profile. What is the workload being profiled, exactly? > > He is running a commit_delay of 80000. Could that be playing a role? > > Merlin > > ---------------------------(end of broadcast)--------------------------- > TIP 1: if posting/reading through Usenet, please send an appropriate > subscribe-nomail command to majordomo@postgresql.org so that your > message can get through to the mailing list cleanly > From pgsql-performance-owner@postgresql.org Mon Aug 8 13:56:52 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D392D528CF for ; Mon, 8 Aug 2005 13:56:50 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 15373-02 for ; Mon, 8 Aug 2005 16:56:42 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 4061C5291E for ; Mon, 8 Aug 2005 13:56:41 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j78GueDg029385; Mon, 8 Aug 2005 12:56:40 -0400 (EDT) To: Kari Lavikka Cc: Merlin Moncure , pgsql-performance@postgresql.org Subject: Re: Finding bottleneck In-reply-to: References: <6EE64EF3AB31D5448D0007DD34EEB3417DD026@Herge.rcsinc.local> Comments: In-reply-to Kari Lavikka message dated "Mon, 08 Aug 2005 19:19:09 +0300" Date: Mon, 08 Aug 2005 12:56:40 -0400 Message-ID: <29384.1123520200@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/67 X-Sequence-Number: 13814 Kari Lavikka writes: > We are having performance problems with some smaller tables and very > simple queries. For example: > SELECT u.uid, u.nick, extract(epoch from uc.stamp) AS stamp FROM > user_channel uc INNER JOIN users u USING (uid) WHERE channel_id = 281321 > AND u.status = 'a' ORDER BY uc.channel_id, upper(uc.nick) > And explain analyze: > Nested Loop (cost=0.00..200.85 rows=35 width=48) (actual time=0.414..38.128 rows=656 loops=1) > -> Index Scan using user_channel_channel_id_nick on user_channel uc (cost=0.00..40.18 rows=47 width=27) (actual time=0.090..0.866 rows=667 loops=1) > Index Cond: (channel_id = 281321) > -> Index Scan using users_pkey on users u (cost=0.00..3.40 rows=1 width=25) (actual time=0.048..0.051 rows=1 loops=667) > Index Cond: ("outer".uid = u.uid) > Filter: (status = 'a'::bpchar) > Total runtime: 38.753 ms > Under heavy load these queries tend to take several minutes to execute > although there's plenty of free cpu available. What that sounds like to me is a machine with inadequate disk I/O bandwidth. Your earlier comment that checkpoint drives the machine into the ground fits right into that theory, too. You said there is "almost no IO-wait" but are you sure you are measuring that correctly? Something else just struck me from your first post: > Queries accumulate and when checkpointing is over, there can be > something like 400 queries running but over 50% of cpu is just idling. 400 queries? Are you launching 400 separate backends to do that? Some sort of connection pooling seems like a good idea, if you don't have it in place already. If the system's effective behavior in the face of heavy load is to start even more concurrent backends, that could easily drive things into the ground. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 8 14:00:25 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 880095291E for ; Mon, 8 Aug 2005 14:00:24 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 60101-01 for ; Mon, 8 Aug 2005 17:00:12 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.198]) by svr1.postgresql.org (Postfix) with ESMTP id CEDB3528FA for ; Mon, 8 Aug 2005 14:00:08 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i5so820164wra for ; Mon, 08 Aug 2005 10:00:09 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=kIhXGdM880DrfcGjiVXAzVb3ra1nuikKkWzgOgV/N/i2AWfAJVH4z7Kr/egp25ybvDqvRouBhOOiL31s74OQJSi+qI4N9oA9CgL8V6Iy/Mj9QxheduQ934d2qou+EVa+vkH4ymVJ5VH/S24vniSbVtU2pHN9TfP8sXMYcTr2lXk= Received: by 10.54.11.20 with SMTP id 20mr4996338wrk; Mon, 08 Aug 2005 10:00:09 -0700 (PDT) Received: by 10.54.119.7 with HTTP; Mon, 8 Aug 2005 10:00:08 -0700 (PDT) Message-ID: <41b0fe89050808100021a2a4fc@mail.gmail.com> Date: Mon, 8 Aug 2005 10:00:08 -0700 From: Rhett Garber Reply-To: Rhett Garber To: Tom Lane Subject: Re: Why hash join instead of nested loop? Cc: =?ISO-8859-1?Q?Havasv=F6lgyi_Ott=F3?= , pgsql-performance@postgresql.org In-Reply-To: <15425.1123304631@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <41b0fe890508051135712a086b@mail.gmail.com> <001801c59a0a$f4be7480$9a00a8c0@OTTO> <41b0fe8905080516164464e04f@mail.gmail.com> <15425.1123304631@sss.pgh.pa.us> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 tagged_above=0 required=5 tests=RCVD_BY_IP X-Spam-Level: X-Archive-Number: 200508/68 X-Sequence-Number: 13815 This is postgres 7.4.1 All the rows involved are integers. Thanks, Rhett On 8/5/05, Tom Lane wrote: > Rhett Garber writes: > > Hash Join (cost=3D5.96..7.04 rows=3D1 width=3D14) (actual > > time=3D10.591..10.609 rows=3D1 loops=3D1) > > Hash Cond: ("outer".id =3D "inner".obj2) > > -> Seq Scan on rtmessagestate (cost=3D0.00..1.05 rows=3D5 width=3D= 14) > > (actual time=3D0.011..0.022 rows=3D5 loops=3D1) > > -> Hash (cost=3D5.96..5.96 rows=3D1 width=3D4) (actual > > time=3D0.109..0.109 rows=3D0 loops=3D1) > > -> Index Scan using connection_regid_obj1_index on > > connection (cost=3D0.00..5.96 rows=3D1 width=3D4) (actual time=3D0.070= ..0.076 > > rows=3D1 loops=3D1) > > Index Cond: ((connection_registry_id =3D 40105) AND (obj= 1 > > =3D 73582)) Total runtime: 11.536 ms > > (7 rows) >=20 > [ scratches head... ] If the hash table build takes only 0.109 msec > and loads only one row into the hash table, and the scan of > rtmessagestate takes only 0.022 msec and produces only 5 rows, it is > real hard to see how the join takes 10.609 msec overall. Unless the id > and obj2 columns are of a datatype with an incredibly slow equality > function. What is the datatype involved here, anyway? And what PG > version are we speaking of? >=20 > regards, tom lane > From pgsql-performance-owner@postgresql.org Mon Aug 8 14:56:28 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0B91D52E41 for ; Mon, 8 Aug 2005 14:56:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 63551-07 for ; Mon, 8 Aug 2005 17:56:21 +0000 (GMT) Received: from purple.bdb.fi (purple.bdb.fi [195.197.212.62]) by svr1.postgresql.org (Postfix) with ESMTP id BE59E52E38 for ; Mon, 8 Aug 2005 14:56:17 -0300 (ADT) Received: by purple.bdb.fi (Postfix, from userid 101) id BE3013B1E; Mon, 8 Aug 2005 20:54:38 +0300 (EETDST) Received: from localhost (localhost [127.0.0.1]) by purple.bdb.fi (Postfix) with ESMTP id B6DBE4972; Mon, 8 Aug 2005 20:54:38 +0300 (EETDST) Date: Mon, 8 Aug 2005 20:54:38 +0300 (EETDST) From: Kari Lavikka To: Tom Lane Cc: pgsql-performance@postgresql.org Subject: Re: Finding bottleneck In-Reply-To: <29384.1123520200@sss.pgh.pa.us> Message-ID: References: <6EE64EF3AB31D5448D0007DD34EEB3417DD026@Herge.rcsinc.local> <29384.1123520200@sss.pgh.pa.us> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/69 X-Sequence-Number: 13816 On Mon, 8 Aug 2005, Tom Lane wrote: > What that sounds like to me is a machine with inadequate disk I/O bandwidth. > Your earlier comment that checkpoint drives the machine into the ground > fits right into that theory, too. You said there is "almost no IO-wait" > but are you sure you are measuring that correctly? Currently there's some iowait caused by "fragmentation" of the comment table. Periodic clustering helps a lot. Disk configurations looks something like this: sda: data (10 spindles, raid10) sdb: xlog & clog (2 spindles, raid1) sdc: os and other stuff Usually iostat (2 second interval) says: avg-cpu: %user %nice %sys %iowait %idle 32.38 0.00 12.88 11.62 43.12 Device: tps kB_read/s kB_wrtn/s kB_read kB_wrtn sda 202.00 1720.00 0.00 3440 0 sdb 152.50 4.00 2724.00 8 5448 sdc 0.00 0.00 0.00 0 0 And during checkpoint: avg-cpu: %user %nice %sys %iowait %idle 31.25 0.00 14.75 54.00 0.00 Device: tps kB_read/s kB_wrtn/s kB_read kB_wrtn sda 3225.50 1562.00 35144.00 3124 70288 sdb 104.50 10.00 2348.00 20 4696 sdc 0.00 0.00 0.00 0 0 I think (insufficiency of) disk IO shouldn't cause those lingering queries because dataset is rather small and it's continuously accessed. It should fit into cache and stay there(?) > 400 queries? Are you launching 400 separate backends to do that? Well yes. That's the common problem with php and persistent connections. > Some sort of connection pooling seems like a good idea, if you don't > have it in place already. pg_pool for example? I'm planning to give it a try. > regards, tom lane |\__/| ( oo ) Kari Lavikka - tuner@bdb.fi - (050) 380 3808 __ooO( )Ooo_______ _____ ___ _ _ _ _ _ _ _ "" From pgsql-performance-owner@postgresql.org Mon Aug 8 16:27:38 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7FBC7529B4 for ; Mon, 8 Aug 2005 16:27:34 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 96590-04 for ; Mon, 8 Aug 2005 19:27:31 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 1B0995290F for ; Mon, 8 Aug 2005 16:27:26 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j78JRP9J008435; Mon, 8 Aug 2005 15:27:25 -0400 (EDT) To: Kari Lavikka Cc: pgsql-performance@postgresql.org Subject: Re: Finding bottleneck In-reply-to: References: <6EE64EF3AB31D5448D0007DD34EEB3417DD026@Herge.rcsinc.local> <29384.1123520200@sss.pgh.pa.us> Comments: In-reply-to Kari Lavikka message dated "Mon, 08 Aug 2005 20:54:38 +0300" Date: Mon, 08 Aug 2005 15:27:25 -0400 Message-ID: <8434.1123529245@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/70 X-Sequence-Number: 13817 Kari Lavikka writes: > Disk configurations looks something like this: > sda: data (10 spindles, raid10) > sdb: xlog & clog (2 spindles, raid1) > sdc: os and other stuff That's definitely wrong. Put clog on the data disk. The entire point of giving xlog its own spindle is that you don't ever want the disk heads moving off the current xlog file. I'm not sure how much this is hurting you, given that clog is relatively low volume, but if you're going to go to the trouble of putting xlog on a separate spindle then it should be a completely dedicated spindle. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 8 21:58:39 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3D34852CBB for ; Mon, 8 Aug 2005 21:58:36 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 06129-06 for ; Tue, 9 Aug 2005 00:58:32 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 3703052803 for ; Mon, 8 Aug 2005 21:58:31 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j790wQPD010931; Mon, 8 Aug 2005 20:58:26 -0400 (EDT) To: Rhett Garber Cc: =?ISO-8859-1?Q?Havasv=F6lgyi_Ott=F3?= , pgsql-performance@postgresql.org Subject: Re: Why hash join instead of nested loop? In-reply-to: <41b0fe89050808100021a2a4fc@mail.gmail.com> References: <41b0fe890508051135712a086b@mail.gmail.com> <001801c59a0a$f4be7480$9a00a8c0@OTTO> <41b0fe8905080516164464e04f@mail.gmail.com> <15425.1123304631@sss.pgh.pa.us> <41b0fe89050808100021a2a4fc@mail.gmail.com> Comments: In-reply-to Rhett Garber message dated "Mon, 08 Aug 2005 10:00:08 -0700" Date: Mon, 08 Aug 2005 20:58:26 -0400 Message-ID: <10930.1123549106@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/71 X-Sequence-Number: 13818 Rhett Garber writes: > This is postgres 7.4.1 > All the rows involved are integers. Hmph. There is something really strange going on here. I tried to duplicate your problem in 7.4.*, thus: regression=# create table rtmessagestate(id int, f1 char(6)); CREATE TABLE regression=# insert into rtmessagestate values(1,'z'); INSERT 559399 1 regression=# insert into rtmessagestate values(2,'z'); INSERT 559400 1 regression=# insert into rtmessagestate values(3,'z'); INSERT 559401 1 regression=# insert into rtmessagestate values(4,'z'); INSERT 559402 1 regression=# insert into rtmessagestate values(5,'z'); INSERT 559403 1 regression=# vacuum analyze rtmessagestate; VACUUM regression=# create table connection(connection_registry_id int, obj1 int, obj2 int); CREATE TABLE regression=# create index connection_regid_obj1_index on connection(connection_registry_id,obj1); CREATE INDEX regression=# insert into connection values(40105,73582,3); INSERT 559407 1 regression=# explain analyze select rtmessagestate.* from rtmessagestate,connection where (connection_registry_id = 40105) AND (obj1 = 73582) and id = obj2; QUERY PLAN ---------------------------------------------------------------------------------------------------------------------------------------------------- Hash Join (cost=4.83..5.91 rows=1 width=14) (actual time=0.498..0.544 rows=1 loops=1) Hash Cond: ("outer".id = "inner".obj2) -> Seq Scan on rtmessagestate (cost=0.00..1.05 rows=5 width=14) (actual time=0.030..0.072 rows=5 loops=1) -> Hash (cost=4.83..4.83 rows=1 width=4) (actual time=0.305..0.305 rows=0 loops=1) -> Index Scan using connection_regid_obj1_index on connection (cost=0.00..4.83 rows=1 width=4) (actual time=0.236..0.264 rows=1 loops=1) Index Cond: ((connection_registry_id = 40105) AND (obj1 = 73582)) Total runtime: 1.119 ms (7 rows) This duplicates your example as to plan and row counts: > Hash Join (cost=5.96..7.04 rows=1 width=14) (actual > time=10.591..10.609 rows=1 loops=1) > Hash Cond: ("outer".id = "inner".obj2) > -> Seq Scan on rtmessagestate (cost=0.00..1.05 rows=5 width=14) > (actual time=0.011..0.022 rows=5 loops=1) > -> Hash (cost=5.96..5.96 rows=1 width=4) (actual > time=0.109..0.109 rows=0 loops=1) > -> Index Scan using connection_regid_obj1_index on > connection (cost=0.00..5.96 rows=1 width=4) (actual time=0.070..0.076 > rows=1 loops=1) > Index Cond: ((connection_registry_id = 40105) AND (obj1 > = 73582)) Total runtime: 11.536 ms > (7 rows) My machine is considerably slower than yours, to judge by the actual elapsed times in the scan nodes ... so why is it beating the pants off yours in the join step? Can you try the above script verbatim in a scratch database and see what you get? (Note it's worth trying the explain two or three times to be sure the values have settled out.) I'm testing a fairly recent 7.4-branch build (7.4.8 plus), so that's one possible reason for the discrepancy between my results and yours, but I do not see anything in the 7.4 CVS logs that looks like it's related to hashjoin performance. I'd be interested to see results from other people using 7.4.* too. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 8 22:09:04 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 98E9452800 for ; Mon, 8 Aug 2005 22:09:01 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 46006-03 for ; Tue, 9 Aug 2005 01:08:50 +0000 (GMT) Received: from cassarossa.samfundet.no (cassarossa.samfundet.no [129.241.93.19]) by svr1.postgresql.org (Postfix) with ESMTP id 245B0528F6 for ; Mon, 8 Aug 2005 22:08:49 -0300 (ADT) Received: from trofast.ipv6.sesse.net ([2001:700:300:dc03:20e:cff:fe36:a766] helo=trofast.sesse.net) by cassarossa.samfundet.no with esmtp (Exim 4.50) id 1E2Ibr-00069y-Cc for pgsql-performance@postgresql.org; Tue, 09 Aug 2005 03:08:51 +0200 Received: from sesse by trofast.sesse.net with local (Exim 3.36 #1 (Debian)) id 1E2Ibq-00052V-00 for ; Tue, 09 Aug 2005 03:08:50 +0200 Date: Tue, 9 Aug 2005 03:08:50 +0200 From: "Steinar H. Gunderson" To: pgsql-performance@postgresql.org Subject: Re: Why hash join instead of nested loop? Message-ID: <20050809010850.GA18938@uio.no> Mail-Followup-To: pgsql-performance@postgresql.org References: <41b0fe890508051135712a086b@mail.gmail.com> <001801c59a0a$f4be7480$9a00a8c0@OTTO> <41b0fe8905080516164464e04f@mail.gmail.com> <15425.1123304631@sss.pgh.pa.us> <41b0fe89050808100021a2a4fc@mail.gmail.com> <10930.1123549106@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <10930.1123549106@sss.pgh.pa.us> X-Operating-System: Linux 2.6.11.8 on a i686 X-Message-Flag: Outlook? --> http://www.mozilla.org/products/thunderbird/ User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.017 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/72 X-Sequence-Number: 13819 On Mon, Aug 08, 2005 at 08:58:26PM -0400, Tom Lane wrote: > Hmph. There is something really strange going on here. I tried to > duplicate your problem in 7.4.*, thus: PostgreSQL 7.4.7 (Debian sarge): regression=# explain analyze select rtmessagestate.* from rtmessagestate,connection where (connection_registry_id = 40105) AND (obj1 = 73582) and id = obj2; QUERY PLAN ---------------------------------------------------------------------------------------------------------------------------------------------------- Hash Join (cost=4.83..5.91 rows=1 width=14) (actual time=0.155..0.159 rows=1 loops=1) Hash Cond: ("outer".id = "inner".obj2) -> Seq Scan on rtmessagestate (cost=0.00..1.05 rows=5 width=14) (actual time=0.003..0.006 rows=5 loops=1) -> Hash (cost=4.83..4.83 rows=1 width=4) (actual time=0.026..0.026 rows=0 loops=1) -> Index Scan using connection_regid_obj1_index on connection (cost=0.00..4.83 rows=1 width=4) (actual time=0.011..0.012 rows=1 loops=1) Index Cond: ((connection_registry_id = 40105) AND (obj1 = 73582)) Total runtime: 0.215 ms (7 rows) This is an Opteron (in 32-bit mode), though. /* Steinar */ -- Homepage: http://www.sesse.net/ From pgsql-performance-owner@postgresql.org Tue Aug 9 01:33:21 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 844D3528DB for ; Tue, 9 Aug 2005 01:33:19 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 35436-03 for ; Tue, 9 Aug 2005 04:33:18 +0000 (GMT) Received: from floppy.pyrenet.fr (news.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id A9624528C1 for ; Tue, 9 Aug 2005 01:33:17 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id 058DC30959; Tue, 9 Aug 2005 06:42:52 +0200 (MET DST) From: "Qingqing Zhou" X-Newsgroups: pgsql.performance Subject: Re: QRY seems not using indexes Date: Tue, 9 Aug 2005 12:30:41 +0800 Organization: Hub.Org Networking Services Lines: 25 Message-ID: References: <1123071402.641229.314250@g43g2000cwa.googlegroups.com> Reply-To: "Qingqing Zhou" X-Complaints-To: usenet@news.hub.org X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 6.00.2800.1506 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1506 To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.184 tagged_above=0 required=5 tests=AWL, PRIORITY_NO_NAME X-Spam-Level: * X-Archive-Number: 200508/73 X-Sequence-Number: 13820 writes > > > so, if I do a qry like "EXPLAIN ANALYZE select * from pridecdr where > idsede=8977758488" it tooks a lot of time before i get back any result: > > Index Scan using prd_id_sede on pridecdr (cost=0.00..699079.90 > rows=181850 width=138) (actual time=51.241..483068.255 rows=150511 > loops=1) > Index Cond: (idsede = 8977758488::bigint) > Total runtime: 483355.325 ms > The query plan looks ok. Try to do EXPLAIN ANALYZE twice and see if there is any difference. This could reduce the IO time to read your index/data since you got enough RAM. Also, if you haven't done VACUUM FULL for a long time, do so and compare the difference. Regards, Qingqing From pgsql-performance-owner@postgresql.org Tue Aug 9 02:05:48 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A016252930 for ; Tue, 9 Aug 2005 02:05:46 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 33565-06 for ; Tue, 9 Aug 2005 05:05:44 +0000 (GMT) Received: from linda-2.paradise.net.nz (bm-2a.paradise.net.nz [202.0.58.21]) by svr1.postgresql.org (Postfix) with ESMTP id 2EACC52928 for ; Tue, 9 Aug 2005 02:05:43 -0300 (ADT) Received: from smtp-3.paradise.net.nz (smtp-3a.paradise.net.nz [202.0.32.196]) by linda-2.paradise.net.nz (Paradise.net.nz) with ESMTP id <0IKX00NR7VHI1I@linda-2.paradise.net.nz> for pgsql-performance@postgresql.org; Tue, 09 Aug 2005 17:05:42 +1200 (NZST) Received: from [192.168.1.11] (218-101-13-165.paradise.net.nz [218.101.13.165]) by smtp-3.paradise.net.nz (Postfix) with ESMTP id F35C5AE067; Tue, 09 Aug 2005 17:05:41 +1200 (NZST) Date: Tue, 09 Aug 2005 17:05:39 +1200 From: Mark Kirkwood Subject: Re: QRY seems not using indexes In-reply-to: To: morandell@gmail.com Cc: Qingqing Zhou , pgsql-performance@postgresql.org Message-id: <42F839A3.2080501@paradise.net.nz> MIME-version: 1.0 Content-type: text/plain; format=flowed; charset=ISO-8859-1 Content-transfer-encoding: 7bit X-Accept-Language: en-us, en User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050726) References: <1123071402.641229.314250@g43g2000cwa.googlegroups.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.152 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/74 X-Sequence-Number: 13821 Qingqing Zhou wrote: > writes > >> >>so, if I do a qry like "EXPLAIN ANALYZE select * from pridecdr where >>idsede=8977758488" it tooks a lot of time before i get back any result: >> >>Index Scan using prd_id_sede on pridecdr (cost=0.00..699079.90 >>rows=181850 width=138) (actual time=51.241..483068.255 rows=150511 >>loops=1) >> Index Cond: (idsede = 8977758488::bigint) >> Total runtime: 483355.325 ms >> > > > The query plan looks ok. Try to do EXPLAIN ANALYZE twice and see if there is > any difference. This could reduce the IO time to read your index/data since > you got enough RAM. > > Also, if you haven't done VACUUM FULL for a long time, do so and compare the > difference. > Could also be libpq buffering all 150000 rows before showing any. It might be worthwhile using a CURSOR and doing 1 FETCH. If that is quick, then buffering is probably the issue. BTW - do you really want all the rows? Cheers Mark From pgsql-performance-owner@postgresql.org Tue Aug 9 02:12:12 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 99F9F528DB for ; Tue, 9 Aug 2005 02:12:10 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 47510-10 for ; Tue, 9 Aug 2005 05:12:01 +0000 (GMT) Received: from tigger.fuhr.org (tigger.fuhr.org [63.214.45.158]) by svr1.postgresql.org (Postfix) with ESMTP id 4249B5287F for ; Tue, 9 Aug 2005 02:12:00 -0300 (ADT) Received: from winnie.fuhr.org (winnie.fuhr.org [10.1.0.1]) by tigger.fuhr.org (8.13.3/8.13.3) with ESMTP id j795Bnnh044438 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK); Mon, 8 Aug 2005 23:11:52 -0600 (MDT) (envelope-from mfuhr@winnie.fuhr.org) Received: from winnie.fuhr.org (localhost [127.0.0.1]) by winnie.fuhr.org (8.13.3/8.13.3) with ESMTP id j795Bn6k085523; Mon, 8 Aug 2005 23:11:49 -0600 (MDT) (envelope-from mfuhr@winnie.fuhr.org) Received: (from mfuhr@localhost) by winnie.fuhr.org (8.13.3/8.13.3/Submit) id j795Bmmj085522; Mon, 8 Aug 2005 23:11:48 -0600 (MDT) (envelope-from mfuhr) Date: Mon, 8 Aug 2005 23:11:48 -0600 From: Michael Fuhr To: Tom Lane Cc: Rhett Garber , =?iso-8859-1?Q?Havasv=F6lgyi_Ott=F3?= , pgsql-performance@postgresql.org Subject: Re: Why hash join instead of nested loop? Message-ID: <20050809051148.GA85494@winnie.fuhr.org> References: <41b0fe890508051135712a086b@mail.gmail.com> <001801c59a0a$f4be7480$9a00a8c0@OTTO> <41b0fe8905080516164464e04f@mail.gmail.com> <15425.1123304631@sss.pgh.pa.us> <41b0fe89050808100021a2a4fc@mail.gmail.com> <10930.1123549106@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <10930.1123549106@sss.pgh.pa.us> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/75 X-Sequence-Number: 13822 On Mon, Aug 08, 2005 at 08:58:26PM -0400, Tom Lane wrote: > I'd be interested to see results from other people using 7.4.* too. I just built 7.4.1 on FreeBSD 4.11-STABLE and ran your test: test=# explain analyze select rtmessagestate.* from rtmessagestate,connection where (connection_registry_id = 40105) AND (obj1 = 73582) and id = obj2; QUERY PLAN ---------------------------------------------------------------------------------------------------------------------------------------------------- Hash Join (cost=4.83..5.91 rows=1 width=14) (actual time=0.220..0.264 rows=1 loops=1) Hash Cond: ("outer".id = "inner".obj2) -> Seq Scan on rtmessagestate (cost=0.00..1.05 rows=5 width=14) (actual time=0.015..0.050 rows=5 loops=1) -> Hash (cost=4.83..4.83 rows=1 width=4) (actual time=0.103..0.103 rows=0 loops=1) -> Index Scan using connection_regid_obj1_index on connection (cost=0.00..4.83 rows=1 width=4) (actual time=0.070..0.081 rows=1 loops=1) Index Cond: ((connection_registry_id = 40105) AND (obj1 = 73582)) Total runtime: 0.495 ms (7 rows) -- Michael Fuhr From pgsql-performance-owner@postgresql.org Tue Aug 9 09:16:08 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6A09852A55 for ; Tue, 9 Aug 2005 09:16:06 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 47906-04 for ; Tue, 9 Aug 2005 12:15:56 +0000 (GMT) Received: from mailhost.intellivid.com (mailhost.intellivid.com [64.32.200.11]) by svr1.postgresql.org (Postfix) with ESMTP id 32B3252A03 for ; Tue, 9 Aug 2005 09:15:53 -0300 (ADT) Received: from [192.168.2.68] (spectre.intellivid.com [192.168.2.68]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client did not present a certificate) by newmail.intellivid.com (Postfix) with ESMTP id 9F268F182AE; Tue, 9 Aug 2005 08:15:56 -0400 (EDT) Subject: Re: Why hash join instead of nested loop? From: Ian Westmacott To: Tom Lane Cc: pgsql-performance@postgresql.org In-Reply-To: <10930.1123549106@sss.pgh.pa.us> References: <41b0fe890508051135712a086b@mail.gmail.com> <001801c59a0a$f4be7480$9a00a8c0@OTTO> <41b0fe8905080516164464e04f@mail.gmail.com> <15425.1123304631@sss.pgh.pa.us> <41b0fe89050808100021a2a4fc@mail.gmail.com> <10930.1123549106@sss.pgh.pa.us> Content-Type: text/plain Organization: Intellivid Corp. Message-Id: <1123589756.15206.4.camel@spectre.intellivid.com> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Tue, 09 Aug 2005 08:15:56 -0400 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/76 X-Sequence-Number: 13823 On Mon, 2005-08-08 at 20:58, Tom Lane wrote: > I'd be interested to see results from other people using 7.4.* too. 7.4.8: QUERY PLAN ---------------------------------------------------------------------------------------------------------------------------------------------------- Hash Join (cost=4.83..5.91 rows=1 width=14) (actual time=0.122..0.126 rows=1 loops=1) Hash Cond: ("outer".id = "inner".obj2) -> Seq Scan on rtmessagestate (cost=0.00..1.05 rows=5 width=14) (actual time=0.003..0.006 rows=5 loops=1) -> Hash (cost=4.83..4.83 rows=1 width=4) (actual time=0.021..0.021 rows=0 loops=1) -> Index Scan using connection_regid_obj1_index on connection (cost=0.00..4.83 rows=1 width=4) (actual time=0.013..0.015 rows=1 loops=1) Index Cond: ((connection_registry_id = 40105) AND (obj1 = 73582)) Total runtime: 0.198 ms 7.4.2: QUERY PLAN ---------------------------------------------------------------------------------------------------------------------------------------------------- Hash Join (cost=4.83..5.91 rows=1 width=14) (actual time=0.577..0.600 rows=1 loops=1) Hash Cond: ("outer".id = "inner".obj2) -> Seq Scan on rtmessagestate (cost=0.00..1.05 rows=5 width=14) (actual time=0.006..0.023 rows=5 loops=1) -> Hash (cost=4.83..4.83 rows=1 width=4) (actual time=0.032..0.032 rows=0 loops=1) -> Index Scan using connection_regid_obj1_index on connection (cost=0.00..4.83 rows=1 width=4) (actual time=0.016..0.020 rows=1 loops=1) Index Cond: ((connection_registry_id = 40105) AND (obj1 = 73582)) Total runtime: 0.697 ms --Ian From pgsql-performance-owner@postgresql.org Tue Aug 9 11:36:25 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9DEE752AE5 for ; Tue, 9 Aug 2005 11:33:40 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 56198-06 for ; Tue, 9 Aug 2005 14:33:33 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 6923552A36 for ; Tue, 9 Aug 2005 11:33:32 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j79EXbji015882; Tue, 9 Aug 2005 10:33:37 -0400 (EDT) To: Ian Westmacott Cc: pgsql-performance@postgresql.org Subject: Re: Why hash join instead of nested loop? In-reply-to: <1123589756.15206.4.camel@spectre.intellivid.com> References: <41b0fe890508051135712a086b@mail.gmail.com> <001801c59a0a$f4be7480$9a00a8c0@OTTO> <41b0fe8905080516164464e04f@mail.gmail.com> <15425.1123304631@sss.pgh.pa.us> <41b0fe89050808100021a2a4fc@mail.gmail.com> <10930.1123549106@sss.pgh.pa.us> <1123589756.15206.4.camel@spectre.intellivid.com> Comments: In-reply-to Ian Westmacott message dated "Tue, 09 Aug 2005 08:15:56 -0400" Date: Tue, 09 Aug 2005 10:33:37 -0400 Message-ID: <15881.1123598017@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/77 X-Sequence-Number: 13824 Ian Westmacott writes: > On Mon, 2005-08-08 at 20:58, Tom Lane wrote: >> I'd be interested to see results from other people using 7.4.* too. > 7.4.8: > Total runtime: 0.198 ms > 7.4.2: > Total runtime: 0.697 ms Just to be clear: those are two different machines of different speeds, right? I don't believe we put any factor-of-three speedups into 7.4.* after release ;-) regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 9 11:50:47 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 084F352806 for ; Tue, 9 Aug 2005 11:43:40 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 97420-01 for ; Tue, 9 Aug 2005 14:43:30 +0000 (GMT) Received: from mailhost.intellivid.com (mailhost.intellivid.com [64.32.200.11]) by svr1.postgresql.org (Postfix) with ESMTP id CAD3A5293D for ; Tue, 9 Aug 2005 11:43:28 -0300 (ADT) Received: from [192.168.2.68] (spectre.intellivid.com [192.168.2.68]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client did not present a certificate) by newmail.intellivid.com (Postfix) with ESMTP id BD829F182AF; Tue, 9 Aug 2005 10:43:33 -0400 (EDT) Subject: Re: Why hash join instead of nested loop? From: Ian Westmacott To: Tom Lane Cc: pgsql-performance@postgresql.org In-Reply-To: <15881.1123598017@sss.pgh.pa.us> References: <41b0fe890508051135712a086b@mail.gmail.com> <001801c59a0a$f4be7480$9a00a8c0@OTTO> <41b0fe8905080516164464e04f@mail.gmail.com> <15425.1123304631@sss.pgh.pa.us> <41b0fe89050808100021a2a4fc@mail.gmail.com> <10930.1123549106@sss.pgh.pa.us> <1123589756.15206.4.camel@spectre.intellivid.com> <15881.1123598017@sss.pgh.pa.us> Content-Type: text/plain Organization: Intellivid Corp. Message-Id: <1123598613.15206.17.camel@spectre.intellivid.com> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Tue, 09 Aug 2005 10:43:33 -0400 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/78 X-Sequence-Number: 13825 Yes, sorry, two totally different machines. The 7.4.8 run was on a dual P4 3.2GHz, and the 7.4.2 run was on a dual hyperthreaded Xeon 2.4GHz. --Ian On Tue, 2005-08-09 at 10:33, Tom Lane wrote: > Ian Westmacott writes: > > On Mon, 2005-08-08 at 20:58, Tom Lane wrote: > >> I'd be interested to see results from other people using 7.4.* too. > > > 7.4.8: > > Total runtime: 0.198 ms > > > 7.4.2: > > Total runtime: 0.697 ms > > Just to be clear: those are two different machines of different speeds, > right? I don't believe we put any factor-of-three speedups into 7.4.* > after release ;-) > > regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 9 13:51:57 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 46BFF5291F for ; Tue, 9 Aug 2005 13:51:55 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 42188-07 for ; Tue, 9 Aug 2005 16:51:52 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.202]) by svr1.postgresql.org (Postfix) with ESMTP id C217A5288D for ; Tue, 9 Aug 2005 13:51:51 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id 36so1239423wra for ; Tue, 09 Aug 2005 09:51:51 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=LzdUZh81WS8v66j92VyOpvduWrFrPEQRslzeS0eBEktjBUyupTqqXFrEemd0HD0chD7pNJPWIgSmaQOGiyHLSsC2gf4zP1PJIe9nDCG6CAuEBTDKjh3PwVfU4hCjjX0vxbu8gHUkQFfWnWxwjCxEG0nqJ1A5smamKzp06EAqYqs= Received: by 10.54.14.29 with SMTP id 29mr687768wrn; Tue, 09 Aug 2005 09:51:51 -0700 (PDT) Received: by 10.54.119.7 with HTTP; Tue, 9 Aug 2005 09:51:51 -0700 (PDT) Message-ID: <41b0fe890508090951546156f7@mail.gmail.com> Date: Tue, 9 Aug 2005 09:51:51 -0700 From: Rhett Garber Reply-To: Rhett Garber To: Tom Lane Subject: Re: Why hash join instead of nested loop? Cc: =?ISO-8859-1?Q?Havasv=F6lgyi_Ott=F3?= , pgsql-performance@postgresql.org In-Reply-To: <10930.1123549106@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <41b0fe890508051135712a086b@mail.gmail.com> <001801c59a0a$f4be7480$9a00a8c0@OTTO> <41b0fe8905080516164464e04f@mail.gmail.com> <15425.1123304631@sss.pgh.pa.us> <41b0fe89050808100021a2a4fc@mail.gmail.com> <10930.1123549106@sss.pgh.pa.us> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 tagged_above=0 required=5 tests=AWL, RCVD_BY_IP X-Spam-Level: X-Archive-Number: 200508/79 X-Sequence-Number: 13826 Duplicated your setup in a separate DB. At least its reproducable for me..... I tested this on a Xeon 2 Ghz, 1 Gig Ram. Its running on some shared storage array that I'm not sure the details of. My production example also shows up on our production machine that is almost the same hardware but has dual zeon and 6 gigs of ram. Rhett Hash Join (cost=3D4.83..5.91 rows=3D1 width=3D14) (actual time=3D7.148..7.= 159 rows=3D1 loops=3D1) Hash Cond: ("outer".id =3D "inner".obj2) -> Seq Scan on rtmessagestate (cost=3D0.00..1.05 rows=3D5 width=3D14) (actual time=3D0.007..0.015 rows=3D5 loops=3D1) -> Hash (cost=3D4.83..4.83 rows=3D1 width=3D4) (actual time=3D0.055..0.055 rows=3D0 loops=3D1) -> Index Scan using connection_regid_obj1_index on connection (cost=3D0.00..4.83 rows=3D1 width=3D4) (actual time=3D0.028..0.= 032 rows=3D1 loops=3D1) Index Cond: ((connection_registry_id =3D 40105) AND (obj1 =3D 73582)) Total runtime: 7.693 ms (7 rows) On 8/8/05, Tom Lane wrote: > Rhett Garber writes: > > This is postgres 7.4.1 > > All the rows involved are integers. >=20 > Hmph. There is something really strange going on here. I tried to > duplicate your problem in 7.4.*, thus: >=20 > regression=3D# create table rtmessagestate(id int, f1 char(6)); > CREATE TABLE > regression=3D# insert into rtmessagestate values(1,'z'); > INSERT 559399 1 > regression=3D# insert into rtmessagestate values(2,'z'); > INSERT 559400 1 > regression=3D# insert into rtmessagestate values(3,'z'); > INSERT 559401 1 > regression=3D# insert into rtmessagestate values(4,'z'); > INSERT 559402 1 > regression=3D# insert into rtmessagestate values(5,'z'); > INSERT 559403 1 > regression=3D# vacuum analyze rtmessagestate; > VACUUM > regression=3D# create table connection(connection_registry_id int, obj1 i= nt, obj2 int); > CREATE TABLE > regression=3D# create index connection_regid_obj1_index on connection(con= nection_registry_id,obj1); > CREATE INDEX > regression=3D# insert into connection values(40105,73582,3); > INSERT 559407 1 > regression=3D# explain analyze select rtmessagestate.* from rtmessagestat= e,connection where (connection_registry_id =3D 40105) AND (obj1 =3D 73582= ) and id =3D obj2; > QUER= Y PLAN > -------------------------------------------------------------------------= --------------------------------------------------------------------------- > Hash Join (cost=3D4.83..5.91 rows=3D1 width=3D14) (actual time=3D0.498.= .0.544 rows=3D1 loops=3D1) > Hash Cond: ("outer".id =3D "inner".obj2) > -> Seq Scan on rtmessagestate (cost=3D0.00..1.05 rows=3D5 width=3D14= ) (actual time=3D0.030..0.072 rows=3D5 loops=3D1) > -> Hash (cost=3D4.83..4.83 rows=3D1 width=3D4) (actual time=3D0.305.= .0.305 rows=3D0 loops=3D1) > -> Index Scan using connection_regid_obj1_index on connection = (cost=3D0.00..4.83 rows=3D1 width=3D4) (actual time=3D0.236..0.264 rows=3D1= loops=3D1) > Index Cond: ((connection_registry_id =3D 40105) AND (obj1 = =3D 73582)) > Total runtime: 1.119 ms > (7 rows) >=20 > This duplicates your example as to plan and row counts: >=20 > > Hash Join (cost=3D5.96..7.04 rows=3D1 width=3D14) (actual > > time=3D10.591..10.609 rows=3D1 loops=3D1) > > Hash Cond: ("outer".id =3D "inner".obj2) > > -> Seq Scan on rtmessagestate (cost=3D0.00..1.05 rows=3D5 width=3D14) > > (actual time=3D0.011..0.022 rows=3D5 loops=3D1) > > -> Hash (cost=3D5.96..5.96 rows=3D1 width=3D4) (actual > > time=3D0.109..0.109 rows=3D0 loops=3D1) > > -> Index Scan using connection_regid_obj1_index on > > connection (cost=3D0.00..5.96 rows=3D1 width=3D4) (actual time=3D0.070= ..0.076 > > rows=3D1 loops=3D1) > > Index Cond: ((connection_registry_id =3D 40105) AND (obj1 > > =3D 73582)) Total runtime: 11.536 ms > > (7 rows) >=20 > My machine is considerably slower than yours, to judge by the actual > elapsed times in the scan nodes ... so why is it beating the pants > off yours in the join step? >=20 > Can you try the above script verbatim in a scratch database and see > what you get? (Note it's worth trying the explain two or three > times to be sure the values have settled out.) >=20 > I'm testing a fairly recent 7.4-branch build (7.4.8 plus), so that's one > possible reason for the discrepancy between my results and yours, but I > do not see anything in the 7.4 CVS logs that looks like it's related to > hashjoin performance. >=20 > I'd be interested to see results from other people using 7.4.* too. >=20 > regards, tom lane > From pgsql-performance-owner@postgresql.org Tue Aug 9 14:13:36 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D686E528BE for ; Tue, 9 Aug 2005 14:13:18 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 19226-09 for ; Tue, 9 Aug 2005 17:13:09 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 6CA9C52A28 for ; Tue, 9 Aug 2005 14:13:06 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j79HCx8M017254; Tue, 9 Aug 2005 13:13:00 -0400 (EDT) To: Rhett Garber Cc: =?ISO-8859-1?Q?Havasv=F6lgyi_Ott=F3?= , pgsql-performance@postgresql.org Subject: Re: Why hash join instead of nested loop? In-reply-to: <41b0fe890508090951546156f7@mail.gmail.com> References: <41b0fe890508051135712a086b@mail.gmail.com> <001801c59a0a$f4be7480$9a00a8c0@OTTO> <41b0fe8905080516164464e04f@mail.gmail.com> <15425.1123304631@sss.pgh.pa.us> <41b0fe89050808100021a2a4fc@mail.gmail.com> <10930.1123549106@sss.pgh.pa.us> <41b0fe890508090951546156f7@mail.gmail.com> Comments: In-reply-to Rhett Garber message dated "Tue, 09 Aug 2005 09:51:51 -0700" Date: Tue, 09 Aug 2005 13:12:59 -0400 Message-ID: <17253.1123607579@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/80 X-Sequence-Number: 13827 Rhett Garber writes: > Duplicated your setup in a separate DB. > At least its reproducable for me..... Hmm. Well, we now have several data points but they seem to be on wildly varying hardware. To try to normalize the results a little, I computed the total actual time for the hash plan divided by the sum of the actual times for the two scan nodes. Thus, for your example: > Hash Join (cost=4.83..5.91 rows=1 width=14) (actual time=7.148..7.159 > rows=1 loops=1) > Hash Cond: ("outer".id = "inner".obj2) > -> Seq Scan on rtmessagestate (cost=0.00..1.05 rows=5 width=14) > (actual time=0.007..0.015 rows=5 loops=1) > -> Hash (cost=4.83..4.83 rows=1 width=4) (actual > time=0.055..0.055 rows=0 loops=1) > -> Index Scan using connection_regid_obj1_index on > connection (cost=0.00..4.83 rows=1 width=4) (actual time=0.028..0.032 > rows=1 loops=1) > Index Cond: ((connection_registry_id = 40105) AND (obj1 > = 73582)) Total runtime: 7.693 ms > (7 rows) this would be 7.159 / (0.015 + 0.032). This is probably not an enormously robust statistic but it at least focuses attention in the right place. Here's what I get (rounded off to 4 digits which is surely as much precision as we have in the numbers): Tom 7.4.8+ 1.619 Ian 7.4.8 6.000 Ian 7.4.2 13.95 Steinar 7.4.7 8.833 Rhett orig 108.3 Rhett test 152.3 Michael 7.4.1 2.015 My number seems to be a bit of an outlier to the low side, but yours are way the heck to the high side. And Michael's test seems to rule out the idea that it's something broken in 7.4.1 in particular. I'm now thinking you've got either a platform- or compiler-specific problem. Exactly what is the hardware (the CPU not the disks)? How did you build or come by the Postgres executables (compiler, configure options, etc)? regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 9 15:02:03 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5B8DF529C0 for ; Tue, 9 Aug 2005 15:02:02 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 77055-05 for ; Tue, 9 Aug 2005 18:01:59 +0000 (GMT) Received: from crestone.coronasolutions.com (crestone.coronasolutions.com [66.45.104.24]) by svr1.postgresql.org (Postfix) with ESMTP id 28BC9529AD for ; Tue, 9 Aug 2005 15:01:53 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by crestone.coronasolutions.com (Postfix) with ESMTP id E0AD7644111 for ; Tue, 9 Aug 2005 12:01:22 -0600 (MDT) Received: from crestone.coronasolutions.com ([127.0.0.1]) by localhost (crestone.coronasolutions.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 01641-01 for ; Tue, 9 Aug 2005 12:01:20 -0600 (MDT) Received: from [192.168.1.12] (c-24-9-24-35.hsd1.co.comcast.net [24.9.24.35]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (No client certificate requested) by crestone.coronasolutions.com (Postfix) with ESMTP id 490A564410F for ; Tue, 9 Aug 2005 12:01:19 -0600 (MDT) Mime-Version: 1.0 (Apple Message framework v730) Content-Transfer-Encoding: 7bit Message-Id: <52CFA236-FC43-4018-B399-1854240110FF@drivefaster.net> Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed To: pgsql-performance@postgresql.org From: Dan Harris Subject: Table locking problems? Date: Tue, 9 Aug 2005 12:04:11 -0600 X-Mailer: Apple Mail (2.730) X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at drivefaster.net X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/81 X-Sequence-Number: 13828 I thought I would send this to pg-performance since so many people helped me with my speed issues recently. I was definitely IO- bottlenecked. Since then, I have installed 2 RAID arrays with 7 15k drives in them in RAID 0+1 as well as add a new controller card with 512MB of cache on it. I also created this new partition on the RAID as XFS instead of ext3. These changes have definitely improved performance, but I am now finding some trouble with UPDATE or DELETE queries "hanging" and never releasing their locks. As this happens, other statements queue up behind it. It seems to occur at times of very high loads on the box. Is my only option to kill the query ( which usually takes down the whole postmaster with it! ouch ). Could these locking issues be related to the other changes I made? I'm really scared that this is related to choosing XFS, but I sure hope not. How should I go about troubleshooting the "problem" queries? They don't seem to be specific to a single table or single database. I'm running 8.0.1 on kernel 2.6.12-3 on 64-bit Opterons if that matters.. -Dan From pgsql-performance-owner@postgresql.org Tue Aug 9 15:10:47 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 73AF1529A5 for ; Tue, 9 Aug 2005 15:10:45 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 87007-09 for ; Tue, 9 Aug 2005 18:10:36 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.204]) by svr1.postgresql.org (Postfix) with ESMTP id 52749528E1 for ; Tue, 9 Aug 2005 15:10:33 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i8so1269794wra for ; Tue, 09 Aug 2005 11:10:34 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=kkJWikFPbgJsk4oitvXDz3LOhmdQ1eSfhBQpUZDRFDst6oFUIFAz2caMk1NUTpe5PcHHbo1EfuQsCKt0p4/vH8UN+5bkWXAJIVzDWVEYehAvPeJJoNBGbNWkmU1FOigNNE0JY/XO+bgQqq7b6Wugq8PCNIgk+Vm8StLo3/nhF10= Received: by 10.54.57.44 with SMTP id f44mr5757960wra; Tue, 09 Aug 2005 11:10:34 -0700 (PDT) Received: by 10.54.119.7 with HTTP; Tue, 9 Aug 2005 11:10:33 -0700 (PDT) Message-ID: <41b0fe89050809111027e1d274@mail.gmail.com> Date: Tue, 9 Aug 2005 11:10:33 -0700 From: Rhett Garber To: Tom Lane Subject: Re: Why hash join instead of nested loop? Cc: =?ISO-8859-1?Q?Havasv=F6lgyi_Ott=F3?= , pgsql-performance@postgresql.org In-Reply-To: <17253.1123607579@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <41b0fe890508051135712a086b@mail.gmail.com> <001801c59a0a$f4be7480$9a00a8c0@OTTO> <41b0fe8905080516164464e04f@mail.gmail.com> <15425.1123304631@sss.pgh.pa.us> <41b0fe89050808100021a2a4fc@mail.gmail.com> <10930.1123549106@sss.pgh.pa.us> <41b0fe890508090951546156f7@mail.gmail.com> <17253.1123607579@sss.pgh.pa.us> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 tagged_above=0 required=5 tests=RCVD_BY_IP X-Spam-Level: X-Archive-Number: 200508/82 X-Sequence-Number: 13829 > I'm now thinking you've got either a platform- or compiler-specific > problem. Exactly what is the hardware (the CPU not the disks)? How did > you build or come by the Postgres executables (compiler, configure > options, etc)? I've tried it on two of our machines, both HP Proliant DL580: Production: Intel(R) Xeon(TM) MP CPU 2.80GHz (I think there are 2 physical CPUs with Hyperthreading, shows up as 4) 6 gigs RAM Development: Intel(R) XEON(TM) MP CPU 2.00GHz (I have vague recollection of disabling hyperthreading on this chip because of some other kernel issue) 1 gig RAM They are both running SuSE 8, 2.4.21-128-smp kernel Compile instructions (I didn't do it myself) indicate we built from source with nothing fancy: tar xpvf postgresql-7.4.1.tar.bz2 cd postgresql-7.4.1 ./configure --prefix=3D/usr/local/postgresql-7.4.1 make make install make install-all-headers If i run 'file' on /usr/local/postgresql-7.4.1/bin/postgres : postgres: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), dynamically linked (uses shared libs), not stripped Thanks for all your help guys, Rhett From pgsql-performance-owner@postgresql.org Tue Aug 9 15:31:46 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CA6C6529F1 for ; Tue, 9 Aug 2005 15:31:44 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 92321-03 for ; Tue, 9 Aug 2005 18:31:34 +0000 (GMT) Received: from hosting.commandprompt.com (128.commandprompt.com [207.173.200.128]) by svr1.postgresql.org (Postfix) with ESMTP id 8F038529D3 for ; Tue, 9 Aug 2005 15:31:34 -0300 (ADT) Received: from [192.168.1.55] (fc1smp [66.93.38.87]) (authenticated bits=0) by hosting.commandprompt.com (8.13.4/8.13.4) with ESMTP id j79IUVsG003359; Tue, 9 Aug 2005 11:30:31 -0700 Message-ID: <42F8F717.6030106@commandprompt.com> Date: Tue, 09 Aug 2005 11:33:59 -0700 From: "Joshua D. Drake" Organization: Command Prompt, Inc. User-Agent: Mozilla Thunderbird 1.0.6-1.1.fc4 (X11/20050720) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Dan Harris Cc: pgsql-performance@postgresql.org Subject: Re: Table locking problems? References: <52CFA236-FC43-4018-B399-1854240110FF@drivefaster.net> In-Reply-To: <52CFA236-FC43-4018-B399-1854240110FF@drivefaster.net> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Greylist: Sender succeded SMTP AUTH authentication, not delayed by milter-greylist-1.6 (hosting.commandprompt.com [192.168.1.101]); Tue, 09 Aug 2005 11:30:31 -0700 (PDT) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.013 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/83 X-Sequence-Number: 13830 > Could these locking issues be related to the other changes I made? I'm > really scared that this is related to choosing XFS, but I sure hope > not. How should I go about troubleshooting the "problem" queries? > They don't seem to be specific to a single table or single database. My experience is that when this type of thing happens it is typically specific queries that cause the problem. If you turn on statement logging you can get the exact queries and debug from there. Here are some things to look for: Is it a large table (and thus large indexes) that it is updating? Is the query using indexes? Is the query modifying ALOT of rows? Of course there is also the RTFM of are you analyzing and vacuuming? Sincerely, Joshua D. Drake > > I'm running 8.0.1 on kernel 2.6.12-3 on 64-bit Opterons if that matters.. > > > -Dan > > ---------------------------(end of broadcast)--------------------------- > TIP 9: In versions below 8.0, the planner will ignore your desire to > choose an index scan if your joining column's datatypes do not > match -- Your PostgreSQL solutions company - Command Prompt, Inc. 1.800.492.2240 PostgreSQL Replication, Consulting, Custom Programming, 24x7 support Managed Services, Shared and Dedicated Hosting Co-Authors: plPHP, plPerlNG - http://www.commandprompt.com/ From pgsql-performance-owner@postgresql.org Tue Aug 9 15:38:55 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CDC07529D2 for ; Tue, 9 Aug 2005 15:37:32 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 54683-10 for ; Tue, 9 Aug 2005 18:37:30 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 7D981529D3 for ; Tue, 9 Aug 2005 15:37:29 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j79IbIfY017826; Tue, 9 Aug 2005 14:37:18 -0400 (EDT) To: Rhett Garber Cc: =?ISO-8859-1?Q?Havasv=F6lgyi_Ott=F3?= , pgsql-performance@postgresql.org Subject: Re: Why hash join instead of nested loop? In-reply-to: <41b0fe89050809111027e1d274@mail.gmail.com> References: <41b0fe890508051135712a086b@mail.gmail.com> <001801c59a0a$f4be7480$9a00a8c0@OTTO> <41b0fe8905080516164464e04f@mail.gmail.com> <15425.1123304631@sss.pgh.pa.us> <41b0fe89050808100021a2a4fc@mail.gmail.com> <10930.1123549106@sss.pgh.pa.us> <41b0fe890508090951546156f7@mail.gmail.com> <17253.1123607579@sss.pgh.pa.us> <41b0fe89050809111027e1d274@mail.gmail.com> Comments: In-reply-to Rhett Garber message dated "Tue, 09 Aug 2005 11:10:33 -0700" Date: Tue, 09 Aug 2005 14:37:18 -0400 Message-ID: <17825.1123612638@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/84 X-Sequence-Number: 13831 Rhett Garber writes: > They are both running SuSE 8, 2.4.21-128-smp kernel > Compile instructions (I didn't do it myself) indicate we built from > source with nothing fancy: You could double-check the configure options by running pg_config. But probably the more interesting question is whether any nondefault CFLAGS were used, and I don't think pg_config records that. (Hmm, maybe it should.) In any case, there's no smoking gun there. I'm now wondering if maybe there's something unusual about your runtime parameters. AFAIR you didn't show us your postgresql.conf settings --- could we see any nondefault entries there? (I looked quickly at the 7.4 hashjoin code, and I see that it uses a hash table sized according to sort_mem even when the input is predicted to be very small ... so an enormous sort_mem setting would account for some plan startup overhead to initialize the table ...) regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 9 15:52:11 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0072452A1D for ; Tue, 9 Aug 2005 15:51:37 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 41563-05 for ; Tue, 9 Aug 2005 18:51:30 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.192]) by svr1.postgresql.org (Postfix) with ESMTP id 555BB52A00 for ; Tue, 9 Aug 2005 15:51:29 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id 36so1268216wra for ; Tue, 09 Aug 2005 11:51:30 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=LVKv668pQGd+/jUq04gr5fEGOXlx0XbtQMIDOYY5tC7UswOxcAm9jbcTHKLYN1vluLDDVEKXuRMtAPYXl9dfIgje2IlLCCviVSL9Xs/ueIl3GjnS87HPHMUJjvV/gDK/YUbiCQFA4shLIDQbypWSiH62r/2ifixjnFYuUS/6Lyg= Received: by 10.54.56.24 with SMTP id e24mr2871416wra; Tue, 09 Aug 2005 11:51:30 -0700 (PDT) Received: by 10.54.119.7 with HTTP; Tue, 9 Aug 2005 11:51:30 -0700 (PDT) Message-ID: <41b0fe89050809115174ae1669@mail.gmail.com> Date: Tue, 9 Aug 2005 11:51:30 -0700 From: Rhett Garber To: Tom Lane Subject: Re: Why hash join instead of nested loop? Cc: =?ISO-8859-1?Q?Havasv=F6lgyi_Ott=F3?= , pgsql-performance@postgresql.org In-Reply-To: <17825.1123612638@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <41b0fe890508051135712a086b@mail.gmail.com> <001801c59a0a$f4be7480$9a00a8c0@OTTO> <41b0fe8905080516164464e04f@mail.gmail.com> <15425.1123304631@sss.pgh.pa.us> <41b0fe89050808100021a2a4fc@mail.gmail.com> <10930.1123549106@sss.pgh.pa.us> <41b0fe890508090951546156f7@mail.gmail.com> <17253.1123607579@sss.pgh.pa.us> <41b0fe89050809111027e1d274@mail.gmail.com> <17825.1123612638@sss.pgh.pa.us> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 tagged_above=0 required=5 tests=RCVD_BY_IP X-Spam-Level: X-Archive-Number: 200508/85 X-Sequence-Number: 13832 Well that could be an issue, is this abnormally large: #shared_buffers =3D 1536 # min 16, at least max_connections*2, 8KB= each shared_buffers =3D 206440 #sort_mem =3D 131072 # min 64, size in KB sort_mem =3D 524288 # min 64, size in KB vacuum_mem =3D 131072 # min 1024, size in K I actually had a lot of trouble finding example values for these... no one wants to give real numbers in any postgres performance tuning articles I saw. What would be appropriate for machines with 1 or 6 gigs of RAM and wanting to maximize performance. Rhett On 8/9/05, Tom Lane wrote: > Rhett Garber writes: > > They are both running SuSE 8, 2.4.21-128-smp kernel >=20 > > Compile instructions (I didn't do it myself) indicate we built from > > source with nothing fancy: >=20 > You could double-check the configure options by running pg_config. > But probably the more interesting question is whether any nondefault > CFLAGS were used, and I don't think pg_config records that. > (Hmm, maybe it should.) >=20 > In any case, there's no smoking gun there. I'm now wondering if maybe > there's something unusual about your runtime parameters. AFAIR you > didn't show us your postgresql.conf settings --- could we see any > nondefault entries there? >=20 > (I looked quickly at the 7.4 hashjoin code, and I see that it uses a > hash table sized according to sort_mem even when the input is predicted > to be very small ... so an enormous sort_mem setting would account for > some plan startup overhead to initialize the table ...) >=20 > regards, tom lane > From pgsql-performance-owner@postgresql.org Tue Aug 9 15:54:40 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6FE26529F1 for ; Tue, 9 Aug 2005 15:52:15 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 12400-04 for ; Tue, 9 Aug 2005 18:52:12 +0000 (GMT) Received: from tigger.fuhr.org (tigger.fuhr.org [63.214.45.158]) by svr1.postgresql.org (Postfix) with ESMTP id CE5DB52A1D for ; Tue, 9 Aug 2005 15:52:08 -0300 (ADT) Received: from winnie.fuhr.org (winnie.fuhr.org [10.1.0.1]) by tigger.fuhr.org (8.13.3/8.13.3) with ESMTP id j79Iq1sJ045377 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK); Tue, 9 Aug 2005 12:52:03 -0600 (MDT) (envelope-from mfuhr@winnie.fuhr.org) Received: from winnie.fuhr.org (localhost [127.0.0.1]) by winnie.fuhr.org (8.13.3/8.13.3) with ESMTP id j79Iq1iZ025404; Tue, 9 Aug 2005 12:52:01 -0600 (MDT) (envelope-from mfuhr@winnie.fuhr.org) Received: (from mfuhr@localhost) by winnie.fuhr.org (8.13.3/8.13.3/Submit) id j79Iq1cN025403; Tue, 9 Aug 2005 12:52:01 -0600 (MDT) (envelope-from mfuhr) Date: Tue, 9 Aug 2005 12:52:01 -0600 From: Michael Fuhr To: Dan Harris Cc: pgsql-performance@postgresql.org Subject: Re: Table locking problems? Message-ID: <20050809185201.GA25325@winnie.fuhr.org> References: <52CFA236-FC43-4018-B399-1854240110FF@drivefaster.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <52CFA236-FC43-4018-B399-1854240110FF@drivefaster.net> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/86 X-Sequence-Number: 13833 On Tue, Aug 09, 2005 at 12:04:11PM -0600, Dan Harris wrote: > These changes have definitely improved performance, but I am now > finding some trouble with UPDATE or DELETE queries "hanging" and > never releasing their locks. As this happens, other statements queue > up behind it. Have you examined pg_locks to see if the UPDATE or DELETE is blocked because of a lock another session holds? Are you using foreign keys? When updating referencing rows, released versions of PostgreSQL acquire a lock on the referenced row that can hurt concurrency or cause deadlock (this will be improved in 8.1). -- Michael Fuhr From pgsql-performance-owner@postgresql.org Tue Aug 9 16:02:09 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 844AC5288C for ; Tue, 9 Aug 2005 16:01:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 73615-02 for ; Tue, 9 Aug 2005 19:01:12 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 81F6C5289C for ; Tue, 9 Aug 2005 16:01:10 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j79J0xm1018144; Tue, 9 Aug 2005 15:00:59 -0400 (EDT) To: Rhett Garber Cc: =?ISO-8859-1?Q?Havasv=F6lgyi_Ott=F3?= , pgsql-performance@postgresql.org Subject: Re: Why hash join instead of nested loop? In-reply-to: <41b0fe89050809115174ae1669@mail.gmail.com> References: <41b0fe890508051135712a086b@mail.gmail.com> <001801c59a0a$f4be7480$9a00a8c0@OTTO> <41b0fe8905080516164464e04f@mail.gmail.com> <15425.1123304631@sss.pgh.pa.us> <41b0fe89050808100021a2a4fc@mail.gmail.com> <10930.1123549106@sss.pgh.pa.us> <41b0fe890508090951546156f7@mail.gmail.com> <17253.1123607579@sss.pgh.pa.us> <41b0fe89050809111027e1d274@mail.gmail.com> <17825.1123612638@sss.pgh.pa.us> <41b0fe89050809115174ae1669@mail.gmail.com> Comments: In-reply-to Rhett Garber message dated "Tue, 09 Aug 2005 11:51:30 -0700" Date: Tue, 09 Aug 2005 15:00:59 -0400 Message-ID: <18143.1123614059@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/87 X-Sequence-Number: 13834 Rhett Garber writes: > Well that could be an issue, is this abnormally large: > #shared_buffers = 1536 # min 16, at least max_connections*2, 8KB each > shared_buffers = 206440 > #sort_mem = 131072 # min 64, size in KB > sort_mem = 524288 # min 64, size in KB > vacuum_mem = 131072 # min 1024, size in K The vacuum_mem number is OK I think, but both of the others seem unreasonably large. Conventional wisdom about shared_buffers is that the sweet spot is maybe 10000 or so buffers, rarely more than 50000. (Particularly in pre-8.0 releases, there are code paths that grovel through all the buffers linearly, so there is a significant cost to making it too large.) Don't worry about it being too small to make effective use of RAM --- we rely on the kernel's disk cache to do that. sort_mem is *per sort*, and so half a gig in a machine with only a couple of gig is far too much except when you know you have only one query running. A couple dozen backends each trying to use half a gig will drive you into the ground in no time. Conventional wisdom here is that the global setting should be conservatively small (perhaps 10Mb to 100Mb depending on how many concurrent backends you expect to have), and then you can explicitly increase it locally with SET for specific queries that need it. In terms of the problem at hand, try the test case with a few different values of sort_mem (use SET to adjust it, you don't need to touch the config file) and see what happens. I think the cost you're seeing is just startup overhead to zero a hash table of a few hundred meg ... regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 9 16:38:14 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3DC095289C for ; Tue, 9 Aug 2005 16:08:22 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 90449-02 for ; Tue, 9 Aug 2005 19:08:10 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 4A84252808 for ; Tue, 9 Aug 2005 16:08:07 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j79J867x018200; Tue, 9 Aug 2005 15:08:06 -0400 (EDT) To: "Joshua D. Drake" Cc: Dan Harris , pgsql-performance@postgresql.org Subject: Re: Table locking problems? In-reply-to: <42F8F717.6030106@commandprompt.com> References: <52CFA236-FC43-4018-B399-1854240110FF@drivefaster.net> <42F8F717.6030106@commandprompt.com> Comments: In-reply-to "Joshua D. Drake" message dated "Tue, 09 Aug 2005 11:33:59 -0700" Date: Tue, 09 Aug 2005 15:08:06 -0400 Message-ID: <18199.1123614486@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/90 X-Sequence-Number: 13837 "Joshua D. Drake" writes: > My experience is that when this type of thing happens it is typically > specific queries that cause the problem. If you turn on statement > logging you can get the exact queries and debug from there. > Here are some things to look for: > Is it a large table (and thus large indexes) that it is updating? > Is the query using indexes? > Is the query modifying ALOT of rows? Another thing to look at is foreign keys. Dan could be running into problems with an update on one side of an FK being blocked by locks on the associated rows on the other side. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 9 16:13:12 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 287B152AB6 for ; Tue, 9 Aug 2005 16:12:00 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 74717-05 for ; Tue, 9 Aug 2005 19:11:52 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.194]) by svr1.postgresql.org (Postfix) with ESMTP id 2F9A052A7E for ; Tue, 9 Aug 2005 16:11:51 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i8so1284116wra for ; Tue, 09 Aug 2005 12:11:52 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=Bkc/M3YoCnjghDEkWifAhSya5hVGOHITTWmasJCoHlaNHUmlsLXSdcmLMb+mYCLcA5SqiTMdDZ6Prt3Y96487+sID5S5zn0N4oLjuWoi1RxYOjiFX1hbOYDy2GdOpNXw+4/aR/8PhnLsoE41g3pxkh2YoxLD4BZ+BXizP29goiE= Received: by 10.54.53.62 with SMTP id b62mr5765154wra; Tue, 09 Aug 2005 12:11:52 -0700 (PDT) Received: by 10.54.119.7 with HTTP; Tue, 9 Aug 2005 12:11:52 -0700 (PDT) Message-ID: <41b0fe8905080912113f381f84@mail.gmail.com> Date: Tue, 9 Aug 2005 12:11:52 -0700 From: Rhett Garber To: Tom Lane Subject: Re: Why hash join instead of nested loop? Cc: =?ISO-8859-1?Q?Havasv=F6lgyi_Ott=F3?= , pgsql-performance@postgresql.org In-Reply-To: <18143.1123614059@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <41b0fe890508051135712a086b@mail.gmail.com> <15425.1123304631@sss.pgh.pa.us> <41b0fe89050808100021a2a4fc@mail.gmail.com> <10930.1123549106@sss.pgh.pa.us> <41b0fe890508090951546156f7@mail.gmail.com> <17253.1123607579@sss.pgh.pa.us> <41b0fe89050809111027e1d274@mail.gmail.com> <17825.1123612638@sss.pgh.pa.us> <41b0fe89050809115174ae1669@mail.gmail.com> <18143.1123614059@sss.pgh.pa.us> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 tagged_above=0 required=5 tests=AWL, RCVD_BY_IP X-Spam-Level: X-Archive-Number: 200508/88 X-Sequence-Number: 13835 Bingo, the smaller the sort_mem, the faster that query is. Thanks a lot to everybody that helped, i'll tweak with these values more when I get a chance now that I have some guidelines that make sense. Rhett On 8/9/05, Tom Lane wrote: > Rhett Garber writes: > > Well that could be an issue, is this abnormally large: > > #shared_buffers =3D 1536 # min 16, at least max_connections*2,= 8KB each > > shared_buffers =3D 206440 > > #sort_mem =3D 131072 # min 64, size in KB > > sort_mem =3D 524288 # min 64, size in KB > > vacuum_mem =3D 131072 # min 1024, size in K >=20 > The vacuum_mem number is OK I think, but both of the others seem > unreasonably large. Conventional wisdom about shared_buffers is that > the sweet spot is maybe 10000 or so buffers, rarely more than 50000. > (Particularly in pre-8.0 releases, there are code paths that grovel > through all the buffers linearly, so there is a significant cost to > making it too large.) Don't worry about it being too small to make > effective use of RAM --- we rely on the kernel's disk cache to do that. >=20 > sort_mem is *per sort*, and so half a gig in a machine with only a > couple of gig is far too much except when you know you have only one > query running. A couple dozen backends each trying to use half a gig > will drive you into the ground in no time. Conventional wisdom here > is that the global setting should be conservatively small (perhaps > 10Mb to 100Mb depending on how many concurrent backends you expect to > have), and then you can explicitly increase it locally with SET for > specific queries that need it. >=20 > In terms of the problem at hand, try the test case with a few different > values of sort_mem (use SET to adjust it, you don't need to touch the > config file) and see what happens. I think the cost you're seeing is > just startup overhead to zero a hash table of a few hundred meg ... >=20 > regards, tom lane > From pgsql-performance-owner@postgresql.org Tue Aug 9 17:43:27 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AE53252C68 for ; Tue, 9 Aug 2005 17:40:43 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 23695-05 for ; Tue, 9 Aug 2005 20:40:34 +0000 (GMT) Received: from crestone.coronasolutions.com (crestone.coronasolutions.com [66.45.104.24]) by svr1.postgresql.org (Postfix) with ESMTP id 153FD52C66 for ; Tue, 9 Aug 2005 17:40:33 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by crestone.coronasolutions.com (Postfix) with ESMTP id 8A5576440E0 for ; Tue, 9 Aug 2005 14:40:04 -0600 (MDT) Received: from crestone.coronasolutions.com ([127.0.0.1]) by localhost (crestone.coronasolutions.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 02670-08 for ; Tue, 9 Aug 2005 14:40:02 -0600 (MDT) Received: from [192.168.1.12] (c-24-9-24-35.hsd1.co.comcast.net [24.9.24.35]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (No client certificate requested) by crestone.coronasolutions.com (Postfix) with ESMTP id 23D5F6440C5 for ; Tue, 9 Aug 2005 14:40:02 -0600 (MDT) Mime-Version: 1.0 (Apple Message framework v730) In-Reply-To: <18199.1123614486@sss.pgh.pa.us> References: <52CFA236-FC43-4018-B399-1854240110FF@drivefaster.net> <42F8F717.6030106@commandprompt.com> <18199.1123614486@sss.pgh.pa.us> Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: <36F56B3E-F6FC-491D-A8E2-B4961CC8AC16@drivefaster.net> Content-Transfer-Encoding: 7bit From: Dan Harris Subject: Re: Table locking problems? Date: Tue, 9 Aug 2005 14:42:57 -0600 To: pgsql-performance@postgresql.org X-Mailer: Apple Mail (2.730) X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at drivefaster.net X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/91 X-Sequence-Number: 13838 On Aug 9, 2005, at 1:08 PM, Tom Lane wrote: > "Joshua D. Drake" writes: > >> My experience is that when this type of thing happens it is typically >> specific queries that cause the problem. If you turn on statement >> logging you can get the exact queries and debug from there. >> > > >> Here are some things to look for: >> > > >> Is it a large table (and thus large indexes) that it is updating? >> Is the query using indexes? >> Is the query modifying ALOT of rows? >> > > Another thing to look at is foreign keys. Dan could be running into > problems with an update on one side of an FK being blocked by locks > on the associated rows on the other side. > > regards, tom lane > Tom, Steve, Josh: Thank you for your ideas. The updates are only on a single table, no joins. I had stats collection turned off. I have turned that on again so that I can try and catch one while the problem is occurring. The last table it did this on was about 3 million records. 4 single-column indexes on it. The problem I had with statement logging is that if the query never finishes, it doesn't get logged as far as I can tell. So everything that did get logged was normal and would run with no isses in psql by copy and pasting it. The rows updated will certainly vary by query. I really need to "catch it in the act" with stats collection on so I can get the query from pg_stat_activity. Once I get it, I will play with explains and see if I can reproduce it outside the wild. Thanks again for your help. -Dan From pgsql-performance-owner@postgresql.org Tue Aug 9 18:33:27 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8B6BB52B05 for ; Tue, 9 Aug 2005 17:49:00 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 40641-03 for ; Tue, 9 Aug 2005 20:48:54 +0000 (GMT) Received: from crestone.coronasolutions.com (crestone.coronasolutions.com [66.45.104.24]) by svr1.postgresql.org (Postfix) with ESMTP id 04EE152A91 for ; Tue, 9 Aug 2005 17:48:53 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by crestone.coronasolutions.com (Postfix) with ESMTP id C3AF06440E0 for ; Tue, 9 Aug 2005 14:48:25 -0600 (MDT) Received: from crestone.coronasolutions.com ([127.0.0.1]) by localhost (crestone.coronasolutions.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 02853-01 for ; Tue, 9 Aug 2005 14:48:23 -0600 (MDT) Received: from [192.168.1.12] (c-24-9-24-35.hsd1.co.comcast.net [24.9.24.35]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (No client certificate requested) by crestone.coronasolutions.com (Postfix) with ESMTP id 77D616440C5 for ; Tue, 9 Aug 2005 14:48:23 -0600 (MDT) Mime-Version: 1.0 (Apple Message framework v730) In-Reply-To: <1123656547.13362.43.camel@amd64-laptop-spoe> References: <52CFA236-FC43-4018-B399-1854240110FF@drivefaster.net> <1123656547.13362.43.camel@amd64-laptop-spoe> Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: <2764B744-4CDC-4F2C-99C4-2948D37A9D23@drivefaster.net> Content-Transfer-Encoding: 7bit From: Dan Harris Subject: Re: Table locking problems? Date: Tue, 9 Aug 2005 14:51:18 -0600 To: pgsql-performance@postgresql.org X-Mailer: Apple Mail (2.730) X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at drivefaster.net X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/92 X-Sequence-Number: 13839 On Aug 10, 2005, at 12:49 AM, Steve Poe wrote: > Dan, > > Do you mean you did RAID 1 + 0 (RAID 10) or RAID 0 + 1? Just a > clarification, since RAID 0 is still a single-point of failure even if > RAID1 is on top of RAID0. Well, you tell me if I stated incorrectly. There are two raid enclosures with 7 drives in each. Each is on its own bus on a dual- channel controller. Each box has a stripe across its drives and the enclosures are mirrors of each other. I understand the controller could be a single point of failure, but I'm not sure I understand your concern about the RAID structure itself. > > How many users are connected when your update / delete queries are > hanging? Have you done an analyze verbose on those queries? Most of the traffic is from programs we run to do analysis of the data and managing changes. At the time I noticed it this morning, there were 10 connections open to the database. That rarely goes above 20 concurrent. As I said in my other response, I believe that the log will only contain the query at the point the query finishes, so if it never finishes... > > Have you made changes to the postgresql.conf? kernel.vm settings? IO > scheduler? I set shmmax appropriately for my shared_buffers setting, but that's the only kernel tweak. > > If you're not doing so already, you may consider running sar > (iostat) to > monitor when the hanging occurs if their is a memory / IO bottleneck > somewhere. > I will try that. Thanks From pgsql-performance-owner@postgresql.org Tue Aug 9 18:52:32 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7D6BB5298E for ; Tue, 9 Aug 2005 18:51:41 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 00804-02 for ; Tue, 9 Aug 2005 21:51:39 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id 4F18552915 for ; Tue, 9 Aug 2005 18:51:33 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050809215131m92009juque>; Tue, 9 Aug 2005 21:51:31 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id C963955FF3; Tue, 9 Aug 2005 16:51:30 -0500 (CDT) Received: from [192.168.1.11] (liliana.arbash-meinel.com [192.168.1.11]) by juju.arbash-meinel.com (Postfix) with ESMTP id 33E1355FCC; Tue, 9 Aug 2005 16:51:24 -0500 (CDT) Message-ID: <42F9255A.4080604@arbash-meinel.com> Date: Tue, 09 Aug 2005 16:51:22 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.2 (Windows/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Dan Harris Cc: pgsql-performance@postgresql.org Subject: Re: Table locking problems? References: <52CFA236-FC43-4018-B399-1854240110FF@drivefaster.net> <1123656547.13362.43.camel@amd64-laptop-spoe> <2764B744-4CDC-4F2C-99C4-2948D37A9D23@drivefaster.net> In-Reply-To: <2764B744-4CDC-4F2C-99C4-2948D37A9D23@drivefaster.net> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig357C927C6DFAE6598CBC821E" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.044 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/93 X-Sequence-Number: 13840 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig357C927C6DFAE6598CBC821E Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Dan Harris wrote: > > On Aug 10, 2005, at 12:49 AM, Steve Poe wrote: > >> Dan, >> >> Do you mean you did RAID 1 + 0 (RAID 10) or RAID 0 + 1? Just a >> clarification, since RAID 0 is still a single-point of failure even if >> RAID1 is on top of RAID0. > > > Well, you tell me if I stated incorrectly. There are two raid > enclosures with 7 drives in each. Each is on its own bus on a dual- > channel controller. Each box has a stripe across its drives and the > enclosures are mirrors of each other. I understand the controller > could be a single point of failure, but I'm not sure I understand your > concern about the RAID structure itself. In this configuration, if you have a drive fail on both controllers, the entire RAID dies. Lets label them A1-7, B1-7, because you stripe within a set, if a single one of A dies, and a single one of B dies, you have lost your entire mirror. The correct way of doing it, is to have A1 be a mirror of B1, and then stripe above that. Since you are using 2 7-disk enclosures, I'm not sure how you can do it well, since it is not an even number of disks. Though if you are using software RAID, there should be no problem. The difference is that in this scenario, *all* of the A drives can die, and you haven't lost any data. The only thing you can't lose is a matched pair (eg losing both A1 and B1 will cause complete data loss) I believe the correct notation for this last form is RAID 1 + 0 (RAID10) since you have a set of RAID1 drives, with a RAID0 on-top of them. > >> >> How many users are connected when your update / delete queries are >> hanging? Have you done an analyze verbose on those queries? > > > Most of the traffic is from programs we run to do analysis of the data > and managing changes. At the time I noticed it this morning, there > were 10 connections open to the database. That rarely goes above 20 > concurrent. As I said in my other response, I believe that the log > will only contain the query at the point the query finishes, so if it > never finishes... > >> >> Have you made changes to the postgresql.conf? kernel.vm settings? IO >> scheduler? > > > I set shmmax appropriately for my shared_buffers setting, but that's > the only kernel tweak. > >> >> If you're not doing so already, you may consider running sar (iostat) to >> monitor when the hanging occurs if their is a memory / IO bottleneck >> somewhere. >> > > I will try that. Thanks > When you discover that an update is hanging, can you get into the database, and see what locks currently exist? (SELECT * FROM pg_locks) That might help you figure out what is being locked and possibly preventing your updates. It is also possible that your UPDATE query is trying to do something funny (someone just recently was talking about an UPDATE that wanted to do a hash join against 12M rows). Which probably meant that it had to spill to disk, where a merge join would have worked better. John =:-> --------------enig357C927C6DFAE6598CBC821E Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC+SVdJdeBCYSNAAMRAnwXAJ9aAP2hg377LNF0ib8RKKrB1j4anQCghKDG Pl7DgEKhRiMBh//x4e0gJCA= =/2aC -----END PGP SIGNATURE----- --------------enig357C927C6DFAE6598CBC821E-- From pgsql-performance-owner@postgresql.org Tue Aug 9 19:03:33 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C0B9D5298F for ; Tue, 9 Aug 2005 19:03:13 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 50121-02 for ; Tue, 9 Aug 2005 22:03:12 +0000 (GMT) Received: from crestone.coronasolutions.com (crestone.coronasolutions.com [66.45.104.24]) by svr1.postgresql.org (Postfix) with ESMTP id 1F96252988 for ; Tue, 9 Aug 2005 19:03:08 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by crestone.coronasolutions.com (Postfix) with ESMTP id 6EFBE6440DC for ; Tue, 9 Aug 2005 16:02:41 -0600 (MDT) Received: from crestone.coronasolutions.com ([127.0.0.1]) by localhost (crestone.coronasolutions.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 03239-06 for ; Tue, 9 Aug 2005 16:02:39 -0600 (MDT) Received: from [192.168.1.12] (c-24-9-24-35.hsd1.co.comcast.net [24.9.24.35]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (No client certificate requested) by crestone.coronasolutions.com (Postfix) with ESMTP id 391856440BC for ; Tue, 9 Aug 2005 16:02:39 -0600 (MDT) Mime-Version: 1.0 (Apple Message framework v730) In-Reply-To: <42F9255A.4080604@arbash-meinel.com> References: <52CFA236-FC43-4018-B399-1854240110FF@drivefaster.net> <1123656547.13362.43.camel@amd64-laptop-spoe> <2764B744-4CDC-4F2C-99C4-2948D37A9D23@drivefaster.net> <42F9255A.4080604@arbash-meinel.com> Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: <17797211-4C5F-4410-B38E-53ECC10D0752@drivefaster.net> Content-Transfer-Encoding: 7bit From: Dan Harris Subject: Re: Table locking problems? Date: Tue, 9 Aug 2005 16:05:34 -0600 To: pgsql-performance@postgresql.org X-Mailer: Apple Mail (2.730) X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at drivefaster.net X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/94 X-Sequence-Number: 13841 On Aug 9, 2005, at 3:51 PM, John A Meinel wrote: > Dan Harris wrote: > >> On Aug 10, 2005, at 12:49 AM, Steve Poe wrote: >> >>> Dan, >>> >>> Do you mean you did RAID 1 + 0 (RAID 10) or RAID 0 + 1? Just a >>> clarification, since RAID 0 is still a single-point of failure >>> even if >>> RAID1 is on top of RAID0. >>> >> Well, you tell me if I stated incorrectly. There are two raid >> enclosures with 7 drives in each. Each is on its own bus on a >> dual- channel controller. Each box has a stripe across its drives >> and the enclosures are mirrors of each other. I understand the >> controller could be a single point of failure, but I'm not sure I >> understand your concern about the RAID structure itself. >> > > In this configuration, if you have a drive fail on both > controllers, the entire RAID dies. Lets label them A1-7, B1-7, > because you stripe within a set, if a single one of A dies, and a > single one of B dies, you have lost your entire mirror. > > The correct way of doing it, is to have A1 be a mirror of B1, and > then stripe above that. Since you are using 2 7-disk enclosures, > I'm not sure how you can do it well, since it is not an even number > of disks. Though if you are using software RAID, there should be no > problem. > > The difference is that in this scenario, *all* of the A drives can > die, and you haven't lost any data. The only thing you can't lose > is a matched pair (eg losing both A1 and B1 will cause complete > data loss) > > I believe the correct notation for this last form is RAID 1 + 0 > (RAID10) since you have a set of RAID1 drives, with a RAID0 on-top > of them. > I have read up on the difference now. I don't understand why it's a "single point of failure". Technically any array could be a "single point" depending on your level of abstraction. In retrospect, I probably should have gone 8 drives in each and used RAID 10 instead for the better fault-tolerance, but it's online now and will require some planning to see if I want to reconfigure that in the future. I wish HP's engineer would have promoted that method instead of 0+1.. -Dan From pgsql-performance-owner@postgresql.org Tue Aug 9 19:30:45 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 576285298E for ; Tue, 9 Aug 2005 19:30:44 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 11724-05 for ; Tue, 9 Aug 2005 22:30:40 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id 5E1B45291E for ; Tue, 9 Aug 2005 19:30:36 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050809223024m9100g5t0oe>; Tue, 9 Aug 2005 22:30:34 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 6EDC355FF3; Tue, 9 Aug 2005 17:30:21 -0500 (CDT) Received: from [192.168.1.11] (liliana.arbash-meinel.com [192.168.1.11]) by juju.arbash-meinel.com (Postfix) with ESMTP id CC65955FCC; Tue, 9 Aug 2005 17:30:15 -0500 (CDT) Message-ID: <42F92E79.5050905@arbash-meinel.com> Date: Tue, 09 Aug 2005 17:30:17 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.2 (Windows/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Dan Harris Cc: pgsql-performance@postgresql.org Subject: Re: Table locking problems? References: <52CFA236-FC43-4018-B399-1854240110FF@drivefaster.net> <1123656547.13362.43.camel@amd64-laptop-spoe> <2764B744-4CDC-4F2C-99C4-2948D37A9D23@drivefaster.net> <42F9255A.4080604@arbash-meinel.com> <17797211-4C5F-4410-B38E-53ECC10D0752@drivefaster.net> In-Reply-To: <17797211-4C5F-4410-B38E-53ECC10D0752@drivefaster.net> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig54CE91A96C702C95B2C308BF" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.044 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/95 X-Sequence-Number: 13842 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig54CE91A96C702C95B2C308BF Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Dan Harris wrote: > > On Aug 9, 2005, at 3:51 PM, John A Meinel wrote: > >> Dan Harris wrote: >> >>> On Aug 10, 2005, at 12:49 AM, Steve Poe wrote: >>> >>>> Dan, >>>> >>>> Do you mean you did RAID 1 + 0 (RAID 10) or RAID 0 + 1? Just a >>>> clarification, since RAID 0 is still a single-point of failure even if >>>> RAID1 is on top of RAID0. >>>> >>> Well, you tell me if I stated incorrectly. There are two raid >>> enclosures with 7 drives in each. Each is on its own bus on a dual- >>> channel controller. Each box has a stripe across its drives and >>> the enclosures are mirrors of each other. I understand the >>> controller could be a single point of failure, but I'm not sure I >>> understand your concern about the RAID structure itself. >>> >> >> In this configuration, if you have a drive fail on both controllers, >> the entire RAID dies. Lets label them A1-7, B1-7, because you stripe >> within a set, if a single one of A dies, and a single one of B dies, >> you have lost your entire mirror. >> >> The correct way of doing it, is to have A1 be a mirror of B1, and >> then stripe above that. Since you are using 2 7-disk enclosures, I'm >> not sure how you can do it well, since it is not an even number of >> disks. Though if you are using software RAID, there should be no >> problem. >> >> The difference is that in this scenario, *all* of the A drives can >> die, and you haven't lost any data. The only thing you can't lose is >> a matched pair (eg losing both A1 and B1 will cause complete data loss) >> >> I believe the correct notation for this last form is RAID 1 + 0 >> (RAID10) since you have a set of RAID1 drives, with a RAID0 on-top of >> them. >> > > I have read up on the difference now. I don't understand why it's a > "single point of failure". Technically any array could be a "single > point" depending on your level of abstraction. In retrospect, I > probably should have gone 8 drives in each and used RAID 10 instead for > the better fault-tolerance, but it's online now and will require some > planning to see if I want to reconfigure that in the future. I wish > HP's engineer would have promoted that method instead of 0+1.. I wouldn't say that it is a single point of failure, but I *can* say that it is much more likely to fail. (2 drives rather than on average n drives) If your devices will hold 8 drives, you could simply do 1 8-drive, and one 6-drive. And then do RAID1 with pairs, and RAID0 across the resultant 7 RAID1 sets. I'm really surprised that someone promoted RAID 0+1 over RAID10. I think I've heard that there is a possible slight performance improvement, but really the failure mode makes it a poor tradeoff. John =:-> > > -Dan > --------------enig54CE91A96C702C95B2C308BF Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC+S55JdeBCYSNAAMRAsF5AJwMi+HzqgRohEH5XENK4+wJc3qOfACfSanr n5bGniMEoSJuHjJNNHqgIHk= =tXav -----END PGP SIGNATURE----- --------------enig54CE91A96C702C95B2C308BF-- From pgsql-patches-owner@postgresql.org Tue Aug 9 21:41:35 2005 X-Original-To: pgsql-patches-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1E30552927 for ; Tue, 9 Aug 2005 21:41:34 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 47137-05 for ; Wed, 10 Aug 2005 00:41:23 +0000 (GMT) Received: from mail.Mi8.com (d01gw01.mi8.com [63.240.6.47]) by svr1.postgresql.org (Postfix) with ESMTP id 3FB5B5295C for ; Tue, 9 Aug 2005 21:41:21 -0300 (ADT) Received: from 172.16.1.110 by mail.Mi8.com with ESMTP (- Welcome to Mi8 Corporation www.Mi8.com (D1)); Tue, 09 Aug 2005 20:41:19 -0400 X-Server-Uuid: 241911D6-425B-44B9-A073-E3FE0F8FC774 Received: from MI8NYCMAIL06.Mi8.com ([172.16.1.175]) by D01SMTP01.Mi8.com with Microsoft SMTPSVC(6.0.3790.211); Tue, 9 Aug 2005 20:41:18 -0400 Received: from 67.103.45.218 ([67.103.45.218]) by MI8NYCMAIL06.Mi8.com ( [172.16.1.219]) via Exchange Front-End Server mi8owa.mi8.com ( [172.16.1.104]) with Microsoft Exchange Server HTTP-DAV ; Tue, 9 Aug 2005 20:41:18 -0500 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Tue, 09 Aug 2005 17:41:18 -0700 Subject: Re: COPY FROM performance improvements From: "Alon Goldshuv" To: pgsql-patches@postgresql.org Message-ID: In-Reply-To: MIME-Version: 1.0 X-OriginalArrivalTime: 10 Aug 2005 00:41:18.0936 (UTC) FILETIME=[38FCDD80:01C59D44] X-WSS-ID: 6EE792A52B416104077-01-01 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.593 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO, RCVD_NUMERIC_HELO X-Spam-Level: * X-Archive-Number: 200508/75 X-Sequence-Number: 17022 I did some performance checks after the recent code commit. The good news is that the parsing speed of COPY is now MUCH faster, which is great. It is about 5 times faster - about 100MB/sec on my machine (previously 20MB/sec at best, usually less). The better news is that my original patch parsing speed reaches 120MB/sec, about 20MB/sec faster than the code that's now in CVS. This can be significant for the long scheme of things and for large data sets. Maybe we can improve the current code a bit more to reach this number. I performed those measurement by executing *only the parsing logic* of the COPY pipeline. All data conversion (functioncall3(string...)) and tuple handling (form_heaptuple etc...) and insertion were manually disabled. So the only code measured is reading from disk and parsing to the attribute level. Cheers, Alon. On 8/7/05 1:21 AM, "Luke Lonergan" wrote: > Tom, > > On 8/6/05 9:08 PM, "Tom Lane" wrote: > >> "Luke Lonergan" writes: >>>> I had some difficulty in generating test cases that weren't largely >>>> I/O-bound, but AFAICT the patch as applied is about the same speed >>>> as what you submitted. >> >>> You achieve the important objective of knocking the parsing stage down a >>> lot, but your parsing code is actually about 20% slower than Alon's. >> >> I would like to see the exact test case you are using to make this >> claim; the tests I did suggested my code is the same speed or faster. > > I showed mine - you show yours :-) Apparently our e-mail crossed. > >> As best I can tell, my version of CopyReadAttributes is significantly >> quicker than Alon's, approximately balancing out the fact that my >> version of CopyReadLine is slower. I did the latter first, and would >> now be tempted to rewrite it in the same style as CopyReadAttributes, >> ie one pass of memory-to-memory copy using pointers rather than buffer >> indexes. > > See previous timings - looks like Alon's parsing is substantially faster. > However, I'd like him to confirm by running with the "shunt" placed at > different stages, in this case between parse and attribute conversion (not > attribute parse). > >> BTW, late today I figured out a way to get fairly reproducible >> non-I/O-bound numbers about COPY FROM: use a trigger that suppresses >> the actual inserts, thus: >> >> create table foo ... >> create function noway() returns trigger as >> 'begin return null; end' language plpgsql; >> create trigger noway before insert on foo >> for each row execute procedure noway(); >> then repeat: >> copy foo from '/tmp/foo.data'; > > Cool! That's a better way than hacking code and inserting shunts. > > Alon will likely hit this tomorrow. > > - Luke > From pgsql-patches-owner@postgresql.org Tue Aug 9 22:01:45 2005 X-Original-To: pgsql-patches-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1B189528EF for ; Tue, 9 Aug 2005 22:01:44 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 32841-03 for ; Wed, 10 Aug 2005 01:01:42 +0000 (GMT) Received: from trolak.mydnsbox2.com (ns1.mydnsbox2.com [207.44.142.118]) by svr1.postgresql.org (Postfix) with ESMTP id 5BE7C528E1 for ; Tue, 9 Aug 2005 22:01:41 -0300 (ADT) Received: from [192.168.1.103] (cpe-024-211-165-134.nc.res.rr.com [24.211.165.134]) (authenticated (0 bits)) by trolak.mydnsbox2.com (8.11.6/8.11.6) with ESMTP id j7A0GoK23122; Tue, 9 Aug 2005 19:16:50 -0500 Message-ID: <42F951F2.8010104@dunslane.net> Date: Tue, 09 Aug 2005 21:01:38 -0400 From: Andrew Dunstan User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.10) Gecko/20050719 Fedora/1.7.10-1.3.1 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Alon Goldshuv Cc: pgsql-patches@postgresql.org Subject: Re: COPY FROM performance improvements References: In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.048 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/76 X-Sequence-Number: 17023 Alon Goldshuv wrote: >I performed those measurement by executing *only the parsing logic* of the >COPY pipeline. All data conversion (functioncall3(string...)) and tuple >handling (form_heaptuple etc...) and insertion were manually disabled. So >the only code measured is reading from disk and parsing to the attribute >level. > > Arguably this might exaggerate the effect quite significantly. Users will want to know the real time effect on a complete COPY. Depending on how much the pasing is in the total time your 20% improvement in parsing might only be a small fraction of 20% improvement in COPY. Like you, I'm happy we have seen a 5 times improvement in parsing. Is it possible you can factor out something smallish from your patch that might make up the balance? cheers andrew From pgsql-patches-owner@postgresql.org Wed Aug 10 01:40:17 2005 X-Original-To: pgsql-patches-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E26F852A1D for ; Wed, 10 Aug 2005 01:40:15 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 97085-01 for ; Wed, 10 Aug 2005 04:40:12 +0000 (GMT) Received: from mail.Mi8.com (nycgw01.mi8.com [63.240.6.42]) by svr1.postgresql.org (Postfix) with ESMTP id A88E1528EF for ; Wed, 10 Aug 2005 01:40:10 -0300 (ADT) Received: from 172.16.1.112 by mail.Mi8.com with ESMTP (- GW01 Welcome to Mi8 Corporation www.Mi8.com); Wed, 10 Aug 2005 00:39:58 -0400 X-Server-Uuid: F1A2E19A-84E4-48DD-8F48-B475613F58B2 Received: from MI8NYCMAIL06.Mi8.com ([172.16.1.175]) by D01SMTP02.Mi8.com with Microsoft SMTPSVC(6.0.3790.211); Wed, 10 Aug 2005 00:39:58 -0400 Received: from 24.5.173.15 ([24.5.173.15]) by MI8NYCMAIL06.Mi8.com ( [172.16.1.219]) via Exchange Front-End Server mi8owa.mi8.com ( [172.16.1.106]) with Microsoft Exchange Server HTTP-DAV ; Wed, 10 Aug 2005 00:39:58 -0500 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Tue, 09 Aug 2005 21:39:55 -0700 Subject: Re: COPY FROM performance improvements From: "Luke Lonergan" To: "Andrew Dunstan" , "Alon Goldshuv" Cc: pgsql-patches@postgresql.org Message-ID: In-Reply-To: <42F951F2.8010104@dunslane.net> MIME-Version: 1.0 X-OriginalArrivalTime: 10 Aug 2005 04:39:58.0822 (UTC) FILETIME=[904E1060:01C59D65] X-WSS-ID: 6EE75A9427K5951281-01-01 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.591 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO, RCVD_NUMERIC_HELO X-Spam-Level: * X-Archive-Number: 200508/80 X-Sequence-Number: 17027 Andrew, > Arguably this might exaggerate the effect quite significantly. Users > will want to know the real time effect on a complete COPY. Depending on > how much the pasing is in the total time your 20% improvement in parsing > might only be a small fraction of 20% improvement in COPY. Arguably has already been argued. We knew this because we profiled the entire COPY process and came up with this approx. breakdown for a specific case: Parsing: 25% Attribute Conversion: 40% Data Insertion: 35% Net copy rate: 8 MB/s on a filesystem that does 240 MB/s So - if we speed up parsing by 500% or 450%, the end result is about a 20-30% speed increase in the overall process. Note that we're still a *long* way from getting anywhere near the limit of the I/O subsystem at 12 MB/s. Oracle can probably get 5-8 times this data rate, if not better. The attribute conversion logic is also very slow and needs similar improvements. The reason we focused first on Parsing is that our MPP version of Bizgres will reach data loading rates approaching the parsing speed, so we needed to improve that part to get it out of the way. We will continue to improve COPY speed in Bizgres so that we can provide comparable COPY performance to Oracle and MySQL. > Like you, I'm happy we have seen a 5 times improvement in parsing. Is it > possible you can factor out something smallish from your patch that > might make up the balance? That parsing was 25% of the workload was traceable to a 3 main things: 1) Per character acquisition from source instead of buffering 2) Frequent interruption of the parsing pipeline to handle attribute conversion 3) Lack of micro-parallelism in the character finding logic Tom's patch took our contribution from (1) and (2) and his improvements, and he rejected (3). The net result is that we lost performance from the lack of (3) but gained performance from his improvements of (1) and (2). I believe that re-introducing (3) may bring us from 100 MB/s to 150 MB/s parsing speed. - Luke From pgsql-patches-owner@postgresql.org Wed Aug 10 01:48:24 2005 X-Original-To: pgsql-patches-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D020A528EE for ; Wed, 10 Aug 2005 01:48:23 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 73571-10 for ; Wed, 10 Aug 2005 04:48:14 +0000 (GMT) Received: from mail.Mi8.com (nycgw01.mi8.com [63.240.6.42]) by svr1.postgresql.org (Postfix) with ESMTP id 49BC95282F for ; Wed, 10 Aug 2005 01:48:13 -0300 (ADT) Received: from 172.16.1.118 by mail.Mi8.com with ESMTP (- GW01 Welcome to Mi8 Corporation www.Mi8.com); Wed, 10 Aug 2005 00:48:08 -0400 X-Server-Uuid: F1A2E19A-84E4-48DD-8F48-B475613F58B2 Received: from MI8NYCMAIL06.Mi8.com ([172.16.1.175]) by d01smtp03.Mi8.com with Microsoft SMTPSVC(6.0.3790.1830); Wed, 10 Aug 2005 00:48:08 -0400 Received: from 24.5.173.15 ([24.5.173.15]) by MI8NYCMAIL06.Mi8.com ( [172.16.1.219]) via Exchange Front-End Server mi8owa.mi8.com ( [172.16.1.106]) with Microsoft Exchange Server HTTP-DAV ; Wed, 10 Aug 2005 00:48:07 -0500 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Tue, 09 Aug 2005 21:48:02 -0700 Subject: Re: COPY FROM performance improvements From: "Luke Lonergan" To: "Tom Lane" Cc: "Alon Goldshuv" , pgsql-patches@postgresql.org Message-ID: In-Reply-To: <29007.1123387696@sss.pgh.pa.us> MIME-Version: 1.0 X-OriginalArrivalTime: 10 Aug 2005 04:48:08.0352 (UTC) FILETIME=[B4166A00:01C59D66] X-WSS-ID: 6EE7588227K5953989-01-01 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.591 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO, RCVD_NUMERIC_HELO X-Spam-Level: * X-Archive-Number: 200508/81 X-Sequence-Number: 17028 Tom, > As best I can tell, my version of CopyReadAttributes is significantly > quicker than Alon's, approximately balancing out the fact that my > version of CopyReadLine is slower. I did the latter first, and would > now be tempted to rewrite it in the same style as CopyReadAttributes, > ie one pass of memory-to-memory copy using pointers rather than buffer > indexes. I think you are right, with the exception that Alon's results prove out that the net result of your patch is 20% slower than his. I think with your speedup of CopyReadAttributes and some additional work on CopyReadLine the net result could be 50% faster than Alon's patch. The key thing that is missing is the lack of micro-parallelism in the character processing in this version. By "inverting the loop", or putting the characters into a buffer on the outside, then doing fast character scanning inside with special "fix-up" cases, we exposed long runs of pipeline-able code to the compiler. I think there is another way to accomplish the same thing and still preserve the current structure, but it requires "strip mining" the character buffer into chunks that can be processed with an explicit loop to check for the different characters. While it may seem artificial (it is), it will provide the compiler with the ability to pipeline the character finding logic over long runs. The other necessary element will have to avoid pipeline stalls from the "if" conditions as much as possible. Anyway, thanks for reviewing this code and improving it - it's important to bring speed increases to our collective customer base. With Bizgres, we're not satisfied with 12 MB/s, we won't stop until we saturate the I/O bus, so we may get more extreme with the code than seems reasonable for the general audience. - Luke From pgsql-performance-owner@postgresql.org Tue Aug 9 16:14:30 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C048A529F6 for ; Tue, 9 Aug 2005 15:58:43 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 99823-07 for ; Tue, 9 Aug 2005 18:58:40 +0000 (GMT) Received: from mail.tecarta.com (66.238.115.135.ptr.us.xo.net [66.238.115.135]) by svr1.postgresql.org (Postfix) with ESMTP id 59927529F4 for ; Tue, 9 Aug 2005 15:58:39 -0300 (ADT) Received: from mail pickup service by mail.tecarta.com with Microsoft SMTPSVC; Tue, 9 Aug 2005 12:02:24 -0700 Received: from mail.tecarta.com ([192.168.160.2]) by mail.tecarta.com with Microsoft SMTPSVC(6.0.3790.0); Tue, 9 Aug 2005 12:02:21 -0700 Received: from barracuda.tecarta.com ([192.168.160.200]) by mail.tecarta.com (SAVSMTP 3.1.0.29) with SMTP id M2005080912022128114 for ; Tue, 09 Aug 2005 12:02:21 -0700 X-ASG-Debug-ID: 1123613917-1115-6-0 X-Barracuda-URL: http://192.168.160.200:8000/cgi-bin/mark.cgi Received: from mail1 (mail1.hq.corp [192.168.160.5]) by barracuda.tecarta.com (Spam Firewall) with SMTP id 622B62014AF5 for ; Tue, 9 Aug 2005 11:58:37 -0700 (PDT) Received: from amd64-laptop-spoe ([63.206.203.145]) by mail.tecarta.com with Microsoft SMTPSVC(6.0.3790.1830); Tue, 9 Aug 2005 12:02:20 -0700 X-ASG-Orig-Subj: Re: [PERFORM] Table locking problems? Subject: Re: Table locking problems? From: Steve Poe To: Dan Harris Cc: pgsql-performance@postgresql.org In-Reply-To: <52CFA236-FC43-4018-B399-1854240110FF@drivefaster.net> References: <52CFA236-FC43-4018-B399-1854240110FF@drivefaster.net> Content-Type: text/plain Date: Wed, 10 Aug 2005 06:49:07 +0000 Message-Id: <1123656547.13362.43.camel@amd64-laptop-spoe> Mime-Version: 1.0 X-Mailer: Evolution 2.2.3 Content-Transfer-Encoding: 7bit X-OriginalArrivalTime: 09 Aug 2005 19:02:20.0428 (UTC) FILETIME=[DE4AD4C0:01C59D14] X-Virus-Scanned: by Barracuda Spam Firewall at tecarta.com X-Barracuda-Spam-Score: 1.30 X-Barracuda-Spam-Status: No, SCORE=1.30 using global scores of TAG_LEVEL=4.0 QUARANTINE_LEVEL=1000.0 KILL_LEVEL=7.0 tests=BAYES_50, DATE_IN_FUTURE_06_12 X-Barracuda-Spam-Report: Code version 3.02, rules version 3.0.3093 Rule breakdown below pts rule name description ---- ---------------------- -------------------------------------------------- 1.30 DATE_IN_FUTURE_06_12 Date: is 6 to 12 hours after Received: date 0.00 BAYES_50 BODY: Bayesian spam probability is 40 to 60% [score: 0.5985] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.75 tagged_above=0 required=5 tests=AWL, DATE_IN_FUTURE_06_12, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/89 X-Sequence-Number: 13836 Dan, Do you mean you did RAID 1 + 0 (RAID 10) or RAID 0 + 1? Just a clarification, since RAID 0 is still a single-point of failure even if RAID1 is on top of RAID0. How many users are connected when your update / delete queries are hanging? Have you done an analyze verbose on those queries? Have you made changes to the postgresql.conf? kernel.vm settings? IO scheduler? If you're not doing so already, you may consider running sar (iostat) to monitor when the hanging occurs if their is a memory / IO bottleneck somewhere. Good luck. Steve Poe On Tue, 2005-08-09 at 12:04 -0600, Dan Harris wrote: > I thought I would send this to pg-performance since so many people > helped me with my speed issues recently. I was definitely IO- > bottlenecked. > > Since then, I have installed 2 RAID arrays with 7 15k drives in them > in RAID 0+1 as well as add a new controller card with 512MB of cache > on it. I also created this new partition on the RAID as XFS instead > of ext3. > > These changes have definitely improved performance, but I am now > finding some trouble with UPDATE or DELETE queries "hanging" and > never releasing their locks. As this happens, other statements queue > up behind it. It seems to occur at times of very high loads on the > box. Is my only option to kill the query ( which usually takes down > the whole postmaster with it! ouch ). > > Could these locking issues be related to the other changes I made? > I'm really scared that this is related to choosing XFS, but I sure > hope not. How should I go about troubleshooting the "problem" > queries? They don't seem to be specific to a single table or single > database. > > I'm running 8.0.1 on kernel 2.6.12-3 on 64-bit Opterons if that > matters.. > > > -Dan > > ---------------------------(end of broadcast)--------------------------- > TIP 9: In versions below 8.0, the planner will ignore your desire to > choose an index scan if your joining column's datatypes do not > match From pgsql-patches-owner@postgresql.org Wed Aug 10 05:16:26 2005 X-Original-To: pgsql-patches-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2EE0652ADB for ; Wed, 10 Aug 2005 05:16:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 96362-04 for ; Wed, 10 Aug 2005 08:16:15 +0000 (GMT) Received: from smtp1.freeserve.com (smtp1.wanadoo.co.uk [193.252.22.158]) by svr1.postgresql.org (Postfix) with ESMTP id A040852C20 for ; Wed, 10 Aug 2005 05:16:14 -0300 (ADT) Received: from me-wanadoo.net (nullmx.kyokofukada.net [127.0.0.1]) by mwinf3004.me.freeserve.com (SMTP Server) with ESMTP id D6C971C001C4; Wed, 10 Aug 2005 10:16:14 +0200 (CEST) Received: from 192.168.0.102 (modem-951.lynx.dialup.pol.co.uk [217.135.195.183]) by mwinf3004.me.freeserve.com (SMTP Server) with ESMTP id 4F61B1C001C3; Wed, 10 Aug 2005 10:16:14 +0200 (CEST) X-ME-UUID: 20050810081614325.4F61B1C001C3@mwinf3004.me.freeserve.com Subject: Re: COPY FROM performance improvements From: Simon Riggs To: Luke Lonergan Cc: Tom Lane , Alon Goldshuv , pgsql-patches@postgresql.org In-Reply-To: References: Content-Type: text/plain Organization: 2nd Quadrant Date: Wed, 10 Aug 2005 09:15:00 +0100 Message-Id: <1123661700.3670.621.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.0.2 (2.0.2-3) Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.052 tagged_above=0 required=5 tests=AWL, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200508/85 X-Sequence-Number: 17032 On Tue, 2005-08-09 at 21:48 -0700, Luke Lonergan wrote: > The key thing that is missing is the lack of micro-parallelism in the > character processing in this version. By "inverting the loop", or putting > the characters into a buffer on the outside, then doing fast character > scanning inside with special "fix-up" cases, we exposed long runs of > pipeline-able code to the compiler. > > I think there is another way to accomplish the same thing and still preserve > the current structure, but it requires "strip mining" the character buffer > into chunks that can be processed with an explicit loop to check for the > different characters. While it may seem artificial (it is), it will provide > the compiler with the ability to pipeline the character finding logic over > long runs. The other necessary element will have to avoid pipeline stalls > from the "if" conditions as much as possible. This is a key point, IMHO. That part of the code was specifically written to take advantage of processing pipelines in the hardware, not because the actual theoretical algorithm for that approach was itself faster. Nobody's said what compiler/hardware they have been using, so since both Alon and Tom say their character finding logic is faster, it is likely to be down to that? Name your platforms gentlemen, please. My feeling is that we may learn something here that applies more widely across many parts of the code. Best Regards, Simon Riggs From pgsql-patches-owner@postgresql.org Wed Aug 10 12:30:06 2005 X-Original-To: pgsql-patches-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id F1016529F1 for ; Wed, 10 Aug 2005 12:30:03 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 08043-04 for ; Wed, 10 Aug 2005 15:29:56 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 000365289C for ; Wed, 10 Aug 2005 12:29:55 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7AFTnha026483; Wed, 10 Aug 2005 11:29:50 -0400 (EDT) To: Simon Riggs Cc: Luke Lonergan , Alon Goldshuv , pgsql-patches@postgresql.org Subject: Re: COPY FROM performance improvements In-reply-to: <1123661700.3670.621.camel@localhost.localdomain> References: <1123661700.3670.621.camel@localhost.localdomain> Comments: In-reply-to Simon Riggs message dated "Wed, 10 Aug 2005 09:15:00 +0100" Date: Wed, 10 Aug 2005 11:29:49 -0400 Message-ID: <26482.1123687789@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/91 X-Sequence-Number: 17038 Simon Riggs writes: > Nobody's said what compiler/hardware they have been using, so since both > Alon and Tom say their character finding logic is faster, it is likely > to be down to that? Name your platforms gentlemen, please. I tested on HPPA with gcc 2.95.3 and on a Pentium 4 with gcc 3.4.3. Got pretty much the same results on both. regards, tom lane From pgsql-performance-owner@postgresql.org Wed Aug 10 14:53:07 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5FC6B52806 for ; Wed, 10 Aug 2005 14:52:32 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 23587-10 for ; Wed, 10 Aug 2005 17:52:22 +0000 (GMT) Received: from mail.nordicbet.com (mail.nordicbet.com [193.69.167.130]) by svr1.postgresql.org (Postfix) with ESMTP id BD32252C3C for ; Wed, 10 Aug 2005 14:52:17 -0300 (ADT) Received: from [84.48.51.94] (helo=tobias.nordicbet.invalid) (Authenticated Sender=tobias@nordicbet.com) by mail.nordicbet.com with esmtpa (Exim 4.50 #1 (Debian)) id 1E2uiM-0004XD-9f for ; Wed, 10 Aug 2005 19:50:13 +0200 Received: by tobias.nordicbet.invalid (Postfix, from userid 500) id 7F3ABE14C2; Wed, 10 Aug 2005 19:52:08 +0200 (CEST) Date: Wed, 10 Aug 2005 19:52:08 +0200 From: Tobias Brox To: pgsql-performance@postgresql.org Subject: partial index regarded more expensive Message-ID: <20050810175208.GD6141@tobias.lan> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Organization: Group Nordicbet User-Agent: Mutt/1.5.8i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/96 X-Sequence-Number: 13843 So, I have a table game with a timestamp attribute 'game_end', ranging from jan-2005 to present. The game table also have an attribute state, with live games beeing in state 2, and ended games beeing in state 4 (so, game_end+delta>now() usually means state=4). There are also an insignificant number of games in states 1,3. This query puzzles me: select * from game where game_end>'2005-07-30' and state in (3,4); Now, one (at least me) should believe that the best index would be a partial index, "resolved_game_by_date" btree (game_end) WHERE ((state = 3) OR (state = 4)) NBET=> explain analyze select * from game where game_end>'2005-07-30' and state in (3,4); QUERY PLAN ---------------------------------------------------------------------------------------------------------------------------------------- Index Scan using resolved_game_by_date on game (cost=0.00..7002.87 rows=7147 width=555) (actual time=0.220..86.234 rows=3852 loops=1) Index Cond: (game_end > '2005-07-30 00:00:00'::timestamp without time zone) Filter: ((state = 3) OR (state = 4)) Total runtime: 90.568 ms (4 rows) Since state has only two significant states, I wouldn't believe this index to be any good: "game_by_state" btree (state) ...and it seems like I'm right: NBET=> explain analyze select * from game where game_end>'2005-07-30' and state in (3,4); QUERY PLAN ------------------------------------------------------------------------------------------------------------------------------------------------ Index Scan using game_by_state, game_by_state on game (cost=0.00..4413.78 rows=7147 width=555) (actual time=0.074..451.771 rows=3851 loops=1) Index Cond: ((state = 3) OR (state = 4)) Filter: (game_end > '2005-07-30 00:00:00'::timestamp without time zone) Total runtime: 457.132 ms (4 rows) Now, how can the planner believe the game_by_state-index to be better? ('vacuum analyze game' did not significantly impact the numbers, and I've tried running the queries some times with and without the game_by_state-index to rule out cacheing effects) -- Tobias Brox This signature has been virus scanned, and is probably safe to read. This mail may contain confidential information, please keep your eyes closed. From pgsql-performance-owner@postgresql.org Wed Aug 10 15:15:19 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D6DD052939 for ; Wed, 10 Aug 2005 15:15:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 92395-04 for ; Wed, 10 Aug 2005 18:15:15 +0000 (GMT) Received: from boutiquenumerique.com (boutiquenumerique.com [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 6850552924 for ; Wed, 10 Aug 2005 15:15:14 -0300 (ADT) Received: (qmail 22982 invoked from network); 10 Aug 2005 20:15:28 +0200 Received: from unknown (HELO localhost) (boutiquenumerique-lists@192.168.0.4) by boutiquenumerique.com with SMTP; 10 Aug 2005 20:15:28 +0200 To: "Tobias Brox" , pgsql-performance@postgresql.org Subject: Re: partial index regarded more expensive References: <20050810175208.GD6141@tobias.lan> Message-ID: From: PFC Organization: =?iso-8859-15?Q?La_Boutique_Num=E9rique?= Content-Type: text/plain; format=flowed; delsp=yes; charset=iso-8859-15 MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Date: Wed, 10 Aug 2005 20:15:13 +0200 In-Reply-To: <20050810175208.GD6141@tobias.lan> User-Agent: Opera M2/8.0 (Linux, build 1095) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200508/97 X-Sequence-Number: 13844 why not simply create an index on (game_end, state) ? From pgsql-performance-owner@postgresql.org Wed Aug 10 15:32:06 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 21DA352991 for ; Wed, 10 Aug 2005 15:32:04 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 19844-04 for ; Wed, 10 Aug 2005 18:31:54 +0000 (GMT) Received: from mail.nordicbet.com (mail.nordicbet.com [193.69.167.130]) by svr1.postgresql.org (Postfix) with ESMTP id 94D5B5280D for ; Wed, 10 Aug 2005 15:31:53 -0300 (ADT) Received: from [84.48.51.94] (helo=tobias.nordicbet.invalid) (Authenticated Sender=tobias@nordicbet.com) by mail.nordicbet.com with esmtpa (Exim 4.50 #1 (Debian)) id 1E2vKf-0006DA-8l; Wed, 10 Aug 2005 20:29:51 +0200 Received: by tobias.nordicbet.invalid (Postfix, from userid 500) id DB648CACF3; Wed, 10 Aug 2005 20:31:42 +0200 (CEST) Date: Wed, 10 Aug 2005 20:31:42 +0200 From: Tobias Brox To: PFC Cc: Tobias Brox , pgsql-performance@postgresql.org Subject: Re: partial index regarded more expensive Message-ID: <20050810183142.GG6141@tobias.lan> References: <20050810175208.GD6141@tobias.lan> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: Organization: Group Nordicbet User-Agent: Mutt/1.5.8i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/98 X-Sequence-Number: 13845 [PFC - Wed at 08:15:13PM +0200] > why not simply create an index on (game_end, state) ? No, the planner prefers to use the partial index (I dropped the index on game(state)). -- Tobias Brox, Nordicbet IT dept This signature has been virus scanned, and is probably safe to read. This mail may contain confidential information, please keep your eyes closed. From pgsql-performance-owner@postgresql.org Wed Aug 10 18:04:09 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CC93A52A4B; Wed, 10 Aug 2005 18:04:03 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 92089-01; Wed, 10 Aug 2005 21:03:58 +0000 (GMT) Received: from mailhost.intellivid.com (mailhost.intellivid.com [64.32.200.11]) by svr1.postgresql.org (Postfix) with ESMTP id 6D35252A44; Wed, 10 Aug 2005 18:03:52 -0300 (ADT) Received: from [192.168.2.68] (spectre.intellivid.com [192.168.2.68]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client did not present a certificate) by newmail.intellivid.com (Postfix) with ESMTP id C0CC8F182AF; Wed, 10 Aug 2005 17:03:53 -0400 (EDT) Subject: Re: Planner doesn't look at LIMIT? From: Ian Westmacott To: Tom Lane Cc: pgsql-performance@postgresql.org, pgsql-hackers@postgresql.org In-Reply-To: <6440.1122049220@sss.pgh.pa.us> References: <758d5e7f0507220210bfa4978@mail.gmail.com> <29473.1122043197@sss.pgh.pa.us> <6440.1122049220@sss.pgh.pa.us> Content-Type: text/plain Organization: Intellivid Corp. Message-Id: <1123707833.17725.133.camel@spectre.intellivid.com> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Wed, 10 Aug 2005 17:03:53 -0400 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests= X-Spam-Level: X-Archive-Number: 200508/99 X-Sequence-Number: 13846 I have a case that I though was an example of this issue, and that this patch would correct. I applied this patch to an 8.0.3 source distribution, but it didn't seem to solve my problem. In a nutshell, I have a LIMIT query where the planner seems to favor a merge join over a nested loop. I've simplified the query as much as possible: itvtrackdata3=> \d tableA Table "public.tableA" Column | Type | Modifiers --------+----------+----------- foo | bigint | not null bar | smallint | not null bap | bigint | not null bip | bigint | not null bom | bigint | not null Indexes: "idx_tableA_bip" btree (bip) WHERE (bip = 9000000000000000000::bigint) "idx_tableA_foo" btree (foo) itvtrackdata3=> \d tableB Table "tableB" Column | Type | Modifiers ---------+----------+----------- bim | bigint | not null bif | smallint | not null baf | smallint | not null bof | smallint | not null buf | smallint | not null foo | bigint | not null Indexes: "idx_tableB_bim" btree ("bim", foo) itvtrackdata3=> set default_statistics_target to 1000; SET Time: 0.448 ms itvtrackdata3=> analyze tableA; ANALYZE Time: 4237.151 ms itvtrackdata3=> analyze tableB; ANALYZE Time: 46672.939 ms itvtrackdata3=> explain analyze SELECT * FROM tableB NATURAL JOIN tableA WHERE bim>=72555896091359 AND bim<72555935412959 AND bim=bap ORDER BY bim ASC LIMIT 1; QUERY PLAN -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Limit (cost=149626.57..252987.71 rows=1 width=50) (actual time=5684.013..5684.013 rows=1 loops=1) -> Merge Join (cost=149626.57..252987.71 rows=1 width=50) (actual time=5684.012..5684.012 rows=1 loops=1) Merge Cond: (("outer"."bim" = "inner"."bap") AND ("outer".foo = "inner".foo)) -> Index Scan using idx_tableB_bim on tableB (cost=0.00..97391.22 rows=55672 width=24) (actual time=0.017..0.059 rows=29 loops=1) Index Cond: (("bim" >= 72555896091359::bigint) AND ("bim" < 72555935412959::bigint)) -> Sort (cost=149626.57..151523.94 rows=758948 width=34) (actual time=5099.300..5442.825 rows=560856 loops=1) Sort Key: tableA."bap", tableA.foo -> Seq Scan on tableA (cost=0.00..47351.48 rows=758948 width=34) (actual time=0.021..1645.204 rows=758948 loops=1) Total runtime: 5706.655 ms (9 rows) Time: 5729.984 ms itvtrackdata3=> set enable_mergejoin to false; SET Time: 0.373 ms itvtrackdata3=> explain analyze SELECT * FROM tableB NATURAL JOIN tableA WHERE bim>=72555896091359 AND bim<72555935412959 AND bim=bap ORDER BY bim ASC LIMIT 1; QUERY PLAN -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Limit (cost=0.00..432619.68 rows=1 width=50) (actual time=11.149..11.150 rows=1 loops=1) -> Nested Loop (cost=0.00..432619.68 rows=1 width=50) (actual time=11.148..11.148 rows=1 loops=1) Join Filter: ("outer"."bim" = "inner"."bap") -> Index Scan using idx_tableB_bim on tableB (cost=0.00..97391.22 rows=55672 width=24) (actual time=0.017..0.062 rows=29 loops=1) Index Cond: (("bim" >= 72555896091359::bigint) AND ("bim" < 72555935412959::bigint)) -> Index Scan using idx_tableA_foo on tableA (cost=0.00..6.01 rows=1 width=34) (actual time=0.007..0.379 rows=1 loops=29) Index Cond: ("outer".foo = tableA.foo) Total runtime: 11.215 ms (8 rows) Time: 32.007 ms Have I just flubbed the patch, or is there something else going on here? Thanks, --Ian On Fri, 2005-07-22 at 12:20, Tom Lane wrote: > I wrote: > > Dawid Kuroczko writes: > >> qnex=# EXPLAIN SELECT * FROM log NATURAL JOIN useragents LIMIT 1; > > >> Limit (cost=15912.20..15912.31 rows=1 width=272) > >> -> Hash Join (cost=15912.20..5328368.96 rows=47044336 width=272) > > >> If I set enable_hashjoin=false: > > >> qnex=# EXPLAIN ANALYZE SELECT * FROM log NATURAL LEFT JOIN useragents LIMIT 1; > > >> Limit (cost=0.00..3.07 rows=1 width=272) (actual time=74.214..74.216 > >> rows=1 loops=1) > >> -> Nested Loop Left Join (cost=0.00..144295895.01 rows=47044336 > >> width=272) (actual time=74.204..74.204 rows=1 loops=1) > > > This is quite strange. The nestloop plan definitely should be preferred > > in the context of the LIMIT, considering that it has far lower estimated > > cost. And it is preferred in simple tests for me. > > After a suitable period of contemplating my navel, I figured out > what is going on here: the total costs involved are large enough that > the still-fairly-high startup cost of the hash is disregarded by > compare_fuzzy_path_costs(), and so the nestloop is discarded as not > having any significant potential advantage in startup time. > > I think that this refutes the original scheme of using the same fuzz > factor for both startup and total cost comparisons, and therefore > propose the attached patch. > > Comments? > > regards, tom lane > > *** src/backend/optimizer/util/pathnode.c.orig Fri Jul 15 13:09:25 2005 > --- src/backend/optimizer/util/pathnode.c Fri Jul 22 12:08:25 2005 > *************** > *** 98,157 **** > static int > compare_fuzzy_path_costs(Path *path1, Path *path2, CostSelector criterion) > { > - Cost fuzz; > - > /* > ! * The fuzz factor is set at one percent of the smaller total_cost, > ! * but not less than 0.01 cost units (just in case total cost is > ! * zero). > * > * XXX does this percentage need to be user-configurable? > */ > - fuzz = Min(path1->total_cost, path2->total_cost) * 0.01; > - fuzz = Max(fuzz, 0.01); > - > if (criterion == STARTUP_COST) > { > ! if (Abs(path1->startup_cost - path2->startup_cost) > fuzz) > ! { > ! if (path1->startup_cost < path2->startup_cost) > ! return -1; > ! else > ! return +1; > ! } > > /* > * If paths have the same startup cost (not at all unlikely), > * order them by total cost. > */ > ! if (Abs(path1->total_cost - path2->total_cost) > fuzz) > ! { > ! if (path1->total_cost < path2->total_cost) > ! return -1; > ! else > ! return +1; > ! } > } > else > { > ! if (Abs(path1->total_cost - path2->total_cost) > fuzz) > ! { > ! if (path1->total_cost < path2->total_cost) > ! return -1; > ! else > ! return +1; > ! } > > /* > * If paths have the same total cost, order them by startup cost. > */ > ! if (Abs(path1->startup_cost - path2->startup_cost) > fuzz) > ! { > ! if (path1->startup_cost < path2->startup_cost) > ! return -1; > ! else > ! return +1; > ! } > } > return 0; > } > --- 98,138 ---- > static int > compare_fuzzy_path_costs(Path *path1, Path *path2, CostSelector criterion) > { > /* > ! * We use a fuzz factor of 1% of the smaller cost. > * > * XXX does this percentage need to be user-configurable? > */ > if (criterion == STARTUP_COST) > { > ! if (path1->startup_cost > path2->startup_cost * 1.01) > ! return +1; > ! if (path2->startup_cost > path1->startup_cost * 1.01) > ! return -1; > > /* > * If paths have the same startup cost (not at all unlikely), > * order them by total cost. > */ > ! if (path1->total_cost > path2->total_cost * 1.01) > ! return +1; > ! if (path2->total_cost > path1->total_cost * 1.01) > ! return -1; > } > else > { > ! if (path1->total_cost > path2->total_cost * 1.01) > ! return +1; > ! if (path2->total_cost > path1->total_cost * 1.01) > ! return -1; > > /* > * If paths have the same total cost, order them by startup cost. > */ > ! if (path1->startup_cost > path2->startup_cost * 1.01) > ! return +1; > ! if (path2->startup_cost > path1->startup_cost * 1.01) > ! return -1; > } > return 0; > } > > ---------------------------(end of broadcast)--------------------------- > TIP 3: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faq From pgsql-performance-owner@postgresql.org Wed Aug 10 19:55:30 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3ABF952922; Wed, 10 Aug 2005 19:55:27 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 17260-06; Wed, 10 Aug 2005 22:55:23 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id DA9CB52CA9; Wed, 10 Aug 2005 19:55:22 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7AMtOWq016041; Wed, 10 Aug 2005 18:55:24 -0400 (EDT) To: Ian Westmacott Cc: pgsql-performance@postgresql.org, pgsql-hackers@postgresql.org Subject: Re: Planner doesn't look at LIMIT? In-reply-to: <1123707833.17725.133.camel@spectre.intellivid.com> References: <758d5e7f0507220210bfa4978@mail.gmail.com> <29473.1122043197@sss.pgh.pa.us> <6440.1122049220@sss.pgh.pa.us> <1123707833.17725.133.camel@spectre.intellivid.com> Comments: In-reply-to Ian Westmacott message dated "Wed, 10 Aug 2005 17:03:53 -0400" Date: Wed, 10 Aug 2005 18:55:24 -0400 Message-ID: <16040.1123714524@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/100 X-Sequence-Number: 13847 Ian Westmacott writes: > In a nutshell, I have a LIMIT query where the planner > seems to favor a merge join over a nested loop. The planner is already estimating only one row out of the join, and so the LIMIT doesn't affect its cost estimates at all. It appears to me that the reason the nestloop plan is fast is just chance: a suitable matching row is found very early in the scan of tableB, so that the indexscan on it can stop after 29 rows, instead of having to go through all 55000 rows in the given range of bim. If it'd have had to go through, say, half of the rows to find a match, the sort/merge plan would show up a lot better. If this wasn't chance, but was expected because there are many matching rows and not only one, then there's a statistical problem. regards, tom lane From pgsql-performance-owner@postgresql.org Wed Aug 10 20:35:22 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D8E7452B89 for ; Wed, 10 Aug 2005 20:35:20 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 13372-02 for ; Wed, 10 Aug 2005 23:35:19 +0000 (GMT) Received: from crestone.coronasolutions.com (crestone.coronasolutions.com [66.45.104.24]) by svr1.postgresql.org (Postfix) with ESMTP id EF41752AD8 for ; Wed, 10 Aug 2005 20:35:17 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by crestone.coronasolutions.com (Postfix) with ESMTP id 775336440A5 for ; Wed, 10 Aug 2005 17:34:49 -0600 (MDT) Received: from crestone.coronasolutions.com ([127.0.0.1]) by localhost (crestone.coronasolutions.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 13086-09 for ; Wed, 10 Aug 2005 17:34:47 -0600 (MDT) Received: from [192.168.1.12] (c-24-9-24-35.hsd1.co.comcast.net [24.9.24.35]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (No client certificate requested) by crestone.coronasolutions.com (Postfix) with ESMTP id 460076440F2 for ; Wed, 10 Aug 2005 17:34:47 -0600 (MDT) Mime-Version: 1.0 (Apple Message framework v730) Content-Transfer-Encoding: 7bit Message-Id: <3E43C7D5-9DB8-4132-BF4F-607B000866E1@drivefaster.net> Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed To: pgsql-performance@postgresql.org From: Dan Harris Subject: Speedier count(*) Date: Wed, 10 Aug 2005 17:37:49 -0600 X-Mailer: Apple Mail (2.730) X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at drivefaster.net X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/101 X-Sequence-Number: 13848 I have a web page for my customers that shows them count of records and some min/max date ranges in each table of a database, as this is how we bill them for service. They can log in and check the counts at any time. I'd like for the counts to be as fresh as possible by keeping this dynamic, but I will use a periodic 'snapshot'/cron job if that is the only option to speed this up. I have thought about using the table statistics, but the estimate error is probably unacceptable because of the billing purposes. For some reason, the SQL Server we migrated the app from can return count(*) in a split second on multi-million row tables, even though it is a MUCH slower box hardware-wise, but it's now taking many seconds to run. I have read in the archives the problems MVCC brings into the count(*) dilemma forcing Pg to run a seq scan to get counts. Does SQLServer not use MVCC or have they found another approach for arriving at this number? Compounding all the min/max and counts from other tables and all those queries take about a minute to run. The tables will contain anywhere from 1 million to 40 million rows. Also, I am using "select ... group by ... order by .. limit 1" to get the min/max since I have already been bit by the issue of min() max() being slower. -Dan From pgsql-performance-owner@postgresql.org Wed Aug 10 21:23:40 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 27FAF52E73 for ; Wed, 10 Aug 2005 21:23:37 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 35937-08 for ; Thu, 11 Aug 2005 00:23:32 +0000 (GMT) Received: from hosting.commandprompt.com (128.commandprompt.com [207.173.200.128]) by svr1.postgresql.org (Postfix) with ESMTP id 6B2CF52E6F for ; Wed, 10 Aug 2005 21:23:32 -0300 (ADT) Received: from [192.168.1.105] (clbb-248.saw.net [64.146.135.248]) (authenticated bits=0) by hosting.commandprompt.com (8.13.4/8.13.4) with ESMTP id j7B0MRef029344; Wed, 10 Aug 2005 17:22:27 -0700 Message-ID: <42FA9A8B.1040400@commandprompt.com> Date: Wed, 10 Aug 2005 17:23:39 -0700 From: "Joshua D. Drake" Organization: Command Prompt, Inc. User-Agent: Mozilla Thunderbird 1.0.2 (Windows/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Dan Harris Cc: pgsql-performance@postgresql.org Subject: Re: Speedier count(*) References: <3E43C7D5-9DB8-4132-BF4F-607B000866E1@drivefaster.net> In-Reply-To: <3E43C7D5-9DB8-4132-BF4F-607B000866E1@drivefaster.net> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Greylist: Sender succeded SMTP AUTH authentication, not delayed by milter-greylist-1.6 (hosting.commandprompt.com [192.168.1.101]); Wed, 10 Aug 2005 17:22:28 -0700 (PDT) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 tagged_above=0 required=5 tests=[AWL=0.024] X-Spam-Level: X-Archive-Number: 200508/102 X-Sequence-Number: 13849 > Also, I am using "select ... group by ... order by .. limit 1" to get > the min/max since I have already been bit by the issue of min() max() > being slower. This specific instance is fixed in 8.1 Sincerely, Joshua D. Drake > > > -Dan > > > > ---------------------------(end of broadcast)--------------------------- > TIP 9: In versions below 8.0, the planner will ignore your desire to > choose an index scan if your joining column's datatypes do not > match From pgsql-performance-owner@postgresql.org Wed Aug 10 21:36:44 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4F04352E6F for ; Wed, 10 Aug 2005 21:36:43 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 03660-02 for ; Thu, 11 Aug 2005 00:36:40 +0000 (GMT) Received: from tigger.fuhr.org (tigger.fuhr.org [63.214.45.158]) by svr1.postgresql.org (Postfix) with ESMTP id 44DAD52E63 for ; Wed, 10 Aug 2005 21:36:40 -0300 (ADT) Received: from winnie.fuhr.org (winnie.fuhr.org [10.1.0.1]) by tigger.fuhr.org (8.13.3/8.13.3) with ESMTP id j7B0aZSg047037 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK); Wed, 10 Aug 2005 18:36:37 -0600 (MDT) (envelope-from mfuhr@winnie.fuhr.org) Received: from winnie.fuhr.org (localhost [127.0.0.1]) by winnie.fuhr.org (8.13.3/8.13.3) with ESMTP id j7B0aZgW069576; Wed, 10 Aug 2005 18:36:35 -0600 (MDT) (envelope-from mfuhr@winnie.fuhr.org) Received: (from mfuhr@localhost) by winnie.fuhr.org (8.13.3/8.13.3/Submit) id j7B0aZIn069575; Wed, 10 Aug 2005 18:36:35 -0600 (MDT) (envelope-from mfuhr) Date: Wed, 10 Aug 2005 18:36:35 -0600 From: Michael Fuhr To: Dan Harris Cc: pgsql-performance@postgresql.org Subject: Re: Speedier count(*) Message-ID: <20050811003635.GA69508@winnie.fuhr.org> References: <3E43C7D5-9DB8-4132-BF4F-607B000866E1@drivefaster.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <3E43C7D5-9DB8-4132-BF4F-607B000866E1@drivefaster.net> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 tagged_above=0 required=5 tests=[AWL=0.005] X-Spam-Level: X-Archive-Number: 200508/103 X-Sequence-Number: 13850 On Wed, Aug 10, 2005 at 05:37:49PM -0600, Dan Harris wrote: > Also, I am using "select ... group by ... order by .. limit 1" to get > the min/max since I have already been bit by the issue of min() max() > being slower. PostgreSQL 8.1 will have optimizations for certain MIN and MAX queries. http://archives.postgresql.org/pgsql-committers/2005-04/msg00163.php http://archives.postgresql.org/pgsql-committers/2005-04/msg00168.php -- Michael Fuhr From pgsql-performance-owner@postgresql.org Wed Aug 10 22:07:56 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C4B5252E84 for ; Wed, 10 Aug 2005 22:07:54 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 62347-04 for ; Thu, 11 Aug 2005 01:07:53 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id BD28152E82 for ; Wed, 10 Aug 2005 22:07:51 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050811010754m92009iclke>; Thu, 11 Aug 2005 01:07:54 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id CA70256008; Wed, 10 Aug 2005 20:07:53 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id EE4C755FCC; Wed, 10 Aug 2005 20:07:49 -0500 (CDT) Message-ID: <42FAA4E6.5070703@arbash-meinel.com> Date: Wed, 10 Aug 2005 20:07:50 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Dan Harris Cc: pgsql-performance@postgresql.org Subject: Re: Speedier count(*) References: <3E43C7D5-9DB8-4132-BF4F-607B000866E1@drivefaster.net> In-Reply-To: <3E43C7D5-9DB8-4132-BF4F-607B000866E1@drivefaster.net> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig9CBF65940C3E69561ED666F5" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.044 tagged_above=0 required=5 tests=[AWL=-0.006, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/104 X-Sequence-Number: 13851 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig9CBF65940C3E69561ED666F5 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Dan Harris wrote: > I have a web page for my customers that shows them count of records and > some min/max date ranges in each table of a database, as this is how we > bill them for service. They can log in and check the counts at any > time. I'd like for the counts to be as fresh as possible by keeping > this dynamic, but I will use a periodic 'snapshot'/cron job if that is > the only option to speed this up. I have thought about using the > table statistics, but the estimate error is probably unacceptable > because of the billing purposes. > > For some reason, the SQL Server we migrated the app from can return > count(*) in a split second on multi-million row tables, even though it > is a MUCH slower box hardware-wise, but it's now taking many seconds to > run. I have read in the archives the problems MVCC brings into the > count(*) dilemma forcing Pg to run a seq scan to get counts. Does > SQLServer not use MVCC or have they found another approach for arriving > at this number? Compounding all the min/max and counts from other > tables and all those queries take about a minute to run. The tables > will contain anywhere from 1 million to 40 million rows. I believe SQL Server doesn't use MVCC in the same way. At the very least, it stores some row information in the index, so it can get some info from just an index, without having to go to the actual page (MVCC requires a main page visit to determine visibility.) Depending on how much it impacts performance, you can create an INSERT/UPDATE trigger so that whenever a new entry is added, it automatically updates a statistics table. It would be maintained as you go, rather than periodically like a cron job. I would go Cron if things can be slightly out of date (like 1 hour at least), and you need updates & inserts to not be slowed down. Otherwise I think the trigger is nicer, since it doesn't do redundant work, and means everything stays up-to-date. > > Also, I am using "select ... group by ... order by .. limit 1" to get > the min/max since I have already been bit by the issue of min() max() > being slower. > > > -Dan John =:-> --------------enig9CBF65940C3E69561ED666F5 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC+qTmJdeBCYSNAAMRAnFjAKCdP7B1WVO3pZtN1xlvJb3p6nbRVwCffPjY 60SZb50g1pbFtihFJ4AoZog= =nZCQ -----END PGP SIGNATURE----- --------------enig9CBF65940C3E69561ED666F5-- From pgsql-performance-owner@postgresql.org Thu Aug 11 00:15:11 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E299E52935 for ; Thu, 11 Aug 2005 00:15:03 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 15238-08 for ; Thu, 11 Aug 2005 03:14:56 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 7D9D152BE5 for ; Thu, 11 Aug 2005 00:14:53 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7B3Eex3017715; Wed, 10 Aug 2005 23:14:40 -0400 (EDT) To: Tobias Brox Cc: pgsql-performance@postgresql.org Subject: Re: partial index regarded more expensive In-reply-to: <20050810175208.GD6141@tobias.lan> References: <20050810175208.GD6141@tobias.lan> Comments: In-reply-to Tobias Brox message dated "Wed, 10 Aug 2005 19:52:08 +0200" Date: Wed, 10 Aug 2005 23:14:40 -0400 Message-ID: <17714.1123730080@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/105 X-Sequence-Number: 13852 Tobias Brox writes: > This query puzzles me: > select * from game where game_end>'2005-07-30' and state in (3,4); > ... > Now, how can the planner believe the game_by_state-index to be better? I suspect the problem has to do with lack of cross-column statistics. The planner does not know that state=4 is correlated with game_end, and it's probably coming up with some bogus guesses about the numbers of index rows visited in each case. You haven't given enough info to quantify this, though. regards, tom lane From pgsql-performance-owner@postgresql.org Thu Aug 11 00:38:14 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D603452A06 for ; Thu, 11 Aug 2005 00:38:06 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 30677-01 for ; Thu, 11 Aug 2005 03:37:58 +0000 (GMT) Received: from smtp-gw.fnbs.net.my (smtp-gw.fnbs.net.my [202.9.108.191]) by svr1.postgresql.org (Postfix) with ESMTP id E376352A92 for ; Thu, 11 Aug 2005 00:37:54 -0300 (ADT) Received: from mail-std.fnbs.net.my (smtp-std.fnbs.net.my [202.9.108.197]) by fnsrvlx9.fnbs.net.my (SMTP Mailer) with ESMTP id AA95752AD2 for ; Thu, 11 Aug 2005 11:37:51 +0800 (MYT) Received: from Beh (unverified [203.106.54.162]) by mail-std.fnbs.net.my (SurgeMail 3.0c2) with ESMTP id 17318585 for ; Thu, 11 Aug 2005 11:37:50 +0800 MYT Message-ID: <002001c59e26$11bc5bf0$a279640a@Beh> From: "Chun Yit(Chronos)" To: Subject: it is always delete temp table will slow down the postmaster? Date: Thu, 11 Aug 2005 11:37:57 +0800 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_001D_01C59E69.1E7A1880" X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 6.00.2900.2180 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.079 tagged_above=0 required=5 tests=[AWL=-0.007, HTML_40_50=0.086, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/106 X-Sequence-Number: 13853 This is a multi-part message in MIME format. ------=_NextPart_000_001D_01C59E69.1E7A1880 Content-Type: text/plain; charset="gb2312" Content-Transfer-Encoding: quoted-printable > >hi, i got one situation here, i create one pl/pgsql function that using = temp table to store temporary data. >wherever i execute my function, i need to delete all the data inside = the temp table, but this will slow down the=20 >searching function if i conitnue to run the server because old tuples = are not really clear if just using delete command. >so i use drop table command and recreate the table. my question is, = would it slow down the postmaster speed if i continue to=20 >run this searching function more than 300 time per day?, cause the = speed for execute searching function will graduatelly increase=20 >after i test it for few day? anyway to test it is causing by the drop = temp table and create temp table command? > >regards >ivan ------=_NextPart_000_001D_01C59E69.1E7A1880 Content-Type: text/html; charset="gb2312" Content-Transfer-Encoding: quoted-printable
>
>hi, i got one situation here, i = create one=20 pl/pgsql function that using temp table to store temporary = data.
>wherever i execute my function, i = need to=20 delete all the data inside the temp table, but this will slow down the=20
>searching function if i conitnue to = run the=20 server because old tuples are not really clear if just using delete=20 command.
>so i use drop table command and = recreate the=20 table. my question is, would it slow down the postmaster speed if i = continue to=20
>run this searching function more = than 300 time=20 per day?, cause the speed for execute searching function will = graduatelly=20 increase
>after i test it for few day? anyway = to test it=20 is causing by the drop temp table and create temp table = command?
>
>regards
>ivan
------=_NextPart_000_001D_01C59E69.1E7A1880-- From pgsql-performance-owner@postgresql.org Thu Aug 11 00:52:24 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 527DD52A90 for ; Thu, 11 Aug 2005 00:52:23 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 30317-08 for ; Thu, 11 Aug 2005 03:52:18 +0000 (GMT) Received: from linuxworld.com.au (unknown [203.34.46.50]) by svr1.postgresql.org (Postfix) with ESMTP id 9183752A2D for ; Thu, 11 Aug 2005 00:52:16 -0300 (ADT) Received: from linuxworld.com.au (IDENT:swm@localhost.localdomain [127.0.0.1]) by linuxworld.com.au (8.13.2/8.13.2) with ESMTP id j7B3q5KC032246; Thu, 11 Aug 2005 13:52:05 +1000 Received: from localhost (swm@localhost) by linuxworld.com.au (8.13.2/8.13.2/Submit) with ESMTP id j7B3q4i6032243; Thu, 11 Aug 2005 13:52:05 +1000 X-Authentication-Warning: linuxworld.com.au: swm owned process doing -bs Date: Thu, 11 Aug 2005 13:52:04 +1000 (EST) From: Gavin Sherry X-X-Sender: swm@linuxworld.com.au To: Dan Harris Cc: pgsql-performance@postgresql.org Subject: Re: Speedier count(*) In-Reply-To: <3E43C7D5-9DB8-4132-BF4F-607B000866E1@drivefaster.net> Message-ID: References: <3E43C7D5-9DB8-4132-BF4F-607B000866E1@drivefaster.net> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.002 tagged_above=0 required=5 tests=[AWL=0.002] X-Spam-Level: X-Archive-Number: 200508/107 X-Sequence-Number: 13854 Hi Dan, On Wed, 10 Aug 2005, Dan Harris wrote: > I have a web page for my customers that shows them count of records > and some min/max date ranges in each table of a database, as this is > how we bill them for service. They can log in and check the counts > at any time. I'd like for the counts to be as fresh as possible by > keeping this dynamic, but I will use a periodic 'snapshot'/cron job > if that is the only option to speed this up. I have thought about > using the table statistics, but the estimate error is probably > unacceptable because of the billing purposes. > > For some reason, the SQL Server we migrated the app from can return > count(*) in a split second on multi-million row tables, even though > it is a MUCH slower box hardware-wise, but it's now taking many > seconds to run. I have read in the archives the problems MVCC brings > into the count(*) dilemma forcing Pg to run a seq scan to get > counts. Does SQLServer not use MVCC or have they found another SQL Server probably jumps through a lot of hoops to do fast count(*)s. I'm sure we could do something similar -- it's just a question of complexity, resources, desirability, etc. The are other solutions, which makes the idea of doing it less attractive still. > approach for arriving at this number? Compounding all the min/max > and counts from other tables and all those queries take about a > minute to run. The tables will contain anywhere from 1 million to 40 > million rows. > > Also, I am using "select ... group by ... order by .. limit 1" to get > the min/max since I have already been bit by the issue of min() max() > being slower. I generally pre generate the results. There are two ways to do this: the 'snapshot'/cronjon you mentioned or using rules and triggers to maintain 'count' tables. The idea is that if data is added, modified or removed from your table, you modify counters in these other tables. Alternatively, feel free to post your schema and sample queries with explain analyze results to this list. Alternatively, jump on irc at irc.freenode.net #postgresql and someone will be more than happy to look through the problem in more detail. Thanks, Gavin From pgsql-performance-owner@postgresql.org Thu Aug 11 01:40:32 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8A6DC52BA3 for ; Thu, 11 Aug 2005 01:40:31 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 62552-07 for ; Thu, 11 Aug 2005 04:40:28 +0000 (GMT) Received: from smtp012.mail.yahoo.com (smtp012.mail.yahoo.com [216.136.173.32]) by svr1.postgresql.org (Postfix) with SMTP id 8619352B8A for ; Thu, 11 Aug 2005 01:40:26 -0300 (ADT) Received: (qmail 51100 invoked from network); 11 Aug 2005 04:40:26 -0000 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Received:User-Agent:Date:Subject:From:To:CC:Message-ID:In-Reply-To:Mime-version:Content-type:Content-transfer-encoding; b=e6eQgLap2fiU6CVjlsiB7XM+zRLwnOq7Yp2xP5KPhtfQv8SsWP1bQjcwLNv6mZlBoJUhi0gvRSAqAzVQg/5YZVbugHx97sHG5K6P4E5/c1mhSUs1lmgUbnpgKSebH7vjmjFMqifBaHFZuvzStvYe+CAeV8x9hOipzf/PcLV6oKQ= ; Received: from unknown (HELO ?192.168.0.103?) (mcotner@66.56.49.200 with login) by smtp012.mail.yahoo.com with SMTP; 11 Aug 2005 04:40:25 -0000 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Thu, 11 Aug 2005 00:40:23 -0400 Subject: Re: Speedier count(*) From: Mark Cotner To: Gavin Sherry , Dan Harris Cc: Message-ID: In-Reply-To: Mime-version: 1.0 Content-type: text/plain; charset="US-ASCII" Content-transfer-encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.374 tagged_above=0 required=5 tests=[DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/108 X-Sequence-Number: 13855 Here's a trigger I wrote to perform essentially the same purpose. The nice thing about this is it keeps the number up to date for you, but you do incur slight overhead. CREATE TABLE test (id serial primary key, name varchar(20)); CREATE TABLE rowcount (tablename varchar(50), rowcount bigint default 0); CREATE INDEX rowcount_tablename ON rowcount(tablename); CREATE OR REPLACE FUNCTION del_rowcount() RETURNS trigger AS $$ BEGIN UPDATE rowcount SET rowcount = rowcount-1 WHERE tablename = TG_RELNAME; RETURN OLD; END; $$ LANGUAGE PLPGSQL; CREATE OR REPLACE FUNCTION add_rowcount() RETURNS trigger AS $$ BEGIN UPDATE rowcount SET rowcount = rowcount+1 WHERE tablename = TG_RELNAME; RETURN NEW; END; $$ LANGUAGE PLPGSQL; CREATE TRIGGER del_rowcount_tr BEFORE DELETE ON test FOR EACH ROW EXECUTE PROCEDURE del_rowcount(); CREATE TRIGGER add_rowcount_tr BEFORE INSERT ON test FOR EACH ROW EXECUTE PROCEDURE add_rowcount(); INSERT INTO rowcount (tablename) VALUES ('test'); root=# select * from test; id | name ----+------ (0 rows) Time: 0.934 ms root=# select * from rowcount; tablename | rowcount -----------+---------- test | 0 (1 row) Time: 0.630 ms root=# insert into test (name) values ('blah'); INSERT 1190671626 1 Time: 3.278 ms root=# select * from test; id | name ----+------ 5 | blah (1 row) Time: 0.612 ms root=# select * from rowcount; tablename | rowcount -----------+---------- test | 1 (1 row) Time: 0.640 ms root=# insert into test (name) values ('blah'); INSERT 1190671627 1 Time: 1.677 ms root=# select * from test; id | name ----+------ 5 | blah 6 | blah (2 rows) Time: 0.653 ms root=# select * from rowcount; tablename | rowcount -----------+---------- test | 2 (1 row) Time: 0.660 ms root=# delete from test where id = 6; DELETE 1 Time: 2.412 ms root=# select * from test; id | name ----+------ 5 | blah (1 row) Time: 0.631 ms root=# select * from rowcount; tablename | rowcount -----------+---------- test | 1 (1 row) Time: 0.609 ms One thing to be mindful of . . . Truncate is NOT accounted for with this, and unfortunately the rule system doesn't allow truncate operations so you can't work around it that way. 'njoy, Mark On 8/10/05 11:52 PM, "Gavin Sherry" wrote: > Hi Dan, > > On Wed, 10 Aug 2005, Dan Harris wrote: > >> I have a web page for my customers that shows them count of records >> and some min/max date ranges in each table of a database, as this is >> how we bill them for service. They can log in and check the counts >> at any time. I'd like for the counts to be as fresh as possible by >> keeping this dynamic, but I will use a periodic 'snapshot'/cron job >> if that is the only option to speed this up. I have thought about >> using the table statistics, but the estimate error is probably >> unacceptable because of the billing purposes. >> >> For some reason, the SQL Server we migrated the app from can return >> count(*) in a split second on multi-million row tables, even though >> it is a MUCH slower box hardware-wise, but it's now taking many >> seconds to run. I have read in the archives the problems MVCC brings >> into the count(*) dilemma forcing Pg to run a seq scan to get >> counts. Does SQLServer not use MVCC or have they found another > > SQL Server probably jumps through a lot of hoops to do fast count(*)s. I'm > sure we could do something similar -- it's just a question of complexity, > resources, desirability, etc. The are other solutions, which makes the > idea of doing it less attractive still. > >> approach for arriving at this number? Compounding all the min/max >> and counts from other tables and all those queries take about a >> minute to run. The tables will contain anywhere from 1 million to 40 >> million rows. >> >> Also, I am using "select ... group by ... order by .. limit 1" to get >> the min/max since I have already been bit by the issue of min() max() >> being slower. > > I generally pre generate the results. There are two ways to do this: the > 'snapshot'/cronjon you mentioned or using rules and triggers to maintain > 'count' tables. The idea is that if data is added, modified or removed > from your table, you modify counters in these other tables. > > Alternatively, feel free to post your schema and sample queries with > explain analyze results to this list. Alternatively, jump on irc at > irc.freenode.net #postgresql and someone will be more than happy to look > through the problem in more detail. > > Thanks, > > Gavin > > ---------------------------(end of broadcast)--------------------------- > TIP 1: if posting/reading through Usenet, please send an appropriate > subscribe-nomail command to majordomo@postgresql.org so that your > message can get through to the mailing list cleanly From pgsql-performance-owner@postgresql.org Thu Aug 11 03:39:54 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6056E52EB5 for ; Thu, 11 Aug 2005 03:39:53 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 19197-09 for ; Thu, 11 Aug 2005 06:39:49 +0000 (GMT) Received: from floppy.pyrenet.fr (floppy.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id EB75452EB2 for ; Thu, 11 Aug 2005 03:39:48 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id 7C3663095A; Thu, 11 Aug 2005 08:49:47 +0200 (MET DST) From: "Qingqing Zhou" X-Newsgroups: pgsql.performance Subject: Re: it is always delete temp table will slow down the postmaster? Date: Thu, 11 Aug 2005 14:37:20 +0800 Organization: Hub.Org Networking Services Lines: 17 Message-ID: References: <002001c59e26$11bc5bf0$a279640a@Beh> Reply-To: "Qingqing Zhou" X-Complaints-To: usenet@news.hub.org X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 6.00.2800.1506 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1506 To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.185 tagged_above=0 required=5 tests=[AWL=-0.000, PRIORITY_NO_NAME=1.185] X-Spam-Level: * X-Archive-Number: 200508/109 X-Sequence-Number: 13856 ""Chun Yit(Chronos)"" writes > >hi, i got one situation here, i create one pl/pgsql function that using temp table to store temporary data. >wherever i execute my function, i need to delete all the data inside the temp table, but this will slow down the >searching function if i conitnue to run the server because old tuples are not really clear if just using delete command. >so i use drop table command and recreate the table. A better way to empty a table fast is "truncate table". Regards, Qingqing From pgsql-performance-owner@postgresql.org Thu Aug 11 04:03:34 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BB0FF52EAF for ; Thu, 11 Aug 2005 04:03:32 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 88485-01 for ; Thu, 11 Aug 2005 07:03:28 +0000 (GMT) Received: from smtp-gw.fnbs.net.my (smtp-gw.fnbs.net.my [202.9.108.191]) by svr1.postgresql.org (Postfix) with ESMTP id 0199252EAB for ; Thu, 11 Aug 2005 04:03:26 -0300 (ADT) Received: from mail-std.fnbs.net.my (smtp-std.fnbs.net.my [202.9.108.197]) by fnsrvlx9.fnbs.net.my (SMTP Mailer) with ESMTP id 5F8C252B51 for ; Thu, 11 Aug 2005 15:03:25 +0800 (MYT) Received: from Beh (unverified [203.106.54.162]) by mail-std.fnbs.net.my (SurgeMail 3.0c2) with ESMTP id 17330523 for multiple; Thu, 11 Aug 2005 15:03:24 +0800 MYT Message-ID: <001001c59e42$c9e9cca0$a279640a@Beh> From: "Chun Yit(Chronos)" To: "Qingqing Zhou" , References: <002001c59e26$11bc5bf0$a279640a@Beh> Subject: Re: it is always delete temp table will slow down the postmaster? Date: Thu, 11 Aug 2005 15:03:30 +0800 MIME-Version: 1.0 Content-Type: text/plain; format=flowed; charset="iso-8859-1"; reply-type=original Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 6.00.2900.2180 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.038 tagged_above=0 required=5 tests=[AWL=0.038] X-Spam-Level: X-Archive-Number: 200508/110 X-Sequence-Number: 13857 ----- Original Message ----- From: "Qingqing Zhou" To: Sent: Thursday, August 11, 2005 2:37 PM Subject: Re: [PERFORM] it is always delete temp table will slow down the postmaster? > > ""Chun Yit(Chronos)"" writes >> >>hi, i got one situation here, i create one pl/pgsql function that using > temp table to store temporary data. >>wherever i execute my function, i need to delete all the data inside the > temp table, but this will slow down the >>searching function if i conitnue to run the server because old tuples are > not really clear if just using delete command. >>so i use drop table command and recreate the table. > > A better way to empty a table fast is "truncate table". > > Regards, > Qingqing > >sorry, but truncate table cannot use inside function, any other way ? >Regards >ivan > > > ---------------------------(end of broadcast)--------------------------- > TIP 3: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faq > > > -- > No virus found in this incoming message. > Checked by AVG Anti-Virus. > Version: 7.0.338 / Virus Database: 267.10.5/68 - Release Date: 10/Aug/05 > From pgsql-performance-owner@postgresql.org Thu Aug 11 04:24:22 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id DFC6952EB3 for ; Thu, 11 Aug 2005 04:24:20 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 19591-04 for ; Thu, 11 Aug 2005 07:24:17 +0000 (GMT) Received: from service-web.de (p15093784.pureserver.info [217.160.106.224]) by svr1.postgresql.org (Postfix) with ESMTP id 7546752EC0 for ; Thu, 11 Aug 2005 04:24:16 -0300 (ADT) Received: from [192.168.178.99] (p548B2D2C.dip0.t-ipconnect.de [84.139.45.44]) by service-web.de (Postfix) with ESMTP id 0078A200039; Thu, 11 Aug 2005 09:24:16 +0200 (CEST) Subject: Re: Speedier count(*) From: Tino Wildenhain To: Mark Cotner Cc: Gavin Sherry , Dan Harris , pgsql-performance@postgresql.org In-Reply-To: References: Content-Type: text/plain Date: Thu, 11 Aug 2005 09:24:08 +0200 Message-Id: <1123745049.27613.1.camel@Andrea.peacock.de> Mime-Version: 1.0 X-Mailer: Evolution 2.0.4 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.234 tagged_above=0 required=5 tests=[AWL=-0.023, FORGED_RCVD_HELO=0.05, UPPERCASE_25_50=0.207] X-Spam-Level: X-Archive-Number: 200508/111 X-Sequence-Number: 13858 Am Donnerstag, den 11.08.2005, 00:40 -0400 schrieb Mark Cotner: > Here's a trigger I wrote to perform essentially the same purpose. The nice > thing about this is it keeps the number up to date for you, but you do incur > slight overhead. ... > > CREATE TRIGGER del_rowcount_tr BEFORE DELETE ON test FOR EACH ROW EXECUTE > PROCEDURE del_rowcount(); > CREATE TRIGGER add_rowcount_tr BEFORE INSERT ON test FOR EACH ROW EXECUTE > PROCEDURE add_rowcount(); > > INSERT INTO rowcount (tablename) VALUES ('test'); ... beware of problems with concurrency and even what happens if transactions roll back. Maybe you can "fix" it a bit by regulary correcting the count via cronjob or so. From pgsql-performance-owner@postgresql.org Thu Aug 11 07:37:19 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id F040C52C7B for ; Thu, 11 Aug 2005 07:37:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 48389-04 for ; Thu, 11 Aug 2005 10:37:13 +0000 (GMT) Received: from linuxworld.com.au (unknown [203.34.46.50]) by svr1.postgresql.org (Postfix) with ESMTP id 9783D52C57 for ; Thu, 11 Aug 2005 07:37:11 -0300 (ADT) Received: from linuxworld.com.au (IDENT:swm@localhost.localdomain [127.0.0.1]) by linuxworld.com.au (8.13.2/8.13.2) with ESMTP id j7BAaxYu001724; Thu, 11 Aug 2005 20:36:59 +1000 Received: from localhost (swm@localhost) by linuxworld.com.au (8.13.2/8.13.2/Submit) with ESMTP id j7BAawZg001721; Thu, 11 Aug 2005 20:36:58 +1000 X-Authentication-Warning: linuxworld.com.au: swm owned process doing -bs Date: Thu, 11 Aug 2005 20:36:58 +1000 (EST) From: Gavin Sherry X-X-Sender: swm@linuxworld.com.au To: Tino Wildenhain Cc: Mark Cotner , Dan Harris , pgsql-performance@postgresql.org Subject: Re: Speedier count(*) In-Reply-To: <1123745049.27613.1.camel@Andrea.peacock.de> Message-ID: References: <1123745049.27613.1.camel@Andrea.peacock.de> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.002 tagged_above=0 required=5 tests=[AWL=0.002] X-Spam-Level: X-Archive-Number: 200508/112 X-Sequence-Number: 13859 On Thu, 11 Aug 2005, Tino Wildenhain wrote: > Am Donnerstag, den 11.08.2005, 00:40 -0400 schrieb Mark Cotner: > > Here's a trigger I wrote to perform essentially the same purpose. The nice > > thing about this is it keeps the number up to date for you, but you do incur > > slight overhead. > ... > > > > CREATE TRIGGER del_rowcount_tr BEFORE DELETE ON test FOR EACH ROW EXECUTE > > PROCEDURE del_rowcount(); > > CREATE TRIGGER add_rowcount_tr BEFORE INSERT ON test FOR EACH ROW EXECUTE > > PROCEDURE add_rowcount(); > > > > INSERT INTO rowcount (tablename) VALUES ('test'); > ... > > beware of problems with concurrency and even what happens > if transactions roll back. Maybe you can "fix" it a bit > by regulary correcting the count via cronjob or so. What problems? MVCC takes care of this. Gavin From pgsql-performance-owner@postgresql.org Thu Aug 11 07:52:21 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 33C9C52C7B for ; Thu, 11 Aug 2005 07:52:18 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 23319-10 for ; Thu, 11 Aug 2005 10:52:16 +0000 (GMT) Received: from service-web.de (p15093784.pureserver.info [217.160.106.224]) by svr1.postgresql.org (Postfix) with ESMTP id 5F7FA52C57 for ; Thu, 11 Aug 2005 07:52:15 -0300 (ADT) Received: from [192.168.178.99] (p548B2D2C.dip0.t-ipconnect.de [84.139.45.44]) by service-web.de (Postfix) with ESMTP id BB354200039; Thu, 11 Aug 2005 12:52:17 +0200 (CEST) Subject: Re: Speedier count(*) From: Tino Wildenhain To: Gavin Sherry Cc: Mark Cotner , Dan Harris , pgsql-performance@postgresql.org In-Reply-To: References: <1123745049.27613.1.camel@Andrea.peacock.de> Content-Type: text/plain Date: Thu, 11 Aug 2005 12:52:16 +0200 Message-Id: <1123757536.27613.10.camel@Andrea.peacock.de> Mime-Version: 1.0 X-Mailer: Evolution 2.0.4 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.106 tagged_above=0 required=5 tests=[AWL=0.056, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/113 X-Sequence-Number: 13860 Am Donnerstag, den 11.08.2005, 20:36 +1000 schrieb Gavin Sherry: > On Thu, 11 Aug 2005, Tino Wildenhain wrote: > > > Am Donnerstag, den 11.08.2005, 00:40 -0400 schrieb Mark Cotner: > > > Here's a trigger I wrote to perform essentially the same purpose. The nice > > > thing about this is it keeps the number up to date for you, but you do incur > > > slight overhead. > > ... > > > > > > CREATE TRIGGER del_rowcount_tr BEFORE DELETE ON test FOR EACH ROW EXECUTE > > > PROCEDURE del_rowcount(); > > > CREATE TRIGGER add_rowcount_tr BEFORE INSERT ON test FOR EACH ROW EXECUTE > > > PROCEDURE add_rowcount(); > > > > > > INSERT INTO rowcount (tablename) VALUES ('test'); > > ... > > > > beware of problems with concurrency and even what happens > > if transactions roll back. Maybe you can "fix" it a bit > > by regulary correcting the count via cronjob or so. > > What problems? MVCC takes care of this. Actually in this case MVCC works against you. Just imagine some competing transactions to insert end delete at will. You could lock the count table to prevent the problem where 2 competing transactions do an insert, read the start value and add 1 to it and then write the result - which is n+1 rather then n+2 - so you are off by one. Think of the same when one transaction inserts 100 and the other 120. Then you could even be off by 100. But locking probably gets your worser performance then simply count(*) all the time if you insert a lot. Also prepare for the likeness of deadlocks. From pgsql-performance-owner@postgresql.org Thu Aug 11 08:33:49 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0762B52822 for ; Thu, 11 Aug 2005 08:33:47 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49088-10 for ; Thu, 11 Aug 2005 11:33:43 +0000 (GMT) Received: from www.almabioinfo.com (www.almabioinfo.com [62.14.232.115]) by svr1.postgresql.org (Postfix) with ESMTP id AFF7B52D1F for ; Thu, 11 Aug 2005 08:33:35 -0300 (ADT) Received: from [192.168.1.60] (stargate.almabioinfo.com [62.14.232.114]) by www.almabioinfo.com (8.11.2/8.11.2/SuSE Linux 8.11.1-0.5) with ESMTP id j7BBXc830625 for ; Thu, 11 Aug 2005 13:33:38 +0200 Message-ID: <42FB3791.2060200@almabioinfo.com> Date: Thu, 11 Aug 2005 13:33:37 +0200 From: Luis Cornide Arce User-Agent: Mozilla Thunderbird 1.0 (X11/20041206) X-Accept-Language: es-es, es MIME-Version: 1.0 To: PGSL-PERFORMANCE LIST Subject: Why is not using the index Content-Type: multipart/alternative; boundary="------------020405020706050801010200" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.073 tagged_above=0 required=5 tests=[HTML_60_70=0.027, HTML_MESSAGE=0.001, HTML_TITLE_EMPTY=0.045] X-Spam-Level: X-Archive-Number: 200508/114 X-Sequence-Number: 13861 This is a multi-part message in MIME format. --------------020405020706050801010200 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Hi everyone, I have some problems with a quite long query and the plan postgreSQL is choosing. The query joins 12 tables and in the WHERE clause I use a IN expression with a lot of identifiers (up to 2000). The problem is that the planner is proposing a seq_scan on two tables 2M rows each (internalexpressionprofile and expressionprofile) I have just try this query (after doing a vacuum analyze), in the 'IN' clause there are 1552 identifiers, and the query should return 14K rows. I'm using a PostgreSQL 8.0.2 on a SuSE 8.1 with 1GB of RAM. explain analyze SELECT DISTINCT rset.replicatesetid, tra.value as value, tra.expressionprofileid, rep.*, epg.expprogeneid, con.ordinal FROM expprogene epg JOIN reporter rep ON (epg.reporterid=rep.reporterid), expressionprofile epro, transformedexpressionprofile tra, internalexpressionprofile int, meanvalues mea, replicateset rset, replicateset_condition rsco, condition con, "CLUSTER" clu, clustertree tre, clusteranalysis an WHERE epg.expprogeneid IN (80174,84567,...) AND epg.expprogeneid=epro.expprogeneid AND epro.expressionprofileid=tra.expressionprofileid AND tra.expressionprofileid=int.expressionprofileid AND int.meanvaluesid=mea.meanvaluesid AND mea.replicatesetid=rset.replicatesetid AND rset.replicatesetid=rsco.replicatesetid AND rsco.conditionid=con.conditionid AND tra.clusterid=clu.clusterid AND clu.clustertreeid=tre.clustertreeid AND tre.clustertreeid=an.genetreeid AND an.clusteranalysisid=1 AND con.clusteranalysisid = an.clusteranalysisid ORDER BY epg.expprogeneid, con.ordinal; The plan... ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Unique (cost=129132.53..129132.59 rows=2 width=150) (actual time=12637.224..12676.016 rows=13968 loops=1) -> Sort (cost=129132.53..129132.54 rows=2 width=150) (actual time=12637.217..12646.484 rows=13968 loops=1) Sort Key: epg.expprogeneid, con.ordinal, rset.replicatesetid, tra.value, tra.expressionprofileid, rep.reporterid, rep.name, rep.anotation, rep.otherinfo, rep.incidences -> Nested Loop (cost=62927.42..129132.52 rows=2 width=150) (actual time=7112.942..12586.314 rows=13968 loops=1) Join Filter: ("outer".genetreeid = "inner".clustertreeid) -> Nested Loop (cost=62927.42..127893.86 rows=409 width=162) (actual time=7112.864..11960.324 rows=41904 loops=1) -> Nested Loop (cost=62927.42..125727.31 rows=369 width=154) (actual time=7112.825..11500.645 rows=13968 loops=1) -> Merge Join (cost=3.02..7.70 rows=1 width=12) (actual time=0.057..0.073 rows=1 loops=1) Merge Cond: ("outer".clustertreeid = "inner".genetreeid) -> Index Scan using clustertree_pk on clustertree tre (cost=0.00..4.35 rows=123 width=4) (actual time=0.017..0.024 rows=2 loops=1) -> Sort (cost=3.02..3.03 rows=1 width=8) (actual time=0.028..0.030 rows=1 loops=1) Sort Key: an.genetreeid -> Index Scan using clusteranalysis_pk on clusteranalysis an (cost=0.00..3.01 rows=1 width=8) (actual time=0.015..0.018 rows=1 loops=1) Index Cond: (clusteranalysisid = 1) -> Hash Join (cost=62924.39..125715.53 rows=408 width=150) (actual time=7112.758..11455.797 rows=13968 loops=1) Hash Cond: ("outer".expressionprofileid = "inner".expressionprofileid) -> Hash Join (cost=15413.58..78079.33 rows=24339 width=134) (actual time=1489.347..5721.306 rows=41904 loops=1) Hash Cond: ("outer".expprogeneid = "inner".expprogeneid) -> Seq Scan on expressionprofile epro (cost=0.00..48263.24 rows=2831824 width=8) (actual time=0.039..3097.656 rows=2839676 loops=1) -> Hash (cost=15409.72..15409.72 rows=1546 width=130) (actual time=43.365..43.365 rows=0 loops=1) -> Nested Loop (cost=0.00..15409.72 rows=1546 width=130) (actual time=0.056..40.637 rows=1552 loops=1) -> Index Scan using expprogene_pk, expprogene_pk, [......] on expprogene epg (cost=0.00..10698.83 rows=1546 width=8) (actual time=0.027..15.907 rows=1552 loops=1) Index Cond: ((expprogeneid = 80174) OR (expprogeneid = 84567) OR (expprogeneid = 83608) OR [OR ....]) -> Index Scan using reporter_pkey on reporter rep (cost=0.00..3.03 rows=1 width=126) (actual time=0.009..0.010 rows=1 loops=1552) Index Cond: ("outer".reporterid = rep.reporterid) -> Hash (cost=47403.68..47403.68 rows=42853 width=16) (actual time=5623.174..5623.174 rows=0 loops=1) -> Hash Join (cost=2369.91..47403.68 rows=42853 width=16) (actual time=346.040..5538.571 rows=75816 loops=1) Hash Cond: ("outer".meanvaluesid = "inner".meanvaluesid) -> Seq Scan on internalexpressionprofile "int" (cost=0.00..34506.16 rows=2019816 width=8) (actual time=0.003..2231.427 rows=2019816 loops=1) -> Hash (cost=2262.78..2262.78 rows=42853 width=16) (actual time=345.803..345.803 rows=0 loops=1) -> Nested Loop (cost=17.49..2262.78 rows=42853 width=16) (actual time=1.965..259.363 rows=75816 loops=1) -> Hash Join (cost=17.49..28.42 rows=6 width=16) (actual time=1.881..2.387 rows=9 loops=1) Hash Cond: ("outer".replicatesetid = "inner".replicatesetid) -> Seq Scan on replicateset rset (cost=0.00..9.58 rows=258 width=4) (actual time=0.003..0.295 rows=258 loops=1) -> Hash (cost=17.47..17.47 rows=6 width=12) (actual time=1.575..1.575 rows=0 loops=1) -> Hash Join (cost=3.17..17.47 rows=6 width=12) (actual time=0.315..1.557 rows=9 loops=1) Hash Cond: ("outer".conditionid = "inner".conditionid) -> Seq Scan on replicateset_condition rsco (cost=0.00..10.83 rows=683 width=8) (actual time=0.004..0.688 rows=683 loops=1) -> Hash (cost=3.14..3.14 rows=9 width=12) (actual time=0.059..0.059 rows=0 loops=1) -> Index Scan using clustering_analysis_fk on condition con (cost=0.00..3.14 rows=9 width=12) (actual time=0.019..0.039 rows=9 loops=1) Index Cond: (clusteranalysisid = 1) -> Index Scan using has_meanvalues_fk on meanvalues mea (cost=0.00..264.03 rows=8669 width=8) (actual time=0.027..13.032 rows=8424 loops=9) Index Cond: ("outer".replicatesetid = mea.replicatesetid) -> Index Scan using comes_from_raw_fk on transformedexpressionprofile tra (cost=0.00..5.86 rows=1 width=16) (actual time=0.010..0.018 rows=3 loops=13968) Index Cond: (tra.expressionprofileid = "outer".expressionprofileid) -> Index Scan using _cluster__pk on "CLUSTER" clu (cost=0.00..3.01 rows=1 width=8) (actual time=0.009..0.010 rows=1 loops=41904) Index Cond: ("outer".clusterid = clu.clusterid) Total runtime: 12696.289 ms (48 rows) I tried setting the enable_seq_scan to off and the query's runtime returned by the explain analyze is 4000ms. Why postgre is not using the indexes? What is the real impact of having such a big 'IN' clause? Thanks in advance, Luis Cornide --------------020405020706050801010200 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Hi everyone,

I have some problems with a quite long query and the plan postgreSQL is choosing. The query joins 12 tables and in the WHERE clause I use a IN expression with a lot of identifiers (up to 2000). The problem is that the planner is proposing a seq_scan on two tables 2M rows each (internalexpressionprofile and expressionprofile)

I have just try this query (after doing a vacuum analyze), in the 'IN' clause there are 1552 identifiers, and the query should return 14K rows.
I'm using a PostgreSQL 8.0.2 on a SuSE 8.1 with 1GB of RAM.

explain analyze SELECT DISTINCT rset.replicatesetid, tra.value as value, tra.expressionprofileid, rep.*, epg.expprogeneid,  con.ordinal
FROM expprogene epg JOIN reporter rep ON  (epg.reporterid=rep.reporterid), expressionprofile epro,
transformedexpressionprofile tra, internalexpressionprofile int,
meanvalues mea, replicateset rset, replicateset_condition rsco, condition con,
"CLUSTER" clu, clustertree tre, clusteranalysis an
WHERE epg.expprogeneid IN (80174,84567,...) AND epg.expprogeneid=epro.expprogeneid
AND epro.expressionprofileid=tra.expressionprofileid AND tra.expressionprofileid=int.expressionprofileid
AND int.meanvaluesid=mea.meanvaluesid AND mea.replicatesetid=rset.replicatesetid
AND rset.replicatesetid=rsco.replicatesetid AND rsco.conditionid=con.conditionid
AND tra.clusterid=clu.clusterid AND clu.clustertreeid=tre.clustertreeid AND tre.clustertreeid=an.genetreeid
AND an.clusteranalysisid=1 AND con.clusteranalysisid = an.clusteranalysisid
ORDER BY epg.expprogeneid, con.ordinal;

The plan...

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
 Unique  (cost=129132.53..129132.59 rows=2 width=150) (actual time=12637.224..12676.016 rows=13968 loops=1)
   ->  Sort  (cost=129132.53..129132.54 rows=2 width=150) (actual time=12637.217..12646.484 rows=13968 loops=1)
         Sort Key: epg.expprogeneid, con.ordinal, rset.replicatesetid, tra.value, tra.expressionprofileid, rep.reporterid, rep.name, rep.anotation, rep.otherinfo, rep.incidences
         ->  Nested Loop  (cost=62927.42..129132.52 rows=2 width=150) (actual time=7112.942..12586.314 rows=13968 loops=1)
               Join Filter: ("outer".genetreeid = "inner".clustertreeid)
               ->  Nested Loop  (cost=62927.42..127893.86 rows=409 width=162) (actual time=7112.864..11960.324 rows=41904 loops=1)
                     ->  Nested Loop  (cost=62927.42..125727.31 rows=369 width=154) (actual time=7112.825..11500.645 rows=13968 loops=1)
                           ->  Merge Join  (cost=3.02..7.70 rows=1 width=12) (actual time=0.057..0.073 rows=1 loops=1)
                                 Merge Cond: ("outer".clustertreeid = "inner".genetreeid)
                                 ->  Index Scan using clustertree_pk on clustertree tre  (cost=0.00..4.35 rows=123 width=4) (actual time=0.017..0.024 rows=2 loops=1)
                                 ->  Sort  (cost=3.02..3.03 rows=1 width=8) (actual time=0.028..0.030 rows=1 loops=1)
                                       Sort Key: an.genetreeid
                                       ->  Index Scan using clusteranalysis_pk on clusteranalysis an  (cost=0.00..3.01 rows=1 width=8) (actual time=0.015..0.018 rows=1 loops=1)
                                             Index Cond: (clusteranalysisid = 1)
                           ->  Hash Join  (cost=62924.39..125715.53 rows=408 width=150) (actual time=7112.758..11455.797 rows=13968 loops=1)
                                 Hash Cond: ("outer".expressionprofileid = "inner".expressionprofileid)
                                 ->  Hash Join  (cost=15413.58..78079.33 rows=24339 width=134) (actual time=1489.347..5721.306 rows=41904 loops=1)
                                       Hash Cond: ("outer".expprogeneid = "inner".expprogeneid)
                                       ->  Seq Scan on expressionprofile epro  (cost=0.00..48263.24 rows=2831824 width=8) (actual time=0.039..3097.656 rows=2839676 loops=1)
                                       ->  Hash  (cost=15409.72..15409.72 rows=1546 width=130) (actual time=43.365..43.365 rows=0 loops=1)
                                             ->  Nested Loop  (cost=0.00..15409.72 rows=1546 width=130) (actual time=0.056..40.637 rows=1552 loops=1)
                                                   ->  Index Scan using expprogene_pk, expprogene_pk, [......] on expprogene epg  (cost=0.00..10698.83 rows=1546 width=8) (actual time=0.027..15.907 rows=1552 loops=1)
                                                         Index Cond: ((expprogeneid = 80174) OR (expprogeneid = 84567) OR (expprogeneid = 83608) OR [OR ....])
                                                   ->  Index Scan using reporter_pkey on reporter rep  (cost=0.00..3.03 rows=1 width=126) (actual time=0.009..0.010 rows=1 loops=1552)
                                                         Index Cond: ("outer".reporterid = rep.reporterid)
                                 ->  Hash  (cost=47403.68..47403.68 rows=42853 width=16) (actual time=5623.174..5623.174 rows=0 loops=1)
                                       ->  Hash Join  (cost=2369.91..47403.68 rows=42853 width=16) (actual time=346.040..5538.571 rows=75816 loops=1)
                                             Hash Cond: ("outer".meanvaluesid = "inner".meanvaluesid)
                                             ->  Seq Scan on internalexpressionprofile "int"  (cost=0.00..34506.16 rows=2019816 width=8) (actual time=0.003..2231.427 rows=2019816 loops=1)
                                             ->  Hash  (cost=2262.78..2262.78 rows=42853 width=16) (actual time=345.803..345.803 rows=0 loops=1)
                                                   ->  Nested Loop  (cost=17.49..2262.78 rows=42853 width=16) (actual time=1.965..259.363 rows=75816 loops=1)
                                                         ->  Hash Join  (cost=17.49..28.42 rows=6 width=16) (actual time=1.881..2.387 rows=9 loops=1)
                                                               Hash Cond: ("outer".replicatesetid = "inner".replicatesetid)
                                                               ->  Seq Scan on replicateset rset  (cost=0.00..9.58 rows=258 width=4) (actual time=0.003..0.295 rows=258 loops=1)
                                                               ->  Hash  (cost=17.47..17.47 rows=6 width=12) (actual time=1.575..1.575 rows=0 loops=1)
                                                                     ->  Hash Join  (cost=3.17..17.47 rows=6 width=12) (actual time=0.315..1.557 rows=9 loops=1)
                                                                           Hash Cond: ("outer".conditionid = "inner".conditionid)
                                                                           ->  Seq Scan on replicateset_condition rsco  (cost=0.00..10.83 rows=683 width=8) (actual time=0.004..0.688 rows=683 loops=1)
                                                                           ->  Hash  (cost=3.14..3.14 rows=9 width=12) (actual time=0.059..0.059 rows=0 loops=1)
                                                                                 ->  Index Scan using clustering_analysis_fk on condition con  (cost=0.00..3.14 rows=9 width=12) (actual time=0.019..0.039 rows=9 loops=1)
                                                                                       Index Cond: (clusteranalysisid = 1)
                                                         ->  Index Scan using has_meanvalues_fk on meanvalues mea  (cost=0.00..264.03 rows=8669 width=8) (actual time=0.027..13.032 rows=8424 loops=9)
                                                               Index Cond: ("outer".replicatesetid = mea.replicatesetid)
                     ->  Index Scan using comes_from_raw_fk on transformedexpressionprofile tra  (cost=0.00..5.86 rows=1 width=16) (actual time=0.010..0.018 rows=3 loops=13968)
                           Index Cond: (tra.expressionprofileid = "outer".expressionprofileid)
               ->  Index Scan using _cluster__pk on "CLUSTER" clu  (cost=0.00..3.01 rows=1 width=8) (actual time=0.009..0.010 rows=1 loops=41904)
                     Index Cond: ("outer".clusterid = clu.clusterid)
 Total runtime: 12696.289 ms
(48 rows)

I tried setting the enable_seq_scan to off and the query's runtime returned by the explain analyze is 4000ms.
Why postgre is not using the indexes?
What is the real impact of having such a big 'IN' clause?


Thanks in advance,

Luis Cornide
 --------------020405020706050801010200-- From pgsql-performance-owner@postgresql.org Thu Aug 11 09:08:38 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1E3935299C for ; Thu, 11 Aug 2005 09:08:37 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 36300-02 for ; Thu, 11 Aug 2005 12:08:34 +0000 (GMT) Received: from boutiquenumerique.com (boutiquenumerique.com [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 04E1052983 for ; Thu, 11 Aug 2005 09:08:32 -0300 (ADT) Received: (qmail 18274 invoked from network); 11 Aug 2005 14:08:50 +0200 Received: from unknown (HELO localhost) (boutiquenumerique-lists@192.168.0.4) by boutiquenumerique.com with SMTP; 11 Aug 2005 14:08:50 +0200 To: "Tino Wildenhain" , "Gavin Sherry" Cc: "Mark Cotner" , "Dan Harris" , pgsql-performance@postgresql.org Subject: Re: Speedier count(*) References: <1123745049.27613.1.camel@Andrea.peacock.de> <1123757536.27613.10.camel@Andrea.peacock.de> Message-ID: Date: Thu, 11 Aug 2005 14:08:34 +0200 From: PFC Organization: =?iso-8859-15?Q?La_Boutique_Num=E9rique?= Content-Type: text/plain; format=flowed; delsp=yes; charset=iso-8859-15 MIME-Version: 1.0 Content-Transfer-Encoding: 8bit In-Reply-To: <1123757536.27613.10.camel@Andrea.peacock.de> User-Agent: Opera M2/8.0 (Linux, build 1095) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 tagged_above=0 required=5 tests=[AWL=0.005] X-Spam-Level: X-Archive-Number: 200508/115 X-Sequence-Number: 13862 > You could lock the count table to prevent the problem > where 2 competing transactions do an insert, read the > start value and add 1 to it and then write the result > - which is n+1 rather then n+2 - so you are off by one. > Think of the same when one transaction inserts 100 > and the other 120. Then you could even be off by 100. Niet. If your trigger does UPDATE counts_cache SET cached_count = cached_count+N WHERE ... Then all locking is taken care of by Postgres. Of course if you use 2 queries then you have locking issues. However the UPDATE counts_cache has a problem, ie. it locks this row FOR UPDATE for the whole transaction, and all transactions which want to update the same row must wait to see if the update commits or rollbacks, so if you have one count cache row for the whole table you get MySQL style scalability... To preserve scalability you could, instead of UPDATE, INSERT the delta of rows inserted/deleted in a table (which has no concurrencies issues) and compute the current count with the sum() of the deltas, then with a cron, consolidate the deltas and update the count_cache table so that the deltas table stays very small. From pgsql-performance-owner@postgresql.org Thu Aug 11 09:23:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BCF1752A19 for ; Thu, 11 Aug 2005 09:23:24 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 74763-06 for ; Thu, 11 Aug 2005 12:23:20 +0000 (GMT) Received: from lon-mail-2.gradwell.net (lon-mail-2.gradwell.net [193.111.201.126]) by svr1.postgresql.org (Postfix) with ESMTP id 5166052A06 for ; Thu, 11 Aug 2005 09:23:19 -0300 (ADT) Received: from www.gradwell.com ([193.111.200.100]) by lon-mail-2.gradwell.net with smtp (Gradwell gwh-smtpd 1.190) id 42fb4339.14d23.f; Thu, 11 Aug 2005 13:23:21 +0100 (envelope-sender ) Received: from 217.45.209.171 (SquirrelMail authenticated user paul@pop3.oxton.com) by www.gradwell.com with HTTP; Thu, 11 Aug 2005 13:23:21 +0100 (BST) Message-ID: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> Date: Thu, 11 Aug 2005 13:23:21 +0100 (BST) Subject: PG8 Tuning From: "Paul Johnson" To: pgsql-performance@postgresql.org Reply-To: paul@oxton.com User-Agent: SquirrelMail/1.4.2 MIME-Version: 1.0 Content-Type: text/plain;charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Priority: 3 Importance: Normal X-Virus-Scanned: by amavisd-new at hub.org X-Archive-Number: 200508/116 X-Sequence-Number: 13863 Hi all, we're running PG8 on a Sun V250 with 8GB RAM and 2*1.3GHz SPARC CPUs running Solaris 10. The DB cluster is on an external fibre-attached Sun T3 array that has 9*36GB drives configured as a single RAID5 LUN. The system is for the sole use of a couple of data warehouse developers, hence we are keen to use 'aggressive' tuning options to maximise performance. So far we have made the following changes and measured the impact on our test suite: 1) Increase checkpoint_segments from 3 to 64. This made a 10x improvement in some cases. 2) Increase work_mem from 1,024 to 524,288. 3) Increase shared_buffers from 1,000 to 262,143 (2 GB). This required setting SHMMAX=4294967295 (4 GB) in /etc/system and re-booting the box. Question - can Postgres only use 2GB RAM, given that shared_buffers can only be set as high as 262,143 (8K pages)? So far so good... 4) Move /pg_xlog to an internal disk within the V250. This has had a severe *negative* impact on performance. Copy job has gone from 2 mins to 12 mins, simple SQL job gone from 1 min to 7 mins. Not even run long SQL jobs. I'm guessing that this is because pg_xlog has gone from a 9 spindle LUN to a single spindle disk? In cases such as this, where an external storage array with a hardware RAID controller is used, the normal advice to separate the data from the pg_xlog seems to come unstuck, or are we missing something? Cheers, Paul Johnson. From pgsql-performance-owner@postgresql.org Thu Aug 11 09:34:05 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0F30C52869 for ; Thu, 11 Aug 2005 09:34:04 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 19364-06 for ; Thu, 11 Aug 2005 12:34:02 +0000 (GMT) Received: from service-web.de (p15093784.pureserver.info [217.160.106.224]) by svr1.postgresql.org (Postfix) with ESMTP id 14F125285A for ; Thu, 11 Aug 2005 09:33:59 -0300 (ADT) Received: from [192.168.178.99] (p548B2D2C.dip0.t-ipconnect.de [84.139.45.44]) by service-web.de (Postfix) with ESMTP id 479EC200039; Thu, 11 Aug 2005 14:34:02 +0200 (CEST) Subject: Re: Speedier count(*) From: Tino Wildenhain To: PFC Cc: Gavin Sherry , Mark Cotner , Dan Harris , pgsql-performance@postgresql.org In-Reply-To: References: <1123745049.27613.1.camel@Andrea.peacock.de> <1123757536.27613.10.camel@Andrea.peacock.de> Content-Type: text/plain Date: Thu, 11 Aug 2005 14:34:00 +0200 Message-Id: <1123763640.27613.18.camel@Andrea.peacock.de> Mime-Version: 1.0 X-Mailer: Evolution 2.0.4 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.098 tagged_above=0 required=5 tests=[AWL=0.048, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/117 X-Sequence-Number: 13864 Am Donnerstag, den 11.08.2005, 14:08 +0200 schrieb PFC: > > > You could lock the count table to prevent the problem > > where 2 competing transactions do an insert, read the > > start value and add 1 to it and then write the result > > - which is n+1 rather then n+2 - so you are off by one. > > Think of the same when one transaction inserts 100 > > and the other 120. Then you could even be off by 100. > > Niet. > > If your trigger does UPDATE counts_cache SET cached_count = > cached_count+N WHERE ... > Then all locking is taken care of by Postgres. > Of course if you use 2 queries then you have locking issues. Yes, in the case you use just the UPDATE statement you are right. This does the locking I was talking about. In either case I'd use an after trigger and not before to minimize the impact. > However the UPDATE counts_cache has a problem, ie. it locks this row FOR > UPDATE for the whole transaction, and all transactions which want to > update the same row must wait to see if the update commits or rollbacks, > so if you have one count cache row for the whole table you get MySQL style > scalability... > > To preserve scalability you could, instead of UPDATE, INSERT the delta of > rows inserted/deleted in a table (which has no concurrencies issues) and > compute the current count with the sum() of the deltas, then with a cron, > consolidate the deltas and update the count_cache table so that the deltas > table stays very small. Yes, this is in fact a better approach to this problem. (All this provided you want an unqualified count() - as the original poster) From pgsql-performance-owner@postgresql.org Thu Aug 11 10:01:24 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CDA3552A06 for ; Thu, 11 Aug 2005 10:01:22 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 12761-05 for ; Thu, 11 Aug 2005 13:01:18 +0000 (GMT) Received: from mail.metronet.co.uk (mail.metronet.co.uk [213.162.97.75]) by svr1.postgresql.org (Postfix) with ESMTP id 52F1E529F9 for ; Thu, 11 Aug 2005 10:01:18 -0300 (ADT) Received: from mainbox.archonet.com (84-51-143-99.archon037.adsl.metronet.co.uk [84.51.143.99]) by smtp.metronet.co.uk (MetroNet Mail) with ESMTP id D3DE040C066; Thu, 11 Aug 2005 14:01:08 +0100 (BST) Received: from localhost (localhost.localdomain [127.0.0.1]) by mainbox.archonet.com (Postfix) with ESMTP id 799B915EDA; Thu, 11 Aug 2005 13:49:15 +0100 (BST) Received: from mainbox.archonet.com ([127.0.0.1]) by localhost (mainbox [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 30982-04; Thu, 11 Aug 2005 13:49:11 +0100 (BST) Received: from [192.168.1.17] (client17.office.archonet.com [192.168.1.17]) by mainbox.archonet.com (Postfix) with ESMTP id 5B67615ED9; Thu, 11 Aug 2005 13:49:11 +0100 (BST) Message-ID: <42FB4946.6030105@archonet.com> Date: Thu, 11 Aug 2005 13:49:10 +0100 From: Richard Huxton User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Luis Cornide Arce Cc: PGSL-PERFORMANCE LIST Subject: Re: Why is not using the index References: <42FB3791.2060200@almabioinfo.com> In-Reply-To: <42FB3791.2060200@almabioinfo.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 tagged_above=0 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/119 X-Sequence-Number: 13866 Luis Cornide Arce wrote: > Hi everyone, > > I have some problems with a quite long query and the plan postgreSQL is > choosing. The query joins 12 tables and in the WHERE clause I use a IN > expression with a lot of identifiers (up to 2000). The problem is that > the planner is proposing a seq_scan on two tables 2M rows each > (internalexpressionprofile and expressionprofile) > > I have just try this query (after doing a vacuum analyze), in the 'IN' > clause there are 1552 identifiers, and the query should return 14K rows. > I'm using a PostgreSQL 8.0.2 on a SuSE 8.1 with 1GB of RAM. > WHERE epg.expprogeneid IN (80174,84567,...) AND > epg.expprogeneid=epro.expprogeneid -> Hash Join (cost=15413.58..78079.33 rows=24339 width=134) (actual time=1489.347..5721.306 rows=41904 loops=1) Hash Cond: ("outer".expprogeneid = "inner".expprogeneid) -> Seq Scan on expressionprofile epro (cost=0.00..48263.24 rows=2831824 width=8) (actual time=0.039..3097.656 rows=2839676 loops=1) -> Index Scan using expprogene_pk, expprogene_pk, [......] on expprogene epg (cost=0.00..10698.83 rows=1546 width=8) (actual time=0.027..15.907 rows=1552 loops=1) Index Cond: ((expprogeneid = 80174) OR (expprogeneid = 84567) OR (expprogeneid = 83608) OR [OR ....]) OK - it looks like the "IN" clause is using your index. The fact that it's using a Seq-scan on "expressionprofile epro" looks odd though, especially since it expects 24339 matches (out of 2.8 million rows - that should favour an index). Of course, I've not considered the context of the rest of the query, but I'd expect the index to be used. Do you have any unusual config settings? -- Richard Huxton Archonet Ltd From pgsql-performance-owner@postgresql.org Thu Aug 11 10:01:18 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6D21C52997 for ; Thu, 11 Aug 2005 10:01:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 06248-08 for ; Thu, 11 Aug 2005 13:01:15 +0000 (GMT) Received: from mail.metronet.co.uk (mail.metronet.co.uk [213.162.97.75]) by svr1.postgresql.org (Postfix) with ESMTP id 30E01529F5 for ; Thu, 11 Aug 2005 10:01:14 -0300 (ADT) Received: from mainbox.archonet.com (84-51-143-99.archon037.adsl.metronet.co.uk [84.51.143.99]) by smtp.metronet.co.uk (MetroNet Mail) with ESMTP id CC416414567; Thu, 11 Aug 2005 14:01:08 +0100 (BST) Received: from localhost (localhost.localdomain [127.0.0.1]) by mainbox.archonet.com (Postfix) with ESMTP id E9AA115EDB; Thu, 11 Aug 2005 13:55:28 +0100 (BST) Received: from mainbox.archonet.com ([127.0.0.1]) by localhost (mainbox [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 01250-01; Thu, 11 Aug 2005 13:55:24 +0100 (BST) Received: from [192.168.1.17] (client17.office.archonet.com [192.168.1.17]) by mainbox.archonet.com (Postfix) with ESMTP id 58A0D15ED9; Thu, 11 Aug 2005 13:55:24 +0100 (BST) Message-ID: <42FB4ABB.1010305@archonet.com> Date: Thu, 11 Aug 2005 13:55:23 +0100 From: Richard Huxton User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: paul@oxton.com Cc: pgsql-performance@postgresql.org Subject: Re: PG8 Tuning References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> In-Reply-To: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 tagged_above=0 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/118 X-Sequence-Number: 13865 Paul Johnson wrote: > Hi all, we're running PG8 on a Sun V250 with 8GB RAM and 2*1.3GHz SPARC > CPUs running Solaris 10. The DB cluster is on an external fibre-attached > Sun T3 array that has 9*36GB drives configured as a single RAID5 LUN. > > The system is for the sole use of a couple of data warehouse developers, > hence we are keen to use 'aggressive' tuning options to maximise > performance. > > So far we have made the following changes and measured the impact on our > test suite: > > 1) Increase checkpoint_segments from 3 to 64. This made a 10x improvement > in some cases. OK > 2) Increase work_mem from 1,024 to 524,288. Don't forget you can use multiples of this in a single query. Might want to reign it back a bit. I *think* you can set it per-query if you want anyway. > 3) Increase shared_buffers from 1,000 to 262,143 (2 GB). This required > setting SHMMAX=4294967295 (4 GB) in /etc/system and re-booting the box. > > Question - can Postgres only use 2GB RAM, given that shared_buffers can > only be set as high as 262,143 (8K pages)? Well, normally you'd want to keep a fair bit for the O.S. to cache data. One quarter of your RAM seems very high. Did you try 5000,10000,50000 too or go straight to the top end? > So far so good... > > 4) Move /pg_xlog to an internal disk within the V250. This has had a > severe *negative* impact on performance. Copy job has gone from 2 mins to > 12 mins, simple SQL job gone from 1 min to 7 mins. Not even run long SQL > jobs. > > I'm guessing that this is because pg_xlog has gone from a 9 spindle LUN to > a single spindle disk? The key limitation will be one commit per rotation of the disk. Multiple spindles, or better still with a battery-backed write-cache will give you peak transactions. > In cases such as this, where an external storage array with a hardware > RAID controller is used, the normal advice to separate the data from the > pg_xlog seems to come unstuck, or are we missing something? Well, I think the advice then is actually "get 2 external arrays..." -- Richard Huxton Archonet Ltd From pgsql-performance-owner@postgresql.org Sun Aug 14 19:56:04 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9124752A97 for ; Thu, 11 Aug 2005 10:19:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 08161-01 for ; Thu, 11 Aug 2005 13:19:10 +0000 (GMT) Received: from smtp4.wanadoo.fr (smtp4.wanadoo.fr [193.252.22.27]) by svr1.postgresql.org (Postfix) with ESMTP id 114A952869 for ; Thu, 11 Aug 2005 10:19:08 -0300 (ADT) Received: from me-wanadoo.net (unknown [127.0.0.1]) by mwinf0401.wanadoo.fr (SMTP Server) with ESMTP id B5DCF1C002A5 for ; Thu, 11 Aug 2005 15:19:11 +0200 (CEST) Received: from NEO (unknown [86.193.99.195]) by mwinf0401.wanadoo.fr (SMTP Server) with ESMTP id 3175A1C0031B for ; Thu, 11 Aug 2005 15:19:11 +0200 (CEST) X-ME-UUID: 20050811131911202.3175A1C0031B@mwinf0401.wanadoo.fr From: =?iso-8859-1?Q?St=E9phane_COEZ?= To: Subject: Performance pb vs SQLServer. Date: Thu, 11 Aug 2005 15:19:06 +0200 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable X-Mailer: Microsoft Office Outlook, Build 11.0.5510 Thread-Index: AcWedz241y3YYG5NRWucxBfjaAmDhQ== X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 Message-Id: <20050811131911.3175A1C0031B@mwinf0401.wanadoo.fr> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/145 X-Sequence-Number: 13892 Hi, I have a perfomance issue : I run PG (8.0.3) and SQLServer2000 on a Windows2000 Server (P4 1,5Ghz = 512Mo) I have a table (3200000 rows) and I run this single query : select cod from mytable group by cod I have an index on cod (char(4) - 88 different values) PG =3D ~ 20 sec. SQLServer =3D < 8 sec the explain is : HashAggregate (cost=3D64410.09..64410.09 rows=3D55 width=3D8) -> Seq Scan on mytable (cost=3D0.00..56325.27 rows=3D3233927 = width=3D8) if I switch to "enable_hashagg =3D false" (just for a try...) the planner will choose my index : Group (cost=3D0.00..76514.01 rows=3D55 width=3D8) -> Index Scan using myindex on mytable (cost=3D0.00..68429.20 = rows=3D3233927 width=3D8) but performance will be comparable to previous test. So with or without using Index I have the same result. Thanks for help. =20 St=E9phane COEZ From pgsql-performance-owner@postgresql.org Thu Aug 11 10:26:39 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 439E152AC9 for ; Thu, 11 Aug 2005 10:26:37 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 98738-07 for ; Thu, 11 Aug 2005 13:26:35 +0000 (GMT) Received: from mailhost.intellivid.com (mailhost.intellivid.com [64.32.200.11]) by svr1.postgresql.org (Postfix) with ESMTP id 6312552A76 for ; Thu, 11 Aug 2005 10:26:34 -0300 (ADT) Received: from [192.168.2.68] (spectre.intellivid.com [192.168.2.68]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client did not present a certificate) by newmail.intellivid.com (Postfix) with ESMTP id 9C213F182B1; Thu, 11 Aug 2005 09:26:37 -0400 (EDT) Subject: Re: Planner doesn't look at LIMIT? From: Ian Westmacott To: Tom Lane Cc: pgsql-performance@postgresql.org In-Reply-To: <16040.1123714524@sss.pgh.pa.us> References: <758d5e7f0507220210bfa4978@mail.gmail.com> <29473.1122043197@sss.pgh.pa.us> <6440.1122049220@sss.pgh.pa.us> <1123707833.17725.133.camel@spectre.intellivid.com> <16040.1123714524@sss.pgh.pa.us> Content-Type: text/plain Organization: Intellivid Corp. Message-Id: <1123766797.21162.93.camel@spectre.intellivid.com> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Thu, 11 Aug 2005 09:26:37 -0400 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/121 X-Sequence-Number: 13868 On Wed, 2005-08-10 at 18:55, Tom Lane wrote: > Ian Westmacott writes: > > In a nutshell, I have a LIMIT query where the planner > > seems to favor a merge join over a nested loop. > > The planner is already estimating only one row out of the join, and so > the LIMIT doesn't affect its cost estimates at all. > > It appears to me that the reason the nestloop plan is fast is just > chance: a suitable matching row is found very early in the scan of > tableB, so that the indexscan on it can stop after 29 rows, instead > of having to go through all 55000 rows in the given range of bim. > If it'd have had to go through, say, half of the rows to find a match, > the sort/merge plan would show up a lot better. Oh, I see. Thanks, that clears up some misconceptions I had about the explain output. > If this wasn't chance, but was expected because there are many matching > rows and not only one, then there's a statistical problem. Well, there are in fact almost 300 of them in this case. So I guess what I need to do is give the planner more information to correctly predict that. Thanks, --Ian From pgsql-performance-owner@postgresql.org Thu Aug 11 10:59:40 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BDB92529D1 for ; Thu, 11 Aug 2005 10:59:38 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 24144-04 for ; Thu, 11 Aug 2005 13:59:35 +0000 (GMT) Received: from www.almabioinfo.com (www.almabioinfo.com [62.14.232.115]) by svr1.postgresql.org (Postfix) with ESMTP id 8BC3152986 for ; Thu, 11 Aug 2005 10:59:30 -0300 (ADT) Received: from [192.168.1.60] (stargate.almabioinfo.com [62.14.232.114]) by www.almabioinfo.com (8.11.2/8.11.2/SuSE Linux 8.11.1-0.5) with ESMTP id j7BDxW831898 for ; Thu, 11 Aug 2005 15:59:32 +0200 Message-ID: <42FB59C3.5020809@almabioinfo.com> Date: Thu, 11 Aug 2005 15:59:31 +0200 From: Luis Cornide Arce User-Agent: Mozilla Thunderbird 1.0 (X11/20041206) X-Accept-Language: es-es, es MIME-Version: 1.0 Cc: PGSL-PERFORMANCE LIST Subject: Re: Why is not using the index References: <42FB3791.2060200@almabioinfo.com> <42FB4946.6030105@archonet.com> In-Reply-To: <42FB4946.6030105@archonet.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.036 tagged_above=0 required=5 tests=[AWL=0.036] X-Spam-Level: X-Archive-Number: 200508/122 X-Sequence-Number: 13869 Well I have change the next setting in the postgresql.conf shared_buffers= 16384 work_mem =32768 maintenance_work_mem= 65536 bgwriter_delay =800 bgwriter_maxpages= 100 wal_buffers =64 efective_cache_size= 20000 The rest of the settings are the default. Thanks, Luis Richard Huxton escribi�: > Luis Cornide Arce wrote: > >> Hi everyone, >> >> I have some problems with a quite long query and the plan postgreSQL >> is choosing. The query joins 12 tables and in the WHERE clause I use >> a IN expression with a lot of identifiers (up to 2000). The problem >> is that the planner is proposing a seq_scan on two tables 2M rows >> each (internalexpressionprofile and expressionprofile) >> >> I have just try this query (after doing a vacuum analyze), in the >> 'IN' clause there are 1552 identifiers, and the query should return >> 14K rows. >> I'm using a PostgreSQL 8.0.2 on a SuSE 8.1 with 1GB of RAM. > > >> WHERE epg.expprogeneid IN (80174,84567,...) AND >> epg.expprogeneid=epro.expprogeneid > > > -> Hash Join > (cost=15413.58..78079.33 rows=24339 width=134) > (actual time=1489.347..5721.306 rows=41904 loops=1) > Hash Cond: ("outer".expprogeneid = "inner".expprogeneid) > -> Seq Scan on expressionprofile epro > (cost=0.00..48263.24 rows=2831824 width=8) > (actual time=0.039..3097.656 rows=2839676 loops=1) > > -> Index Scan using > expprogene_pk, expprogene_pk, [......] on expprogene epg > (cost=0.00..10698.83 rows=1546 width=8) (actual time=0.027..15.907 > rows=1552 loops=1) > Index Cond: ((expprogeneid = 80174) OR (expprogeneid = 84567) > OR (expprogeneid = 83608) OR [OR ....]) > > OK - it looks like the "IN" clause is using your index. The fact that > it's using a Seq-scan on "expressionprofile epro" looks odd though, > especially since it expects 24339 matches (out of 2.8 million rows - > that should favour an index). > > Of course, I've not considered the context of the rest of the query, > but I'd expect the index to be used. > > Do you have any unusual config settings? From pgsql-performance-owner@postgresql.org Thu Aug 11 12:02:02 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9A8C5529D1 for ; Thu, 11 Aug 2005 11:09:49 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 29124-08 for ; Thu, 11 Aug 2005 14:09:46 +0000 (GMT) Received: from vms042pub.verizon.net (vms042pub.verizon.net [206.46.252.42]) by svr1.postgresql.org (Postfix) with ESMTP id AD59652C12 for ; Thu, 11 Aug 2005 11:09:41 -0300 (ADT) Received: from osgiliath.mathom.us ([70.108.53.154]) by vms042.mailsrvcs.net (Sun Java System Messaging Server 6.2 HotFix 0.04 (built Dec 24 2004)) with ESMTPA id <0IL200HM59VT7C57@vms042.mailsrvcs.net> for pgsql-performance@postgresql.org; Thu, 11 Aug 2005 09:07:06 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by osgiliath.mathom.us (Postfix) with ESMTP id 2087D70882D for ; Thu, 11 Aug 2005 10:07:05 -0400 (EDT) Received: from osgiliath.mathom.us ([127.0.0.1]) by localhost (osgiliath [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 01836-01 for ; Thu, 11 Aug 2005 10:07:04 -0400 (EDT) Received: by osgiliath.mathom.us (Postfix, from userid 1000) id E470B702202; Thu, 11 Aug 2005 10:07:04 -0400 (EDT) Date: Thu, 11 Aug 2005 10:07:04 -0400 From: Michael Stone Subject: Re: PG8 Tuning In-reply-to: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> To: pgsql-performance@postgresql.org Mail-Followup-To: pgsql-performance@postgresql.org Message-id: <20050811140704.GL19080@mathom.us> MIME-version: 1.0 Content-type: text/plain; charset=us-ascii; format=flowed Content-disposition: inline X-Pgp-Fingerprint: 53 FF 38 00 E7 DD 0A 9C 84 52 84 C5 EE DF 7C 88 X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mathom.us References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.007 tagged_above=0 required=5 tests=[AWL=0.007] X-Spam-Level: X-Archive-Number: 200508/123 X-Sequence-Number: 13870 On Thu, Aug 11, 2005 at 01:23:21PM +0100, Paul Johnson wrote: >I'm guessing that this is because pg_xlog has gone from a 9 spindle LUN to >a single spindle disk? > >In cases such as this, where an external storage array with a hardware >RAID controller is used, the normal advice to separate the data from the >pg_xlog seems to come unstuck Yes. That's the downside to dogma. If you're writing pg_xlog to a battery-backed ram buffer you'll see faster commits than you will with a write to a disk, even if you've got a dedicated spindle, unless you've got constant write activity. (Because once the buffer fills you're limited to disk speed as you wait for buffer flushes.) If you've got a lot of system RAM, a battery-backed disk buffer, an OS/filesystem than effectively delays writes, and bursty transactional writes it's quite possible you'll get better performance putting everything on one array rather than breaking it up to follow the "rules". You might get a performance boost by putting the transaction log on a seperate partition or lun on the external array, depending on how the fs implements syncs or whether you can optimize the filsystem choice for each partition. The correct approach is to run comparative benchmarks of each configuration. :-) Mike Stone From pgsql-performance-owner@postgresql.org Sun Aug 14 19:55:08 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5E1B252AC0 for ; Thu, 11 Aug 2005 11:30:21 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49682-02 for ; Thu, 11 Aug 2005 14:30:14 +0000 (GMT) Received: from brmea-mail-4.sun.com (brmea-mail-4.Sun.COM [192.18.98.36]) by svr1.postgresql.org (Postfix) with ESMTP id 5D6D9528D7 for ; Thu, 11 Aug 2005 11:30:13 -0300 (ADT) Received: from phys-bur-1 ([129.148.9.72]) by brmea-mail-4.sun.com (8.12.10/8.12.9) with ESMTP id j7BEUDx9010848 for ; Thu, 11 Aug 2005 08:30:14 -0600 (MDT) Received: from conversion-daemon.bur-mail1.east.sun.com by bur-mail1.east.sun.com (iPlanet Messaging Server 5.2 HotFix 1.24 (built Dec 19 2003)) id <0IL200801ATJW4@bur-mail1.east.sun.com> (original mail from J.K.Shah@Sun.COM) for pgsql-performance@postgresql.org; Thu, 11 Aug 2005 10:30:13 -0400 (EDT) Received: from bur-mail1.east.sun.com (phys-bur-1 [129.148.9.72]) by bur-mail1.east.sun.com (iPlanet Messaging Server 5.2 HotFix 1.24 (built Dec 19 2003)) with ESMTP id <0IL200AOPAYDEM@bur-mail1.east.sun.com>; Thu, 11 Aug 2005 10:30:13 -0400 (EDT) Received: from [129.148.168.2] by bur-mail1.east.sun.com (mshttpd); Thu, 11 Aug 2005 10:30:13 -0400 Date: Thu, 11 Aug 2005 10:30:13 -0400 From: Jignesh Shah Subject: Re: [Fwd: PG8 Tuning] To: paul@oxton.com, pgsql-performance@postgresql.org Cc: Robert Lor , Donald Courtney Message-id: <7833474d78.74d7878334@bur-mail1.east.sun.com> MIME-version: 1.0 X-Mailer: iPlanet Messenger Express 5.2 HotFix 1.24 (built Dec 19 2003) Content-type: multipart/mixed; boundary="Boundary_(ID_EpePzI99TlR0Dh/+EexxvA)" Content-language: en X-Accept-Language: en X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/144 X-Sequence-Number: 13891 This is a multi-part message in MIME format. --Boundary_(ID_EpePzI99TlR0Dh/+EexxvA) Content-type: text/plain; charset=us-ascii Content-transfer-encoding: 7BIT Content-disposition: inline Hi Paul, I was passed your message... regarding DSS workload with Postgres on Solaris. (I am not in the alias). Performance is relative to your workload. Can you actually send us what you are doing in your queries, updates etc? I have been running few tests myself and here are my rules of thumbs, your mileage can vary.. http://blogs.sun.com/roller/page/jkshah?entry=tuning_postgresql_8_0_2 * Increasing checkpoint certainly helps. (I went as far as actually going to increase LOGFILE size from 16MB to 256MB and recompiling it and then using lower number of checkpoints (appropriately).. (file rotations also decreases performance) * Moving pg_xlog to a different file system and mounting that file system with "forcedirectio" also helps a lot (This increases the througput by another 2x to 5x or more.) (This can be done either by adding forcedirectio in your /etc/vfstab mount options or for existing mounts as follows: mount -o remount,forcedirectio /filesystem (Note: Database files should not be using forcedirectio otherwise file system cache will not be used for it) * I actually reduced the PG Bufferpool to 1G or less since it seemed to decrease performance as I increased its bufferpool size (depending on your workload) * If you are using SPARC then following etc commands will help.. set segmap_percent=60 set ufs:freebehind=0 This will allocate 60% of RAM for file system buffer (database files) and also cache all files (since PostgreSQL files are 1G by default) This will help your repeat queries significantly. Other things depends on what you queries you are running? If you send me few samples, I can send you appropriate DTrace scripts (Solaris 10 or higher) to run to figure out what's happening Regards, Jignesh ____________________________________________________ Jignesh K. Shah MTS Software Engineer Sun Microsystems, Inc MDE-Horizontal Technologies Email: J.K.Shah@sun.com Phone: (781) 442 3052 http://blogs.sun.com/jkshah ____________________________________________________ ----- Original Message ----- >From Paul Johnson Date Thu, 11 Aug 2005 13:23:21 +0100 (BST) To pgsql-performance@postgresql.org Subject [PERFORM] PG8 Tuning Hi all, we're running PG8 on a Sun V250 with 8GB RAM and 2*1.3GHz SPARC CPUs running Solaris 10. The DB cluster is on an external fibre-attached Sun T3 array that has 9*36GB drives configured as a single RAID5 LUN. The system is for the sole use of a couple of data warehouse developers, hence we are keen to use 'aggressive' tuning options to maximise performance. So far we have made the following changes and measured the impact on our test suite: 1) Increase checkpoint_segments from 3 to 64. This made a 10x improvement in some cases. 2) Increase work_mem from 1,024 to 524,288. 3) Increase shared_buffers from 1,000 to 262,143 (2 GB). This required setting SHMMAX=4294967295 (4 GB) in /etc/system and re-booting the box. Question - can Postgres only use 2GB RAM, given that shared_buffers can only be set as high as 262,143 (8K pages)? So far so good... 4) Move /pg_xlog to an internal disk within the V250. This has had a severe *negative* impact on performance. Copy job has gone from 2 mins to 12 mins, simple SQL job gone from 1 min to 7 mins. Not even run long SQL jobs. I'm guessing that this is because pg_xlog has gone from a 9 spindle LUN to a single spindle disk? In cases such as this, where an external storage array with a hardware RAID controller is used, the normal advice to separate the data from the pg_xlog seems to come unstuck, or are we missing something? Cheers, Paul Johnson. --Boundary_(ID_EpePzI99TlR0Dh/+EexxvA) Content-type: message/rfc822; name="[PERFORM] PG8 Tuning" Return-path: Received: from conversion-daemon.bur-mail1.east.sun.com by bur-mail1.east.sun.com (iPlanet Messaging Server 5.2 HotFix 1.24 (built Dec 19 2003)) id <0IL200C015ME3X@bur-mail1.east.sun.com> (original mail from pgsql-performance-owner+M13863@postgresql.org) for djc@bur-mail1.East.Sun.COM; Thu, 11 Aug 2005 08:38:53 -0400 (EDT) Received: from eastmail2bur.East.Sun.COM (eastmail2bur.East.Sun.COM [129.148.13.40]) by bur-mail1.east.sun.com (iPlanet Messaging Server 5.2 HotFix 1.24 (built Dec 19 2003)) with ESMTP id <0IL200MH15STJS@bur-mail1.east.sun.com> for djc@bur-mail1.East.Sun.COM (ORCPT donald.courtney@sun.com); Thu, 11 Aug 2005 08:38:53 -0400 (EDT) Received: from sunmail1brm.Central.Sun.COM (sunmail1brm.Central.Sun.COM [129.147.62.17]) by eastmail2bur.East.Sun.COM (8.12.10+Sun/8.12.10/ENSMAIL,v2.2) with ESMTP id j7BCcrlM025431 for ; Thu, 11 Aug 2005 08:38:53 -0400 (EDT) Received: from nwkea-mail-2.sun.com (nwkea-mail-2.Sun.COM [192.18.42.14]) by sunmail1brm.Central.Sun.COM (8.11.7p1+Sun/8.11.7/ENSMAIL,v2.2) with ESMTP id j7BCcpb07056 for ; Thu, 11 Aug 2005 06:38:52 -0600 (MDT) Received: from relay23.sun.com (relay23.sun.com [192.12.251.54] (may be forged)) by nwkea-mail-2.sun.com (8.12.10/8.12.9) with ESMTP id j7BCcp1d027727 for ; Thu, 11 Aug 2005 05:38:51 -0700 (PDT) Received: from mms21es.sun.com (mms21es.sun.com [150.143.232.14]) by relay23.sun.com with ESMTP for donald.courtney@sun.com; Thu, 11 Aug 2005 12:38:51 +0000 (Z) Received: from relay21.sun.com (relay21.sun.com [192.12.251.14]) by mms21es.sun.com with ESMTP for donald.courtney@sun.com; Thu, 11 Aug 2005 12:38:50 +0000 (Z) Received: from hosting.commandprompt.com ([207.173.200.128] [207.173.200.128]) by relay21.sun.com with ESMTP for donald.courtney@sun.com; Thu, 11 Aug 2005 12:38:50 +0000 (Z) Received: from postgresql.org (svr1.postgresql.org [200.46.204.71]) by hosting.commandprompt.com (8.13.4/8.13.4) with ESMTP id j7BCPn0Q014340; Thu, 11 Aug 2005 05:26:04 -0700 Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BCF1752A19 for ; Thu, 11 Aug 2005 09:23:24 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 74763-06 for ; Thu, 11 Aug 2005 12:23:20 +0000 (GMT) Received: from lon-mail-2.gradwell.net (lon-mail-2.gradwell.net [193.111.201.126]) by svr1.postgresql.org (Postfix) with ESMTP id 5166052A06 for ; Thu, 11 Aug 2005 09:23:19 -0300 (ADT) Received: from www.gradwell.com ([193.111.200.100]) by lon-mail-2.gradwell.net with smtp (Gradwell gwh-smtpd 1.190) id 42fb4339.14d23.f; Thu, 11 Aug 2005 13:23:21 +0100 (envelope-sender ) Received: from 217.45.209.171 (SquirrelMail authenticated user paul@pop3.oxton.com) by www.gradwell.com with HTTP; Thu, 11 Aug 2005 13:23:21 +0100 (BST) Date: Thu, 11 Aug 2005 13:23:21 +0100 (BST) From: Paul Johnson Subject: [PERFORM] PG8 Tuning Sender: pgsql-performance-owner@postgresql.org To: pgsql-performance@postgresql.org Reply-to: paul@oxton.com Message-id: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> MIME-version: 1.0 Content-type: text/plain; charset=iso-8859-1 Content-transfer-encoding: 8BIT Importance: Normal X-Priority: 3 Precedence: bulk User-Agent: SquirrelMail/1.4.2 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-1.6 (hosting.commandprompt.com [192.168.1.101]); Thu, 11 Aug 2005 05:26:08 -0700 (PDT) X-Mailing-List: pgsql-performance List-Owner: List-Post: List-Subscribe: List-Unsubscribe: List-Archive: List-Help: List-Id: Original-recipient: rfc822;donald.courtney@sun.com Hi all, we're running PG8 on a Sun V250 with 8GB RAM and 2*1.3GHz SPARC CPUs running Solaris 10. The DB cluster is on an external fibre-attached Sun T3 array that has 9*36GB drives configured as a single RAID5 LUN. The system is for the sole use of a couple of data warehouse developers, hence we are keen to use 'aggressive' tuning options to maximise performance. So far we have made the following changes and measured the impact on our test suite: 1) Increase checkpoint_segments from 3 to 64. This made a 10x improvement in some cases. 2) Increase work_mem from 1,024 to 524,288. 3) Increase shared_buffers from 1,000 to 262,143 (2 GB). This required setting SHMMAX=4294967295 (4 GB) in /etc/system and re-booting the box. Question - can Postgres only use 2GB RAM, given that shared_buffers can only be set as high as 262,143 (8K pages)? So far so good... 4) Move /pg_xlog to an internal disk within the V250. This has had a severe *negative* impact on performance. Copy job has gone from 2 mins to 12 mins, simple SQL job gone from 1 min to 7 mins. Not even run long SQL jobs. I'm guessing that this is because pg_xlog has gone from a 9 spindle LUN to a single spindle disk? In cases such as this, where an external storage array with a hardware RAID controller is used, the normal advice to separate the data from the pg_xlog seems to come unstuck, or are we missing something? Cheers, Paul Johnson. ---------------------------(end of broadcast)--------------------------- TIP 4: Have you searched our list archives? http://archives.postgresql.org --Boundary_(ID_EpePzI99TlR0Dh/+EexxvA)-- From pgsql-performance-owner@postgresql.org Thu Aug 11 12:28:16 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 19B0752E95 for ; Thu, 11 Aug 2005 12:25:33 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 95179-02 for ; Thu, 11 Aug 2005 15:25:31 +0000 (GMT) Received: from unicorn.rentec.com (unicorn.rentec.com [216.223.240.9]) by svr1.postgresql.org (Postfix) with ESMTP id E412652E8A for ; Thu, 11 Aug 2005 12:25:29 -0300 (ADT) Received: from wren.rentec.com (wren.rentec.com [192.5.35.106]) by unicorn.rentec.com (8.13.1/8.12.1) with ESMTP id j7BFPRfE020212 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO) for ; Thu, 11 Aug 2005 11:25:28 -0400 (EDT) X-Rentec: external Received: from [172.26.132.145] (hoopoe.rentec.com [172.26.132.145]) by wren.rentec.com (8.13.1/8.12.1) with ESMTP id j7BFPRto023192 for ; Thu, 11 Aug 2005 11:25:27 -0400 (EDT) Message-ID: <42FB6DE7.7060107@rentec.com> Date: Thu, 11 Aug 2005 11:25:27 -0400 From: Alan Stange Reply-To: stange@rentec.com Organization: Renaissance Technologies Corp. User-Agent: Mozilla Thunderbird 1.0+ (X11/20050712) MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: BG writer question? Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Logged: Logged by unicorn.rentec.com as j7BFPRfE020212 at Thu Aug 11 11:25:28 2005 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/124 X-Sequence-Number: 13871 Hello all, I just was running strace in the writer process and I noticed this pattern: select(0, NULL, NULL, NULL, {0, 200000}) = 0 (Timeout) getppid() = 4240 time(NULL) = 1123773324 mmap2(NULL, 528384, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0x81000) = 0x69ea3000 semop(1409034, 0xffffc0bc, 1) = 0 <...seeks and writes...> munmap(0x69ea3000, 528384) = 0 select(0, NULL, NULL, NULL, {0, 200000}) = 0 (Timeout) getppid() = 4240 time(NULL) = 1123773324 mmap2(NULL, 528384, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0x81000) = 0x69ea3000 semop(1605648, 0xffffc0bc, 1) = 0 <...seeks and writes...> munmap(0x69ea3000, 528384) = 0 select(0, NULL, NULL, NULL, {0, 200000}) = 0 (Timeout) why mmap and munmap each time? mmap and munmap are fairly expensive operations (on some systems), especially on multi cpu machines. munmap in particular generally needs to issue cross calls to the other cpus to ensure any page mappings are invalidated. Just curious. Thanks! -- Alan From pgsql-performance-owner@postgresql.org Thu Aug 11 12:59:51 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A5EA152E94 for ; Thu, 11 Aug 2005 12:30:32 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 96349-07 for ; Thu, 11 Aug 2005 15:30:29 +0000 (GMT) Received: from crestone.coronasolutions.com (crestone.coronasolutions.com [66.45.104.24]) by svr1.postgresql.org (Postfix) with ESMTP id 8A7B152E72 for ; Thu, 11 Aug 2005 12:30:28 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by crestone.coronasolutions.com (Postfix) with ESMTP id E55426440E6 for ; Thu, 11 Aug 2005 09:29:55 -0600 (MDT) Received: from crestone.coronasolutions.com ([127.0.0.1]) by localhost (crestone.coronasolutions.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 19615-05 for ; Thu, 11 Aug 2005 09:29:54 -0600 (MDT) Received: from [192.168.1.12] (c-24-9-24-35.hsd1.co.comcast.net [24.9.24.35]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (No client certificate requested) by crestone.coronasolutions.com (Postfix) with ESMTP id F03C16440B8 for ; Thu, 11 Aug 2005 09:29:53 -0600 (MDT) Mime-Version: 1.0 (Apple Message framework v730) In-Reply-To: <1123763640.27613.18.camel@Andrea.peacock.de> References: <1123745049.27613.1.camel@Andrea.peacock.de> <1123757536.27613.10.camel@Andrea.peacock.de> <1123763640.27613.18.camel@Andrea.peacock.de> Content-Type: text/plain; charset=US-ASCII; format=flowed Message-Id: <40E2616B-1BCB-4B1D-9B00-9FE52DEBD85B@drivefaster.net> Content-Transfer-Encoding: 7bit From: Dan Harris Subject: Re: Speedier count(*) Date: Thu, 11 Aug 2005 09:33:00 -0600 To: pgsql-performance@postgresql.org X-Mailer: Apple Mail (2.730) X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at drivefaster.net X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/125 X-Sequence-Number: 13872 Thanks for all the great ideas. I have more options to evaluate now. -Dan From pgsql-performance-owner@postgresql.org Thu Aug 11 13:43:35 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9D9B452FE3 for ; Thu, 11 Aug 2005 13:42:51 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 37945-07 for ; Thu, 11 Aug 2005 16:42:49 +0000 (GMT) Received: from alvh.no-ip.org (200-85-218-206.bk4-dsl.surnet.cl [200.85.218.206]) by svr1.postgresql.org (Postfix) with ESMTP id 5373052FF4 for ; Thu, 11 Aug 2005 13:42:48 -0300 (ADT) Received: by alvh.no-ip.org (Postfix, from userid 1000) id B32CCC2D450; Thu, 11 Aug 2005 12:43:05 -0400 (CLT) Date: Thu, 11 Aug 2005 12:43:05 -0400 From: Alvaro Herrera To: Alan Stange Cc: pgsql-performance@postgresql.org Subject: Re: BG writer question? Message-ID: <20050811164305.GB20172@alvh.no-ip.org> Mail-Followup-To: Alan Stange , pgsql-performance@postgresql.org References: <42FB6DE7.7060107@rentec.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <42FB6DE7.7060107@rentec.com> User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.789 tagged_above=0 required=5 tests=[AWL=0.365, DNS_FROM_RFC_ABUSE=0.374, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/126 X-Sequence-Number: 13873 On Thu, Aug 11, 2005 at 11:25:27AM -0400, Alan Stange wrote: > why mmap and munmap each time? mmap and munmap are fairly expensive > operations (on some systems), especially on multi cpu machines. munmap > in particular generally needs to issue cross calls to the other cpus to > ensure any page mappings are invalidated. There are no mmap/munmap calls in our code. The problematic code is probably somewhere in the libc. Maybe it'd be useful to figure out where it's called and why, with an eye on working around that. -- Alvaro Herrera () "I love the Postgres community. It's all about doing things _properly_. :-)" (David Garamond) From pgsql-performance-owner@postgresql.org Thu Aug 11 14:04:57 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AC9CF52C50 for ; Thu, 11 Aug 2005 14:01:38 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 58442-06 for ; Thu, 11 Aug 2005 16:58:55 +0000 (GMT) Received: from gghcwest.com (adsl-71-128-90-172.dsl.pltn13.pacbell.net [71.128.90.172]) by svr1.postgresql.org (Postfix) with ESMTP id 5412252936 for ; Thu, 11 Aug 2005 13:58:53 -0300 (ADT) Received: from toonses.gghcwest.com (toonses.gghcwest.com [192.168.168.115]) by gghcwest.com (8.12.10/8.12.9) with ESMTP id j7BGwhmd016814; Thu, 11 Aug 2005 09:58:44 -0700 Received: from jwb by toonses.gghcwest.com with local (Exim 4.50) id 1E3GOC-0001l7-Df; Thu, 11 Aug 2005 09:58:44 -0700 Subject: Re: [SPAM?] Re: PG8 Tuning From: "Jeffrey W. Baker" To: Steve Poe Cc: paul@oxton.com, pgsql-performance@postgresql.org In-Reply-To: <1123836428.19976.26.camel@amd64-laptop-spoe> References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> <1123836428.19976.26.camel@amd64-laptop-spoe> Content-Type: text/plain Content-Transfer-Encoding: 7bit Date: Thu, 11 Aug 2005 09:58:44 -0700 Message-Id: <1123779524.6664.1.camel@toonses.gghcwest.com> Mime-Version: 1.0 X-Mailer: Evolution 2.3.6.1 X-Virus-Scanned: by amavisd-new at hub.org X-Archive-Number: 200508/127 X-Sequence-Number: 13874 On Fri, 2005-08-12 at 08:47 +0000, Steve Poe wrote: > Paul, > > Before I say anything else, one online document which may be of > assistance to you is: > http://www.powerpostgresql.com/PerfList/ > > Some thoughts I have: > > 3) You're shared RAM setting seems overkill to me. Part of the challenge > is you're going from 1000 to 262K with no assessment in between. Each > situation can be different, but try in the range of 10 - 50K. > > 4) pg_xlog: If you're pg_xlog is on a spindle is *only* for pg_xlog > you're better off. Like Mr. Stone said earlier, this is pure dogma. In my experience, xlogs on the same volume with data is much faster if both are on battery-backed write-back RAID controller memory. Moving from this situation to xlogs on a single normal disk is going to be much slower in most cases. -jwb From pgsql-performance-owner@postgresql.org Thu Aug 11 16:04:57 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0C6BD5303E for ; Thu, 11 Aug 2005 14:20:18 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 68973-01 for ; Thu, 11 Aug 2005 17:20:14 +0000 (GMT) Received: from mir3-fs.mir3.com (mail.mir3.com [65.208.188.100]) by svr1.postgresql.org (Postfix) with ESMTP id 581965301E for ; Thu, 11 Aug 2005 14:20:11 -0300 (ADT) Received: from archimedes ([172.16.2.68]) by mir3-fs.mir3.com with Microsoft SMTPSVC(5.0.2195.6713); Thu, 11 Aug 2005 10:21:42 -0700 Subject: Re: PG8 Tuning From: Mark Lewis To: Steve Poe Cc: paul@oxton.com, pgsql-performance@postgresql.org In-Reply-To: <1123836428.19976.26.camel@amd64-laptop-spoe> References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> <1123836428.19976.26.camel@amd64-laptop-spoe> Content-Type: text/plain Organization: MIR3, Inc. Date: Thu, 11 Aug 2005 10:18:44 -0700 Message-Id: <1123780724.14573.108.camel@archimedes> Mime-Version: 1.0 X-Mailer: Evolution 2.0.2 (2.0.2-16) Content-Transfer-Encoding: 7bit X-OriginalArrivalTime: 11 Aug 2005 17:21:42.0945 (UTC) FILETIME=[247F8910:01C59E99] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/129 X-Sequence-Number: 13876 (Musing, trying to think of a general-purpose performance-tuning rule that applies here): Actually, it seems to me that with the addition of the WAL in PostgreSQL and the subsequent decreased need to fsync the data files themselves (only during checkpoints?), that the only time a battery-backed write cache would make a really large performance difference would be on the drive(s) hosting the WAL. So although it is in general good to have a dedicated spindle for the WAL, for many workloads it is in fact significantly better to have the WAL written to a battery-backed write cache. The exception would be for applications with fewer, larger transactions, in which case you could actually use the dedicated spindle. Hmmm, on second thought, now I think I understand the rationale behind having a non-zero commit delay setting-- the problem with putting pg_xlog on a single disk without a write cache is that frequent fsync() calls might cause it to spend most of its time seeking instead of writing (as seems to be happening to Paul here). Then again, the OS IO scheduler should take care of this for you, making this a non-issue. Perhaps Solaris 10 just has really poor IO scheduling performance with this particular hardware and workload? Ah well. Thought myself in circles and have no real conclusions to show for it. Posting anyway, maybe this will give somebody some ideas to work with. -- Mark Lewis On Fri, 2005-08-12 at 08:47 +0000, Steve Poe wrote: > Paul, > > Before I say anything else, one online document which may be of > assistance to you is: > http://www.powerpostgresql.com/PerfList/ > > Some thoughts I have: > > 3) You're shared RAM setting seems overkill to me. Part of the challenge > is you're going from 1000 to 262K with no assessment in between. Each > situation can be different, but try in the range of 10 - 50K. > > 4) pg_xlog: If you're pg_xlog is on a spindle is *only* for pg_xlog > you're better off. If it is sharing with any other OS/DB resource, the > performance will be impacted. > > >From what I have learned from others on this list, RAID5 is not the best > choice for the database. RAID10 would be a better solution (using 8 of > your disks) then take the remaining disk and do mirror with your pg_xlog > if possible. > > Best of luck, > > Steve Poe > > On Thu, 2005-08-11 at 13:23 +0100, Paul Johnson wrote: > > Hi all, we're running PG8 on a Sun V250 with 8GB RAM and 2*1.3GHz SPARC > > CPUs running Solaris 10. The DB cluster is on an external fibre-attached > > Sun T3 array that has 9*36GB drives configured as a single RAID5 LUN. > > > > The system is for the sole use of a couple of data warehouse developers, > > hence we are keen to use 'aggressive' tuning options to maximise > > performance. > > > > So far we have made the following changes and measured the impact on our > > test suite: > > > > 1) Increase checkpoint_segments from 3 to 64. This made a 10x improvement > > in some cases. > > > > 2) Increase work_mem from 1,024 to 524,288. > > > > 3) Increase shared_buffers from 1,000 to 262,143 (2 GB). This required > > setting SHMMAX=4294967295 (4 GB) in /etc/system and re-booting the box. > > > > Question - can Postgres only use 2GB RAM, given that shared_buffers can > > only be set as high as 262,143 (8K pages)? > > > > So far so good... > > > > 4) Move /pg_xlog to an internal disk within the V250. This has had a > > severe *negative* impact on performance. Copy job has gone from 2 mins to > > 12 mins, simple SQL job gone from 1 min to 7 mins. Not even run long SQL > > jobs. > > > > I'm guessing that this is because pg_xlog has gone from a 9 spindle LUN to > > a single spindle disk? > > > > In cases such as this, where an external storage array with a hardware > > RAID controller is used, the normal advice to separate the data from the > > pg_xlog seems to come unstuck, or are we missing something? > > > > Cheers, > > > > Paul Johnson. > > > > ---------------------------(end of broadcast)--------------------------- > > TIP 4: Have you searched our list archives? > > > > http://archives.postgresql.org > > > ---------------------------(end of broadcast)--------------------------- > TIP 3: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faq From pgsql-performance-owner@postgresql.org Thu Aug 11 15:21:09 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0EB2952F01 for ; Thu, 11 Aug 2005 15:00:30 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 64033-08 for ; Thu, 11 Aug 2005 18:00:26 +0000 (GMT) Received: from mail.jobflash.com (mail.jobflash.com [64.62.211.41]) by svr1.postgresql.org (Postfix) with ESMTP id D0D4052E9C for ; Thu, 11 Aug 2005 15:00:25 -0300 (ADT) Received: from [192.168.2.22] (user-38lc0mu.dialup.mindspring.com [209.86.2.222]) by mail.jobflash.com (Postfix) with ESMTP id 689DC420C1; Thu, 11 Aug 2005 11:00:23 -0700 (PDT) Message-ID: <42FB9235.4090505@jobflash.com> Date: Thu, 11 Aug 2005 11:00:21 -0700 From: Tom Arthurs User-Agent: Mozilla Thunderbird 1.0.2 (Windows/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Steve Poe Cc: paul@oxton.com, pgsql-performance@postgresql.org Subject: Re: [SPAM?] Re: PG8 Tuning References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> <1123836428.19976.26.camel@amd64-laptop-spoe> In-Reply-To: <1123836428.19976.26.camel@amd64-laptop-spoe> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/128 X-Sequence-Number: 13875 I think the T-3 RAID at least breaks some of these rules -- I've got 2 T-3's, 1 configured as RAID-10 and the other as RAID5, and they both seem to perform about the same. I use RAID5 with a hot spare, so it's using 8 spindles. I got a lot of performance improvement out of mount the fs noatime and turning journaling off. Of course it takes a *long* time to recover from a crash. Steve Poe wrote: > Paul, > > Before I say anything else, one online document which may be of > assistance to you is: > http://www.powerpostgresql.com/PerfList/ > > Some thoughts I have: > > 3) You're shared RAM setting seems overkill to me. Part of the challenge > is you're going from 1000 to 262K with no assessment in between. Each > situation can be different, but try in the range of 10 - 50K. > > 4) pg_xlog: If you're pg_xlog is on a spindle is *only* for pg_xlog > you're better off. If it is sharing with any other OS/DB resource, the > performance will be impacted. > >>From what I have learned from others on this list, RAID5 is not the best > choice for the database. RAID10 would be a better solution (using 8 of > your disks) then take the remaining disk and do mirror with your pg_xlog > if possible. > > Best of luck, > > Steve Poe > > On Thu, 2005-08-11 at 13:23 +0100, Paul Johnson wrote: > >>Hi all, we're running PG8 on a Sun V250 with 8GB RAM and 2*1.3GHz SPARC >>CPUs running Solaris 10. The DB cluster is on an external fibre-attached >>Sun T3 array that has 9*36GB drives configured as a single RAID5 LUN. >> >>The system is for the sole use of a couple of data warehouse developers, >>hence we are keen to use 'aggressive' tuning options to maximise >>performance. >> >>So far we have made the following changes and measured the impact on our >>test suite: >> >>1) Increase checkpoint_segments from 3 to 64. This made a 10x improvement >>in some cases. >> >>2) Increase work_mem from 1,024 to 524,288. >> >>3) Increase shared_buffers from 1,000 to 262,143 (2 GB). This required >>setting SHMMAX=4294967295 (4 GB) in /etc/system and re-booting the box. >> >>Question - can Postgres only use 2GB RAM, given that shared_buffers can >>only be set as high as 262,143 (8K pages)? >> >>So far so good... >> >>4) Move /pg_xlog to an internal disk within the V250. This has had a >>severe *negative* impact on performance. Copy job has gone from 2 mins to >>12 mins, simple SQL job gone from 1 min to 7 mins. Not even run long SQL >>jobs. >> >>I'm guessing that this is because pg_xlog has gone from a 9 spindle LUN to >>a single spindle disk? >> >>In cases such as this, where an external storage array with a hardware >>RAID controller is used, the normal advice to separate the data from the >>pg_xlog seems to come unstuck, or are we missing something? >> >>Cheers, >> >>Paul Johnson. >> >>---------------------------(end of broadcast)--------------------------- >>TIP 4: Have you searched our list archives? >> >> http://archives.postgresql.org > > > > ---------------------------(end of broadcast)--------------------------- > TIP 3: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faq > > > From pgsql-performance-owner@postgresql.org Thu Aug 11 17:36:49 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9218752EBF for ; Thu, 11 Aug 2005 17:36:47 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 14547-03 for ; Thu, 11 Aug 2005 20:36:42 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id D4E7352ECD for ; Thu, 11 Aug 2005 17:36:40 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050811203641m9100g6g0be>; Thu, 11 Aug 2005 20:36:41 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id B920E56008; Thu, 11 Aug 2005 15:36:40 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id 003A955FCC for ; Thu, 11 Aug 2005 15:36:35 -0500 (CDT) Message-ID: <42FBB6CF.6000404@arbash-meinel.com> Date: Thu, 11 Aug 2005 15:36:31 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Postgresql Performance Subject: Odd Locking Problem X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigE9A54294F5ECE755761E8BF4" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.044 tagged_above=0 required=5 tests=[AWL=-0.006, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/130 X-Sequence-Number: 13877 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigE9A54294F5ECE755761E8BF4 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit I'm having an odd case where my system is locking such that if I insert into a table during a transaction, if I start a new connection and transaction, it blocks while trying to do a similar insert until the first transaction is committed or rolled back. The schema is rather complex (currently 157 tables, 200 views), and I still haven't been able to create a small test case. Everything I've tried so far just works. The data is private, but the schema is open source, so I probably could work with someone on it. When I look at the pg_locks table, I seem to be blocked on: SELECT * FROM pg_locks WHERE granted = false; relation | database | transaction | pid | mode | granted ----------+----------+-------------+-------+------------------+--------- | | 1525932 | 30175 | ShareLock | f ... Which if I understand correctly, means that the current transaction is intentionally blocking waiting for the other transaction to finish. I'm currently running 8.0.3, but the database was first created under 7.4.? I confirmed this behavior on both systems. Under what circumstances would this occur? To try and outline the situation there is a main object table, which is the root object. It contains a group column which is used for access rights. There is a groupref table, which keeps track of the group rights for each user. (Each user has specific insert,update,select rights per group). The select rights are enforced by views (the tables are not publicly accessible, the views join against the groupref table to check for select permission). Insert and update rights are validated by BEFORE INSERT triggers. Most tables references the object table. Basically it is OO, but doesn't use the postgres inheritance (in our testing postgres inheritance didn't scale well for deep inheritance, and wasn't able to enforce uniqueness anyway.) The views present an OO appearance, and behind the scenes direct table foreign keys maintain referential integrity. I have checked using RAISE NOTICE and the BEFORE INSERT trigger gets all the way to the RETURN statement before things hang, so I haven't figured out what is actually hanging. I have a bzip'd version of the schema and just enough data to be useful available here: http://www.arbash-meinel.com/extras/schema_and_data.sql.bz2 This is the commands to replicate the locking: -- Connect as postgres -- Required before any inserts, so that the TEMP env table is -- created and filled out. select mf_setup_env(); -- Begin a transaction and insert some data BEGIN; INSERT INTO object(vgroup,otype,oname) VALUES ('test',1,'test'); -- Start a new shell, and connect again and do exactly the same thing -- as the above. -- It should hang until you either do END/ROLLBACK in the first -- connection. Thanks for any help, John =:-> --------------enigE9A54294F5ECE755761E8BF4 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC+7bTJdeBCYSNAAMRAoobAKCe4p+4YDnVkl63HS3HQjr6WH3/EgCeKo5Z NbtuOQJIdn3+cWTbBC4glsc= =BvPL -----END PGP SIGNATURE----- --------------enigE9A54294F5ECE755761E8BF4-- From pgsql-performance-owner@postgresql.org Thu Aug 11 17:39:00 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 49BD65286F for ; Thu, 11 Aug 2005 17:38:58 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 05208-06 for ; Thu, 11 Aug 2005 20:38:55 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id 964625284E for ; Thu, 11 Aug 2005 17:38:52 -0300 (ADT) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: PG8 Tuning Date: Thu, 11 Aug 2005 16:38:53 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3417DD095@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] PG8 Tuning Thread-Index: AcWeqVEJJSwD/9NeSQGbxvRSDEigkgACZ18w From: "Merlin Moncure" To: "Mark Lewis" Cc: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 tagged_above=0 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/131 X-Sequence-Number: 13878 > Actually, it seems to me that with the addition of the WAL in PostgreSQL > and the subsequent decreased need to fsync the data files themselves > (only during checkpoints?), that the only time a battery-backed write > cache would make a really large performance difference would be on the > drive(s) hosting the WAL. It still helps. In my experience a good BBU Raid controller is only slightly slower than fsync=3Dfalse. Fear the checkpoint storm if you don't have some write caching. Beyond that I don't really care about write delay. Another thing to watch out for is that some sync modes (varying by platform) can do >1 seeks per sync. This will absolutely kill your commit performance on the WAL without write caching. =20 > So although it is in general good to have a dedicated spindle for the > WAL, for many workloads it is in fact significantly better to have the > WAL written to a battery-backed write cache. The exception would be for > applications with fewer, larger transactions, in which case you could > actually use the dedicated spindle. Exactly. =20 > Hmmm, on second thought, now I think I understand the rationale behind > having a non-zero commit delay setting-- the problem with putting I don't trust commit_delay. Get a good raid controller and make sure pg is properly using it. Now, if you can't (or won't) do some type of write caching bbu or no, your system has to be very carefully designed to get any performance at all, especially with high transaction volumes. Merlin From pgsql-performance-owner@postgresql.org Thu Aug 11 18:08:40 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D9A0E52EE4 for ; Thu, 11 Aug 2005 18:08:30 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 16618-04 for ; Thu, 11 Aug 2005 21:08:24 +0000 (GMT) Received: from alvh.no-ip.org (200-85-218-206.bk4-dsl.surnet.cl [200.85.218.206]) by svr1.postgresql.org (Postfix) with ESMTP id F160F52E8A for ; Thu, 11 Aug 2005 18:08:21 -0300 (ADT) Received: by alvh.no-ip.org (Postfix, from userid 1000) id C46F6C2D450; Thu, 11 Aug 2005 17:08:42 -0400 (CLT) Date: Thu, 11 Aug 2005 17:08:42 -0400 From: Alvaro Herrera To: John A Meinel Cc: Postgresql Performance Subject: Re: Odd Locking Problem Message-ID: <20050811210842.GA28253@alvh.no-ip.org> Mail-Followup-To: John A Meinel , Postgresql Performance References: <42FBB6CF.6000404@arbash-meinel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: <42FBB6CF.6000404@arbash-meinel.com> User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.776 tagged_above=0 required=5 tests=[AWL=0.353, DNS_FROM_RFC_ABUSE=0.374, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/132 X-Sequence-Number: 13879 On Thu, Aug 11, 2005 at 03:36:31PM -0500, John A Meinel wrote: > I'm having an odd case where my system is locking such that if I insert > into a table during a transaction, if I start a new connection and > transaction, it blocks while trying to do a similar insert until the > first transaction is committed or rolled back. Are there foreign keys here? I can duplicate the problem easily with them: -- session 1 create table a (a serial primary key); create table b (a int references a); insert into a values (1); begin; insert into b values (1); -- session 2 insert into b values (1); -- hangs If I commit on session 1, session 2 is unlocked. This is a known problem, solved in 8.1. A workaround for previous releases is to defer FK checks until commit: create table b (a int references a initially deferred); -- Alvaro Herrera () Dios hizo a Ad�n, pero fue Eva quien lo hizo hombre. From pgsql-performance-owner@postgresql.org Thu Aug 11 18:14:28 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BA08052EF5 for ; Thu, 11 Aug 2005 18:12:09 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 85097-09 for ; Thu, 11 Aug 2005 21:12:07 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id 5D5EF52EF4 for ; Thu, 11 Aug 2005 18:12:02 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050811211204m92009ij23e>; Thu, 11 Aug 2005 21:12:04 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 346F256008; Thu, 11 Aug 2005 16:12:04 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id 3C7C655FCC; Thu, 11 Aug 2005 16:11:59 -0500 (CDT) Message-ID: <42FBBF1E.6040403@arbash-meinel.com> Date: Thu, 11 Aug 2005 16:11:58 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Alvaro Herrera Cc: Postgresql Performance Subject: Re: Odd Locking Problem References: <42FBB6CF.6000404@arbash-meinel.com> <20050811210842.GA28253@alvh.no-ip.org> In-Reply-To: <20050811210842.GA28253@alvh.no-ip.org> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig85C8C97A4CA324ACA27CB4AC" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.044 tagged_above=0 required=5 tests=[AWL=-0.006, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/133 X-Sequence-Number: 13880 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig85C8C97A4CA324ACA27CB4AC Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Alvaro Herrera wrote: > On Thu, Aug 11, 2005 at 03:36:31PM -0500, John A Meinel wrote: > >>I'm having an odd case where my system is locking such that if I insert >>into a table during a transaction, if I start a new connection and >>transaction, it blocks while trying to do a similar insert until the >>first transaction is committed or rolled back. > > > Are there foreign keys here? I can duplicate the problem easily with > them: > > -- session 1 > create table a (a serial primary key); > create table b (a int references a); > insert into a values (1); > > begin; > insert into b values (1); > > > -- session 2 > insert into b values (1); > -- hangs > Actually, there are but the insert is occurring into table 'a' not table 'b'. 'a' refers to other tables, but these should not be modified. > > If I commit on session 1, session 2 is unlocked. > > This is a known problem, solved in 8.1. A workaround for previous > releases is to defer FK checks until commit: > > create table b (a int references a initially deferred); I'll try one of the CVS entries and see if it happens there. Good to hear there has been work done. John =:-> > --------------enig85C8C97A4CA324ACA27CB4AC Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC+78eJdeBCYSNAAMRAi8AAJ9N+E4aJXZvG+SadflT5Mml4DpX1ACcC4j1 FLEvQdCoRCup3WTHkgg+dNk= =dHX/ -----END PGP SIGNATURE----- --------------enig85C8C97A4CA324ACA27CB4AC-- From pgsql-performance-owner@postgresql.org Thu Aug 11 19:21:49 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9035452983 for ; Thu, 11 Aug 2005 19:21:46 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 37218-06 for ; Thu, 11 Aug 2005 22:21:42 +0000 (GMT) Received: from rwcrmhc11.comcast.net (rwcrmhc11.comcast.net [204.127.198.35]) by svr1.postgresql.org (Postfix) with ESMTP id E382D5301B for ; Thu, 11 Aug 2005 19:21:41 -0300 (ADT) Received: from [192.168.1.52] (c-24-60-119-214.hsd1.ma.comcast.net[24.60.119.214]) by comcast.net (rwcrmhc11) with ESMTP id <2005081122210201300ohcjhe>; Thu, 11 Aug 2005 22:21:02 +0000 Message-ID: <42FBCF4D.6060002@comcast.net> Date: Thu, 11 Aug 2005 18:21:01 -0400 From: Jeffrey Tenny User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7) Gecko/20040616 X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Mostly read performance Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.544 tagged_above=0 required=5 tests=[AWL=-0.698, DNS_FROM_RFC_ABUSE=0.374, DNS_FROM_RFC_POST=1.376, DNS_FROM_RFC_WHOIS=0.492] X-Spam-Level: * X-Archive-Number: 200508/134 X-Sequence-Number: 13881 I have a largely table-append-only application where most transactions are read-intensive and many are read-only. The transactions may span many tables, and in some cases might need to pull 70 MB of data out of a couple of the larger tables. In 7.3, I don't seem to see any file system or other caching that helps with repeated reads of the 70MB of data. Secondary fetches are pretty much as slow as the first fetch. (The 70MB in this example might take place via 2000 calls to a parameterized statement via JDBC). Were there changes after 7.3 w.r.t. caching of data? I read this list and see people saying that 8.0 will use the native file system cache to good effect. Is this true? Is it supposed to work with 7.3? Is there something I need to do to get postgresql to take advatage of large ram systems? Thanks for any advice. From pgsql-performance-owner@postgresql.org Thu Aug 11 19:39:02 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 55A815284E for ; Thu, 11 Aug 2005 19:39:00 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 89183-05 for ; Thu, 11 Aug 2005 22:38:57 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id 832E952E7E for ; Thu, 11 Aug 2005 19:38:56 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050811223859m92009hurme>; Thu, 11 Aug 2005 22:38:59 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id A959E56008; Thu, 11 Aug 2005 17:38:56 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id 6B1E755FCC; Thu, 11 Aug 2005 17:38:50 -0500 (CDT) Message-ID: <42FBD377.8040006@arbash-meinel.com> Date: Thu, 11 Aug 2005 17:38:47 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Jeffrey Tenny Cc: pgsql-performance@postgresql.org Subject: Re: Mostly read performance References: <42FBCF4D.6060002@comcast.net> In-Reply-To: <42FBCF4D.6060002@comcast.net> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigA0AF2BF7B9B582267E8407E5" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.044 tagged_above=0 required=5 tests=[AWL=-0.006, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/135 X-Sequence-Number: 13882 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigA0AF2BF7B9B582267E8407E5 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Jeffrey Tenny wrote: > I have a largely table-append-only application where most transactions > are read-intensive and many are read-only. The transactions may span > many tables, and in some cases might need to pull 70 MB of data out of a > couple of the larger tables. > > > In 7.3, I don't seem to see any file system or other caching that helps > with repeated reads of the 70MB of data. Secondary fetches are pretty > much as slow as the first fetch. (The 70MB in this example might take > place via 2000 calls to a parameterized statement via JDBC). > > Were there changes after 7.3 w.r.t. caching of data? I read this list > and see people saying that 8.0 will use the native file system cache to > good effect. Is this true? Is it supposed to work with 7.3? Is there > something I need to do to get postgresql to take advatage of large ram > systems? > > Thanks for any advice. > Well, first off, the general recommendation is probably that 7.3 is really old, and you should try to upgrade to at least 7.4, though recommended to 8.0. The bigger questions: How much RAM do you have? How busy is your system? 8.0 doesn't really do anything to do make the system cache the data. What kernel are you using? Also, if your tables are small enough, and your RAM is big enough, you might already have everything cached. One way to flush the caches, is to allocate a bunch of memory, and then scan through it. Or maybe mmap a really big file, and access every byte. But if your kernel is smart enough, it could certainly deallocate pages after you stopped accessing them, so I can't say for sure that you can flush the memory cache. Usually, I believe these methods are sufficient. John =:-> --------------enigA0AF2BF7B9B582267E8407E5 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC+9N5JdeBCYSNAAMRAqWnAKCftXMEdupw2BcPfFF//U2sPrt8+ACgi1fz ClIX7xWItItgq2NhnLskcQM= =P+9r -----END PGP SIGNATURE----- --------------enigA0AF2BF7B9B582267E8407E5-- From pgsql-performance-owner@postgresql.org Thu Aug 11 20:13:31 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C74D552F05 for ; Thu, 11 Aug 2005 20:13:29 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 12193-08 for ; Thu, 11 Aug 2005 23:13:27 +0000 (GMT) Received: from sccrmhc14.comcast.net (sccrmhc14.comcast.net [204.127.202.59]) by svr1.postgresql.org (Postfix) with ESMTP id F307252F3C for ; Thu, 11 Aug 2005 20:13:25 -0300 (ADT) Received: from [192.168.1.52] (c-24-60-119-214.hsd1.ma.comcast.net[24.60.119.214]) by comcast.net (sccrmhc14) with ESMTP id <2005081123132801400k49ene>; Thu, 11 Aug 2005 23:13:28 +0000 Message-ID: <42FBDB97.1010906@comcast.net> Date: Thu, 11 Aug 2005 19:13:27 -0400 From: Jeffrey Tenny User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7) Gecko/20040616 X-Accept-Language: en-us, en MIME-Version: 1.0 To: John A Meinel Cc: pgsql-performance@postgresql.org Subject: Re: Mostly read performance References: <42FBCF4D.6060002@comcast.net> <42FBD377.8040006@arbash-meinel.com> In-Reply-To: <42FBD377.8040006@arbash-meinel.com> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.562 tagged_above=0 required=5 tests=[AWL=-0.680, DNS_FROM_RFC_ABUSE=0.374, DNS_FROM_RFC_POST=1.376, DNS_FROM_RFC_WHOIS=0.492] X-Spam-Level: * X-Archive-Number: 200508/136 X-Sequence-Number: 13883 John A Meinel wrote: > Well, first off, the general recommendation is probably that 7.3 is > really old, and you should try to upgrade to at least 7.4, though > recommended to 8.0. There have been issues with each release that led me to wait. Even now I'm waiting for some things to settle in the 8.0 JDBC driver (timezones), and 7.3 has behaved well for me. But yes, I'd like to upgrade. > > The bigger questions: How much RAM do you have? How busy is your system? The system for testing was 512MB. I'm in the process of buying some additional memory. However there was no swap activity on that system, so I doubt memory was the limiting factor. > > 8.0 doesn't really do anything to do make the system cache the data. > What kernel are you using? 2.4.X for various large x. (Multiple systems). Gonna try 2.6.x soon. > > Also, if your tables are small enough, and your RAM is big enough, you > might already have everything cached. Well, that's what you'd expect. But a first time 70MB fetch on a freshly rebooted system took just as long as all secondary times. (Took over a minute to fetch, which is too long for my needs, at least on secondary attempts). > One way to flush the caches, is to allocate a bunch of memory, and then > scan through it. Or maybe mmap a really big file, and access every byte. > But if your kernel is smart enough, it could certainly deallocate pages > after you stopped accessing them, so I can't say for sure that you can > flush the memory cache. Usually, I believe these methods are sufficient. Not sure how that would really help. It doesn't seem like the database or file system is caching the table content either way, which led me to this inquiry. From pgsql-performance-owner@postgresql.org Thu Aug 11 20:30:46 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1A84852B87 for ; Thu, 11 Aug 2005 20:30:45 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 54180-07 for ; Thu, 11 Aug 2005 23:30:40 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id E6DE552822 for ; Thu, 11 Aug 2005 20:30:36 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050811233036m9100g7hq6e>; Thu, 11 Aug 2005 23:30:36 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 41E2056008; Thu, 11 Aug 2005 18:30:36 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id BFBF355FCC; Thu, 11 Aug 2005 18:30:29 -0500 (CDT) Message-ID: <42FBDF94.5060104@arbash-meinel.com> Date: Thu, 11 Aug 2005 18:30:28 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Alvaro Herrera Cc: Postgresql Performance Subject: Re: Odd Locking Problem References: <42FBB6CF.6000404@arbash-meinel.com> <20050811210842.GA28253@alvh.no-ip.org> In-Reply-To: <20050811210842.GA28253@alvh.no-ip.org> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigA572A149154C7AAEB8BE044E" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.044 tagged_above=0 required=5 tests=[AWL=-0.006, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/137 X-Sequence-Number: 13884 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigA572A149154C7AAEB8BE044E Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Alvaro Herrera wrote: > On Thu, Aug 11, 2005 at 03:36:31PM -0500, John A Meinel wrote: > ... > > This is a known problem, solved in 8.1. A workaround for previous > releases is to defer FK checks until commit: So I don't know exactly what the fix was, but I just tested, and my problem is indeed fixed with the latest CVS head. It no longer blocks. > > create table b (a int references a initially deferred); > John =:-> --------------enigA572A149154C7AAEB8BE044E Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC+9+UJdeBCYSNAAMRAmrrAJ4xvUnHizNwntKnQvWdUeFoXJvsgACgrH3a xypks5LPA/BuZ3T9h2oZlpo= =b9YD -----END PGP SIGNATURE----- --------------enigA572A149154C7AAEB8BE044E-- From pgsql-performance-owner@postgresql.org Thu Aug 11 20:54:52 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4F04852ECA for ; Thu, 11 Aug 2005 20:54:50 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 09410-10 for ; Thu, 11 Aug 2005 23:54:47 +0000 (GMT) Received: from vms044pub.verizon.net (vms044pub.verizon.net [206.46.252.44]) by svr1.postgresql.org (Postfix) with ESMTP id DACD352EC7 for ; Thu, 11 Aug 2005 20:54:41 -0300 (ADT) Received: from osgiliath.mathom.us ([70.108.53.154]) by vms044.mailsrvcs.net (Sun Java System Messaging Server 6.2 HotFix 0.04 (built Dec 24 2004)) with ESMTPA id <0IL30088P1380PV1@vms044.mailsrvcs.net> for pgsql-performance@postgresql.org; Thu, 11 Aug 2005 18:54:45 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by osgiliath.mathom.us (Postfix) with ESMTP id 6861761C7F4 for ; Thu, 11 Aug 2005 19:54:44 -0400 (EDT) Received: from osgiliath.mathom.us ([127.0.0.1]) by localhost (osgiliath [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 27237-01-3 for ; Thu, 11 Aug 2005 19:54:44 -0400 (EDT) Received: by osgiliath.mathom.us (Postfix, from userid 1000) id 4C5346005C0; Thu, 11 Aug 2005 19:54:44 -0400 (EDT) Date: Thu, 11 Aug 2005 19:54:44 -0400 From: Michael Stone Subject: Re: PG8 Tuning In-reply-to: <1123780724.14573.108.camel@archimedes> To: pgsql-performance@postgresql.org Mail-Followup-To: pgsql-performance@postgresql.org Message-id: <20050811235444.GO19080@mathom.us> MIME-version: 1.0 Content-type: text/plain; charset=us-ascii; format=flowed Content-disposition: inline X-Pgp-Fingerprint: 53 FF 38 00 E7 DD 0A 9C 84 52 84 C5 EE DF 7C 88 X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mathom.us References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> <1123836428.19976.26.camel@amd64-laptop-spoe> <1123780724.14573.108.camel@archimedes> User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/138 X-Sequence-Number: 13885 On Thu, Aug 11, 2005 at 10:18:44AM -0700, Mark Lewis wrote: >Actually, it seems to me that with the addition of the WAL in PostgreSQL >and the subsequent decreased need to fsync the data files themselves >(only during checkpoints?), that the only time a battery-backed write >cache would make a really large performance difference would be on the >drive(s) hosting the WAL. Write cache on a raid array helps in the general case too, because it allows the controller to aggregate & reorder write requests. The OS probably tries to do this to some degree, but it can't do as well as the raid controller because it doesn't know the physical disk layout. >Hmmm, on second thought, now I think I understand the rationale behind >having a non-zero commit delay setting-- the problem with putting >pg_xlog on a single disk without a write cache is that frequent fsync() >calls might cause it to spend most of its time seeking instead of >writing (as seems to be happening to Paul here). Then again, the OS IO >scheduler should take care of this for you, making this a non-issue. The OS can't do anything much in terms of IO scheduling for synchronous writes. Either it handles them in suboptimal order or you get hideous latency while requests are queued for reordering. Neither option is really great. Mike Stone From pgsql-performance-owner@postgresql.org Thu Aug 11 21:34:38 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 63EEC53051 for ; Thu, 11 Aug 2005 21:30:33 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 20880-10 for ; Fri, 12 Aug 2005 00:30:30 +0000 (GMT) Received: from vms044pub.verizon.net (vms044pub.verizon.net [206.46.252.44]) by svr1.postgresql.org (Postfix) with ESMTP id 3B2E352EBC for ; Thu, 11 Aug 2005 21:30:29 -0300 (ADT) Received: from osgiliath.mathom.us ([70.108.53.154]) by vms044.mailsrvcs.net (Sun Java System Messaging Server 6.2 HotFix 0.04 (built Dec 24 2004)) with ESMTPA id <0IL300BRX2QW7OZ5@vms044.mailsrvcs.net> for pgsql-performance@postgresql.org; Thu, 11 Aug 2005 19:30:33 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by osgiliath.mathom.us (Postfix) with ESMTP id 207CF6005E0 for ; Thu, 11 Aug 2005 20:30:32 -0400 (EDT) Received: from osgiliath.mathom.us ([127.0.0.1]) by localhost (osgiliath [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 27559-05-3 for ; Thu, 11 Aug 2005 20:30:32 -0400 (EDT) Received: by osgiliath.mathom.us (Postfix, from userid 1000) id 05DF16005C0; Thu, 11 Aug 2005 20:30:32 -0400 (EDT) Date: Thu, 11 Aug 2005 20:30:31 -0400 From: Michael Stone Subject: Re: Mostly read performance In-reply-to: <42FBDB97.1010906@comcast.net> To: pgsql-performance@postgresql.org Mail-Followup-To: pgsql-performance@postgresql.org Message-id: <20050812003031.GP19080@mathom.us> MIME-version: 1.0 Content-type: text/plain; charset=us-ascii; format=flowed Content-disposition: inline X-Pgp-Fingerprint: 53 FF 38 00 E7 DD 0A 9C 84 52 84 C5 EE DF 7C 88 X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mathom.us References: <42FBCF4D.6060002@comcast.net> <42FBD377.8040006@arbash-meinel.com> <42FBDB97.1010906@comcast.net> User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 tagged_above=0 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/139 X-Sequence-Number: 13886 On Thu, Aug 11, 2005 at 07:13:27PM -0400, Jeffrey Tenny wrote: >The system for testing was 512MB That's definately *not* a "large ram" system. If you're reading a subset of data that totals 70MB I'm going to guess that your data set is larger than or at least a large fraction of 512MB. >additional memory. However there was no swap activity on that system, >so I doubt memory was the limiting factor. The system won't swap if your data set is larger than your memory, it just won't cache the data. >Well, that's what you'd expect. But a first time 70MB fetch on a >freshly rebooted system took just as long as all secondary times. (Took >over a minute to fetch, which is too long for my needs, at least on >secondary attempts). If the query involves a table scan and the data set is larger than your available memory, you'll need a full scan every time. If you do a table scan and the table fits in RAM, subsequent runs should be faster. If you have an index and only need to look at a subset of the table, subsequent runs should be faster. Without knowing more about your queries it's not clear what your situation is. Mike Stone From pgsql-performance-owner@postgresql.org Sun Aug 14 19:52:31 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id ACBD853073 for ; Fri, 12 Aug 2005 03:53:47 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 62983-08 for ; Fri, 12 Aug 2005 06:53:45 +0000 (GMT) Received: from mailrelay1.nefonline.de (mailrelay1.nefonline.de [212.114.153.196]) by svr1.postgresql.org (Postfix) with ESMTP id 023D553069 for ; Fri, 12 Aug 2005 03:53:43 -0300 (ADT) Received: from mail (mail.marekmicro.de [212.114.149.81]) by mailrelay1.nefonline.de (NEFkom Mailservice) with ESMTP id j7C6rgp24815 for ; Fri, 12 Aug 2005 08:53:43 +0200 Received: from [83.208.155.174] by mail.marekmicro.de (GMS 11.00.3335/NY4279.00.ab72d4bc) with SMTP id pqmfzhaa for pgsql-performance@postgresql.org; Fri, 12 Aug 2005 08:53:34 +0200 Received: from 127.0.0.1 (AVG SMTP 7.0.338 [267.10.7]); Fri, 12 Aug 2005 08:53:34 +0200 Message-ID: <001a01c59f0a$8e8d9560$de01a8c0@kavan> From: "Petr Kavan" To: Subject: How many views is ok? Date: Fri, 12 Aug 2005 08:53:33 +0200 Organization: MarekMicro MIME-Version: 1.0 Content-Type: text/plain; format=flowed; charset="iso-8859-1"; reply-type=original Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 6.00.2900.2180 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 X-Originating-IP: [83.208.155.174] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 tagged_above=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/143 X-Sequence-Number: 13890 I have database of company data, and some of them is table of information about employees. I need each employee to have access only to his own row. Postgre cannot do this by system of privileges, because that can give privileges only to whole tables. Possibility is to create a view for each employee that chooses only his data and give employee privileges to this view. But I am not sure if such number of views does not have some performance drawbacks or even if postgre can support it (I expect i can). I would need several tables protected like this and it can result in, say 1000 views in maximum. Because access to DB will go through PHP information system, other possibility to protect data is to let IS connect as more privileged than user really is, but let it retrieve only data for that user. View-approach seems far more clear than this, but im not sure if postgre can handle it without problems. Thanks for any reply :-) ----------------------------------------------------------- Petr Kavan Database Development From pgsql-performance-owner@postgresql.org Fri Aug 12 04:32:48 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B1FCD53061 for ; Fri, 12 Aug 2005 04:32:46 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 02730-09 for ; Fri, 12 Aug 2005 07:32:45 +0000 (GMT) Received: from stark.xeocode.com (stark.xeocode.com [216.58.44.227]) by svr1.postgresql.org (Postfix) with ESMTP id 13FFE53022 for ; Fri, 12 Aug 2005 04:32:44 -0300 (ADT) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1E3U1t-00048Q-00; Fri, 12 Aug 2005 03:32:37 -0400 To: Michael Stone Cc: pgsql-performance@postgresql.org Subject: Re: Mostly read performance References: <42FBCF4D.6060002@comcast.net> <42FBD377.8040006@arbash-meinel.com> <42FBDB97.1010906@comcast.net> <20050812003031.GP19080@mathom.us> In-Reply-To: <20050812003031.GP19080@mathom.us> From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 12 Aug 2005 03:32:37 -0400 Message-ID: <871x4za9uy.fsf@stark.xeocode.com> Lines: 25 User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.3 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.003 tagged_above=0 required=5 tests=[AWL=0.003] X-Spam-Level: X-Archive-Number: 200508/140 X-Sequence-Number: 13887 Michael Stone writes: > > Well, that's what you'd expect. But a first time 70MB fetch on a freshly > > rebooted system took just as long as all secondary times. (Took over a > > minute to fetch, which is too long for my needs, at least on secondary > > attempts). That's not impressively fast even for the disk. You should get up to about 40Mbit/s or 5MByte/s from the disk. Add some overhead for postgres; so I would expect a full table scan of 70MB to take more like 15-30s, not over a minute. What is your shared_buffers setting? Perhaps you have it set way too high or way too low? Also, you probably should post the "explain analyze" output of the actual query you're trying to optimize. Even if you're not looking for a better plan having hard numbers is better than guessing. And the best way to tell if the data is cached is having a "vmstat 1" running in another window. Start the query and look at the bi/bo columns. If you see bi spike upwards then it's reading from disk. -- greg From pgsql-performance-owner@postgresql.org Thu Aug 11 10:19:48 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9EC9552AFE for ; Thu, 11 Aug 2005 10:19:30 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 99406-05 for ; Thu, 11 Aug 2005 13:19:25 +0000 (GMT) Received: from mail.tecarta.com (66.238.115.135.ptr.us.xo.net [66.238.115.135]) by svr1.postgresql.org (Postfix) with ESMTP id 5E93652AE5 for ; Thu, 11 Aug 2005 10:19:24 -0300 (ADT) Received: from mail pickup service by mail.tecarta.com with Microsoft SMTPSVC; Thu, 11 Aug 2005 06:23:13 -0700 Received: from mail.tecarta.com ([192.168.160.2]) by mail.tecarta.com with Microsoft SMTPSVC(6.0.3790.0); Thu, 11 Aug 2005 06:23:09 -0700 Received: from barracuda.tecarta.com ([192.168.160.200]) by mail.tecarta.com (SAVSMTP 3.1.0.29) with SMTP id M2005081106230901800 for ; Thu, 11 Aug 2005 06:23:09 -0700 X-ASG-Debug-ID: 1123766364-13015-0-0 X-Barracuda-URL: http://192.168.160.200:8000/cgi-bin/mark.cgi Received: from mail1 (mail1.hq.corp [192.168.160.5]) by barracuda.tecarta.com (Spam Firewall) with SMTP id 1BD2D20171B3 for ; Thu, 11 Aug 2005 06:19:24 -0700 (PDT) Received: from [192.168.0.100] ([63.206.203.145]) by mail.tecarta.com with Microsoft SMTPSVC(6.0.3790.1830); Thu, 11 Aug 2005 06:23:02 -0700 X-ASG-Orig-Subj: Re: [PERFORM] PG8 Tuning Subject: [SPAM?] Re: PG8 Tuning From: Steve Poe To: paul@oxton.com Cc: pgsql-performance@postgresql.org In-Reply-To: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> Content-Type: text/plain Date: Fri, 12 Aug 2005 08:47:08 +0000 Message-Id: <1123836428.19976.26.camel@amd64-laptop-spoe> Mime-Version: 1.0 X-Mailer: Evolution 2.2.3 Content-Transfer-Encoding: 7bit X-OriginalArrivalTime: 11 Aug 2005 13:23:02.0302 (UTC) FILETIME=[CCBAD3E0:01C59E77] X-Virus-Scanned: by Barracuda Spam Firewall at tecarta.com X-Barracuda-Spam-Score: 4.03 X-Barracuda-Spam-Status: Yes, SCORE=4.03 using global scores of TAG_LEVEL=4.0 QUARANTINE_LEVEL=1000.0 KILL_LEVEL=7.0 tests=BAYES_60, DATE_IN_FUTURE_12_24 X-Barracuda-Spam-Report: Code version 3.02, rules version 3.0.3136 Rule breakdown below pts rule name description ---- ---------------------- -------------------------------------------------- 3.03 DATE_IN_FUTURE_12_24 Date: is 12 to 24 hours after Received: date 1.00 BAYES_60 BODY: Bayesian spam probability is 60 to 80% [score: 0.6148] X-Priority: 5 (Lowest) X-MSMail-Priority: Low Importance: Low X-Barracuda-Spam-Flag: YES X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.329 tagged_above=0 required=5 tests=[AWL=-1.050, DATE_IN_FUTURE_12_24=2.329, FORGED_RCVD_HELO=0.05, MISSING_MIMEOLE=0] X-Spam-Level: * X-Archive-Number: 200508/120 X-Sequence-Number: 13867 Paul, Before I say anything else, one online document which may be of assistance to you is: http://www.powerpostgresql.com/PerfList/ Some thoughts I have: 3) You're shared RAM setting seems overkill to me. Part of the challenge is you're going from 1000 to 262K with no assessment in between. Each situation can be different, but try in the range of 10 - 50K. 4) pg_xlog: If you're pg_xlog is on a spindle is *only* for pg_xlog you're better off. If it is sharing with any other OS/DB resource, the performance will be impacted. >From what I have learned from others on this list, RAID5 is not the best choice for the database. RAID10 would be a better solution (using 8 of your disks) then take the remaining disk and do mirror with your pg_xlog if possible. Best of luck, Steve Poe On Thu, 2005-08-11 at 13:23 +0100, Paul Johnson wrote: > Hi all, we're running PG8 on a Sun V250 with 8GB RAM and 2*1.3GHz SPARC > CPUs running Solaris 10. The DB cluster is on an external fibre-attached > Sun T3 array that has 9*36GB drives configured as a single RAID5 LUN. > > The system is for the sole use of a couple of data warehouse developers, > hence we are keen to use 'aggressive' tuning options to maximise > performance. > > So far we have made the following changes and measured the impact on our > test suite: > > 1) Increase checkpoint_segments from 3 to 64. This made a 10x improvement > in some cases. > > 2) Increase work_mem from 1,024 to 524,288. > > 3) Increase shared_buffers from 1,000 to 262,143 (2 GB). This required > setting SHMMAX=4294967295 (4 GB) in /etc/system and re-booting the box. > > Question - can Postgres only use 2GB RAM, given that shared_buffers can > only be set as high as 262,143 (8K pages)? > > So far so good... > > 4) Move /pg_xlog to an internal disk within the V250. This has had a > severe *negative* impact on performance. Copy job has gone from 2 mins to > 12 mins, simple SQL job gone from 1 min to 7 mins. Not even run long SQL > jobs. > > I'm guessing that this is because pg_xlog has gone from a 9 spindle LUN to > a single spindle disk? > > In cases such as this, where an external storage array with a hardware > RAID controller is used, the normal advice to separate the data from the > pg_xlog seems to come unstuck, or are we missing something? > > Cheers, > > Paul Johnson. > > ---------------------------(end of broadcast)--------------------------- > TIP 4: Have you searched our list archives? > > http://archives.postgresql.org From pgsql-performance-owner@postgresql.org Fri Aug 12 09:18:33 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0C3ED52A9D for ; Fri, 12 Aug 2005 09:18:32 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 62989-02 for ; Fri, 12 Aug 2005 12:18:28 +0000 (GMT) Received: from rwcrmhc11.comcast.net (rwcrmhc11.comcast.net [204.127.198.35]) by svr1.postgresql.org (Postfix) with ESMTP id 9317852A44 for ; Fri, 12 Aug 2005 09:18:27 -0300 (ADT) Received: from jefftrout.com ([24.147.120.205]) by comcast.net (rwcrmhc11) with SMTP id <2005081212182901300oiarje>; Fri, 12 Aug 2005 12:18:29 +0000 Received: (qmail 87675 invoked from network); 12 Aug 2005 12:19:12 -0000 Received: from c-24-147-120-205.hsd1.ma.comcast.net (HELO ?192.168.0.106?) (24.147.120.205) by 192.168.0.109 with SMTP; 12 Aug 2005 12:19:12 -0000 In-Reply-To: <1123779524.6664.1.camel@toonses.gghcwest.com> References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> <1123836428.19976.26.camel@amd64-laptop-spoe> <1123779524.6664.1.camel@toonses.gghcwest.com> Mime-Version: 1.0 (Apple Message framework v733) Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: Cc: Steve Poe , paul@oxton.com, pgsql-performance@postgresql.org Content-Transfer-Encoding: 7bit From: Jeff Trout Subject: Re: [SPAM?] Re: PG8 Tuning Date: Fri, 12 Aug 2005 08:18:27 -0400 To: Jeffrey W.Baker X-Mailer: Apple Mail (2.733) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 tagged_above=0 required=5 tests=[RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/141 X-Sequence-Number: 13888 On Aug 11, 2005, at 12:58 PM, Jeffrey W. Baker wrote: > Like Mr. Stone said earlier, this is pure dogma. In my experience, > xlogs on the same volume with data is much faster if both are on > battery-backed write-back RAID controller memory. Moving from this > situation to xlogs on a single normal disk is going to be much > slower in > most cases. > This does also point one important point about performance. Which is a touch unfortunate (and expensive to test): Your milage may vary on any of these improvements. Some people have 0 problems and incredible performance with say, 1000 shared_bufs and the WAL on the same disk.. Others need 10k shared bufs and wal split over a 900 spindle raid with data spread across 18 SAN's... Unfortunately there is no one true way :( The best bet (which is great if you can): Try out various settings.. if you still run into problems look into some more hardware.. see if you can borrow any or fabricate a "poor man"'s equivalent for testing. -- Jeff Trout http://www.jefftrout.com/ http://www.stuarthamm.net/ From pgsql-performance-owner@postgresql.org Fri Aug 12 19:54:32 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6D10F52C54 for ; Fri, 12 Aug 2005 19:37:49 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 64189-04 for ; Fri, 12 Aug 2005 22:37:47 +0000 (GMT) Received: from rwcrmhc12.comcast.net (rwcrmhc13.comcast.net [216.148.227.118]) by svr1.postgresql.org (Postfix) with ESMTP id C5FFB52C4B for ; Fri, 12 Aug 2005 19:37:45 -0300 (ADT) Received: from [192.168.1.52] (c-24-60-119-214.hsd1.ma.comcast.net[24.60.119.214]) by comcast.net (rwcrmhc13) with ESMTP id <20050812223747015004omnde>; Fri, 12 Aug 2005 22:37:48 +0000 Message-ID: <42FD24B9.8080209@comcast.net> Date: Fri, 12 Aug 2005 18:37:45 -0400 From: Jeffrey Tenny User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7) Gecko/20040616 X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Re: Mostly read performance (2 replies) Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.576 tagged_above=0 required=5 tests=[AWL=-0.666, DNS_FROM_RFC_ABUSE=0.374, DNS_FROM_RFC_POST=1.376, DNS_FROM_RFC_WHOIS=0.492] X-Spam-Level: * X-Archive-Number: 200508/142 X-Sequence-Number: 13889 (Pardon my replying two two replies at once, I only get the digest and this was easier). Michael Stone wrote: [...] >> Well, that's what you'd expect. But a first time 70MB fetch on a freshly rebooted system took just as long as all secondary times. (Took over a minute to fetch, which is too long for my needs, at least on secondary attempts). > > > If the query involves a table scan and the data set is larger than your > available memory, you'll need a full scan every time. If you do a table > scan and the table fits in RAM, subsequent runs should be faster. If you > have an index and only need to look at a subset of the table, subsequent > runs should be faster. Without knowing more about your queries it's not > clear what your situation is. I must amend my original statement. I'm not using a parameterized statement. The system is effectively fetching file content stored in the database for portions of one or more files. It attempts to batch the records being fetched into as few non-parameterized queries as possible, while balancing the rowset retrieval memory impact. Currently that means it will request up to 16K records in a query that is assembled using a combination of IN (recids...) , BETWEEN ranges, and UNION ALL for multiple file IDs. I do this to minimize the latency of dbclient/dbserver requests, while at the same time capping the maximum data returned by a DBIO to about 1.2MB per maximum retrieved record set. (I'm trying not to pound the java app server via jdbc memory usage). There's an ORDER BY on the file id column too. It sounds like a simple enough thing to do, but this "pieces of many files in a database" problem is actually pretty hard to optimize. Fetching all records for all files, even though I don't need all records, is both inefficient and likely to use too much memory. Fetching 1 file at a time is likely to result in too many queries (latency overhead). So right now I err on the side of large but record limited queries. That let's me process many files in one query, unless the pieces of the files I need are substantial. (I've been burned by trying to use setFetchSize so many times it isn't funny, I never count on that any more). An index is in place to assist with record selection, I'll double check that it's being used. It's a joint index on file-id and record-id-within-the-file. I'll check to be sure it's being used. ------------------------ Greg Stark wrote: [...] > What is your shared_buffers setting? Perhaps you have it set way too high or > way too low? I generally run with the conservative installation default. I did some experimenting with larger values but didn't see any improvement (and yes, I restarted postmaster). This testing was done a while ago, I don't have the numbers in memory any more so I can't tell you what they were. > > Also, you probably should post the "explain analyze" output of the actual > query you're trying to optimize. Even if you're not looking for a better plan > having hard numbers is better than guessing. A good suggestion. I'll look into it. > And the best way to tell if the data is cached is having a "vmstat 1" running > in another window. Start the query and look at the bi/bo columns. If you see > bi spike upwards then it's reading from disk. Another good suggestion. I'll look into getting further data from the above suggestions. I'm also looking into getting a gig or two of ram to make sure that isn't an issue. The basis of my post originally was to make sure that, all things being equal, there's no reason those disk I/Os on behalf of the database shouldn't be cached by the operating/file system so that repeated reads might benefit from in-memory data. From pgsql-performance-owner@postgresql.org Sun Aug 14 21:22:29 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6762A52B76 for ; Sun, 14 Aug 2005 21:22:26 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 13700-07 for ; Mon, 15 Aug 2005 00:22:24 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id AC12D52B45 for ; Sun, 14 Aug 2005 21:22:22 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050815002225m9100g5h18e>; Mon, 15 Aug 2005 00:22:25 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id E8DDE55FBA; Sun, 14 Aug 2005 19:22:24 -0500 (CDT) Received: from [192.168.1.11] (liliana.arbash-meinel.com [192.168.1.11]) by juju.arbash-meinel.com (Postfix) with ESMTP id 0EA0C55FAD; Sun, 14 Aug 2005 19:22:17 -0500 (CDT) Message-ID: <42FFE034.2020809@arbash-meinel.com> Date: Sun, 14 Aug 2005 19:22:12 -0500 From: John Arbash Meinel User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050727) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Petr Kavan Cc: pgsql-performance@postgresql.org Subject: Re: How many views is ok? References: <001a01c59f0a$8e8d9560$de01a8c0@kavan> In-Reply-To: <001a01c59f0a$8e8d9560$de01a8c0@kavan> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig147DE0C9C9250FD6A8D13C72" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.044 required=5 tests=[AWL=-0.006, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/146 X-Sequence-Number: 13893 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig147DE0C9C9250FD6A8D13C72 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Petr Kavan wrote: > I have database of company data, and some of them is table of > information about employees. I need each employee to have access only > to his own row. Postgre cannot do this by system of privileges, > because that can give privileges only to whole tables. > > Possibility is to create a view for each employee that chooses only > his data and give employee privileges to this view. But I am not sure > if such number of views does not have some performance drawbacks or > even if postgre can support it (I expect i can). I would need several > tables protected like this and it can result in, say 1000 views in > maximum. > > Because access to DB will go through PHP information system, other > possibility to protect data is to let IS connect as more privileged > than user really is, but let it retrieve only data for that user. > > View-approach seems far more clear than this, but im not sure if > postgre can handle it without problems. We do a similar thing tying user to per-row permissions. We have 1 view per table, and it works fine. I would recommend that you do something similar. Basically, just make the view: CREATE VIEW just_me SECURITY DEFINER AS SELECT * FROM user_table WHERE username=session_user; REVOKE ALL FROM user_table; GRANT SELECT TO just_me TO PUBLIC; security definer, means that the 'just_me' view will be executed as the user who created the function (superuser). The REVOKE ALL (my syntax might be wrong) prevents users from querying the user tables directly. The 'session_user' makes the view use the name of the actual connected user (because of security definer, plain 'user' is the superuser) This should allow a user to see only their own row in the database. (Whichever rows that have username matching the connected name). Now, this only works if the php front end connects specifically as the given user (our system is setup to do this). If you don't do it this way, you'll always be stuck with the IS layer doing the restriction. Even if you create a view per user, if your PHP layer has the right to look at other tables/views, it doesn't really help. Good luck, John =:-> > > Thanks for any reply :-) > > ----------------------------------------------------------- > Petr Kavan > Database Development > > > > ---------------------------(end of broadcast)--------------------------- > TIP 3: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faq > --------------enig147DE0C9C9250FD6A8D13C72 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC/+A4JdeBCYSNAAMRAmf1AJsHZghn/X3xlqi5TEbUnmQ+q3gH6wCfVbc/ oYGPbncq79Eq/Y/Pgp3aB/g= =BVAg -----END PGP SIGNATURE----- --------------enig147DE0C9C9250FD6A8D13C72-- From pgsql-performance-owner@postgresql.org Sun Aug 14 21:27:57 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3275452DCC for ; Sun, 14 Aug 2005 21:27:55 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 13076-04 for ; Mon, 15 Aug 2005 00:27:52 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id A700552DC4 for ; Sun, 14 Aug 2005 21:27:50 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050815002743m92009jakke>; Mon, 15 Aug 2005 00:27:53 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id F3CF455FBA; Sun, 14 Aug 2005 19:27:42 -0500 (CDT) Received: from [192.168.1.11] (liliana.arbash-meinel.com [192.168.1.11]) by juju.arbash-meinel.com (Postfix) with ESMTP id 2F1F455FAD; Sun, 14 Aug 2005 19:27:39 -0500 (CDT) Message-ID: <42FFE17A.6010208@arbash-meinel.com> Date: Sun, 14 Aug 2005 19:27:38 -0500 From: John Arbash Meinel User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050727) X-Accept-Language: en-us, en MIME-Version: 1.0 To: =?ISO-8859-1?Q?St=E9phane_COEZ?= Cc: pgsql-performance@postgresql.org Subject: Re: Performance pb vs SQLServer. References: <20050811131911.3175A1C0031B@mwinf0401.wanadoo.fr> In-Reply-To: <20050811131911.3175A1C0031B@mwinf0401.wanadoo.fr> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigD250D4B21DC4E3093825C750" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.043 required=5 tests=[AWL=-0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/147 X-Sequence-Number: 13894 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigD250D4B21DC4E3093825C750 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable St=E9phane COEZ wrote: >Hi, > >I have a perfomance issue : > >I run PG (8.0.3) and SQLServer2000 on a Windows2000 Server (P4 1,5Ghz 51= 2Mo) >I have a table (3200000 rows) and I run this single query : > >select cod from mytable group by cod >I have an index on cod (char(4) - 88 different values) > >PG =3D ~ 20 sec. >SQLServer =3D < 8 sec > > >the explain is : > >HashAggregate (cost=3D64410.09..64410.09 rows=3D55 width=3D8) > -> Seq Scan on mytable (cost=3D0.00..56325.27 rows=3D3233927 width=3D= 8) > > >if I switch to "enable_hashagg =3D false" (just for a try...) >the planner will choose my index : > >Group (cost=3D0.00..76514.01 rows=3D55 width=3D8) > -> Index Scan using myindex on mytable (cost=3D0.00..68429.20 rows=3D= 3233927 >width=3D8) > >but performance will be comparable to previous test. > >So with or without using Index I have the same result. > =20 > My guess is that this is part of a larger query. There isn't really much you can do. If you want all 3.2M rows, then you have to wait for them to be pulled in. What you generally can do for performance, is to restructure things, so that you *don't* have to touch all 3.2M rows. If you are just trying to determine what the unique entries are for cod, you probably are better off doing some normalization, and keeping a separate table of cod values. I'm guessing the reason your query is faster with SQLServer is because of how postgres handles MVCC. Basically, it still has to fetch the main page to determine if a row exists. While SQL server doesn't do MVCC, so it can just look things up in the index. You might also try a different query, something like: SELECT DISTINCT cod FROM mytable ORDER BY cod GROUP BY cod; (You may or may not want order by, or group by, try the different combinations.) It might be possible to have the planner realize that all you want is unique rows, just doing a group by doesn't give you that. John =3D:-> > >Thanks for help. >=20 >St=E9phane COEZ > > > > >---------------------------(end of broadcast)---------------------------= >TIP 9: In versions below 8.0, the planner will ignore your desire to > choose an index scan if your joining column's datatypes do not > match > > =20 > --------------enigD250D4B21DC4E3093825C750 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC/+F6JdeBCYSNAAMRAjKFAJ9N79+cvK394sIPq4gIvwfFu8x2XgCaAwnm QJu7OTqD6J+QWxIeIEj0jvk= =I7DZ -----END PGP SIGNATURE----- --------------enigD250D4B21DC4E3093825C750-- From pgsql-performance-owner@postgresql.org Sun Aug 14 21:38:05 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 68B8853186 for ; Sun, 14 Aug 2005 21:38:02 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 11850-08 for ; Mon, 15 Aug 2005 00:38:00 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 4BEF153180 for ; Sun, 14 Aug 2005 21:38:00 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7F0c0t3005060; Sun, 14 Aug 2005 20:38:00 -0400 (EDT) To: "Petr Kavan" Cc: pgsql-performance@postgresql.org Subject: Re: How many views is ok? In-reply-to: <001a01c59f0a$8e8d9560$de01a8c0@kavan> References: <001a01c59f0a$8e8d9560$de01a8c0@kavan> Comments: In-reply-to "Petr Kavan" message dated "Fri, 12 Aug 2005 08:53:33 +0200" Date: Sun, 14 Aug 2005 20:38:00 -0400 Message-ID: <5059.1124066280@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/148 X-Sequence-Number: 13895 "Petr Kavan" writes: > Possibility is to create a view for each employee that chooses only his data > and give employee privileges to this view. But I am not sure if such number > of views does not have some performance drawbacks or even if postgre can > support it (I expect i can). Do you really need more than one view? I'd consider something like create view emp_view as select * from emp where name = current_user; This requires that your Postgres usernames match up with something in the underlying table, of course. regards, tom lane From pgsql-performance-owner@postgresql.org Sun Aug 14 22:02:07 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 96B5552B08 for ; Sun, 14 Aug 2005 22:02:04 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 20018-08 for ; Mon, 15 Aug 2005 01:02:01 +0000 (GMT) Received: from cassarossa.samfundet.no (cassarossa.samfundet.no [129.241.93.19]) by svr1.postgresql.org (Postfix) with ESMTP id 06B2452B0A for ; Sun, 14 Aug 2005 22:01:59 -0300 (ADT) Received: from trofast.sesse.net ([129.241.93.32]) by cassarossa.samfundet.no with esmtp (Exim 4.50) id 1E4TMW-0002cN-Md for pgsql-performance@postgresql.org; Mon, 15 Aug 2005 03:02:02 +0200 Received: from sesse by trofast.sesse.net with local (Exim 3.36 #1 (Debian)) id 1E4TMV-0007Lx-00 for ; Mon, 15 Aug 2005 03:01:59 +0200 Date: Mon, 15 Aug 2005 03:01:59 +0200 From: "Steinar H. Gunderson" To: pgsql-performance@postgresql.org Subject: Re: Performance pb vs SQLServer. Message-ID: <20050815010159.GA27795@uio.no> Mail-Followup-To: pgsql-performance@postgresql.org References: <20050811131911.3175A1C0031B@mwinf0401.wanadoo.fr> <42FFE17A.6010208@arbash-meinel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <42FFE17A.6010208@arbash-meinel.com> X-Operating-System: Linux 2.6.11.8 on a i686 X-Message-Flag: Outlook? --> http://www.mozilla.org/products/thunderbird/ User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.017 required=5 tests=[AWL=0.017] X-Spam-Level: X-Archive-Number: 200508/149 X-Sequence-Number: 13896 On Sun, Aug 14, 2005 at 07:27:38PM -0500, John Arbash Meinel wrote: > My guess is that this is part of a larger query. There isn't really much > you can do. If you want all 3.2M rows, then you have to wait for them to > be pulled in. To me, it looks like he'll get 88 rows, not 3.2M. Surely we must be able to do something better than a full sequential scan in this case? test=# create table foo ( bar char(4) ); CREATE TABLE test=# insert into foo values ('0000'); INSERT 24773320 1 test=# insert into foo values ('0000'); INSERT 24773321 1 test=# insert into foo values ('1111'); INSERT 24773322 1 test=# select * from foo group by bar; bar ------ 1111 0000 (2 rows) I considered doing some odd magic with generate_series() and subqueries with LIMIT 1, but it was a bit too weird in the end :-) /* Steinar */ -- Homepage: http://www.sesse.net/ From pgsql-performance-owner@postgresql.org Sun Aug 14 22:04:18 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AC02B52B0A for ; Sun, 14 Aug 2005 22:04:15 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 18411-07 for ; Mon, 15 Aug 2005 01:04:12 +0000 (GMT) Received: from cassarossa.samfundet.no (cassarossa.samfundet.no [129.241.93.19]) by svr1.postgresql.org (Postfix) with ESMTP id 83B9852B08 for ; Sun, 14 Aug 2005 22:04:12 -0300 (ADT) Received: from trofast.ipv6.sesse.net ([2001:700:300:dc03:20e:cff:fe36:a766] helo=trofast.sesse.net) by cassarossa.samfundet.no with esmtp (Exim 4.50) id 1E4TOh-0002cw-PO for pgsql-performance@postgresql.org; Mon, 15 Aug 2005 03:04:16 +0200 Received: from sesse by trofast.sesse.net with local (Exim 3.36 #1 (Debian)) id 1E4TOh-0007QW-00 for ; Mon, 15 Aug 2005 03:04:15 +0200 Date: Mon, 15 Aug 2005 03:04:15 +0200 From: "Steinar H. Gunderson" To: pgsql-performance@postgresql.org Subject: Re: Performance pb vs SQLServer. Message-ID: <20050815010415.GB27795@uio.no> Mail-Followup-To: pgsql-performance@postgresql.org References: <20050811131911.3175A1C0031B@mwinf0401.wanadoo.fr> <42FFE17A.6010208@arbash-meinel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <42FFE17A.6010208@arbash-meinel.com> X-Operating-System: Linux 2.6.11.8 on a i686 X-Message-Flag: Outlook? --> http://www.mozilla.org/products/thunderbird/ User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.017 required=5 tests=[AWL=0.017] X-Spam-Level: X-Archive-Number: 200508/150 X-Sequence-Number: 13897 On Sun, Aug 14, 2005 at 07:27:38PM -0500, John Arbash Meinel wrote: > If you are just trying to determine what the unique entries are for cod, > you probably are better off doing some normalization, and keeping a > separate table of cod values. Pah, I missed this part of the e-mail -- you can ignore most of my (other) reply, then :-) /* Steinar */ -- Homepage: http://www.sesse.net/ From pgsql-performance-owner@postgresql.org Sun Aug 14 22:06:07 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D8CCA52EBF for ; Sun, 14 Aug 2005 22:06:03 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 20648-06 for ; Mon, 15 Aug 2005 01:05:59 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id 4A01352ABA for ; Sun, 14 Aug 2005 22:05:58 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050815010602m9100g6bcte>; Mon, 15 Aug 2005 01:06:02 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 25A1155FBA; Sun, 14 Aug 2005 20:06:02 -0500 (CDT) Received: from [192.168.1.11] (liliana.arbash-meinel.com [192.168.1.11]) by juju.arbash-meinel.com (Postfix) with ESMTP id 07EFE55FAD; Sun, 14 Aug 2005 20:05:59 -0500 (CDT) Message-ID: <42FFEA76.6070109@arbash-meinel.com> Date: Sun, 14 Aug 2005 20:05:58 -0500 From: John Arbash Meinel User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050727) X-Accept-Language: en-us, en MIME-Version: 1.0 To: "Steinar H. Gunderson" Cc: pgsql-performance@postgresql.org Subject: Re: Performance pb vs SQLServer. References: <20050811131911.3175A1C0031B@mwinf0401.wanadoo.fr> <42FFE17A.6010208@arbash-meinel.com> <20050815010159.GA27795@uio.no> In-Reply-To: <20050815010159.GA27795@uio.no> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig600AA473DF119A810C53F557" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.043 required=5 tests=[AWL=-0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/151 X-Sequence-Number: 13898 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig600AA473DF119A810C53F557 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Steinar H. Gunderson wrote: >On Sun, Aug 14, 2005 at 07:27:38PM -0500, John Arbash Meinel wrote: > > >>My guess is that this is part of a larger query. There isn't really much >>you can do. If you want all 3.2M rows, then you have to wait for them to >>be pulled in. >> >> > >To me, it looks like he'll get 88 rows, not 3.2M. Surely we must be able to >do something better than a full sequential scan in this case? > >test=# create table foo ( bar char(4) ); >CREATE TABLE >test=# insert into foo values ('0000'); >INSERT 24773320 1 >test=# insert into foo values ('0000'); >INSERT 24773321 1 >test=# insert into foo values ('1111'); >INSERT 24773322 1 >test=# select * from foo group by bar; > bar >------ > 1111 > 0000 >(2 rows) > >I considered doing some odd magic with generate_series() and subqueries with >LIMIT 1, but it was a bit too weird in the end :-) > >/* Steinar */ > > I think a plain "GROUP BY" is not smart enough to detect it doesn't need all rows (since it is generally used because you want to get aggregate values of other columns). I think you would want something like SELECT DISTINCT, possibly with an ORDER BY rather than a GROUP BY (which was my final suggestion). John =:-> --------------enig600AA473DF119A810C53F557 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFC/+p2JdeBCYSNAAMRAnhQAJ4oOR2RxZyaLi4yJsrSCHJObxtfewCfQZIV 6mg+ebmDTj6dK5q4doe4hj4= =6vR1 -----END PGP SIGNATURE----- --------------enig600AA473DF119A810C53F557-- From pgsql-performance-owner@postgresql.org Sun Aug 14 22:18:49 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5138252B71 for ; Sun, 14 Aug 2005 22:18:47 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 24135-01 for ; Mon, 15 Aug 2005 01:18:46 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 832AF52974 for ; Sun, 14 Aug 2005 22:18:44 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7F1IjgU014406; Sun, 14 Aug 2005 21:18:45 -0400 (EDT) To: "Steinar H. Gunderson" Cc: pgsql-performance@postgresql.org Subject: Re: Performance pb vs SQLServer. In-reply-to: <20050815010159.GA27795@uio.no> References: <20050811131911.3175A1C0031B@mwinf0401.wanadoo.fr> <42FFE17A.6010208@arbash-meinel.com> <20050815010159.GA27795@uio.no> Comments: In-reply-to "Steinar H. Gunderson" message dated "Mon, 15 Aug 2005 03:01:59 +0200" Date: Sun, 14 Aug 2005 21:18:45 -0400 Message-ID: <14401.1124068725@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/152 X-Sequence-Number: 13899 "Steinar H. Gunderson" writes: > To me, it looks like he'll get 88 rows, not 3.2M. Surely we must be able to > do something better than a full sequential scan in this case? Not really. There's been some speculation about implementing index "skip search" --- once you've verified there's at least one visible row of a given index value, tell the index to skip to the next different value instead of handing back any of the remaining entries of the current value. But it'd be a lot of work and AFAICS not useful for very many kinds of queries besides this. regards, tom lane From pgsql-performance-owner@postgresql.org Sun Aug 14 22:37:47 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A05B35318D for ; Sun, 14 Aug 2005 22:37:43 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 26437-04 for ; Mon, 15 Aug 2005 01:37:41 +0000 (GMT) Received: from cassarossa.samfundet.no (cassarossa.samfundet.no [129.241.93.19]) by svr1.postgresql.org (Postfix) with ESMTP id D9E8F53186 for ; Sun, 14 Aug 2005 22:37:40 -0300 (ADT) Received: from trofast.sesse.net ([129.241.93.32]) by cassarossa.samfundet.no with esmtp (Exim 4.50) id 1E4Tv4-0002hi-AJ for pgsql-performance@postgresql.org; Mon, 15 Aug 2005 03:37:43 +0200 Received: from sesse by trofast.sesse.net with local (Exim 3.36 #1 (Debian)) id 1E4Tv3-0008Vq-00 for ; Mon, 15 Aug 2005 03:37:41 +0200 Date: Mon, 15 Aug 2005 03:37:41 +0200 From: "Steinar H. Gunderson" To: pgsql-performance@postgresql.org Subject: Re: Performance pb vs SQLServer. Message-ID: <20050815013741.GA32106@uio.no> Mail-Followup-To: pgsql-performance@postgresql.org References: <20050811131911.3175A1C0031B@mwinf0401.wanadoo.fr> <42FFE17A.6010208@arbash-meinel.com> <20050815010159.GA27795@uio.no> <14401.1124068725@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <14401.1124068725@sss.pgh.pa.us> X-Operating-System: Linux 2.6.11.8 on a i686 X-Message-Flag: Outlook? --> http://www.mozilla.org/products/thunderbird/ User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.016 required=5 tests=[AWL=0.016] X-Spam-Level: X-Archive-Number: 200508/153 X-Sequence-Number: 13900 On Sun, Aug 14, 2005 at 09:18:45PM -0400, Tom Lane wrote: > Not really. There's been some speculation about implementing index > "skip search" --- once you've verified there's at least one visible > row of a given index value, tell the index to skip to the next different > value instead of handing back any of the remaining entries of the > current value. But it'd be a lot of work and AFAICS not useful for > very many kinds of queries besides this. This is probably a completely wrong way of handling it all, but could it be done in a PL/PgSQL query like this? (Pseudo-code, sort of; I'm not very well versed in the actual syntax, but I'd guess you get the idea.) x = ( SELECT foo FROM table ORDER BY foo LIMIT 1 ); WHILE x IS NOT NULL RETURN NEXT x; x = ( SELECT foo FROM table WHERE foo > x ORDER BY foo LIMIT 1 ); END; (Replace with max() and min() for 8.1, of course.) /* Steinar */ - fond of horrible hacks :-) -- Homepage: http://www.sesse.net/ From pgsql-performance-owner@postgresql.org Mon Aug 15 02:23:32 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1810752A95 for ; Mon, 15 Aug 2005 02:23:31 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83171-07 for ; Mon, 15 Aug 2005 05:23:28 +0000 (GMT) Received: from smtp.seznam.cz (smtp.seznam.cz [212.80.76.43]) by svr1.postgresql.org (Postfix) with SMTP id 2413C528CC for ; Mon, 15 Aug 2005 02:23:25 -0300 (ADT) Received: (qmail 15477 invoked from network); 15 Aug 2005 05:23:24 -0000 Received: from unknown (HELO ?192.168.1.222?) (kavan.petr@83.208.155.174) by cetus.go.seznam.cz with ESMTPA; 15 Aug 2005 05:23:24 -0000 Received: from 127.0.0.1 (AVG SMTP 7.0.338 [267.10.9]); Mon, 15 Aug 2005 07:23:24 +0200 Message-ID: <005301c5a159$75a2ec20$de01a8c0@kavan> From: "Petr Kavan" To: "John Arbash Meinel" Cc: References: <001a01c59f0a$8e8d9560$de01a8c0@kavan> <42FFE034.2020809@arbash-meinel.com> Subject: Re: How many views is ok? Date: Mon, 15 Aug 2005 07:23:23 +0200 MIME-Version: 1.0 Content-Type: text/plain; format=flowed; charset="iso-8859-1"; reply-type=original Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 6.00.2900.2180 X-Mimeole: Produced By Microsoft MimeOLE V6.00.2900.2180 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.376 required=5 tests=[DNS_FROM_RFC_POST=1.376] X-Spam-Level: * X-Archive-Number: 200508/155 X-Sequence-Number: 13902 Hey, that trick with session_user is great! :-) Thank you all very much, this will certainly help. ----------------------------------------------------------- Petr Kavan Database Development ----- Original Message ----- From: "John Arbash Meinel" To: "Petr Kavan" Cc: Sent: Monday, August 15, 2005 2:22 AM Subject: Re: [PERFORM] How many views is ok? From pgsql-performance-owner@postgresql.org Mon Aug 15 01:54:35 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 438B2529B8 for ; Mon, 15 Aug 2005 01:54:34 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 77139-10 for ; Mon, 15 Aug 2005 04:54:32 +0000 (GMT) Received: from netbox.unitech.com.ar (unknown [200.32.92.34]) by svr1.postgresql.org (Postfix) with ESMTP id 7E28E5298F for ; Mon, 15 Aug 2005 01:54:29 -0300 (ADT) Received: (from dariop@localhost) by netbox.unitech.com.ar (8.11.6/8.11.6) id j7F5ptH26582; Mon, 15 Aug 2005 01:51:55 -0400 Date: Mon, 15 Aug 2005 01:51:55 -0400 From: dario_d_s@unitech.com.ar X-Authentication-Warning: netbox.unitech.com.ar: dariop set sender to dario_d_s@unitech.com.ar using -f Subject: Re: Performance pb vs SQLServer. To: scoez@harrysoftware.com Cc: pgsql-performance@postgresql.org X-Originating-IP: 201.254.9.56 X-Mailer: Usermin 1.110 X-Priority: 4 Message-Id: <1124085115.26580@netbox.unitech.com.ar> MIME-Version: 1.0 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.154 required=5 tests=[AWL=-0.024, NO_REAL_NAME=0.178] X-Spam-Level: Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 7bit X-Archive-Number: 200508/154 X-Sequence-Number: 13901 This is a multi-part message in MIME format. --bound1124085115 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 7bit One little thing. Did you shutdown sql2000 while testing postgresql? Remember that postgresql uses system cache. Sql2000 uses a large part of memory as buffer and it will not be available to operating system. I must say that, probably, results will be the same, but it will be a better test. > I'm guessing the reason your query is faster with SQLServer is because > of how postgres handles MVCC. Basically, it still has to fetch the main > page to determine if a row exists. While SQL server doesn't do MVCC, so > it can just look things up in the index. Another thing [almost offtopic]: I would like to add something to understand what does MVCC means and what are the consecuences. MVCC: multiversion concurrency control. (ehhh...) Just do this. Open two psql sessions. Do this: Session 1: begin; update any_table set any_column = 'value_a' where other_column = 'value_b' -- do not commit Session 2: select any_table where other_column = 'value_b' Watch the result. Session 1: commit; Session 2: select any_table where other_column = 'value_b' Watch the result. Now open two session in query analyzer. Do the same thing: Session 1: begin tran update any_table set any_column = 'value_a' where other_column = 'value_b' -- do not commit Session 2: select any_table where other_column = 'value_b' Wait for result. Wait... wait... (Oh, a lock! Ok, when you get tired, go back to session 1.) Session 1: commit Session 2: Then watch the result. Which one was faster? ["very, very offtopic"] Ok. This comparition is just as useless as the other one, because it's comparing oranges with apples (It's funny anyway). I was just choosing an example in which you can see the best of postgresql against 'not so nice' behavior of mssql2000 (no service pack, it's my desktop system, I'll do the same test later with SP4 and different isolation levels and I'll check results). Furthermore, MSSQL2000 is 5 years old now. Does anybody has the same cellular phone, or computer? (I don't want to know :-) ). The big question is 'What do you need?'. No system can give you all. That's marketing 'sarasa'. Sorry for my english and the noise. [End of offtopic] Long life, little spam and prosperity. --bound1124085115-- From pgsql-performance-owner@postgresql.org Mon Aug 15 05:18:07 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id ABC23530C6 for ; Mon, 15 Aug 2005 05:18:05 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 28805-02 for ; Mon, 15 Aug 2005 08:18:02 +0000 (GMT) Received: from mx-2.sollentuna.net (mx-2.sollentuna.net [195.84.163.199]) by svr1.postgresql.org (Postfix) with ESMTP id 727C453021 for ; Mon, 15 Aug 2005 05:18:02 -0300 (ADT) Received: from ALGOL.sollentuna.se (janus.sollentuna.se [62.65.68.67]) by mx-2.sollentuna.net (Postfix) with ESMTP id C1EC18F289; Mon, 15 Aug 2005 10:18:03 +0200 (CEST) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Subject: Re: Performance pb vs SQLServer. X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Date: Mon, 15 Aug 2005 10:18:03 +0200 Message-ID: <6BCB9D8A16AC4241919521715F4D8BCE6C789B@algol.sollentuna.se> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Performance pb vs SQLServer. Thread-Index: AcWedz241y3YYG5NRWucxBfjaAmDhQC+oSZA From: "Magnus Hagander" To: =?iso-8859-1?Q?St=E9phane_COEZ?= , X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.061 required=5 tests=[AWL=0.011, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/156 X-Sequence-Number: 13903 > Hi, >=20 > I have a perfomance issue : >=20 > I run PG (8.0.3) and SQLServer2000 on a Windows2000 Server=20 > (P4 1,5Ghz 512Mo) I have a table (3200000 rows) and I run=20 > this single query : >=20 > select cod from mytable group by cod > I have an index on cod (char(4) - 88 different values) >=20 > PG =3D ~ 20 sec. > SQLServer =3D < 8 sec >=20 >=20 > the explain is : >=20 > HashAggregate (cost=3D64410.09..64410.09 rows=3D55 width=3D8) > -> Seq Scan on mytable (cost=3D0.00..56325.27 rows=3D3233927 = width=3D8) >=20 >=20 > if I switch to "enable_hashagg =3D false" (just for a try...)=20 > the planner will choose my index : >=20 > Group (cost=3D0.00..76514.01 rows=3D55 width=3D8) > -> Index Scan using myindex on mytable =20 > (cost=3D0.00..68429.20 rows=3D3233927 > width=3D8) >=20 > but performance will be comparable to previous test. >=20 > So with or without using Index I have the same result. Out of curiosity, what plan do you get from SQLServer? I bet it's a = clustered index scan... //Magnus From pgsql-performance-owner@postgresql.org Mon Aug 15 05:25:51 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 800295311B for ; Mon, 15 Aug 2005 05:25:48 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 23969-08 for ; Mon, 15 Aug 2005 08:25:46 +0000 (GMT) Received: from mx-2.sollentuna.net (mx-2.sollentuna.net [195.84.163.199]) by svr1.postgresql.org (Postfix) with ESMTP id D290D53115 for ; Mon, 15 Aug 2005 05:25:45 -0300 (ADT) Received: from ALGOL.sollentuna.se (janus.sollentuna.se [62.65.68.67]) by mx-2.sollentuna.net (Postfix) with ESMTP id 26F978F289; Mon, 15 Aug 2005 10:25:47 +0200 (CEST) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: quoted-printable Subject: Re: Performance pb vs SQLServer. X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Date: Mon, 15 Aug 2005 10:25:47 +0200 Message-ID: <6BCB9D8A16AC4241919521715F4D8BCE6C789C@algol.sollentuna.se> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Performance pb vs SQLServer. Thread-Index: AcWhV2CLRhRsDMJrSC6MdQEWvUnh7wAGpBzQ From: "Magnus Hagander" To: , Cc: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.061 required=5 tests=[AWL=0.011, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/157 X-Sequence-Number: 13904 > ["very, very offtopic"] > Ok. This comparition is just as useless as the other one,=20 > because it's comparing oranges with apples (It's funny=20 > anyway). I was just choosing an example in which you can see=20 > the best of postgresql against 'not so nice' behavior of=20 > mssql2000 (no service pack, it's my desktop system, I'll do=20 > the same test later with SP4 and different isolation levels=20 > and I'll check results). There will be no difference in the service packs. SQL 2005 has "MVCC" (they call it something different, of course, but that's basicallyi what it is) > Furthermore, MSSQL2000 is 5 years=20 > old now. Does anybody has the same cellular phone, or=20 > computer? (I don't want to know :-) ). The big question is There is a big difference between your database and your cellphone. There are a lot of systems out there running very solidly on older products like MSSQL 7 (probably even some on 6.x), as well as Oracle 7,8 and 9... I'd say there is generally a huge difference in reliabilty in your cellphone hw/sw than there is in your db hw/sw. I have yet to see a cellphone that can run for a year without a reboot (or with a lot of brands, complete replacement). =20 //Magnus From pgsql-performance-owner@postgresql.org Fri Aug 19 01:33:54 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2AF9353115 for ; Mon, 15 Aug 2005 06:05:20 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 34799-08 for ; Mon, 15 Aug 2005 09:05:15 +0000 (GMT) Received: from smtp7.wanadoo.fr (smtp7.wanadoo.fr [193.252.22.24]) by svr1.postgresql.org (Postfix) with ESMTP id 7DBFE530D7 for ; Mon, 15 Aug 2005 06:05:14 -0300 (ADT) Received: from me-wanadoo.net (unknown [127.0.0.1]) by mwinf0703.wanadoo.fr (SMTP Server) with ESMTP id 413831000099 for ; Mon, 15 Aug 2005 11:05:15 +0200 (CEST) Received: from NEO (AToulon-151-1-27-207.w83-197.abo.wanadoo.fr [83.197.78.207]) by mwinf0703.wanadoo.fr (SMTP Server) with ESMTP id 0D75E100008F; Mon, 15 Aug 2005 11:05:15 +0200 (CEST) X-ME-UUID: 20050815090515552.0D75E100008F@mwinf0703.wanadoo.fr From: =?iso-8859-1?Q?St=E9phane_COEZ?= To: "'John Arbash Meinel'" Cc: Subject: Re: Performance pb vs SQLServer. Date: Mon, 15 Aug 2005 11:05:11 +0200 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable X-Mailer: Microsoft Office Outlook, Build 11.0.5510 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 Thread-Index: AcWhMC+opyYYrHMVSf2/AMvq1xq1rAAR/teQ In-Reply-To: <42FFE17A.6010208@arbash-meinel.com> Message-Id: <20050815090515.0D75E100008F@mwinf0703.wanadoo.fr> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/238 X-Sequence-Number: 13985 John Arbash Meinel wrote :=20 >=20 > You might also try a different query, something like: >=20 > SELECT DISTINCT cod FROM mytable ORDER BY cod GROUP BY cod;=20 > (You may or may not want order by, or group by, try the different > combinations.) > It might be possible to have the planner realize that all you=20 > want is unique rows, just doing a group by doesn't give you that. >=20 > John > =3D:-> >=20 Thanks John, but using SELECT DISTINCT with or without Order nor Group = by is worth... 30 sec (with index) - stopped at 200 sec without index... So Hash Aggregate is much better than index scan ... > > > >Thanks for help. > >=20 > >St=E9phane COEZ > > > > > > > > > >---------------------------(end of=20 > >broadcast)--------------------------- > >TIP 9: In versions below 8.0, the planner will ignore your desire to > > choose an index scan if your joining column's datatypes do not > > match > > > > =20 > > >=20 >=20 >=20 From pgsql-performance-owner@postgresql.org Fri Aug 19 01:37:41 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 903E653115 for ; Mon, 15 Aug 2005 06:08:09 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 41087-02 for ; Mon, 15 Aug 2005 09:08:08 +0000 (GMT) Received: from smtp7.wanadoo.fr (smtp7.wanadoo.fr [193.252.22.24]) by svr1.postgresql.org (Postfix) with ESMTP id DE86B530D7 for ; Mon, 15 Aug 2005 06:08:07 -0300 (ADT) Received: from me-wanadoo.net (localhost [127.0.0.1]) by mwinf0709.wanadoo.fr (SMTP Server) with ESMTP id BFC8E1C000AA for ; Mon, 15 Aug 2005 11:08:09 +0200 (CEST) Received: from NEO (AToulon-151-1-27-207.w83-197.abo.wanadoo.fr [83.197.78.207]) by mwinf0709.wanadoo.fr (SMTP Server) with ESMTP id 87D941C000A4; Mon, 15 Aug 2005 11:08:09 +0200 (CEST) X-ME-UUID: 20050815090809556.87D941C000A4@mwinf0709.wanadoo.fr From: =?iso-8859-1?Q?St=E9phane_COEZ?= To: Cc: Subject: Re: Performance pb vs SQLServer. Date: Mon, 15 Aug 2005 11:08:06 +0200 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Mailer: Microsoft Office Outlook, Build 11.0.5510 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 Thread-Index: AcWhVWPiS4kw4PktQu2478X/qr7OIAAIyOCg In-Reply-To: <1124085115.26580@netbox.unitech.com.ar> Message-Id: <20050815090809.87D941C000A4@mwinf0709.wanadoo.fr> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/245 X-Sequence-Number: 13992 > > One little thing. Did you shutdown sql2000 while testing > postgresql? Remember that postgresql uses system cache. > Sql2000 uses a large part of memory as buffer and it will not > be available to operating system. I must say that, probably, > results will be the same, but it will be a better test. > Shutting done SQL2000 has no effect on PG performancies. Stephane. From pgsql-performance-owner@postgresql.org Fri Aug 19 01:33:56 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 79FC6530E5 for ; Mon, 15 Aug 2005 06:08:09 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 33767-07 for ; Mon, 15 Aug 2005 09:08:08 +0000 (GMT) Received: from smtp7.wanadoo.fr (smtp7.wanadoo.fr [193.252.22.24]) by svr1.postgresql.org (Postfix) with ESMTP id D66A252F3B for ; Mon, 15 Aug 2005 06:08:07 -0300 (ADT) Received: from me-wanadoo.net (localhost [127.0.0.1]) by mwinf0709.wanadoo.fr (SMTP Server) with ESMTP id 6C2951C000A5 for ; Mon, 15 Aug 2005 11:08:09 +0200 (CEST) Received: from NEO (AToulon-151-1-27-207.w83-197.abo.wanadoo.fr [83.197.78.207]) by mwinf0709.wanadoo.fr (SMTP Server) with ESMTP id 3607E1C000A4; Mon, 15 Aug 2005 11:08:09 +0200 (CEST) X-ME-UUID: 20050815090809221.3607E1C000A4@mwinf0709.wanadoo.fr From: =?iso-8859-1?Q?St=E9phane_COEZ?= To: "'Magnus Hagander'" , Subject: Re: Performance pb vs SQLServer. Date: Mon, 15 Aug 2005 11:08:06 +0200 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Mailer: Microsoft Office Outlook, Build 11.0.5510 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 Thread-Index: AcWedz241y3YYG5NRWucxBfjaAmDhQC+oSZAAAG5cjA= In-Reply-To: <6BCB9D8A16AC4241919521715F4D8BCE6C789B@algol.sollentuna.se> Message-Id: <20050815090809.3607E1C000A4@mwinf0709.wanadoo.fr> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/239 X-Sequence-Number: 13986 > De : Magnus Hagander [mailto:mha@sollentuna.net] > Out of curiosity, what plan do you get from SQLServer? I bet > it's a clustered index scan... > > > //Magnus > I have a Table scan and Hashaggregate... Stephane From pgsql-performance-owner@postgresql.org Mon Aug 15 12:25:35 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8EC1252AC9 for ; Mon, 15 Aug 2005 12:19:35 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 35063-02 for ; Mon, 15 Aug 2005 15:19:34 +0000 (GMT) Received: from email.aon.at (warsl404pip7.highway.telekom.at [195.3.96.91]) by svr1.postgresql.org (Postfix) with ESMTP id 9A36C52AC4 for ; Mon, 15 Aug 2005 12:19:32 -0300 (ADT) Received: (qmail 8838 invoked from network); 15 Aug 2005 15:19:27 -0000 Received: from m148p003.dipool.highway.telekom.at (HELO fermat.Koizar.inf) ([62.46.8.99]) (envelope-sender ) by smarthub73.highway.telekom.at (qmail-ldap-1.03) with SMTP for ; 15 Aug 2005 15:19:27 -0000 From: Manfred Koizar To: John A Meinel Cc: Alvaro Herrera , Postgresql Performance Subject: Re: Odd Locking Problem Date: Mon, 15 Aug 2005 17:19:44 +0200 Message-ID: References: <42FBB6CF.6000404@arbash-meinel.com> <20050811210842.GA28253@alvh.no-ip.org> <42FBBF1E.6040403@arbash-meinel.com> In-Reply-To: <42FBBF1E.6040403@arbash-meinel.com> X-Mailer: Forte Agent 2.0/32.652 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.217 required=5 tests=[AWL=0.167, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/158 X-Sequence-Number: 13905 On Thu, 11 Aug 2005 16:11:58 -0500, John A Meinel wrote: >the insert is occurring into table 'a' not table 'b'. >'a' refers to other tables, but these should not be modified. So your "a" is Alvaro's "b", and one of your referenced tables is Alvaro's "a". This is further supported by the fact that the problem doesn't occur with 8.1. Servus Manfred From pgsql-performance-owner@postgresql.org Mon Aug 15 13:22:09 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5EB9A52B42 for ; Mon, 15 Aug 2005 13:18:36 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 51230-03 for ; Mon, 15 Aug 2005 16:18:35 +0000 (GMT) Received: from alvh.no-ip.org (200-85-218-206.bk4-dsl.surnet.cl [200.85.218.206]) by svr1.postgresql.org (Postfix) with ESMTP id A872D52ADD for ; Mon, 15 Aug 2005 13:18:34 -0300 (ADT) Received: by alvh.no-ip.org (Postfix, from userid 1000) id 872A2C2DC09; Mon, 15 Aug 2005 12:18:59 -0400 (CLT) Date: Mon, 15 Aug 2005 12:18:59 -0400 From: Alvaro Herrera To: Magnus Hagander Cc: dario_d_s@unitech.com.ar, scoez@harrysoftware.com, pgsql-performance@postgresql.org Subject: Re: Performance pb vs SQLServer. Message-ID: <20050815161859.GB22386@alvh.no-ip.org> Mail-Followup-To: Magnus Hagander , dario_d_s@unitech.com.ar, scoez@harrysoftware.com, pgsql-performance@postgresql.org References: <6BCB9D8A16AC4241919521715F4D8BCE6C789C@algol.sollentuna.se> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <6BCB9D8A16AC4241919521715F4D8BCE6C789C@algol.sollentuna.se> User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.736 required=5 tests=[AWL=0.312, DNS_FROM_RFC_ABUSE=0.374, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/159 X-Sequence-Number: 13906 On Mon, Aug 15, 2005 at 10:25:47AM +0200, Magnus Hagander wrote: > SQL 2005 has "MVCC" (they call it something different, of course, but > that's basicallyi what it is) Interesting; do they use an overwriting storage manager like Oracle, or a non-overwriting one like Postgres? -- Alvaro Herrera () "The Postgresql hackers have what I call a "NASA space shot" mentality. Quite refreshing in a world of "weekend drag racer" developers." (Scott Marlowe) From pgsql-performance-owner@postgresql.org Fri Aug 19 01:37:44 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E6940528E6 for ; Mon, 15 Aug 2005 13:25:03 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 52821-03 for ; Mon, 15 Aug 2005 16:25:01 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id BDB06528CB for ; Mon, 15 Aug 2005 13:24:58 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7FGOv0j022541 for ; Mon, 15 Aug 2005 11:24:57 -0500 Subject: I'm configuraing a new system (Bigish) and need some advice. From: Jeremiah Jahn To: postgres performance Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="=-ZiOwI4AYpsu7FZjWOMAk" Date: Mon, 15 Aug 2005 11:24:57 -0500 Message-Id: <1124123097.27881.59.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Received-SPF: pass (penguin.goodinassociates.com: domain of goodinassociates.com designates 206.80.71.242 as permitted sender) client-ip=206.80.71.242; envelope-from=jeremiah@goodinassociates.com; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1021/Mon Aug 15 02:05:53 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 required=5 tests=[FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/246 X-Sequence-Number: 13993 --=-ZiOwI4AYpsu7FZjWOMAk Content-Type: text/plain Content-Transfer-Encoding: quoted-printable The system is a dual Xenon with 6Gig of ram and 14 73Gig 15K u320 scsi drives. Plus 2 raid 1 system dives. RedHat EL ES4 is the OS.=20 Any1 have any suggestions as to the configuration? The database is about 60 Gig's. Should jump to 120 here quite soon. Mus of the searches involve people's names. Through a website. My current setup just doesn't seem to have resulted in the performance kick I wanted. I don't know if it's LVM or what. The strang thing is that My Memory usage stays very LOW for some reason. While on my current production server it stays very high. Also looking for ideas on stipe and extent size. The below is run off of a RAID 10. I have not moved my WAL file yet, but there were no incoming transactions at the time the query was run. My stats on the identity table are set to 1000. > explain analyze select distinct case_category,identity_id,court.name,liti= gant_details.case_id,case_year,date_of_birth,assigned_case_role,litigant_de= tails.court_ori,full_name,litigant_details.actor_id,case_data.type_code,cas= e_data.subtype_code,litigant_details.impound_litigant_data, to_number(trim(= leading case_data.type_code from trim(leading case_data.case_year from case= _data.case_id)),'999999') as seq from identity,court,litigant_details,case_= data where identity.court_ori =3D litigant_details.court_ori and identity.c= ase_id =3D litigant_details.case_id and identity.actor_id =3D litigant_deta= ils.actor_id and court.id =3D identity.court_ori and identity.court_ori =3D= case_data.court_ori and case_data.case_id =3D identity.case_id and identi= ty.court_ori =3D 'IL081025J' and full_name like 'SMITH%' order by full_nam= e; > = = = QUERY PLAN=20 > -------------------------------------------------------------------------= ---------------------------------------------------------------------------= ---------------------------------------------------------------------------= ---------------------------------------------------------------------------= ---------------------------------------------------------------------------= ---------------------------------------------------------------------------= ----------------------- > Unique (cost=3D34042.46..34042.57 rows=3D3 width=3D173) (actual time=3D= 63696.896..63720.193 rows=3D8086 loops=3D1) > -> Sort (cost=3D34042.46..34042.47 rows=3D3 width=3D173) (actual tim= e=3D63696.892..63702.239 rows=3D8086 loops=3D1) > Sort Key: identity.full_name, case_data.case_category, identity.= identity_id, court.name, litigant_details.case_id, case_data.case_year, ide= ntity.date_of_birth, litigant_details.assigned_case_role, litigant_details.= court_ori, litigant_details.actor_id, case_data.type_code, case_data.subtyp= e_code, litigant_details.impound_litigant_data, to_number(ltrim(ltrim((case= _data.case_id)::text, (case_data.case_year)::text), (case_data.type_code)::= text), '999999'::text) > -> Nested Loop (cost=3D0.00..34042.43 rows=3D3 width=3D173) (a= ctual time=3D135.498..63655.542 rows=3D8086 loops=3D1) > -> Nested Loop (cost=3D0.00..34037.02 rows=3D1 width=3D1= 59) (actual time=3D95.760..34637.611 rows=3D8086 loops=3D1) > -> Nested Loop (cost=3D0.00..34033.72 rows=3D1 wid= th=3D138) (actual time=3D89.222..34095.763 rows=3D8086 loops=3D1) > Join Filter: (("outer".case_id)::text =3D ("in= ner".case_id)::text) > -> Index Scan using name_speed on identity (= cost=3D0.00..1708.26 rows=3D8152 width=3D82) (actual time=3D42.589..257.818= rows=3D8092 loops=3D1) > Index Cond: (((full_name)::text >=3D 'SM= ITH'::character varying) AND ((full_name)::text < 'SMITI'::character varyin= g)) > Filter: (((court_ori)::text =3D 'IL08102= 5J'::text) AND ((full_name)::text ~~ 'SMITH%'::text)) > -> Index Scan using lit_actor_speed on litiga= nt_details (cost=3D0.00..3.95 rows=3D1 width=3D81) (actual time=3D4.157..4= .170 rows=3D1 loops=3D8092) > Index Cond: (("outer".actor_id)::text = =3D (litigant_details.actor_id)::text) > Filter: ('IL081025J'::text =3D (court_or= i)::text) > -> Seq Scan on court (cost=3D0.00..3.29 rows=3D1 w= idth=3D33) (actual time=3D0.051..0.058 rows=3D1 loops=3D8086) > Filter: ('IL081025J'::text =3D (id)::text) > -> Index Scan using case_data_pkey on case_data (cost=3D= 0.00..5.36 rows=3D2 width=3D53) (actual time=3D3.569..3.572 rows=3D1 loops= =3D8086) > Index Cond: (('IL081025J'::text =3D (case_data.court= _ori)::text) AND ((case_data.case_id)::text =3D ("outer".case_id)::text)) > Total runtime: 63727.873 ms >=20 >=20 > tcpip_socket =3D true > max_connections =3D 100 > shared_buffers =3D 50000 # min 16, at least max_connections*2, 8= KB each > sort_mem =3D 2024000 # min 64, size in KB > vacuum_mem =3D 819200 # min 1024, size in KB > checkpoint_segments =3D 20 # in logfile segments, min 1, 16MB each > effective_cache_size =3D 3600000 # typically 8KB each > random_page_cost =3D 2 # units are one sequential page fetch c= ost > log_min_duration_statement =3D 10000 # Log all statements whose > lc_messages =3D 'C' # locale for system error message strin= gs > lc_monetary =3D 'C' # locale for monetary formatting > lc_numeric =3D 'C' # locale for number formatting > lc_time =3D 'C' # locale for time formatting Ingrate, n.: A man who bites the hand that feeds him, and then complains of indigestion. --=-ZiOwI4AYpsu7FZjWOMAk Content-Type: application/pgp-signature; name=signature.asc Content-Description: This is a digitally signed message part -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (GNU/Linux) iD8DBQBDAMHZ6caZY8jbHDkRAsGSAJ4q1QMU9LIsj3aaOQ/0jZFnGRStagCeLEDl NYMEcN6rFLROJ+MF0riM0MY= =FjzE -----END PGP SIGNATURE----- --=-ZiOwI4AYpsu7FZjWOMAk-- From pgsql-performance-owner@postgresql.org Fri Aug 19 01:37:07 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3ED6E5293C for ; Mon, 15 Aug 2005 13:29:08 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 51782-08 for ; Mon, 15 Aug 2005 16:29:05 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id E0B4052B48 for ; Mon, 15 Aug 2005 13:29:04 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7FGT4vW022635 for ; Mon, 15 Aug 2005 11:29:04 -0500 Subject: Re: I'm configuraing a new system (Bigish) and need some advice. From: Jeremiah Jahn To: postgres performance Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="=-Ay129qIYc2pXgE16W2hX" Date: Mon, 15 Aug 2005 11:29:04 -0500 Message-Id: <1124123344.27881.63.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Received-SPF: pass (penguin.goodinassociates.com: domain of goodinassociates.com designates 206.80.71.242 as permitted sender) client-ip=206.80.71.242; envelope-from=jeremiah@goodinassociates.com; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1021/Mon Aug 15 02:05:53 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 required=5 tests=[FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/244 X-Sequence-Number: 13991 --=-Ay129qIYc2pXgE16W2hX Content-Type: text/plain Content-Transfer-Encoding: quoted-printable 7.4 is the pg version BTW....going to switch to 8 if it's worth it. Ingrate, n.: A man who bites the hand that feeds him, and then complains of indigestion. --=20 "Don't say yes until I finish talking." -- Darryl F. Zanuck --=-Ay129qIYc2pXgE16W2hX Content-Type: application/pgp-signature; name=signature.asc Content-Description: This is a digitally signed message part -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (GNU/Linux) iD8DBQBDAMLQ6caZY8jbHDkRAj8jAJ4+aN0q3ZjnHNAbAuqlFg3mw6NClwCeK2eS 4SJtEpsubfDREI+LzDJpwJU= =yRiq -----END PGP SIGNATURE----- --=-Ay129qIYc2pXgE16W2hX-- From pgsql-performance-owner@postgresql.org Mon Aug 15 13:29:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0900D52B48 for ; Mon, 15 Aug 2005 13:29:26 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 53761-03 for ; Mon, 15 Aug 2005 16:29:24 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id 33C2352B45 for ; Mon, 15 Aug 2005 13:29:24 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7FGTO3o022664 for ; Mon, 15 Aug 2005 11:29:24 -0500 Subject: Re: I'm configuraing a new system (Bigish) and need some advice. From: Jeremiah Jahn To: postgres performance Content-Type: text/plain Date: Mon, 15 Aug 2005 11:29:23 -0500 Message-Id: <1124123363.27881.65.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Content-Transfer-Encoding: 7bit Received-SPF: none (penguin.goodinassociates.com: 206.80.71.242 is neither permitted nor denied by domain of cs.earlham.edu) client-ip=206.80.71.242; envelope-from=jeremiah@cs.earlham.edu; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1021/Mon Aug 15 02:05:53 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.027 required=5 tests=[AWL=-0.023, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/160 X-Sequence-Number: 13907 7.4 is the pg version BTW....going to switch to 8 if it's worth it. Ingrate, n.: A man who bites the hand that feeds him, and then complains of indigestion. -- "Don't say yes until I finish talking." -- Darryl F. Zanuck -- "Don't say yes until I finish talking." -- Darryl F. Zanuck From pgsql-performance-owner@postgresql.org Mon Aug 15 17:35:16 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id ADE5E52C1A for ; Mon, 15 Aug 2005 17:35:14 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 19831-02 for ; Mon, 15 Aug 2005 20:35:11 +0000 (GMT) Received: from vt-pe2550-001.VANTAGE.vantage.com (unknown [64.80.203.244]) by svr1.postgresql.org (Postfix) with ESMTP id B225552AD8 for ; Mon, 15 Aug 2005 17:35:07 -0300 (ADT) X-MimeOLE: Produced By Microsoft Exchange V6.0.6249.0 content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----_=_NextPart_001_01C5A1D8.D1E68745" Subject: choosing RAID level for xlogs Date: Mon, 15 Aug 2005 16:35:05 -0400 Message-ID: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BDF@vt-pe2550-001.VANTAGE.vantage.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: choosing RAID level for xlogs Thread-Index: AcWh2NDgCbXrlD2dSvuZJkSyYWP7KA== From: "Anjan Dave" To: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.034 required=5 tests=[AWL=0.033, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/161 X-Sequence-Number: 13908 This is a multi-part message in MIME format. ------_=_NextPart_001_01C5A1D8.D1E68745 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Hi, =20 One simple question. For 125 or more checkpoint segments (checkpoint_timeout is 600 seconds, shared_buffers are at 21760 or 170MB) on a very busy database, what is more suitable, a separate 6 disk RAID5 volume, or a RAID10 volume? Databases will be on separate spindles. Disks are 36GB 15KRPM, 2Gb Fiber Channel. Performance is paramount, but I don't want to use RAID0. =20 PG7.4.7 on RHAS 4.0 =20 I can provide more info if needed. =20 Appreciate some recommendations! =20 Thanks, Anjan =20 =20 --- This email message and any included attachments constitute confidential and privileged information intended exclusively for the listed addressee(s). If you are not the intended recipient, please notify Vantage by immediately telephoning 215-579-8390, extension 1158. In addition, please reply to this message confirming your receipt of the same in error. A copy of your email reply can also be sent to support@vantage.com. Please do not disclose, copy, distribute or take any action in reliance on the contents of this information. Kindly destroy all copies of this message and any attachments. Any other use of this email is prohibited. Thank you for your cooperation. For more information about Vantage, please visit our website at http://www.vantage.com . --- =20 ------_=_NextPart_001_01C5A1D8.D1E68745 Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable

Hi,

 

One simple question. For 125 or more checkpoint = segments (checkpoint_timeout is 600 seconds, shared_buffers are at 21760 or 170MB) on a very busy = database, what is more suitable, a separate 6 disk RAID5 volume, or a RAID10 = volume? Databases will be on separate spindles. Disks are 36GB 15KRPM, 2Gb Fiber Channel. Performance is paramount, but I don’t want to use = RAID0.

 

PG7.4.7 on RHAS 4.0

 

I can provide more info if needed.

 

Appreciate some = recommendations!

 

Thanks,

Anjan

 

 
---=

This =
email message and any included attachments constitute confidential and =
privileged information intended exclusively for the =
listed addressee(s). If you are not the intended recipient, please =
notify Vantage by immediately telephoning 215-579-8390, extension 1158. =
In addition, please reply to this message confirming your receipt of the =
same in error. A copy of your email reply can also be sent to =
support@vantage.com. Please do not disclose, copy, distribute or take =
any action in reliance on the contents of this information. Kindly destroy all copies of =
this message and any attachments. Any other use of this email is =
prohibited. Thank you for your cooperation. For more information about Vantage, please visit our =
website at http://www.vantage.com.
---=

 

------_=_NextPart_001_01C5A1D8.D1E68745-- From pgsql-performance-owner@postgresql.org Tue Aug 16 04:01:06 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0DB3C5281F for ; Tue, 16 Aug 2005 04:01:04 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 82895-04 for ; Tue, 16 Aug 2005 07:01:00 +0000 (GMT) Received: from hotmail.com (bay18-f21.bay18.hotmail.com [65.54.187.71]) by svr1.postgresql.org (Postfix) with ESMTP id 68D1B52A78 for ; Tue, 16 Aug 2005 04:00:59 -0300 (ADT) Received: from mail pickup service by hotmail.com with Microsoft SMTPSVC; Tue, 16 Aug 2005 00:01:00 -0700 Message-ID: Received: from 203.155.1.244 by by18fd.bay18.hotmail.msn.com with HTTP; Tue, 16 Aug 2005 07:01:00 GMT X-Originating-IP: [203.155.1.244] X-Originating-Email: [maccran@hotmail.com] X-Sender: maccran@hotmail.com From: "wisan watcharinporn" To: pgsql-performance@postgresql.org Subject: database encoding with index search problem Date: Tue, 16 Aug 2005 07:01:00 +0000 Mime-Version: 1.0 Content-Type: text/plain; format=flowed X-OriginalArrivalTime: 16 Aug 2005 07:01:00.0962 (UTC) FILETIME=[429CA820:01C5A230] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.75 required=5 tests=[DNS_FROM_RFC_ABUSE=0.374, DNS_FROM_RFC_POST=1.376, MSGID_FROM_MTA_HEADER=0] X-Spam-Level: * X-Archive-Number: 200508/162 X-Sequence-Number: 13909 i have problem in database encoding with indexing case :1 when i initdb -E win874 mydb and create database with createdb -E win874 dbname and create table emp with index in empname field i can select sorting name correct in thai alphabet such select empName from emp order by empName; but in select empName from emp where empname like 'xxx%' it not using index scan , it use seq scan so it may slow in find name case :2 when i initdb mydb (use default) and create database with createdb -E win874 dbname and create table emp with index in empname field i can not select sorting name correct in thai alphabet but in select empName from emp where empname like 'xxx%' it using index scan , very fast in find name problem: how can i configure database that can correct in sorting name and using index scan in like 'xxxx%' search using FreeBSD 5.4 postgreql 8.0. _________________________________________________________________ Express yourself instantly with MSN Messenger! Download today it's FREE! http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/ From pgsql-performance-owner@postgresql.org Tue Aug 16 04:29:39 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 85A2352A72 for ; Tue, 16 Aug 2005 04:29:37 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 87417-06 for ; Tue, 16 Aug 2005 07:29:34 +0000 (GMT) Received: from dd01.profihoster.net (dd01.profihoster.net [84.233.130.40]) by svr1.postgresql.org (Postfix) with ESMTP id B4AEE52A3C for ; Tue, 16 Aug 2005 04:29:32 -0300 (ADT) Received: from [192.168.54.114] (Iba10.i.pppool.de [85.73.186.16]) by dd01.profihoster.net (Postfix) with ESMTP id C147A2380EF for ; Tue, 16 Aug 2005 09:29:28 +0200 (CEST) Message-ID: <430195DC.70104@laliluna.de> Date: Tue, 16 Aug 2005 09:29:32 +0200 From: Sebastian Hennebrueder User-Agent: Mozilla Thunderbird 1.0.6 (Windows/20050716) X-Accept-Language: de-DE, de, en-us, en MIME-Version: 1.0 To: postgres performance Subject: Looking for a large database for testing Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/163 X-Sequence-Number: 13910 Hello, I would like to test the performance of my Java/PostgreSQL applications especially when making full text searches. For this I am looking for a database with 50 to 300 MB having text fields. e.g. A table with books with fields holding a comment, table of content or example chapters or what ever else. Does anybody have an idea where I can find a database like this or does even have something like this? -- Best Regards / Viele Gr��e Sebastian Hennebrueder ---- http://www.laliluna.de Tutorials for JSP, JavaServer Faces, Struts, Hibernate and EJB Get support, education and consulting for these technologies - uncomplicated and cheap. From pgsql-performance-owner@postgresql.org Tue Aug 16 05:08:14 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8DB1D52AAE for ; Tue, 16 Aug 2005 05:08:12 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 02082-07 for ; Tue, 16 Aug 2005 08:08:08 +0000 (GMT) Received: from service-web.de (p15093784.pureserver.info [217.160.106.224]) by svr1.postgresql.org (Postfix) with ESMTP id 7E85D52862 for ; Tue, 16 Aug 2005 05:08:07 -0300 (ADT) Received: from [193.97.247.16] (074-016-066-080.eggenet.de [80.66.16.74]) by service-web.de (Postfix) with ESMTP id B125020003A; Tue, 16 Aug 2005 10:08:08 +0200 (CEST) Message-ID: <43019ED4.7070509@wildenhain.de> Date: Tue, 16 Aug 2005 10:07:48 +0200 From: Tino Wildenhain User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050802) X-Accept-Language: de-DE, de, en-us, en MIME-Version: 1.0 To: Sebastian Hennebrueder Cc: postgres performance Subject: Re: Looking for a large database for testing References: <430195DC.70104@laliluna.de> In-Reply-To: <430195DC.70104@laliluna.de> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.058 required=5 tests=[AWL=0.008, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/164 X-Sequence-Number: 13911 Sebastian Hennebrueder schrieb: > Hello, > > I would like to test the performance of my Java/PostgreSQL applications > especially when making full text searches. > For this I am looking for a database with 50 to 300 MB having text fields. > e.g. A table with books with fields holding a comment, table of content > or example chapters > or what ever else. > > Does anybody have an idea where I can find a database like this or does > even have something like this? > You can download the wikipedia content. Just browse the wikimedia site. Its some work to change the data to be able to import into postgres, but at least you have a lot real world data - in many languages. From pgsql-performance-owner@postgresql.org Tue Aug 16 05:24:02 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4D66052AAE for ; Tue, 16 Aug 2005 05:24:00 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 05425-02 for ; Tue, 16 Aug 2005 08:23:56 +0000 (GMT) Received: from dd01.profihoster.net (dd01.profihoster.net [84.233.130.40]) by svr1.postgresql.org (Postfix) with ESMTP id 73D9F52AAB for ; Tue, 16 Aug 2005 05:23:55 -0300 (ADT) Received: from [192.168.54.114] (Iba10.i.pppool.de [85.73.186.16]) by dd01.profihoster.net (Postfix) with ESMTP id 7A5822380EF for ; Tue, 16 Aug 2005 10:23:52 +0200 (CEST) Message-ID: <4301A29C.9010003@laliluna.de> Date: Tue, 16 Aug 2005 10:23:56 +0200 From: Sebastian Hennebrueder User-Agent: Mozilla Thunderbird 1.0.6 (Windows/20050716) X-Accept-Language: de-DE, de, en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Re: Looking for a large database for testing References: <430195DC.70104@laliluna.de> <43019ED4.7070509@wildenhain.de> In-Reply-To: <43019ED4.7070509@wildenhain.de> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/165 X-Sequence-Number: 13912 Tino Wildenhain schrieb: > Sebastian Hennebrueder schrieb: > >> Hello, >> >> I would like to test the performance of my Java/PostgreSQL applications >> especially when making full text searches. >> For this I am looking for a database with 50 to 300 MB having text >> fields. >> e.g. A table with books with fields holding a comment, table of content >> or example chapters >> or what ever else. >> >> Does anybody have an idea where I can find a database like this or does >> even have something like this? >> > You can download the wikipedia content. Just browse the wikimedia site. > Its some work to change the data to be able to import into postgres, > but at least you have a lot real world data - in many languages. I have just found it. Here there is a link http://download.wikimedia.org/ They have content in multiple languages and dumps up to 20 GB. -- Best Regards / Viele Gr��e Sebastian Hennebrueder ---- http://www.laliluna.de Tutorials for JSP, JavaServer Faces, Struts, Hibernate and EJB Get support, education and consulting for these technologies - uncomplicated and cheap. From pgsql-performance-owner@postgresql.org Tue Aug 16 05:25:28 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 90F3F52AA3 for ; Tue, 16 Aug 2005 05:25:26 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 05425-04 for ; Tue, 16 Aug 2005 08:25:22 +0000 (GMT) Received: from mail.metronet.co.uk (mail.metronet.co.uk [213.162.97.75]) by svr1.postgresql.org (Postfix) with ESMTP id 67AE552AB6 for ; Tue, 16 Aug 2005 05:25:21 -0300 (ADT) Received: from mainbox.archonet.com (84-51-143-99.archon037.adsl.metronet.co.uk [84.51.143.99]) by smtp.metronet.co.uk (MetroNet Mail) with ESMTP id D4FDF4163D1; Tue, 16 Aug 2005 09:25:16 +0100 (BST) Received: from localhost (localhost.localdomain [127.0.0.1]) by mainbox.archonet.com (Postfix) with ESMTP id B8D3415EDA; Tue, 16 Aug 2005 09:24:02 +0100 (BST) Received: from mainbox.archonet.com ([127.0.0.1]) by localhost (mainbox [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 18203-07; Tue, 16 Aug 2005 09:24:00 +0100 (BST) Received: from [192.168.1.17] (client17.office.archonet.com [192.168.1.17]) by mainbox.archonet.com (Postfix) with ESMTP id C88D115ED9; Tue, 16 Aug 2005 09:23:59 +0100 (BST) Message-ID: <4301A29E.30703@archonet.com> Date: Tue, 16 Aug 2005 09:23:58 +0100 From: Richard Huxton User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: wisan watcharinporn Cc: pgsql-performance@postgresql.org Subject: Re: database encoding with index search problem References: In-Reply-To: Content-Type: text/plain; charset=ISO-8859-15; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/166 X-Sequence-Number: 13913 wisan watcharinporn wrote: > problem: > how can i configure database that can correct in sorting name and using > index scan in like 'xxxx%' search I think you'll want to read the following then have a quick search of the mailing list archives for "opclass" for some examples. http://www.postgresql.org/docs/8.0/static/sql-createindex.html http://www.postgresql.org/docs/8.0/static/indexes-opclass.html HTH -- Richard Huxton Archonet Ltd From pgsql-performance-owner@postgresql.org Tue Aug 16 05:39:31 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0F39E52A36 for ; Tue, 16 Aug 2005 05:39:30 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 08564-03 for ; Tue, 16 Aug 2005 08:39:26 +0000 (GMT) Received: from heisenberg.zen.co.uk (heisenberg.zen.co.uk [212.23.3.141]) by svr1.postgresql.org (Postfix) with ESMTP id 4476752A2F for ; Tue, 16 Aug 2005 05:39:25 -0300 (ADT) Received: from [62.3.112.246] (helo=gweek.purplebat.com) by heisenberg.zen.co.uk with esmtp (Exim 4.30) id 1E4wyj-0003fP-JW; Tue, 16 Aug 2005 08:39:25 +0000 Received: from mrae by gweek.purplebat.com with local (Exim 4.50) id 1E4wyU-0008C9-Vd; Tue, 16 Aug 2005 09:39:10 +0100 Date: Tue, 16 Aug 2005 09:39:10 +0100 To: Sebastian Hennebrueder Cc: postgres performance Subject: Re: Looking for a large database for testing Message-ID: <20050816083910.GA31460@purplebat.com> References: <430195DC.70104@laliluna.de> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <430195DC.70104@laliluna.de> User-Agent: Mutt/1.5.9i From: Mark Rae X-Originating-Heisenberg-IP: [62.3.112.246] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.033 required=5 tests=[AWL=-0.017, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/167 X-Sequence-Number: 13914 On Tue, Aug 16, 2005 at 09:29:32AM +0200, Sebastian Hennebrueder wrote: > I would like to test the performance of my Java/PostgreSQL applications > especially when making full text searches. > For this I am looking for a database with 50 to 300 MB having text fields. > e.g. A table with books with fields holding a comment, table of content > or example chapters > or what ever else. You could try the OMIM database, which is currently 100M It contains both journal references and large sections of 'plain' text. It also contains a large amount of technical terms which will really test any kind of soundex matching if you are using that. http://www.ncbi.nlm.nih.gov/Omim/omimfaq.html#download Unfortunately it only comes as a flat text file, but is very easy to parse. And if you start reading it, you'll probably learn quite a lot of things you really didn't want to know!! :-D -Mark From pgsql-performance-owner@postgresql.org Tue Aug 16 06:38:58 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BC1D65297B for ; Tue, 16 Aug 2005 06:38:56 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 22542-06 for ; Tue, 16 Aug 2005 09:38:52 +0000 (GMT) Received: from ra.sai.msu.su (ra.sai.msu.su [158.250.29.2]) by svr1.postgresql.org (Postfix) with ESMTP id 8E88252983 for ; Tue, 16 Aug 2005 06:38:42 -0300 (ADT) Received: from ra (ra [158.250.29.2]) by ra.sai.msu.su (8.13.4/8.13.4) with ESMTP id j7G9cf3d018056; Tue, 16 Aug 2005 13:38:41 +0400 (MSD) Date: Tue, 16 Aug 2005 13:38:41 +0400 (MSD) From: Oleg Bartunov X-X-Sender: megera@ra.sai.msu.su To: Sebastian Hennebrueder Cc: postgres performance Subject: Re: Looking for a large database for testing In-Reply-To: <430195DC.70104@laliluna.de> Message-ID: References: <430195DC.70104@laliluna.de> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.326 required=5 tests=[AWL=-0.048, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/168 X-Sequence-Number: 13915 Sebastian, you can try document generator. I used http://www.cs.rmit.edu.au/~jz/resources/finnegan.zip yuo can play with freq. of words and document length distribution. Also, I have SentenceGenerator.java which could be used for generation of synthetic texts. Oleg On Tue, 16 Aug 2005, Sebastian Hennebrueder wrote: > Hello, > > I would like to test the performance of my Java/PostgreSQL applications > especially when making full text searches. > For this I am looking for a database with 50 to 300 MB having text fields. > e.g. A table with books with fields holding a comment, table of content > or example chapters > or what ever else. > > Does anybody have an idea where I can find a database like this or does > even have something like this? > > Regards, Oleg _____________________________________________________________ Oleg Bartunov, sci.researcher, hostmaster of AstroNet, Sternberg Astronomical Institute, Moscow University (Russia) Internet: oleg@sai.msu.su, http://www.sai.msu.su/~megera/ phone: +007(095)939-16-83, +007(095)939-23-83 From pgsql-performance-owner@postgresql.org Tue Aug 16 12:37:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4481552903 for ; Tue, 16 Aug 2005 12:37:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 13965-07 for ; Tue, 16 Aug 2005 15:37:10 +0000 (GMT) Received: from thor.netera.se (thor.netera.se [85.112.172.11]) by svr1.postgresql.org (Postfix) with ESMTP id 943F8528F0 for ; Tue, 16 Aug 2005 12:37:08 -0300 (ADT) Received: from [192.168.2.240] (1-1-1-41a.o.sth.bostream.se [81.26.246.14]) by thor.netera.se (Postfix) with ESMTP id A664713AC078 for ; Tue, 16 Aug 2005 17:37:07 +0200 (CEST) Message-ID: <430208AE.7080100@relevanttraffic.se> Date: Tue, 16 Aug 2005 17:39:26 +0200 From: Ulrich Wisser Organization: Relevant Traffic AB User-Agent: Mozilla Thunderbird 1.0.6 (Windows/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Need for speed Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/169 X-Sequence-Number: 13916 Hello, one of our services is click counting for on line advertising. We do this by importing Apache log files every five minutes. This results in a lot of insert and delete statements. At the same time our customers shall be able to do on line reporting. We have a box with Linux Fedora Core 3, Postgres 7.4.2 Intel(R) Pentium(R) 4 CPU 2.40GHz 2 scsi 76GB disks (15.000RPM, 2ms) I did put pg_xlog on another file system on other discs. Still when several users are on line the reporting gets very slow. Queries can take more then 2 min. I need some ideas how to improve performance in some orders of magnitude. I already thought of a box with the whole database on a ram disc. So really any idea is welcome. Ulrich -- Ulrich Wisser / System Developer RELEVANT TRAFFIC SWEDEN AB, Riddarg 17A, SE-114 57 Sthlm, Sweden Direct (+46)86789755 || Cell (+46)704467893 || Fax (+46)86789769 ________________________________________________________________ http://www.relevanttraffic.com From pgsql-performance-owner@postgresql.org Tue Aug 16 12:59:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AFE7C52800; Tue, 16 Aug 2005 12:59:23 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 19138-05; Tue, 16 Aug 2005 15:59:21 +0000 (GMT) Received: from notes.beauchamp.loxane.fr (unknown [217.167.112.209]) by svr1.postgresql.org (Postfix) with ESMTP id 551C352A07; Tue, 16 Aug 2005 12:59:17 -0300 (ADT) To: Ulrich Wisser Cc: pgsql-performance@postgresql.org, pgsql-performance-owner@postgresql.org Subject: Re. : Need for speed MIME-Version: 1.0 X-Mailer: Lotus Notes Release 5.0.10 March 22, 2002 Message-ID: From: bsimon@loxane.com Date: Tue, 16 Aug 2005 18:02:53 +0200 X-MIMETrack: Serialize by Router on notes/Loxane(Release 5.0.10 |March 22, 2002) at 16/08/2005 18:02:58, Serialize complete at 16/08/2005 18:02:58 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.21 required=5 tests=[AWL=0.032, NO_REAL_NAME=0.178] X-Spam-Level: X-Archive-Number: 200508/170 X-Sequence-Number: 13917 Hi, How much Ram do you have ? Could you give us your postgresql.conf ? (shared buffer parameter) If you do lots of deletes/inserts operations you HAVE to vacuum analyze=20 your table (especially if you have indexes).=20 I'm not sure if vacuuming locks your table with pg 7.4.2 (it doesn't with=20 8.0), you might consider upgrading your pg version.=20 Anyway, your "SELECT" performance while vacuuming is going to be altered. = I don't know your application but I would certainly try to split your=20 table. it would result in one table for inserts/vaccum and one for=20 selects. You would have to switch from one to the other every five=20 minutes. Benjamin. Ulrich Wisser Envoy=E9 par : pgsql-performance-owner@postgresql.org 16/08/2005 17:39 =20 Pour : pgsql-performance@postgresql.org cc :=20 Objet : [PERFORM] Need for speed Hello, one of our services is click counting for on line advertising. We do=20 this by importing Apache log files every five minutes. This results in a=20 lot of insert and delete statements. At the same time our customers=20 shall be able to do on line reporting. We have a box with Linux Fedora Core 3, Postgres 7.4.2 Intel(R) Pentium(R) 4 CPU 2.40GHz 2 scsi 76GB disks (15.000RPM, 2ms) I did put pg=5Fxlog on another file system on other discs. Still when several users are on line the reporting gets very slow.=20 Queries can take more then 2 min. I need some ideas how to improve performance in some orders of=20 magnitude. I already thought of a box with the whole database on a ram=20 disc. So really any idea is welcome. Ulrich --=20 Ulrich Wisser / System Developer RELEVANT TRAFFIC SWEDEN AB, Riddarg 17A, SE-114 57 Sthlm, Sweden Direct (+46)86789755 || Cell (+46)704467893 || Fax (+46)86789769 =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F= =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F= =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F http://www.relevanttraffic.com ---------------------------(end of broadcast)--------------------------- TIP 1: if posting/reading through Usenet, please send an appropriate subscribe-nomail command to majordomo@postgresql.org so that your message can get through to the mailing list cleanly From pgsql-performance-owner@postgresql.org Tue Aug 16 13:05:39 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 405E552A02 for ; Tue, 16 Aug 2005 13:05:37 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 21148-06 for ; Tue, 16 Aug 2005 16:05:33 +0000 (GMT) Received: from mail.metronet.co.uk (mail.metronet.co.uk [213.162.97.75]) by svr1.postgresql.org (Postfix) with ESMTP id F30C952800 for ; Tue, 16 Aug 2005 13:05:32 -0300 (ADT) Received: from mainbox.archonet.com (84-51-143-99.archon037.adsl.metronet.co.uk [84.51.143.99]) by smtp.metronet.co.uk (MetroNet Mail) with ESMTP id B0C10417CD6; Tue, 16 Aug 2005 17:05:18 +0100 (BST) Received: from localhost (localhost.localdomain [127.0.0.1]) by mainbox.archonet.com (Postfix) with ESMTP id AA63715EDB; Tue, 16 Aug 2005 17:03:58 +0100 (BST) Received: from mainbox.archonet.com ([127.0.0.1]) by localhost (mainbox [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 04938-01; Tue, 16 Aug 2005 17:03:56 +0100 (BST) Received: from [192.168.1.17] (client17.office.archonet.com [192.168.1.17]) by mainbox.archonet.com (Postfix) with ESMTP id 0EC0D15ED9; Tue, 16 Aug 2005 17:03:56 +0100 (BST) Message-ID: <43020E6B.9080607@archonet.com> Date: Tue, 16 Aug 2005 17:03:55 +0100 From: Richard Huxton User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Ulrich Wisser Cc: pgsql-performance@postgresql.org Subject: Re: Need for speed References: <430208AE.7080100@relevanttraffic.se> In-Reply-To: <430208AE.7080100@relevanttraffic.se> Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/171 X-Sequence-Number: 13918 Ulrich Wisser wrote: > Hello, > > one of our services is click counting for on line advertising. We do > this by importing Apache log files every five minutes. This results in a > lot of insert and delete statements. At the same time our customers > shall be able to do on line reporting. > I need some ideas how to improve performance in some orders of > magnitude. I already thought of a box with the whole database on a ram > disc. So really any idea is welcome. So what's the problem - poor query plans? CPU saturated? I/O saturated? Too much context-switching? What makes it worse - adding another reporting user, or importing another logfile? -- Richard Huxton Archonet Ltd From pgsql-performance-owner@postgresql.org Tue Aug 16 13:13:27 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1478C52A17 for ; Tue, 16 Aug 2005 13:12:49 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 25765-02 for ; Tue, 16 Aug 2005 16:12:46 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id 7482552A0E for ; Tue, 16 Aug 2005 13:12:45 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050816161244m92009jgjhe>; Tue, 16 Aug 2005 16:12:45 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 17E8355FBA; Tue, 16 Aug 2005 11:12:44 -0500 (CDT) Received: from [192.168.1.119] (71-32-69-23.cdrr.qwest.net [71.32.69.23]) by juju.arbash-meinel.com (Postfix) with ESMTP id 5612255FAD; Tue, 16 Aug 2005 11:12:29 -0500 (CDT) Message-ID: <43021068.7080301@arbash-meinel.com> Date: Tue, 16 Aug 2005 11:12:24 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Ulrich Wisser Cc: pgsql-performance@postgresql.org Subject: Re: Need for speed References: <430208AE.7080100@relevanttraffic.se> In-Reply-To: <430208AE.7080100@relevanttraffic.se> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig6CD64640031BC136768E861A" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 required=5 tests=[FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/173 X-Sequence-Number: 13920 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig6CD64640031BC136768E861A Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Ulrich Wisser wrote: > Hello, > > one of our services is click counting for on line advertising. We do > this by importing Apache log files every five minutes. This results in a > lot of insert and delete statements. At the same time our customers > shall be able to do on line reporting. What are you deleting? I can see having a lot of updates and inserts, but I'm trying to figure out what the deletes would be. Is it just that you completely refill the table based on the apache log, rather than doing only appending? Or are you deleting old rows? > > We have a box with > Linux Fedora Core 3, Postgres 7.4.2 > Intel(R) Pentium(R) 4 CPU 2.40GHz > 2 scsi 76GB disks (15.000RPM, 2ms) > > I did put pg_xlog on another file system on other discs. > > Still when several users are on line the reporting gets very slow. > Queries can take more then 2 min. If it only gets slow when you have multiple clients it sounds like your select speed is the issue, more than conflicting with your insert/deletes. > > I need some ideas how to improve performance in some orders of > magnitude. I already thought of a box with the whole database on a ram > disc. So really any idea is welcome. How much ram do you have in the system? It sounds like you only have 1 CPU, so there is a lot you can do to make the box scale. A dual Opteron (possibly a dual motherboard with dual core (but only fill one for now)), with 16GB of ram, and an 8-drive RAID10 system would perform quite a bit faster. How big is your database on disk? Obviously it isn't very large if you are thinking to hold everything in RAM (and only have 76GB of disk storage to put it in anyway). If your machine only has 512M, an easy solution would be to put in a bunch more memory. In general, your hardware is pretty low in overall specs. So if you are willing to throw money at the problem, there is a lot you can do. Alternatively, turn on statement logging, and then post the queries that are slow. This mailing list is pretty good at fixing poor queries. One thing you are probably hitting is a lot of sequential scans on the main table. If you are doing mostly inserting, make sure you are in a transaction, and think about doing a COPY. There is a lot more that can be said, we just need to have more information about what you want. John =:-> > > Ulrich > > > --------------enig6CD64640031BC136768E861A Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDAhBqJdeBCYSNAAMRAo/0AJ9Y8kXiQfm6BYbWa8enp7/7aqSCAwCeJ0p0 PE1EpGW1XZMH+C7p5+nFTpg= =9tf0 -----END PGP SIGNATURE----- --------------enig6CD64640031BC136768E861A-- From pgsql-performance-owner@postgresql.org Tue Aug 16 13:11:42 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id DB02D52A0F for ; Tue, 16 Aug 2005 13:11:38 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 21537-09 for ; Tue, 16 Aug 2005 16:11:31 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id A9A7552A72 for ; Tue, 16 Aug 2005 13:11:30 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [63.195.55.98] (account josh@agliodbs.com HELO spooky) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7733922; Tue, 16 Aug 2005 09:13:47 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: [SPAM?] Re: PG8 Tuning Date: Tue, 16 Aug 2005 09:12:31 -0700 User-Agent: KMail/1.8 Cc: "Jeffrey W. Baker" , Steve Poe , paul@oxton.com References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> <1123836428.19976.26.camel@amd64-laptop-spoe> <1123779524.6664.1.camel@toonses.gghcwest.com> In-Reply-To: <1123779524.6664.1.camel@toonses.gghcwest.com> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Message-Id: <200508160912.32263.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.012 required=5 tests=[AWL=0.012] X-Spam-Level: X-Archive-Number: 200508/172 X-Sequence-Number: 13919 Jeff, > > 4) pg_xlog: If you're pg_xlog is on a spindle is *only* for pg_xlog > > you're better off. > > Like Mr. Stone said earlier, this is pure dogma. =C2=A0In my experience, > xlogs on the same volume with data is much faster if both are on > battery-backed write-back RAID controller memory. =C2=A0Moving from this > situation to xlogs on a single normal disk is going to be much slower in > most cases. The advice on separate drives for xlog (as is all advice on that web page) = is=20 based on numerous, repeatable tests at OSDL. =20 However, you are absolutely correct in that it's *relative* advice, not=20 absolute advice. If, for example, you're using a $100,000 EMC SAN as your= =20 storage you'll probably be better off giving it everything and letting its= =20 controller and cache handle disk allocation etc. On the other hand, if=20 you're dealing with the 5 drives in a single Dell 6650, I've yet to encount= er=20 a case where a separate xlog disk did not benefit an OLTP application. =46or Solaris, the advantage of using a separate disk or partition is that = the=20 mount options you want for the xlog (including forcedirectio) are=20 considerably different from what you'd use with the main database. =2D-=20 Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Tue Aug 16 13:25:04 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4B0E752888 for ; Tue, 16 Aug 2005 13:25:01 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 27354-07 for ; Tue, 16 Aug 2005 16:24:59 +0000 (GMT) Received: from alvh.no-ip.org (200-85-218-206.bk4-dsl.surnet.cl [200.85.218.206]) by svr1.postgresql.org (Postfix) with ESMTP id B03C05288E for ; Tue, 16 Aug 2005 13:24:56 -0300 (ADT) Received: by alvh.no-ip.org (Postfix, from userid 1000) id 310C4C2DC09; Tue, 16 Aug 2005 12:25:31 -0400 (CLT) Date: Tue, 16 Aug 2005 12:25:31 -0400 From: Alvaro Herrera To: Josh Berkus Cc: pgsql-performance@postgresql.org, "Jeffrey W. Baker" , Steve Poe , paul@oxton.com Subject: Re: [SPAM?] Re: PG8 Tuning Message-ID: <20050816162531.GB32695@alvh.no-ip.org> Mail-Followup-To: Josh Berkus , pgsql-performance@postgresql.org, "Jeffrey W. Baker" , Steve Poe , paul@oxton.com References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> <1123836428.19976.26.camel@amd64-laptop-spoe> <1123779524.6664.1.camel@toonses.gghcwest.com> <200508160912.32263.josh@agliodbs.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <200508160912.32263.josh@agliodbs.com> User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.725 required=5 tests=[AWL=0.301, DNS_FROM_RFC_ABUSE=0.374, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/174 X-Sequence-Number: 13921 On Tue, Aug 16, 2005 at 09:12:31AM -0700, Josh Berkus wrote: > However, you are absolutely correct in that it's *relative* advice, not > absolute advice. If, for example, you're using a $100,000 EMC SAN as your > storage you'll probably be better off giving it everything and letting its > controller and cache handle disk allocation etc. On the other hand, if > you're dealing with the 5 drives in a single Dell 6650, I've yet to encounter > a case where a separate xlog disk did not benefit an OLTP application. I've been asked this a couple of times and I don't know the answer: what happens if you give XLog a single drive (unmirrored single spindle), and that drive dies? So the question really is, should you be giving two disks to XLog? -- Alvaro Herrera () "[PostgreSQL] is a great group; in my opinion it is THE best open source development communities in existence anywhere." (Lamar Owen) From pgsql-performance-owner@postgresql.org Tue Aug 16 13:30:48 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 506305288A for ; Tue, 16 Aug 2005 13:29:27 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 28351-09 for ; Tue, 16 Aug 2005 16:29:23 +0000 (GMT) Received: from hosting.commandprompt.com (128.commandprompt.com [207.173.200.128]) by svr1.postgresql.org (Postfix) with ESMTP id 4126952888 for ; Tue, 16 Aug 2005 13:29:23 -0300 (ADT) Received: from [192.168.1.55] (fc1smp [66.93.38.87]) (authenticated bits=0) by hosting.commandprompt.com (8.13.4/8.13.4) with ESMTP id j7GGRd4c024499; Tue, 16 Aug 2005 09:27:39 -0700 Message-ID: <430214F3.7010503@commandprompt.com> Date: Tue, 16 Aug 2005 09:31:47 -0700 From: "Joshua D. Drake" Organization: Command Prompt, Inc. User-Agent: Mozilla Thunderbird 1.0.6-1.1.fc4 (X11/20050720) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Alvaro Herrera Cc: Josh Berkus , pgsql-performance@postgresql.org, "Jeffrey W. Baker" , Steve Poe , paul@oxton.com Subject: Re: [SPAM?] Re: PG8 Tuning References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> <1123836428.19976.26.camel@amd64-laptop-spoe> <1123779524.6664.1.camel@toonses.gghcwest.com> <200508160912.32263.josh@agliodbs.com> <20050816162531.GB32695@alvh.no-ip.org> In-Reply-To: <20050816162531.GB32695@alvh.no-ip.org> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Greylist: Sender succeded SMTP AUTH authentication, not delayed by milter-greylist-1.6 (hosting.commandprompt.com [192.168.1.101]); Tue, 16 Aug 2005 09:27:39 -0700 (PDT) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.013 required=5 tests=[AWL=0.013] X-Spam-Level: X-Archive-Number: 200508/175 X-Sequence-Number: 13922 > I've been asked this a couple of times and I don't know the answer: what > happens if you give XLog a single drive (unmirrored single spindle), and > that drive dies? So the question really is, should you be giving two > disks to XLog? If that drive dies your restoring from backup. You would need to run at least RAID 1, preferrably RAID 10. Sincerely, Joshua D. Drkae > -- Your PostgreSQL solutions company - Command Prompt, Inc. 1.800.492.2240 PostgreSQL Replication, Consulting, Custom Programming, 24x7 support Managed Services, Shared and Dedicated Hosting Co-Authors: plPHP, plPerlNG - http://www.commandprompt.com/ From pgsql-performance-owner@postgresql.org Tue Aug 16 13:46:52 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2725C52991 for ; Tue, 16 Aug 2005 13:34:06 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 32349-01 for ; Tue, 16 Aug 2005 16:34:03 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id A971A52967 for ; Tue, 16 Aug 2005 13:33:58 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050816163352m92009jfg4e>; Tue, 16 Aug 2005 16:33:58 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 7525D55FBA; Tue, 16 Aug 2005 11:33:52 -0500 (CDT) Received: from [192.168.1.119] (71-32-69-23.cdrr.qwest.net [71.32.69.23]) by juju.arbash-meinel.com (Postfix) with ESMTP id 8373A55FAD; Tue, 16 Aug 2005 11:33:46 -0500 (CDT) Message-ID: <43021567.1000101@arbash-meinel.com> Date: Tue, 16 Aug 2005 11:33:43 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Alvaro Herrera Cc: Josh Berkus , pgsql-performance@postgresql.org, "Jeffrey W. Baker" , Steve Poe , paul@oxton.com Subject: Re: PG8 Tuning References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> <1123836428.19976.26.camel@amd64-laptop-spoe> <1123779524.6664.1.camel@toonses.gghcwest.com> <200508160912.32263.josh@agliodbs.com> <20050816162531.GB32695@alvh.no-ip.org> In-Reply-To: <20050816162531.GB32695@alvh.no-ip.org> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigB40C2E18404ED09A81E307A1" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.038 required=5 tests=[AWL=-0.013, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/177 X-Sequence-Number: 13924 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigB40C2E18404ED09A81E307A1 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Alvaro Herrera wrote: > On Tue, Aug 16, 2005 at 09:12:31AM -0700, Josh Berkus wrote: > > >>However, you are absolutely correct in that it's *relative* advice, not >>absolute advice. If, for example, you're using a $100,000 EMC SAN as your >>storage you'll probably be better off giving it everything and letting its >>controller and cache handle disk allocation etc. On the other hand, if >>you're dealing with the 5 drives in a single Dell 6650, I've yet to encounter >>a case where a separate xlog disk did not benefit an OLTP application. > > > I've been asked this a couple of times and I don't know the answer: what > happens if you give XLog a single drive (unmirrored single spindle), and > that drive dies? So the question really is, should you be giving two > disks to XLog? > I can propose a simple test. Create a test database. Run postgres, insert a bunch of stuff. Stop postgres. Delete everything in the pg_xlog directory. Start postgres again, what does it do? I suppose to simulate more of a failure mode, you could kill -9 the postmaster (and all children processes) perhaps during an insert, and then delete pg_xlog. But I would like to hear from the postgres folks what they *expect* would happen if you ever lost pg_xlog. What about something like keeping pg_xlog on a ramdisk, and then rsyncing it to a hard-disk every 5 minutes. If you die in the middle, does it just restore back to the 5-minutes ago point, or does it get more thoroughly messed up? For some people, a 5-minute old restore would be okay, as long as you still have transaction safety, so that you can figure out what needs to be restored. John =:-> --------------enigB40C2E18404ED09A81E307A1 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDAhVnJdeBCYSNAAMRAsGuAKC2vFtT4mOyAH5nKLN4kggMF7hUeQCgyYRa yN/jfacttQyc3tNPvQTK15o= =XI9u -----END PGP SIGNATURE----- --------------enigB40C2E18404ED09A81E307A1-- From pgsql-performance-owner@postgresql.org Tue Aug 16 13:46:14 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 65C0252BA7 for ; Tue, 16 Aug 2005 13:43:12 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 34733-03 for ; Tue, 16 Aug 2005 16:43:06 +0000 (GMT) Received: from gghcwest.com (adsl-71-128-90-172.dsl.pltn13.pacbell.net [71.128.90.172]) by svr1.postgresql.org (Postfix) with ESMTP id 3B41152B75 for ; Tue, 16 Aug 2005 13:43:04 -0300 (ADT) Received: from toonses.gghcwest.com (toonses.gghcwest.com [192.168.168.115]) by gghcwest.com (8.12.10/8.12.9) with ESMTP id j7GGh0md027531; Tue, 16 Aug 2005 09:43:02 -0700 Received: from jwb by toonses.gghcwest.com with local (Exim 4.50) id 1E54Wj-00049f-TV; Tue, 16 Aug 2005 09:43:01 -0700 Subject: Re: Need for speed From: "Jeffrey W. Baker" To: Ulrich Wisser Cc: pgsql-performance@postgresql.org In-Reply-To: <430208AE.7080100@relevanttraffic.se> References: <430208AE.7080100@relevanttraffic.se> Content-Type: text/plain Content-Transfer-Encoding: 7bit Date: Tue, 16 Aug 2005 09:43:01 -0700 Message-Id: <1124210581.15880.10.camel@toonses.gghcwest.com> Mime-Version: 1.0 X-Mailer: Evolution 2.3.6.1 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=-1.019 required=5 tests=[AWL=-1.069, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/176 X-Sequence-Number: 13923 On Tue, 2005-08-16 at 17:39 +0200, Ulrich Wisser wrote: > Hello, > > one of our services is click counting for on line advertising. We do > this by importing Apache log files every five minutes. This results in a > lot of insert and delete statements. At the same time our customers > shall be able to do on line reporting. > > We have a box with > Linux Fedora Core 3, Postgres 7.4.2 > Intel(R) Pentium(R) 4 CPU 2.40GHz This is not a good CPU for this workload. Try an Opteron or Xeon. Also of major importance is the amount of memory. If possible, you would like to have memory larger than the size of your database. > 2 scsi 76GB disks (15.000RPM, 2ms) If you decide your application is I/O bound, here's an obvious place for improvement. More disks == faster. > I did put pg_xlog on another file system on other discs. Did that have a beneficial effect? > Still when several users are on line the reporting gets very slow. > Queries can take more then 2 min. Is this all the time or only during the insert? > I need some ideas how to improve performance in some orders of > magnitude. I already thought of a box with the whole database on a ram > disc. So really any idea is welcome. You don't need a RAM disk, just a lot of RAM. Your operating system will cache disk contents in memory if possible. You have a very small configuration, so more CPU, more memory, and especially more disks will probably all yield improvements. From pgsql-performance-owner@postgresql.org Fri Aug 19 01:35:07 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 35F7852C86 for ; Tue, 16 Aug 2005 14:01:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 42827-07 for ; Tue, 16 Aug 2005 17:01:20 +0000 (GMT) Received: from sj1-exch-01.us.corp.kailea.com (mail.ragingnet.net [209.249.149.2]) by svr1.postgresql.org (Postfix) with ESMTP id 013CE52C50 for ; Tue, 16 Aug 2005 14:01:14 -0300 (ADT) X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: base64 Subject: Re: Need for speed Date: Tue, 16 Aug 2005 10:01:14 -0700 Message-ID: X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Need for speed Thread-Index: AcWifb5K1mERc9seQ8KmxIu2FCJWUgABcK1Q From: "Roger Hand" To: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 required=5 tests=[FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/243 X-Sequence-Number: 13990 PiBVbHJpY2ggV2lzc2VyIHdyb3RlOg0KPiA+DQo+ID4gb25lIG9mIG91ciBzZXJ2aWNlcyBpcyBj bGljayBjb3VudGluZyBmb3Igb24gbGluZSBhZHZlcnRpc2luZy4gV2UgZG8NCj4gPiB0aGlzIGJ5 IGltcG9ydGluZyBBcGFjaGUgbG9nIGZpbGVzIGV2ZXJ5IGZpdmUgbWludXRlcy4gVGhpcyByZXN1 bHRzIGluIGENCj4gPiBsb3Qgb2YgaW5zZXJ0IGFuZCBkZWxldGUgc3RhdGVtZW50cy4gDQouLi4N Cj4gSWYgeW91IGFyZSBkb2luZyBtb3N0bHkgaW5zZXJ0aW5nLCBtYWtlIHN1cmUgeW91IGFyZSBp biBhIHRyYW5zYWN0aW9uLA0KDQpXZWxsLCB5ZXMsIGJ1dCB5b3UgbWF5IG5lZWQgdG8gbWFrZSBz dXJlIHRoYXQgYSBzaW5nbGUgdHJhbnNhY3Rpb24gZG9lc24ndCBoYXZlIHRvbyBtYW55IGluc2Vy dHMgaW4gaXQuDQpJIHdhcyBoYXZpbmcgYSBwZXJmb3JtYW5jZSBwcm9ibGVtIHdoZW4gZG9pbmcg dHJhbnNhY3Rpb25zIHdpdGggYSBodWdlIG51bWJlciBvZiBpbnNlcnRzDQoodGVucyBvZiB0aG91 c2FuZHMpLCBhbmQgSSBzb2x2ZWQgdGhlIHByb2JsZW0gYnkgcHV0dGluZyBhIHNpbXBsZSBjb3Vu dGVyIGluIHRoZSBsb29wIChpbiB0aGUgSmF2YSBpbXBvcnQgY29kZSwgDQp0aGF0IGlzKSBhbmQg ZG9pbmcgYSBjb21taXQgZXZlcnkgMTAwIG9yIHNvIGluc2VydHMuDQoNCi1Sb2dlcg0KDQo+IEpv aG4NCj4NCj4gPiBVbHJpY2gNCg== From pgsql-performance-owner@postgresql.org Tue Aug 16 14:38:03 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 05B3552927 for ; Tue, 16 Aug 2005 14:38:02 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 48386-10 for ; Tue, 16 Aug 2005 17:37:59 +0000 (GMT) Received: from vms044pub.verizon.net (vms044pub.verizon.net [206.46.252.44]) by svr1.postgresql.org (Postfix) with ESMTP id B1FD65291A for ; Tue, 16 Aug 2005 14:37:58 -0300 (ADT) Received: from osgiliath.mathom.us ([70.108.53.154]) by vms044.mailsrvcs.net (Sun Java System Messaging Server 6.2 HotFix 0.04 (built Dec 24 2004)) with ESMTPA id <0ILB00BMGSNB0RD4@vms044.mailsrvcs.net> for pgsql-performance@postgresql.org; Tue, 16 Aug 2005 12:30:49 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by osgiliath.mathom.us (Postfix) with ESMTP id 76F896B811B for ; Tue, 16 Aug 2005 13:30:47 -0400 (EDT) Received: from osgiliath.mathom.us ([127.0.0.1]) by localhost (osgiliath [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 07205-01-2 for ; Tue, 16 Aug 2005 13:30:47 -0400 (EDT) Received: by osgiliath.mathom.us (Postfix, from userid 1000) id 5955E6B8118; Tue, 16 Aug 2005 13:30:47 -0400 (EDT) Date: Tue, 16 Aug 2005 13:30:47 -0400 From: Michael Stone Subject: Re: [SPAM?] Re: PG8 Tuning In-reply-to: <200508160912.32263.josh@agliodbs.com> To: pgsql-performance@postgresql.org Mail-Followup-To: pgsql-performance@postgresql.org Message-id: <20050816173047.GC19080@mathom.us> MIME-version: 1.0 Content-type: text/plain; charset=us-ascii; format=flowed Content-disposition: inline X-Pgp-Fingerprint: 53 FF 38 00 E7 DD 0A 9C 84 52 84 C5 EE DF 7C 88 X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mathom.us References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> <1123836428.19976.26.camel@amd64-laptop-spoe> <1123779524.6664.1.camel@toonses.gghcwest.com> <200508160912.32263.josh@agliodbs.com> User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/178 X-Sequence-Number: 13925 On Tue, Aug 16, 2005 at 09:12:31AM -0700, Josh Berkus wrote: >However, you are absolutely correct in that it's *relative* advice, not >absolute advice. If, for example, you're using a $100,000 EMC SAN as your >storage you'll probably be better off giving it everything and letting its >controller and cache handle disk allocation etc. Well, you don't have to spend *quite* that much to get a decent storage array. :) >On the other hand, if you're dealing with the 5 drives in a single Dell >6650, I've yet to encounter a case where a separate xlog disk did not >benefit an OLTP application. IIRC, that's an older raid controller that tops out at 128MB write cache, and 5 spindles ain't a lot--so it makes sense that it would benefit from a seperate spindle for xlog. Also note that I said the write cache advice goes out the window if you have a workload that involves constant writing (or if your xlog writes come in faster than your write cache can drain) because at that point you essentially drop back to raw disk speed; I assume the OLTP apps you mention are fairly write-intensive. OTOH, in a reasonably safe configuration I suppose you'd end up with a 3 disk raid5 / 2 disk raid1 or 2 raid 1 pairs on that dell 6650; is that how you test? Once you're down to that small a data set I'd expect the system's ram cache to be a much larger percentage of the working set, which would tend to make the xlog just about the *only* latency-critical i/o. That's a different creature from a data mining app that might really benefit from having additional spindles to accelerate read performance from indices much larger than RAM. At any rate, this just underscores the need for testing a particular workload on particular hardware. Things like the disk speed, raid configuration, write cache size, transaction size, data set size, working set size, concurrent transactions, read vs write emphasis, etc., are each going to have a fairly large impact on performance. >For Solaris, the advantage of using a separate disk or partition is that the >mount options you want for the xlog (including forcedirectio) are >considerably different from what you'd use with the main database. Yeah, having a seperate partition is often good even if you do have everything on the same disks. Mike Stone From pgsql-performance-owner@postgresql.org Fri Aug 19 01:34:47 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CB2FB5291A for ; Tue, 16 Aug 2005 14:46:49 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 52752-07 for ; Tue, 16 Aug 2005 17:46:46 +0000 (GMT) Received: from sj1-exch-01.us.corp.kailea.com (mail.ragingnet.net [209.249.149.2]) by svr1.postgresql.org (Postfix) with ESMTP id 18D6652899 for ; Tue, 16 Aug 2005 14:46:45 -0300 (ADT) X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Subject: Query plan looks OK, but slow I/O - settings advice? Date: Tue, 16 Aug 2005 10:46:46 -0700 Message-ID: X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: Query plan looks OK, but slow I/O - settings advice? Thread-Index: AcWiiQlymkx6ErqSQZWErqyzbLeYrg== From: "Roger Hand" To: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.054 required=5 tests=[AWL=0.003, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/241 X-Sequence-Number: 13988 Summary =3D=3D=3D=3D=3D=3D=3D We are writing to the db pretty much 24 hours a day. Recently the amount of data we write has increased, and the query speed, = formerly okay, has taken a dive. The query is using the indexes as expected, so I don't _think_ I have a = query tuning issue, just an io problem.=20 The first time a query is done it takes about 60 seconds. The second = time it runs in about 6 seconds. What I know I need advice on is io settings and various buffer settings. = I may also need advice on other things, but just don't know it yet! Below is ... - an explain analyze - details of the db setup and hardware - some vmstat and iostat output showing the disks are very busy - the SHOW ALL output for the db config. Details =3D=3D=3D=3D=3D=3D=3D Postgres 8.0.3 Below is a sample query. (This is actually implemented as a prepared = statement. Here I fill in the '?'s with actual values.) electric=3D# EXPLAIN ANALYZE electric-# SELECT datavalue, logfielddatatype, timestamp FROM = logdata_recent=20 electric-# WHERE (logfielddatatype =3D 70 OR logfielddatatype =3D 71 OR = logfielddatatype =3D 69)=20 electric-# AND graphtargetlog =3D 1327=20 electric-# AND timestamp >=3D 1123052400 AND timestamp <=3D 1123138800=20 electric-# ORDER BY timestamp; = = QUERY PLAN = = = =20 -------------------------------------------------- Sort (cost=3D82.48..82.50 rows=3D6 width=3D14) (actual = time=3D60208.968..60211.232 rows=3D2625 loops=3D1) Sort Key: public.logdata_recent."timestamp" -> Result (cost=3D0.00..82.41 rows=3D6 width=3D14) (actual = time=3D52.483..60200.868 rows=3D2625 loops=3D1) -> Append (cost=3D0.00..82.41 rows=3D6 width=3D14) (actual = time=3D52.476..60189.929 rows=3D2625 loops=3D1) -> Seq Scan on logdata_recent (cost=3D0.00..46.25 = rows=3D1 width=3D14) (actual time=3D0.003..0.003 rows=3D0 loops=3D1) Filter: (((logfielddatatype =3D 70) OR = (logfielddatatype =3D 71) OR (logfielddatatype =3D 69)) AND = (graphtargetlog =3D 1327) AND ("timestamp" >=3D 1123052400) AND = ("timestamp" <=3D 1123138800)) -> Index Scan using = logdata_recent_1123085306_ix_t_fld_gtl, = logdata_recent_1123085306_ix_t_fld_gtl, = logdata_recent_1123085306_ix_t_fld_gtl on logdata_recent_stale = logdata_recent (cost=3D0.00..18.08 rows=3D3 width=3D14) (actual = time=3D52.465..60181.624 rows=3D2625 loops=3D1) Index Cond: ((("timestamp" >=3D 1123052400) AND = ("timestamp" <=3D 1123138800) AND (logfielddatatype =3D 70) AND = (graphtargetlog =3D 1327)) OR (("timestamp" >=3D 1123052400) AND = ("timestamp" <=3D 1123138800) AND (logfielddatatype =3D 71) AND = (graphtargetlog =3D 1327)) OR (("timestamp" >=3D 1123052400) AND = ("timestamp" <=3D 1123138800) AND (logfielddatatype =3D 69) AND = (graphtargetlog =3D 1327))) Filter: (((logfielddatatype =3D 70) OR = (logfielddatatype =3D 71) OR (logfielddatatype =3D 69)) AND = (graphtargetlog =3D 1327) AND ("timestamp" >=3D 1123052400) AND = ("timestamp" <=3D 1123138800)) -> Index Scan using = logdata_recent_1123139634_ix_t_fld_gtl, = logdata_recent_1123139634_ix_t_fld_gtl, = logdata_recent_1123139634_ix_t_fld_gtl on logdata_recent_active = logdata_recent (cost=3D0.00..18.08 rows=3D2 width=3D14) (actual = time=3D0.178..0.178 rows=3D0 loops=3D1) Index Cond: ((("timestamp" >=3D 1123052400) AND = ("timestamp" <=3D 1123138800) AND (logfielddatatype =3D 70) AND = (graphtargetlog =3D 1327)) OR (("timestamp" >=3D 1123052400) AND = ("timestamp" <=3D 1123138800) AND (logfielddatatype =3D 71) AND = (graphtargetlog =3D 1327)) OR (("timestamp" >=3D 1123052400) AND = ("timestamp" <=3D 1123138800) AND (logfielddatatype =3D 69) AND = (graphtargetlog =3D 1327))) Filter: (((logfielddatatype =3D 70) OR = (logfielddatatype =3D 71) OR (logfielddatatype =3D 69)) AND = (graphtargetlog =3D 1327) AND ("timestamp" >=3D 1123052400) AND = ("timestamp" <=3D 1123138800)) Total runtime: 60214.545 ms (13 rows) 60 seconds is much longer than it used to be. I would guess it used to = be under 10 seconds. The second time the above query is run we see the = magic of caching as the time goes down to 6 seconds. logdata_recent_active and logdata_recent_stale are inherited tables of = logdata_recent, which never has any data. (This is pseudo-partitioning = in action!) So the very quick seq_scan on the empty logdata_recent parent table is = okay with me. The index is built on timestamp, logfielddatatype, graphtargetlog. I am = curious as to why the same index shows up 3 times in the "using" clause, = but can live without knowing the details as long as it doesn't indicate = that something's wrong. The logdata_recent_stale table has 5 millions rows. The size of the = table itself, on disk, is 324MB. The size of the index is 210MB. The disks are ext3 with journalling type of ordered, but this was later = changed to writeback with no apparent change in speed. They're on a Dell poweredge 6650 with LSI raid card, setup as follows: 4 disks raid 10 for indexes (145GB) - sdc1 6 disks raid 10 for data (220GB) - sdd1 2 mirrored disks for logs - sdb1 stripe size is 32k cache policy: cached io (am told the controller has bbu) write policy: write-back read policy: readahead The partition names do what they say ... [root@rage-db2 /dbdata01]$ df Filesystem 1K-blocks Used Available Use% Mounted on /dev/sdb1 70430588 729324 66123592 2% /dblog01 /dev/sdc1 140861236 19472588 114233300 15% /dbindex01 /dev/sdd1 211299960 157159988 43406548 79% /dbdata01 ... Using iostat (the version from http://linux.inet.hr/) I saw at one point = that the data disk was 100% busy. I believe this was when running the above query, or similar, but in any = case the system is always busy with both reads and (usually) writes. device mgr/s mgw/s r/s w/s kr/s kw/s size queue wait = svc_t %b=20 sdb1 0 61 0.1 15.5 0.4 305.7 19.6 0.1 5.8 = 4.9 8=20 sdc1 21 22 20.6 17.7 164.6 158.6 8.4 1.1 28.3 = 6.2 24=20 sdd1 1742 11 1904.7 6.6 14585.6 71.5 7.7 20.6 10.8 = 0.5 100=20 Another time, when I was running the query above, the index partition = went to 90+% busy for 40 seconds: device mgr/s mgw/s r/s w/s kr/s kw/s size queue wait = svc_t %b=20 sdb1 0 0 0.0 0.2 0.0 0.8 4.0 0.0 20.0 = 15.0 0=20 sdc1 366 53 687.0 66.1 4213.1 483.0 6.2 11.8 15.7 = 1.3 96=20 sdd1 8 17 16.6 13.9 99.5 125.4 7.4 0.7 23.0 = 1.9 6=20 On another occasion (when the query took 24 seconds) I ran vmstat and = iostat every 5 seconds from just before the query until just after. About the first two outputs = are before the query. In this case the index disk is maxed. [root@rage-db2 ~]$ vmstat 5 16 procs memory swap io system = cpu r b swpd free buff cache si so bi bo in cs us sy = wa id 0 0 92 1233500 225692 9578564 0 0 0 0 1 1 2 = 2 1 1 0 1 92 1218460 225748 9595136 0 0 3322 18 655 898 0 = 0 20 79 0 1 92 1202124 225780 9616140 0 0 4204 58 920 1291 0 = 1 24 76 0 1 92 1172876 225820 9645348 0 0 5847 120 1053 1482 0 = 1 23 76 1 0 92 1151712 225836 9666504 0 0 4234 7 847 1239 2 = 1 18 78 1 0 92 1140860 225844 9677436 0 0 2153 500 575 2027 13 = 2 11 73 1 0 92 1140852 225848 9677636 0 0 0 506 213 442 10 = 1 0 89 [root@rage-db2 ~]$ /usr/local/bin/iostat -Px 5 16 device mgr/s mgw/s r/s w/s kr/s kw/s size queue wait = svc_t %b=20 sdb1 1 243 0.1 105.5 2.7 37.6 0.4 0.0 0.2 = 0.1 1=20 sdc1 6 111 3.7 75.9 38.3 769.3 10.1 0.0 0.3 = 0.2 1=20 sdd1 255 107 85.2 37.6 4.9 581.0 4.8 0.0 0.1 = 0.0 1=20 device mgr/s mgw/s r/s w/s kr/s kw/s size queue wait = svc_t %b=20 sdb1 0 0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 = 0.0 0=20 sdc1 273 0 414.0 0.4 2747.7 2.4 6.6 1.6 3.9 = 1.7 69=20 sdd1 0 1 1.4 0.4 7.2 5.6 7.1 0.0 10.0 = 6.7 1=20 device mgr/s mgw/s r/s w/s kr/s kw/s size queue wait = svc_t %b=20 sdb1 0 0 0.0 0.4 0.0 1.6 4.0 0.0 10.0 = 5.0 0=20 sdc1 225 4 777.1 4.6 4011.0 35.1 5.2 2.5 3.2 = 1.3 99=20 sdd1 0 2 0.0 2.6 0.0 16.8 6.5 0.0 8.5 = 0.8 0=20 device mgr/s mgw/s r/s w/s kr/s kw/s size queue wait = svc_t %b=20 sdb1 0 0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 = 0.0 0=20 sdc1 508 7 917.8 7.4 5703.0 58.3 6.2 2.2 2.4 = 1.1 98=20 sdd1 0 4 0.0 6.8 0.0 44.7 6.6 0.1 15.6 = 0.6 0=20 device mgr/s mgw/s r/s w/s kr/s kw/s size queue wait = svc_t %b=20 sdb1 0 0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 = 0.0 0=20 sdc1 361 0 737.5 0.4 4391.7 2.4 6.0 1.8 2.4 = 1.0 76=20 sdd1 0 0 0.0 0.4 0.0 2.4 6.0 0.0 0.0 = 0.0 0=20 device mgr/s mgw/s r/s w/s kr/s kw/s size queue wait = svc_t %b=20 sdb1 0 87 0.0 17.8 0.0 418.3 23.6 0.0 1.3 = 1.2 2=20 sdc1 216 2 489.5 0.4 2821.7 11.2 5.8 1.2 2.4 = 1.1 56=20 sdd1 2 4 7.2 0.6 37.5 18.4 7.2 0.0 6.2 = 3.3 3=20 device mgr/s mgw/s r/s w/s kr/s kw/s size queue wait = svc_t %b=20 sdb1 0 89 0.0 22.4 0.0 446.3 20.0 0.0 1.1 = 0.8 2=20 sdc1 0 4 0.0 1.0 0.0 18.4 18.4 0.0 0.0 = 0.0 0=20 sdd1 0 6 0.0 1.0 0.0 27.1 27.2 0.0 0.0 = 0.0 0=20 device mgr/s mgw/s r/s w/s kr/s kw/s size queue wait = svc_t %b=20 sdb1 0 89 0.0 22.5 0.0 446.2 19.8 0.0 0.4 = 0.3 1=20 sdc1 0 2 0.0 0.4 0.0 9.6 24.0 0.0 0.0 = 0.0 0=20 sdd1 0 4 0.0 0.6 0.0 20.0 33.3 0.0 0.0 = 0.0 0=20 Finally, here's a show all: add_missing_from | on archive_command | unset australian_timezones | off authentication_timeout | 60 bgwriter_delay | 200 bgwriter_maxpages | 100 bgwriter_percent | 1 block_size | 8192 check_function_bodies | on checkpoint_segments | 20 checkpoint_timeout | 300 checkpoint_warning | 30 client_encoding | UNICODE client_min_messages | notice commit_delay | 350 commit_siblings | 5 config_file | /dbdata01/pgdata/postgresql.conf cpu_index_tuple_cost | 0.001 cpu_operator_cost | 0.0025 cpu_tuple_cost | 0.01 custom_variable_classes | unset data_directory | /dbdata01/pgdata DateStyle | ISO, MDY db_user_namespace | off deadlock_timeout | 1000 debug_pretty_print | off debug_print_parse | off debug_print_plan | off debug_print_rewritten | off debug_shared_buffers | 0 default_statistics_target | 50 default_tablespace | unset default_transaction_isolation | read committed default_transaction_read_only | off default_with_oids | off dynamic_library_path | $libdir effective_cache_size | 48000 enable_hashagg | on enable_hashjoin | on enable_indexscan | on enable_mergejoin | on enable_nestloop | on enable_seqscan | on enable_sort | on enable_tidscan | on explain_pretty_print | on external_pid_file | unset extra_float_digits | 0 from_collapse_limit | 8 fsync | on geqo | on geqo_effort | 5 geqo_generations | 0 geqo_pool_size | 0 geqo_selection_bias | 2 geqo_threshold | 12 hba_file | /dbdata01/pgdata/pg_hba.conf ident_file | /dbdata01/pgdata/pg_ident.conf integer_datetimes | off join_collapse_limit | 8 krb_server_keyfile | unset lc_collate | en_US.UTF-8 lc_ctype | en_US.UTF-8 lc_messages | en_US.UTF-8 lc_monetary | en_US.UTF-8 lc_numeric | en_US.UTF-8 lc_time | en_US.UTF-8 listen_addresses | * log_connections | off log_destination | stderr log_directory | /dblog01 log_disconnections | off log_duration | off log_error_verbosity | default log_executor_stats | off log_filename | postgresql-%Y-%m-%d_%H%M%S.log log_hostname | off log_line_prefix | unset log_min_duration_statement | -1 log_min_error_statement | panic log_min_messages | notice log_parser_stats | off log_planner_stats | off log_rotation_age | 1440 log_rotation_size | 10240 log_statement | none log_statement_stats | off log_truncate_on_rotation | off maintenance_work_mem | 262144 max_connections | 40 max_files_per_process | 1000 max_fsm_pages | 100000 max_fsm_relations | 1000 max_function_args | 32 max_identifier_length | 63 max_index_keys | 32 max_locks_per_transaction | 64 max_stack_depth | 2048 password_encryption | on port | 5432 pre_auth_delay | 0 preload_libraries | unset random_page_cost | 4 redirect_stderr | on regex_flavor | advanced rendezvous_name | unset search_path | $user,public server_encoding | UNICODE server_version | 8.0.3 shared_buffers | 10240 silent_mode | off sql_inheritance | on ssl | off statement_timeout | 0 stats_block_level | off stats_command_string | off stats_reset_on_server_start | on stats_row_level | on stats_start_collector | on superuser_reserved_connections | 2 syslog_facility | LOCAL0 syslog_ident | postgres TimeZone | PST8PDT trace_notify | off transaction_isolation | read committed transaction_read_only | off transform_null_equals | off unix_socket_directory | unset unix_socket_group | unset unix_socket_permissions | 511 vacuum_cost_delay | 180 vacuum_cost_limit | 200 vacuum_cost_page_dirty | 20 vacuum_cost_page_hit | 1 vacuum_cost_page_miss | 10 wal_buffers | 8 wal_sync_method | fdatasync work_mem | 98304 zero_damaged_pages | off From pgsql-performance-owner@postgresql.org Tue Aug 16 14:59:57 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A93FF52862 for ; Tue, 16 Aug 2005 14:59:55 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 54479-08 for ; Tue, 16 Aug 2005 17:59:54 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.200]) by svr1.postgresql.org (Postfix) with ESMTP id E144E5282E for ; Tue, 16 Aug 2005 14:59:53 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i22so1232900wra for ; Tue, 16 Aug 2005 10:59:54 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=mHxzDy9dqCkijBt/ujQWiXMjEQeiB3ahk3mVaSeSOrBm1YdVKmlEjlZoVWSQQHJEo2XmdE6Wa4JF//3vwX2i65EFJanqxLmjNPenDT3Mqf2ZgLJf4HLSl0pmmBgPVb0rGkZ7Q7WEV79NYZuHXr+mXv7KJvYjy7Xac3NiDjNVKTs= Received: by 10.54.144.9 with SMTP id r9mr4401173wrd; Tue, 16 Aug 2005 10:59:54 -0700 (PDT) Received: by 10.54.86.15 with HTTP; Tue, 16 Aug 2005 10:59:53 -0700 (PDT) Message-ID: <33c6269f050816105957ae5d6f@mail.gmail.com> Date: Tue, 16 Aug 2005 13:59:53 -0400 From: Alex Turner To: Ulrich Wisser Subject: Re: Need for speed Cc: pgsql-performance@postgresql.org In-Reply-To: <430208AE.7080100@relevanttraffic.se> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <430208AE.7080100@relevanttraffic.se> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.106 required=5 tests=[AWL=0.082, RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/179 X-Sequence-Number: 13926 Are you calculating aggregates, and if so, how are you doing it (I ask the question from experience of a similar application where I found that my aggregating PGPLSQL triggers were bogging the system down, and changed them so scheduled jobs instead). Alex Turner NetEconomist On 8/16/05, Ulrich Wisser wrote: > Hello, >=20 > one of our services is click counting for on line advertising. We do > this by importing Apache log files every five minutes. This results in a > lot of insert and delete statements. At the same time our customers > shall be able to do on line reporting. >=20 > We have a box with > Linux Fedora Core 3, Postgres 7.4.2 > Intel(R) Pentium(R) 4 CPU 2.40GHz > 2 scsi 76GB disks (15.000RPM, 2ms) >=20 > I did put pg_xlog on another file system on other discs. >=20 > Still when several users are on line the reporting gets very slow. > Queries can take more then 2 min. >=20 > I need some ideas how to improve performance in some orders of > magnitude. I already thought of a box with the whole database on a ram > disc. So really any idea is welcome. >=20 > Ulrich >=20 >=20 >=20 > -- > Ulrich Wisser / System Developer >=20 > RELEVANT TRAFFIC SWEDEN AB, Riddarg 17A, SE-114 57 Sthlm, Sweden > Direct (+46)86789755 || Cell (+46)704467893 || Fax (+46)86789769 > ________________________________________________________________ > http://www.relevanttraffic.com >=20 > ---------------------------(end of broadcast)--------------------------- > TIP 1: if posting/reading through Usenet, please send an appropriate > subscribe-nomail command to majordomo@postgresql.org so that your > message can get through to the mailing list cleanly > From pgsql-performance-owner@postgresql.org Tue Aug 16 15:00:17 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id DB270528C6 for ; Tue, 16 Aug 2005 15:00:12 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 58363-05 for ; Tue, 16 Aug 2005 18:00:09 +0000 (GMT) Received: from mail0.rawbw.com (mail0.rawbw.com [198.144.192.41]) by svr1.postgresql.org (Postfix) with ESMTP id 941BA528B7 for ; Tue, 16 Aug 2005 15:00:08 -0300 (ADT) Received: (from www@localhost) by mail0.rawbw.com (8.11.6p2/8.11.6) id j7GI08k76560 for pgsql-performance@postgresql.org; Tue, 16 Aug 2005 11:00:08 -0700 (PDT) Received: from cybs-gw.ic3.com (cybs-gw.ic3.com [66.185.177.10]) by webmail.rawbw.com (IMP) with HTTP for ; Tue, 16 Aug 2005 11:00:08 -0700 Message-ID: <1124215208.430229a8a9cdd@webmail.rawbw.com> Date: Tue, 16 Aug 2005 11:00:08 -0700 From: mudfoot@rawbw.com To: pgsql-performance@postgresql.org Subject: Re: choosing RAID level for xlogs References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BDF@vt-pe2550-001.VANTAGE.vantage.com> In-Reply-To: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BDF@vt-pe2550-001.VANTAGE.vantage.com> MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit User-Agent: Internet Messaging Program (IMP) 3.2.1 X-Originating-IP: 66.185.177.10 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.178 required=5 tests=[NO_REAL_NAME=0.178] X-Spam-Level: X-Archive-Number: 200508/180 X-Sequence-Number: 13927 Quoting Anjan Dave : > Hi, > > > > One simple question. For 125 or more checkpoint segments > (checkpoint_timeout is 600 seconds, shared_buffers are at 21760 or > 170MB) on a very busy database, what is more suitable, a separate 6 disk > RAID5 volume, or a RAID10 volume? Databases will be on separate > spindles. Disks are 36GB 15KRPM, 2Gb Fiber Channel. Performance is > paramount, but I don't want to use RAID0. > RAID10 -- no question. xlog activity is overwhelmingly sequential 8KB writes. In order for RAID5 to perform a write, the host (or controller) needs to perform extra calculations for parity. This turns into latency. RAID10 does not perform those extra calculations. > > > PG7.4.7 on RHAS 4.0 > > > > I can provide more info if needed. > > > > Appreciate some recommendations! > > > > Thanks, > > Anjan > > > > > --- > This email message and any included attachments constitute confidential > and privileged information intended exclusively for the listed > addressee(s). If you are not the intended recipient, please notify > Vantage by immediately telephoning 215-579-8390, extension 1158. In > addition, please reply to this message confirming your receipt of the > same in error. A copy of your email reply can also be sent to > support@vantage.com. Please do not disclose, copy, distribute or take > any action in reliance on the contents of this information. Kindly > destroy all copies of this message and any attachments. Any other use of > this email is prohibited. Thank you for your cooperation. For more > information about Vantage, please visit our website at > http://www.vantage.com . > --- > > > > From pgsql-performance-owner@postgresql.org Tue Aug 16 15:38:07 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3618F52971 for ; Tue, 16 Aug 2005 15:38:06 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 64908-07 for ; Tue, 16 Aug 2005 18:38:03 +0000 (GMT) Received: from vt-pe2550-001.VANTAGE.vantage.com (unknown [64.80.203.244]) by svr1.postgresql.org (Postfix) with ESMTP id 86FBF52927 for ; Tue, 16 Aug 2005 15:37:57 -0300 (ADT) X-MimeOLE: Produced By Microsoft Exchange V6.0.6249.0 content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Subject: Re: choosing RAID level for xlogs Date: Tue, 16 Aug 2005 14:37:38 -0400 Message-ID: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BE0@vt-pe2550-001.VANTAGE.vantage.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] choosing RAID level for xlogs Thread-Index: AcWijK9TdlHgfGpnS/CevndAeVpYfgAAkMAQ From: "Anjan Dave" To: , X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.033 required=5 tests=[AWL=0.033] X-Spam-Level: X-Archive-Number: 200508/181 X-Sequence-Number: 13928 Yes, that's true, though, I am a bit confused because the Clariion array document I am reading talks about how the write cache can eliminate the RAID5 Write Penalty for sequential and large IOs...resulting in better sequential write performance than RAID10. anjan -----Original Message----- From: mudfoot@rawbw.com [mailto:mudfoot@rawbw.com]=20 Sent: Tuesday, August 16, 2005 2:00 PM To: pgsql-performance@postgresql.org Subject: Re: [PERFORM] choosing RAID level for xlogs Quoting Anjan Dave : > Hi, >=20 > =20 >=20 > One simple question. For 125 or more checkpoint segments > (checkpoint_timeout is 600 seconds, shared_buffers are at 21760 or > 170MB) on a very busy database, what is more suitable, a separate 6 disk > RAID5 volume, or a RAID10 volume? Databases will be on separate > spindles. Disks are 36GB 15KRPM, 2Gb Fiber Channel. Performance is > paramount, but I don't want to use RAID0. >=20 RAID10 -- no question. xlog activity is overwhelmingly sequential 8KB writes.=20 In order for RAID5 to perform a write, the host (or controller) needs to perform extra calculations for parity. This turns into latency. RAID10 does not perform those extra calculations. > =20 >=20 > PG7.4.7 on RHAS 4.0 >=20 > =20 >=20 > I can provide more info if needed. >=20 > =20 >=20 > Appreciate some recommendations! >=20 > =20 >=20 > Thanks, >=20 > Anjan >=20 > =20 >=20 > =20 > --- > This email message and any included attachments constitute confidential > and privileged information intended exclusively for the listed > addressee(s). If you are not the intended recipient, please notify > Vantage by immediately telephoning 215-579-8390, extension 1158. In > addition, please reply to this message confirming your receipt of the > same in error. A copy of your email reply can also be sent to > support@vantage.com. Please do not disclose, copy, distribute or take > any action in reliance on the contents of this information. Kindly > destroy all copies of this message and any attachments. Any other use of > this email is prohibited. Thank you for your cooperation. For more > information about Vantage, please visit our website at > http://www.vantage.com . > --- >=20 > =20 >=20 >=20 ---------------------------(end of broadcast)--------------------------- TIP 1: if posting/reading through Usenet, please send an appropriate subscribe-nomail command to majordomo@postgresql.org so that your message can get through to the mailing list cleanly From pgsql-performance-owner@postgresql.org Tue Aug 16 15:48:03 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B3A75529C4 for ; Tue, 16 Aug 2005 15:48:01 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 66712-09 for ; Tue, 16 Aug 2005 18:47:59 +0000 (GMT) Received: from zigo.dhs.org (ua-83-227-204-174.cust.bredbandsbolaget.se [83.227.204.174]) by svr1.postgresql.org (Postfix) with ESMTP id 5EC35529A9 for ; Tue, 16 Aug 2005 15:47:56 -0300 (ADT) Received: from zigo.zigo.dhs.org (zigo.zigo.dhs.org [192.168.0.1]) by zigo.dhs.org (Postfix) with ESMTP id 9C9138467; Tue, 16 Aug 2005 20:47:50 +0200 (CEST) Date: Tue, 16 Aug 2005 20:47:50 +0200 (CEST) From: Dennis Bjorklund To: Ulrich Wisser Cc: pgsql-performance@postgresql.org Subject: Re: Need for speed In-Reply-To: <430208AE.7080100@relevanttraffic.se> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=ISO-8859-1 Content-Transfer-Encoding: 8BIT X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.325 required=5 tests=[AWL=-0.099, DNS_FROM_RFC_ABUSE=0.374, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/182 X-Sequence-Number: 13929 On Tue, 16 Aug 2005, Ulrich Wisser wrote: > Still when several users are on line the reporting gets very slow. > Queries can take more then 2 min. Could you show an exampleof such a query and the output of EXPLAIN ANALYZE on that query (preferably done when the database is slow). It's hard to say what is wrong without more information. -- /Dennis Bj�rklund From pgsql-performance-owner@postgresql.org Tue Aug 16 16:04:28 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 67B6A52A14 for ; Tue, 16 Aug 2005 16:04:26 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 74942-03 for ; Tue, 16 Aug 2005 19:04:19 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id 33D5952A01 for ; Tue, 16 Aug 2005 16:04:18 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050816190418m9100g6rdhe>; Tue, 16 Aug 2005 19:04:18 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id F361B55FBA; Tue, 16 Aug 2005 14:04:17 -0500 (CDT) Received: from [192.168.1.119] (71-32-69-23.cdrr.qwest.net [71.32.69.23]) by juju.arbash-meinel.com (Postfix) with ESMTP id D6AE555FAD; Tue, 16 Aug 2005 14:04:10 -0500 (CDT) Message-ID: <430238A5.5000002@arbash-meinel.com> Date: Tue, 16 Aug 2005 14:04:05 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Anjan Dave Cc: mudfoot@rawbw.com, pgsql-performance@postgresql.org Subject: Re: choosing RAID level for xlogs References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BE0@vt-pe2550-001.VANTAGE.vantage.com> In-Reply-To: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BE0@vt-pe2550-001.VANTAGE.vantage.com> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig8FC2F801ED54EF8FE7F18F89" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.038 required=5 tests=[AWL=-0.013, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/183 X-Sequence-Number: 13930 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig8FC2F801ED54EF8FE7F18F89 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Anjan Dave wrote: > Yes, that's true, though, I am a bit confused because the Clariion array > document I am reading talks about how the write cache can eliminate the > RAID5 Write Penalty for sequential and large IOs...resulting in better > sequential write performance than RAID10. > > anjan > Well, if your stripe size is 128k, and you have N disks in the RAID (N must be even and > 4 for RAID10). With RAID5 you have a stripe across N-1 disks, and 1 parity entry. With RAID10 you have a stripe across N/2 disks, replicated on the second set. So if the average write size is >128k*N/2, then you will generally be using all of the disks during a write, and you can expect a the maximum scale up of about N/2 for RAID10. If your average write size is >128k*(N-1) then you can again write an entire stripe at a time and even the parity since you already know all of the information you don't have to do any reading. So you can get a maximum speed up of N-1. If you are doing infrequent smallish writes, it can be buffered by the write cache, and isn't disk limited at all. And the controller can write it out when it feels like it. So it should be able to do more buffered all-at-once writes. If you are writing a little bit more often (such that the cache fills up), depending on your write pattern, it is possible that all of the stripes are already in the cache, so again there is little penalty for the parity stripe. I suppose the worst case is if you were writing lots of very small chunks, all over the disk in random order. In which case each write encounters a 2x read penalty for a smart controller, or a Nx read penalty if you are going for more safety than speed. (You can read the original value, and the parity, and re-compute the parity with the new value (2x read penalty), but if there is corruption it would not be detected, so you might want to read all of the stripes in the block, and recompute the parity with the new data (Nx read penalty)). I think the issue for Postgres is that it writes 8k pages, which is quite small relative to the stripe size. So you don't tend to build up big buffers to write out the entire stripe at once. So if you aren't filling up your write buffer, RAID5 can do quite well with bulk loads. I also don't know about the penalties for a read followed immediately by a write. Since you will be writing to the same location, you know that you have to wait for the disk to spin back to the same location. At 10k rpm that is a 6ms wait time. For 7200rpm disks, it is 8.3ms. Just to say that there are some specific extra penalties when you are reading the location that you are going to write right away. Now a really smart controller with lots of data to write could read the whole circle on the disk, and then start writing out the entire circle, and not have any spin delay. But you would have to know the size of the circle, and that depends on what block you are on, and the heads arrangement and everything else. Though since hard-drives also have small caches in them, you could hide some of the spin delay, but not a lot, since you have to leave the head there until you are done writing, so while the current command would finish quickly, the next command couldn't start until the first actually finished. Writing large buffers hides all of these seek/spin based latencies, so you can get really good throughput. But a lot of DB action is small buffers randomly distributed, so you really do need low seek time, of which RAID10 is probably better than RAID5. John =:-> --------------enig8FC2F801ED54EF8FE7F18F89 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDAjioJdeBCYSNAAMRAmrxAJ0d/LOOzTPlN8aX0HhbpAD+U9JzBQCcChof FBZU85r9+ojYJKWqBzGLAc0= =d4vK -----END PGP SIGNATURE----- --------------enig8FC2F801ED54EF8FE7F18F89-- From pgsql-performance-owner@postgresql.org Tue Aug 16 16:04:41 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2530D52883 for ; Tue, 16 Aug 2005 16:04:39 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 73272-04 for ; Tue, 16 Aug 2005 19:04:36 +0000 (GMT) Received: from yertle.kcilink.com (yertle.kcilink.com [65.205.34.180]) by svr1.postgresql.org (Postfix) with ESMTP id 1DFBC5286E for ; Tue, 16 Aug 2005 16:04:30 -0300 (ADT) Received: from [192.168.7.103] (host-103.int.kcilink.com [192.168.7.103]) by yertle.kcilink.com (Postfix) with ESMTP id 05EA6B813 for ; Tue, 16 Aug 2005 15:04:24 -0400 (EDT) Mime-Version: 1.0 (Apple Message framework v733) In-Reply-To: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BE0@vt-pe2550-001.VANTAGE.vantage.com> References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BE0@vt-pe2550-001.VANTAGE.vantage.com> Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: <32D5016B-C7E3-4EEC-AA03-A924521F1DE8@khera.org> Content-Transfer-Encoding: 7bit From: Vivek Khera Subject: Re: choosing RAID level for xlogs Date: Tue, 16 Aug 2005 15:04:23 -0400 To: Postgresql Performance X-Mailer: Apple Mail (2.733) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.019 required=5 tests=[AWL=0.019] X-Spam-Level: X-Archive-Number: 200508/184 X-Sequence-Number: 13931 On Aug 16, 2005, at 2:37 PM, Anjan Dave wrote: > Yes, that's true, though, I am a bit confused because the Clariion > array > document I am reading talks about how the write cache can eliminate > the > RAID5 Write Penalty for sequential and large IOs...resulting in better > sequential write performance than RAID10. > well, then run your own tests and find out :-) if I were using LSI MegaRAID controllers, I'd probalby go RAID10, but I don't see why you need 6 disks for this... perhaps just 4 would be enough? Or are your logs really that big? Vivek Khera, Ph.D. +1-301-869-4449 x806 From pgsql-performance-owner@postgresql.org Tue Aug 16 16:16:57 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6B9DD52942 for ; Tue, 16 Aug 2005 16:16:56 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78044-01 for ; Tue, 16 Aug 2005 19:16:54 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id 7CBD752903 for ; Tue, 16 Aug 2005 16:16:53 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050816191654m92009jj78e>; Tue, 16 Aug 2005 19:16:54 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 2947255FBA; Tue, 16 Aug 2005 14:16:54 -0500 (CDT) Received: from [192.168.1.119] (71-32-69-23.cdrr.qwest.net [71.32.69.23]) by juju.arbash-meinel.com (Postfix) with ESMTP id A7CB355FB6; Tue, 16 Aug 2005 14:16:49 -0500 (CDT) Message-ID: <43023BA0.5030608@arbash-meinel.com> Date: Tue, 16 Aug 2005 14:16:48 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Anjan Dave Cc: mudfoot@rawbw.com, pgsql-performance@postgresql.org Subject: Re: choosing RAID level for xlogs References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BE0@vt-pe2550-001.VANTAGE.vantage.com> In-Reply-To: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BE0@vt-pe2550-001.VANTAGE.vantage.com> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig0BC0F093EED0FFC93551C378" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.037 required=5 tests=[AWL=-0.013, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/185 X-Sequence-Number: 13932 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig0BC0F093EED0FFC93551C378 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Anjan Dave wrote: > Yes, that's true, though, I am a bit confused because the Clariion array > document I am reading talks about how the write cache can eliminate the > RAID5 Write Penalty for sequential and large IOs...resulting in better > sequential write performance than RAID10. > > anjan > To give a shorter statement after my long one... If you have enough cache that the controller can write out big chunks to the disk at a time, you can get very good sequential RAID5 performance, because the stripe size is large (so it can do a parallel write to all disks). But for small chunk writes, you suffer the penalty of the read before write, and possible multi-disk read (depends on what is in cache). RAID10 generally handles small writes better, and I would guess that 4disks would perform almost identically to 6disks, since you aren't usually writing enough data to span multiple stripes. If your battery-backed cache is big enough that you don't fill it, they probably perform about the same (superfast) since the cache hides the latency of the disks. If you start filling up your cache, RAID5 probably can do better because of the parallelization. But small writes followed by an fsync do favor RAID10 over RAID5. John =:-> --------------enig0BC0F093EED0FFC93551C378 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDAjugJdeBCYSNAAMRAuRYAKDGHPWXaxLjCySei3LLZ+O/9DuTaQCeIVpM sHxz9K3jdXGZb1lDVGjk7lk= =+TkO -----END PGP SIGNATURE----- --------------enig0BC0F093EED0FFC93551C378-- From pgsql-performance-owner@postgresql.org Tue Aug 16 16:21:14 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1C75952A37 for ; Tue, 16 Aug 2005 16:21:12 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 79279-01 for ; Tue, 16 Aug 2005 19:21:08 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.195]) by svr1.postgresql.org (Postfix) with ESMTP id 9248052A14 for ; Tue, 16 Aug 2005 16:21:07 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id 71so14431wra for ; Tue, 16 Aug 2005 12:21:09 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=fAQ5ZmK0OTPRthgYzwyNmyEnwPEurvQdjERrm2DQ3AVAvwegCDwgnzTeawRuaJ3orYCEy9YOlEb2yCUG5RNACcY0W8BU8Uj8DRTtypCSek/pqG2cEQ/7ywHacxYHz/O+wj0weQAJh3bIik15ZoAe6ZcS0J4PHs/86TB4sUNcxUY= Received: by 10.54.30.54 with SMTP id d54mr4462120wrd; Tue, 16 Aug 2005 12:21:09 -0700 (PDT) Received: by 10.54.86.15 with HTTP; Tue, 16 Aug 2005 12:21:09 -0700 (PDT) Message-ID: <33c6269f0508161221703fc52d@mail.gmail.com> Date: Tue, 16 Aug 2005 15:21:09 -0400 From: Alex Turner To: Anjan Dave Subject: Re: choosing RAID level for xlogs Cc: mudfoot@rawbw.com, pgsql-performance@postgresql.org In-Reply-To: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BE0@vt-pe2550-001.VANTAGE.vantage.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BE0@vt-pe2550-001.VANTAGE.vantage.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.105 required=5 tests=[AWL=0.081, RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/186 X-Sequence-Number: 13933 Theoretically RAID 5 can perform better than RAID 10 over the same number of drives (more members form the stripe in RAID 5 than in RAID 10). All you have to do is calculate parity faster than the drives can write. Doesn't seem like a hard task really, although most RAID controllers seem incapable of doing so, it is possible that Clariion might be able to acheive it. The other factor is that for partial block writes, the array has to first read the original block in order to recalculate the parity, so small random writes are very slow. If you are writing chunks that are larger than your stripe size*(n-1), then in theory the controller doesn't have to re-read a block, and can just overwrite the parity with the new info. Consider just four drives. in RAID 10, it is a stripe of two mirrors, forming two independant units to write to. in RAID 5, it is a 3 drive stripe with parity giving three independant units to write to.=20 Theoretically the RAID 5 should be faster, but I've yet to benchmark a controler where this holds to be true. Of course if you ever do have a drive failure, your array grinds to a halt because rebuilding a raid 5 requires reading (n-1) blocks to rebuild just one block where n is the number of drives in the array, whereas a mirror only required to read from a single spindle of the RAID. I would suggest running some benchmarks at RAID 5 and RAID 10 to see what the _real_ performance actualy is, thats the only way to really tell. Alex Turner NetEconomist On 8/16/05, Anjan Dave wrote: > Yes, that's true, though, I am a bit confused because the Clariion array > document I am reading talks about how the write cache can eliminate the > RAID5 Write Penalty for sequential and large IOs...resulting in better > sequential write performance than RAID10. >=20 > anjan >=20 >=20 > -----Original Message----- > From: mudfoot@rawbw.com [mailto:mudfoot@rawbw.com] > Sent: Tuesday, August 16, 2005 2:00 PM > To: pgsql-performance@postgresql.org > Subject: Re: [PERFORM] choosing RAID level for xlogs >=20 > Quoting Anjan Dave : >=20 > > Hi, > > > > > > > > One simple question. For 125 or more checkpoint segments > > (checkpoint_timeout is 600 seconds, shared_buffers are at 21760 or > > 170MB) on a very busy database, what is more suitable, a separate 6 > disk > > RAID5 volume, or a RAID10 volume? Databases will be on separate > > spindles. Disks are 36GB 15KRPM, 2Gb Fiber Channel. Performance is > > paramount, but I don't want to use RAID0. > > >=20 > RAID10 -- no question. xlog activity is overwhelmingly sequential 8KB > writes. > In order for RAID5 to perform a write, the host (or controller) needs to > perform > extra calculations for parity. This turns into latency. RAID10 does > not > perform those extra calculations. >=20 > > > > > > PG7.4.7 on RHAS 4.0 > > > > > > > > I can provide more info if needed. > > > > > > > > Appreciate some recommendations! > > > > > > > > Thanks, > > > > Anjan > > > > > > > > > > --- > > This email message and any included attachments constitute > confidential > > and privileged information intended exclusively for the listed > > addressee(s). If you are not the intended recipient, please notify > > Vantage by immediately telephoning 215-579-8390, extension 1158. In > > addition, please reply to this message confirming your receipt of the > > same in error. A copy of your email reply can also be sent to > > support@vantage.com. Please do not disclose, copy, distribute or take > > any action in reliance on the contents of this information. Kindly > > destroy all copies of this message and any attachments. Any other use > of > > this email is prohibited. Thank you for your cooperation. For more > > information about Vantage, please visit our website at > > http://www.vantage.com . > > --- > > > > > > > > >=20 >=20 >=20 > ---------------------------(end of broadcast)--------------------------- > TIP 1: if posting/reading through Usenet, please send an appropriate > subscribe-nomail command to majordomo@postgresql.org so that your > message can get through to the mailing list cleanly >=20 >=20 > ---------------------------(end of broadcast)--------------------------- > TIP 4: Have you searched our list archives? >=20 > http://archives.postgresql.org > From pgsql-performance-owner@postgresql.org Tue Aug 16 16:52:51 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8D67E529C4 for ; Tue, 16 Aug 2005 16:52:48 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 79971-10 for ; Tue, 16 Aug 2005 19:52:45 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.203]) by svr1.postgresql.org (Postfix) with ESMTP id F03E3529E5 for ; Tue, 16 Aug 2005 16:52:44 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id 70so14956wra for ; Tue, 16 Aug 2005 12:52:45 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=PoUtsl1+zUyT13R5ETmzREg2wGQN2XBh27QvU8i+0BdquaEgwtsU69bMFYU5hW7Czf76uzXaZxAWOwlfIM3XPJmkBvbbOZ3JmLHnyayih5TKuphk7Xnl3/X1o21lL6y8Lf3lqGvEmnDQNrWC4ZU7dcnEWxBzD6bywsjB4tWZhNw= Received: by 10.54.22.71 with SMTP id 71mr1092582wrv; Tue, 16 Aug 2005 12:52:45 -0700 (PDT) Received: by 10.54.86.15 with HTTP; Tue, 16 Aug 2005 12:52:45 -0700 (PDT) Message-ID: <33c6269f050816125266a62d5d@mail.gmail.com> Date: Tue, 16 Aug 2005 15:52:45 -0400 From: Alex Turner To: John A Meinel Subject: Re: choosing RAID level for xlogs Cc: Anjan Dave , mudfoot@rawbw.com, pgsql-performance@postgresql.org In-Reply-To: <43023BA0.5030608@arbash-meinel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BE0@vt-pe2550-001.VANTAGE.vantage.com> <43023BA0.5030608@arbash-meinel.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.105 required=5 tests=[AWL=0.081, RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/187 X-Sequence-Number: 13934 Don't forget that often controlers don't obey fsyncs like a plain drive does. thats the point of having a BBU ;) Alex Turner NetEconomist On 8/16/05, John A Meinel wrote: > Anjan Dave wrote: > > Yes, that's true, though, I am a bit confused because the Clariion arra= y > > document I am reading talks about how the write cache can eliminate the > > RAID5 Write Penalty for sequential and large IOs...resulting in better > > sequential write performance than RAID10. > > > > anjan > > >=20 > To give a shorter statement after my long one... > If you have enough cache that the controller can write out big chunks to > the disk at a time, you can get very good sequential RAID5 performance, > because the stripe size is large (so it can do a parallel write to all > disks). >=20 > But for small chunk writes, you suffer the penalty of the read before > write, and possible multi-disk read (depends on what is in cache). >=20 > RAID10 generally handles small writes better, and I would guess that > 4disks would perform almost identically to 6disks, since you aren't > usually writing enough data to span multiple stripes. >=20 > If your battery-backed cache is big enough that you don't fill it, they > probably perform about the same (superfast) since the cache hides the > latency of the disks. >=20 > If you start filling up your cache, RAID5 probably can do better because > of the parallelization. >=20 > But small writes followed by an fsync do favor RAID10 over RAID5. >=20 > John > =3D:-> >=20 >=20 > From pgsql-performance-owner@postgresql.org Tue Aug 16 19:04:11 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D9D60528C6 for ; Tue, 16 Aug 2005 18:59:02 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 14829-08 for ; Tue, 16 Aug 2005 21:58:57 +0000 (GMT) Received: from dd01.profihoster.net (dd01.profihoster.net [84.233.130.40]) by svr1.postgresql.org (Postfix) with ESMTP id A445552896 for ; Tue, 16 Aug 2005 18:58:55 -0300 (ADT) Received: from [192.168.54.114] (Iba10.i.pppool.de [85.73.186.16]) by dd01.profihoster.net (Postfix) with ESMTP id 376002380EF for ; Tue, 16 Aug 2005 23:58:50 +0200 (CEST) Message-ID: <430261A0.3040708@laliluna.de> Date: Tue, 16 Aug 2005 23:58:56 +0200 From: Sebastian Hennebrueder User-Agent: Mozilla Thunderbird 1.0.6 (Windows/20050716) X-Accept-Language: de-DE, de, en-us, en MIME-Version: 1.0 To: postgres performance Subject: Re: Looking for a large database for testing References: <430195DC.70104@laliluna.de> <43019ED4.7070509@wildenhain.de> <4301A29C.9010003@laliluna.de> In-Reply-To: <4301A29C.9010003@laliluna.de> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/188 X-Sequence-Number: 13935 Sebastian Hennebrueder schrieb: >Tino Wildenhain schrieb: > > > > >>You can download the wikipedia content. Just browse the wikimedia site. >>Its some work to change the data to be able to import into postgres, >>but at least you have a lot real world data - in many languages. >> >> > >I have just found it. Here there is a link >http://download.wikimedia.org/ >They have content in multiple languages and dumps up to 20 GB. > > > Just if anybody wants to import the wikipedia data. I had considerable problems to get the proper encoding working. I downloaded the german content from wikipedia, which is a dump of a unicode encoded database of mysql (utf8) I used MySql 4.1 on Windows 2000 to read the dump and then copied the data with a small application to postgreSQL In mysql.ini you should configure the setting max_allowed_packet = 10M I set it to 10, wich worked out. Else you can not import the dump into mysql. The error message was something like lost connection .... The default encoding of mysql was latin1 which worked. Then I imported the dump mysql -uYourUserName -pPassword --default-character-set=utf8 database < downloadedAndUnzippedFile The default-character-set is very important Create table in postgres (not with all the columns) CREATE TABLE content ( cur_id int4 NOT NULL DEFAULT nextval('public.cur_cur_id_seq'::text), cur_namespace int2 NOT NULL DEFAULT (0)::smallint, cur_title varchar(255) NOT NULL DEFAULT ''::character varying, cur_text text NOT NULL, cur_comment text, cur_user int4 NOT NULL DEFAULT 0, cur_user_text varchar(255) NOT NULL DEFAULT ''::character varying, cur_timestamp varchar(14) NOT NULL DEFAULT ''::character varying ) ; After this I copied the data from mySql to postgres with a small Java application. The code is not beautiful. private void copyEntries() throws Exception { Class.forName("org.postgresql.Driver"); Class.forName("com.mysql.jdbc.Driver"); Connection conMySQL = DriverManager.getConnection( "jdbc:mysql://localhost/wikidb", "root", "mysql"); Connection conPostgreSQL = DriverManager.getConnection( "jdbc:postgresql://localhost/wiki", "postgres", "p"); Statement selectStatement = conMySQL.createStatement(); StringBuffer sqlQuery = new StringBuffer(); sqlQuery.append("insert into content ("); sqlQuery .append("cur_id, cur_namespace, cur_title, cur_text, cur_comment, cur_user, "); sqlQuery.append("cur_user_text , cur_timestamp) "); sqlQuery.append("values (?,?,?,?,?,?,?,?)"); PreparedStatement insertStatement = conPostgreSQL .prepareStatement(sqlQuery.toString()); // get total rows java.sql.ResultSet resultSet = selectStatement .executeQuery("select count(*) from cur"); resultSet.next(); int iMax = resultSet.getInt(1); int i = 0; while (i < iMax) { resultSet = selectStatement .executeQuery("select * from cur limit "+i +", 2000"); while (resultSet.next()) { i++; if (i % 100 == 0) System.out.println("" + i + " von " + iMax); insertStatement.setInt(1, resultSet.getInt(1)); insertStatement.setInt(2, resultSet.getInt(2)); insertStatement.setString(3, resultSet.getString(3)); insertStatement.setString(4, resultSet.getString(4)); // this blob field is utf-8 encoded byte comment[] = resultSet.getBytes(5); insertStatement.setString(5, new String(comment, "UTF-8")); insertStatement.setInt(6, resultSet.getInt(6)); insertStatement.setString(7, resultSet.getString(7)); insertStatement.setString(8, resultSet.getString(8)); insertStatement.execute(); } } } -- Best Regards / Viele Gr��e Sebastian Hennebrueder ---- http://www.laliluna.de Tutorials for JSP, JavaServer Faces, Struts, Hibernate and EJB Get support, education and consulting for these technologies. From pgsql-performance-owner@postgresql.org Tue Aug 16 19:42:08 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 462F25283C for ; Tue, 16 Aug 2005 19:22:59 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 23684-03 for ; Tue, 16 Aug 2005 22:22:56 +0000 (GMT) Received: from loki.globexplorer.com (loki.globexplorer.com [208.35.14.101]) by svr1.postgresql.org (Postfix) with ESMTP id 380655281E for ; Tue, 16 Aug 2005 19:22:55 -0300 (ADT) X-MIMEOLE: Produced By Microsoft Exchange V6.0.6603.0 Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="Windows-1252" Content-Transfer-Encoding: quoted-printable Subject: Re: choosing RAID level for xlogs Date: Tue, 16 Aug 2005 15:22:55 -0700 Message-ID: <71E37EF6B7DCC1499CEA0316A2568328024BB79D@loki.wc.globexplorer.net> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: choosing RAID level for xlogs Thread-Index: AcWh2NDgCbXrlD2dSvuZJkSyYWP7KAA2AksO From: "Gregory S. Williamson" To: "Anjan Dave" , X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/189 X-Sequence-Number: 13936 I would be very cautious about ever using RAID5, despite manufacturers' = claims to the contrary. The link below is authored by a very = knowledgable fellow whose posts I know (and trust) from Informix land. Greg Williamson DBA GlobeXplorer LLC -----Original Message----- From: pgsql-performance-owner@postgresql.org on behalf of Anjan Dave Sent: Mon 8/15/2005 1:35 PM To: pgsql-performance@postgresql.org Cc:=09 Subject: [PERFORM] choosing RAID level for xlogs Hi, =20 One simple question. For 125 or more checkpoint segments (checkpoint_timeout is 600 seconds, shared_buffers are at 21760 or 170MB) on a very busy database, what is more suitable, a separate 6 disk RAID5 volume, or a RAID10 volume? Databases will be on separate spindles. Disks are 36GB 15KRPM, 2Gb Fiber Channel. Performance is paramount, but I don't want to use RAID0. =20 PG7.4.7 on RHAS 4.0 =20 I can provide more info if needed. =20 Appreciate some recommendations! =20 Thanks, Anjan =20 =20 --- This email message and any included attachments constitute confidential and privileged information intended exclusively for the listed addressee(s). If you are not the intended recipient, please notify Vantage by immediately telephoning 215-579-8390, extension 1158. In addition, please reply to this message confirming your receipt of the same in error. A copy of your email reply can also be sent to support@vantage.com. Please do not disclose, copy, distribute or take any action in reliance on the contents of this information. Kindly destroy all copies of this message and any attachments. Any other use of this email is prohibited. Thank you for your cooperation. For more information about Vantage, please visit our website at http://www.vantage.com . --- =20 !DSPAM:4300fd35105094125621296! From pgsql-performance-owner@postgresql.org Tue Aug 16 19:52:55 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id F18235283C for ; Tue, 16 Aug 2005 19:50:24 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 32212-01 for ; Tue, 16 Aug 2005 22:50:21 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 5716C529BC for ; Tue, 16 Aug 2005 19:50:19 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7GMnxXp007517; Tue, 16 Aug 2005 18:49:59 -0400 (EDT) To: John A Meinel Cc: Alvaro Herrera , Josh Berkus , pgsql-performance@postgresql.org, "Jeffrey W. Baker" , Steve Poe , paul@oxton.com Subject: Re: PG8 Tuning In-reply-to: <43021567.1000101@arbash-meinel.com> References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> <1123836428.19976.26.camel@amd64-laptop-spoe> <1123779524.6664.1.camel@toonses.gghcwest.com> <200508160912.32263.josh@agliodbs.com> <20050816162531.GB32695@alvh.no-ip.org> <43021567.1000101@arbash-meinel.com> Comments: In-reply-to John A Meinel message dated "Tue, 16 Aug 2005 11:33:43 -0500" Date: Tue, 16 Aug 2005 18:49:59 -0400 Message-ID: <7516.1124232599@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/190 X-Sequence-Number: 13937 John A Meinel writes: > Alvaro Herrera wrote: >> I've been asked this a couple of times and I don't know the answer: what >> happens if you give XLog a single drive (unmirrored single spindle), and >> that drive dies? So the question really is, should you be giving two >> disks to XLog? > I can propose a simple test. Create a test database. Run postgres, > insert a bunch of stuff. Stop postgres. Delete everything in the pg_xlog > directory. Start postgres again, what does it do? That test would really be completely unrelated to the problem. If you are able to shut down the database cleanly, then you do not need pg_xlog anymore --- everything is on disk in the data area. You might have to use pg_resetxlog to get going again, but you won't lose anything by doing so. The question of importance is: if the xlog drive dies while the database is running, are you going to be able to get the postmaster to shut down cleanly? My suspicion is "no" --- if the kernel is reporting write failures on WAL, that's going to prevent writes to the data drives (good ol' WAL-before-data rule). You could imagine failure modes where the drive is toast but isn't actually reporting any errors ... but one hopes that's not a common scenario. In a scenario like this, it might be interesting to have a shutdown mode that deliberately ignores writing to WAL and just does its best to get all the dirty pages down onto the data drives. In the meantime ... use a mirrored drive for WAL. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 16 20:22:04 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7292D52870 for ; Tue, 16 Aug 2005 20:15:06 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 36806-02 for ; Tue, 16 Aug 2005 23:15:00 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id 2E235529E6 for ; Tue, 16 Aug 2005 20:14:58 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050816231500m9100g6eeae>; Tue, 16 Aug 2005 23:15:00 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id C048155FBA; Tue, 16 Aug 2005 18:14:59 -0500 (CDT) Received: from [192.168.1.119] (71-32-69-23.cdrr.qwest.net [71.32.69.23]) by juju.arbash-meinel.com (Postfix) with ESMTP id 46D3755FAD; Tue, 16 Aug 2005 18:14:45 -0500 (CDT) Message-ID: <43027361.2070303@arbash-meinel.com> Date: Tue, 16 Aug 2005 18:14:41 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane , Postgresql Performance Subject: Re: PG8 Tuning References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> <1123836428.19976.26.camel@amd64-laptop-spoe> <1123779524.6664.1.camel@toonses.gghcwest.com> <200508160912.32263.josh@agliodbs.com> <20050816162531.GB32695@alvh.no-ip.org> <43021567.1000101@arbash-meinel.com> <7516.1124232599@sss.pgh.pa.us> In-Reply-To: <7516.1124232599@sss.pgh.pa.us> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig5E2D1FE484D3F3971760F402" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.038 required=5 tests=[AWL=-0.013, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/191 X-Sequence-Number: 13938 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig5E2D1FE484D3F3971760F402 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Tom Lane wrote: > John A Meinel writes: > >>Alvaro Herrera wrote: >> >>>I've been asked this a couple of times and I don't know the answer: what >>>happens if you give XLog a single drive (unmirrored single spindle), and >>>that drive dies? So the question really is, should you be giving two >>>disks to XLog? > > >>I can propose a simple test. Create a test database. Run postgres, >>insert a bunch of stuff. Stop postgres. Delete everything in the pg_xlog >>directory. Start postgres again, what does it do? > > > That test would really be completely unrelated to the problem. > > If you are able to shut down the database cleanly, then you do not need > pg_xlog anymore --- everything is on disk in the data area. You might > have to use pg_resetxlog to get going again, but you won't lose anything > by doing so. So pg_xlog is really only needed for a dirty shutdown. So what about the idea of having pg_xlog on a ramdisk that is syncronized periodically to a real disk. I'm guessing you would get corruption of the database, or at least you don't know what is clean and what is dirty, since there would be no WAL entry for some of the things that completed, but also no WAL entry for things that were not completed. So what is actually written to the WAL? Is it something like: "I am writing these pages, and when page X has a certain value, I am finished" I'm just curious, because I don't believe you write to the WAL when you complete the writing the data, you only make a note about what you are going to do before you do it. So there needs to be a way to detect if you actually finished (which would be in the actual data). John =:-> > > The question of importance is: if the xlog drive dies while the database > is running, are you going to be able to get the postmaster to shut down > cleanly? My suspicion is "no" --- if the kernel is reporting write > failures on WAL, that's going to prevent writes to the data drives (good > ol' WAL-before-data rule). You could imagine failure modes where the > drive is toast but isn't actually reporting any errors ... but one hopes > that's not a common scenario. > > In a scenario like this, it might be interesting to have a shutdown mode > that deliberately ignores writing to WAL and just does its best to get > all the dirty pages down onto the data drives. > > In the meantime ... use a mirrored drive for WAL. > > regards, tom lane > --------------enig5E2D1FE484D3F3971760F402 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDAnNkJdeBCYSNAAMRAvbYAJ4oWsYrHLDMpZwN54Hj0iar6SSsGACgx59f 9DeX5ViF+I9WodBxGD3/4lw= =MeG4 -----END PGP SIGNATURE----- --------------enig5E2D1FE484D3F3971760F402-- From pgsql-performance-owner@postgresql.org Tue Aug 16 22:12:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7C99E5282E for ; Tue, 16 Aug 2005 22:12:24 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 68348-01 for ; Wed, 17 Aug 2005 01:12:20 +0000 (GMT) Received: from vt-pe2550-001.VANTAGE.vantage.com (unknown [64.80.203.244]) by svr1.postgresql.org (Postfix) with ESMTP id 809E45287D for ; Tue, 16 Aug 2005 22:12:15 -0300 (ADT) X-MimeOLE: Produced By Microsoft Exchange V6.0.6249.0 content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: base64 Subject: Re: choosing RAID level for xlogs Date: Tue, 16 Aug 2005 21:12:14 -0400 Message-ID: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BE1@vt-pe2550-001.VANTAGE.vantage.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: choosing RAID level for xlogs Thread-Index: AcWh2NDgCbXrlD2dSvuZJkSyYWP7KAA2AksOAASmv34= From: "Anjan Dave" To: "Gregory S. Williamson" , X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.032 required=5 tests=[AWL=0.032] X-Spam-Level: X-Archive-Number: 200508/192 X-Sequence-Number: 13939 VGhhbmtzLCBldmVyeW9uZS4gSSBnb3Qgc29tZSBleGNlbGxlbnQgcmVwbGllcywgaW5jbHVkaW5n IHNvbWUgbG9uZyBleHBsYW5hdGlvbnMuIEFwcHJlY2lhdGUgdGhlIHRpbWUgeW91IGd1eXMgdG9v ayBvdXQgZm9yIHRoZSByZXNwb25zZXMuDQogDQpUaGUgZ2lzdCBvZiBpdCBpIHRha2UsIGlzIHRv IHVzZSBSQUlEMTAuIEkgaGF2ZSA0MDBNQisgb2Ygd3JpdGUgY2FjaGUgb24gdGhlIGNvbnRyb2xs ZXIocyksIHRoYXQgdGhlIFJBSUQ1IExVTihzKSBjb3VsZCBiZW5lZml0IGZyb20gYnkgZmlsbGlu ZyBpdCB1cCBhbmQgd3JpdGluZyBvdXQgdGhlIGNvbXBsZXRlIHN0cmlwZSwgYnV0IGNvbWUgdG8g dGhpbmsgb2YgaXQsIGl0J3Mgc2hhcmVkIGFtb25nIHRoZSB0d28gU3RvcmFnZSBQcm9jZXNzb3Jz LCBhbGwgdGhlIExVTnMsIG5vdCBqdXN0IHRoZSBvbmVzIGhvbGRpbmcgdGhlIHBnX3hsb2cgZGly ZWN0b3J5LiBUaGUgb3RoZXIgdGhpbmcgKHdpdGggQ2xhcmlpb24pIGlzIHRoZSB3cml0ZSBjYWNo ZSBtaXJyb3JpbmcuIFdyaXRlIGlzbid0IHNpZ25hbGxlZCBjb21wbGV0ZSB0byB0aGUgaG9zdCB1 bnRpbCB0aGUgY2FjaGUgY29udGVudCBpcyBtaXJyb3JlZCBhY3Jvc3MgdGhlIG90aGVyIFNQIChh bmQgdmljZS12ZXJzYSksIHdoaWNoIGlzIGEgZ29vZCB0aGluZywgYnV0IHRoaXMgb3BlcmF0aW9u IGNvdWxkIHBvdGVudGlhbGx5IGJlY29tZSBhIGJvdHRsZW5lY2sgd2l0aCB2ZXJ5IGhpZ2ggbG9h ZCBvbiB0aGUgU1BzLg0KIA0KQWxzbywgb25lIHdvdWxkIGhhdmUgdG8gZnVsbHkgdHJ1c3QgdGhl IGNvbnRyb2xsZXIvbWFudWZhY3R1cmVyJ3MgY2xhaW0gb24gc2lnbmFsbGluZyB0aGUgd3JpdGUg Y29tcGxldGlvbi4gQW5kLCBwZXJmb3JtYW5jZSBpcyBhIHByaW9yaXR5IG92ZXIgdGhlIGRyaXZl IHNwYWNlIGxvc3QgaW4gUkFJRDEwIGZvciBtZS4NCiANCkkgY2FuIHVzZSA0IGRyaXZlcyBpbnN0 ZWFkIG9mIDYuDQogDQpUaGFua3MsDQpBbmphbiAgDQoNCgl0LS0tLS1PcmlnaW5hbCBNZXNzYWdl LS0tLS0gDQoJRnJvbTogR3JlZ29yeSBTLiBXaWxsaWFtc29uIFttYWlsdG86Z3N3QGdsb2JleHBs b3Jlci5jb21dIA0KCVNlbnQ6IFR1ZSA4LzE2LzIwMDUgNjoyMiBQTSANCglUbzogQW5qYW4gRGF2 ZTsgcGdzcWwtcGVyZm9ybWFuY2VAcG9zdGdyZXNxbC5vcmcgDQoJQ2M6IA0KCVN1YmplY3Q6IFJF OiBbUEVSRk9STV0gY2hvb3NpbmcgUkFJRCBsZXZlbCBmb3IgeGxvZ3MNCgkNCgkNCg0KCUkgd291 bGQgYmUgdmVyeSBjYXV0aW91cyBhYm91dCBldmVyIHVzaW5nIFJBSUQ1LCBkZXNwaXRlIG1hbnVm YWN0dXJlcnMnIGNsYWltcyB0byB0aGUgY29udHJhcnkuIFRoZSBsaW5rIGJlbG93IGlzIGF1dGhv cmVkIGJ5IGEgdmVyeSBrbm93bGVkZ2FibGUgZmVsbG93IHdob3NlIHBvc3RzIEkga25vdyAoYW5k IHRydXN0KSBmcm9tIEluZm9ybWl4IGxhbmQuDQoNCgk8aHR0cDovL3d3dy5taXJhY2xlYXMuY29t L0JBQVJGL1JBSUQ1X3ZlcnN1c19SQUlEMTAudHh0PiANCg0KCUdyZWcgV2lsbGlhbXNvbiANCglE QkEgDQoJR2xvYmVYcGxvcmVyIExMQyANCg0KDQoJLS0tLS1PcmlnaW5hbCBNZXNzYWdlLS0tLS0g DQoJRnJvbTogICBwZ3NxbC1wZXJmb3JtYW5jZS1vd25lckBwb3N0Z3Jlc3FsLm9yZyBvbiBiZWhh bGYgb2YgQW5qYW4gRGF2ZSANCglTZW50OiAgIE1vbiA4LzE1LzIwMDUgMTozNSBQTSANCglUbzog ICAgIHBnc3FsLXBlcmZvcm1hbmNlQHBvc3RncmVzcWwub3JnIA0KCUNjOiAgICAgDQoJU3ViamVj dDogICAgICAgIFtQRVJGT1JNXSBjaG9vc2luZyBSQUlEIGxldmVsIGZvciB4bG9ncyANCglIaSwg DQoNCgkNCg0KCU9uZSBzaW1wbGUgcXVlc3Rpb24uIEZvciAxMjUgb3IgbW9yZSBjaGVja3BvaW50 IHNlZ21lbnRzIA0KCShjaGVja3BvaW50X3RpbWVvdXQgaXMgNjAwIHNlY29uZHMsIHNoYXJlZF9i dWZmZXJzIGFyZSBhdCAyMTc2MCBvciANCgkxNzBNQikgb24gYSB2ZXJ5IGJ1c3kgZGF0YWJhc2Us IHdoYXQgaXMgbW9yZSBzdWl0YWJsZSwgYSBzZXBhcmF0ZSA2IGRpc2sgDQoJUkFJRDUgdm9sdW1l LCBvciBhIFJBSUQxMCB2b2x1bWU/IERhdGFiYXNlcyB3aWxsIGJlIG9uIHNlcGFyYXRlIA0KCXNw aW5kbGVzLiBEaXNrcyBhcmUgMzZHQiAxNUtSUE0sIDJHYiBGaWJlciBDaGFubmVsLiBQZXJmb3Jt YW5jZSBpcyANCglwYXJhbW91bnQsIGJ1dCBJIGRvbid0IHdhbnQgdG8gdXNlIFJBSUQwLiANCg0K CQ0KDQoJUEc3LjQuNyBvbiBSSEFTIDQuMCANCg0KCQ0KDQoJSSBjYW4gcHJvdmlkZSBtb3JlIGlu Zm8gaWYgbmVlZGVkLiANCg0KCQ0KDQoJQXBwcmVjaWF0ZSBzb21lIHJlY29tbWVuZGF0aW9ucyEg DQoNCgkNCg0KCVRoYW5rcywgDQoNCglBbmphbiANCg0KCQ0KDQoJDQoJLS0tIA0KCVRoaXMgZW1h aWwgbWVzc2FnZSBhbmQgYW55IGluY2x1ZGVkIGF0dGFjaG1lbnRzIGNvbnN0aXR1dGUgY29uZmlk ZW50aWFsIA0KCWFuZCBwcml2aWxlZ2VkIGluZm9ybWF0aW9uIGludGVuZGVkIGV4Y2x1c2l2ZWx5 IGZvciB0aGUgbGlzdGVkIA0KCWFkZHJlc3NlZShzKS4gSWYgeW91IGFyZSBub3QgdGhlIGludGVu ZGVkIHJlY2lwaWVudCwgcGxlYXNlIG5vdGlmeSANCglWYW50YWdlIGJ5IGltbWVkaWF0ZWx5IHRl bGVwaG9uaW5nIDIxNS01NzktODM5MCwgZXh0ZW5zaW9uIDExNTguIEluIA0KCWFkZGl0aW9uLCBw bGVhc2UgcmVwbHkgdG8gdGhpcyBtZXNzYWdlIGNvbmZpcm1pbmcgeW91ciByZWNlaXB0IG9mIHRo ZSANCglzYW1lIGluIGVycm9yLiBBIGNvcHkgb2YgeW91ciBlbWFpbCByZXBseSBjYW4gYWxzbyBi ZSBzZW50IHRvIA0KCXN1cHBvcnRAdmFudGFnZS5jb20uIFBsZWFzZSBkbyBub3QgZGlzY2xvc2Us IGNvcHksIGRpc3RyaWJ1dGUgb3IgdGFrZSANCglhbnkgYWN0aW9uIGluIHJlbGlhbmNlIG9uIHRo ZSBjb250ZW50cyBvZiB0aGlzIGluZm9ybWF0aW9uLiBLaW5kbHkgDQoJZGVzdHJveSBhbGwgY29w aWVzIG9mIHRoaXMgbWVzc2FnZSBhbmQgYW55IGF0dGFjaG1lbnRzLiBBbnkgb3RoZXIgdXNlIG9m IA0KCXRoaXMgZW1haWwgaXMgcHJvaGliaXRlZC4gVGhhbmsgeW91IGZvciB5b3VyIGNvb3BlcmF0 aW9uLiBGb3IgbW9yZSANCglpbmZvcm1hdGlvbiBhYm91dCBWYW50YWdlLCBwbGVhc2UgdmlzaXQg b3VyIHdlYnNpdGUgYXQgDQoJaHR0cDovL3d3dy52YW50YWdlLmNvbSA8aHR0cDovL3d3dy52YW50 YWdlLmNvbS8+IC4gDQoJLS0tIA0KDQoJDQoNCg0KDQoJIURTUEFNOjQzMDBmZDM1MTA1MDk0MTI1 NjIxMjk2ISANCg0KDQoNCg== From pgsql-performance-owner@postgresql.org Tue Aug 16 23:28:23 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BC73B52A80 for ; Tue, 16 Aug 2005 23:27:19 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 82224-09 for ; Wed, 17 Aug 2005 02:27:18 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 0337552AB9 for ; Tue, 16 Aug 2005 23:27:16 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7H2RKtN010093; Tue, 16 Aug 2005 22:27:20 -0400 (EDT) To: John A Meinel Cc: Postgresql Performance Subject: Re: PG8 Tuning In-reply-to: <43027361.2070303@arbash-meinel.com> References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> <1123836428.19976.26.camel@amd64-laptop-spoe> <1123779524.6664.1.camel@toonses.gghcwest.com> <200508160912.32263.josh@agliodbs.com> <20050816162531.GB32695@alvh.no-ip.org> <43021567.1000101@arbash-meinel.com> <7516.1124232599@sss.pgh.pa.us> <43027361.2070303@arbash-meinel.com> Comments: In-reply-to John A Meinel message dated "Tue, 16 Aug 2005 18:14:41 -0500" Date: Tue, 16 Aug 2005 22:27:20 -0400 Message-ID: <10092.1124245640@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/193 X-Sequence-Number: 13940 John A Meinel writes: > So pg_xlog is really only needed for a dirty shutdown. So what about the > idea of having pg_xlog on a ramdisk that is syncronized periodically to > a real disk. Well, if "periodically" means "at every transaction commit", that's pretty much what we do now. regards, tom lane From pgsql-performance-owner@postgresql.org Wed Aug 17 00:15:27 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 54C4E52AD2 for ; Wed, 17 Aug 2005 00:10:53 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 04050-02 for ; Wed, 17 Aug 2005 03:10:46 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.203]) by svr1.postgresql.org (Postfix) with ESMTP id 7EBB4528D4 for ; Wed, 17 Aug 2005 00:10:39 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i22so77519wra for ; Tue, 16 Aug 2005 20:10:39 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=Iy+Iwb6xoYEhyU8axm2jXwctLiAGr2jpzRmhUmeu/cCFSAUGYalFIH9NhVWcUB5LqEL0EVqdgSVIrM/rHE0WKxXbuS8/KBcAHZ91fbAY7pjqOzE4yKFX0czNiAoBEeoCKNTE8oq8VtTonQZwLPZROO30UbbgIvYw/Onv9lKVD2k= Received: by 10.54.26.63 with SMTP id 63mr149251wrz; Tue, 16 Aug 2005 20:10:38 -0700 (PDT) Received: by 10.54.86.15 with HTTP; Tue, 16 Aug 2005 20:10:38 -0700 (PDT) Message-ID: <33c6269f0508162010519a7f12@mail.gmail.com> Date: Tue, 16 Aug 2005 23:10:38 -0400 From: Alex Turner To: Anjan Dave Subject: Re: choosing RAID level for xlogs Cc: "Gregory S. Williamson" , pgsql-performance@postgresql.org In-Reply-To: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BE1@vt-pe2550-001.VANTAGE.vantage.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098BE1@vt-pe2550-001.VANTAGE.vantage.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.104 required=5 tests=[AWL=0.080, RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/194 X-Sequence-Number: 13941 The other point that is well made is that with enough drives you will max out the PCI bus before you max out the drives. 64-bit 66Mhz can do about 400MB/sec, which can be acheived by two 3 drive stripes (6 drive in RAID 10). A true PCI-X card can do better, but can your controller? Remember, U320 is only 320MB/channel... Alex Turner NetEconomist On 8/16/05, Anjan Dave wrote: > Thanks, everyone. I got some excellent replies, including some long expla= nations. Appreciate the time you guys took out for the responses. >=20 > The gist of it i take, is to use RAID10. I have 400MB+ of write cache on = the controller(s), that the RAID5 LUN(s) could benefit from by filling it u= p and writing out the complete stripe, but come to think of it, it's shared= among the two Storage Processors, all the LUNs, not just the ones holding = the pg_xlog directory. The other thing (with Clariion) is the write cache m= irroring. Write isn't signalled complete to the host until the cache conten= t is mirrored across the other SP (and vice-versa), which is a good thing, = but this operation could potentially become a bottleneck with very high loa= d on the SPs. >=20 > Also, one would have to fully trust the controller/manufacturer's claim o= n signalling the write completion. And, performance is a priority over the = drive space lost in RAID10 for me. >=20 > I can use 4 drives instead of 6. >=20 > Thanks, > Anjan >=20 > t-----Original Message----- > From: Gregory S. Williamson [mailto:gsw@globexplorer.com] > Sent: Tue 8/16/2005 6:22 PM > To: Anjan Dave; pgsql-performance@postgresql.org > Cc: > Subject: RE: [PERFORM] choosing RAID level for xlogs >=20 >=20 >=20 > I would be very cautious about ever using RAID5, despite manufact= urers' claims to the contrary. The link below is authored by a very knowled= gable fellow whose posts I know (and trust) from Informix land. >=20 > >=20 > Greg Williamson > DBA > GlobeXplorer LLC >=20 >=20 > -----Original Message----- > From: pgsql-performance-owner@postgresql.org on behalf of Anjan= Dave > Sent: Mon 8/15/2005 1:35 PM > To: pgsql-performance@postgresql.org > Cc: > Subject: [PERFORM] choosing RAID level for xlogs > Hi, >=20 >=20 >=20 > One simple question. For 125 or more checkpoint segments > (checkpoint_timeout is 600 seconds, shared_buffers are at 21760 o= r > 170MB) on a very busy database, what is more suitable, a separate= 6 disk > RAID5 volume, or a RAID10 volume? Databases will be on separate > spindles. Disks are 36GB 15KRPM, 2Gb Fiber Channel. Performance i= s > paramount, but I don't want to use RAID0. >=20 >=20 >=20 > PG7.4.7 on RHAS 4.0 >=20 >=20 >=20 > I can provide more info if needed. >=20 >=20 >=20 > Appreciate some recommendations! >=20 >=20 >=20 > Thanks, >=20 > Anjan >=20 >=20 >=20 >=20 > --- > This email message and any included attachments constitute confid= ential > and privileged information intended exclusively for the listed > addressee(s). If you are not the intended recipient, please notif= y > Vantage by immediately telephoning 215-579-8390, extension 1158. = In > addition, please reply to this message confirming your receipt of= the > same in error. A copy of your email reply can also be sent to > support@vantage.com. Please do not disclose, copy, distribute or = take > any action in reliance on the contents of this information. Kindl= y > destroy all copies of this message and any attachments. Any other= use of > this email is prohibited. Thank you for your cooperation. For mor= e > information about Vantage, please visit our website at > http://www.vantage.com . > --- >=20 >=20 >=20 >=20 >=20 > !DSPAM:4300fd35105094125621296! >=20 >=20 >=20 >=20 > ---------------------------(end of broadcast)--------------------------- > TIP 9: In versions below 8.0, the planner will ignore your desire to > choose an index scan if your joining column's datatypes do not > match > From pgsql-jdbc-owner@postgresql.org Wed Aug 17 01:42:39 2005 X-Original-To: pgsql-jdbc-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4624252927; Wed, 17 Aug 2005 01:42:37 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 26067-01; Wed, 17 Aug 2005 04:42:33 +0000 (GMT) Received: from NSNOVPS00411.nacio.xythos.com (212-59.84.64.master-link.com [64.84.59.212]) by svr1.postgresql.org (Postfix) with ESMTP id 55B6552897; Wed, 17 Aug 2005 01:42:30 -0300 (ADT) X-MimeOLE: Produced By Microsoft Exchange V6.0.6487.1 Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----_=_NextPart_001_01C5A2E6.133B18A8" Subject: Performance problem using V3 protocol in jdbc driver Date: Tue, 16 Aug 2005 21:42:29 -0700 Message-ID: <03E7D3E231BB7B4A915A6581D4296CC6015D534E@NSNOVPS00411.nacio.xythos.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: Performance problem using V3 protocol in jdbc driver Thread-Index: AcWi5hZNs712lLDwQT+PcpWjcgBCsQ== From: "Barry Lind" To: Cc: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.074 required=5 tests=[AWL=-0.004, FORGED_RCVD_HELO=0.05, HTML_60_70=0.027, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/108 X-Sequence-Number: 13677 This is a multi-part message in MIME format. ------_=_NextPart_001_01C5A2E6.133B18A8 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable We just moved a large production instance of ours from Oracle to Postgres 8.0.3 on linux. When running on Oracle the machine hummed along using about 5% of the CPU easily handling the fairly constant load, after moving the data to Postgres the machine was pretty much maxed out on CPU and could no longer keep up with the transaction volume. On a hunch I switched the jdbc driver to using the V2 protocol and the load on the machine dropped down to what it was when using Oracle and everything was fine. =20 Now obviously I have found a work around for the performance problem, but I really don't want to rely on using the V2 protocol forever, and don't want to have to recommend to our customers that they need to run with the V2 protocol. So I would like to resolve the problem and be able to move back to a default configuration with the V3 protocol and the benefits thereof. =20 The problem is that I don't really know where to begin to debug a problem like this. In development environments and testing environments we have not seen performance problems with the V3 protocol in the jdbc driver. But they don't come close to approaching the transaction volume of this production instance. =20 What I see when running the V3 protocol under 'top' is that the postgres processes are routinely using 15% or more of the CPU each, when running the V2 protocol they use more like 0.3%. =20 Does anyone have any suggestions on an approach to debug a problem like this? =20 Thanks, --Barry ------_=_NextPart_001_01C5A2E6.133B18A8 Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable

We just moved a large production instance of ours = from Oracle to Postgres 8.0.3 on linux.  When running on Oracle the = machine hummed along using about 5% of the CPU easily handling the fairly = constant load, after moving the data to Postgres the machine was pretty much maxed out = on CPU and could no longer keep up with the transaction volume.  On a = hunch I switched the jdbc driver to using the V2 protocol and the load on the = machine dropped down to what it was when using Oracle and everything was = fine.

 

Now obviously I have found a work around for the = performance problem, but I really don’t want to rely on using the V2 protocol forever, and don’t want to have to recommend to our customers that = they need to run with the V2 protocol.  So I would like to resolve the = problem and be able to move back to a default configuration with the V3 protocol = and the benefits thereof.

 

The problem is that I don’t really know where = to begin to debug a problem like this.  In development environments and = testing environments we have not seen performance problems with the V3 protocol = in the jdbc driver.  But they don’t come close to approaching the transaction volume of this production = instance.

 

What I see when running the V3 protocol under = ‘top’ is that the postgres processes are routinely using 15% or more of the = CPU each, when running the V2 protocol they use more like = 0.3%.

 

Does anyone have any suggestions on an approach to = debug a problem like this?

 

Thanks,

--Barry

------_=_NextPart_001_01C5A2E6.133B18A8-- From pgsql-performance-owner@postgresql.org Wed Aug 17 01:49:02 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 147825297F for ; Wed, 17 Aug 2005 01:48:31 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 27495-03 for ; Wed, 17 Aug 2005 04:48:29 +0000 (GMT) Received: from smtpauth04.mail.atl.earthlink.net (smtpauth04.mail.atl.earthlink.net [209.86.89.64]) by svr1.postgresql.org (Postfix) with ESMTP id 333B952956 for ; Wed, 17 Aug 2005 01:48:27 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=Q9qkzjv+vdZ9REwC0rN4tbIK6jhXxXcmmVLliM9zMl/0Bpojdx2VaB3q3bkERWy+; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:Cc:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth04.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E5Fqk-00019n-QJ; Wed, 17 Aug 2005 00:48:26 -0400 Message-Id: <6.2.3.4.0.20050817001405.01fc3cd8@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Wed, 17 Aug 2005 00:48:26 -0400 To: "Kari Lavikka" From: Ron Subject: Re: Finding bottleneck Cc: pgsql-performance@postgresql.org Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bc963dc7ab735b43a6dce1f6a9b0f1afd0350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.374 required=5 tests=[DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/196 X-Sequence-Number: 13943 I think I have a solution for you. You have posted that you presently have these RAID volumes and behaviors: sda: data (10 spindles, raid10) sdb: xlog & clog (2 spindles, raid1) sdc: os and other stuff Usually iostat (2 second interval) says: avg-cpu: %user %nice %sys %iowait %idle 32.38 0.00 12.88 11.62 43.12 Device: tps kB_read/s kB_wrtn/s kB_read kB_wrtn sda 202.00 1720.00 0.00 3440 0 sdb 152.50 4.00 2724.00 8 5448 sdc 0.00 0.00 0.00 0 0 And during checkpoint: avg-cpu: %user %nice %sys %iowait %idle 31.25 0.00 14.75 54.00 0.00 Device: tps kB_read/s kB_wrtn/s kB_read kB_wrtn sda 3225.50 1562.00 35144.00 3124 70288 sdb 104.50 10.00 2348.00 20 4696 sdc 0.00 0.00 0.00 0 0 During checkpoints sda is becoming saturated, essentially halting all other DB activity involving sda. A lesser version of the porblem is probably occurring every time multiple entities on sda are being accessed simultaneously, particularly simultaneous writes. My Proposed Solution: Put comment and its index on it's own dedicated RAID volume. Put comment_archive and its index on its own dedicated RAID volume. Put the rest of the tables currently part of "data" on their own dedicated RAID volume. Put the rest if the indexes to the tables currently part of "data" on their own dedicated RAID volume. Put xlog on its own dedicated RAID volume. The general idea here is to put any tables or indexes that tend to require simultaneous access, particularly write access, on different spindles. Like all things, there's a point of diminishing returns that is dependent on the HW used and the DB load. If you must wring every last bit of IO out of the HD subsystem, a more exact set of spindle assignments can be made by analyzing your queries and then 1) make sure writes that tend to be simultaneous are to different spindles, then (if you still need better IO) 2) make sure reads that tend to be simultaneous are to different spindles. At some point, your controller will become the bottleneck. At some point beyond that, the IO channels on the mainboard will become the bottleneck. My suggestion should get you to within 80-90% of optimal if I've understood the implications of your posts correctly. The other suggestion I'd make is to bump your RAM from 16GB to 32GB as soon as you can afford it and then tune your PostgreSQL parameters to make best use of it. The more RAM resident your DB, the better. Hope this helps, Ron Peacetree ===========Original Message Follows=========== From: Kari Lavikka To: Merlin Moncure Subject: Re: Finding bottleneck Date: Mon, 8 Aug 2005 19:19:09 +0300 (EETDST) ---------- Actually I modified postgresql.conf a bit and there isn't commit delay any more. That didn't make noticeable difference though.. Workload is generated by a website with about 1000 dynamic page views a second. Finland's biggest site among youths btw. Anyway, there are about 70 tables and here's some of the most important: relname | reltuples ----------------------------------+------------- comment | 1.00723e+08 comment_archive | 9.12764e+07 channel_comment | 6.93912e+06 image | 5.80314e+06 admin_event | 5.1936e+06 user_channel | 3.36877e+06 users | 325929 channel | 252267 Queries to "comment" table are mostly IO-bound but are performing quite well. Here's an example: (SELECT u.nick, c.comment, c.private, c.admin, c.visible, c.parsable, c.uid_sender, to_char(c.stamp, 'DD.MM.YY HH24:MI') AS stamp, c.comment_id FROM comment c INNER JOIN users u ON u.uid = c.uid_sender WHERE u.status = 'a' AND c.image_id = 15500900 AND c.uid_target = 780345 ORDER BY uid_target DESC, image_id DESC, c.comment_id DESC) LIMIT 36 And explain analyze: Limit (cost=0.00..6.81 rows=1 width=103) (actual time=0.263..17.522 rows=12 loops=1) -> Nested Loop (cost=0.00..6.81 rows=1 width=103) (actual time=0.261..17.509 rows=12 loops=1) -> Index Scan Backward using comment_uid_target_image_id_comment_id_20050527 on "comment" c (cost=0.00..3.39 rows=1 width=92) (actual time=0.129..16.213 rows=12 loops=1) Index Cond: ((uid_target = 780345) AND (image_id = 15500900)) -> Index Scan using users_pkey on users u (cost=0.00..3.40 rows=1 width=15) (actual time=0.084..0.085 rows=1 loops=12) Index Cond: (u.uid = "outer".uid_sender) Filter: (status = 'a'::bpchar) Total runtime: 17.653 ms We are having performance problems with some smaller tables and very simple queries. For example: SELECT u.uid, u.nick, extract(epoch from uc.stamp) AS stamp FROM user_channel uc INNER JOIN users u USING (uid) WHERE channel_id = 281321 AND u.status = 'a' ORDER BY uc.channel_id, upper(uc.nick) And explain analyze: Nested Loop (cost=0.00..200.85 rows=35 width=48) (actual time=0.414..38.128 rows=656 loops=1) -> Index Scan using user_channel_channel_id_nick on user_channel uc (cost=0.00..40.18 rows=47 width=27) (actual time=0.090..0.866 rows=667 loops=1) Index Cond: (channel_id = 281321) -> Index Scan using users_pkey on users u (cost=0.00..3.40 rows=1 width=25) (actual time=0.048..0.051 rows=1 loops=667) Index Cond: ("outer".uid = u.uid) Filter: (status = 'a'::bpchar) Total runtime: 38.753 ms Under heavy load these queries tend to take several minutes to execute although there's plenty of free cpu available. There aren't any blocking locks in pg_locks. |\__/| ( oo ) Kari Lavikka - tuner ( at ) bdb ( dot ) fi - (050) 380 3808 __ooO( )Ooo_______ _____ ___ _ _ _ _ _ _ _ "" From pgsql-jdbc-owner@postgresql.org Wed Aug 17 02:01:55 2005 X-Original-To: pgsql-jdbc-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 43575529DA; Wed, 17 Aug 2005 02:01:53 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 32118-03; Wed, 17 Aug 2005 05:01:48 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 1C195529C6; Wed, 17 Aug 2005 02:01:45 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7H51j41012038; Wed, 17 Aug 2005 01:01:45 -0400 (EDT) To: "Barry Lind" Cc: pgsql-performance@postgresql.org, pgsql-jdbc@postgresql.org Subject: Re: Performance problem using V3 protocol in jdbc driver In-reply-to: <03E7D3E231BB7B4A915A6581D4296CC6015D534E@NSNOVPS00411.nacio.xythos.com> References: <03E7D3E231BB7B4A915A6581D4296CC6015D534E@NSNOVPS00411.nacio.xythos.com> Comments: In-reply-to "Barry Lind" message dated "Tue, 16 Aug 2005 21:42:29 -0700" Date: Wed, 17 Aug 2005 01:01:45 -0400 Message-ID: <12037.1124254905@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/109 X-Sequence-Number: 13678 "Barry Lind" writes: > ... On a hunch I switched the jdbc driver to using the V2 protocol > and the load on the machine dropped down to what it was when using > Oracle and everything was fine. First knee-jerk reaction is that it's an optimization problem stemming from V3 protocol feeding parameterized queries to the backend where V2 did not, and the planner being unable to cope :-( Can you identify the specific queries causing the problem? regards, tom lane From pgsql-jdbc-owner@postgresql.org Wed Aug 17 02:43:48 2005 X-Original-To: pgsql-jdbc-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8161D529EC; Wed, 17 Aug 2005 02:43:45 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 36727-09; Wed, 17 Aug 2005 05:43:43 +0000 (GMT) Received: from NSNOVPS00411.nacio.xythos.com (212-59.84.64.master-link.com [64.84.59.212]) by svr1.postgresql.org (Postfix) with ESMTP id 275EF529D6; Wed, 17 Aug 2005 02:43:40 -0300 (ADT) X-MimeOLE: Produced By Microsoft Exchange V6.0.6487.1 Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Subject: Re: Performance problem using V3 protocol in jdbc driver Date: Tue, 16 Aug 2005 22:43:40 -0700 Message-ID: <03E7D3E231BB7B4A915A6581D4296CC6015D5352@NSNOVPS00411.nacio.xythos.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [JDBC] Performance problem using V3 protocol in jdbc driver Thread-Index: AcWi6MdL6nw11NNHQ9q4dDVy3dz7XQAA4nBw From: "Barry Lind" To: "Tom Lane" Cc: , X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.061 required=5 tests=[AWL=0.011, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/110 X-Sequence-Number: 13679 That was my suspicion as well, which is why I tried the V2 protocol. =20 I do not know of any specific queries that are causing the problem. As I monitor 'top' I see processes utilizing a significant amount of CPU running SELECT, UPDATE and DELETE, which would lead me to believe that it isn't any one specific query. How does one identify on a live system specific queries that are running slow, especially with the V3 protocol and when the system is executing about a 100 queries a second (which makes turning on any sort of logging very very verbose)? (I just subscribed to the performance list, so this is probably something that has been answered many times before on this list). I haven't tried to track down a performance problem like this before on postgres. Since most of our large customers run Oracle that is where I have the knowledge to figure something like this out. Thanks, --Barry -----Original Message----- From: Tom Lane [mailto:tgl@sss.pgh.pa.us]=20 Sent: Tuesday, August 16, 2005 10:02 PM To: Barry Lind Cc: pgsql-performance@postgresql.org; pgsql-jdbc@postgresql.org Subject: Re: [JDBC] Performance problem using V3 protocol in jdbc driver "Barry Lind" writes: > ... On a hunch I switched the jdbc driver to using the V2 protocol > and the load on the machine dropped down to what it was when using > Oracle and everything was fine. First knee-jerk reaction is that it's an optimization problem stemming from V3 protocol feeding parameterized queries to the backend where V2 did not, and the planner being unable to cope :-( Can you identify the specific queries causing the problem? regards, tom lane From pgsql-performance-owner@postgresql.org Wed Aug 17 05:16:31 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CA54652B1E for ; Wed, 17 Aug 2005 05:14:26 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91460-05 for ; Wed, 17 Aug 2005 08:14:25 +0000 (GMT) Received: from mail0.rawbw.com (mail0.rawbw.com [198.144.192.41]) by svr1.postgresql.org (Postfix) with ESMTP id BC77452B38 for ; Wed, 17 Aug 2005 05:14:23 -0300 (ADT) Received: (from www@localhost) by mail0.rawbw.com (8.11.6p2/8.11.6) id j7H8EPA52792 for pgsql-performance@postgresql.org; Wed, 17 Aug 2005 01:14:25 -0700 (PDT) Received: from 198.144.203.173 ([198.144.203.173]) by webmail.rawbw.com (IMP) with HTTP for ; Wed, 17 Aug 2005 01:14:24 -0700 Message-ID: <1124266464.4302f1e0d5d30@webmail.rawbw.com> Date: Wed, 17 Aug 2005 01:14:24 -0700 From: mudfoot@rawbw.com To: pgsql-performance@postgresql.org Subject: Re: Performance problem using V3 protocol in jdbc driver References: <03E7D3E231BB7B4A915A6581D4296CC6015D534E@NSNOVPS00411.nacio.xythos.com> In-Reply-To: <03E7D3E231BB7B4A915A6581D4296CC6015D534E@NSNOVPS00411.nacio.xythos.com> MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit User-Agent: Internet Messaging Program (IMP) 3.2.1 X-Originating-IP: 198.144.203.173 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.178 required=5 tests=[NO_REAL_NAME=0.178] X-Spam-Level: X-Archive-Number: 200508/199 X-Sequence-Number: 13946 Quoting Barry Lind : > > > What I see when running the V3 protocol under 'top' is that the postgres > processes are routinely using 15% or more of the CPU each, when running > the V2 protocol they use more like 0.3%. > > > > Does anyone have any suggestions on an approach to debug a problem like > this? > > Tracing system calls is a good starting point--truss on Solaris, strace on Linux (Redhat anyway), ktrace on BSD. The difference between 0.3% and 15% CPU utilization under similar load will very likely (though not with complete certainty) be showing very noticeably different system call activity. If you notice a difference in system call activity, then that would probably provide a hint as to what's going on--where the inefficiency lies. It's possible to spin the CPU up without any system calls, but system call tracing can be done pretty quickly and you should be able to see any interesting patterns emerge quite quickly. ^ | This method is a good starting point for troubleshooting just about any funny process activity. And it comes with the added benefit of not having to know ahead of time about the specific matter at hand (JDBC implementation, in this case). :-) That's having your cake and eating it, too. > > Thanks, > > --Barry > > From pgsql-performance-owner@postgresql.org Wed Aug 17 06:13:23 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5695252B42 for ; Wed, 17 Aug 2005 06:13:22 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 15607-04 for ; Wed, 17 Aug 2005 09:13:20 +0000 (GMT) Received: from thor.netera.se (thor.netera.se [85.112.172.11]) by svr1.postgresql.org (Postfix) with ESMTP id 7401452B48 for ; Wed, 17 Aug 2005 06:13:18 -0300 (ADT) Received: from [192.168.2.240] (1-1-1-41a.o.sth.bostream.se [81.26.246.14]) by thor.netera.se (Postfix) with ESMTP id 7E29313AC06F for ; Wed, 17 Aug 2005 11:13:19 +0200 (CEST) Message-ID: <4303003B.3020407@relevanttraffic.se> Date: Wed, 17 Aug 2005 11:15:39 +0200 From: Ulrich Wisser Organization: Relevant Traffic AB User-Agent: Mozilla Thunderbird 1.0.6 (Windows/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Re: Need for speed References: In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/200 X-Sequence-Number: 13947 Hello, thanks for all your suggestions. I can see that the Linux system is 90% waiting for disc io. At that time all my queries are *very* slow. My scsi raid controller and disc are already the fastest available. The query plan uses indexes and "vacuum analyze" is run once a day. To avoid aggregating to many rows, I already made some aggregation tables which will be updated after the import from the Apache logfiles. That did help, but only to a certain level. I believe the biggest problem is disc io. Reports for very recent data are quite fast, these are used very often and therefor already in the cache. But reports can contain (and regulary do) very old data. In that case the whole system slows down. To me this sounds like the recent data is flushed out of the cache and now all data for all queries has to be fetched from disc. My machine has 2GB memory, please find postgresql.conf below. Ulrich #--------------------------------------------------------------------------- # RESOURCE USAGE (except WAL) #--------------------------------------------------------------------------- # - Memory - shared_buffers = 20000 # min 16, at least max_connections*2, sort_mem = 4096 # min 64, size in KB vacuum_mem = 8192 # min 1024, size in KB # - Free Space Map - max_fsm_pages = 50000 # min max_fsm_relations*16, 6 bytes each max_fsm_relations = 3000 # min 100, ~50 bytes each # - Kernel Resource Usage - #max_files_per_process = 1000 # min 25 #preload_libraries = '' #--------------------------------------------------------------------------- # WRITE AHEAD LOG #--------------------------------------------------------------------------- # - Settings - fsync = false # turns forced synchronization on or off #wal_sync_method = fsync # the default varies across platforms: wal_buffers = 128 # min 4, 8KB each # - Checkpoints - checkpoint_segments = 16 # in logfile segments, min 1, 16MB each #checkpoint_timeout = 300 # range 30-3600, in seconds #checkpoint_warning = 30 # 0 is off, in seconds #commit_delay = 0 # range 0-100000, in microseconds #commit_siblings = 5 # range 1-1000 From pgsql-jdbc-owner@postgresql.org Wed Aug 17 06:31:12 2005 X-Original-To: pgsql-jdbc-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5D3EB52A43; Wed, 17 Aug 2005 06:30:56 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 20237-07; Wed, 17 Aug 2005 09:30:53 +0000 (GMT) Received: from mail.ecircle.de (mail.ecircle.de [195.140.186.200]) by svr1.postgresql.org (Postfix) with ESMTP id D0C8E529E6; Wed, 17 Aug 2005 06:30:51 -0300 (ADT) Received: from [192.168.1.110] (unknown [192.168.1.110]) by mail.ecircle.de (READY) with ESMTP id CE0FF120567; Wed, 17 Aug 2005 11:30:51 +0200 (CEST) Subject: Re: Performance problem using V3 protocol in jdbc driver From: Csaba Nagy To: Barry Lind Cc: pgsql-performance@postgresql.org, Postgres JDBC In-Reply-To: <03E7D3E231BB7B4A915A6581D4296CC6015D534E@NSNOVPS00411.nacio.xythos.com> References: <03E7D3E231BB7B4A915A6581D4296CC6015D534E@NSNOVPS00411.nacio.xythos.com> Content-Type: text/plain; charset=UTF-8 Message-Id: <1124271051.24337.56.camel@coppola.muc.ecircle.de> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Date: Wed, 17 Aug 2005 11:30:51 +0200 Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.011 required=5 tests=[AWL=0.011] X-Spam-Level: X-Archive-Number: 200508/111 X-Sequence-Number: 13680 Barry, I have made a similar experience, moving a big Oracle data base to Postgres 8.03 on linux. The first impact was similar, huge performance problems. The main problem was bad planner choices. The cause in our case: bad parameter types in the jdbc set methods (I guess you use Java). For oracle we used the NUMERIC type to set primary keys, but the postgres id type used was BIGINT, and it just refused to use the index in this case. Imagine that kicking in on a 100 million rows table... a sequential scan started a few times a second, now that made the DB unusable. So we fixed the code that for oracle continues to use NUMERIC and for postgres it uses BIGINT, and that is very important on setNull calls too. One very useful tool was the following query: prepare ps as SELECT procpid, substring(current_query for 97), to_char((now()-query_start), 'HH24:MI:SS') as t FROM pg_stat_activity where current_query not like '% We just moved a large production instance of ours from Oracle to > Postgres 8.0.3 on linux. When running on Oracle the machine hummed > along using about 5% of the CPU easily handling the fairly constant > load, after moving the data to Postgres the machine was pretty much > maxed out on CPU and could no longer keep up with the transaction > volume. On a hunch I switched the jdbc driver to using the V2 > protocol and the load on the machine dropped down to what it was when > using Oracle and everything was fine. > > > > Now obviously I have found a work around for the performance problem, > but I really don’t want to rely on using the V2 protocol forever, and > don’t want to have to recommend to our customers that they need to run > with the V2 protocol. So I would like to resolve the problem and be > able to move back to a default configuration with the V3 protocol and > the benefits thereof. > > > > The problem is that I don’t really know where to begin to debug a > problem like this. In development environments and testing > environments we have not seen performance problems with the V3 > protocol in the jdbc driver. But they don’t come close to approaching > the transaction volume of this production instance. > > > > What I see when running the V3 protocol under ‘top’ is that the > postgres processes are routinely using 15% or more of the CPU each, > when running the V2 protocol they use more like 0.3%. > > > > Does anyone have any suggestions on an approach to debug a problem > like this? > > > > Thanks, > > --Barry > > From pgsql-jdbc-owner@postgresql.org Wed Aug 17 06:48:35 2005 X-Original-To: pgsql-jdbc-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6490552956; Wed, 17 Aug 2005 06:48:32 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 24053-07; Wed, 17 Aug 2005 09:48:30 +0000 (GMT) Received: from smtp.bizmail.net4india.com (smtp.bizmail.net4india.com [202.71.129.102]) by svr1.postgresql.org (Postfix) with ESMTP id 43A6C528F0; Wed, 17 Aug 2005 06:48:26 -0300 (ADT) Received: from [61.11.17.20] (helo=devp20) by smtp.bizmail.net4india.com with smtp (Exim 3.36 #2) id 1E5KbJ-0000Y3-00; Wed, 17 Aug 2005 15:22:50 +0530 Message-ID: <009501c5a311$3706ef20$dd0aa8c0@codecindia.com> From: "Mahesh Shinde" To: Cc: Subject: Data Selection Slow From VB 6.0 Date: Wed, 17 Aug 2005 02:50:45 -0700 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0092_01C5A2D6.771DF440" X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 6.00.2800.1106 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1106 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.048 required=5 tests=[AWL=0.020, HTML_60_70=0.027, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/112 X-Sequence-Number: 13681 This is a multi-part message in MIME format. ------=_NextPart_000_0092_01C5A2D6.771DF440 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Hi I am using Postgres version=20 *PostgreSQL 7.4.5 on i686-pc-linux-gnu, compiled by GCC gcc = (GCC) 3.2.2 20030222 (Red Hat Linux 3.2.2-5).*=20 for an multy user desktop application using VB 6.0 as a front = end toll. To connect To the PostgreSQL I am using *PostgreSQL Win32 ODBC and = OLEDB client drivers 1.0.0.2*=20 The files included are=20 Version 1.0 of the PGW32CLI Installer will install the following = file versions. Files are installed in a separate PGW32CLI directory so = should not conflict with existing applications. libpq.dll 8.0.2.5098 (PostgreSQL library) libintl-2.dll 0.11.5.1189 (GNU Text Utils) libiconv-2.dll 1.8.1134.7927 (GNU Text Utils) psqlodbc.dll 8.0.0.4 (PG ODBC) pgoledb.dll 1.0.0.19 (PgOleDB) libeay32.dll 0.9.7.f (OpenSSL) ssleay32.dll 0.9.7.f (OpenSSL) =20 I have server configuration as=20 P4 3 GHz HT Tech 2 GB DDR RAM, Intel Original 875 Chipset Motherboard, 73 GB 10 K RPM SCSI HDD x 2 Nos. Adp SCSI Controller, (You can do software RAID on it) Server Class Cabinet =20 Since in the database I have one Major table that Debtor table which = is master table and having around 55 lac records. I have set debtorId as = a primary key having index on it.I am developing a search screen to = search a specific debtor info using this table.=20 When I fire a query to search a debtor id, it took around 5 seconds to = return an answer for a query whether entered debtor id is present in the = database or not using ODBC. Where as when Explian the query on the = database=20 Index Scan using tbmstban_debtorid on tbmstbandetails = (cost=3D0.00..6.01 rows=3D2 width=3D143) Index Cond: ((debtorid)::text =3D '234'::text) Query for the search criteria is=20 select * from tbmstdebtordetails where debtorid =3D'234'=20 Where as when I am using a like query to search a record starting with = debtor id having a characters then it took around 10-15 sec to return a = record set having records. query is=20 select * from tbmstdebtordetails where debtorid like '234%'=20 Explain output on the database Index Scan using tbmstban_debtorid on tbmstbandetails = (cost=3D0.00..6.01 rows=3D2 width=3D143) Index Cond: ((debtorid)::text =3D '234%'::text) Thanks & regards, Mahesh Shinde ------------------------------------------------------------------ Codec Communications (I) Pvt. Ltd. PUNE (INDIA) T # 91-20-24221460/70(Ext 43) Desk No. 25143 Email - mahesh.shinde@codecindia.com ------=_NextPart_000_0092_01C5A2D6.771DF440 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Hi
    I am using Postgres = version=20
       =20 *PostgreSQL 7.4.5 on i686-pc-linux-gnu, compiled by GCC gcc = (GCC) 3.2.2=20 20030222 (Red Hat Linux 3.2.2-5).*
        = for an multy=20 user desktop application using VB 6.0 as a = front end=20 toll.
 
    To connect To the PostgreSQL I am using = *PostgreSQL Win32=20 ODBC and OLEDB client drivers 1.0.0.2*
 
    The files included = are=20
    =    =20 Version 1.0 of the PGW32CLI Installer will install the following file = versions.=20 Files are installed in a separate PGW32CLI directory so should not = conflict with=20 existing            =20 applications.
    =    =20     libpq.dll 8.0.2.5098 = (PostgreSQL=20 library)
            = libintl-2.dll=20 0.11.5.1189 (GNU Text Utils)
            = libiconv-2.dll=20 1.8.1134.7927 (GNU Text Utils)
            = psqlodbc.dll=20 8.0.0.4 (PG ODBC)
            = pgoledb.dll=20 1.0.0.19 (PgOleDB)
            = libeay32.dll=20 0.9.7.f (OpenSSL)
            = ssleay32.dll=20 0.9.7.f (OpenSSL)
   =20
I have server configuration as =
    =    =20     P4 3 GHz HT = Tech
   =20         2 GB DDR RAM,
   =20         Intel Original 875 Chipset=20 Motherboard,
   =20         73 GB  10 K RPM SCSI HDD = x =20 2 Nos.
   =20         Adp SCSI Controller, (You can do = software=20 RAID on it)
   =20         Server Class = Cabinet
   =20
    = Since in the=20 database I have one Major table that Debtor table which is master table = and=20 having around 55 lac records. I have set debtorId as a primary key = having index=20 on it.I am developing a search screen to search a specific debtor=20 info using this table.
 
When I fire a query to = search a=20 debtor id,  it took around 5 seconds to return an answer for a = query=20 whether entered debtor id is present in the database or not using ODBC. = Where as=20 when Explian  the query on the database
 Index Scan using=20 tbmstban_debtorid on tbmstbandetails  (cost=3D0.00..6.01 rows=3D2=20 width=3D143)
   Index Cond: ((debtorid)::text =3D=20 '234'::text)

Query for the=20 search criteria is
 select * from = tbmstdebtordetails=20 where debtorid =3D'234'
 
 Where as when I = am using a like=20 query to search a record starting with debtor id having a characters = then it=20 took around 10-15 sec to return a record set having = records.
query is 
select * from = tbmstdebtordetails where=20 debtorid like '234%'
 
Explain output on the = database
 Index Scan=20 using tbmstban_debtorid on tbmstbandetails  (cost=3D0.00..6.01 = rows=3D2=20 width=3D143)
   Index Cond: ((debtorid)::text =3D=20 '234%'::text)
Thanks & regards,
Mahesh=20 Shinde
---------------------------------------------------------------= ---
Codec=20 Communications (I) Pvt. Ltd.
PUNE (INDIA)
T # = 91-20-24221460/70(Ext=20 43)
Desk No. 25143
Email =96 mahesh.shinde@codecindia.com=
------=_NextPart_000_0092_01C5A2D6.771DF440-- From pgsql-performance-owner@postgresql.org Wed Aug 17 07:29:08 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E6AEA52917 for ; Wed, 17 Aug 2005 07:29:06 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 34451-06 for ; Wed, 17 Aug 2005 10:29:03 +0000 (GMT) Received: from my.endian.it (unknown [62.146.87.34]) by svr1.postgresql.org (Postfix) with ESMTP id F1989528F0 for ; Wed, 17 Aug 2005 07:29:00 -0300 (ADT) Received: from dell.1006.org (host58-128.pool8251.interbusiness.it [82.51.128.58]) (authenticated (0 bits)) by my.endian.it (8.11.6/8.11.6) with ESMTP id j7HAKZx06697; Wed, 17 Aug 2005 12:20:35 +0200 Subject: Re: Data Selection Slow From VB 6.0 From: Chris Mair To: Mahesh Shinde Cc: pgsql-performance@postgresql.org In-Reply-To: <009501c5a311$3706ef20$dd0aa8c0@codecindia.com> References: <009501c5a311$3706ef20$dd0aa8c0@codecindia.com> Content-Type: text/plain Date: Wed, 17 Aug 2005 12:28:16 +0200 Message-Id: <1124274496.2651.9.camel@dell.1006.org> Mime-Version: 1.0 X-Mailer: Evolution 2.2.3 (2.2.3-2.fc4) Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 required=5 tests=[FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/202 X-Sequence-Number: 13949 > When I fire a query to search a debtor id, it took around 5 seconds > to return an answer for a query [...] Are you sure that time is actually spent in the database engine? Maybe there are DNS resolving issues or something... Did you try to execute the queries directly on the server from the psql shell? Bye, Chris. > From pgsql-jdbc-owner@postgresql.org Wed Aug 17 09:29:25 2005 X-Original-To: pgsql-jdbc-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 443915288A; Wed, 17 Aug 2005 09:29:18 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 74393-04; Wed, 17 Aug 2005 12:29:14 +0000 (GMT) Received: from yass.opencloud.co.nz (yass.opencloud.co.nz [203.79.85.162]) by svr1.postgresql.org (Postfix) with ESMTP id B827B528B7; Wed, 17 Aug 2005 09:29:09 -0300 (ADT) Received: from 203-79-104-60.cable.paradise.net.nz ([203.79.104.60] helo=[192.168.0.21]) by yass.opencloud.co.nz with asmtp (TLS-1.0:DHE_RSA_AES_128_CBC_SHA:16) (Exim 4.34) id 1E5N2e-0005ds-Lv; Thu, 18 Aug 2005 00:29:12 +1200 Message-ID: <43032D99.9050606@opencloud.com> Date: Wed, 17 Aug 2005 12:29:13 +0000 From: Oliver Jowett User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Mahesh Shinde Cc: pgsql-performance@postgresql.org, pgsql-jdbc@postgresql.org Subject: Re: Data Selection Slow From VB 6.0 References: <009501c5a311$3706ef20$dd0aa8c0@codecindia.com> In-Reply-To: <009501c5a311$3706ef20$dd0aa8c0@codecindia.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.009 required=5 tests=[AWL=0.009] X-Spam-Level: X-Archive-Number: 200508/113 X-Sequence-Number: 13682 Mahesh Shinde wrote: > Hi > I am using Postgres version > **PostgreSQL 7.4.5 on i686-pc-linux-gnu, compiled by GCC gcc > (GCC) 3.2.2 20030222 (Red Hat Linux 3.2.2-5).* * > for an multy user desktop application using VB 6.0 as a front > end toll. > > To connect To the PostgreSQL I am using **PostgreSQL Win32 ODBC and > OLEDB client drivers 1.0.0.2** pgsql-jdbc isn't relevant, then -- the JDBC driver is not involved. -O From pgsql-jdbc-owner@postgresql.org Wed Aug 17 10:00:20 2005 X-Original-To: pgsql-jdbc-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4DA06529DA for ; Wed, 17 Aug 2005 09:59:33 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 76722-10 for ; Wed, 17 Aug 2005 12:59:30 +0000 (GMT) Received: from net2.micro-automation.com (net2.micro-automation.com [64.7.141.29]) by svr1.postgresql.org (Postfix) with SMTP id 1F1F3528DB for ; Wed, 17 Aug 2005 09:59:28 -0300 (ADT) Received: (qmail 9823 invoked from network); 17 Aug 2005 12:59:32 -0000 Received: from dcdsl.ebox.com (HELO ?192.168.1.2?) (davec@64.7.143.116) by net2.micro-automation.com with SMTP; 17 Aug 2005 12:59:32 -0000 In-Reply-To: <03E7D3E231BB7B4A915A6581D4296CC6015D5352@NSNOVPS00411.nacio.xythos.com> References: <03E7D3E231BB7B4A915A6581D4296CC6015D5352@NSNOVPS00411.nacio.xythos.com> Mime-Version: 1.0 (Apple Message framework v733) Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: <76FC392F-3B3D-409E-A790-1D15B784C03E@fastcrypt.com> Cc: "Tom Lane" , , Content-Transfer-Encoding: 7bit From: Dave Cramer Subject: Re: [PERFORM] Performance problem using V3 protocol in jdbc driver Date: Wed, 17 Aug 2005 08:59:30 -0400 To: Barry Lind X-Mailer: Apple Mail (2.733) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.013 required=5 tests=[AWL=0.013] X-Spam-Level: X-Archive-Number: 200508/114 X-Sequence-Number: 13683 Barry, One way to do this is to turn logging on for calls over a certain duration log_duration in the config file. This will only log calls over n milliseconds. There's a tool called iron eye SQL that monitors JDBC calls. http://www.irongrid.com/ unfortunately I am getting DNS errors from that site right now. I do have a copy of their code if you need it. Dave On 17-Aug-05, at 1:43 AM, Barry Lind wrote: > That was my suspicion as well, which is why I tried the V2 protocol. > > I do not know of any specific queries that are causing the > problem. As > I monitor 'top' I see processes utilizing a significant amount of CPU > running SELECT, UPDATE and DELETE, which would lead me to believe that > it isn't any one specific query. > > How does one identify on a live system specific queries that are > running > slow, especially with the V3 protocol and when the system is executing > about a 100 queries a second (which makes turning on any sort of > logging > very very verbose)? (I just subscribed to the performance list, so > this > is probably something that has been answered many times before on this > list). > > I haven't tried to track down a performance problem like this > before on > postgres. Since most of our large customers run Oracle that is > where I > have the knowledge to figure something like this out. > > Thanks, > --Barry > > > -----Original Message----- > From: Tom Lane [mailto:tgl@sss.pgh.pa.us] > Sent: Tuesday, August 16, 2005 10:02 PM > To: Barry Lind > Cc: pgsql-performance@postgresql.org; pgsql-jdbc@postgresql.org > Subject: Re: [JDBC] Performance problem using V3 protocol in jdbc > driver > > > "Barry Lind" writes: > >> ... On a hunch I switched the jdbc driver to using the V2 protocol >> and the load on the machine dropped down to what it was when using >> Oracle and everything was fine. >> > > First knee-jerk reaction is that it's an optimization problem stemming > from V3 protocol feeding parameterized queries to the backend where V2 > did not, and the planner being unable to cope :-( > > Can you identify the specific queries causing the problem? > > regards, tom lane > > > ---------------------------(end of > broadcast)--------------------------- > TIP 3: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faq > > From pgsql-performance-owner@postgresql.org Wed Aug 17 11:52:50 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A783152938 for ; Wed, 17 Aug 2005 11:52:46 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 15317-02 for ; Wed, 17 Aug 2005 14:52:45 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id AEEB6528E6 for ; Wed, 17 Aug 2005 11:52:44 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7HEqmF4018976; Wed, 17 Aug 2005 10:52:49 -0400 (EDT) To: Ulrich Wisser Cc: pgsql-performance@postgresql.org Subject: Re: Need for speed In-reply-to: <4303003B.3020407@relevanttraffic.se> References: <4303003B.3020407@relevanttraffic.se> Comments: In-reply-to Ulrich Wisser message dated "Wed, 17 Aug 2005 11:15:39 +0200" Date: Wed, 17 Aug 2005 10:52:48 -0400 Message-ID: <18975.1124290368@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/204 X-Sequence-Number: 13951 Ulrich Wisser writes: > My machine has 2GB memory, please find postgresql.conf below. > max_fsm_pages = 50000 # min max_fsm_relations*16, 6 bytes each FWIW, that index I've been groveling through in connection with your other problem contains an astonishingly large amount of dead space --- almost 50%. I suspect that you need a much larger max_fsm_pages setting, and possibly more-frequent vacuuming, in order to keep a lid on the amount of wasted space. regards, tom lane From pgsql-jdbc-owner@postgresql.org Wed Aug 17 12:27:20 2005 X-Original-To: pgsql-jdbc-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A4FFF529AA; Wed, 17 Aug 2005 12:18:59 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 21041-05; Wed, 17 Aug 2005 15:18:56 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id BD111529D6; Wed, 17 Aug 2005 12:18:54 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050817151852m92009ic8le>; Wed, 17 Aug 2005 15:18:52 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 0484255FBA; Wed, 17 Aug 2005 10:18:52 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id E303555FAD; Wed, 17 Aug 2005 10:18:37 -0500 (CDT) Message-ID: <4303554B.2020500@arbash-meinel.com> Date: Wed, 17 Aug 2005 10:18:35 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Mahesh Shinde Cc: pgsql-performance@postgresql.org, pgsql-jdbc@postgresql.org Subject: Re: [PERFORM] Data Selection Slow From VB 6.0 References: <009501c5a311$3706ef20$dd0aa8c0@codecindia.com> In-Reply-To: <009501c5a311$3706ef20$dd0aa8c0@codecindia.com> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig620020274E7B6ABB6664657C" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.043 required=5 tests=[AWL=-0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/115 X-Sequence-Number: 13684 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig620020274E7B6ABB6664657C Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable Mahesh Shinde wrote: > Hi =2E.. > To connect To the PostgreSQL I am using **PostgreSQL Win32 ODBC and= > OLEDB client drivers 1.0.0.2** > =20 =2E.. > Since in the database I have one Major table that Debtor table whic= h > is master table and having around 55 lac records. I have set debtorId a= s > a primary key having index on it.I am developing a search screen to > search a specific debtor info using this table. > =20 > When I fire a query to search a debtor id, it took around 5 seconds to= > return an answer for a query whether entered debtor id is present in th= e > database or not using ODBC. Where as when Explian the query on the > database > Index Scan using tbmstban_debtorid on tbmstbandetails (cost=3D0.00..6= =2E01 > rows=3D2 width=3D143) > Index Cond: ((debtorid)::text =3D '234'::text) Are you checking this from the VB App? Or just going onto the server and running psql? (I'm guessing there is some way to run a flat query using VB. In which case you can just have the query run EXPLAIN ANALYZE, the return value is just the text, one line after another.) What I'm thinking is that it might be a locale/encoding issue. What is the encoding of your database? And what is the default locale and the locale that you are connecting as? Can you give us the "EXPLAIN ANALYZE" output so that we can see if the planner is doing what it thinks it is? It certainly sounds like either it is always doing a sequential scan, or something else is going on. 5 sec is a really long time for the type of query you are doing. Oh, and can you run the win32 psql client to see if it might be ODBC which is causing the problem? John =3D:-> >=20 > Query for the search criteria is > *select * from tbmstdebtordetails where debtorid =3D'234'* > =20 > Where as when I am using a like query to search a record starting with= > debtor id having a characters then it took around 10-15 sec to return a= > record set having records. > query is=20 > *select * from tbmstdebtordetails where debtorid like '234%'* > =20 > Explain output on the database > Index Scan using tbmstban_debtorid on tbmstbandetails (cost=3D0.00..6= =2E01 > rows=3D2 width=3D143) > Index Cond: ((debtorid)::text =3D '234%'::text) > Thanks & regards, > Mahesh Shinde > ------------------------------------------------------------------ > Codec Communications (I) Pvt. Ltd. > PUNE (INDIA) > T # 91-20-24221460/70(Ext 43) > Desk No. 25143 > Email =E2=80=93 mahesh.shinde@codecindia.com --------------enig620020274E7B6ABB6664657C Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDA1VLJdeBCYSNAAMRAswzAKCd7lERmivZb8CinkVh8ognvDjoqACdETdw EY5dlmeNaDewSiibDeK+vFQ= =hTR3 -----END PGP SIGNATURE----- --------------enig620020274E7B6ABB6664657C-- From pgsql-performance-owner@postgresql.org Wed Aug 17 13:50:47 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3B21E528DA for ; Wed, 17 Aug 2005 13:49:36 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 52195-05 for ; Wed, 17 Aug 2005 16:49:34 +0000 (GMT) Received: from mail24.sea5.speakeasy.net (mail24.sea5.speakeasy.net [69.17.117.26]) by svr1.postgresql.org (Postfix) with ESMTP id 227385288A for ; Wed, 17 Aug 2005 13:49:33 -0300 (ADT) Received: (qmail 27532 invoked from network); 17 Aug 2005 16:49:33 -0000 Received: from dsl081-060-184.sfo1.dsl.speakeasy.net (HELO noodles) ([64.81.60.184]) (envelope-sender ) by mail24.sea5.speakeasy.net (qmail-ldap-1.03) with RC4-MD5 encrypted SMTP for ; 17 Aug 2005 16:49:33 -0000 Subject: Re: Need for speed From: "Jeffrey W. Baker" To: Ulrich Wisser Cc: pgsql-performance@postgresql.org In-Reply-To: <4303003B.3020407@relevanttraffic.se> References: <4303003B.3020407@relevanttraffic.se> Content-Type: text/plain Date: Wed, 17 Aug 2005 09:49:37 -0700 Message-Id: <1124297377.423.2.camel@noodles> Mime-Version: 1.0 X-Mailer: Evolution 2.2.1.1 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/206 X-Sequence-Number: 13953 On Wed, 2005-08-17 at 11:15 +0200, Ulrich Wisser wrote: > Hello, > > thanks for all your suggestions. > > I can see that the Linux system is 90% waiting for disc io. At that time > all my queries are *very* slow. My scsi raid controller and disc are > already the fastest available. What RAID controller? Initially you said you have only 2 disks, and since you have your xlog on a separate spindle, I assume you have 1 disk for the xlog and 1 for the data. Even so, if you have a RAID, I'm going to further assume you are using RAID 1, since no sane person would use RAID 0. In those cases you are getting the performance of a single disk, which is never going to be very impressive. You need a RAID. Please be more precise when describing your system to this list. -jwb From pgsql-performance-owner@postgresql.org Wed Aug 17 14:53:58 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id EDB1C5288A for ; Wed, 17 Aug 2005 14:14:35 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 54975-07 for ; Wed, 17 Aug 2005 17:14:32 +0000 (GMT) Received: from zproxy.gmail.com (zproxy.gmail.com [64.233.162.197]) by svr1.postgresql.org (Postfix) with ESMTP id 225325284B for ; Wed, 17 Aug 2005 14:14:30 -0300 (ADT) Received: by zproxy.gmail.com with SMTP id 8so143042nzo for ; Wed, 17 Aug 2005 10:14:30 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:mime-version:content-type:content-transfer-encoding:content-disposition; b=le9asLzXXz7zEFvtRk44EHzPLW2cgBQvgngZgJxQ2X6HCLnEu/QzoH0orrM/3Tuq4kbKhtbibPWp6aSIh+YU3XLmI2xlguh1hvSiFrOah2RscgP1yPyR2KyPj2LWNOcRKjGX7rmo+Xvgr1F1acYF9t76rJiJ+hvu7C71w/lyp88= Received: by 10.36.215.10 with SMTP id n10mr661285nzg; Wed, 17 Aug 2005 10:14:30 -0700 (PDT) Received: by 10.36.24.2 with HTTP; Wed, 17 Aug 2005 10:14:29 -0700 (PDT) Message-ID: <1d219a6f05081710146acb9d40@mail.gmail.com> Date: Wed, 17 Aug 2005 13:14:29 -0400 From: Chris Hoover To: pgsql-performance@postgresql.org Subject: Tuning Effective Cache Question Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[AWL=-0.000, RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/209 X-Sequence-Number: 13956 I have some questions about tuning my effective_cache_size I have a RHEL 2.1 box running with dual Xeon (2.6 GHz I believe and they have HT on). The box has 8GB memory. In my postgresql.conf, I have set the effective_cache_size =3D 530505 (~4GB). However, I am noticing on this machine, top is telling me that I have ~3.5GB in the buff and only 3GB in cached. Here are the exact numbers: Mem: 7720040K av, 7714364K used, 5676K free, 314816K shrd, 3737540K bu= ff Swap: 2096440K av, 119448K used, 1976992K free 3188192K ca= ched 1. Is the configuration that linux is running in hurting PostgreSQL in any= way? 2. Is there a negative impact of the effective_cache_size being larger than the actual cached memory? 3. What effect postive or negative does the buff memory have on PostgreSQL= .\ 4. What exactly is this buff memory used for? We have had a time trying to find a good explanation of what it means. Overall, this system appears to be running fine. However, I was taken back when I saw the current memory configuration. We have been walking a fine line performance wise, and can quickly become i/o starved. So I want to make sure that I am not pushing the db towards the i/o starved side. Thanks for any insight, Chris From pgsql-performance-owner@postgresql.org Wed Aug 17 14:52:47 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E51FA5288A for ; Wed, 17 Aug 2005 14:17:37 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 60870-01 for ; Wed, 17 Aug 2005 17:17:35 +0000 (GMT) Received: from zproxy.gmail.com (zproxy.gmail.com [64.233.162.202]) by svr1.postgresql.org (Postfix) with ESMTP id 55B5A52BD7 for ; Wed, 17 Aug 2005 14:17:34 -0300 (ADT) Received: by zproxy.gmail.com with SMTP id 8so143435nzo for ; Wed, 17 Aug 2005 10:17:34 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=biNdqOqNLbMTnuoY+edDq6gJ8kgNC++vypfQ1YlVt/c+URjP5uqdlLDeMfKRxy1D8iVfvgLK9dEs8fKv64b51ABxNMplkhGVhFDNXxn2e2w0eqPdU1Sk3vWNUyhlg44AJGMXvkThwX55ZzPf+63u2iBOH812gSMkTdby8g9ak0w= Received: by 10.36.9.2 with SMTP id 2mr657392nzi; Wed, 17 Aug 2005 10:17:34 -0700 (PDT) Received: by 10.36.24.2 with HTTP; Wed, 17 Aug 2005 10:17:33 -0700 (PDT) Message-ID: <1d219a6f050817101729be31a0@mail.gmail.com> Date: Wed, 17 Aug 2005 13:17:33 -0400 From: Chris Hoover To: pgsql-performance@postgresql.org Subject: Re: Tuning Effective Cache Question In-Reply-To: <1d219a6f05081710146acb9d40@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <1d219a6f05081710146acb9d40@mail.gmail.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/208 X-Sequence-Number: 13955 Sorry, forgot to state that we are still on PG 7.3.4. On 8/17/05, Chris Hoover wrote: > I have some questions about tuning my effective_cache_size >=20 > I have a RHEL 2.1 box running with dual Xeon (2.6 GHz I believe and > they have HT on). The box has 8GB memory. In my postgresql.conf, I > have set the effective_cache_size =3D 530505 (~4GB). >=20 > However, I am noticing on this machine, top is telling me that I have > ~3.5GB in the buff and only 3GB in cached. >=20 > Here are the exact numbers: > Mem: 7720040K av, 7714364K used, 5676K free, 314816K shrd, 3737540K = buff > Swap: 2096440K av, 119448K used, 1976992K free 3188192K = cached >=20 >=20 > 1. Is the configuration that linux is running in hurting PostgreSQL in a= ny way? >=20 > 2. Is there a negative impact of the effective_cache_size being > larger than the actual cached memory? >=20 > 3. What effect postive or negative does the buff memory have on PostgreS= QL.\ >=20 > 4. What exactly is this buff memory used for? We have had a time > trying to find a good explanation of what it means. >=20 > Overall, this system appears to be running fine. However, I was taken > back when I saw the current memory configuration. We have been > walking a fine line performance wise, and can quickly become i/o > starved. So I want to make sure that I am not pushing the db towards > the i/o starved side. >=20 > Thanks for any insight, >=20 > Chris > From pgsql-performance-owner@postgresql.org Wed Aug 17 14:31:58 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4E39152BDF for ; Wed, 17 Aug 2005 14:27:10 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 62916-03 for ; Wed, 17 Aug 2005 17:27:05 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id 8599752C09 for ; Wed, 17 Aug 2005 14:27:02 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [63.195.55.98] (account josh@agliodbs.com HELO spooky) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7742017; Wed, 17 Aug 2005 10:29:18 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Need for speed Date: Wed, 17 Aug 2005 10:28:04 -0700 User-Agent: KMail/1.8 Cc: Ulrich Wisser References: <4303003B.3020407@relevanttraffic.se> In-Reply-To: <4303003B.3020407@relevanttraffic.se> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200508171028.05028.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.012 required=5 tests=[AWL=0.012] X-Spam-Level: X-Archive-Number: 200508/207 X-Sequence-Number: 13954 Ulrich, > I believe the biggest problem is disc io. Reports for very recent data > are quite fast, these are used very often and therefor already in the > cache. But reports can contain (and regulary do) very old data. In that > case the whole system slows down. To me this sounds like the recent data > is flushed out of the cache and now all data for all queries has to be > fetched from disc. How large is the database on disk? > My machine has 2GB memory, please find postgresql.conf below. hmmmm ... effective_cache_size? random_page_cost? cpu_tuple_cost? etc. -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Wed Aug 17 15:37:16 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CC5EC528CD for ; Wed, 17 Aug 2005 15:00:52 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 74095-04 for ; Wed, 17 Aug 2005 18:00:48 +0000 (GMT) Received: from smtpauth01.mail.atl.earthlink.net (smtpauth01.mail.atl.earthlink.net [209.86.89.61]) by svr1.postgresql.org (Postfix) with ESMTP id CE74A528BB for ; Wed, 17 Aug 2005 15:00:47 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=WE8seZa9OEPVkg00hy6KujpVn4H4sMJhsIwypfw+53FFOFucg3IScwEl0xKexqsZ; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth01.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E5SDX-00020O-0I; Wed, 17 Aug 2005 14:00:47 -0400 Message-Id: <6.2.3.4.0.20050817130842.05d31a40@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Wed, 17 Aug 2005 14:00:14 -0400 To: Ulrich Wisser , pgsql-performance@postgresql.org From: Ron Subject: Re: Need for speed In-Reply-To: <4303003B.3020407@relevanttraffic.se> References: <4303003B.3020407@relevanttraffic.se> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bcb2c633390d1eadcf586367faf0372854350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.484 required=5 tests=[AWL=-1.298, BUY_DIRECT=1.779, DNS_FROM_RFC_ABUSE=0.374, URIBL_SBL=0.629] X-Spam-Level: * X-Archive-Number: 200508/210 X-Sequence-Number: 13957 At 05:15 AM 8/17/2005, Ulrich Wisser wrote: >Hello, > >thanks for all your suggestions. > >I can see that the Linux system is 90% waiting for disc io. A clear indication that you need to improve your HD IO subsystem. >At that time all my queries are *very* slow. To be more precise, your server performance at that point is essentially equal to your HD IO subsystem performance. > My scsi raid controller and disc are already the fastest available. Oh, REALLY? This is the description of the system you gave us: "We have a box with Linux Fedora Core 3, Postgres 7.4.2 Intel(R) Pentium(R) 4 CPU 2.40GHz 2 scsi 76GB disks (15.000RPM, 2ms)" The is far, Far, FAR from the "the fastest available" in terms of SW, OS, CPU host, _or_ HD subsystem. The "fastest available" means 1= you should be running 8.0.3 2= you should be running the latest stable 2.6 based kernel 3= you should be running an Opteron based server 4= Fibre Channel HDs are higher performance than SCSI ones. 5= (and this is the big one) YOU NEED MORE SPINDLES AND A HIGHER END RAID CONTROLLER. The absolute "top of the line" for RAID controllers is something based on Fibre Channel from Xyratex (who make the RAID engines for EMC and NetApps), Engino (the enterprise division of LSI Logic who sell mostly to IBM. Apple has a server based on an Engino card), dot-hill (who bought Chaparral among others). I suspect you can't afford them even if they would do business with you. The ante for a FC-based RAID subsystem in this class is in the ~$32K to ~$128K range, even if you buy direct from the actual RAID HW manufacturer rather than an OEM like In the retail commodity market, the current best RAID controllers are probably the 16 and 24 port versions of the Areca cards ( www.areca.us ). They come darn close to saturating the the Real World Peak Bandwidth of a 64b 133MHz PCI-X bus. I did put pg_xlog on another file system on other discs. > The query plan uses indexes and "vacuum analyze" is run once a day. That >To avoid aggregating to many rows, I already made some aggregation >tables which will be updated after the import from the Apache >logfiles. That did help, but only to a certain level. > >I believe the biggest problem is disc io. Reports for very recent >data are quite fast, these are used very often and therefor already >in the cache. But reports can contain (and regulary do) very old >data. In that case the whole system slows down. To me this sounds >like the recent data is flushed out of the cache and now all data >for all queries has to be fetched from disc. > >My machine has 2GB memory, From pgsql-performance-owner@postgresql.org Wed Aug 17 16:42:15 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id DB0B352968 for ; Wed, 17 Aug 2005 15:36:58 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83253-05 for ; Wed, 17 Aug 2005 18:36:51 +0000 (GMT) Received: from smtpauth07.mail.atl.earthlink.net (smtpauth07.mail.atl.earthlink.net [209.86.89.67]) by svr1.postgresql.org (Postfix) with ESMTP id DA86252ABF for ; Wed, 17 Aug 2005 15:36:50 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=Emjj+fu/ul/LU/IpUZ9l9asLa9VX5Wq6zPi3KrFaG1aE47o5TPqV/xrqdgxhAtbb; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth07.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E5SmN-0005rx-LQ; Wed, 17 Aug 2005 14:36:47 -0400 Message-Id: <6.2.3.4.0.20050817140116.05d317b0@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Wed, 17 Aug 2005 14:33:20 -0400 To: Ulrich Wisser , pgsql-performance@postgresql.org From: Ron Subject: Re: Need for speed Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bc9ef602ea55fa4e9e544bee205330bd47350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.917 required=5 tests=[AWL=-0.865, BUY_DIRECT=1.779, DNS_FROM_RFC_ABUSE=0.374, URIBL_SBL=0.629] X-Spam-Level: * X-Archive-Number: 200508/211 X-Sequence-Number: 13958 At 05:15 AM 8/17/2005, Ulrich Wisser wrote: >Hello, > >thanks for all your suggestions. > >I can see that the Linux system is 90% waiting for disc io. A clear indication that you need to improve your HD IO subsystem if possible. >At that time all my queries are *very* slow. To be more precise, your server performance at that point is essentially equal to your HD IO subsystem performance. > My scsi raid controller and disc are already the fastest available. Oh, REALLY? This is the description of the system you gave us: "We have a box with Linux Fedora Core 3, Postgres 7.4.2 Intel(R) Pentium(R) 4 CPU 2.40GHz 2 scsi 76GB disks (15.000RPM, 2ms)" The is far, Far, FAR from the "the fastest available" in terms of SW, OS, CPU host, _or_ HD subsystem. The "fastest available" means 1= you should be running PostgreSQL 8.0.3 2= you should be running the latest stable 2.6 based kernel 3= you should be running an Opteron based server 4= Fibre Channel HDs are slightly higher performance than SCSI ones. 5= (and this is the big one) YOU NEED MORE SPINDLES AND A HIGHER END RAID CONTROLLER. Your description of you workload was: "one of our services is click counting for on line advertising. We do this by importing Apache log files every five minutes. This results in a lot of insert and delete statements. At the same time our customers shall be able to do on line reporting." There are two issues here: 1= your primary usage is OLTP-like, but you are also expecting to do reports against the same schema that is supporting your OLTP-like usage. Bad Idea. Schemas that are optimized for reporting and other data mining like operation are pessimal for OLTP-like applications and vice versa. You need two schemas: one optimized for lots of inserts and deletes (OLTP-like), and one optimized for reporting (data-mining like). 2= 2 spindles, even 15K rpm spindles, is minuscule. Real enterprise class RAID subsystems have at least 10-20x that many spindles, usually split into 6-12 sets dedicated to different groups of tables in the DB. Putting xlog on its own dedicated spindles is just the first step. The absolute "top of the line" for RAID controllers is something based on Fibre Channel from Xyratex (who make the RAID engines for EMC and NetApps), Engino (the enterprise division of LSI Logic who sell mostly to IBM. Apple has a server based on an Engino card), or dot-hill (who bought Chaparral among others). I suspect you can't afford them even if they would do business with you. The ante for a FC-based RAID subsystem in this class is in the ~$32K to ~$128K range, even if you buy direct from the actual RAID HW manufacturer rather than an OEM like EMC, IBM, or NetApp who will 2x or 4x the price. OTOH, these subsystems will provide OLTP or OLTP-like DB apps with performance that is head-and-shoulders better than anything else to be found. Numbers like 50K-200K IOPS. You get what you pay for. In the retail commodity market where you are more realistically going to be buying, the current best RAID controllers are probably the Areca cards ( www.areca.us ). They come darn close to saturating the Real World Peak Bandwidth of a 64b 133MHz PCI-X bus and have better IOPS numbers than their commodity brethren. However, _none_ of the commodity RAID cards have IOPS numbers anywhere near as high as those mentioned above. >To avoid aggregating to many rows, I already made some aggregation >tables which will be updated after the import from the Apache >logfiles. That did help, but only to a certain level. > >I believe the biggest problem is disc io. Reports for very recent >data are quite fast, these are used very often and therefor already >in the cache. But reports can contain (and regulary do) very old >data. In that case the whole system slows down. To me this sounds >like the recent data is flushed out of the cache and now all data >for all queries has to be fetched from disc. I completely agree. Hopefully my above suggestions make sense and are of use to you. >My machine has 2GB memory, ...and while we are at it, OLTP like apps benefit less from RAM than data mining ones, but still 2GB of RAM is just not that much for a real DB server... Ron Peacetree From pgsql-performance-owner@postgresql.org Wed Aug 17 18:57:12 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id DFC485288A for ; Wed, 17 Aug 2005 17:25:36 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 39461-01 for ; Wed, 17 Aug 2005 20:25:32 +0000 (GMT) Received: from email.aon.at (warsl404pip7.highway.telekom.at [195.3.96.91]) by svr1.postgresql.org (Postfix) with ESMTP id 42D47528CD for ; Wed, 17 Aug 2005 17:25:30 -0300 (ADT) Received: (qmail 3258 invoked from network); 17 Aug 2005 20:25:29 -0000 Received: from m160p019.dipool.highway.telekom.at (HELO fermat.Koizar.inf) ([62.46.9.243]) (envelope-sender ) by smarthub73.highway.telekom.at (qmail-ldap-1.03) with SMTP for ; 17 Aug 2005 20:25:29 -0000 From: Manfred Koizar To: Richard Huxton Cc: Yves Vindevogel , 'Postgresql Performance' Subject: Re: Insert performance (OT?) Date: Wed, 17 Aug 2005 22:25:09 +0200 Message-ID: References: <42DCCA3B.8070100@archonet.com> <42DCDB47.3060009@archonet.com> In-Reply-To: <42DCDB47.3060009@archonet.com> X-Mailer: Forte Agent 2.0/32.652 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.317 required=5 tests=[AWL=0.060, FORGED_RCVD_HELO=0.05, UPPERCASE_25_50=0.207] X-Spam-Level: X-Archive-Number: 200508/216 X-Sequence-Number: 13963 On Tue, 19 Jul 2005 11:51:51 +0100, Richard Huxton wrote: >You could get away with one query if you converted them to left-joins: >INSERT INTO ... >SELECT * FROM upload LEFT JOIN ... WHERE f3 IS NULL >UNION >SELECT * FROM upload LEFT JOIN ... WHERE f4 IS NULL For the archives: This won't work. Each of the two SELECTs eliminates rows violating one of the two constraints but includes rows violating the other constraint. After the UNION you are back to violating both constraints :-( Servus Manfred From pgsql-performance-owner@postgresql.org Wed Aug 17 18:40:50 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BDDD352802 for ; Wed, 17 Aug 2005 17:28:34 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 30915-09 for ; Wed, 17 Aug 2005 20:28:31 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id D170E52968 for ; Wed, 17 Aug 2005 17:28:30 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [64.81.245.111] (account josh@agliodbs.com HELO temoku.sf.agliodbs.com) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7742974; Wed, 17 Aug 2005 13:30:47 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Tuning Effective Cache Question Date: Wed, 17 Aug 2005 13:31:38 -0700 User-Agent: KMail/1.8 Cc: Chris Hoover References: <1d219a6f05081710146acb9d40@mail.gmail.com> In-Reply-To: <1d219a6f05081710146acb9d40@mail.gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200508171331.38980.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.049 required=5 tests=[AWL=-0.001, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/215 X-Sequence-Number: 13962 Chris, > I have a RHEL 2.1 box running with dual Xeon (2.6 GHz I believe and > they have HT on). The box has 8GB memory. In my postgresql.conf, I > have set the effective_cache_size = 530505 (~4GB). > > However, I am noticing on this machine, top is telling me that I have > ~3.5GB in the buff and only 3GB in cached. effective_cache_size is just information for the query planner; it does not affect the actually system file cache. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Wed Aug 17 18:26:29 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 37CE252AE9 for ; Wed, 17 Aug 2005 17:34:01 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 44765-02 for ; Wed, 17 Aug 2005 20:33:56 +0000 (GMT) Received: from xproxy.gmail.com (xproxy.gmail.com [66.249.82.198]) by svr1.postgresql.org (Postfix) with ESMTP id 4FA5052ABD for ; Wed, 17 Aug 2005 17:33:50 -0300 (ADT) Received: by xproxy.gmail.com with SMTP id i31so37428wxd for ; Wed, 17 Aug 2005 13:33:52 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=uFoByAS4pzqfRhu7po/hVPrlklTJx80+Sj8gBZlk5wJlaSHBn5md7AhSyD4SyyvGEKlt5aBn7WkIhsr6Qjnhv1AXRjEZVar707ptdgMEVHdpa+vF6fvwG3Kk2D+8L/kOnfjhRu/tkJQeJwFW6sw8DO1OTX+DszqmFp+KlGeksf0= Received: by 10.70.73.8 with SMTP id v8mr5591wxa; Wed, 17 Aug 2005 13:33:52 -0700 (PDT) Received: by 10.70.128.19 with HTTP; Wed, 17 Aug 2005 13:33:52 -0700 (PDT) Message-ID: Date: Wed, 17 Aug 2005 15:33:52 -0500 From: Matthew Nuzum Reply-To: newz@bearfruit.org To: pgsql-performance@postgresql.org Subject: Re: Need for speed Cc: Ulrich Wisser In-Reply-To: <6.2.3.4.0.20050817140116.05d317b0@pop.earthlink.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <6.2.3.4.0.20050817140116.05d317b0@pop.earthlink.net> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/214 X-Sequence-Number: 13961 On 8/17/05, Ron wrote: > At 05:15 AM 8/17/2005, Ulrich Wisser wrote: > >Hello, > > > >thanks for all your suggestions. > > > >I can see that the Linux system is 90% waiting for disc io. ... > 1=3D your primary usage is OLTP-like, but you are also expecting to do > reports against the same schema that is supporting your OLTP-like > usage. Bad Idea. Schemas that are optimized for reporting and other > data mining like operation are pessimal for OLTP-like applications > and vice versa. You need two schemas: one optimized for lots of > inserts and deletes (OLTP-like), and one optimized for reporting > (data-mining like). Ulrich, If you meant that your disc/scsi system is already the fastest available *with your current budget* then following Ron's advise I quoted above will be a good step. I have some systems very similar to yours. What I do is import in batches and then immediately pre-process the batch data into tables optimized for quick queries. For example, if your reports frequenly need to find the total number of views per hour for each customer, create a table whose data contains just the totals for each customer for each hour of the day. This will make it a tiny fraction of the size, allowing it to fit largely in RAM for the query and making the indexes more efficient. This is a tricky job, but if you do it right, your company will be a big success and buy you more hardware to work with. Of course, they'll also ask you to create dozens of new reports, but that's par for the course. Even if you have the budget for more hardware, I feel that creating an effective db structure is a much more elegant solution than to throw more hardware. (I admit, sometimes its cheaper to throw more hardware) If you have particular queries that are too slow, posting the explain analyze for each on the list should garner some help. --=20 Matthew Nuzum www.bearfruit.org From pgsql-performance-owner@postgresql.org Wed Aug 17 18:25:33 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E26695288A for ; Wed, 17 Aug 2005 17:36:45 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 39818-07 for ; Wed, 17 Aug 2005 20:36:38 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id 9E29752ACD for ; Wed, 17 Aug 2005 17:36:38 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [64.81.245.111] (account josh@agliodbs.com HELO temoku.sf.agliodbs.com) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7743019; Wed, 17 Aug 2005 13:38:55 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: PG8 Tuning Date: Wed, 17 Aug 2005 13:39:47 -0700 User-Agent: KMail/1.8 Cc: Michael Stone References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com> <200508160912.32263.josh@agliodbs.com> <20050816173047.GC19080@mathom.us> In-Reply-To: <20050816173047.GC19080@mathom.us> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200508171339.47399.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.049 required=5 tests=[AWL=-0.001, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/213 X-Sequence-Number: 13960 Michael, > Well, you don't have to spend *quite* that much to get a decent storage > array. :) Yes, I'm just pointing out that it's only the extreme cases which are clear-cut. Middle cases are a lot harder to define. For example, we've found that on DBT2 running of a 14-drive JBOD, seperating off WAL boosts performance about 8% to 14%. On DBT3 (DSS) seperate WAL (and seperate tablespaces) helps considerably during data load,but not otherwise. So it all depends. > That's a different creature from > a data mining app that might really benefit from having additional > spindles to accelerate read performance from indices much larger than > RAM. Yes, although the data mining app benefits from the special xlog disk during ETL. So it's a tradeoff. > At any rate, this just underscores the need for testing a > particular workload on particular hardware Yes, absolutely. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Wed Aug 17 18:09:23 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 135A0529E6; Wed, 17 Aug 2005 17:40:39 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 44884-08; Wed, 17 Aug 2005 20:40:37 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id CF9E5529C2; Wed, 17 Aug 2005 17:40:34 -0300 (ADT) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: bitmap scan issues 8.1 devel Date: Wed, 17 Aug 2005 16:40:34 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3417DD11C@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: bitmap scan issues 8.1 devel Thread-Index: AcWja+mdyQz95+VbTreJShGw9lPYFg== From: "Merlin Moncure" To: Cc: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/212 X-Sequence-Number: 13959 Hello, Doing some testing on upcoming 8.1 devel and am having serious issues with new bitmap index scan feature. It is easy to work around (just disable it) but IMO the planner is using it when a regular index scan should be strongly favored. The performance of the bitmapscan in my usage is actually quite a bit worse than a full sequential scan. here is a query which does this: explain analyze execute data1_read_next_product_structure_file_0('012241', '', '', '002', 1); Here is the 8.0/bitmap off plan: Limit (cost=3D0.00..45805.23 rows=3D5722 width=3D288) (actual time=3D0.070..0.072 rows=3D1 loops=3D1) -> Index Scan using product_structure_file_pkey on product_structure_file (cost=3D0.00..45805.23 rows=3D5722 width=3D288) (actual time=3D0.063..0.063 row s=3D1 loops=3D1) Index Cond: ((ps_parent_code)::text >=3D ($1)::text) Filter: ((((ps_parent_code)::text > ($1)::text) OR (ps_group_code >=3D $2)) AND (((ps_parent_code)::text > ($1)::text) OR (ps_group_code > $2) OR ((ps_section_code)::text >=3D ($3)::text)) AND = (((ps_parent_code)::text > ($1)::text) OR (ps_group_code > $2) OR ((ps_section_code)::text > ($3)::tex t) OR ((ps_seq_no)::smallint > $4))) Total runtime: 0.185 ms Here is the 8.1 with bitamp on: Limit (cost=3D3768.32..3782.63 rows=3D5722 width=3D288) (actual time=3D2287.488..2287.490 rows=3D1 loops=3D1) -> Sort (cost=3D3768.32..3782.63 rows=3D5722 width=3D288) (actual time=3D2287.480..2287.480 rows=3D1 loops=3D1) Sort Key: ps_parent_code, ps_group_code, ps_section_code, ps_seq_no -> Bitmap Heap Scan on product_structure_file (cost=3D187.84..3411.20 rows=3D5722 width=3D288) (actual = time=3D19.977..514.532 rows=3D47355 loops=3D1) Recheck Cond: ((ps_parent_code)::text >=3D ($1)::text) Filter: ((((ps_parent_code)::text > ($1)::text) OR (ps_group_code >=3D $2)) AND (((ps_parent_code)::text > ($1)::text) OR (ps_group_code > $2) OR ((ps_section_code)::text >=3D ($3)::text)) AND (((ps_parent_code)::text > ($1)::text) OR (ps_group_code > $2) OR ((ps_section_code)::text > ($3 )::text) OR ((ps_seq_no)::smallint > $4))) -> Bitmap Index Scan on product_structure_file_pkey (cost=3D0.00..187.84 rows=3D18239 width=3D0) (actual = time=3D19.059..19.059 rows=3D47356 loo ps=3D1) Index Cond: ((ps_parent_code)::text >=3D = ($1)::text) Total runtime: 2664.034 ms Here is the prepared statement definition: prepare data1_read_next_product_structure_file_0 (character varying, character, character varying, int4, int4) as select 1::int4, * from data1.product_structure_file where ps_parent_code >=3D $1 and=20 (ps_parent_code > $1 or ps_group_code >=3D $2) and=20 (ps_parent_code > $1 or ps_group_code > $2 or ps_section_code >=3D $3) and=20 (ps_parent_code > $1 or ps_group_code > $2 or ps_section_code > $3 or ps_seq_no > $4)=20 order by ps_parent_code, ps_group_code, ps_section_code, ps_seq_no limit $5 Aside: this is the long way of writing select 1::int4, * from data1.product_structure_file where (ps_parent_code, ps_group_code, ps_section_code, ps_seq_no) > ($1, $2, $3, $4) limit %5 which is allowed in pg but returns the wrong answer. Merlin From pgsql-performance-owner@postgresql.org Wed Aug 17 20:09:30 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6553C52CBF; Wed, 17 Aug 2005 18:30:13 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 16068-02; Wed, 17 Aug 2005 21:30:08 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id AF9C452E46; Wed, 17 Aug 2005 18:30:06 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [64.81.245.111] (account josh@agliodbs.com HELO temoku.sf.agliodbs.com) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7743300; Wed, 17 Aug 2005 14:32:24 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-hackers@postgresql.org Subject: Re: [HACKERS] bitmap scan issues 8.1 devel Date: Wed, 17 Aug 2005 14:33:15 -0700 User-Agent: KMail/1.8 Cc: "Merlin Moncure" , pgsql-performance@postgresql.org References: <6EE64EF3AB31D5448D0007DD34EEB3417DD11C@Herge.rcsinc.local> In-Reply-To: <6EE64EF3AB31D5448D0007DD34EEB3417DD11C@Herge.rcsinc.local> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Message-Id: <200508171433.15614.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.049 required=5 tests=[AWL=-0.001, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/219 X-Sequence-Number: 13966 Merlin, > =A0 =A0-> =A0Index Scan using product_structure_file_pkey on > product_structure_file =A0(cost=3D0.00..45805.23 rows=3D5722 width=3D288) > (actual time=3D0.063..0.063 row > s=3D1 loops=3D1) It appears that your DB is estimating the number of rows returned much too= =20 high (5722 instead of 1). Please raise the statistics on all columns to=20 about 500, analyze, and try your test again. Thanks! =2D-=20 =2D-Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Wed Aug 17 19:46:39 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AED4A52AAE; Wed, 17 Aug 2005 18:54:39 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 77510-04; Wed, 17 Aug 2005 21:54:34 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 455E052A7E; Wed, 17 Aug 2005 18:54:33 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7HLsa0t017375; Wed, 17 Aug 2005 17:54:36 -0400 (EDT) To: "Merlin Moncure" Cc: pgsql-performance@postgresql.org, pgsql-hackers@postgresql.org Subject: Re: [HACKERS] bitmap scan issues 8.1 devel In-reply-to: <6EE64EF3AB31D5448D0007DD34EEB3417DD11C@Herge.rcsinc.local> References: <6EE64EF3AB31D5448D0007DD34EEB3417DD11C@Herge.rcsinc.local> Comments: In-reply-to "Merlin Moncure" message dated "Wed, 17 Aug 2005 16:40:34 -0400" Date: Wed, 17 Aug 2005 17:54:36 -0400 Message-ID: <17374.1124315676@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/218 X-Sequence-Number: 13965 "Merlin Moncure" writes: > Doing some testing on upcoming 8.1 devel and am having serious issues > with new bitmap index scan feature. It is easy to work around (just > disable it) but IMO the planner is using it when a regular index scan > should be strongly favored. I think blaming the bitmap code is the wrong response. What I see in your example is that the planner doesn't know what the LIMIT value is, and accordingly is favoring a plan that isn't going to get blown out of the water if the LIMIT is large. I'd suggest not parameterizing the LIMIT. (But hmm ... I wonder if we could use estimate_expression_value for LIMIT items, instead of handling only simple Consts as the code does now?) regards, tom lane From pgsql-performance-owner@postgresql.org Wed Aug 17 19:35:17 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 169F35285B for ; Wed, 17 Aug 2005 19:02:42 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 84472-09 for ; Wed, 17 Aug 2005 22:02:38 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id 04BC452B37 for ; Wed, 17 Aug 2005 19:02:34 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050817220235m9100g6hbue>; Wed, 17 Aug 2005 22:02:35 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 2133B55FBA; Wed, 17 Aug 2005 17:02:35 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id 55F1C55FAD; Wed, 17 Aug 2005 17:02:28 -0500 (CDT) Message-ID: <4303B3F4.9010306@arbash-meinel.com> Date: Wed, 17 Aug 2005 17:02:28 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Manfred Koizar , Postgresql Performance Subject: Re: Insert performance (OT?) References: <42DCCA3B.8070100@archonet.com> <42DCDB47.3060009@archonet.com> In-Reply-To: X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig314C86187A1349552BC0B1FC" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.043 required=5 tests=[AWL=-0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/217 X-Sequence-Number: 13964 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig314C86187A1349552BC0B1FC Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Manfred Koizar wrote: > On Tue, 19 Jul 2005 11:51:51 +0100, Richard Huxton > wrote: > >>You could get away with one query if you converted them to left-joins: >>INSERT INTO ... >>SELECT * FROM upload LEFT JOIN ... WHERE f3 IS NULL >>UNION >>SELECT * FROM upload LEFT JOIN ... WHERE f4 IS NULL > > > For the archives: This won't work. Each of the two SELECTs > eliminates rows violating one of the two constraints but includes rows > violating the other constraint. After the UNION you are back to > violating both constraints :-( Couldn't you use "INTERSECT" then? To only get the rows that *both* queries return? John =:-> > > Servus > Manfred > > > ---------------------------(end of broadcast)--------------------------- > TIP 5: don't forget to increase your free space map settings > --------------enig314C86187A1349552BC0B1FC Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDA7P0JdeBCYSNAAMRAjhsAJ9pCTOStHkdEkOEV4uwj2RSuZm4+QCffk+7 Y6mRPRvjR4RKPGu4R8sg/GY= =F/S/ -----END PGP SIGNATURE----- --------------enig314C86187A1349552BC0B1FC-- From pgsql-performance-owner@postgresql.org Wed Aug 17 23:02:22 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3B3DE52848 for ; Wed, 17 Aug 2005 23:02:19 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 79120-04 for ; Thu, 18 Aug 2005 02:02:17 +0000 (GMT) Received: from unicorn.rentec.com (unicorn.rentec.com [216.223.240.9]) by svr1.postgresql.org (Postfix) with ESMTP id E17F552802 for ; Wed, 17 Aug 2005 23:02:15 -0300 (ADT) Received: from ram.rentec.com (mailhost [192.5.35.66]) by unicorn.rentec.com (8.13.1/8.12.1) with ESMTP id j7I22HTs010784 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO) for ; Wed, 17 Aug 2005 22:02:19 -0400 (EDT) X-Rentec: external Received: from [172.16.160.108] (stange-dhcp2.rentec.com [172.16.160.108]) by ram.rentec.com (8.13.1/8.12.1) with ESMTP id j7I22Ge2022033 for ; Wed, 17 Aug 2005 22:02:16 -0400 (EDT) Message-ID: <4303E704.4030200@rentec.com> Date: Wed, 17 Aug 2005 21:40:20 -0400 From: Alan Stange Organization: Renaissance Technologies Corp. User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: limit number of concurrent callers to a stored proc? Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Logged: Logged by unicorn.rentec.com as j7I22HTs010784 at Wed Aug 17 22:02:19 2005 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/220 X-Sequence-Number: 13967 Hello all, is there a simple way to limit the number of concurrent callers to a stored proc? The problem we have is about 50 clients come and perform the same operation at nearly the same time. Typically, this query takes a few seconds to run, but in the case of this thundering herd the query time drops to 70 seconds or much more. The query can return up to 15MB of data. The machine is a dual opteron, 8 GB memory, lots of fiber channel disk, Linux 2.6, etc. So, I'm thinking that a semaphore than will block more than N clients from being in the core of the function at one time would be a good thing. Thanks! -- Alan From pgsql-performance-owner@postgresql.org Wed Aug 17 23:11:51 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id DE68352848 for ; Wed, 17 Aug 2005 23:11:48 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 00760-09 for ; Thu, 18 Aug 2005 02:11:46 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id D50165282E for ; Wed, 17 Aug 2005 23:11:45 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7I2BnVi004050 for ; Wed, 17 Aug 2005 21:11:49 -0500 Subject: extremly low memory usage From: Jeremiah Jahn To: postgres performance Content-Type: text/plain Date: Wed, 17 Aug 2005 21:11:48 -0500 Message-Id: <1124331108.27881.96.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Content-Transfer-Encoding: 7bit Received-SPF: none (penguin.goodinassociates.com: 206.80.71.242 is neither permitted nor denied by domain of cs.earlham.edu) client-ip=206.80.71.242; envelope-from=jeremiah@cs.earlham.edu; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1027/Tue Aug 16 18:44:00 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.03 required=5 tests=[AWL=-0.020, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/221 X-Sequence-Number: 13968 I just put together a system with 6GB of ram on a 14 disk raid 10 array. When I run my usual big painful queries, I get very little to know memory usage. My production box (raid 5 4GB ram) hovers at 3.9GB used most of the time. the new devel box sits at around 250MB. I've switched to an 8.0 system on the new devel box, but the .conf really didn't change. Index usage is the same. Something seems wrong and I'm not sure why. any thoughts, -jj- shared_buffers = 32768 # min 16, at least max_connections*2, 8KB each work_mem = 2097151 # min 64, size in KB maintenance_work_mem = 819200 # min 1024, size in KB max_fsm_pages = 80000 # min max_fsm_relations*16, 6 bytes each checkpoint_segments = 30 # in logfile segments, min 1, 16MB each effective_cache_size = 3600000 <-----this is a little out of control, but would it have any real effect? random_page_cost = 2 # units are one sequential page fetch cost log_min_duration_statement = 10000 # -1 is disabled, in milliseconds. lc_messages = 'C' # locale for system error message strings lc_monetary = 'C' # locale for monetary formatting lc_numeric = 'C' # locale for number formatting lc_time = 'C' # locale for time formatting -- "Now this is a totally brain damaged algorithm. Gag me with a smurfette." -- P. Buhr, Computer Science 354 From pgsql-performance-owner@postgresql.org Wed Aug 17 23:23:33 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 05A8E52A14 for ; Wed, 17 Aug 2005 23:23:30 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 06571-10 for ; Thu, 18 Aug 2005 02:23:25 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id 6CFDD52A06 for ; Wed, 17 Aug 2005 23:23:24 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050818022327m92009id64e>; Thu, 18 Aug 2005 02:23:27 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 9A9E055FAD; Wed, 17 Aug 2005 21:23:26 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id C92CB55FAD; Wed, 17 Aug 2005 21:21:09 -0500 (CDT) Message-ID: <4303F090.40602@arbash-meinel.com> Date: Wed, 17 Aug 2005 21:21:04 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Jeremiah Jahn Cc: postgres performance Subject: Re: extremly low memory usage References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> In-Reply-To: <1124331108.27881.96.camel@bluejay.goodinassociates.com> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig6476E14D6CEC3DB757D1CFC5" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.043 required=5 tests=[AWL=-0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/222 X-Sequence-Number: 13969 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig6476E14D6CEC3DB757D1CFC5 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Jeremiah Jahn wrote: > I just put together a system with 6GB of ram on a 14 disk raid 10 array. > When I run my usual big painful queries, I get very little to know > memory usage. My production box (raid 5 4GB ram) hovers at 3.9GB used > most of the time. the new devel box sits at around 250MB. > > I've switched to an 8.0 system on the new devel box, but the .conf > really didn't change. Index usage is the same. Something seems wrong and > I'm not sure why. > How big is your actual database on disk? And how much of it is actually touched by your queries? It seems that your tough queries might only be exercising a portion of the database. If you really want to make memory usage increase try something like: find . -type f -print0 | xargs -0 cat >/dev/null Which should read all the files. After doing that, does the memory usage increase? > > any thoughts, > -jj- > > > shared_buffers = 32768 # min 16, at least max_connections*2, 8KB each > work_mem = 2097151 # min 64, size in KB This seems awfully high. 2GB Per sort? This might actually be flushing some of your ram, since it would get allocated and filled, and then freed when finished. Remember, depending on what you are doing, this amount can get allocated more than once per query. > maintenance_work_mem = 819200 # min 1024, size in KB > max_fsm_pages = 80000 # min max_fsm_relations*16, 6 bytes each > checkpoint_segments = 30 # in logfile segments, min 1, 16MB each > effective_cache_size = 3600000 <-----this is a little out of control, but would it have any real effect? It should just tell the planner that it is more likely to have buffers in cache, so index scans are slightly cheaper than they would otherwise be. > random_page_cost = 2 # units are one sequential page fetch cost > log_min_duration_statement = 10000 # -1 is disabled, in milliseconds. > lc_messages = 'C' # locale for system error message strings > lc_monetary = 'C' # locale for monetary formatting > lc_numeric = 'C' # locale for number formatting > lc_time = 'C' # locale for time formatting > John =:-> --------------enig6476E14D6CEC3DB757D1CFC5 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDA/CTJdeBCYSNAAMRAkkFAJ4mPFWO1cTQXZzSN8a3PYd8PbilvACgoRex HWfH4vnvrPDt9COuwFW/y+8= =0N+Z -----END PGP SIGNATURE----- --------------enig6476E14D6CEC3DB757D1CFC5-- From pgsql-performance-owner@postgresql.org Wed Aug 17 23:30:34 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1BAC85294D for ; Wed, 17 Aug 2005 23:30:27 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 03288-01 for ; Thu, 18 Aug 2005 02:30:21 +0000 (GMT) Received: from linuxworld.com.au (unknown [203.34.46.50]) by svr1.postgresql.org (Postfix) with ESMTP id 33B4C5282E for ; Wed, 17 Aug 2005 23:30:19 -0300 (ADT) Received: from linuxworld.com.au (IDENT:swm@localhost.localdomain [127.0.0.1]) by linuxworld.com.au (8.13.2/8.13.2) with ESMTP id j7I2UORY022238; Thu, 18 Aug 2005 12:30:24 +1000 Received: from localhost (swm@localhost) by linuxworld.com.au (8.13.2/8.13.2/Submit) with ESMTP id j7I2UOqn022235; Thu, 18 Aug 2005 12:30:24 +1000 X-Authentication-Warning: linuxworld.com.au: swm owned process doing -bs Date: Thu, 18 Aug 2005 12:30:24 +1000 (EST) From: Gavin Sherry X-X-Sender: swm@linuxworld.com.au To: Alan Stange Cc: pgsql-performance@postgresql.org Subject: Re: limit number of concurrent callers to a stored proc? In-Reply-To: <4303E704.4030200@rentec.com> Message-ID: References: <4303E704.4030200@rentec.com> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.002 required=5 tests=[AWL=0.002] X-Spam-Level: X-Archive-Number: 200508/223 X-Sequence-Number: 13970 Hi Alan, On Wed, 17 Aug 2005, Alan Stange wrote: > Hello all, > > is there a simple way to limit the number of concurrent callers to a > stored proc? > > The problem we have is about 50 clients come and perform the same > operation at nearly the same time. Typically, this query takes a few > seconds to run, but in the case of this thundering herd the query time > drops to 70 seconds or much more. The query can return up to 15MB of data. > > The machine is a dual opteron, 8 GB memory, lots of fiber channel disk, > Linux 2.6, etc. > > So, I'm thinking that a semaphore than will block more than N clients > from being in the core of the function at one time would be a good thing. There is no PostgreSQL feature which will do this for you. It should be possible to implement this yourself, without too much pain. If you're using PL/PgSQL, write another function in C or one of the other more sophisticated PLs to implement the logic for you. At the beginning of the function, execute the function to increment the count; at the end, execute a function to decrement it. If you're writing the function in C or one of those more sophisticated PLs, it's even easier. As an aside, using semaphores might be a little painful. I'd just grab some shared memory and keep a counter in it. If the counter is greater than your desired number of concurrent executions, you sleep and try again soon. That being said, did you want to give us a look at your function and data and see if we can improve the performance at all? Thanks, Gavin From pgsql-performance-owner@postgresql.org Wed Aug 17 23:32:19 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 72CF45294D for ; Wed, 17 Aug 2005 23:32:07 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91406-02 for ; Thu, 18 Aug 2005 02:32:04 +0000 (GMT) Received: from houston.familyhealth.com.au (houston.au.fhnetwork.com [203.22.197.21]) by svr1.postgresql.org (Postfix) with ESMTP id A4C4F52A14 for ; Wed, 17 Aug 2005 23:32:02 -0300 (ADT) Received: from houston.familyhealth.com.au (localhost [127.0.0.1]) by houston.familyhealth.com.au (Postfix) with ESMTP id 8CCBE24FF7; Thu, 18 Aug 2005 10:32:04 +0800 (WST) Received: from [127.0.0.1] (work-40.internal [192.168.0.40]) by houston.familyhealth.com.au (Postfix) with ESMTP id A44A724FF1; Thu, 18 Aug 2005 10:32:02 +0800 (WST) Message-ID: <4303F417.3090307@familyhealth.com.au> Date: Thu, 18 Aug 2005 10:36:07 +0800 From: Christopher Kings-Lynne User-Agent: Mozilla Thunderbird 1.0.6 (Windows/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Alan Stange Cc: pgsql-performance@postgresql.org Subject: Re: limit number of concurrent callers to a stored proc? References: <4303E704.4030200@rentec.com> In-Reply-To: <4303E704.4030200@rentec.com> Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit X-familyhealth-MailScanner-Information: Please contact the ISP for more information X-familyhealth-MailScanner: Found to be clean X-familyhealth-MailScanner-From: chriskl@familyhealth.com.au X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.062 required=5 tests=[AWL=0.012, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/224 X-Sequence-Number: 13971 You could use a 1 column/1 row table perhaps. Use some sort of locking mechanism. Also, check out contrib/userlock Chris Alan Stange wrote: > Hello all, > > is there a simple way to limit the number of concurrent callers to a > stored proc? > > The problem we have is about 50 clients come and perform the same > operation at nearly the same time. Typically, this query takes a few > seconds to run, but in the case of this thundering herd the query time > drops to 70 seconds or much more. The query can return up to 15MB of data. > > The machine is a dual opteron, 8 GB memory, lots of fiber channel disk, > Linux 2.6, etc. > > So, I'm thinking that a semaphore than will block more than N clients > from being in the core of the function at one time would be a good thing. > Thanks! > > -- Alan > > ---------------------------(end of broadcast)--------------------------- > TIP 3: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faq From pgsql-performance-owner@postgresql.org Thu Aug 18 00:21:54 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 82E3B528DA for ; Thu, 18 Aug 2005 00:19:15 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 47452-05 for ; Thu, 18 Aug 2005 03:19:12 +0000 (GMT) Received: from smtpauth07.mail.atl.earthlink.net (smtpauth07.mail.atl.earthlink.net [209.86.89.67]) by svr1.postgresql.org (Postfix) with ESMTP id 5E86E528D4 for ; Thu, 18 Aug 2005 00:19:11 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=eYiuLLIWlc4fnricpidY+eXfUNn96MPKcPPUCYrCRm0sm+Q78nemMqJeMw0UQpuB; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth07.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E5avu-0006tW-Ew; Wed, 17 Aug 2005 23:19:10 -0400 Message-Id: <6.2.3.4.0.20050817224934.05c09228@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Wed, 17 Aug 2005 23:19:04 -0400 To: Alan Stange , pgsql-performance@postgresql.org From: Ron Subject: Re: limit number of concurrent callers to a stored In-Reply-To: <4303E704.4030200@rentec.com> References: <4303E704.4030200@rentec.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bcdba9d9f157450d079aba40b5a5011883350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.083 required=5 tests=[AWL=0.709, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: * X-Archive-Number: 200508/225 X-Sequence-Number: 13972 At 09:40 PM 8/17/2005, Alan Stange wrote: >is there a simple way to limit the number of concurrent callers to a >stored proc? > >The problem we have is about 50 clients come and perform the same >operation at nearly the same time. Typically, this query takes a >few seconds to run, but in the case of this thundering herd the >query time drops to 70 seconds or much more. The query can return >up to 15MB of data. I'm assuming there is some significant write activity going on at some point as a result of the query, since MVCC should not care about concurrent read activity? Is that "a few seconds each query" or "a few seconds total if we run 50 queries sequentially but 70+ seconds per query if we try to run 50 queries concurrently"? A) If the former, "a few seconds" * 50 can easily be 70+ seconds, and things are what you should expect. Getting higher performance in that situation means reducing per query times, which may or may not be easy. Looking at the stored procedure code with an eye towards optimization would be a good place to start. B) If the later, then table access contention is driving performance into the ground, and there are a few things you can try: 1= lock the table(s) under these circumstances so only one query of the 50 can be acting on it at a time. If the table(s) is/are small enough to be made RAM resident, this may be a particularly low-cost, low-effort, reasonable solution. 2= put a queue into place and only let some small number n of queries run against the table(s) concurrently. Adjust n until you get best performance. There are a few ways this could be done. 3= Buy a SSD and put the table(s) in question on it. IIRC, 3.5" format SSDs that can "drop in" replace HDs are available in up to 147GB capacities. >The machine is a dual opteron, 8 GB memory, lots of fiber channel >disk, Linux 2.6, etc. > >So, I'm thinking that a semaphore than will block more than N >clients from being in the core of the function at one time would be >a good thing. This will only help in case "B" above. If you go the "hard" route of using systems programming, you will have a lot of details that must be paid attention to correctly or Bad Things (tm) will happen. Putting the semaphore in place is the tip of the iceberg. Hope this helps, Ron Peacetree From pgsql-performance-owner@postgresql.org Thu Aug 18 05:59:22 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4335D52AF7 for ; Thu, 18 Aug 2005 05:59:20 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 46818-05 for ; Thu, 18 Aug 2005 08:59:16 +0000 (GMT) Received: from floppy.pyrenet.fr (floppy.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id 4107352ABF for ; Thu, 18 Aug 2005 05:59:15 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id A75D930B42; Thu, 18 Aug 2005 11:10:30 +0200 (MET DST) From: "Qingqing Zhou" X-Newsgroups: pgsql.performance Subject: Re: Performance pb vs SQLServer. Date: Thu, 18 Aug 2005 16:56:41 +0800 Organization: Hub.Org Networking Services Lines: 16 Message-ID: References: <6BCB9D8A16AC4241919521715F4D8BCE6C789C@algol.sollentuna.se> <20050815161859.GB22386@alvh.no-ip.org> Reply-To: "Qingqing Zhou" X-Complaints-To: usenet@news.hub.org X-Priority: 3 X-MSMail-Priority: Normal X-Newsreader: Microsoft Outlook Express 6.00.2800.1506 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1506 To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.185 required=5 tests=[AWL=-0.000, PRIORITY_NO_NAME=1.185] X-Spam-Level: * X-Archive-Number: 200508/226 X-Sequence-Number: 13973 "Alvaro Herrera" writes > > Interesting; do they use an overwriting storage manager like Oracle, or > a non-overwriting one like Postgres? > They call this MVCC "RLV(row level versioning)". I think they use rollback segment like Oracle (a.k.a "version store" or tempdb in SQL Server). Some details are explained in their white paper:"Database concurrency and row level versioning in SQL Server 2005". Regards, Qingqing From pgsql-performance-owner@postgresql.org Thu Aug 18 10:00:44 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C542B529C6 for ; Thu, 18 Aug 2005 10:00:41 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 81956-07 for ; Thu, 18 Aug 2005 13:00:38 +0000 (GMT) Received: from sccrmhc13.comcast.net (sccrmhc13.comcast.net [63.240.76.28]) by svr1.postgresql.org (Postfix) with ESMTP id 66755528DC for ; Thu, 18 Aug 2005 10:00:36 -0300 (ADT) Received: from jefftrout.com ([24.147.120.205]) by comcast.net (sccrmhc13) with SMTP id <2005081813003201300l7cmfe>; Thu, 18 Aug 2005 13:00:33 +0000 Received: (qmail 74410 invoked from network); 18 Aug 2005 13:01:22 -0000 Received: from c-24-147-120-205.hsd1.ma.comcast.net (HELO ?192.168.0.106?) (24.147.120.205) by 192.168.0.109 with SMTP; 18 Aug 2005 13:01:22 -0000 In-Reply-To: <1124331108.27881.96.camel@bluejay.goodinassociates.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 (Apple Message framework v733) Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: Cc: postgres performance Content-Transfer-Encoding: 7bit From: Jeff Trout Subject: Re: extremly low memory usage Date: Thu, 18 Aug 2005 09:00:31 -0400 To: Jeremiah Jahn X-Mailer: Apple Mail (2.733) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/227 X-Sequence-Number: 13974 On Aug 17, 2005, at 10:11 PM, Jeremiah Jahn wrote: > I just put together a system with 6GB of ram on a 14 disk raid 10 > array. > When I run my usual big painful queries, I get very little to know > memory usage. My production box (raid 5 4GB ram) hovers at 3.9GB used > most of the time. the new devel box sits at around 250MB. > Is the system performing fine? Are you touching as much data as the production box? If the system is performing fine don't worry about it. > work_mem = 2097151 # min 64, size in KB This is EXTREMELY high. You realize this is the amount of memory that can be used per-sort and per-hash build in a query? You can end up with multiples of this on a single query. If you have some big queries that are run infrequently have them set it manually. > effective_cache_size = 3600000 <-----this is a little out of > control, but would it have any real effect? This doesn't allocate anything - it is a hint to the planner about how much data it can assume is cached. -- Jeff Trout http://www.jefftrout.com/ http://www.stuarthamm.net/ From pgsql-performance-owner@postgresql.org Thu Aug 18 11:47:02 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3677B52A98 for ; Thu, 18 Aug 2005 11:46:56 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 19925-08 for ; Thu, 18 Aug 2005 14:46:53 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id 43ABC52A6E for ; Thu, 18 Aug 2005 11:46:50 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050818144653m9100g61s9e>; Thu, 18 Aug 2005 14:46:53 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id C47D655FAD; Thu, 18 Aug 2005 09:46:52 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id 3FC5855FAD; Thu, 18 Aug 2005 09:44:39 -0500 (CDT) Message-ID: <43049ED2.1080606@arbash-meinel.com> Date: Thu, 18 Aug 2005 09:44:34 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Qingqing Zhou Cc: pgsql-performance@postgresql.org Subject: Re: Performance pb vs SQLServer. References: <6BCB9D8A16AC4241919521715F4D8BCE6C789C@algol.sollentuna.se> <20050815161859.GB22386@alvh.no-ip.org> In-Reply-To: X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigC08C6AF0E589E59A3E6ACFBA" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.043 required=5 tests=[AWL=-0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/228 X-Sequence-Number: 13975 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigC08C6AF0E589E59A3E6ACFBA Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Qingqing Zhou wrote: > "Alvaro Herrera" writes > >>Interesting; do they use an overwriting storage manager like Oracle, or >>a non-overwriting one like Postgres? >> > > > They call this MVCC "RLV(row level versioning)". I think they use rollback > segment like Oracle (a.k.a "version store" or tempdb in SQL Server). Some > details are explained in their white paper:"Database concurrency and row > level versioning in SQL Server 2005". > > Regards, > Qingqing > I found the paper here: http://www.microsoft.com/technet/prodtechnol/sql/2005/cncrrncy.mspx And it does sound like they are doing it the Oracle way: When a record in a table or index is updated, the new record is stamped with the transaction sequence_number of the transaction that is doing the update. The previous version of the record is stored in the version store, and the new record contains a pointer to the old record in the version store. Old records in the version store may contain pointers to even older versions. All the old versions of a particular record are chained in a linked list, and SQL Server may need to follow several pointers in a list to reach the right version. Version records need to be kept in the version store only as long as there are there are operations that might require them. John =:-> --------------enigC08C6AF0E589E59A3E6ACFBA Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDBJ7WJdeBCYSNAAMRAj+pAJ9sYAV896c6AjMlDUrVyhJldhNaFwCfUnMI UZAct14ESLSKVMjdK9fGTQc= =k73F -----END PGP SIGNATURE----- --------------enigC08C6AF0E589E59A3E6ACFBA-- From pgsql-performance-owner@postgresql.org Thu Aug 18 12:01:27 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1F95D529CC for ; Thu, 18 Aug 2005 12:01:26 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 03401-02 for ; Thu, 18 Aug 2005 15:01:20 +0000 (GMT) Received: from urano.datatransfer.com.ar (host202.200.80.31.ifxnw.com.ar [200.80.31.202]) by svr1.postgresql.org (Postfix) with ESMTP id D259B52984 for ; Thu, 18 Aug 2005 12:01:16 -0300 (ADT) Content-Class: urn:content-classes:message MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----_=_NextPart_001_01C5A405.B3701AEC" Subject: FW: Tx forecast improving harware capabilities. X-MimeOLE: Produced By Microsoft Exchange V6.0.6556.0 Date: Thu, 18 Aug 2005 12:01:24 -0300 Message-ID: <722B1B9CFE0ECE488342E3EC2BA920B86CC443@mailserver> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: Tx forecast improving harware capabilities. thread-index: AcWjb2iCgsLuHIzSSjO5/yF8LaVzUQAiVH5AAAMSZgA= From: "Sebastian Lallana" To: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.051 required=5 tests=[FORGED_RCVD_HELO=0.05, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/229 X-Sequence-Number: 13976 This is a multi-part message in MIME format. ------_=_NextPart_001_01C5A405.B3701AEC Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Hello: We are having serious performance problems using JBOSS and PGSQL. I'm sure the problem has to do with the application itself (and neither with JBOSS nor PGSQL) but the fact is that we are using desktop equipment to run both Jboss and Postgresql (An Athlon 2600, 1 Gb Ram, IDE HDD with 60 Mb/sec Transfer Rate), and the answers arise: If we upgrade our hardware to a Dual Processor would the transactions per second increase significantly? Would Postgresql take advantage from SMP? Presumably yes, but can we do a forecast about the number of tps? What we need is a paper with some figures showing the expected performance in different environments. Some study about the "degree of correlation" between TPS and Number of Processors, Cache, Frequency, Word Size, Architecture, etc.=20 It exists something like this? Does anybody has experience about this subject? =20 Thanks in advance and best regards. =20 P.S. I've been looking at www.tpc.org but I could't find anything valuable. =20 =20 ------_=_NextPart_001_01C5A405.B3701AEC Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable

Hello:

We are having serious performance problems = using JBOSS and PGSQL.

I’m sure the problem has to do with the application itself (and neither with JBOSS nor PGSQL) but the fact is that we are using = desktop equipment to run both Jboss and Postgresql (An = Athlon 2600, 1 Gb Ram, IDE HDD with 60 Mb/sec Transfer Rate), and the answers = arise:

If we upgrade our hardware to a Dual Processor = would the transactions per second increase significantly? Would Postgresql take advantage from SMP? = Presumably yes, but can we do a forecast about the number of tps? What we need is a = paper with some figures showing the expected performance in different environments. = Some study about the “degree of correlation” between TPS and = Number of Processors, Cache, Frequency, Word Size, Architecture, etc. =

It exists = something like this? Does anybody has experience about this subject?

 =

Thanks in advance and best regards.

 =

P.S. I’ve = been looking at www.tpc.org but I could’t find anything valuable.  =

 =

------_=_NextPart_001_01C5A405.B3701AEC-- From pgsql-performance-owner@postgresql.org Thu Aug 18 13:29:50 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 66498529ED for ; Thu, 18 Aug 2005 13:29:49 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 70624-07 for ; Thu, 18 Aug 2005 16:29:46 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id 4B951529B8 for ; Thu, 18 Aug 2005 13:29:46 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [63.195.55.98] (account josh@agliodbs.com HELO spooky) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7751994; Thu, 18 Aug 2005 09:32:01 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: FW: Tx forecast improving harware capabilities. Date: Thu, 18 Aug 2005 09:30:47 -0700 User-Agent: KMail/1.8 Cc: "Sebastian Lallana" References: <722B1B9CFE0ECE488342E3EC2BA920B86CC443@mailserver> In-Reply-To: <722B1B9CFE0ECE488342E3EC2BA920B86CC443@mailserver> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200508180930.48089.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.012 required=5 tests=[AWL=0.012] X-Spam-Level: X-Archive-Number: 200508/230 X-Sequence-Number: 13977 Sebastian, > We are having serious performance problems using JBOSS and PGSQL. How about some information about your application? Performance tuning approaches vary widely according to what you're doing with the database. Also, read this: http://www.powerpostgresql.com/PerfList > I'm sure the problem has to do with the application itself (and neither > with JBOSS nor PGSQL) but the fact is that we are using desktop > equipment to run both Jboss and Postgresql (An Athlon 2600, 1 Gb Ram, > IDE HDD with 60 Mb/sec Transfer Rate), and the answers arise: Well, first off, the IDE HDD is probably killing performance unless your application is 95% read or greater. > If we upgrade our hardware to a Dual Processor would the transactions > per second increase significantly? Would Postgresql take advantage from > SMP? Presumably yes, but can we do a forecast about the number of tps? If this is an OLTP application, chance are that nothing is going to improve performance until you get decent disk support. > What we need is a paper with some figures showing the expected > performance in different environments. Some study about the "degree of > correlation" between TPS and Number of Processors, Cache, Frequency, > Word Size, Architecture, etc. I don't think such a thing exists even for Oracle. Hardware configuration for maximum performance is almost entirely dependant on your application. If it helps, running DBT2 (an OLTP test devised by OSDL after TPC-C), I can easily get 1700 new orders per minute (NOTPM) (about 3000 total multiple-write transactions per minute) on a quad-pentium-III with 4GB RAM and 14 drives, and 6500 notpm on a dual-Itanium machine. > P.S. I've been looking at www.tpc.org but I could't find anything > valuable. Nor would you for any real-world situation even if we had a TPC benchmark (which are involved and expensive, give us a couple of years). The TPC benchmarks are more of a litmus test that your database system & platform are "competitive"; they don't really relate to real-world performance (unless you have budget for an 112-disk system!) -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Thu Aug 18 13:35:41 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0639F52C05 for ; Thu, 18 Aug 2005 13:35:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 85947-06 for ; Thu, 18 Aug 2005 16:35:18 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id C764452A7E for ; Thu, 18 Aug 2005 13:35:16 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7IGZCrc014488; Thu, 18 Aug 2005 11:35:14 -0500 Subject: Re: extremly low memory usage From: Jeremiah Jahn To: John A Meinel Cc: postgres performance In-Reply-To: <4303F090.40602@arbash-meinel.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <4303F090.40602@arbash-meinel.com> Content-Type: text/plain Date: Thu, 18 Aug 2005 11:35:11 -0500 Message-Id: <1124382911.27881.113.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Content-Transfer-Encoding: 7bit Received-SPF: none (penguin.goodinassociates.com: 206.80.71.242 is neither permitted nor denied by domain of cs.earlham.edu) client-ip=206.80.71.242; envelope-from=jeremiah@cs.earlham.edu; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1030/Wed Aug 17 10:53:46 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.031 required=5 tests=[AWL=-0.019, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/231 X-Sequence-Number: 13978 On Wed, 2005-08-17 at 21:21 -0500, John A Meinel wrote: > Jeremiah Jahn wrote: > > I just put together a system with 6GB of ram on a 14 disk raid 10 array. > > When I run my usual big painful queries, I get very little to know > > memory usage. My production box (raid 5 4GB ram) hovers at 3.9GB used > > most of the time. the new devel box sits at around 250MB. > > > > I've switched to an 8.0 system on the new devel box, but the .conf > > really didn't change. Index usage is the same. Something seems wrong and > > I'm not sure why. > > > > How big is your actual database on disk? And how much of it is actually > touched by your queries? The DB is about 60GB. About 10GB is actually used in real queries, versus get me this single record with this ID. I have a large query that finds court cases based on certain criteria that is name based. I get a full seq scan on the name table in about 7 seconds, This table has about 6 million names (most being 'smith, something'). The index scan takes much less time of course, once it's been cached (somewhere but not apparently memory). The really query can take 60 seconds on a first run. And 1.3 seconds on a second run. I'm very happy with the cached results, just not really sure where that caching is happening since it doesn't show up as memory usage. I do know that the caching that happens seems to be independent of the DB. I can restart the DB and my speeds are still the same as the cached second query. Is there some way to pre-cache some of the tables/files on the file system? If I switch my query to search for 'jones%' instead of 'smith%', I take a hit. But if I then rerun the smith search, I still get cached speed. I only have two tables essentially names and events that have to do any real work ie. not very atomic data. I'd love to be able to force these two tables into a cache somewhere. This is a linux system (RHEL ES4) by the way. > > It seems that your tough queries might only be exercising a portion of > the database. If you really want to make memory usage increase try > something like: > find . -type f -print0 | xargs -0 cat >/dev/null > Which should read all the files. After doing that, does the memory usage > increase? > > > > > any thoughts, > > -jj- > > > > > > shared_buffers = 32768 # min 16, at least max_connections*2, 8KB each > > work_mem = 2097151 # min 64, size in KB > > This seems awfully high. 2GB Per sort? This might actually be flushing > some of your ram, since it would get allocated and filled, and then > freed when finished. Remember, depending on what you are doing, this > amount can get allocated more than once per query. What's a good way to determine the optimal size? > > > maintenance_work_mem = 819200 # min 1024, size in KB > > max_fsm_pages = 80000 # min max_fsm_relations*16, 6 bytes each > > checkpoint_segments = 30 # in logfile segments, min 1, 16MB each > > effective_cache_size = 3600000 <-----this is a little out of control, but would it have any real effect? > > It should just tell the planner that it is more likely to have buffers > in cache, so index scans are slightly cheaper than they would otherwise be. > > > random_page_cost = 2 # units are one sequential page fetch cost > > log_min_duration_statement = 10000 # -1 is disabled, in milliseconds. > > lc_messages = 'C' # locale for system error message strings > > lc_monetary = 'C' # locale for monetary formatting > > lc_numeric = 'C' # locale for number formatting > > lc_time = 'C' # locale for time formatting > > > > John > =:-> -- "Now this is a totally brain damaged algorithm. Gag me with a smurfette." -- P. Buhr, Computer Science 354 From pgsql-performance-owner@postgresql.org Thu Aug 18 14:14:19 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9EDED528D1 for ; Thu, 18 Aug 2005 14:14:16 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 17681-01 for ; Thu, 18 Aug 2005 17:14:13 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id 5DF44528CF for ; Thu, 18 Aug 2005 14:14:11 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050818171412m9100g65l2e>; Thu, 18 Aug 2005 17:14:12 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id AC60355FBA; Thu, 18 Aug 2005 12:14:11 -0500 (CDT) Received: from [192.168.1.11] (liliana.arbash-meinel.com [192.168.1.11]) by juju.arbash-meinel.com (Postfix) with ESMTP id A57FB55FAD; Thu, 18 Aug 2005 12:14:05 -0500 (CDT) Message-ID: <4304C1D9.40107@arbash-meinel.com> Date: Thu, 18 Aug 2005 12:14:01 -0500 From: John Arbash Meinel User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050727) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Jeremiah Jahn Cc: postgres performance Subject: Re: extremly low memory usage References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <4303F090.40602@arbash-meinel.com> <1124382911.27881.113.camel@bluejay.goodinassociates.com> In-Reply-To: <1124382911.27881.113.camel@bluejay.goodinassociates.com> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig6C68D44715E31BBBCA7F4421" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.043 required=5 tests=[AWL=-0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/232 X-Sequence-Number: 13979 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig6C68D44715E31BBBCA7F4421 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Jeremiah Jahn wrote: >On Wed, 2005-08-17 at 21:21 -0500, John A Meinel wrote: > > >>Jeremiah Jahn wrote: >> >> >>>I just put together a system with 6GB of ram on a 14 disk raid 10 array. >>>When I run my usual big painful queries, I get very little to know >>>memory usage. My production box (raid 5 4GB ram) hovers at 3.9GB used >>>most of the time. the new devel box sits at around 250MB. >>> >>>I've switched to an 8.0 system on the new devel box, but the .conf >>>really didn't change. Index usage is the same. Something seems wrong and >>>I'm not sure why. >>> >>> >>> >>How big is your actual database on disk? And how much of it is actually >>touched by your queries? >> >> >The DB is about 60GB. About 10GB is actually used in real queries, >versus get me this single record with this ID. I have a large query that >finds court cases based on certain criteria that is name based. I get a >full seq scan on the name table in about 7 seconds, This table has about >6 million names (most being 'smith, something'). The index scan takes >much less time of course, once it's been cached (somewhere but not >apparently memory). The really query can take 60 seconds on a first run. >And 1.3 seconds on a second run. I'm very happy with the cached results, >just not really sure where that caching is happening since it doesn't >show up as memory usage. I do know that the caching that happens seems >to be independent of the DB. I can restart the DB and my speeds are >still the same as the cached second query. Is there some way to >pre-cache some of the tables/files on the file system? If I switch my >query to search for 'jones%' instead of 'smith%', I take a hit. But if I >then rerun the smith search, I still get cached speed. I only have two >tables essentially names and events that have to do any real work ie. >not very atomic data. I'd love to be able to force these two tables into >a cache somewhere. This is a linux system (RHEL ES4) by the way. > > I think what is happening is that *some* of the index pages are being cached, just not all of them. Most indexes (if you didn't specify anything special) are btree, so that you load the root page, and then determine what pages need to be loaded from there. So the "jones%" pages aren't anywhere near the "smith%" pages. And don't need to be loaded if you aren't accessing them. So the required memory usage might be smaller than you think. At least until all of the index pages have been accessed. The reason it is DB independent is because the OS is caching a file access (you read a file, it keeps the old pages in RAM in case you ask for it again). Part of the trick, is that as you use the database, it will cache what has been used. So you may not need to do anything. It should sort itself out with time. However, if you have to have cached performance as soon as your machine reboots, you could figure out what files on disk represent your indexes and tables, and then just "cat $files >/dev/null" That should cause a read on those files, which should pull them into the memory cache. *However* this will fail if the size of those files is greater than available memory, so you may want to be a little bit stingy about what you preload. Alternatively, you could just write an SQL script which runs a bunch of indexed queries to make sure all the pages get loaded. Something like: FOR curname IN SELECT DISTINCT name FROM users LOOP SELECT name FROM users WHERE name=curname; END LOOP; That should make the database go through the entire table, and load the index for every user. This is overkill, and will probably take a long time to execute. But you could do it if you wanted. >>It seems that your tough queries might only be exercising a portion of >>the database. If you really want to make memory usage increase try >>something like: >>find . -type f -print0 | xargs -0 cat >/dev/null >>Which should read all the files. After doing that, does the memory usage >>increase? >> >> >> >>>any thoughts, >>>-jj- >>> >>> >>>shared_buffers = 32768 # min 16, at least max_connections*2, 8KB each >>>work_mem = 2097151 # min 64, size in KB >>> >>> >>This seems awfully high. 2GB Per sort? This might actually be flushing >>some of your ram, since it would get allocated and filled, and then >>freed when finished. Remember, depending on what you are doing, this >>amount can get allocated more than once per query. >> >> >What's a good way to determine the optimal size? > > Practice. :) A few questions I guess... How many concurrent connections are you expecting? How many joins does a standard query have? How big are the joins? In general, I would tend to make this a smaller number, so that the os has more room to cache tables, rather than having big buffers for joins. If someone is requesting a join that requires a lot of rows, I would rather *that* query be slower, than impacting everyone else. I would put it more with a maximum in the 20-100MB range. John =:-> > > >>>maintenance_work_mem = 819200 # min 1024, size in KB >>>max_fsm_pages = 80000 # min max_fsm_relations*16, 6 bytes each >>>checkpoint_segments = 30 # in logfile segments, min 1, 16MB each >>>effective_cache_size = 3600000 <-----this is a little out of control, but would it have any real effect? >>> >>> >>It should just tell the planner that it is more likely to have buffers >>in cache, so index scans are slightly cheaper than they would otherwise be. >> >> >> >>>random_page_cost = 2 # units are one sequential page fetch cost >>>log_min_duration_statement = 10000 # -1 is disabled, in milliseconds. >>>lc_messages = 'C' # locale for system error message strings >>>lc_monetary = 'C' # locale for monetary formatting >>>lc_numeric = 'C' # locale for number formatting >>>lc_time = 'C' # locale for time formatting >>> >>> >>> >>John >>=:-> >> >> --------------enig6C68D44715E31BBBCA7F4421 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDBMHdJdeBCYSNAAMRAjBbAJ47D8yVbYeYhjL6JHPmJKHrn5M4sgCgiAhT MAh3/BrF05iFf9uGUe8CkQI= =BksX -----END PGP SIGNATURE----- --------------enig6C68D44715E31BBBCA7F4421-- From pgsql-performance-owner@postgresql.org Thu Aug 18 14:39:28 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CA34452BC5 for ; Thu, 18 Aug 2005 14:39:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78359-04 for ; Thu, 18 Aug 2005 17:39:23 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id CFA6352B98 for ; Thu, 18 Aug 2005 14:39:21 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7IHdMMx016018; Thu, 18 Aug 2005 12:39:22 -0500 Subject: Re: extremly low memory usage From: Jeremiah Jahn To: Jeff Trout Cc: postgres performance In-Reply-To: References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> Content-Type: text/plain Date: Thu, 18 Aug 2005 12:39:21 -0500 Message-Id: <1124386761.27881.119.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Content-Transfer-Encoding: 7bit Received-SPF: none (penguin.goodinassociates.com: 206.80.71.242 is neither permitted nor denied by domain of cs.earlham.edu) client-ip=206.80.71.242; envelope-from=jeremiah@cs.earlham.edu; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1030/Wed Aug 17 10:53:46 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.032 required=5 tests=[AWL=-0.018, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/233 X-Sequence-Number: 13980 here's an example standard query. Ireally have to make the first hit go faster. The table is clustered as well on full_name as well. 'Smith%' took 87 seconds on the first hit. I wonder if I set up may array wrong. I remeber see something about DMA access versus something else, and choose DMA access. LVM maybe? explain analyze select distinct case_category,identity_id,court.name,litigant_details.case_id,case_year,date_of_birth,assigned_case_role,litigant_details.court_ori,full_name,litigant_details.actor_id,case_data.type_code,case_data.subtype_code,litigant_details.impound_litigant_data, to_number(trim(leading case_data.type_code from trim(leading case_data.case_year from case_data.case_id)),'999999') as seq from identity,court,litigant_details,case_data where identity.court_ori = litigant_details.court_ori and identity.case_id = litigant_details.case_id and identity.actor_id = litigant_details.actor_id and court.id = identity.court_ori and identity.court_ori = case_data.court_ori and case_data.case_id = identity.case_id and identity.court_ori = 'IL081025J' and full_name like 'MILLER%' order by full_name; QUERY PLAN --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Unique (cost=20411.84..20411.91 rows=2 width=173) (actual time=38340.231..38355.120 rows=4906 loops=1) -> Sort (cost=20411.84..20411.84 rows=2 width=173) (actual time=38340.227..38343.667 rows=4906 loops=1) Sort Key: identity.full_name, case_data.case_category, identity.identity_id, court.name, litigant_details.case_id, case_data.case_year, identity.date_of_birth, litigant_details.assigned_case_role, litigant_details.court_ori, litigant_details.actor_id, case_data.type_code, case_data.subtype_code, litigant_details.impound_litigant_data, to_number(ltrim(ltrim((case_data.case_id)::text, (case_data.case_year)::text), (case_data.type_code)::text), '999999'::text) -> Nested Loop (cost=0.00..20411.83 rows=2 width=173) (actual time=12.891..38317.017 rows=4906 loops=1) -> Nested Loop (cost=0.00..20406.48 rows=1 width=159) (actual time=12.826..23232.106 rows=4906 loops=1) -> Nested Loop (cost=0.00..20403.18 rows=1 width=138) (actual time=12.751..22885.439 rows=4906 loops=1) Join Filter: (("outer".case_id)::text = ("inner".case_id)::text) -> Index Scan using name_speed on identity (cost=0.00..1042.34 rows=4868 width=82) (actual time=0.142..52.538 rows=4915 loops=1) Index Cond: (((full_name)::text >= 'MILLER'::character varying) AND ((full_name)::text < 'MILLES'::character varying)) Filter: (((court_ori)::text = 'IL081025J'::text) AND ((full_name)::text ~~ 'MILLER%'::text)) -> Index Scan using lit_actor_speed on litigant_details (cost=0.00..3.96 rows=1 width=81) (actual time=4.631..4.635 rows=1 loops=4915) Index Cond: (("outer".actor_id)::text = (litigant_details.actor_id)::text) Filter: ('IL081025J'::text = (court_ori)::text) -> Seq Scan on court (cost=0.00..3.29 rows=1 width=33) (actual time=0.053..0.062 rows=1 loops=4906) Filter: ('IL081025J'::text = (id)::text) -> Index Scan using case_speed on case_data (cost=0.00..5.29 rows=3 width=53) (actual time=3.049..3.058 rows=1 loops=4906) Index Cond: (('IL081025J'::text = (case_data.court_ori)::text) AND ((case_data.case_id)::text = ("outer".case_id)::text)) Total runtime: 38359.722 ms (18 rows) copa=> explain analyze select distinct case_category,identity_id,court.name,litigant_details.case_id,case_year,date_of_birth,assigned_case_role,litigant_details.court_ori,full_name,litigant_details.actor_id,case_data.type_code,case_data.subtype_code,litigant_details.impound_litigant_data, to_number(trim(leading case_data.type_code from trim(leading case_data.case_year from case_data.case_id)),'999999') as seq from identity,court,litigant_details,case_data where identity.court_ori = litigant_details.court_ori and identity.case_id = litigant_details.case_id and identity.actor_id = litigant_details.actor_id and court.id = identity.court_ori and identity.court_ori = case_data.court_ori and case_data.case_id = identity.case_id and identity.court_ori = 'IL081025J' and full_name like 'MILLER%' order by full_name; QUERY PLAN --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Unique (cost=20411.84..20411.91 rows=2 width=173) (actual time=666.832..688.081 rows=4906 loops=1) -> Sort (cost=20411.84..20411.84 rows=2 width=173) (actual time=666.825..671.833 rows=4906 loops=1) Sort Key: identity.full_name, case_data.case_category, identity.identity_id, court.name, litigant_details.case_id, case_data.case_year, identity.date_of_birth, litigant_details.assigned_case_role, litigant_details.court_ori, litigant_details.actor_id, case_data.type_code, case_data.subtype_code, litigant_details.impound_litigant_data, to_number(ltrim(ltrim((case_data.case_id)::text, (case_data.case_year)::text), (case_data.type_code)::text), '999999'::text) -> Nested Loop (cost=0.00..20411.83 rows=2 width=173) (actual time=0.216..641.366 rows=4906 loops=1) -> Nested Loop (cost=0.00..20406.48 rows=1 width=159) (actual time=0.149..477.063 rows=4906 loops=1) -> Nested Loop (cost=0.00..20403.18 rows=1 width=138) (actual time=0.084..161.045 rows=4906 loops=1) Join Filter: (("outer".case_id)::text = ("inner".case_id)::text) -> Index Scan using name_speed on identity (cost=0.00..1042.34 rows=4868 width=82) (actual time=0.047..37.898 rows=4915 loops=1) Index Cond: (((full_name)::text >= 'MILLER'::character varying) AND ((full_name)::text < 'MILLES'::character varying)) Filter: (((court_ori)::text = 'IL081025J'::text) AND ((full_name)::text ~~ 'MILLER%'::text)) -> Index Scan using lit_actor_speed on litigant_details (cost=0.00..3.96 rows=1 width=81) (actual time=0.015..0.017 rows=1 loops=4915) Index Cond: (("outer".actor_id)::text = (litigant_details.actor_id)::text) Filter: ('IL081025J'::text = (court_ori)::text) -> Seq Scan on court (cost=0.00..3.29 rows=1 width=33) (actual time=0.049..0.056 rows=1 loops=4906) Filter: ('IL081025J'::text = (id)::text) -> Index Scan using case_speed on case_data (cost=0.00..5.29 rows=3 width=53) (actual time=0.017..0.020 rows=1 loops=4906) Index Cond: (('IL081025J'::text = (case_data.court_ori)::text) AND ((case_data.case_id)::text = ("outer".case_id)::text)) Total runtime: 694.639 ms (18 rows) On Thu, 2005-08-18 at 09:00 -0400, Jeff Trout wrote: > On Aug 17, 2005, at 10:11 PM, Jeremiah Jahn wrote: > > > I just put together a system with 6GB of ram on a 14 disk raid 10 > > array. > > When I run my usual big painful queries, I get very little to know > > memory usage. My production box (raid 5 4GB ram) hovers at 3.9GB used > > most of the time. the new devel box sits at around 250MB. > > > > Is the system performing fine? Are you touching as much data as the > production box? > > If the system is performing fine don't worry about it. > > > work_mem = 2097151 # min 64, size in KB > > This is EXTREMELY high. You realize this is the amount of memory > that can be used per-sort and per-hash build in a query? You can end > up with multiples of this on a single query. If you have some big > queries that are run infrequently have them set it manually. > > > effective_cache_size = 3600000 <-----this is a little out of > > control, but would it have any real effect? > > This doesn't allocate anything - it is a hint to the planner about > how much data it can assume is cached. > > -- > Jeff Trout > http://www.jefftrout.com/ > http://www.stuarthamm.net/ > > > > ---------------------------(end of broadcast)--------------------------- > TIP 9: In versions below 8.0, the planner will ignore your desire to > choose an index scan if your joining column's datatypes do not > match -- "Now this is a totally brain damaged algorithm. Gag me with a smurfette." -- P. Buhr, Computer Science 354 From pgsql-performance-owner@postgresql.org Thu Aug 18 14:55:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E39AB52C67 for ; Thu, 18 Aug 2005 14:55:18 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 53635-01 for ; Thu, 18 Aug 2005 17:55:12 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id 0E34252BE2 for ; Thu, 18 Aug 2005 14:55:10 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050818175508m92009jleke>; Thu, 18 Aug 2005 17:55:09 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 572B955FBA; Thu, 18 Aug 2005 12:55:08 -0500 (CDT) Received: from [192.168.1.11] (liliana.arbash-meinel.com [192.168.1.11]) by juju.arbash-meinel.com (Postfix) with ESMTP id B882E55FAD; Thu, 18 Aug 2005 12:55:03 -0500 (CDT) Message-ID: <4304CB77.4070403@arbash-meinel.com> Date: Thu, 18 Aug 2005 12:55:03 -0500 From: John Arbash Meinel User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050727) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Jeremiah Jahn Cc: Jeff Trout , postgres performance Subject: Re: extremly low memory usage References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> In-Reply-To: <1124386761.27881.119.camel@bluejay.goodinassociates.com> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigEB715DABC17B44CB3B41A889" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.043 required=5 tests=[AWL=-0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/234 X-Sequence-Number: 13981 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigEB715DABC17B44CB3B41A889 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Jeremiah Jahn wrote: >here's an example standard query. Ireally have to make the first hit go >faster. The table is clustered as well on full_name as well. 'Smith%' >took 87 seconds on the first hit. I wonder if I set up may array wrong. >I remeber see something about DMA access versus something else, and >choose DMA access. LVM maybe? > > It would be nice if you would format your queries to be a little bit easier to read before posting them. However, I believe I am reading it correctly, to say that the index scan on identity is not your slow point. In fact, as near as I can tell, it only takes 52ms to complete. The expensive parts are the 4915 lookups into the litigant_details (each one takes approx 4ms for a total of ~20s). And then you do it again on case_data (average 3ms each * 4906 loops = ~15s). So there is no need for preloading your indexes on the identity table. It is definitely not the bottleneck. So a few design bits, which may help your database. Why is "actor_id" a text field instead of a number? You could try creating an index on "litigant_details (actor_id, count_ori)" so that it can do just an index lookup, rather than an index + filter. More importantly, though, the planner seems to think the join of identity to litigant_details will only return 1 row, not 5000. Do you regularly vacuum analyze your tables? Just as a test, try running: set enable_nested_loop to off; And then run EXPLAIN ANALYZE again, just to see if it is faster. You probably need to increase some statistics targets, so that the planner can design better plans. > -> Nested Loop (cost=0.00..20411.83 rows=2 width=173) > (actual time=12.891..38317.017 rows=4906 loops=1) > -> Nested Loop (cost=0.00..20406.48 rows=1 width=159) > (actual time=12.826..23232.106 rows=4906 loops=1) > -> Nested Loop (cost=0.00..20403.18 rows=1 > width=138) (actual time=12.751..22885.439 rows=4906 loops=1) > Join Filter: (("outer".case_id)::text = > ("inner".case_id)::text) > -> Index Scan using name_speed on > identity (cost=0.00..1042.34 rows=4868 width=82) (actual > time=0.142..52.538 rows=4915 loops=1) > Index Cond: (((full_name)::text >= > 'MILLER'::character varying) AND ((full_name)::text < > 'MILLES'::character varying)) > Filter: (((court_ori)::text = > 'IL081025J'::text) AND ((full_name)::text ~~ 'MILLER%'::text)) > -> Index Scan using lit_actor_speed on > litigant_details (cost=0.00..3.96 rows=1 width=81) (actual > time=4.631..4.635 rows=1 loops=4915) > Index Cond: (("outer".actor_id)::text > = (litigant_details.actor_id)::text) > Filter: ('IL081025J'::text = > (court_ori)::text) > -> Seq Scan on court (cost=0.00..3.29 rows=1 > width=33) (actual time=0.053..0.062 rows=1 loops=4906) > Filter: ('IL081025J'::text = (id)::text) > -> Index Scan using case_speed on case_data > (cost=0.00..5.29 rows=3 width=53) (actual time=3.049..3.058 rows=1 > loops=4906) > Index Cond: (('IL081025J'::text = > (case_data.court_ori)::text) AND ((case_data.case_id)::text = > ("outer".case_id)::text)) John =:-> --------------enigEB715DABC17B44CB3B41A889 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDBMt3JdeBCYSNAAMRAj33AJ9qq96tbNdUfwLtoN5YdV8t8mRgmQCgp/rk mZgsruuYF/yOK8Smgkk+XQE= =CRGt -----END PGP SIGNATURE----- --------------enigEB715DABC17B44CB3B41A889-- From pgsql-performance-owner@postgresql.org Thu Aug 18 16:23:09 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1064352B28 for ; Thu, 18 Aug 2005 16:22:52 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 34278-02 for ; Thu, 18 Aug 2005 19:22:48 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id 9147C52AB8 for ; Thu, 18 Aug 2005 16:22:46 -0300 (ADT) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: limit number of concurrent callers to a stored proc? Date: Thu, 18 Aug 2005 15:22:46 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3417DD147@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] limit number of concurrent callers to a stored proc? Thread-Index: AcWjnWbJ7u0/UiC6RFm84YoIU7CBagAi8AjA From: "Merlin Moncure" To: "Christopher Kings-Lynne" Cc: , X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/235 X-Sequence-Number: 13982 Christopher > You could use a 1 column/1 row table perhaps. Use some sort of locking > mechanism. >=20 > Also, check out contrib/userlock userlock is definitely the way to go for this type of problem. =20 The are really the only way to provide locking facilities that live outside transactions. You are provided with 48 bits of lock space in the form of offset/block in 32 bit field and a 16 bit field. The 16 bit field could be the pid of the locker and the 32 bit field the oid of the function. Unfortunately, userlocks are not really easy to query via the pg_locks() view. However this has been addressed for 8.1. In 8.1, it will be trivial to create a function which checked the number of lockers on the function oid and acquire a lock if less than a certain amount. Merlin From pgsql-performance-owner@postgresql.org Thu Aug 18 16:57:24 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id EFDB352A71 for ; Thu, 18 Aug 2005 16:57:07 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 01995-09 for ; Thu, 18 Aug 2005 19:57:02 +0000 (GMT) Received: from smtpauth08.mail.atl.earthlink.net (smtpauth08.mail.atl.earthlink.net [209.86.89.68]) by svr1.postgresql.org (Postfix) with ESMTP id 44A4052A44 for ; Thu, 18 Aug 2005 16:57:00 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=sylsv3mtKhWcyFo9YODu3cbtw8zheltuLTFbrPQ0QgC+3r/BALJVpx/IIA/Uuq18; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:Cc:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth08.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E5qVX-0006nl-SX; Thu, 18 Aug 2005 15:57:00 -0400 Message-Id: <6.2.3.4.0.20050818141059.05c2fac0@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Thu, 18 Aug 2005 15:56:53 -0400 To: Jeremiah Jahn From: Ron Subject: Re: extremly low memory usage Cc: postgres performance In-Reply-To: <4304CB77.4070403@arbash-meinel.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bc6a30abf5fc3b1b4640a4a4551cdf78ba350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.882 required=5 tests=[AWL=0.508, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/236 X-Sequence-Number: 13983 At 01:55 PM 8/18/2005, John Arbash Meinel wrote: >Jeremiah Jahn wrote: > > >here's an example standard query. Ireally have to make the first hit go > >faster. The table is clustered as well on full_name as well. 'Smith%' > >took 87 seconds on the first hit. I wonder if I set up may array wrong. > >I remeber see something about DMA access versus something else, and > >choose DMA access. LVM maybe? > > > > >It would be nice if you would format your queries to be a little bit >easier to read before posting them. >However, I believe I am reading it correctly, to say that the index scan >on identity is not your slow point. In fact, as near as I can tell, it >only takes 52ms to complete. > >The expensive parts are the 4915 lookups into the litigant_details (each >one takes approx 4ms for a total of ~20s). >And then you do it again on case_data (average 3ms each * 4906 loops = >~15s). How big are litigant_details and case_data? If they can fit in RAM, preload them using methods like the "cat to /dev/null" trick and those table lookups will be ~100-1000x faster. If they won't fit into RAM but the machine can be expanded to hold enough RAM to fit the tables, it's well worth the ~$75-$150/GB to upgrade the server so that the tables will fit into RAM. If they can't be made to fit into RAM as atomic entities, you have a few choices: A= Put the data tables and indexes on separate dedicated spindles and put litigant_details and case_data each on their own dedicated spindles. This will lower seek conflicts. Again it this requires buying some more HDs, it's well worth it. B= Break litigant_details and case_data into a set of smaller tables (based on something sane like the first n characters of the primary key) such that the smaller tables easily fit into RAM. Given that you've said only 10GB/60GB is "hot", this could work very well. Combine it with "A" above (put all the litigant_details sub tables on one dedicated spindle set and all the case_data sub tables on another spindle set) for added oomph. C= Buy a SSD big enough to hold litigant_details and case_data and put them there. Again, this can be combined with "A" and "B" above to lessen the size of the SSD needed. >So there is no need for preloading your indexes on the identity >table. It is definitely not the bottleneck. > >So a few design bits, which may help your database. Why is >"actor_id" a text field instead of a number? >You could try creating an index on "litigant_details (actor_id, >count_ori)" so that it can do just an index lookup, rather than an >index+ filter. Yes, that certainly sounds like it would be more efficient. >More importantly, though, the planner seems to think the join of >identity to litigant_details will only return 1 row, not 5000. >Do you regularly vacuum analyze your tables? >Just as a test, try running: >set enable_nested_loop to off; >And then run EXPLAIN ANALYZE again, just to see if it is faster. > >You probably need to increase some statistics targets, so that the >planner can design better plans. > > > -> Nested Loop (cost=0.00..20411.83 rows=2 width=173) (actual > time=12.891..38317.017 rows=4906 loops=1) > > -> Nested Loop (cost=0.00..20406.48 rows=1 width=159)(actual > time=12.826..23232.106 rows=4906 loops=1) > > -> Nested Loop (cost=0.00..20403.18 rows=1 width=138) > (actual time=12.751..22885.439 rows=4906 loops=1) > > Join Filter: (("outer".case_id)::text = > ("inner".case_id)::text) > > -> Index Scan using name_speed on > identity (cost=0.00..1042.34 rows=4868 width=82) (actual time=0.142..52.538 > > rows=4915 loops=1) > > Index Cond: (((full_name)::text >= > 'MILLER'::character varying) AND ((full_name)::text < > 'MILLES'::character varying)) > > Filter: (((court_ori)::text = > 'IL081025J'::text) AND ((full_name)::text ~~ 'MILLER%'::text)) > > -> Index Scan using lit_actor_speed on > litigant_details (cost=0.00..3.96 rows=1 width=81) (actual > > time=4.631..4.635 rows=1 loops=4915) > > Index Cond: (("outer".actor_id)::text = > (litigant_details.actor_id)::text) > > Filter: ('IL081025J'::text = (court_ori)::text) > > -> Seq Scan on court (cost=0.00..3.29 > rows=1 width=33) (actual time=0.053..0.062 rows=1 loops=4906) > > Filter: ('IL081025J'::text = (id)::text) > > -> Index Scan using case_speed on > case_data (cost=0.00..5.29 rows=3 width=53) (actual time=3.049..3.058 > > rows=1 loops=4906) > > Index Cond: (('IL081025J'::text > = (case_data.court_ori)::text) AND ((case_data.case_id)::text = > > ("outer".case_id)::text)) From pgsql-performance-owner@postgresql.org Thu Aug 18 19:08:56 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 495AF528C3 for ; Thu, 18 Aug 2005 19:08:54 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 43853-02 for ; Thu, 18 Aug 2005 22:08:50 +0000 (GMT) Received: from dns.deep-purple.com (dns.deep-purple.com [209.61.158.121]) by svr1.postgresql.org (Postfix) with ESMTP id AEE5652BFA for ; Thu, 18 Aug 2005 19:08:46 -0300 (ADT) Received: from [192.168.5.2] (195-112-29-123.dyn.gotadsl.co.uk [195.112.29.123]) (authenticated) by dns.deep-purple.com (8.11.6/8.11.6) with ESMTP id j7IM8ZT07600; Thu, 18 Aug 2005 17:08:35 -0500 In-Reply-To: <722B1B9CFE0ECE488342E3EC2BA920B86CC443@mailserver> References: <722B1B9CFE0ECE488342E3EC2BA920B86CC443@mailserver> Mime-Version: 1.0 (Apple Message framework v734) Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: <701BBF0B-1E25-4BC0-BF7C-0D53EAD2784D@hodgkinson.org> Cc: Content-Transfer-Encoding: 7bit From: David Hodgkinson Subject: Re: FW: Tx forecast improving harware capabilities. Date: Thu, 18 Aug 2005 23:08:29 +0100 To: Sebastian Lallana X-Mailer: Apple Mail (2.734) X-RIF3-MailScanner-Information: Please contact the ISP for more information X-RIF3-MailScanner: Found to be clean X-RIF3-MailScanner-SpamCheck: not spam, SpamAssassin (score=-4.871, required 5, autolearn=not spam, AWL 0.03, BAYES_00 -4.90) X-MailScanner-From: daveh@hodgkinson.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/237 X-Sequence-Number: 13984 On 18 Aug 2005, at 16:01, Sebastian Lallana wrote: > It exists something like this? Does anybody has experience about > this subject? I've just been through this with a client with both a badly tuned Pg and an application being less than optimal. First, find a benchmark. Just something you can hold on to. For us, it was the generation time of the site's home page. In this case, 7 seconds. We looked hard at postgresql.conf, planned the memory usage, sort_memory and all that. That was a boost. Then we looked at the queries that were being thrown at the database. Over 200 to build one page! So, a layer of caching was built into the web server layer. Finally, some frequently occurring combinations of queries were pushed down into stored procs. We got the page gen time down to 1.5 seconds AND the server being stable under extreme stress. So, a fair win. Thanks to cms for several clues. So, without understanding your application and were it's taking the time, you can't begin to estimate hardware usage. From pgsql-performance-owner@postgresql.org Fri Aug 19 03:56:12 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7009052D18 for ; Fri, 19 Aug 2005 03:55:33 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 87395-08 for ; Fri, 19 Aug 2005 06:55:21 +0000 (GMT) Received: from mail23.sea5.speakeasy.net (mail23.sea5.speakeasy.net [69.17.117.25]) by svr1.postgresql.org (Postfix) with ESMTP id 15F8352DE8 for ; Fri, 19 Aug 2005 03:55:18 -0300 (ADT) Received: (qmail 5497 invoked from network); 19 Aug 2005 06:55:18 -0000 Received: from dsl081-060-184.sfo1.dsl.speakeasy.net (HELO noodles) ([64.81.60.184]) (envelope-sender ) by mail23.sea5.speakeasy.net (qmail-ldap-1.03) with RC4-MD5 encrypted SMTP for ; 19 Aug 2005 06:55:18 -0000 Subject: Re: Query plan looks OK, but slow I/O - settings advice? From: "Jeffrey W. Baker" To: Roger Hand Cc: pgsql-performance@postgresql.org In-Reply-To: References: Content-Type: text/plain Date: Thu, 18 Aug 2005 23:55:35 -0700 Message-Id: <1124434535.3257.7.camel@noodles> Mime-Version: 1.0 X-Mailer: Evolution 2.2.1.1 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/247 X-Sequence-Number: 13994 On Tue, 2005-08-16 at 10:46 -0700, Roger Hand wrote: > The disks are ext3 with journalling type of ordered, but this was later changed to writeback with no apparent change in speed. > > They're on a Dell poweredge 6650 with LSI raid card, setup as follows: > 4 disks raid 10 for indexes (145GB) - sdc1 > 6 disks raid 10 for data (220GB) - sdd1 > 2 mirrored disks for logs - sdb1 > > stripe size is 32k > cache policy: cached io (am told the controller has bbu) > write policy: write-back > read policy: readahead I assume you are using Linux 2.6. Have you considered booting your machine with elevator=deadline? You can also change this at runtime using sysfs. These read speeds are not too impressive. Perhaps this is a slow controller. Alternately you might need bigger CPUs. There's a lot of possibilities, obviously :) I'd start with the elevator, since that's easily tested. -jwb From pgsql-performance-owner@postgresql.org Fri Aug 19 04:18:05 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9CF4152CEA for ; Fri, 19 Aug 2005 04:18:03 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 45563-10 for ; Fri, 19 Aug 2005 07:17:59 +0000 (GMT) Received: from linuxworld.com.au (unknown [203.34.46.50]) by svr1.postgresql.org (Postfix) with ESMTP id 71FEE52CBB for ; Fri, 19 Aug 2005 04:17:56 -0300 (ADT) Received: from linuxworld.com.au (IDENT:swm@localhost.localdomain [127.0.0.1]) by linuxworld.com.au (8.13.2/8.13.2) with ESMTP id j7J7HlIO031072; Fri, 19 Aug 2005 17:17:48 +1000 Received: from localhost (swm@localhost) by linuxworld.com.au (8.13.2/8.13.2/Submit) with ESMTP id j7J7HjEF031069; Fri, 19 Aug 2005 17:17:47 +1000 X-Authentication-Warning: linuxworld.com.au: swm owned process doing -bs Date: Fri, 19 Aug 2005 17:17:45 +1000 (EST) From: Gavin Sherry X-X-Sender: swm@linuxworld.com.au To: Roger Hand Cc: pgsql-performance@postgresql.org Subject: Re: Query plan looks OK, but slow I/O - settings advice? In-Reply-To: Message-ID: References: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.002 required=5 tests=[AWL=0.002] X-Spam-Level: X-Archive-Number: 200508/248 X-Sequence-Number: 13995 The query plan does *not* look okay. > electric=# EXPLAIN ANALYZE > electric-# SELECT datavalue, logfielddatatype, timestamp FROM logdata_recent > electric-# WHERE (logfielddatatype = 70 OR logfielddatatype = 71 OR logfielddatatype = 69) > electric-# AND graphtargetlog = 1327 > electric-# AND timestamp >= 1123052400 AND timestamp <= 1123138800 > electric-# ORDER BY timestamp; > QUERY PLAN > -------------------------------------------------- > Sort (cost=82.48..82.50 rows=6 width=14) (actual time=60208.968..60211.232 rows=2625 loops=1) > Sort Key: public.logdata_recent."timestamp" > -> Result (cost=0.00..82.41 rows=6 width=14) (actual time=52.483..60200.868 rows=2625 loops=1) > -> Append (cost=0.00..82.41 rows=6 width=14) (actual time=52.476..60189.929 rows=2625 loops=1) > -> Seq Scan on logdata_recent (cost=0.00..46.25 rows=1 width=14) (actual time=0.003..0.003 rows=0 loops=1) > Filter: (((logfielddatatype = 70) OR (logfielddatatype = 71) OR (logfielddatatype = 69)) AND (graphtargetlog = 1327) AND ("timestamp" >= 1123052400) AND ("timestamp" <= 1123138800)) > -> Index Scan using logdata_recent_1123085306_ix_t_fld_gtl, logdata_recent_1123085306_ix_t_fld_gtl, logdata_recent_1123085306_ix_t_fld_gtl on logdata_recent_stale logdata_recent (cost=0.00..18.08 rows=3 width=14) (actual time=52.465..60181.624 rows=2625 loops=1) Notice here that expected rows is 3, but actual rows is a hell of a lot higher. Try increasing stats collections for the columns on which logdata_recent_1123085306_ix_t_fld_gtl is declared. Also, the actual index scan is taking a long time. How recently have you vacuum full'd? Thanks, Gavin From pgsql-performance-owner@postgresql.org Fri Aug 19 04:35:47 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5E1BE52DD0 for ; Fri, 19 Aug 2005 04:35:31 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91837-09 for ; Fri, 19 Aug 2005 07:35:25 +0000 (GMT) Received: from sj1-exch-01.us.corp.kailea.com (mail.ragingnet.net [209.249.149.2]) by svr1.postgresql.org (Postfix) with ESMTP id 5CBE252DFC for ; Fri, 19 Aug 2005 04:35:22 -0300 (ADT) X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Subject: Re: Query plan looks OK, but slow I/O - settings advice? Date: Fri, 19 Aug 2005 00:35:23 -0700 Message-ID: X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Query plan looks OK, but slow I/O - settings advice? Thread-Index: AcWkiviNpjnrx5NjS8eZ5LEY435vXgAA/ycw From: "Roger Hand" To: "Jeffrey W. Baker" Cc: , "Leszek Kotzian" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.344 required=5 tests=[AWL=-0.292, FORGED_RCVD_HELO=0.05, HOT_NASTY=0.586] X-Spam-Level: X-Archive-Number: 200508/249 X-Sequence-Number: 13996 Jeffrey W. Baker wrote: > On Tue, 2005-08-16 at 10:46 -0700, Roger Hand wrote: >> The disks are ext3 with journalling type of ordered, but this was = later changed to writeback with no apparent change in speed. >>=20 >> They're on a Dell poweredge 6650 with LSI raid card, setup as = follows: >> 4 disks raid 10 for indexes (145GB) - sdc1 >> 6 disks raid 10 for data (220GB) - sdd1 >> 2 mirrored disks for logs - sdb1 >>=20 >> stripe size is 32k >> cache policy: cached io (am told the controller has bbu) >> write policy: write-back >> read policy: readahead >=20 > I assume you are using Linux 2.6. =20 Oops, sorry I left that out. Nope, we're on 2.4: [root@rage-db2 ~]$ uname -a Linux xxx.xxx.xxx 2.4.21-27.0.2.ELsmp #1 SMP Wed Jan 12 23:35:44 EST = 2005 i686 i686 i386 GNU/Linux It's RedHat Enterprise AS3.0 Fri Nov 5 17:55:14 PST 2004 > Have you considered booting your > machine with elevator=3Ddeadline?=20 I just did a little Googling and see that the 2.4 kernel didn't have a = decent elevator tuning system, and that was fixed in 2.6. Hmmm .... Thanks for the ideas ... -Roger From pgsql-performance-owner@postgresql.org Fri Aug 19 05:24:10 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id EA32B52D58 for ; Fri, 19 Aug 2005 05:24:08 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 29025-05 for ; Fri, 19 Aug 2005 08:24:04 +0000 (GMT) Received: from web32915.mail.mud.yahoo.com (web32915.mail.mud.yahoo.com [68.142.206.62]) by svr1.postgresql.org (Postfix) with SMTP id A39C352CEA for ; Fri, 19 Aug 2005 05:24:02 -0300 (ADT) Received: (qmail 53753 invoked by uid 60001); 19 Aug 2005 08:24:04 -0000 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Message-ID:Received:Date:From:Subject:To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=QEFBL9/jqKbUl/eHyZ2CWh/w//9L6anHWaJUkX/5A9UHFP9xuUFfIPQF22XtubMz6C+UUk08f3Lq3crR+DP0V2Jtuw++n1iTzgQDh8V37Gvtf4+n35KAoxvN9T5uZQZSYL4sWAA3Ibwf7u9lPsRM/TULTmjSj+ta2lISmxoVHCk= ; Message-ID: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> Received: from [66.56.49.200] by web32915.mail.mud.yahoo.com via HTTP; Fri, 19 Aug 2005 01:24:04 PDT Date: Fri, 19 Aug 2005 01:24:04 -0700 (PDT) From: Mark Cotner Subject: sustained update load of 1-2k/sec To: pgsql-performance@postgresql.org MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.28 required=5 tests=[AWL=-0.093, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/250 X-Sequence-Number: 13997 Hi all, I bet you get tired of the same ole questions over and over. I'm currently working on an application that will poll thousands of cable modems per minute and I would like to use PostgreSQL to maintain state between polls of each device. This requires a very heavy amount of updates in place on a reasonably large table(100k-500k rows, ~7 columns mostly integers/bigint). Each row will be refreshed every 15 minutes, or at least that's how fast I can poll via SNMP. I hope I can tune the DB to keep up. The app is threaded and will likely have well over 100 concurrent db connections. Temp tables for storage aren't a preferred option since this is designed to be a shared nothing approach and I will likely have several polling processes. Here are some of my assumptions so far . . . HUGE WAL Vacuum hourly if not more often I'm getting 1700tx/sec from MySQL and I would REALLY prefer to use PG. I don't need to match the number, just get close. Is there a global temp table option? In memory tables would be very beneficial in this case. I could just flush it to disk occasionally with an insert into blah select from memory table. Any help or creative alternatives would be greatly appreciated. :) 'njoy, Mark -- Writing software requires an intelligent person, creating functional art requires an artist. -- Unknown From pgsql-performance-owner@postgresql.org Fri Aug 19 06:09:47 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2477C52C59 for ; Fri, 19 Aug 2005 06:09:45 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 46670-05 for ; Fri, 19 Aug 2005 09:09:37 +0000 (GMT) Received: from redivi.com (redivi.com [64.207.133.54]) by svr1.postgresql.org (Postfix) with ESMTP id E885C52B52 for ; Fri, 19 Aug 2005 06:09:35 -0300 (ADT) Received: from [64.203.23.17] (helo=[192.168.0.8]) by redivi.com with esmtpsa (TLSv1:RC4-SHA:128) (Exim 4.43) id 1E62sa-0001E6-6n; Fri, 19 Aug 2005 02:09:37 -0700 In-Reply-To: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> References: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> Mime-Version: 1.0 (Apple Message framework v733) Message-Id: <67788A4A-4011-49AB-B329-683FD9532661@redivi.com> Cc: pgsql-performance@postgresql.org From: Bob Ippolito Date: Thu, 18 Aug 2005 23:09:27 -1000 To: Mark Cotner X-Mailer: Apple Mail (2.733) X-SA-Exim-Connect-IP: 64.203.23.17 X-SA-Exim-Mail-From: bob@redivi.com Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: sustained update load of 1-2k/sec X-SA-Exim-Version: 4.1+cvs (built Mon, 23 Aug 2004 08:44:05 -0700) X-SA-Exim-Scanned: Yes (on redivi.com) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/251 X-Sequence-Number: 13998 On Aug 18, 2005, at 10:24 PM, Mark Cotner wrote: > I'm currently working on an application that will poll > thousands of cable modems per minute and I would like > to use PostgreSQL to maintain state between polls of > each device. This requires a very heavy amount of > updates in place on a reasonably large table(100k-500k > rows, ~7 columns mostly integers/bigint). Each row > will be refreshed every 15 minutes, or at least that's > how fast I can poll via SNMP. I hope I can tune the > DB to keep up. > > The app is threaded and will likely have well over 100 > concurrent db connections. Temp tables for storage > aren't a preferred option since this is designed to be > a shared nothing approach and I will likely have > several polling processes. Somewhat OT, but.. The easiest way to speed that up is to use less threads. You're adding a whole TON of overhead with that many threads that you just don't want or need. You should probably be using something event- driven to solve this problem, with just a few database threads to store all that state. Less is definitely more in this case. See (and there's plenty of other literature out there saying that event driven is an extremely good way to do this sort of thing). Here are some frameworks to look at for this kind of network code: (Python) Twisted - (Perl) POE - (Java) java.nio (not familiar enough with the Java thing to know whether or not there's a high-level wrapper) (C++) ACE - (Ruby) IO::Reactor - (C) libevent - .. and of course, you have select/poll/kqueue/WaitNextEvent/whatever that you could use directly, if you wanted to roll your own solution, but don't do that. If you don't want to optimize the whole application, I'd at least just push the DB operations down to a very small number of connections (*one* might even be optimal!), waiting on some kind of thread-safe queue for updates from the rest of the system. This way you can easily batch those updates into transactions and you won't be putting so much unnecessary synchronization overhead into your application and the database. Generally, once you have more worker threads (or processes) than CPUs, you're going to get diminishing returns in a bad way, assuming those threads are making good use of their time. -bob From pgsql-performance-owner@postgresql.org Fri Aug 19 07:15:06 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0D7545291A for ; Fri, 19 Aug 2005 07:14:57 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 07604-01 for ; Fri, 19 Aug 2005 10:14:55 +0000 (GMT) Received: from smtp106.mail.sc5.yahoo.com (smtp106.mail.sc5.yahoo.com [66.163.169.226]) by svr1.postgresql.org (Postfix) with SMTP id A5150528C3 for ; Fri, 19 Aug 2005 07:14:53 -0300 (ADT) Received: (qmail 2178 invoked from network); 19 Aug 2005 10:14:56 -0000 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Received:User-Agent:Date:Subject:From:To:CC:Message-ID:In-Reply-To:Mime-version:Content-type:Content-transfer-encoding; b=lyv/h+YDXeUfE66KEq88++EvoTA4z3Ep96c+jUOoujdvTuJXRREDskLQyFb9wotC68frqppIejCjs8PS9xDdB0HzLgqfxm0bAVo3vIvjuGcoLG21yKo6Uy/MJgu3WL7au3faEefHlmqDSGLf+k7uh/efgQltKn2jdRmCTa6hZwk= ; Received: from unknown (HELO ?192.168.0.103?) (mcotner@66.56.49.200 with login) by smtp106.mail.sc5.yahoo.com with SMTP; 19 Aug 2005 10:14:55 -0000 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Fri, 19 Aug 2005 06:14:54 -0400 Subject: Re: sustained update load of 1-2k/sec From: Mark Cotner To: Bob Ippolito Cc: Message-ID: In-Reply-To: <67788A4A-4011-49AB-B329-683FD9532661@redivi.com> Mime-version: 1.0 Content-type: text/plain; charset="US-ASCII" Content-transfer-encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.28 required=5 tests=[AWL=-0.093, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/252 X-Sequence-Number: 13999 Excellent feedback. Thank you. Please do keep in mind I'm storing the results of SNMP queries. The majority of the time each thread is in a wait state, listening on a UDP port for return packet. The number of threads is high because in order to sustain poll speed I need to minimize the impact of timeouts and all this waiting for return packets. I had intended to have a fallback plan which would build a thread safe queue for db stuffs, but the application isn't currently architected that way. It's not completely built yet so now is the time for change. I hadn't thought of building up a batch of queries and creating a transaction from them. I've been looking into memcached as a persistent object store as well and hadn't seen the reactor pattern yet. Still trying to get my puny brain around that one. Again, thanks for the help. 'njoy, Mark On 8/19/05 5:09 AM, "Bob Ippolito" wrote: > > On Aug 18, 2005, at 10:24 PM, Mark Cotner wrote: > >> I'm currently working on an application that will poll >> thousands of cable modems per minute and I would like >> to use PostgreSQL to maintain state between polls of >> each device. This requires a very heavy amount of >> updates in place on a reasonably large table(100k-500k >> rows, ~7 columns mostly integers/bigint). Each row >> will be refreshed every 15 minutes, or at least that's >> how fast I can poll via SNMP. I hope I can tune the >> DB to keep up. >> >> The app is threaded and will likely have well over 100 >> concurrent db connections. Temp tables for storage >> aren't a preferred option since this is designed to be >> a shared nothing approach and I will likely have >> several polling processes. > > Somewhat OT, but.. > > The easiest way to speed that up is to use less threads. You're > adding a whole TON of overhead with that many threads that you just > don't want or need. You should probably be using something event- > driven to solve this problem, with just a few database threads to > store all that state. Less is definitely more in this case. See > (and there's plenty of other > literature out there saying that event driven is an extremely good > way to do this sort of thing). > > Here are some frameworks to look at for this kind of network code: > (Python) Twisted - > (Perl) POE - > (Java) java.nio (not familiar enough with the Java thing to know > whether or not there's a high-level wrapper) > (C++) ACE - > (Ruby) IO::Reactor - > (C) libevent - > > .. and of course, you have select/poll/kqueue/WaitNextEvent/whatever > that you could use directly, if you wanted to roll your own solution, > but don't do that. > > If you don't want to optimize the whole application, I'd at least > just push the DB operations down to a very small number of > connections (*one* might even be optimal!), waiting on some kind of > thread-safe queue for updates from the rest of the system. This way > you can easily batch those updates into transactions and you won't be > putting so much unnecessary synchronization overhead into your > application and the database. > > Generally, once you have more worker threads (or processes) than > CPUs, you're going to get diminishing returns in a bad way, assuming > those threads are making good use of their time. > > -bob > From pgsql-performance-owner@postgresql.org Fri Aug 19 08:28:33 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A4E8A529EF for ; Fri, 19 Aug 2005 08:28:31 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 76808-07 for ; Fri, 19 Aug 2005 11:28:27 +0000 (GMT) Received: from redivi.com (redivi.com [64.207.133.54]) by svr1.postgresql.org (Postfix) with ESMTP id A516C529EE for ; Fri, 19 Aug 2005 08:28:27 -0300 (ADT) Received: from [64.203.23.17] (helo=[192.168.0.8]) by redivi.com with esmtpsa (TLSv1:RC4-SHA:128) (Exim 4.43) id 1E652x-0001el-CE; Fri, 19 Aug 2005 04:28:30 -0700 In-Reply-To: References: Mime-Version: 1.0 (Apple Message framework v733) Message-Id: <9852B15E-8F9D-46F5-B4B4-9EAAE26F1AF7@redivi.com> Cc: From: Bob Ippolito Date: Fri, 19 Aug 2005 01:28:07 -1000 To: Mark Cotner X-Mailer: Apple Mail (2.733) X-SA-Exim-Connect-IP: 64.203.23.17 X-SA-Exim-Mail-From: bob@redivi.com Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: sustained update load of 1-2k/sec X-SA-Exim-Version: 4.1+cvs (built Mon, 23 Aug 2004 08:44:05 -0700) X-SA-Exim-Scanned: Yes (on redivi.com) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/253 X-Sequence-Number: 14000 On Aug 19, 2005, at 12:14 AM, Mark Cotner wrote: > Excellent feedback. Thank you. Please do keep in mind I'm storing > the > results of SNMP queries. The majority of the time each thread is > in a wait > state, listening on a UDP port for return packet. The number of > threads is > high because in order to sustain poll speed I need to minimize the > impact of > timeouts and all this waiting for return packets. Asynchronous IO via select/poll/etc. basically says: "given these 100 sockets, wake me up when any of them has something to tell me, or wake me up anyway in N milliseconds". From one thread, you can usually deal with thousands of connections without breaking a sweat, where with thread-per-connection you have so much overhead just for the threads that you probably run out of RAM before your network is throttled. The reactor pattern basically just abstracts this a bit so that you worry about what do to when the sockets have something to say, and also allow you to schedule timed events, rather than having to worry about how to implement that correctly *and* write your application. With 100 threads you are basically invoking a special-case of the same mechanism that only looks at one socket, but this makes for 100 different data structures that end up in both userspace and kernel space, plus the thread stacks (which can easily be a few megs each) and context switching when any of them wakes up.. You're throwing a lot of RAM and CPU cycles out the window by using this design. Also, preemptive threads are hard. > I had intended to have a fallback plan which would build a thread > safe queue > for db stuffs, but the application isn't currently architected that > way. > It's not completely built yet so now is the time for change. I hadn't > thought of building up a batch of queries and creating a > transaction from > them. It should be *really* easy to just swap out the implementation of your "change this record" function with one that simply puts its arguments on a queue, with another thread that gets them from the queue and actually does the work. > I've been looking into memcached as a persistent object store as > well and > hadn't seen the reactor pattern yet. Still trying to get my puny > brain > around that one. memcached is RAM based, it's not persistent at all... unless you are sure all of your nodes will be up at all times and will never go down. IIRC, it also just starts throwing away data once you hit its size limit. If course, this isn't really any different than MySQL's MyISAM tables if you hit the row limit, but I think that memcached might not even give you an error when this happens. Also, memcached is just key/value pairs over a network, not much of a database going on there. If you can fit all this data in RAM and you don't care so much about the integrity, you might not benefit much from a RDBMS at all. However, I don't really know what you're doing with the data once you have it so I might be very wrong here... -bob > > Again, thanks for the help. > > 'njoy, > Mark > > > On 8/19/05 5:09 AM, "Bob Ippolito" wrote: > > >> >> On Aug 18, 2005, at 10:24 PM, Mark Cotner wrote: >> >> >>> I'm currently working on an application that will poll >>> thousands of cable modems per minute and I would like >>> to use PostgreSQL to maintain state between polls of >>> each device. This requires a very heavy amount of >>> updates in place on a reasonably large table(100k-500k >>> rows, ~7 columns mostly integers/bigint). Each row >>> will be refreshed every 15 minutes, or at least that's >>> how fast I can poll via SNMP. I hope I can tune the >>> DB to keep up. >>> >>> The app is threaded and will likely have well over 100 >>> concurrent db connections. Temp tables for storage >>> aren't a preferred option since this is designed to be >>> a shared nothing approach and I will likely have >>> several polling processes. >>> >> >> Somewhat OT, but.. >> >> The easiest way to speed that up is to use less threads. You're >> adding a whole TON of overhead with that many threads that you just >> don't want or need. You should probably be using something event- >> driven to solve this problem, with just a few database threads to >> store all that state. Less is definitely more in this case. See >> (and there's plenty of other >> literature out there saying that event driven is an extremely good >> way to do this sort of thing). >> >> Here are some frameworks to look at for this kind of network code: >> (Python) Twisted - >> (Perl) POE - >> (Java) java.nio (not familiar enough with the Java thing to know >> whether or not there's a high-level wrapper) >> (C++) ACE - >> (Ruby) IO::Reactor - >> (C) libevent - >> >> .. and of course, you have select/poll/kqueue/WaitNextEvent/whatever >> that you could use directly, if you wanted to roll your own solution, >> but don't do that. >> >> If you don't want to optimize the whole application, I'd at least >> just push the DB operations down to a very small number of >> connections (*one* might even be optimal!), waiting on some kind of >> thread-safe queue for updates from the rest of the system. This way >> you can easily batch those updates into transactions and you won't be >> putting so much unnecessary synchronization overhead into your >> application and the database. >> >> Generally, once you have more worker threads (or processes) than >> CPUs, you're going to get diminishing returns in a bad way, assuming >> those threads are making good use of their time. >> >> -bob >> >> > > > > ---------------------------(end of > broadcast)--------------------------- > TIP 5: don't forget to increase your free space map settings > From pgsql-performance-owner@postgresql.org Fri Aug 19 08:36:47 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1E826529F5 for ; Fri, 19 Aug 2005 08:36:46 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 95556-10 for ; Fri, 19 Aug 2005 11:36:42 +0000 (GMT) Received: from purple.bdb.fi (purple.bdb.fi [195.197.212.62]) by svr1.postgresql.org (Postfix) with ESMTP id C77F452899 for ; Fri, 19 Aug 2005 08:36:38 -0300 (ADT) Received: by purple.bdb.fi (Postfix, from userid 101) id 58CDE31D2; Fri, 19 Aug 2005 14:34:47 +0300 (EETDST) Received: from localhost (localhost [127.0.0.1]) by purple.bdb.fi (Postfix) with ESMTP id 4D2A1292D; Fri, 19 Aug 2005 14:34:47 +0300 (EETDST) Date: Fri, 19 Aug 2005 14:34:47 +0300 (EETDST) From: Kari Lavikka To: Tom Lane Cc: pgsql-performance@postgresql.org, rjpeace@earthlink.net Subject: Re: Finding bottleneck In-Reply-To: <29384.1123520200@sss.pgh.pa.us> Message-ID: References: <6EE64EF3AB31D5448D0007DD34EEB3417DD026@Herge.rcsinc.local> <29384.1123520200@sss.pgh.pa.us> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/254 X-Sequence-Number: 14001 On Mon, 8 Aug 2005, Tom Lane wrote: > What that sounds like to me is a machine with inadequate disk I/O bandwidth. > Your earlier comment that checkpoint drives the machine into the ground > fits right into that theory, too. You said there is "almost no IO-wait" > but are you sure you are measuring that correctly? Reducing checkpoint_timeout to 600 seconds had a positive effect. Previous value was 1800 seconds. We have a spare disk array from the old server and I'm planning to use it as a tablespace for the comment table (the 100M+ rows one) as Ron suggested. >> Queries accumulate and when checkpointing is over, there can be >> something like 400 queries running but over 50% of cpu is just idling. > > 400 queries? Are you launching 400 separate backends to do that? > Some sort of connection pooling seems like a good idea, if you don't > have it in place already. If the system's effective behavior in the > face of heavy load is to start even more concurrent backends, that > could easily drive things into the ground. Ok, I implemented connection pooling using pgpool and it increased performance a lot! We are now delivering about 1500 dynamic pages a second without problems. Each of the eight single-cpu webservers are running a pgpool instance with 20 connections. However, those configuration changes didn't have significant effect to oprofile results. AtEOXact_CatCache consumes even more cycles. This isn't a problem right now but it may be in the future... CPU: AMD64 processors, speed 2190.23 MHz (estimated) Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit mask of 0x00 (No unit mask) count 100000 samples % symbol name 1147870 21.1602 AtEOXact_CatCache 187466 3.4558 hash_seq_search 174357 3.2142 AllocSetAlloc 170896 3.1504 nocachegetattr 131724 2.4282 ExecMakeFunctionResultNoSets 125292 2.3097 SearchCatCache 117264 2.1617 StrategyDirtyBufferList 105741 1.9493 hash_search 98245 1.8111 FunctionCall2 97878 1.8043 yyparse 90932 1.6763 LWLockAcquire 83555 1.5403 LWLockRelease 81045 1.4940 _bt_compare ... and so on ... ----->8 Signigicant rows from current postgresql.conf 8<----- max_connections = 768 # unnecessarily large with connection pooling shared_buffers = 15000 work_mem = 2048 maintenance_work_mem = 32768 max_fsm_pages = 1000000 max_fsm_relations = 5000 bgwriter_percent = 2 fsync = true wal_buffers = 512 checkpoint_segments = 200 # less would probably be enuff with 600sec timeout checkpoint_timeout = 600 effective_cache_size = 500000 random_page_cost = 1.5 default_statistics_target = 150 stats_start_collector = true stats_command_string = true |\__/| ( oo ) Kari Lavikka - tuner@bdb.fi - (050) 380 3808 __ooO( )Ooo_______ _____ ___ _ _ _ _ _ _ _ "" From pgsql-performance-owner@postgresql.org Fri Aug 19 09:50:14 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B81CA52C1F for ; Fri, 19 Aug 2005 09:50:09 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 80161-04 for ; Fri, 19 Aug 2005 12:50:03 +0000 (GMT) Received: from floppy.pyrenet.fr (news.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id C1E8952B3D for ; Fri, 19 Aug 2005 09:50:02 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id 85A9530B42; Fri, 19 Aug 2005 15:01:30 +0200 (MET DST) From: Christopher Browne X-Newsgroups: pgsql.performance Subject: Re: Need for speed Date: Fri, 19 Aug 2005 08:00:26 -0400 Organization: cbbrowne Computing Inc Lines: 33 Message-ID: References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Complaints-To: usenet@news.hub.org X-message-flag: Outlook is rather hackable, isn't it? X-Home-Page: http://www.cbbrowne.com/info/ X-Affero: http://svcs.affero.net/rm.php?r=cbbrowne User-Agent: Gnus/5.1006 (Gnus v5.10.6) XEmacs/21.4 (Jumbo Shrimp, berkeley-unix) Cancel-Lock: sha1:2WATvvUFlUBNDukomBEVrYVzI0c= To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.116 required=5 tests=[AWL=0.116] X-Spam-Level: X-Archive-Number: 200508/256 X-Sequence-Number: 14003 >> Ulrich Wisser wrote: >> > >> > one of our services is click counting for on line advertising. We do >> > this by importing Apache log files every five minutes. This results in a >> > lot of insert and delete statements. > ... >> If you are doing mostly inserting, make sure you are in a transaction, > > Well, yes, but you may need to make sure that a single transaction > doesn't have too many inserts in it. I was having a performance > problem when doing transactions with a huge number of inserts (tens > of thousands), and I solved the problem by putting a simple counter > in the loop (in the Java import code, that is) and doing a commit > every 100 or so inserts. Are you sure that was an issue with PostgreSQL? I have certainly observed that issue with Oracle, but NOT with PostgreSQL. I have commonly done data loads where they loaded 50K rows at a time, the reason for COMMITting at that point being "programming paranoia" at the possibility that some data might fail to load and need to be retried, and I'd rather have less fail... It would seem more likely that the issue would be on the Java side; it might well be that the data being loaded might bloat JVM memory usage, and that the actions taken at COMMIT time might keep the size of the Java-side memory footprint down. -- (reverse (concatenate 'string "moc.liamg" "@" "enworbbc")) http://cbbrowne.com/info/ If we were meant to fly, we wouldn't keep losing our luggage. From pgsql-performance-owner@postgresql.org Fri Aug 19 09:41:08 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A14C552BCA for ; Fri, 19 Aug 2005 09:40:57 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 66529-05 for ; Fri, 19 Aug 2005 12:40:46 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.199]) by svr1.postgresql.org (Postfix) with ESMTP id A4BFB52E45 for ; Fri, 19 Aug 2005 09:40:38 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i22so507916wra for ; Fri, 19 Aug 2005 05:40:39 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=akHVlRYJ02H0PDPuDv/e84h115C5Zk6FwUE3USQDOkBsM4YcNMf88fIhmrd2Zvhoz9maRWiLsqbwx2/CMcITT2t7sp8wDx/A61nCNtwTbf+Wxc/gTNgnxZDMSNwRYZvk1Kn/qR43yl03/X1YCwt6y9Rw4PjEnCG8IPiEj0Io8L0= Received: by 10.54.32.60 with SMTP id f60mr1248568wrf; Fri, 19 Aug 2005 05:40:39 -0700 (PDT) Received: by 10.54.86.15 with HTTP; Fri, 19 Aug 2005 05:40:39 -0700 (PDT) Message-ID: <33c6269f050819054038af9886@mail.gmail.com> Date: Fri, 19 Aug 2005 08:40:39 -0400 From: Alex Turner To: Mark Cotner Subject: Re: sustained update load of 1-2k/sec Cc: pgsql-performance@postgresql.org In-Reply-To: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.104 required=5 tests=[AWL=0.080, RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/255 X-Sequence-Number: 14002 I have managed tx speeds that high from postgresql going even as high as 2500/sec for small tables, but it does require a good RAID controler card (yes I'm even running with fsync on). I'm using 3ware 9500S-8MI with Raptor drives in multiple RAID 10s. The box wasn't too $$$ at just around $7k. I have two independant controlers on two independant PCI buses to give max throughput. on with a 6 drive RAID 10 and the other with two 4 drive RAID 10s. Alex Turner NetEconomist On 8/19/05, Mark Cotner wrote: > Hi all, > I bet you get tired of the same ole questions over and > over. >=20 > I'm currently working on an application that will poll > thousands of cable modems per minute and I would like > to use PostgreSQL to maintain state between polls of > each device. This requires a very heavy amount of > updates in place on a reasonably large table(100k-500k > rows, ~7 columns mostly integers/bigint). Each row > will be refreshed every 15 minutes, or at least that's > how fast I can poll via SNMP. I hope I can tune the > DB to keep up. >=20 > The app is threaded and will likely have well over 100 > concurrent db connections. Temp tables for storage > aren't a preferred option since this is designed to be > a shared nothing approach and I will likely have > several polling processes. >=20 > Here are some of my assumptions so far . . . >=20 > HUGE WAL > Vacuum hourly if not more often >=20 > I'm getting 1700tx/sec from MySQL and I would REALLY > prefer to use PG. I don't need to match the number, > just get close. >=20 > Is there a global temp table option? In memory tables > would be very beneficial in this case. I could just > flush it to disk occasionally with an insert into blah > select from memory table. >=20 > Any help or creative alternatives would be greatly > appreciated. :) >=20 > 'njoy, > Mark >=20 >=20 > -- > Writing software requires an intelligent person, > creating functional art requires an artist. > -- Unknown >=20 >=20 > ---------------------------(end of broadcast)--------------------------- > TIP 3: Have you checked our extensive FAQ? >=20 > http://www.postgresql.org/docs/faq > From pgsql-performance-owner@postgresql.org Fri Aug 19 10:06:05 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BC10752E60 for ; Fri, 19 Aug 2005 10:05:48 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 24697-07 for ; Fri, 19 Aug 2005 13:05:46 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 12D4552E65 for ; Fri, 19 Aug 2005 10:05:45 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7JD5loZ017465; Fri, 19 Aug 2005 09:05:48 -0400 (EDT) To: Kari Lavikka Cc: pgsql-performance@postgresql.org, rjpeace@earthlink.net Subject: Re: Finding bottleneck In-reply-to: References: <6EE64EF3AB31D5448D0007DD34EEB3417DD026@Herge.rcsinc.local> <29384.1123520200@sss.pgh.pa.us> Comments: In-reply-to Kari Lavikka message dated "Fri, 19 Aug 2005 14:34:47 +0300" Date: Fri, 19 Aug 2005 09:05:47 -0400 Message-ID: <17464.1124456747@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/257 X-Sequence-Number: 14004 Kari Lavikka writes: > However, those configuration changes didn't have significant effect to > oprofile results. AtEOXact_CatCache consumes even more cycles. I believe I've fixed that for 8.1. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Aug 19 10:37:05 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B76BE52E68 for ; Fri, 19 Aug 2005 10:35:58 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 77688-07 for ; Fri, 19 Aug 2005 13:35:51 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 92E7B52E1E for ; Fri, 19 Aug 2005 10:35:49 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7JDZro9017691; Fri, 19 Aug 2005 09:35:53 -0400 (EDT) To: Bob Ippolito Cc: Mark Cotner , pgsql-performance@postgresql.org Subject: Re: sustained update load of 1-2k/sec In-reply-to: <67788A4A-4011-49AB-B329-683FD9532661@redivi.com> References: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> <67788A4A-4011-49AB-B329-683FD9532661@redivi.com> Comments: In-reply-to Bob Ippolito message dated "Thu, 18 Aug 2005 23:09:27 -1000" Date: Fri, 19 Aug 2005 09:35:53 -0400 Message-ID: <17690.1124458553@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/258 X-Sequence-Number: 14005 Bob Ippolito writes: > If you don't want to optimize the whole application, I'd at least > just push the DB operations down to a very small number of > connections (*one* might even be optimal!), waiting on some kind of > thread-safe queue for updates from the rest of the system. While I agree that hundreds of threads seems like overkill, I think the above advice might be going too far in the other direction. The problem with single-threaded operation is that any delay affects the whole system --- eg, if you're blocked waiting for disk I/O, the CPU doesn't get anything done either. You want enough DB connections doing things in parallel to make sure that there's always something else useful to do for each major component. This is particularly important for Postgres, which doesn't do any internal query parallelization (not that it would help much anyway for the sorts of trivial queries you are worried about). If you have, say, a 4-way CPU you want at least 4 active connections to make good use of the CPUs. I'd suggest trying to build the system so that it uses a dozen or two active database connections. If that doesn't match up to the number of polling activities you want to have in flight at any instant, then you can do something like what Bob suggested on the client side to bridge the gap. As far as the question "can PG do 1-2k xact/sec", the answer is "yes if you throw enough hardware at it". Spending enough money on the disk subsystem is the key ... regards, tom lane From pgsql-performance-owner@postgresql.org Fri Aug 19 10:37:37 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C21FE52E14 for ; Fri, 19 Aug 2005 10:37:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 82454-05 for ; Fri, 19 Aug 2005 13:37:22 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id 029E152E07 for ; Fri, 19 Aug 2005 10:37:16 -0300 (ADT) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: Finding bottleneck Date: Fri, 19 Aug 2005 09:37:19 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3417DD150@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Finding bottleneck Thread-Index: AcWkvxCeRZU6M6WETVCl/asOphrnNwAAjKuw From: "Merlin Moncure" To: "Tom Lane" Cc: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/259 X-Sequence-Number: 14006 > Kari Lavikka writes: > > However, those configuration changes didn't have significant effect to > > oprofile results. AtEOXact_CatCache consumes even more cycles. >=20 > I believe I've fixed that for 8.1. Relative to 8.0, I am seeing a dramatic, almost miraculous reduction in CPU load times in 8.1devel. This is for ISAM style access patterns over the parse/bind interface. (IOW one record at a time, 90% read, 10% write). Relative to commercial dedicated ISAM storage engines, pg holds up very well except in cpu load, but 8.1 is a huge step towards addressing that. So far, except for one minor (and completely understandable) issue with bitmap issues, 8.1 has been a stellar performer. Also great is the expansion of pg_locks view (which I didn't see mentioned in Bruce's TODO list, just FYI). Merlin From pgsql-performance-owner@postgresql.org Fri Aug 19 10:58:45 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3E19152C34 for ; Fri, 19 Aug 2005 10:58:42 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 18998-05 for ; Fri, 19 Aug 2005 13:58:37 +0000 (GMT) Received: from moutng.kundenserver.de (moutng.kundenserver.de [212.227.126.188]) by svr1.postgresql.org (Postfix) with ESMTP id C64FD52C1F for ; Fri, 19 Aug 2005 10:58:35 -0300 (ADT) Received: from p548F1B85.dip0.t-ipconnect.de [84.143.27.133] (helo=pse.dyndns.org) by mrelayeu.kundenserver.de with ESMTP (Nemesis), id 0MKwh2-1E67OH3mTv-00034N; Fri, 19 Aug 2005 15:58:37 +0200 Received: from p548f1e1e.dip0.t-ipconnect.de ([84.143.30.30] helo=[10.10.11.120]) by pse.dyndns.org with esmtpsa (TLSv1:AES256-SHA:256) (Exim 4.44) id 1E67OF-0000dE-Ht; Fri, 19 Aug 2005 15:58:36 +0200 Message-ID: <4305E57E.1010604@pse-consulting.de> Date: Fri, 19 Aug 2005 15:58:22 +0200 From: Andreas Pflug User-Agent: Mozilla Thunderbird 1.0.6 (Windows/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: Bob Ippolito , Mark Cotner , pgsql-performance@postgresql.org Subject: Re: sustained update load of 1-2k/sec References: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> <67788A4A-4011-49AB-B329-683FD9532661@redivi.com> <17690.1124458553@sss.pgh.pa.us> In-Reply-To: <17690.1124458553@sss.pgh.pa.us> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Provags-ID: kundenserver.de abuse@kundenserver.de login:0ce7ee5c3478b8d72edd8e05ccd40b70 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.073 required=5 tests=[AWL=-0.769, FORGED_RCVD_HELO=0.05, RCVD_IN_NJABL_DUL=1.655, RCVD_IN_SORBS_DUL=0.137] X-Spam-Level: * X-Archive-Number: 200508/260 X-Sequence-Number: 14007 Tom Lane wrote: >Bob Ippolito writes: > > >>If you don't want to optimize the whole application, I'd at least >>just push the DB operations down to a very small number of >>connections (*one* might even be optimal!), waiting on some kind of >>thread-safe queue for updates from the rest of the system. >> >> > >While I agree that hundreds of threads seems like overkill, I think the >above advice might be going too far in the other direction. The problem >with single-threaded operation is that any delay affects the whole >system --- eg, if you're blocked waiting for disk I/O, the CPU doesn't >get anything done either. You want enough DB connections doing things >in parallel to make sure that there's always something else useful to do >for each major component. This is particularly important for Postgres, >which doesn't do any internal query parallelization (not that it would >help much anyway for the sorts of trivial queries you are worried about). >If you have, say, a 4-way CPU you want at least 4 active connections to >make good use of the CPUs. > >I'd suggest trying to build the system so that it uses a dozen or two >active database connections. If that doesn't match up to the number of >polling activities you want to have in flight at any instant, then you >can do something like what Bob suggested on the client side to bridge >the gap. > >As far as the question "can PG do 1-2k xact/sec", the answer is "yes >if you throw enough hardware at it". Spending enough money on the >disk subsystem is the key ... > > The 1-2k xact/sec for MySQL seems suspicious, sounds very much like write-back cached, not write-through, esp. considering that heavy concurrent write access isn't said to be MySQLs strength... I wonder if preserving the database after a fatal crash is really necessary, since the data stored sounds quite volatile; in this case, fsync=false might be sufficient. Regards, Andreas From pgsql-performance-owner@postgresql.org Fri Aug 19 11:03:13 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 08F0F52C19 for ; Fri, 19 Aug 2005 11:03:11 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 38325-01 for ; Fri, 19 Aug 2005 14:03:09 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id A901A52BE2 for ; Fri, 19 Aug 2005 11:03:08 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7JE3D3K017940; Fri, 19 Aug 2005 10:03:13 -0400 (EDT) To: "Merlin Moncure" Cc: pgsql-performance@postgresql.org Subject: Re: Finding bottleneck In-reply-to: <6EE64EF3AB31D5448D0007DD34EEB3417DD150@Herge.rcsinc.local> References: <6EE64EF3AB31D5448D0007DD34EEB3417DD150@Herge.rcsinc.local> Comments: In-reply-to "Merlin Moncure" message dated "Fri, 19 Aug 2005 09:37:19 -0400" Date: Fri, 19 Aug 2005 10:03:12 -0400 Message-ID: <17939.1124460192@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/261 X-Sequence-Number: 14008 "Merlin Moncure" writes: > Relative to 8.0, I am seeing a dramatic, almost miraculous reduction in > CPU load times in 8.1devel. This is for ISAM style access patterns over > the parse/bind interface. (IOW one record at a time, 90% read, 10% > write). > Relative to commercial dedicated ISAM storage engines, pg holds up very > well except in cpu load, but 8.1 is a huge step towards addressing that. Cool --- we've done a fair amount of work on squeezing out internal inefficiencies during this devel cycle, but it's always hard to predict just how much anyone will notice in the real world. Care to do some oprofile or gprof profiles to see where it's still bad? regards, tom lane From pgsql-performance-owner@postgresql.org Fri Aug 19 11:09:07 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 89D3352BCE for ; Fri, 19 Aug 2005 11:09:05 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 40276-04 for ; Fri, 19 Aug 2005 14:09:03 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 681E252AD6 for ; Fri, 19 Aug 2005 11:09:03 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7JE96GO018001; Fri, 19 Aug 2005 10:09:06 -0400 (EDT) To: Andreas Pflug Cc: Bob Ippolito , Mark Cotner , pgsql-performance@postgresql.org Subject: Re: sustained update load of 1-2k/sec In-reply-to: <4305E57E.1010604@pse-consulting.de> References: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> <67788A4A-4011-49AB-B329-683FD9532661@redivi.com> <17690.1124458553@sss.pgh.pa.us> <4305E57E.1010604@pse-consulting.de> Comments: In-reply-to Andreas Pflug message dated "Fri, 19 Aug 2005 15:58:22 +0200" Date: Fri, 19 Aug 2005 10:09:05 -0400 Message-ID: <18000.1124460545@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/262 X-Sequence-Number: 14009 Andreas Pflug writes: > Tom Lane wrote: >> As far as the question "can PG do 1-2k xact/sec", the answer is "yes >> if you throw enough hardware at it". Spending enough money on the >> disk subsystem is the key ... >> > The 1-2k xact/sec for MySQL seems suspicious, sounds very much like > write-back cached, not write-through, esp. considering that heavy > concurrent write access isn't said to be MySQLs strength... > I wonder if preserving the database after a fatal crash is really > necessary, since the data stored sounds quite volatile; in this case, > fsync=false might be sufficient. Yeah, that's something to think about. If you do need full transaction safety, then you *must* have a decent battery-backed-write-cache setup, else your transaction commit rate will be limited by disk rotation speed --- for instance, a single connection can commit at most 250 xacts per second if the WAL log is on a 15000RPM drive. (You can improve this to the extent that you can spread activity across multiple connections, but I'm not sure you can expect to reliably have 8 or more connections ready to commit each time the disk goes 'round.) regards, tom lane From pgsql-performance-owner@postgresql.org Fri Aug 19 11:55:30 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1E20752E0C for ; Fri, 19 Aug 2005 11:55:04 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 31597-04 for ; Fri, 19 Aug 2005 14:54:58 +0000 (GMT) Received: from smtpauth08.mail.atl.earthlink.net (smtpauth08.mail.atl.earthlink.net [209.86.89.68]) by svr1.postgresql.org (Postfix) with ESMTP id 1F99652A2B for ; Fri, 19 Aug 2005 11:54:57 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=FqVgtOHasABKey78n2Y8/jE2SspFW36HGI69f1rJDsyfgMUGA+lFzJZ/BOEqDf45; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth08.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E68Gs-0005PC-0O; Fri, 19 Aug 2005 10:55:02 -0400 Message-Id: <6.2.3.4.0.20050819093011.05c1b9d8@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Fri, 19 Aug 2005 10:54:57 -0400 To: postgres performance From: Ron Subject: Re: sustained update load of 1-2k/sec In-Reply-To: <33c6269f050819054038af9886@mail.gmail.com> References: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> <33c6269f050819054038af9886@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bc8e05e127e111a9d4886337fbf3f87140350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.762 required=5 tests=[AWL=0.388, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/263 X-Sequence-Number: 14010 Alex mentions a nice setup, but I'm pretty sure I know how to beat that IO subsystems HW's performance by at least 1.5x or 2x. Possibly more. (No, I do NOT work for any vendor I'm about to discuss.) Start by replacing the WD Raptors with Maxtor Atlas 15K II's. At 5.5ms average access, 97.4MB/s outer track throughput, 85.9MB/s average, and 74.4 MB/s inner track throughput, they have the best performance characteristics of any tested shipping HDs I know of. (Supposedly the new SAS versions will _sustain_ ~98MB/s, but I'll believe that only if I see it under independent testing). In comparison, the numbers on the WD740GD are 8.1ms average access, 71.8, 62.9, and 53.9 MB/s outer, average and inner track throughputs respectively. Be prepared to use as many of them as possible (read: as many you can afford) if you want to maximize transaction rates, particularly for small transactions like this application seems to be mentioning. Next, use a better RAID card. The TOL enterprise stuff (Xyratex, Engino, Dot-hill) is probably too expensive, but in the commodity market benchmarks indicate that that Areca's 1GB buffer RAID cards currently outperform all the other commodity RAID stuff. 9 Atlas II's per card in a RAID 5 set, or 16 per card in a RAID 10 set, should max the RAID card's throughput and come very close to, if not attaining, the real world peak bandwidth of the 64b 133MHz PCI-X bus they are plugged into. Say somewhere in the 700-800MB/s range. Repeat the above for as many independent PCI-X buses as you have for a very fast commodity RAID IO subsystem. Two such configured cards used in the dame manner as mentioned by Alex should easily attain 1.5x - 2x the transaction numbers mentioned by Alex unless there's a bottleneck somewhere else in the system design. Hope this helps, Ron Peacetree At 08:40 AM 8/19/2005, Alex Turner wrote: >I have managed tx speeds that high from postgresql going even as high >as 2500/sec for small tables, but it does require a good RAID >controler card (yes I'm even running with fsync on). I'm using 3ware >9500S-8MI with Raptor drives in multiple RAID 10s. The box wasn't too >$$$ at just around $7k. I have two independant controlers on two >independant PCI buses to give max throughput. on with a 6 drive RAID >10 and the other with two 4 drive RAID 10s. > >Alex Turner >NetEconomist > >On 8/19/05, Mark Cotner wrote: > > Hi all, > > I bet you get tired of the same ole questions over and > > over. > > > > I'm currently working on an application that will poll > > thousands of cable modems per minute and I would like > > to use PostgreSQL to maintain state between polls of > > each device. This requires a very heavy amount of > > updates in place on a reasonably large table(100k-500k > > rows, ~7 columns mostly integers/bigint). Each row > > will be refreshed every 15 minutes, or at least that's > > how fast I can poll via SNMP. I hope I can tune the > > DB to keep up. > > > > The app is threaded and will likely have well over 100 > > concurrent db connections. Temp tables for storage > > aren't a preferred option since this is designed to be > > a shared nothing approach and I will likely have > > several polling processes. > > > > Here are some of my assumptions so far . . . > > > > HUGE WAL > > Vacuum hourly if not more often > > > > I'm getting 1700tx/sec from MySQL and I would REALLY > > prefer to use PG. I don't need to match the number, > > just get close. > > > > Is there a global temp table option? In memory tables > > would be very beneficial in this case. I could just > > flush it to disk occasionally with an insert into blah > > select from memory table. > > > > Any help or creative alternatives would be greatly > > appreciated. :) > > > > 'njoy, > > Mark > > > > > > -- > > Writing software requires an intelligent person, > > creating functional art requires an artist. > > -- Unknown > > > > > > ---------------------------(end of broadcast)--------------------------- > > TIP 3: Have you checked our extensive FAQ? > > > > http://www.postgresql.org/docs/faq > > > >---------------------------(end of broadcast)--------------------------- >TIP 2: Don't 'kill -9' the postmaster From pgsql-performance-owner@postgresql.org Fri Aug 19 12:04:13 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B3FB752E2E for ; Fri, 19 Aug 2005 12:04:05 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 44686-08 for ; Fri, 19 Aug 2005 15:04:00 +0000 (GMT) Received: from smtpauth08.mail.atl.earthlink.net (smtpauth08.mail.atl.earthlink.net [209.86.89.68]) by svr1.postgresql.org (Postfix) with ESMTP id 8B81A52E2D for ; Fri, 19 Aug 2005 12:04:00 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=qpS1QwNjCFcGxob02hmALGmzcUbRGYopOoj7fhKy88feLFPT77pYxEN6YamwYhVq; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:Cc:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth08.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E68PX-0003y9-9r; Fri, 19 Aug 2005 11:03:59 -0400 Message-Id: <6.2.3.4.0.20050819105751.05c1a760@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Fri, 19 Aug 2005 11:03:54 -0400 To: Andreas Pflug From: Ron Subject: Re: sustained update load of 1-2k/sec Cc: pgsql-performance@postgresql.org In-Reply-To: <4305E57E.1010604@pse-consulting.de> References: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> <67788A4A-4011-49AB-B329-683FD9532661@redivi.com> <17690.1124458553@sss.pgh.pa.us> <4305E57E.1010604@pse-consulting.de> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bc1a6d27d7b2cbdd61554e775236ab66aa350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.682 required=5 tests=[AWL=0.308, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/264 X-Sequence-Number: 14011 At 09:58 AM 8/19/2005, Andreas Pflug wrote: >The 1-2k xact/sec for MySQL seems suspicious, sounds very much like >write-back cached, not write-through, esp. considering that heavy >concurrent write access isn't said to be MySQLs strength... Don't be suspicious. I haven't seen the code under discussion, but I have seen mySQL easily achieve these kinds of numbers using the myISAM storage engine in write-through cache mode. myISAM can be =FAST=. Particularly when decent HW is thrown at it. Ron From pgsql-performance-owner@postgresql.org Fri Aug 19 12:27:45 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id F136152C3A for ; Fri, 19 Aug 2005 12:22:53 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 81433-02 for ; Fri, 19 Aug 2005 15:22:49 +0000 (GMT) Received: from birao.terra.com.br (birao.terra.com.br [200.176.10.197]) by svr1.postgresql.org (Postfix) with ESMTP id 62A1B52ACB for ; Fri, 19 Aug 2005 12:22:48 -0300 (ADT) Received: from donga.terra.com.br (donga.terra.com.br [200.176.10.10]) by birao.terra.com.br (Postfix) with ESMTP id 22BA0108C0AD for ; Fri, 19 Aug 2005 12:22:47 -0300 (BRT) X-Terra-Karma: -2% X-Terra-Hash: 4cf10e32f89f4380a51ac919858cb02c Received-SPF: none (donga.terra.com.br: 200.176.10.10 is neither permitted nor denied by domain of sistemica.info) client-ip=200.176.10.10; envelope-from=diego@sistemica.info; helo=modem.sistemica; Received: from modem.sistemica (201-25-44-253.paemt704.dsl.brasiltelecom.net.br [201.25.44.253]) (authenticated user sistemic) by donga.terra.com.br (Postfix) with ESMTP id 952BF9C4071 for ; Fri, 19 Aug 2005 12:22:46 -0300 (BRT) Received: from diretoria02 (unknown [192.168.200.61]) by modem.sistemica (Postfix) with SMTP id 0FA2813F610 for ; Fri, 19 Aug 2005 12:24:00 -0300 (BRT) Message-ID: <002a01c5a4d1$d36f6420$3dc8a8c0@diretoria02> Reply-To: "Diego de Lima" From: "Diego de Lima" To: Subject: LEFT JOIN ON vs. LEFT JOIN USING performance Date: Fri, 19 Aug 2005 12:22:35 -0300 Organization: Sistemica.info MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0027_01C5A4B8.AE118A50" X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 6.00.2800.1437 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.146 required=5 tests=[FORGED_RCVD_HELO=0.05, HTML_50_60=0.095, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/266 X-Sequence-Number: 14013 This is a multi-part message in MIME format. ------=_NextPart_000_0027_01C5A4B8.AE118A50 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Hi list, I=B4m using Pg 8.0.3 on Linux FC2. This question may have a very simple answer (I hope), but I=B4m having = lots of trouble solving it, and I counldn=B4t find any other post about = it or anything in the pg docs. I have some very complex select statements on 4 million rows tables. = When using LEFT JOIN ON, some select statements takes about 2 minutes. = When I write exactly the same statement but with LEFT JOIN USING, it = takes only 1 minute. Comparing to Oracle, the same statement takes 1 = minute also, but with LEFT JOIN ON. Sometimes tables have the same column names and I can use LEFT JOIN = USING, but in some other cases I MUST use LEFT JOIN ON, because the = tables have different column names. So my question is: is there a way to make LEFT JOIN ON uses the same = plan of LEFT JOIN USING? Thanks, Diego de Lima ------=_NextPart_000_0027_01C5A4B8.AE118A50 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Hi list,
 
I=B4m using Pg 8.0.3 on Linux = FC2.
 
This question may have a very simple = answer (I=20 hope), but I=B4m having lots of trouble solving it, and I counldn=B4t = find any other=20 post about it or anything in the pg docs.
 
I have some very complex select = statements on 4=20 million rows tables. When using LEFT JOIN ON, some select = statements=20 takes about 2 minutes. When I write exactly the same statement but with = LEFT=20 JOIN USING, it takes only 1 minute. Comparing to Oracle, the same = statement=20 takes 1 minute also, but with LEFT JOIN ON.
 
Sometimes tables have the = same column=20 names and I can use LEFT JOIN USING, but in some other cases I MUST use = LEFT=20 JOIN ON, because the tables have different column names.
 
So my question is: is there a way to = make LEFT JOIN=20 ON uses the same plan of LEFT JOIN USING?
 
Thanks,
 
Diego de Lima
 
 
------=_NextPart_000_0027_01C5A4B8.AE118A50-- From pgsql-performance-owner@postgresql.org Fri Aug 19 12:27:35 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9DA0852969 for ; Fri, 19 Aug 2005 12:26:33 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 81896-07 for ; Fri, 19 Aug 2005 15:26:30 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id 26B0D5288E for ; Fri, 19 Aug 2005 12:26:29 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7JFQQEU031777 for ; Fri, 19 Aug 2005 10:26:28 -0500 Subject: Re: I'm configuraing a new system (Bigish) and need some From: Jeremiah Jahn To: postgres performance In-Reply-To: <1124123097.27881.59.camel@bluejay.goodinassociates.com> References: <1124123097.27881.59.camel@bluejay.goodinassociates.com> Content-Type: text/plain Date: Fri, 19 Aug 2005 10:26:25 -0500 Message-Id: <1124465185.27881.127.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Content-Transfer-Encoding: 7bit Received-SPF: none (penguin.goodinassociates.com: 206.80.71.242 is neither permitted nor denied by domain of cs.earlham.edu) client-ip=206.80.71.242; envelope-from=jeremiah@cs.earlham.edu; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1034/Thu Aug 18 15:07:58 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.033 required=5 tests=[AWL=-0.018, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/265 X-Sequence-Number: 14012 that took a little while to get through the system didn't it. Please ignore. > Ingrate, n.: A man who bites the hand that feeds him, and then complains > of indigestion. -- A free society is one where it is safe to be unpopular. -- Adlai Stevenson From pgsql-performance-owner@postgresql.org Fri Aug 19 12:44:36 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id EABF652EC7 for ; Fri, 19 Aug 2005 12:41:23 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 05428-05 for ; Fri, 19 Aug 2005 15:41:17 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 428B352DD0 for ; Fri, 19 Aug 2005 12:40:50 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7JFeneQ025397; Fri, 19 Aug 2005 11:40:50 -0400 (EDT) To: "Diego de Lima" Cc: pgsql-performance@postgresql.org Subject: Re: LEFT JOIN ON vs. LEFT JOIN USING performance In-reply-to: <002a01c5a4d1$d36f6420$3dc8a8c0@diretoria02> References: <002a01c5a4d1$d36f6420$3dc8a8c0@diretoria02> Comments: In-reply-to "Diego de Lima" message dated "Fri, 19 Aug 2005 12:22:35 -0300" Date: Fri, 19 Aug 2005 11:40:49 -0400 Message-ID: <25396.1124466049@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.246 required=5 tests=[AWL=-0.235, INFO_TLD=0.481] X-Spam-Level: X-Archive-Number: 200508/268 X-Sequence-Number: 14015 "Diego de Lima" writes: > I have some very complex select statements on 4 million rows tables. = > When using LEFT JOIN ON, some select statements takes about 2 minutes. = > When I write exactly the same statement but with LEFT JOIN USING, it = > takes only 1 minute. Could we see details please? Like the table schemas, the query itself, and EXPLAIN ANALYZE results for both cases. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Aug 19 12:44:25 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C7FFB52BD8 for ; Fri, 19 Aug 2005 12:41:32 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 11714-04 for ; Fri, 19 Aug 2005 15:41:24 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id 5464D52B54 for ; Fri, 19 Aug 2005 12:41:06 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050819154105m92009i8pue>; Fri, 19 Aug 2005 15:41:05 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id CA2FB55FBA; Fri, 19 Aug 2005 10:41:04 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id 1527655FAD; Fri, 19 Aug 2005 10:41:00 -0500 (CDT) Message-ID: <4305FD8B.9090803@arbash-meinel.com> Date: Fri, 19 Aug 2005 10:40:59 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Diego de Lima Cc: pgsql-performance@postgresql.org Subject: Re: LEFT JOIN ON vs. LEFT JOIN USING performance References: <002a01c5a4d1$d36f6420$3dc8a8c0@diretoria02> In-Reply-To: <002a01c5a4d1$d36f6420$3dc8a8c0@diretoria02> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigF91D8A06A6C8BD5262A8C6FA" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.043 required=5 tests=[AWL=-0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/267 X-Sequence-Number: 14014 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigF91D8A06A6C8BD5262A8C6FA Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable Diego de Lima wrote: > Hi list, > =20 > I=C2=B4m using Pg 8.0.3 on Linux FC2. > =20 > This question may have a very simple answer (I hope), but I=C2=B4m havi= ng > lots of trouble solving it, and I counldn=C2=B4t find any other post ab= out it > or anything in the pg docs. > =20 > I have some very complex select statements on 4 million rows > tables. When using LEFT JOIN ON, some select statements takes about 2 > minutes. When I write exactly the same statement but with LEFT JOIN > USING, it takes only 1 minute. Comparing to Oracle, the same statement > takes 1 minute also, but with LEFT JOIN ON. > =20 > Sometimes tables have the same column names and I can use LEFT JOIN > USING, but in some other cases I MUST use LEFT JOIN ON, because the > tables have different column names. > =20 > So my question is: is there a way to make LEFT JOIN ON uses the same > plan of LEFT JOIN USING? > =20 > Thanks, > =20 > Diego de Lima > =20 > =20 I'm guessing that ON/USING isn't the specific problem. It's probably more an issue of how the planner is deciding to do the joins (merge join, hash join, nested loop, etc.) Can you send the results of EXPLAIN ANALYZE ? Also, any sort of join where you have to join against millions of rows is going to be slow. I don't know your specific design, but likely you could change the design to be more selective at an earlier level, which means that you can cut the size of the join by a lot. If you post you query, a lot of times people here can help optimize your query. (But make sure to explain what you are trying to do, so the optimizations make sense.) John =3D:-> --------------enigF91D8A06A6C8BD5262A8C6FA Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDBf2LJdeBCYSNAAMRArdNAJ4msAEWqcEZRGvoYoGanRvDpL8YiACguaIY oA3YYom0rtxPpTetl5SXeMM= =vuUc -----END PGP SIGNATURE----- --------------enigF91D8A06A6C8BD5262A8C6FA-- From pgsql-performance-owner@postgresql.org Fri Aug 19 13:34:57 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C681D52A06 for ; Fri, 19 Aug 2005 13:34:52 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 76579-06 for ; Fri, 19 Aug 2005 16:34:50 +0000 (GMT) Received: from gghcwest.com (adsl-71-128-90-172.dsl.pltn13.pacbell.net [71.128.90.172]) by svr1.postgresql.org (Postfix) with ESMTP id 121C952990 for ; Fri, 19 Aug 2005 13:34:48 -0300 (ADT) Received: from toonses.gghcwest.com (toonses.gghcwest.com [192.168.168.115]) by gghcwest.com (8.12.10/8.12.9) with ESMTP id j7JGYimd027052; Fri, 19 Aug 2005 09:34:44 -0700 Received: from jwb by toonses.gghcwest.com with local (Exim 4.50) id 1E69pN-0007nY-4t; Fri, 19 Aug 2005 09:34:45 -0700 Subject: Re: sustained update load of 1-2k/sec From: "Jeffrey W. Baker" To: Ron Cc: pgsql-performance@postgresql.org In-Reply-To: <6.2.3.4.0.20050819093011.05c1b9d8@pop.earthlink.net> References: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> <33c6269f050819054038af9886@mail.gmail.com> <6.2.3.4.0.20050819093011.05c1b9d8@pop.earthlink.net> Content-Type: text/plain Content-Transfer-Encoding: 7bit Date: Fri, 19 Aug 2005 09:34:44 -0700 Message-Id: <1124469284.29778.4.camel@toonses.gghcwest.com> Mime-Version: 1.0 X-Mailer: Evolution 2.3.6.1 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=-0.841 required=5 tests=[AWL=-0.891, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/269 X-Sequence-Number: 14016 On Fri, 2005-08-19 at 10:54 -0400, Ron wrote: > Maxtor Atlas 15K II's. > Areca's 1GB buffer RAID cards The former are SCSI disks and the latter is an SATA controller. The combination would have a transaction rate of approximately 0. I can vouch for the Areca controllers, however. You can certainly achieve pgbench transaction rates in the hundreds per second even with only 5 7200RPM disks and 128MB cache. Don't forget to buy the battery. -jwb From pgsql-performance-owner@postgresql.org Fri Aug 19 13:48:38 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D987052869 for ; Fri, 19 Aug 2005 13:48:36 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 99034-04 for ; Fri, 19 Aug 2005 16:48:34 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id 93DFB5281F for ; Fri, 19 Aug 2005 13:48:32 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7JGmUlN001513; Fri, 19 Aug 2005 11:48:30 -0500 Subject: Re: extremly low memory usage From: Jeremiah Jahn To: John Arbash Meinel Cc: Jeff Trout , postgres performance In-Reply-To: <4304CB77.4070403@arbash-meinel.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> Content-Type: text/plain Date: Fri, 19 Aug 2005 11:48:29 -0500 Message-Id: <1124470109.27881.152.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Content-Transfer-Encoding: 7bit Received-SPF: none (penguin.goodinassociates.com: 206.80.71.242 is neither permitted nor denied by domain of cs.earlham.edu) client-ip=206.80.71.242; envelope-from=jeremiah@cs.earlham.edu; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1034/Thu Aug 18 15:07:58 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.033 required=5 tests=[AWL=-0.017, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/270 X-Sequence-Number: 14017 Sorry about the formatting. On Thu, 2005-08-18 at 12:55 -0500, John Arbash Meinel wrote: > Jeremiah Jahn wrote: > > >here's an example standard query. Ireally have to make the first hit go > >faster. The table is clustered as well on full_name as well. 'Smith%' > >took 87 seconds on the first hit. I wonder if I set up may array wrong. > >I remeber see something about DMA access versus something else, and > >choose DMA access. LVM maybe? > > > > > It would be nice if you would format your queries to be a little bit > easier to read before posting them. > However, I believe I am reading it correctly, to say that the index scan > on identity is not your slow point. In fact, as near as I can tell, it > only takes 52ms to complete. > > The expensive parts are the 4915 lookups into the litigant_details (each > one takes approx 4ms for a total of ~20s). > And then you do it again on case_data (average 3ms each * 4906 loops = > ~15s). Is there some way to avoid this? > > So there is no need for preloading your indexes on the identity table. > It is definitely not the bottleneck. > > So a few design bits, which may help your database. > Why is "actor_id" a text field instead of a number? This is simply due to the nature of the data. > You could try creating an index on "litigant_details (actor_id, > count_ori)" so that it can do just an index lookup, rather than an index > + filter. I have one, but it doesn't seem to like to use it. Don't really need it though, I can just drop the court_id out of the query. It's redundant, since each actor_id is also unique in litigant details. I had run vac full and analyze but I ran them again anyway and the planning improved. However, my 14 disk raid 10 array is still slower than my 3 disk raid 5 on my production box. 46sec vs 30sec (with live traffic on the production) One of the strange things is that when I run the cat command on my index and tables that are "HOT" it has no effect on memory usage. Right now I'm running ext3 on LVM. I'm still in a position to redo the file system and everything. Is this a good way to do it or should I switch to something else? What about stripe and extent sizes...? kernel parameters to change? ---------------devel box:----------------------- copa=# EXPLAIN ANALYZE select full_name,identity_id,identity.case_id,court.id,date_of_birth,assigned_case_role,litigant_details.impound_litigant_data copa-# from identity copa-# join litigant_details on identity.actor_id = litigant_details.actor_id copa-# join case_data on litigant_details.case_id = case_data.case_id and litigant_details.court_ori = case_data.court_ori copa-# join court on identity.court_ori = court.id copa-# where identity.court_ori = 'IL081025J' and full_name like 'JONES%' order by full_name; QUERY PLAN -------------------------------------------------------------------------------------------------------------------------------------------------------- Nested Loop (cost=3.29..29482.22 rows=3930 width=86) (actual time=114.060..46001.480 rows=5052 loops=1) -> Nested Loop (cost=3.29..16193.27 rows=3820 width=112) (actual time=93.038..24584.275 rows=5052 loops=1) -> Nested Loop (cost=0.00..16113.58 rows=3820 width=113) (actual time=85.778..24536.489 rows=5052 loops=1) -> Index Scan using name_speed on identity (cost=0.00..824.72 rows=3849 width=82) (actual time=50.284..150.133 rows=5057 loops=1) Index Cond: (((full_name)::text >= 'JONES'::character varying) AND ((full_name)::text < 'JONET'::character varying)) Filter: (((court_ori)::text = 'IL081025J'::text) AND ((full_name)::text ~~ 'JONES%'::text)) -> Index Scan using lit_actor_speed on litigant_details (cost=0.00..3.96 rows=1 width=81) (actual time=4.788..4.812 rows=1 loops=5057) Index Cond: (("outer".actor_id)::text = (litigant_details.actor_id)::text) -> Materialize (cost=3.29..3.30 rows=1 width=12) (actual time=0.002..0.003 rows=1 loops=5052) -> Seq Scan on court (cost=0.00..3.29 rows=1 width=12) (actual time=7.248..7.257 rows=1 loops=1) Filter: ('IL081025J'::text = (id)::text) -> Index Scan using case_speed on case_data (cost=0.00..3.46 rows=1 width=26) (actual time=4.222..4.230 rows=1 loops=5052) Index Cond: ((("outer".court_ori)::text = (case_data.court_ori)::text) AND (("outer".case_id)::text = (case_data.case_id)::text)) Total runtime: 46005.994 ms > > More importantly, though, the planner seems to think the join of > identity to litigant_details will only return 1 row, not 5000. > Do you regularly vacuum analyze your tables? > Just as a test, try running: > set enable_nested_loop to off; not quite acceptable Total runtime: 221486.149 ms > And then run EXPLAIN ANALYZE again, just to see if it is faster. > > You probably need to increase some statistics targets, so that the > planner can design better plans. ---------------------this is the output from the production box------------------ LOG: duration: 27213.068 ms statement: EXPLAIN ANALYZE select full_name,identity_id,identity.case_id,court.id,date_of_birth,assigned_case_role,litigant_details.impound_litigant_data from identity join litigant_details on identity.actor_id = litigant_details.actor_id join case_data on litigant_details.case_id = case_data.case_id and litigant_details.court_ori = case_data.court_ori join court on identity.court_ori = court.id where identity.court_ori = 'IL081025J' and full_name like 'JONES%' order by full_name; QUERY PLAN -------------------------------------------------------------------------------------------------------------------------------------------------------- Nested Loop (cost=3.29..43498.76 rows=2648 width=86) (actual time=17.106..27192.000 rows=5052 loops=1) -> Nested Loop (cost=0.00..43442.53 rows=2647 width=87) (actual time=16.947..27120.619 rows=5052 loops=1) -> Nested Loop (cost=0.00..23061.79 rows=3827 width=113) (actual time=16.801..17390.682 rows=5052 loops=1) -> Index Scan using name_speed on identity (cost=0.00..1277.39 rows=3858 width=82) (actual time=9.842..213.424 rows=5057 loops=1) Index Cond: (((full_name)::text >= 'JONES'::character varying) AND ((full_name)::text < 'JONET'::character varying)) Filter: (((court_ori)::text = 'IL081025J'::text) AND ((full_name)::text ~~ 'JONES%'::text)) -> Index Scan using lit_actor_speed on litigant_details (cost=0.00..5.63 rows=1 width=81) (actual time=3.355..3.364 rows=1 loops=5057) Index Cond: (("outer".actor_id)::text = (litigant_details.actor_id)::text) -> Index Scan using case_data_pkey on case_data (cost=0.00..5.31 rows=1 width=26) (actual time=1.897..1.904 rows=1 loops=5052) Index Cond: ((("outer".court_ori)::text = (case_data.court_ori)::text) AND (("outer".case_id)::text = (case_data.case_id)::text)) -> Materialize (cost=3.29..3.30 rows=1 width=12) (actual time=0.002..0.003 rows=1 loops=5052) -> Seq Scan on court (cost=0.00..3.29 rows=1 width=12) (actual time=0.142..0.165 rows=1 loops=1) Filter: ('IL081025J'::text = (id)::text) Total runtime: 27205.060 ms > > > John > =:-> > -- "I didn't know it was impossible when I did it." From pgsql-performance-owner@postgresql.org Fri Aug 19 14:12:51 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 542CA5288B for ; Fri, 19 Aug 2005 14:12:49 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 14926-08 for ; Fri, 19 Aug 2005 17:12:47 +0000 (GMT) Received: from mx1.neopolitan.us (mx1.neopolitan.us [65.87.16.224]) by svr1.postgresql.org (Postfix) with ESMTP id D365A5288E for ; Fri, 19 Aug 2005 14:12:43 -0300 (ADT) Received: from [65.87.16.98] (HELO [10.0.0.221]) by mx1.neopolitan.us (CommuniGate Pro SMTP 4.2.10) with ESMTP id 10387635; Fri, 19 Aug 2005 10:12:43 -0700 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Fri, 19 Aug 2005 10:12:42 -0700 Subject: Re: sustained update load of 1-2k/sec From: "J. Andrew Rogers" To: Mark Cotner , Message-ID: In-Reply-To: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> Mime-version: 1.0 Content-type: text/plain; charset="US-ASCII" Content-transfer-encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.312 required=5 tests=[AWL=-0.438, DNS_FROM_RFC_ABUSE=0.374, DNS_FROM_RFC_POST=1.376] X-Spam-Level: * X-Archive-Number: 200508/271 X-Sequence-Number: 14018 On 8/19/05 1:24 AM, "Mark Cotner" wrote: > I'm currently working on an application that will poll > thousands of cable modems per minute and I would like > to use PostgreSQL to maintain state between polls of > each device. This requires a very heavy amount of > updates in place on a reasonably large table(100k-500k > rows, ~7 columns mostly integers/bigint). Each row > will be refreshed every 15 minutes, or at least that's > how fast I can poll via SNMP. I hope I can tune the > DB to keep up. > > The app is threaded and will likely have well over 100 > concurrent db connections. Temp tables for storage > aren't a preferred option since this is designed to be > a shared nothing approach and I will likely have > several polling processes. Mark, We have PostgreSQL databases on modest hardware doing exactly what you are attempting to (massive scalable SNMP monitoring system). The monitoring volume for a single database server appears to exceed what you are trying to do by a few orders of magnitude with no scaling or performance issues, so I can state without reservation that PostgreSQL can easily handle your application in theory. However, that is predicated on having a well-architected system that minimizes resource contention and unnecessary blocking, and based on your description you may be going about it a bit wrong. The biggest obvious bottleneck is the use of threads and massive process-level parallelization. As others have pointed out, async queues are your friends, as is partitioning the workload horizontally rather than vertically through the app stack. A very scalable high-throughput engine for SNMP polling only requires two or three threads handling different parts of the workload to saturate the network, and by choosing what each thread does carefully you can all but eliminate blocking when there is work to be done. We only use a single database connection to insert all the data into PostgreSQL, and that process/thread receives its data from a work queue. Depending on how you design your system, you can batch many records in your queue as a single transaction. In our case, we also use very few updates, mostly just inserts, which is probably advantageous in terms of throughput if you have the disk for it. The insert I/O load is easily handled, and our disk array is a modest 10k SCSI rig. The only thing that really hammers the server is when multiple reporting processes are running, which frequently touch several million rows each (the database is much larger than the system memory), and even this is manageable with clever database design. In short, what you are trying to do is easily doable on PostgreSQL in theory. However, restrictions on design choices may pose significant hurdles. We did not start out with an ideal system either; it took a fair amount of re-engineering to solve all the bottlenecks and problems that pop up. Good luck, J. Andrew Rogers jrogers@neopolitan.com From pgsql-performance-owner@postgresql.org Fri Aug 19 14:24:52 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 215665288B for ; Fri, 19 Aug 2005 14:18:56 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 23083-04 for ; Fri, 19 Aug 2005 17:18:52 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id D027A52BB8 for ; Fri, 19 Aug 2005 14:18:50 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050819171849m92009j6cke>; Fri, 19 Aug 2005 17:18:50 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 251F955FBA; Fri, 19 Aug 2005 12:18:49 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id A9DE955FAD; Fri, 19 Aug 2005 12:18:42 -0500 (CDT) Message-ID: <43061472.2080201@arbash-meinel.com> Date: Fri, 19 Aug 2005 12:18:42 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Jeremiah Jahn Cc: Jeff Trout , postgres performance Subject: Re: extremly low memory usage References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> In-Reply-To: <1124470109.27881.152.camel@bluejay.goodinassociates.com> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig5193DC20985C1AA5C3A20100" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.042 required=5 tests=[AWL=-0.008, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/273 X-Sequence-Number: 14020 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig5193DC20985C1AA5C3A20100 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Jeremiah Jahn wrote: > Sorry about the formatting. > > On Thu, 2005-08-18 at 12:55 -0500, John Arbash Meinel wrote: > >>Jeremiah Jahn wrote: >> >> ... >>The expensive parts are the 4915 lookups into the litigant_details (each >>one takes approx 4ms for a total of ~20s). >>And then you do it again on case_data (average 3ms each * 4906 loops = >>~15s). > > Is there some way to avoid this? > Well, in general, 3ms for a single lookup seems really long. Maybe your index is bloated by not vacuuming often enough. Do you tend to get a lot of updates to litigant_details? There are a couple possibilities at this point. First, you can REINDEX the appropriate index, and see if that helps. However, if this is a test box, it sounds like you just did a dump and reload, which wouldn't have bloat in an index. Another possibility. Is this the column that you usually use when pulling information out of litigant_details? If so, you can CLUSTER litigant_details on the appropriate index. This will help things be close together that should be, which decreases the index lookup costs. However, if this is not the common column, then you probably will slow down whatever other accesses you may have on this table. After CLUSTER, the current data will stay clustered, but new data will not, so you have to continually CLUSTER, the same way that you might VACUUM. *However*, IIRC CLUSTER grabs an Exclusive lock, so it is as expensive as a VACUUM FULL. Be aware of this, but it might vastly improve your performance, so it would be worth it. > > >>So there is no need for preloading your indexes on the identity table. >>It is definitely not the bottleneck. >> >>So a few design bits, which may help your database. >>Why is "actor_id" a text field instead of a number? > > This is simply due to the nature of the data. > I'm just wondering if changing into a number, and using a number->name lookup would be faster for you. It may not be. In general, I prefer to use numbers for references. I may be over paranoid, but I know that some locales are bad with string -> string comparisons. And since the data in your database is stored as UNICODE, I'm not sure if it has to do any translating or not. Again, something to consider, it may not make any difference. > >>You could try creating an index on "litigant_details (actor_id, >>count_ori)" so that it can do just an index lookup, rather than an index >>+ filter. > > I have one, but it doesn't seem to like to use it. Don't really need it > though, I can just drop the court_id out of the query. It's redundant, > since each actor_id is also unique in litigant details. I had run vac > full and analyze but I ran them again anyway and the planning improved. > However, my 14 disk raid 10 array is still slower than my 3 disk raid 5 > on my production box. 46sec vs 30sec (with live traffic on the > production) One of the strange things is that when I run the cat command > on my index and tables that are "HOT" it has no effect on memory usage. > Right now I'm running ext3 on LVM. I'm still in a position to redo the > file system and everything. Is this a good way to do it or should I > switch to something else? What about stripe and extent sizes...? kernel > parameters to change? Well, the plans are virtually identical. There is one small difference as to whether it joins against case_data or court first. But 'court' is very tiny (small enough to use a seqscan instead of index scan) I'm a little surprised with court being this small that it doesn't do something like a hash aggregation, but court takes no time anyway. The real problem is that your nested loop index time is *much* slower. Devel: -> Index Scan using lit_actor_speed on litigant_details (cost=0.00..3.96 rows=1 width=81) (actual time=4.788..4.812 rows=1 loops=5057) Production: -> Index Scan using lit_actor_speed on litigant_details (cost=0.00..5.63 rows=1 width=81) (actual time=3.355..3.364 rows=1 loops=5057) Devel: -> Index Scan using case_speed on case_data (cost=0.00..3.46 rows=1 width=26) (actual time=4.222..4.230 rows=1 loops=5052) Production: -> Index Scan using case_data_pkey on case_data (cost=0.00..5.31 rows=1 width=26) (actual time=1.897..1.904 rows=1 loops=5052) Notice that the actual per-row cost is as much as 1/2 less than on your devel box. As a test, can you do "time cat $index_file >/dev/null" a couple of times. And then determine the MB/s. Alternatively run vmstat in another shell. If the read/s doesn't change, then you know the "cat" is being served from RAM, and thus it really is cached. I can point you to REINDEX and CLUSTER, but if it is caching in ram, I honestly can't say why the per loop would be that much slower. Are both systems running the same postgres version? It sounds like it is different (since you say something about switching to 8.0). I doubt it, but you might try an 8.1devel version. ... >>Do you regularly vacuum analyze your tables? >>Just as a test, try running: >>set enable_nested_loop to off; > > not quite acceptable > Total runtime: 221486.149 ms > Well, the estimates are now at least closer (3k vs 5k instead of 1), and it is still choosing nested loops. So they probably are faster. I would still be interested in the actual EXPLAIN ANALYZE with nested loops disabled. It is possible that *some* of the nested loops are performing worse than they have to. But really, you have worse index speed, and that needs to be figured out. John =:-> --------------enig5193DC20985C1AA5C3A20100 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDBhRyJdeBCYSNAAMRAnHPAKDURdzSBGw7mBWHd/tSfQn0N2xvRACgry6Z gXkO0btpFCO0AoY/gH79tyY= =5qCi -----END PGP SIGNATURE----- --------------enig5193DC20985C1AA5C3A20100-- From pgsql-performance-owner@postgresql.org Fri Aug 19 14:24:27 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C0E1152C60 for ; Fri, 19 Aug 2005 14:20:57 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 23451-05 for ; Fri, 19 Aug 2005 17:20:50 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id F179B52BA7 for ; Fri, 19 Aug 2005 14:20:49 -0300 (ADT) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: Finding bottleneck Date: Fri, 19 Aug 2005 13:20:49 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3417DD153@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Finding bottleneck Thread-Index: AcWkxr5EndVUITMSSs6KP/tRXihkJAAF32MA From: "Merlin Moncure" To: "Tom Lane" Cc: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/272 X-Sequence-Number: 14019 > Cool --- we've done a fair amount of work on squeezing out internal > inefficiencies during this devel cycle, but it's always hard to predict > just how much anyone will notice in the real world. >=20 > Care to do some oprofile or gprof profiles to see where it's still bad? >=20 Since release of 8.0, we are a strictly windows shop :). I tried building pg with -pg flag and got errors in some of the satellite libraries. I think this is solvable though at some point I'll spend more time on it. Anyways, just so you know the #s that I'm seein, I've run several benchmarks of various programs that access pg via our ISAM bridge. The results are as consistent as they are good. These tests are on the same box using the same .conf on the same freshly loaded data. The disk doesn't play a major role in these tests. All data access is through ExecPrepared libpq C interface. Benchmark is run from a separate box on a LAN. Bill of Materials Traversal ( ~ 62k records). ISAM* pg 8.0 pg 8.1 devel delta 8.0->8.1 running time 63 sec 90 secs 71 secs 21% cpu load 17% 45% 32% 29% =20 loadsecs** 10.71 40.5 22.72 44% recs/sec 984 688 873 recs/loadsec 5882 1530 2728 *ISAM is an anonymous commercial ISAM library in an optimized server architecture (pg smokes the non-optimized flat file version). **Loadsecs being seconds of CPU at 100% load. =20 IOW cpu load drop is around 44%. Amazing! Merlin From pgsql-performance-owner@postgresql.org Fri Aug 19 14:58:06 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B284452A2F for ; Fri, 19 Aug 2005 14:58:04 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 56198-03 for ; Fri, 19 Aug 2005 17:58:02 +0000 (GMT) Received: from smtpauth08.mail.atl.earthlink.net (smtpauth08.mail.atl.earthlink.net [209.86.89.68]) by svr1.postgresql.org (Postfix) with ESMTP id AA77C528DC for ; Fri, 19 Aug 2005 14:57:56 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=M2/N7rt7vzYLvcvQXdvkzrN143YkQfUJ2LYIBdPOQ4HwyadmXh3l51+Mc5N1WEM2; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:Cc:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth08.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E6B7o-0000SA-Uk; Fri, 19 Aug 2005 13:57:53 -0400 Message-Id: <6.2.3.4.0.20050819124201.05c15e40@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Fri, 19 Aug 2005 13:57:46 -0400 To: "Jeffrey W. Baker" From: Ron Subject: Re: sustained update load of 1-2k/sec Cc: pgsql-performance@postgresql.org In-Reply-To: <1124469284.29778.4.camel@toonses.gghcwest.com> References: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> <33c6269f050819054038af9886@mail.gmail.com> <6.2.3.4.0.20050819093011.05c1b9d8@pop.earthlink.net> <1124469284.29778.4.camel@toonses.gghcwest.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bc240ef543a1a701d2484258297a4c1a65350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.625 required=5 tests=[AWL=0.251, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/274 X-Sequence-Number: 14021 At 12:34 PM 8/19/2005, Jeffrey W. Baker wrote: >On Fri, 2005-08-19 at 10:54 -0400, Ron wrote: > > Maxtor Atlas 15K II's. > > > Areca's 1GB buffer RAID cards > >The former are SCSI disks and the latter is an SATA controller. The >combination would have a transaction rate of approximately 0. You are evidently thinking of the Areca ARC-11xx controllers (and you are certainly right for that HW combination ;-) ). Those are not the only product Areca makes that can be upgraded to a 1GB cache. Until SAS infrastructure is good enough, U320 SCSI and FC HD's remain the top performing HD's realistically available. At the most fundamental, your DBMS is only as good as your HD IO subsystem, and your HD IO subsystem is only as good as your HDs. As others have said here, skimping on your HDs is _not_ a good design choice where DBMSs are concerned. As an aside, the Atlas 15K II's are now available in SAS: http://www.maxtor.com/portal/site/Maxtor/menuitem.ba88f6d7cf664718376049b291346068/?channelpath=/en_us/Products/SCSI%20Hard%20Drives/Atlas%2015K%20Family/Atlas%2015K%20II%20SAS I haven't seen independent benches on them, so I explicitly referenced the U320 Atlas 15K II's known performance numbers instead. As I said, Maxtor is claiming even better for the SAS version of the Atlas 15K II. None of the SAS <-> PCI-X or PCI-E RAID cards I know of are ready for mass market yet, although a few are in beta.. >I can vouch for the Areca controllers, however. You can certainly >achieve pgbench transaction rates in the hundreds per second even with >only 5 7200RPM disks and 128MB cache. > >Don't forget to buy the battery. Agreed. Hope this is helpful, Ron Peacetree From pgsql-performance-owner@postgresql.org Fri Aug 19 15:12:36 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7159352BE2 for ; Fri, 19 Aug 2005 15:12:02 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 64054-07 for ; Fri, 19 Aug 2005 18:11:57 +0000 (GMT) Received: from boutiquenumerique.com (boutiquenumerique.com [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 8F21752B87 for ; Fri, 19 Aug 2005 15:11:55 -0300 (ADT) Received: (qmail 5869 invoked from network); 19 Aug 2005 20:12:15 +0200 Received: from unknown (HELO localhost) (boutiquenumerique-lists@192.168.0.4) by boutiquenumerique.com with SMTP; 19 Aug 2005 20:12:15 +0200 Date: Fri, 19 Aug 2005 20:11:54 +0200 To: "Tom Lane" , "Bob Ippolito" Subject: Re: sustained update load of 1-2k/sec Cc: "Mark Cotner" , pgsql-performance@postgresql.org References: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> <67788A4A-4011-49AB-B329-683FD9532661@redivi.com> <17690.1124458553@sss.pgh.pa.us> From: PFC Organization: =?iso-8859-15?Q?La_Boutique_Num=E9rique?= Content-Type: text/plain; format=flowed; delsp=yes; charset=iso-8859-15 MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Message-ID: In-Reply-To: <17690.1124458553@sss.pgh.pa.us> User-Agent: Opera M2/8.0 (Linux, build 1095) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 required=5 tests=[AWL=0.005] X-Spam-Level: X-Archive-Number: 200508/275 X-Sequence-Number: 14022 > While I agree that hundreds of threads seems like overkill, I think the > above advice might be going too far in the other direction. The problem > with single-threaded operation is that any delay affects the whole > system --- eg, if you're blocked waiting for disk I/O, the CPU doesn't You use UDP which is a connectionless protocol... then why use threads ? I'd advise this : Use asynchronous network code (one thread) to do your network stuff. This will lower the CPU used by this code immensely. Every minute, dump a file contianing everything to insert into the table. Use another thread to COPY it into the DB, in a temporary table if you wish, and then INSERT INTO ... SELECT. This should be well adapted to your requirements. From pgsql-performance-owner@postgresql.org Fri Aug 19 15:43:38 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id F256F52E4D for ; Fri, 19 Aug 2005 15:42:59 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 88795-05 for ; Fri, 19 Aug 2005 18:42:52 +0000 (GMT) Received: from smtpauth06.mail.atl.earthlink.net (smtpauth06.mail.atl.earthlink.net [209.86.89.66]) by svr1.postgresql.org (Postfix) with ESMTP id B5FA052C17 for ; Fri, 19 Aug 2005 15:42:50 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=Vy/G/MIfCVF79yAHkmLlW+/SYCIs8lezgttlbB9lPn4IWmdTLxUEiAF0FK8zbhtZ; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:Cc:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth06.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E6BpJ-0000Pm-1m; Fri, 19 Aug 2005 14:42:49 -0400 Message-Id: <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Fri, 19 Aug 2005 14:42:43 -0400 To: Jeremiah Jahn From: Ron Subject: Re: extremly low memory usage Cc: postgres performance In-Reply-To: <43061472.2080201@arbash-meinel.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bcadf749b503341dcf81cdd621d335eccc350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.582 required=5 tests=[AWL=0.208, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/276 X-Sequence-Number: 14023 At 01:18 PM 8/19/2005, John A Meinel wrote: >Jeremiah Jahn wrote: > > Sorry about the formatting. > > > > On Thu, 2005-08-18 at 12:55 -0500, John Arbash Meinel wrote: > > > >>Jeremiah Jahn wrote: > >> > >> > >... > > >>The expensive parts are the 4915 lookups into the litigant_details (each > >>one takes approx 4ms for a total of ~20s). > >>And then you do it again on case_data (average 3ms each * 4906 loops = > >>~15s). > > > > Is there some way to avoid this? > > > >Well, in general, 3ms for a single lookup seems really long. Maybe your >index is bloated by not vacuuming often enough. Do you tend to get a lot >of updates to litigant_details? Given that the average access time for a 15Krpm HD is in the 5.5-6ms range (7.5-8ms for a 10Krpm HD), having an average of 3ms for a single lookup implies that ~1/2 (the 15Krpm case) or ~1/3 (the 10Krpm case) table accesses is requiring a seek. This implies a poor match between physical layout and access pattern. If I understand correctly, the table should not be very fragmented given that this is a reasonably freshly loaded DB? That implies that the fields being looked up are not well sorted in the table compared to the query pattern. If the entire table could fit in RAM, this would be far less of a consideration. Failing that, the physical HD layout has to be improved or the query pattern has to be changed to reduce seeks. >There are a couple possibilities at this point. First, you can REINDEX >the appropriate index, and see if that helps. However, if this is a test >box, it sounds like you just did a dump and reload, which wouldn't have >bloat in an index. > >Another possibility. Is this the column that you usually use when >pulling information out of litigant_details? If so, you can CLUSTER >litigant_details on the appropriate index. This will help things be >close together that should be, which decreases the index lookup costs. > >However, if this is not the common column, then you probably will slow >down whatever other accesses you may have on this table. > >After CLUSTER, the current data will stay clustered, but new data will >not, so you have to continually CLUSTER, the same way that you might >VACUUM. *However*, IIRC CLUSTER grabs an Exclusive lock, so it is as >expensive as a VACUUM FULL. Be aware of this, but it might vastly >improve your performance, so it would be worth it. CLUSTER can be a very large maintenance overhead/problem if the table(s) in question actually need to be "continually" re CLUSTER ed. If there is no better solution available, then you do what you have to, but it feels like there should be a better answer here. Perhaps the DB schema needs examining to see if it matches up well with its real usage? Ron Peacetree From pgsql-performance-owner@postgresql.org Fri Aug 19 16:23:57 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8645C52873 for ; Fri, 19 Aug 2005 16:23:11 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 16256-06 for ; Fri, 19 Aug 2005 19:23:08 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id CE6045285B for ; Fri, 19 Aug 2005 16:23:06 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050819192307m92009htcne>; Fri, 19 Aug 2005 19:23:07 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 0925755FBA; Fri, 19 Aug 2005 14:23:07 -0500 (CDT) Received: from [192.168.1.12] (Jigglypuff.arbash-meinel.com [192.168.1.12]) by juju.arbash-meinel.com (Postfix) with ESMTP id 5B4AA55FAD; Fri, 19 Aug 2005 14:23:02 -0500 (CDT) Message-ID: <43063196.9000001@arbash-meinel.com> Date: Fri, 19 Aug 2005 14:23:02 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Ron Cc: Jeremiah Jahn , postgres performance Subject: Re: extremly low memory usage References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> In-Reply-To: <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig83629B4E59FE3046462F8DEF" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.042 required=5 tests=[AWL=-0.008, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/277 X-Sequence-Number: 14024 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig83629B4E59FE3046462F8DEF Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Ron wrote: > At 01:18 PM 8/19/2005, John A Meinel wrote: > >> Jeremiah Jahn wrote: >> > Sorry about the formatting. >> > >> > On Thu, 2005-08-18 at 12:55 -0500, John Arbash Meinel wrote: >> > >> >>Jeremiah Jahn wrote: >> >> >> >> >> >> ... >> >> >>The expensive parts are the 4915 lookups into the litigant_details >> (each >> >>one takes approx 4ms for a total of ~20s). >> >>And then you do it again on case_data (average 3ms each * 4906 loops = >> >>~15s). >> > >> > Is there some way to avoid this? >> > >> >> Well, in general, 3ms for a single lookup seems really long. Maybe your >> index is bloated by not vacuuming often enough. Do you tend to get a lot >> of updates to litigant_details? > > > Given that the average access time for a 15Krpm HD is in the 5.5-6ms > range (7.5-8ms for a 10Krpm HD), having an average of 3ms for a single > lookup implies that ~1/2 (the 15Krpm case) or ~1/3 (the 10Krpm case) > table accesses is requiring a seek. > Well, from what he has said, the total indexes are < 1GB and he has 6GB of ram. So everything should fit. Not to mention he is only accessing 5000/several million rows. > This implies a poor match between physical layout and access pattern. This seems to be the case. But since this is not the only query, it may be that other access patterns are more important to optimize for. > > If I understand correctly, the table should not be very fragmented given > that this is a reasonably freshly loaded DB? That implies that the > fields being looked up are not well sorted in the table compared to the > query pattern. > > If the entire table could fit in RAM, this would be far less of a > consideration. Failing that, the physical HD layout has to be improved > or the query pattern has to be changed to reduce seeks. > > ... >> After CLUSTER, the current data will stay clustered, but new data will >> not, so you have to continually CLUSTER, the same way that you might >> VACUUM. *However*, IIRC CLUSTER grabs an Exclusive lock, so it is as >> expensive as a VACUUM FULL. Be aware of this, but it might vastly >> improve your performance, so it would be worth it. > > > CLUSTER can be a very large maintenance overhead/problem if the table(s) > in question actually need to be "continually" re CLUSTER ed. > > If there is no better solution available, then you do what you have to, > but it feels like there should be a better answer here. > > Perhaps the DB schema needs examining to see if it matches up well with > its real usage? > > Ron Peacetree > I certainly agree that CLUSTER is expensive, and is an on-going maintenance issue. If it is the normal access pattern, though, it may be worth it. I also wonder, though, if his table is properly normalized. Which, as you mentioned, might lead to improved access patterns. John =:-> --------------enig83629B4E59FE3046462F8DEF Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDBjGWJdeBCYSNAAMRAsisAKDMPTuigYc5k2BjdPS3vsBPcxJ3DwCgjE0f 6hzEvFAReO5GlSMzwbZeUf0= =ARWA -----END PGP SIGNATURE----- --------------enig83629B4E59FE3046462F8DEF-- From pgsql-performance-owner@postgresql.org Fri Aug 19 16:32:37 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1A3FC52A16 for ; Fri, 19 Aug 2005 16:32:15 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 23242-05 for ; Fri, 19 Aug 2005 19:32:12 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.202]) by svr1.postgresql.org (Postfix) with ESMTP id A40DF52A80 for ; Fri, 19 Aug 2005 16:32:09 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i27so556008wra for ; Fri, 19 Aug 2005 12:31:58 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=SrTQn70CE1NPLM2hyk61e9tHpd5kT368hdaOtlJKBSxX+x4NyH1p1ZeFS2+MtSjvnc67BVgsk8pFmAMW/nmWSi/c0TlbgMPq7yeCEd51AGQIM0ZCqAWHjiWU9ELOJbc44qsJKcfFFmC9dAcvw0UAeJ8AE0BgxkuWotxFlMQugyk= Received: by 10.54.153.7 with SMTP id a7mr2038071wre; Fri, 19 Aug 2005 12:31:58 -0700 (PDT) Received: by 10.54.86.15 with HTTP; Fri, 19 Aug 2005 12:31:58 -0700 (PDT) Message-ID: <33c6269f0508191231b12f5b@mail.gmail.com> Date: Fri, 19 Aug 2005 15:31:58 -0400 From: Alex Turner To: Ron Subject: Re: sustained update load of 1-2k/sec Cc: postgres performance In-Reply-To: <6.2.3.4.0.20050819093011.05c1b9d8@pop.earthlink.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> <33c6269f050819054038af9886@mail.gmail.com> <6.2.3.4.0.20050819093011.05c1b9d8@pop.earthlink.net> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.103 required=5 tests=[AWL=0.079, RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/278 X-Sequence-Number: 14025 Don't forget that Ultra 320 is the speed of the bus, not each drive.=20 No matter how many honking 15k disks you put on a 320MB bus, you can only get 320MB/sec! and have so many outstanding IO/s on the bus. Not so with SATA! Each drive is on it's own bus, and you are only limited by the speed of your PCI-X Bus, which can be as high as 800MB/sec at 133Mhz/64bit. It's cheap and it's fast - all you have to do is pay for the enclosure, which can be a bit pricey, but there are some nice 24bay and even 40bay enclosures out there for SATA. Yes a 15k RPM drive will give you better seek time and better peak through put, but put them all on a single U320 bus and you won't see much return past a stripe size of 3 or 4. If it's raw transactions per second data warehouse style, it's all about the xlog baby which is sequential writes, and all about large block reads, which is sequential reads. Alex Turner NetEconomist P.S. Sorry if i'm a bit punchy, I've been up since yestarday with server upgrade nightmares that continue ;) On 8/19/05, Ron wrote: > Alex mentions a nice setup, but I'm pretty sure I know how to beat > that IO subsystems HW's performance by at least 1.5x or 2x. Possibly > more. (No, I do NOT work for any vendor I'm about to discuss.) >=20 > Start by replacing the WD Raptors with Maxtor Atlas 15K II's. > At 5.5ms average access, 97.4MB/s outer track throughput, 85.9MB/s > average, and 74.4 MB/s inner track throughput, they have the best > performance characteristics of any tested shipping HDs I know > of. (Supposedly the new SAS versions will _sustain_ ~98MB/s, but > I'll believe that only if I see it under independent testing). > In comparison, the numbers on the WD740GD are 8.1ms average access, > 71.8, 62.9, and 53.9 MB/s outer, average and inner track throughputs > respectively. >=20 > Be prepared to use as many of them as possible (read: as many you can > afford) if you want to maximize transaction rates, particularly for > small transactions like this application seems to be mentioning. >=20 > Next, use a better RAID card. The TOL enterprise stuff (Xyratex, > Engino, Dot-hill) is probably too expensive, but in the commodity > market benchmarks indicate that that Areca's 1GB buffer RAID cards > currently outperform all the other commodity RAID stuff. >=20 > 9 Atlas II's per card in a RAID 5 set, or 16 per card in a RAID 10 > set, should max the RAID card's throughput and come very close to, if > not attaining, the real world peak bandwidth of the 64b 133MHz PCI-X > bus they are plugged into. Say somewhere in the 700-800MB/s range. >=20 > Repeat the above for as many independent PCI-X buses as you have for > a very fast commodity RAID IO subsystem. >=20 > Two such configured cards used in the dame manner as mentioned by > Alex should easily attain 1.5x - 2x the transaction numbers mentioned > by Alex unless there's a bottleneck somewhere else in the system design. >=20 > Hope this helps, > Ron Peacetree >=20 > At 08:40 AM 8/19/2005, Alex Turner wrote: > >I have managed tx speeds that high from postgresql going even as high > >as 2500/sec for small tables, but it does require a good RAID > >controler card (yes I'm even running with fsync on). I'm using 3ware > >9500S-8MI with Raptor drives in multiple RAID 10s. The box wasn't too > >$$$ at just around $7k. I have two independant controlers on two > >independant PCI buses to give max throughput. on with a 6 drive RAID > >10 and the other with two 4 drive RAID 10s. > > > >Alex Turner > >NetEconomist > > > >On 8/19/05, Mark Cotner wrote: > > > Hi all, > > > I bet you get tired of the same ole questions over and > > > over. > > > > > > I'm currently working on an application that will poll > > > thousands of cable modems per minute and I would like > > > to use PostgreSQL to maintain state between polls of > > > each device. This requires a very heavy amount of > > > updates in place on a reasonably large table(100k-500k > > > rows, ~7 columns mostly integers/bigint). Each row > > > will be refreshed every 15 minutes, or at least that's > > > how fast I can poll via SNMP. I hope I can tune the > > > DB to keep up. > > > > > > The app is threaded and will likely have well over 100 > > > concurrent db connections. Temp tables for storage > > > aren't a preferred option since this is designed to be > > > a shared nothing approach and I will likely have > > > several polling processes. > > > > > > Here are some of my assumptions so far . . . > > > > > > HUGE WAL > > > Vacuum hourly if not more often > > > > > > I'm getting 1700tx/sec from MySQL and I would REALLY > > > prefer to use PG. I don't need to match the number, > > > just get close. > > > > > > Is there a global temp table option? In memory tables > > > would be very beneficial in this case. I could just > > > flush it to disk occasionally with an insert into blah > > > select from memory table. > > > > > > Any help or creative alternatives would be greatly > > > appreciated. :) > > > > > > 'njoy, > > > Mark > > > > > > > > > -- > > > Writing software requires an intelligent person, > > > creating functional art requires an artist. > > > -- Unknown > > > > > > > > > ---------------------------(end of broadcast)------------------------= --- > > > TIP 3: Have you checked our extensive FAQ? > > > > > > http://www.postgresql.org/docs/faq > > > > > > >---------------------------(end of broadcast)--------------------------- > >TIP 2: Don't 'kill -9' the postmaster >=20 >=20 >=20 >=20 > ---------------------------(end of broadcast)--------------------------- > TIP 9: In versions below 8.0, the planner will ignore your desire to > choose an index scan if your joining column's datatypes do not > match > From pgsql-performance-owner@postgresql.org Fri Aug 19 16:57:03 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E37C152A72 for ; Fri, 19 Aug 2005 16:56:35 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 42580-07 for ; Fri, 19 Aug 2005 19:56:31 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id 7ED2E52858 for ; Fri, 19 Aug 2005 16:56:29 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7JJuRVs005330; Fri, 19 Aug 2005 14:56:29 -0500 Subject: Re: extremly low memory usage From: Jeremiah Jahn To: John A Meinel Cc: Jeff Trout , postgres performance In-Reply-To: <43061472.2080201@arbash-meinel.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> Content-Type: text/plain Date: Fri, 19 Aug 2005 14:56:26 -0500 Message-Id: <1124481386.27881.167.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Content-Transfer-Encoding: 7bit Received-SPF: none (penguin.goodinassociates.com: 206.80.71.242 is neither permitted nor denied by domain of cs.earlham.edu) client-ip=206.80.71.242; envelope-from=jeremiah@cs.earlham.edu; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1034/Thu Aug 18 15:07:58 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.033 required=5 tests=[AWL=-0.017, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/279 X-Sequence-Number: 14026 On Fri, 2005-08-19 at 12:18 -0500, John A Meinel wrote: > Jeremiah Jahn wrote: > > Sorry about the formatting. > > > > On Thu, 2005-08-18 at 12:55 -0500, John Arbash Meinel wrote: > > > >>Jeremiah Jahn wrote: > >> > >> > > ... > > >>The expensive parts are the 4915 lookups into the litigant_details (each > >>one takes approx 4ms for a total of ~20s). > >>And then you do it again on case_data (average 3ms each * 4906 loops = > >>~15s). > > > > Is there some way to avoid this? > > > > Well, in general, 3ms for a single lookup seems really long. Maybe your > index is bloated by not vacuuming often enough. Do you tend to get a lot > of updates to litigant_details? I have vacuumed this already. I get lots of updates, but this data is mostly unchanging. > > There are a couple possibilities at this point. First, you can REINDEX > the appropriate index, and see if that helps. However, if this is a test > box, it sounds like you just did a dump and reload, which wouldn't have > bloat in an index. I loaded it using slony > > Another possibility. Is this the column that you usually use when > pulling information out of litigant_details? If so, you can CLUSTER > litigant_details on the appropriate index. This will help things be > close together that should be, which decreases the index lookup costs. clustering on this right now. Most of the other things are already clustered. name and case_data > > However, if this is not the common column, then you probably will slow > down whatever other accesses you may have on this table. > > After CLUSTER, the current data will stay clustered, but new data will > not, so you have to continually CLUSTER, the same way that you might > VACUUM. *However*, IIRC CLUSTER grabs an Exclusive lock, so it is as > expensive as a VACUUM FULL. Be aware of this, but it might vastly > improve your performance, so it would be worth it. I generally re-cluster once a week. > > > > > > >>So there is no need for preloading your indexes on the identity table. > >>It is definitely not the bottleneck. > >> > >>So a few design bits, which may help your database. > >>Why is "actor_id" a text field instead of a number? > > > > This is simply due to the nature of the data. > > > > I'm just wondering if changing into a number, and using a number->name > lookup would be faster for you. It may not be. In general, I prefer to > use numbers for references. I may be over paranoid, but I know that some > locales are bad with string -> string comparisons. And since the data in > your database is stored as UNICODE, I'm not sure if it has to do any > translating or not. Again, something to consider, it may not make any > difference. I don't believe so. I initialze the DB as 'lang=C'. I used to have the problem where things were being inited as en_US. this would prevent any text based index from working. This doesn't seem to be the case here, so I'm not worried about it. > > > > > >>You could try creating an index on "litigant_details (actor_id, > >>count_ori)" so that it can do just an index lookup, rather than an index > >>+ filter. > > > > I have one, but it doesn't seem to like to use it. Don't really need it > > though, I can just drop the court_id out of the query. It's redundant, > > since each actor_id is also unique in litigant details. I had run vac > > full and analyze but I ran them again anyway and the planning improved. > > However, my 14 disk raid 10 array is still slower than my 3 disk raid 5 > > on my production box. 46sec vs 30sec (with live traffic on the > > production) One of the strange things is that when I run the cat command > > on my index and tables that are "HOT" it has no effect on memory usage. > > Right now I'm running ext3 on LVM. I'm still in a position to redo the > > file system and everything. Is this a good way to do it or should I > > switch to something else? What about stripe and extent sizes...? kernel > > parameters to change? > > Well, the plans are virtually identical. There is one small difference > as to whether it joins against case_data or court first. But 'court' is > very tiny (small enough to use a seqscan instead of index scan) I'm a > little surprised with court being this small that it doesn't do > something like a hash aggregation, but court takes no time anyway. > > The real problem is that your nested loop index time is *much* slower. > > Devel: > -> Index Scan using lit_actor_speed on litigant_details > (cost=0.00..3.96 rows=1 width=81) > (actual time=4.788..4.812 rows=1 loops=5057) > > Production: > -> Index Scan using lit_actor_speed on litigant_details > (cost=0.00..5.63 rows=1 width=81) > (actual time=3.355..3.364 rows=1 loops=5057) > > Devel: > -> Index Scan using case_speed on case_data > (cost=0.00..3.46 rows=1 width=26) > (actual time=4.222..4.230 rows=1 loops=5052) > > Production: > -> Index Scan using case_data_pkey on case_data > (cost=0.00..5.31 rows=1 width=26) > (actual time=1.897..1.904 rows=1 loops=5052) > > Notice that the actual per-row cost is as much as 1/2 less than on your > devel box. > > As a test, can you do "time cat $index_file >/dev/null" a couple of > times. And then determine the MB/s. > Alternatively run vmstat in another shell. If the read/s doesn't change, > then you know the "cat" is being served from RAM, and thus it really is > cached. it's cached alright. I'm getting a read rate of about 150MB/sec. I would have thought is would be faster with my raid setup. I think I'm going to scrap the whole thing and get rid of LVM. I'll just do a straight ext3 system. Maybe that will help. Still trying to get suggestions for a stripe size. > > I can point you to REINDEX and CLUSTER, but if it is caching in ram, I > honestly can't say why the per loop would be that much slower. > Are both systems running the same postgres version? It sounds like it is > different (since you say something about switching to 8.0). These had little or no effect. The production machine is running 7.4 while the devel machine is running 8.0 > I doubt it, but you might try an 8.1devel version. > > ... > > >>Do you regularly vacuum analyze your tables? > >>Just as a test, try running: > >>set enable_nested_loop to off; > > > > not quite acceptable > > Total runtime: 221486.149 ms > > > > Well, the estimates are now at least closer (3k vs 5k instead of 1), and > it is still choosing nested loops. So they probably are faster. > I would still be interested in the actual EXPLAIN ANALYZE with nested > loops disabled. It is possible that *some* of the nested loops are > performing worse than they have to. this is a cached version. > copa=> explain analyze select full_name,identity_id,identity.case_id,court.id,date_of_birth,assigned_case_role,litigant_details.impound_litigant_data > copa-> from identity > copa-> join litigant_details on identity.actor_id = litigant_details.actor_id > copa-> join case_data on litigant_details.case_id = case_data.case_id and litigant_details.court_ori = case_data.court_ori > copa-> join court on identity.court_ori = court.id > copa-> where identity.court_ori = 'IL081025J' and full_name like 'SMITH%' order by full_name; > QUERY PLAN > ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- > Sort (cost=100502560.72..100502583.47 rows=9099 width=86) (actual time=17843.876..17849.401 rows=8094 loops=1) > Sort Key: identity.full_name > -> Merge Join (cost=100311378.72..100501962.40 rows=9099 width=86) (actual time=15195.816..17817.847 rows=8094 loops=1) > Merge Cond: ((("outer".court_ori)::text = "inner"."?column10?") AND (("outer".case_id)::text = "inner"."?column11?")) > -> Index Scan using case_speed on case_data (cost=0.00..170424.73 rows=3999943 width=26) (actual time=0.015..4540.525 rows=3018284 loops=1) > -> Sort (cost=100311378.72..100311400.82 rows=8839 width=112) (actual time=9594.985..9601.174 rows=8094 loops=1) > Sort Key: (litigant_details.court_ori)::text, (litigant_details.case_id)::text > -> Nested Loop (cost=100002491.43..100310799.34 rows=8839 width=112) (actual time=6892.755..9555.828 rows=8094 loops=1) > -> Seq Scan on court (cost=0.00..3.29 rows=1 width=12) (actual time=0.085..0.096 rows=1 loops=1) > Filter: ('IL081025J'::text = (id)::text) > -> Merge Join (cost=2491.43..310707.66 rows=8839 width=113) (actual time=6892.656..9519.680 rows=8094 loops=1) > Merge Cond: (("outer".actor_id)::text = "inner"."?column7?") > -> Index Scan using lit_actor_speed on litigant_details (cost=0.00..295722.00 rows=4956820 width=81) (actual time=0.027..5613.814 rows=3736703 loops=1) > -> Sort (cost=2491.43..2513.71 rows=8913 width=82) (actual time=116.071..122.272 rows=8100 loops=1) > Sort Key: (identity.actor_id)::text > -> Index Scan using name_speed on identity (cost=0.00..1906.66 rows=8913 width=82) (actual time=0.133..81.104 rows=8100 loops=1) > Index Cond: (((full_name)::text >= 'SMITH'::character varying) AND ((full_name)::text < 'SMITI'::character varying)) > Filter: (((court_ori)::text = 'IL081025J'::text) AND ((full_name)::text ~~ 'SMITH%'::text)) > Total runtime: 17859.917 ms > But really, you have worse index speed, and that needs to be figured out. > > John > =:-> -- Speak softly and carry a +6 two-handed sword. From pgsql-performance-owner@postgresql.org Fri Aug 19 17:11:55 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id DC9C152858 for ; Fri, 19 Aug 2005 17:11:33 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 53255-07 for ; Fri, 19 Aug 2005 20:11:32 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id BFC8552869 for ; Fri, 19 Aug 2005 17:11:29 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7JKBSis005565; Fri, 19 Aug 2005 15:11:28 -0500 Subject: Re: extremly low memory usage From: Jeremiah Jahn To: John A Meinel Cc: Ron , postgres performance In-Reply-To: <43063196.9000001@arbash-meinel.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> <43063196.9000001@arbash-meinel.com> Content-Type: text/plain Date: Fri, 19 Aug 2005 15:11:28 -0500 Message-Id: <1124482288.27881.182.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Content-Transfer-Encoding: 7bit Received-SPF: none (penguin.goodinassociates.com: 206.80.71.242 is neither permitted nor denied by domain of cs.earlham.edu) client-ip=206.80.71.242; envelope-from=jeremiah@cs.earlham.edu; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1034/Thu Aug 18 15:07:58 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.034 required=5 tests=[AWL=-0.016, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/280 X-Sequence-Number: 14027 On Fri, 2005-08-19 at 14:23 -0500, John A Meinel wrote: > Ron wrote: > > At 01:18 PM 8/19/2005, John A Meinel wrote: > > > >> Jeremiah Jahn wrote: > >> > Sorry about the formatting. > >> > > >> > On Thu, 2005-08-18 at 12:55 -0500, John Arbash Meinel wrote: > >> > > >> >>Jeremiah Jahn wrote: > >> >> > >> >> > >> > >> ... > >> > >> >>The expensive parts are the 4915 lookups into the litigant_details > >> (each > >> >>one takes approx 4ms for a total of ~20s). > >> >>And then you do it again on case_data (average 3ms each * 4906 loops = > >> >>~15s). > >> > > >> > Is there some way to avoid this? > >> > > >> > >> Well, in general, 3ms for a single lookup seems really long. Maybe your > >> index is bloated by not vacuuming often enough. Do you tend to get a lot > >> of updates to litigant_details? > > > > > > Given that the average access time for a 15Krpm HD is in the 5.5-6ms > > range (7.5-8ms for a 10Krpm HD), having an average of 3ms for a single > > lookup implies that ~1/2 (the 15Krpm case) or ~1/3 (the 10Krpm case) > > table accesses is requiring a seek. > > I think LVM may be a problem, since it also seems to break things up on the file system. My access time on the seek should be around 1/7th the 15Krpm I believe since it's a 14 disk raid 10 array. And no other traffic at the moment. > > > Well, from what he has said, the total indexes are < 1GB and he has 6GB > of ram. So everything should fit. Not to mention he is only accessing > 5000/several million rows. I table spaced some of the indexes and they are around 211066880 bytes for the name_speed index and 149825330 for the lit_actor_speed index tables seem to be about a gig. > > > > This implies a poor match between physical layout and access pattern. > > This seems to be the case. But since this is not the only query, it may > be that other access patterns are more important to optimize for. > > > > > If I understand correctly, the table should not be very fragmented given > > that this is a reasonably freshly loaded DB? That implies that the > > fields being looked up are not well sorted in the table compared to the > > query pattern. > > > > If the entire table could fit in RAM, this would be far less of a > > consideration. Failing that, the physical HD layout has to be improved > > or the query pattern has to be changed to reduce seeks. > > > > > > ... > > >> After CLUSTER, the current data will stay clustered, but new data will > >> not, so you have to continually CLUSTER, the same way that you might > >> VACUUM. *However*, IIRC CLUSTER grabs an Exclusive lock, so it is as > >> expensive as a VACUUM FULL. Be aware of this, but it might vastly > >> improve your performance, so it would be worth it. > > > > > > CLUSTER can be a very large maintenance overhead/problem if the table(s) > > in question actually need to be "continually" re CLUSTER ed. > > > > If there is no better solution available, then you do what you have to, > > but it feels like there should be a better answer here. > > > > Perhaps the DB schema needs examining to see if it matches up well with > > its real usage? > > > > Ron Peacetree > > > > I certainly agree that CLUSTER is expensive, and is an on-going > maintenance issue. If it is the normal access pattern, though, it may be > worth it. The query I've sent you is one of the most common I get just change the name. I handle about 180K of them a day mostly between 8 and 5. The clustering has never really been a problem. Like I said before I do it about once a week. I handle about 3000 update an hour consisting of about 1000-3000 statement per update. ie about 2.5 million updates per hour. In the last few months or so I've filtered these down to about 400K update/delete/insert statements per hour. > > I also wonder, though, if his table is properly normalized. Which, as > you mentioned, might lead to improved access patterns. The system is about as normalized as I can get it. In general the layout is the following: courts have cases, cases have litigant_details. Actors have identities and litigant_details. > > John > =:-> -- Speak softly and carry a +6 two-handed sword. From pgsql-performance-owner@postgresql.org Fri Aug 19 18:06:23 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CAF8E52C17 for ; Fri, 19 Aug 2005 18:01:50 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 13616-01 for ; Fri, 19 Aug 2005 21:01:43 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id 3CC9E52BF3 for ; Fri, 19 Aug 2005 18:01:42 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7JL1etb006646; Fri, 19 Aug 2005 16:01:40 -0500 Subject: Re: extremly low memory usage From: Jeremiah Jahn To: John A Meinel Cc: Ron , postgres performance In-Reply-To: <43063196.9000001@arbash-meinel.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> <43063196.9000001@arbash-meinel.com> Content-Type: text/plain Date: Fri, 19 Aug 2005 16:01:39 -0500 Message-Id: <1124485299.27881.192.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Content-Transfer-Encoding: 7bit Received-SPF: none (penguin.goodinassociates.com: 206.80.71.242 is neither permitted nor denied by domain of cs.earlham.edu) client-ip=206.80.71.242; envelope-from=jeremiah@cs.earlham.edu; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1034/Thu Aug 18 15:07:58 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.034 required=5 tests=[AWL=-0.016, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/281 X-Sequence-Number: 14028 Rebuild in progress with just ext3 on the raid array...will see if this helps the access times. If it doesn't I'll mess with the stripe size. I have REINDEXED, CLUSTERED, tablespaced and cached with 'cat table/index > /dev/null' none of this seems to have helped, or even increased my memory usage. argh! The only thing about this new system that I'm unfamiliar with is the array setup and LVM, which is why I think that's where the issue is. clustering and indexing as well as vacuum etc are things that I do and have been aware of for sometime. Perhaps slony is a factor, but I really don't see it causing problems on index read speed esp. when it's not running. thanx for your help, I really appreciate it. -jj- On Fri, 2005-08-19 at 14:23 -0500, John A Meinel wrote: > Ron wrote: > > At 01:18 PM 8/19/2005, John A Meinel wrote: > > > >> Jeremiah Jahn wrote: > >> > Sorry about the formatting. > >> > > >> > On Thu, 2005-08-18 at 12:55 -0500, John Arbash Meinel wrote: > >> > > >> >>Jeremiah Jahn wrote: > >> >> > >> >> > >> > >> ... > >> > >> >>The expensive parts are the 4915 lookups into the litigant_details > >> (each > >> >>one takes approx 4ms for a total of ~20s). > >> >>And then you do it again on case_data (average 3ms each * 4906 loops = > >> >>~15s). > >> > > >> > Is there some way to avoid this? > >> > > >> > >> Well, in general, 3ms for a single lookup seems really long. Maybe your > >> index is bloated by not vacuuming often enough. Do you tend to get a lot > >> of updates to litigant_details? > > > > > > Given that the average access time for a 15Krpm HD is in the 5.5-6ms > > range (7.5-8ms for a 10Krpm HD), having an average of 3ms for a single > > lookup implies that ~1/2 (the 15Krpm case) or ~1/3 (the 10Krpm case) > > table accesses is requiring a seek. > > > > > Well, from what he has said, the total indexes are < 1GB and he has 6GB > of ram. So everything should fit. Not to mention he is only accessing > 5000/several million rows. > > > > This implies a poor match between physical layout and access pattern. > > This seems to be the case. But since this is not the only query, it may > be that other access patterns are more important to optimize for. > > > > > If I understand correctly, the table should not be very fragmented given > > that this is a reasonably freshly loaded DB? That implies that the > > fields being looked up are not well sorted in the table compared to the > > query pattern. > > > > If the entire table could fit in RAM, this would be far less of a > > consideration. Failing that, the physical HD layout has to be improved > > or the query pattern has to be changed to reduce seeks. > > > > > > ... > > >> After CLUSTER, the current data will stay clustered, but new data will > >> not, so you have to continually CLUSTER, the same way that you might > >> VACUUM. *However*, IIRC CLUSTER grabs an Exclusive lock, so it is as > >> expensive as a VACUUM FULL. Be aware of this, but it might vastly > >> improve your performance, so it would be worth it. > > > > > > CLUSTER can be a very large maintenance overhead/problem if the table(s) > > in question actually need to be "continually" re CLUSTER ed. > > > > If there is no better solution available, then you do what you have to, > > but it feels like there should be a better answer here. > > > > Perhaps the DB schema needs examining to see if it matches up well with > > its real usage? > > > > Ron Peacetree > > > > I certainly agree that CLUSTER is expensive, and is an on-going > maintenance issue. If it is the normal access pattern, though, it may be > worth it. > > I also wonder, though, if his table is properly normalized. Which, as > you mentioned, might lead to improved access patterns. > > John > =:-> -- Speak softly and carry a +6 two-handed sword. From pgsql-performance-owner@postgresql.org Fri Aug 19 18:06:28 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 69D1452B9D for ; Fri, 19 Aug 2005 18:03:21 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 00678-09 for ; Fri, 19 Aug 2005 21:03:18 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id C248052B98 for ; Fri, 19 Aug 2005 18:03:16 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050819210318m92009jnkpe>; Fri, 19 Aug 2005 21:03:18 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 0F23255FBA; Fri, 19 Aug 2005 16:03:18 -0500 (CDT) Received: from [192.168.1.103] (63-230-159-164.cdrr.qwest.net [63.230.159.164]) by juju.arbash-meinel.com (Postfix) with ESMTP id 55CA255FAD; Fri, 19 Aug 2005 16:03:07 -0500 (CDT) Message-ID: <43064908.7010505@arbash-meinel.com> Date: Fri, 19 Aug 2005 16:03:04 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Jeremiah Jahn Cc: Jeff Trout , postgres performance Subject: Re: extremly low memory usage References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <1124481386.27881.167.camel@bluejay.goodinassociates.com> In-Reply-To: <1124481386.27881.167.camel@bluejay.goodinassociates.com> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigE5E6EAB78A66B0E023BE7ECF" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.025 required=5 tests=[AWL=-0.025, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/282 X-Sequence-Number: 14029 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigE5E6EAB78A66B0E023BE7ECF Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Jeremiah Jahn wrote: > On Fri, 2005-08-19 at 12:18 -0500, John A Meinel wrote: > >>Jeremiah Jahn wrote: >> ... >> >>Well, in general, 3ms for a single lookup seems really long. Maybe your >>index is bloated by not vacuuming often enough. Do you tend to get a lot >>of updates to litigant_details? > > I have vacuumed this already. I get lots of updates, but this data is > mostly unchanging. > > >>There are a couple possibilities at this point. First, you can REINDEX >>the appropriate index, and see if that helps. However, if this is a test >>box, it sounds like you just did a dump and reload, which wouldn't have >>bloat in an index. > > > I loaded it using slony I don't know that slony versus pg_dump/pg_restore really matters. The big thing is that Updates wouldn't be trashing your index. But if you are saying that you cluster once/wk your index can't be that messed up anyway. (Unless CLUSTER messes up the non-clustered indexes, but that would make cluster much less useful, so I would have guessed this was not the case) > > >>Another possibility. Is this the column that you usually use when >>pulling information out of litigant_details? If so, you can CLUSTER >>litigant_details on the appropriate index. This will help things be >>close together that should be, which decreases the index lookup costs. > > clustering on this right now. Most of the other things are already > clustered. name and case_data Just as a reality check, they are clustered on the columns in question, right? (I don't know if this column is a primary key or not, but any index can be used for clustering). > > >>However, if this is not the common column, then you probably will slow >>down whatever other accesses you may have on this table. >> >>After CLUSTER, the current data will stay clustered, but new data will >>not, so you have to continually CLUSTER, the same way that you might >>VACUUM. *However*, IIRC CLUSTER grabs an Exclusive lock, so it is as >>expensive as a VACUUM FULL. Be aware of this, but it might vastly >>improve your performance, so it would be worth it. > > I generally re-cluster once a week. > >>> >>>>So there is no need for preloading your indexes on the identity table. >>>>It is definitely not the bottleneck. >>>> >>>>So a few design bits, which may help your database. >>>>Why is "actor_id" a text field instead of a number? >>> >>>This is simply due to the nature of the data. >>> >> >>I'm just wondering if changing into a number, and using a number->name >>lookup would be faster for you. It may not be. In general, I prefer to >>use numbers for references. I may be over paranoid, but I know that some >>locales are bad with string -> string comparisons. And since the data in >>your database is stored as UNICODE, I'm not sure if it has to do any >>translating or not. Again, something to consider, it may not make any >>difference. > > I don't believe so. I initialze the DB as 'lang=C'. I used to have the > problem where things were being inited as en_US. this would prevent any > text based index from working. This doesn't seem to be the case here, so > I'm not worried about it. > Sorry, I think I was confusing you with someone else who posted SHOW ALL. > > >> ... > it's cached alright. I'm getting a read rate of about 150MB/sec. I would > have thought is would be faster with my raid setup. I think I'm going to > scrap the whole thing and get rid of LVM. I'll just do a straight ext3 > system. Maybe that will help. Still trying to get suggestions for a > stripe size. > I don't think 150MB/s is out of the realm for a 14 drive array. How fast is time dd if=/dev/zero of=testfile bs=8192 count=1000000 (That should create a 8GB file, which is too big to cache everything) And then how fast is: time dd if=testfile of=/dev/null bs=8192 count=1000000 That should give you a semi-decent way of measuring how fast the RAID system is, since it should be too big to cache in ram. > >>I can point you to REINDEX and CLUSTER, but if it is caching in ram, I >>honestly can't say why the per loop would be that much slower. >>Are both systems running the same postgres version? It sounds like it is >>different (since you say something about switching to 8.0). > > These had little or no effect. > The production machine is running 7.4 while the devel machine is running > 8.0 > Well, my concern is that maybe some portion of the 8.0 code actually slowed things down for you. You could try reverting to 7.4 on the devel box, though I think playing with upgrading to 8.1 might be more worthwhile. ... > > this is a cached version. > I assume that you mean this is the second run of the query. I can't compare it too much, since this is "smith" rather than "jones". But this one is 17s rather than the other one being 46s. And that includes having 8k rows instead of having 5k rows. Have you tried other values with disabled nested loops? Because this query (at least in cached form) seems to be *way* faster than with nested loops. I know that you somehow managed to get 200s in your testing, but it might just be that whatever needed to be loaded is now loaded, and you would get better performance. If this is true, it means you might need to tweak some settings, and make sure your statistics are decent, so that postgres can actually pick the optimal plan. > >>copa=> explain analyze select full_name,identity_id,identity.case_id,court.id,date_of_birth,assigned_case_role,litigant_details.impound_litigant_data >>copa-> from identity >>copa-> join litigant_details on identity.actor_id = litigant_details.actor_id >>copa-> join case_data on litigant_details.case_id = case_data.case_id and litigant_details.court_ori = case_data.court_ori >>copa-> join court on identity.court_ori = court.id >>copa-> where identity.court_ori = 'IL081025J' and full_name like 'SMITH%' order by full_name; >> QUERY PLAN >>------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- >> Sort (cost=100502560.72..100502583.47 rows=9099 width=86) (actual time=17843.876..17849.401 rows=8094 loops=1) >> Sort Key: identity.full_name >> -> Merge Join (cost=100311378.72..100501962.40 rows=9099 width=86) (actual time=15195.816..17817.847 rows=8094 loops=1) >> Merge Cond: ((("outer".court_ori)::text = "inner"."?column10?") AND (("outer".case_id)::text = "inner"."?column11?")) >> -> Index Scan using case_speed on case_data (cost=0.00..170424.73 rows=3999943 width=26) (actual time=0.015..4540.525 rows=3018284 loops=1) >> -> Sort (cost=100311378.72..100311400.82 rows=8839 width=112) (actual time=9594.985..9601.174 rows=8094 loops=1) >> Sort Key: (litigant_details.court_ori)::text, (litigant_details.case_id)::text >> -> Nested Loop (cost=100002491.43..100310799.34 rows=8839 width=112) (actual time=6892.755..9555.828 rows=8094 loops=1) >> -> Seq Scan on court (cost=0.00..3.29 rows=1 width=12) (actual time=0.085..0.096 rows=1 loops=1) >> Filter: ('IL081025J'::text = (id)::text) What I don't really understand is the next part. It seems to be doing an index scan on 3.7M rows, and getting very decent performance (5s), and then merging against a table which returns only 8k rows. Why is it having to look through all of those rows? I may be missing something, but this says it is able to do 600 index lookups / millisecond. Which seems superfast. (Compared to your earlier 4ms / lookup) Something fishy is going on here. >> -> Merge Join (cost=2491.43..310707.66 rows=8839 width=113) (actual time=6892.656..9519.680 rows=8094 loops=1) >> Merge Cond: (("outer".actor_id)::text = "inner"."?column7?") >> -> Index Scan using lit_actor_speed on litigant_details (cost=0.00..295722.00 rows=4956820 width=81) (actual time=0.027..5613.814 rows=3736703 loops=1) >> -> Sort (cost=2491.43..2513.71 rows=8913 width=82) (actual time=116.071..122.272 rows=8100 loops=1) >> Sort Key: (identity.actor_id)::text >> -> Index Scan using name_speed on identity (cost=0.00..1906.66 rows=8913 width=82) (actual time=0.133..81.104 rows=8100 loops=1) >> Index Cond: (((full_name)::text >= 'SMITH'::character varying) AND ((full_name)::text < 'SMITI'::character varying)) >> Filter: (((court_ori)::text = 'IL081025J'::text) AND ((full_name)::text ~~ 'SMITH%'::text)) >> Total runtime: 17859.917 ms > > >>But really, you have worse index speed, and that needs to be figured out. >> >>John >>=:-> I'm assuming your data is private (since it looks like legal stuff). Unless maybe that makes it part of the public record. Anyway, I'm not able to, but sometimes someone like Tom can profile stuff to see what is going on. I might just be messing up my ability to read the explain output. But somehow things don't seem to be lining up with the cost of a single index lookup. On my crappy Celeron 450 box, an index lookup is 0.06ms once things are cached in ram. John =:-> --------------enigE5E6EAB78A66B0E023BE7ECF Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDBkkIJdeBCYSNAAMRAvkXAKDT8tRVb2b85F6RjhGvMorKv39YfQCaA4HD uFJc5g6grmnac0RjuIt1vp4= =3LUc -----END PGP SIGNATURE----- --------------enigE5E6EAB78A66B0E023BE7ECF-- From pgsql-performance-owner@postgresql.org Fri Aug 19 18:12:56 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D639552E04 for ; Fri, 19 Aug 2005 18:07:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 07214-09 for ; Fri, 19 Aug 2005 21:07:24 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id E7B3352C61 for ; Fri, 19 Aug 2005 18:07:23 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050819210724m9100g7aoge>; Fri, 19 Aug 2005 21:07:24 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 54A0355FBA; Fri, 19 Aug 2005 16:07:24 -0500 (CDT) Received: from [192.168.1.103] (63-230-159-164.cdrr.qwest.net [63.230.159.164]) by juju.arbash-meinel.com (Postfix) with ESMTP id 37DCA55FAD; Fri, 19 Aug 2005 16:07:20 -0500 (CDT) Message-ID: <43064A06.8020102@arbash-meinel.com> Date: Fri, 19 Aug 2005 16:07:18 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Jeremiah Jahn Cc: Ron , postgres performance Subject: Re: extremly low memory usage References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> <43063196.9000001@arbash-meinel.com> <1124485299.27881.192.camel@bluejay.goodinassociates.com> In-Reply-To: <1124485299.27881.192.camel@bluejay.goodinassociates.com> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig0B481C51B574C85D434A19BC" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.027 required=5 tests=[AWL=-0.023, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/283 X-Sequence-Number: 14030 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig0B481C51B574C85D434A19BC Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Jeremiah Jahn wrote: > Rebuild in progress with just ext3 on the raid array...will see if this > helps the access times. If it doesn't I'll mess with the stripe size. I > have REINDEXED, CLUSTERED, tablespaced and cached with 'cat table/index > >>/dev/null' none of this seems to have helped, or even increased my > > memory usage. argh! The only thing about this new system that I'm > unfamiliar with is the array setup and LVM, which is why I think that's > where the issue is. clustering and indexing as well as vacuum etc are > things that I do and have been aware of for sometime. Perhaps slony is a > factor, but I really don't see it causing problems on index read speed > esp. when it's not running. > > thanx for your help, I really appreciate it. > -jj- > By the way, how are you measuring memory usage? Can you give the output of that command, just to make sure you are reading it correctly. John =:-> --------------enig0B481C51B574C85D434A19BC Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDBkoGJdeBCYSNAAMRAkUiAKCAOu6b6h8oee4Tv/WB0VDtVOF2OACfaDp3 Cx9Hw+sPlGNxaw1/tZ8Mxv4= =IxL1 -----END PGP SIGNATURE----- --------------enig0B481C51B574C85D434A19BC-- From pgsql-performance-owner@postgresql.org Fri Aug 19 18:21:25 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A50CD52B9D for ; Fri, 19 Aug 2005 18:19:07 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 26812-05 for ; Fri, 19 Aug 2005 21:19:04 +0000 (GMT) Received: from smtpauth07.mail.atl.earthlink.net (smtpauth07.mail.atl.earthlink.net [209.86.89.67]) by svr1.postgresql.org (Postfix) with ESMTP id 144C152B6D for ; Fri, 19 Aug 2005 18:19:03 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=oqD3jsLMocgjvyXVw03LXiQNwxXiPl6I2cw0zve2vXW22gUwyIHeNjE9/+vrMzeP; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:Cc:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth07.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E6EGX-000856-0z; Fri, 19 Aug 2005 17:19:05 -0400 Message-Id: <6.2.3.4.0.20050819163215.05ba0a98@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Fri, 19 Aug 2005 17:19:00 -0400 To: Alex Turner From: Ron Subject: Re: sustained update load of 1-2k/sec Cc: postgres performance In-Reply-To: <33c6269f0508191231b12f5b@mail.gmail.com> References: <20050819082404.53751.qmail@web32915.mail.mud.yahoo.com> <33c6269f050819054038af9886@mail.gmail.com> <6.2.3.4.0.20050819093011.05c1b9d8@pop.earthlink.net> <33c6269f0508191231b12f5b@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bc8b6ab3b984c581f56246ba50e4529136350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.548 required=5 tests=[AWL=0.174, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/284 X-Sequence-Number: 14031 At 03:31 PM 8/19/2005, Alex Turner wrote: >Don't forget that Ultra 320 is the speed of the bus, not each drive. >No matter how many honking 15k disks you put on a 320MB bus, you can >only get 320MB/sec! and have so many outstanding IO/s on the bus. Of course. This is exactly why multi-channel SCSI and multichannel Fibre Channel cards exist; and why external RAID enclosures usually have multiple such cards in them... Even moderately acceptable U320 SCSI cards are dual channel at this point (think Adaptec dual channel AHAxxxx's), and Quad channel ones are just as common. The Quads will, of course, saturate a 64b 133MHz PCI-X bus. _IF_ the chipset on them can keep up. The current kings of RAID card performance are all Fibre Channel based, and all the ones I know of are theoretically capable of saturating a 64b 133MHz PCI-X bus. Again, _IF_ the chipset on them can keep up. Most commodity RAID card have neither adequate CPU nor enough buffer. Regardless of the peripheral IO technology they use. >Not so with SATA! Each drive is on it's own bus, and you are only >limited by the speed of your PCI-X Bus, which can be as high as >800MB/sec at 133Mhz/64bit. That's the Theory anyway, and latency should be lower as well. OTOH, as my wife likes to say "In theory, Theory and Practice are the same. In practice, they almost never are." You are only getting the performance you mention as long as your card can keep up with multiplexing N IO streams, crunching RAID 5 XORs (assuming you are using RAID 5), etc, etc. As I'm sure you know, "The chain is only as strong as its weakest link.". Most commodity SATA RAID cards brag about being able to pump 300MB/s (they were all over LW SF bragging about this!?), which in this context is woefully unimpressive. Sigh. I'm impressed with the Areca cards because they usually have CPUs that actually can come close to pushing the theoretical IO limit of the bus they are plugged into; and they can be upgraded to (barely) acceptable buffer amounts (come on, manufacturers! 4GB of DDR PC3200 is only -2- DIMMs, and shortly that will be enough to hold 8GB of DDR PC3200. Give us more buffer!). >It's cheap and it's fast - all you have to do is pay for the >enclosure, which can be a bit pricey, but there are some nice 24bay >and even 40bay enclosures out there for SATA. I've even seen 48 bay ones. However, good enclosures, particularly for larger numbers of HDs, are examples of non-trivial engineering and priced accordingly. Too many times I see people buy "bargain" enclosures and set themselves and their organizations up for some _very_ unpleasant times that could easily have been avoided by being careful to buy quality products. "Pay when you buy or pay much more later." >Yes a 15k RPM drive will give you better seek time and better peak >through put, but put them all on a single U320 bus and you won't see >much return past a stripe size of 3 or 4 Agreed. Same holds for 2Gbps FC. Haven't tested 4Gbps FC personally yet, but I'm told the limit is higher in the manner you'd expect. >If it's raw transactions per second data warehouse style, it's all >about the xlog baby which is sequential writes, and all about large >block reads, which is sequential reads. > >Alex Turner >NetEconomist >P.S. Sorry if i'm a bit punchy, I've been up since yestarday with >server upgrade nightmares that continue ;) My condolences and sympathies. I've definitely been there and done that. Ron Peacetree From pgsql-performance-owner@postgresql.org Fri Aug 19 20:02:43 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C2DE952CB2 for ; Fri, 19 Aug 2005 20:02:40 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 05681-09 for ; Fri, 19 Aug 2005 23:02:36 +0000 (GMT) Received: from smtp106.mail.sc5.yahoo.com (smtp106.mail.sc5.yahoo.com [66.163.169.226]) by svr1.postgresql.org (Postfix) with SMTP id 7C89952B6D for ; Fri, 19 Aug 2005 20:02:35 -0300 (ADT) Received: (qmail 86809 invoked from network); 19 Aug 2005 23:02:37 -0000 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Received:User-Agent:Date:Subject:From:To:CC:Message-ID:In-Reply-To:Mime-version:Content-type:Content-transfer-encoding; b=1OToURhp9PxWy6HIdRVIavt+C59i+wcyRtNCFkXeA3GBrqajs+NcNQkPc5WoHrEYBhMYsOvQ+df5uJRtiAbGp/R3r85PBggLlTJ1J3pLMDCRLWcEPrfrqp7R8SoAFansDwd1yH7JGGY7BCyuIzYqXC91s00EtkHmK0VsKSr13r4= ; Received: from unknown (HELO ?10.3.101.47?) (mcotner@24.248.72.254 with login) by smtp106.mail.sc5.yahoo.com with SMTP; 19 Aug 2005 23:02:37 -0000 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Fri, 19 Aug 2005 19:02:35 -0400 Subject: Re: sustained update load of 1-2k/sec From: Mark Cotner To: PFC , Tom Lane , Bob Ippolito Cc: Message-ID: In-Reply-To: Mime-version: 1.0 Content-type: text/plain; charset="US-ASCII" Content-transfer-encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.374 required=5 tests=[DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/285 X-Sequence-Number: 14032 :) Most of the ppl on this list are systems programmers, however I am not. The tool of choice for this app is Ruby and the libraries don't support async SNMP at the moment. I've done a good deal of async snmp and the libraries that actually pull it off generally aren't that good(Net-SNMP and Perl's Net::SNMP). Granted, UDP is connectionless to an extent, but you still have to send the PDU, and bind to the return socket and wait. If you batch the outgoing PDUs then you can get away with sending them out synchronously and listening on the returning socket synchronously, but it would require that your libraries support this. I understand the concepts well enough, maybe I'll put together a patch. It would be much lower overhead than managing all those threads. Looks like it's gonna be a fun weekend. Thanks again for all the great feedback. 'njoy, Mark On 8/19/05 2:11 PM, "PFC" wrote: > >> While I agree that hundreds of threads seems like overkill, I think the >> above advice might be going too far in the other direction. The problem >> with single-threaded operation is that any delay affects the whole >> system --- eg, if you're blocked waiting for disk I/O, the CPU doesn't > > You use UDP which is a connectionless protocol... then why use threads ? > > I'd advise this : > > Use asynchronous network code (one thread) to do your network stuff. This > will lower the CPU used by this code immensely. > Every minute, dump a file contianing everything to insert into the table. > Use another thread to COPY it into the DB, in a temporary table if you > wish, and then INSERT INTO ... SELECT. > This should be well adapted to your requirements. > > ---------------------------(end of broadcast)--------------------------- > TIP 2: Don't 'kill -9' the postmaster From pgsql-performance-owner@postgresql.org Sat Aug 20 03:52:19 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9F67B52E31 for ; Sat, 20 Aug 2005 03:52:16 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 92759-05 for ; Sat, 20 Aug 2005 06:52:13 +0000 (GMT) Received: from crestone.coronasolutions.com (crestone.coronasolutions.com [66.45.104.24]) by svr1.postgresql.org (Postfix) with ESMTP id 2960B52A04 for ; Sat, 20 Aug 2005 03:52:10 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by crestone.coronasolutions.com (Postfix) with ESMTP id 338B464413E for ; Sat, 20 Aug 2005 00:51:38 -0600 (MDT) Received: from crestone.coronasolutions.com ([127.0.0.1]) by localhost (crestone.coronasolutions.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 12088-02 for ; Sat, 20 Aug 2005 00:51:36 -0600 (MDT) Received: from [192.168.1.12] (c-24-9-24-35.hsd1.co.comcast.net [24.9.24.35]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (No client certificate requested) by crestone.coronasolutions.com (Postfix) with ESMTP id 192C164411F for ; Sat, 20 Aug 2005 00:51:36 -0600 (MDT) Mime-Version: 1.0 (Apple Message framework v733) In-Reply-To: <1124434535.3257.7.camel@noodles> References: <1124434535.3257.7.camel@noodles> Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: Content-Transfer-Encoding: 7bit From: Dan Harris Subject: Re: Query plan looks OK, but slow I/O - settings advice? Date: Sat, 20 Aug 2005 00:52:08 -0600 To: pgsql-performance@postgresql.org X-Mailer: Apple Mail (2.733) X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at drivefaster.net X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/286 X-Sequence-Number: 14033 On Aug 19, 2005, at 12:55 AM, Jeffrey W. Baker wrote: > On Tue, 2005-08-16 at 10:46 -0700, Roger Hand wrote: > Have you considered booting your > machine with elevator=deadline? Although I'm not the OP for this problem, I thought I'd try it out. WOW.. this should be in a Pg tuning guide somewhere. I added this to my server tonight just for kicks and saw a pronounced improvement in IO performance. Thank you very much for mentioning this on the list. I didn't have a long enough maintenance window to do solid benchmarking, but I can say for certain that the change was noticeable, especially in VACUUM operations. Specs for the server: PG 8.0.1 Linux 2.6.12-3 kernel 4xOpteron 2.2 12GB RAM 16-drive RAID 10 XFS mounted with noatime pg_xlog on separate RAID controller -Dan From pgsql-performance-owner@postgresql.org Sat Aug 20 04:12:29 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0B50D52E4E for ; Sat, 20 Aug 2005 04:12:20 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 54839-06 for ; Sat, 20 Aug 2005 07:12:17 +0000 (GMT) Received: from crestone.coronasolutions.com (crestone.coronasolutions.com [66.45.104.24]) by svr1.postgresql.org (Postfix) with ESMTP id 838DB52E4A for ; Sat, 20 Aug 2005 04:12:16 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by crestone.coronasolutions.com (Postfix) with ESMTP id 5A28664413E for ; Sat, 20 Aug 2005 01:11:45 -0600 (MDT) Received: from crestone.coronasolutions.com ([127.0.0.1]) by localhost (crestone.coronasolutions.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 12088-04 for ; Sat, 20 Aug 2005 01:11:43 -0600 (MDT) Received: from [192.168.1.12] (c-24-9-24-35.hsd1.co.comcast.net [24.9.24.35]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (No client certificate requested) by crestone.coronasolutions.com (Postfix) with ESMTP id 45A7E64411F for ; Sat, 20 Aug 2005 01:11:43 -0600 (MDT) Mime-Version: 1.0 (Apple Message framework v733) In-Reply-To: <1124485299.27881.192.camel@bluejay.goodinassociates.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> <43063196.9000001@arbash-meinel.com> <1124485299.27881.192.camel@bluejay.goodinassociates.com> Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: Content-Transfer-Encoding: 7bit From: Dan Harris Subject: Re: extremly low memory usage Date: Sat, 20 Aug 2005 01:12:15 -0600 To: pgsql-performance@postgresql.org X-Mailer: Apple Mail (2.733) X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at drivefaster.net X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/287 X-Sequence-Number: 14034 On Aug 19, 2005, at 3:01 PM, Jeremiah Jahn wrote: > Rebuild in progress with just ext3 on the raid array...will see if > this > helps the access times. From my recent experiences, I can say ext3 is probably not a great choice for Pg databases. If you check the archives you'll see there's a lot of discussion about various journalling filesystems and ext3 usually(always?) comes up on the bottom as far as performance goes. If you insist on using it, I would at least recommend the noatime option in fstab and using data=writeback to get the faster of the journal modes. XFS seems to be a trusted choice, followed by Reiser and JFS both with the occasional controversy when the comparisons pop up. -Dan From pgsql-performance-owner@postgresql.org Sat Aug 20 08:18:05 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 76CCB52DEF for ; Sat, 20 Aug 2005 08:18:04 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 75198-07 for ; Sat, 20 Aug 2005 11:18:00 +0000 (GMT) Received: from fep32-app.kolumbus.fi (fep32-0.kolumbus.fi [193.229.0.63]) by svr1.postgresql.org (Postfix) with ESMTP id 60CCE52DE2 for ; Sat, 20 Aug 2005 08:17:56 -0300 (ADT) Received: from [192.168.1.101] (really [80.186.78.229]) by fep32-app.kolumbus.fi with ESMTP id <20050820111758.QSZF27630.fep32-app.kolumbus.fi@[192.168.1.101]> for ; Sat, 20 Aug 2005 14:17:58 +0300 Message-ID: <43071162.6050405@kolumbus.fi> Date: Sat, 20 Aug 2005 14:17:54 +0300 From: Marko Ristola User-Agent: Debian Thunderbird 1.0.2 (X11/20050331) X-Accept-Language: en-us, en MIME-Version: 1.0 Cc: pgsql-performance@postgresql.org Subject: Re: extremly low memory usage References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> <43063196.9000001@arbash-meinel.com> <1124485299.27881.192.camel@bluejay.goodinassociates.com> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.128 required=5 tests=[AWL=0.078, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/288 X-Sequence-Number: 14035 Dan Harris wrote: > From my recent experiences, I can say ext3 is probably not a great > choice for Pg databases. If you check the archives you'll see > there's a lot of discussion about various journalling filesystems and > ext3 usually(always?) comes up on the bottom as far as performance > goes. If you insist on using it, I would at least recommend the > noatime option in fstab and using data=writeback to get the faster Based on my knoledge, Ext3 is good with keeping filesystem integrity AND data integrity while pressing the reset button. However, by selecting data=writeback, you gain more speed, but you risk the data integrity during a crash: Ext3 garantees only filesystem integrity. This means with database transaction logs: The last transactions are not guaranteed to be written into the hard drives during a hardware reset, meaning of a loss of some committed transactions. Reiserfs is known to do things this false way also. Is there a way with a Reiserfs filesystem to fulfill both filesystem AND data integrity requirements nowadays? See for example "man mount" to see the effects of data=journal, data=ordered(default) and data=writeback for Ext3. Only the writeback risks data integrity. Ext3 is the only journaled filesystem, that I know that fulfills these fundamental data integrity guarantees. Personally I like about such filesystems, even though it means less speed. Marko Ristola From pgsql-performance-owner@postgresql.org Sat Aug 20 09:15:24 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4594C52AE9 for ; Sat, 20 Aug 2005 09:15:08 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 77442-02 for ; Sat, 20 Aug 2005 12:15:05 +0000 (GMT) Received: from vms048pub.verizon.net (vms048pub.verizon.net [206.46.252.48]) by svr1.postgresql.org (Postfix) with ESMTP id 9A53052A38 for ; Sat, 20 Aug 2005 09:15:02 -0300 (ADT) Received: from osgiliath.mathom.us ([70.108.53.154]) by vms048.mailsrvcs.net (Sun Java System Messaging Server 6.2 HotFix 0.04 (built Dec 24 2004)) with ESMTPA id <0ILI00F9BSP59W47@vms048.mailsrvcs.net> for pgsql-performance@postgresql.org; Sat, 20 Aug 2005 07:15:06 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by osgiliath.mathom.us (Postfix) with ESMTP id 1C09E6050CD; Sat, 20 Aug 2005 08:15:05 -0400 (EDT) Received: from osgiliath.mathom.us ([127.0.0.1]) by localhost (osgiliath [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 27694-04; Sat, 20 Aug 2005 08:15:04 -0400 (EDT) Received: by osgiliath.mathom.us (Postfix, from userid 1000) id DCD24600498; Sat, 20 Aug 2005 08:15:04 -0400 (EDT) Date: Sat, 20 Aug 2005 08:15:04 -0400 From: Michael Stone Subject: Re: Query plan looks OK, but slow I/O - settings advice? In-reply-to: To: Dan Harris Cc: pgsql-performance@postgresql.org Mail-Followup-To: Dan Harris , pgsql-performance@postgresql.org Message-id: <20050820121504.GL19080@mathom.us> MIME-version: 1.0 Content-type: text/plain; charset=us-ascii; format=flowed Content-disposition: inline X-Pgp-Fingerprint: 53 FF 38 00 E7 DD 0A 9C 84 52 84 C5 EE DF 7C 88 X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mathom.us References: <1124434535.3257.7.camel@noodles> User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 required=5 tests=[AWL=0.005] X-Spam-Level: X-Archive-Number: 200508/289 X-Sequence-Number: 14036 On Sat, Aug 20, 2005 at 12:52:08AM -0600, Dan Harris wrote: >On Aug 19, 2005, at 12:55 AM, Jeffrey W. Baker wrote: >> Have you considered booting your >>machine with elevator=deadline? > >Although I'm not the OP for this problem, I thought I'd try it out. >WOW.. this should be in a Pg tuning guide somewhere. [snip] >16-drive RAID 10 Yeah, the default scheduler tries to optimize disk access patterns for a single-spindle setup, and actually makes things worse if you have a device with multiple spindles. Mike Stone From pgsql-performance-owner@postgresql.org Sat Aug 20 09:18:37 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2D71A5299B for ; Sat, 20 Aug 2005 09:18:35 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 77264-09 for ; Sat, 20 Aug 2005 12:18:30 +0000 (GMT) Received: from floppy.pyrenet.fr (news.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id 2871052936 for ; Sat, 20 Aug 2005 09:18:29 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id 5244130B42; Sat, 20 Aug 2005 14:30:08 +0200 (MET DST) From: ohp@pyrenet.fr X-Newsgroups: pgsql.performance Subject: index as large as table Date: Sat, 20 Aug 2005 14:18:29 +0200 Organization: PYRENET Midi-pyrenees Provider Lines: 26 Message-ID: Reply-To: ohp@pyrenet.fr Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Complaints-To: abuse@pyrenet.fr X-X-Sender: ohp@sun.pyrenet To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.178 required=5 tests=[NO_REAL_NAME=0.178] X-Spam-Level: X-Archive-Number: 200508/291 X-Sequence-Number: 14038 Hi, While testing 8.1dev I came to this: CREATE TABLE t ( a int, b int PRIMARY KEY (a,b)); In that case, the index is as big as the table. My question is is it worthwhile to have such index peformance wise. I understand I'd loose uniqness buthas such an index any chance to be used against seq scan. Is there any chance we have a "btree table" in the future for that case? Regards, -- Olivier PRENANT Tel: +33-5-61-50-97-00 (Work) 15, Chemin des Monges +33-5-61-50-97-01 (Fax) 31190 AUTERIVE +33-6-07-63-80-64 (GSM) FRANCE Email: ohp@pyrenet.fr ------------------------------------------------------------------------------ Make your life a dream, make your dream a reality. (St Exupery) From pgsql-performance-owner@postgresql.org Sat Aug 20 09:18:36 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7F76A52996 for ; Sat, 20 Aug 2005 09:18:34 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 89706-04 for ; Sat, 20 Aug 2005 12:18:31 +0000 (GMT) Received: from vms044pub.verizon.net (vms044pub.verizon.net [206.46.252.44]) by svr1.postgresql.org (Postfix) with ESMTP id 08C6952945 for ; Sat, 20 Aug 2005 09:18:30 -0300 (ADT) Received: from osgiliath.mathom.us ([70.108.53.154]) by vms044.mailsrvcs.net (Sun Java System Messaging Server 6.2 HotFix 0.04 (built Dec 24 2004)) with ESMTPA id <0ILI00BC7SUX0QII@vms044.mailsrvcs.net> for pgsql-performance@postgresql.org; Sat, 20 Aug 2005 07:18:34 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by osgiliath.mathom.us (Postfix) with ESMTP id 604036050CD for ; Sat, 20 Aug 2005 08:18:33 -0400 (EDT) Received: from osgiliath.mathom.us ([127.0.0.1]) by localhost (osgiliath [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 27878-03-2 for ; Sat, 20 Aug 2005 08:18:33 -0400 (EDT) Received: by osgiliath.mathom.us (Postfix, from userid 1000) id 360D8600498; Sat, 20 Aug 2005 08:18:33 -0400 (EDT) Date: Sat, 20 Aug 2005 08:18:33 -0400 From: Michael Stone Subject: Re: extremly low memory usage In-reply-to: To: pgsql-performance@postgresql.org Mail-Followup-To: pgsql-performance@postgresql.org Message-id: <20050820121833.GM19080@mathom.us> MIME-version: 1.0 Content-type: text/plain; charset=us-ascii; format=flowed Content-disposition: inline X-Pgp-Fingerprint: 53 FF 38 00 E7 DD 0A 9C 84 52 84 C5 EE DF 7C 88 X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mathom.us References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> <43063196.9000001@arbash-meinel.com> <1124485299.27881.192.camel@bluejay.goodinassociates.com> User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 required=5 tests=[AWL=0.005] X-Spam-Level: X-Archive-Number: 200508/290 X-Sequence-Number: 14037 On Sat, Aug 20, 2005 at 01:12:15AM -0600, Dan Harris wrote: >XFS seems to be a trusted choice, followed by Reiser and JFS both >with the occasional controversy when the comparisons pop up. And don't put the xlog on a journaled filesystem. There is no advantage to doing so, and it will slow things down. (Assuming a sane seperate xlog partition configuration, sized reasonably.) Mike Stone From pgsql-performance-owner@postgresql.org Sat Aug 20 09:23:09 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 519C2528E7 for ; Sat, 20 Aug 2005 09:23:07 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 90582-06 for ; Sat, 20 Aug 2005 12:23:03 +0000 (GMT) Received: from vms046pub.verizon.net (vms046pub.verizon.net [206.46.252.46]) by svr1.postgresql.org (Postfix) with ESMTP id ACB5452849 for ; Sat, 20 Aug 2005 09:23:02 -0300 (ADT) Received: from osgiliath.mathom.us ([70.108.53.154]) by vms046.mailsrvcs.net (Sun Java System Messaging Server 6.2 HotFix 0.04 (built Dec 24 2004)) with ESMTPA id <0ILI0044IT2H0RD6@vms046.mailsrvcs.net> for pgsql-performance@postgresql.org; Sat, 20 Aug 2005 07:23:06 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by osgiliath.mathom.us (Postfix) with ESMTP id 91A3B6050CD for ; Sat, 20 Aug 2005 08:23:05 -0400 (EDT) Received: from osgiliath.mathom.us ([127.0.0.1]) by localhost (osgiliath [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 27694-05 for ; Sat, 20 Aug 2005 08:23:05 -0400 (EDT) Received: by osgiliath.mathom.us (Postfix, from userid 1000) id 5F459600498; Sat, 20 Aug 2005 08:23:05 -0400 (EDT) Date: Sat, 20 Aug 2005 08:23:05 -0400 From: Michael Stone Subject: Re: extremly low memory usage In-reply-to: <43071162.6050405@kolumbus.fi> To: pgsql-performance@postgresql.org Mail-Followup-To: pgsql-performance@postgresql.org Message-id: <20050820122305.GN19080@mathom.us> MIME-version: 1.0 Content-type: text/plain; charset=us-ascii; format=flowed Content-disposition: inline X-Pgp-Fingerprint: 53 FF 38 00 E7 DD 0A 9C 84 52 84 C5 EE DF 7C 88 X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mathom.us References: <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> <43063196.9000001@arbash-meinel.com> <1124485299.27881.192.camel@bluejay.goodinassociates.com> <43071162.6050405@kolumbus.fi> User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 required=5 tests=[AWL=0.005] X-Spam-Level: X-Archive-Number: 200508/292 X-Sequence-Number: 14039 On Sat, Aug 20, 2005 at 02:17:54PM +0300, Marko Ristola wrote: >Based on my knoledge, Ext3 is good with keeping filesystem integrity >AND data integrity while pressing the reset button. However, by >selecting data=writeback, you gain more speed, but you risk the data >integrity during a crash: Ext3 garantees only filesystem integrity. That's why postgres keeps its own transaction log. Any of these filesystems guarantee data integrity for data that's been synced to disk, and postgres keeps track of what data has and has not been committed so it can recover gracefully from a failure. That's why most filesystems are designed the way they are; the application can determine what things need better data integrity and which need better performance on a case-by-case basis. Mike Stone From pgsql-performance-owner@postgresql.org Sat Aug 20 10:08:20 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E637A529C7 for ; Sat, 20 Aug 2005 10:08:18 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 24690-05 for ; Sat, 20 Aug 2005 13:08:14 +0000 (GMT) Received: from linuxworld.com.au (unknown [203.34.46.50]) by svr1.postgresql.org (Postfix) with ESMTP id BE0ED5299B for ; Sat, 20 Aug 2005 10:08:10 -0300 (ADT) Received: from linuxworld.com.au (IDENT:swm@localhost.localdomain [127.0.0.1]) by linuxworld.com.au (8.13.2/8.13.2) with ESMTP id j7KD8ES8004254; Sat, 20 Aug 2005 23:08:14 +1000 Received: from localhost (swm@localhost) by linuxworld.com.au (8.13.2/8.13.2/Submit) with ESMTP id j7KD8D3q004251; Sat, 20 Aug 2005 23:08:13 +1000 X-Authentication-Warning: linuxworld.com.au: swm owned process doing -bs Date: Sat, 20 Aug 2005 23:08:13 +1000 (EST) From: Gavin Sherry X-X-Sender: swm@linuxworld.com.au To: ohp@pyrenet.fr Cc: pgsql-performance@postgresql.org Subject: Re: index as large as table In-Reply-To: Message-ID: References: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.002 required=5 tests=[AWL=0.002] X-Spam-Level: X-Archive-Number: 200508/293 X-Sequence-Number: 14040 On Sat, 20 Aug 2005 ohp@pyrenet.fr wrote: > Hi, > > While testing 8.1dev I came to this: > > CREATE TABLE t ( > a int, > b int > PRIMARY KEY (a,b)); > > In that case, the index is as big as the table. Right. Think about it: the index must store a, b, a reference to the data in the table itself and index meta data. If an index is defined across all columns of the table, it must be bigger than the table itself. (In PostgreSQL, when the table is small, the index will be smaller still. This is because of each entry in the table itself has meta data. But the amount of data per row of a table remains constant, whereas, the amount of metadata in an index grows.) > My question is is it worthwhile to have such index peformance wise. > I understand I'd loose uniqness buthas such an index any chance to be used > against seq scan. Of course. The idea is that, generally speaking, you're only interested in a small portion of the data stored in the table. Indexes store extra data so that they can locate the portion you're interested in faster. Gavin From pgsql-performance-owner@postgresql.org Sat Aug 20 11:41:07 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2484252E20 for ; Sat, 20 Aug 2005 11:40:50 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 90422-07 for ; Sat, 20 Aug 2005 14:40:41 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id E0C9352C63 for ; Sat, 20 Aug 2005 11:40:39 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7KEef86023045; Sat, 20 Aug 2005 10:40:44 -0400 (EDT) To: Michael Stone Cc: pgsql-performance@postgresql.org Subject: Re: extremly low memory usage In-reply-to: <20050820122305.GN19080@mathom.us> References: <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> <43063196.9000001@arbash-meinel.com> <1124485299.27881.192.camel@bluejay.goodinassociates.com> <43071162.6050405@kolumbus.fi> <20050820122305.GN19080@mathom.us> Comments: In-reply-to Michael Stone message dated "Sat, 20 Aug 2005 08:23:05 -0400" Date: Sat, 20 Aug 2005 10:40:41 -0400 Message-ID: <23044.1124548841@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/294 X-Sequence-Number: 14041 Michael Stone writes: > On Sat, Aug 20, 2005 at 02:17:54PM +0300, Marko Ristola wrote: >> Based on my knoledge, Ext3 is good with keeping filesystem integrity >> AND data integrity while pressing the reset button. However, by >> selecting data=writeback, you gain more speed, but you risk the data >> integrity during a crash: Ext3 garantees only filesystem integrity. > That's why postgres keeps its own transaction log. Any of these > filesystems guarantee data integrity for data that's been synced to > disk, and postgres keeps track of what data has and has not been > committed so it can recover gracefully from a failure. Right. I think the optimal setting for a Postgres data directory is journaled metadata, non-journaled file content. Postgres can take care of the data integrity for itself, but it does assume that the filesystem stays structurally sane (eg, data blocks don't get reassigned to the wrong file), so you need a filesystem guarantee about the metadata. WAL files are handled in a much more conservative way (created, filled with zeroes, and fsync'd before we ever put any valuable data in 'em). If you have WAL on its own drive then I think Mike's recommendation of no filesystem journalling at all for that drive is probably OK. Or you can do same as above (journal metadata only) if you want a little extra protection. And of course all this reasoning depends on the assumption that the drive tells the truth about write-completion. If the drive does write caching it had better be able to complete all its accepted writes before dying in a power failure. (Hence, battery-backed write cache is OK, any other kind is evil.) regards, tom lane From pgsql-performance-owner@postgresql.org Sat Aug 20 12:11:05 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5DE1052F0F for ; Sat, 20 Aug 2005 12:10:49 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 70818-04 for ; Sat, 20 Aug 2005 15:10:31 +0000 (GMT) Received: from cassarossa.samfundet.no (cassarossa.samfundet.no [129.241.93.19]) by svr1.postgresql.org (Postfix) with ESMTP id 5230D52DC1 for ; Sat, 20 Aug 2005 12:10:29 -0300 (ADT) Received: from trofast.sesse.net ([129.241.93.32]) by cassarossa.samfundet.no with esmtp (Exim 4.50) id 1E6UzK-00036J-Bt for pgsql-performance@postgresql.org; Sat, 20 Aug 2005 17:10:27 +0200 Received: from sesse by trofast.sesse.net with local (Exim 3.36 #1 (Debian)) id 1E6UzJ-0003Zr-00 for ; Sat, 20 Aug 2005 17:10:25 +0200 Date: Sat, 20 Aug 2005 17:10:25 +0200 From: "Steinar H. Gunderson" To: pgsql-performance@postgresql.org Subject: Re: index as large as table Message-ID: <20050820151025.GA12890@uio.no> Mail-Followup-To: pgsql-performance@postgresql.org References: Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: X-Operating-System: Linux 2.6.11.8 on a i686 X-Message-Flag: Outlook? --> http://www.mozilla.org/products/thunderbird/ User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.016 required=5 tests=[AWL=0.016] X-Spam-Level: X-Archive-Number: 200508/295 X-Sequence-Number: 14042 On Sat, Aug 20, 2005 at 11:08:13PM +1000, Gavin Sherry wrote: > Of course. The idea is that, generally speaking, you're only interested in > a small portion of the data stored in the table. Indexes store extra data > so that they can locate the portion you're interested in faster. I think his question was more why you needed the data in itself, when you had everything you needed in the index anyway. (Actually, you don't -- indexes don't carry MVCC information, but I guess that's a bit beside the point.) There has been discussion on "heap tables" or whatever you'd want to call them (ie. tables that are organized as a B+-tree on some index) here before; I guess the archives would be a reasonable place to start looking. /* Steinar */ -- Homepage: http://www.sesse.net/ From pgsql-performance-owner@postgresql.org Sat Aug 20 12:59:56 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CB7B952C91 for ; Sat, 20 Aug 2005 12:59:50 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 84932-07 for ; Sat, 20 Aug 2005 15:59:47 +0000 (GMT) Received: from smtpauth08.mail.atl.earthlink.net (smtpauth08.mail.atl.earthlink.net [209.86.89.68]) by svr1.postgresql.org (Postfix) with ESMTP id 07F8B52C17 for ; Sat, 20 Aug 2005 12:59:45 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=RdLPRkFJX3Pw0igNikvZ0LZeJGMM2ClgT2bxDZYtgUFK6lCF4MgQph6NNBxA1kFD; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:Cc:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth08.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E6Vku-0007IP-L5; Sat, 20 Aug 2005 11:59:36 -0400 Message-Id: <6.2.3.4.0.20050819233655.05c993a0@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Sat, 20 Aug 2005 11:59:32 -0400 To: Jeremiah Jahn From: Ron Subject: Re: extremly low memory usage Cc: postgres performance In-Reply-To: <1124482288.27881.182.camel@bluejay.goodinassociates.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> <43063196.9000001@arbash-meinel.com> <1124482288.27881.182.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bc076e50b2875ad8579b2fef1c8c724c89350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.521 required=5 tests=[AWL=0.147, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/296 X-Sequence-Number: 14043 At 04:11 PM 8/19/2005, Jeremiah Jahn wrote: >On Fri, 2005-08-19 at 14:23 -0500, John A Meinel wrote: > > Ron wrote: > > > At 01:18 PM 8/19/2005, John A Meinel wrote: > > > > > >> Jeremiah Jahn wrote: > > >> > Sorry about the formatting. > > >> > > > >> > On Thu, 2005-08-18 at 12:55 -0500, John Arbash Meinel wrote: > > >> > > > >> >>Jeremiah Jahn wrote: > > >> >> > > >> >> > > >> > > >> ... > > >> > > >> >>The expensive parts are the 4915 lookups into the litigant_details > > >> (each > > >> >>one takes approx 4ms for a total of ~20s). > > >> >>And then you do it again on case_data (average 3ms each * 4906 loops = > > >> >>~15s). > > >> > > > >> > Is there some way to avoid this? > > >> > > > >> > > >> Well, in general, 3ms for a single lookup seems really long. Maybe your > > >> index is bloated by not vacuuming often enough. Do you tend to get a lot > > >> of updates to litigant_details? > > > > > > > > > Given that the average access time for a 15Krpm HD is in the 5.5-6ms > > > range (7.5-8ms for a 10Krpm HD), having an average of 3ms for a single > > > lookup implies that ~1/2 (the 15Krpm case) or ~1/3 (the 10Krpm case) > > > table accesses is requiring a seek. > > > >I think LVM may be a problem, since it also seems to break things up on >the file system. My access time on the seek should be around 1/7th the >15Krpm I believe since it's a 14 disk raid 10 array. And no other >traffic at the moment. Oops. There's a misconception here. RAID arrays increase _throughput_ AKA _bandwidth_ through parallel access to HDs. OTOH, access time is _latency_, and that is not changed. Access time for a RAID set is equal to that of the slowest access time, AKA highest latency, HD in the RAID set. > > Well, from what he has said, the total indexes are < 1GB and he has 6GB > > of ram. So everything should fit. Not to mention he is only accessing > > 5000/several million rows. >I table spaced some of the indexes and they are around 211066880 bytes >for the name_speed index and 149825330 for the lit_actor_speed index >tables seem to be about a gig. Hmm. And you think you are only using 250MB out of your 6GB of RAM? Something doesn't seem to add up here. From what's been posted, I'd expect much more RAM to be in use. > > > This implies a poor match between physical layout and access pattern. > > > > This seems to be the case. But since this is not the only query, it may > > be that other access patterns are more important to optimize for. > > > > > > > > If I understand correctly, the table should not be very fragmented given > > > that this is a reasonably freshly loaded DB? That implies that the > > > fields being looked up are not well sorted in the table compared to the > > > query pattern. > > > > > > If the entire table could fit in RAM, this would be far less of a > > > consideration. Failing that, the physical HD layout has to be improved > > > or the query pattern has to be changed to reduce seeks. > > > > > > > > > > ... > > > > >> After CLUSTER, the current data will stay clustered, but new data will > > >> not, so you have to continually CLUSTER, the same way that you might > > >> VACUUM. *However*, IIRC CLUSTER grabs an Exclusive lock, so it is as > > >> expensive as a VACUUM FULL. Be aware of this, but it might vastly > > >> improve your performance, so it would be worth it. > > > > > > > > > CLUSTER can be a very large maintenance overhead/problem if the table(s) > > > in question actually need to be "continually" re CLUSTER ed. > > > > > > If there is no better solution available, then you do what you have to, > > > but it feels like there should be a better answer here. > > > > > > Perhaps the DB schema needs examining to see if it matches up well with > > > its real usage? > > > > > > Ron Peacetree > > > > > > > I certainly agree that CLUSTER is expensive, and is an on-going > > maintenance issue. If it is the normal access pattern, though, it may be > > worth it. > >The query I've sent you is one of the most common I get just change the >name. I handle about 180K of them a day mostly between 8 and 5. The >clustering has never really been a problem. Like I said before I do it >about once a week. I handle about 3000 update an hour consisting of >about 1000-3000 statement per update. ie about 2.5 million updates per >hour. In the last few months or so I've filtered these down to about >400K update/delete/insert statements per hour. 2.5M updates per hour = ~695 updates per second. 400K per hour = ~112 updates per sec. These should be well within the capabilities of a RAID 10 subsystem based on 14 15Krpm HDs assuming a decent RAID card. What is the exact HW of the RAID subsystem involved and how is it configured? You shouldn't be having a performance problem AFAICT... > > I also wonder, though, if his table is properly normalized. > Which, as you mentioned, might lead to improved access patterns. >The system is about as normalized as I can get it. In general the >layout is the following: >courts have cases, cases have litigant_details. Actors have >identities and litigant_details. Hmmm. Can you tell us more about the actual schema, I may have an idea... > > > > John > > =:-> >-- >Speak softly and carry a +6 two-handed sword. Nah. A wand of 25th level automatic Magic Missile Fire ;-) Ron Peacetree From pgsql-performance-owner@postgresql.org Sat Aug 20 15:16:31 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id DE65852AB0 for ; Sat, 20 Aug 2005 15:16:23 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 44105-09 for ; Sat, 20 Aug 2005 18:16:17 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id 76CBC52F06 for ; Sat, 20 Aug 2005 15:16:13 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7KIGA1D017691; Sat, 20 Aug 2005 13:16:10 -0500 Subject: Re: extremly low memory usage From: Jeremiah Jahn To: John A Meinel Cc: Ron , postgres performance In-Reply-To: <43064A06.8020102@arbash-meinel.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> <43063196.9000001@arbash-meinel.com> <1124485299.27881.192.camel@bluejay.goodinassociates.com> <43064A06.8020102@arbash-meinel.com> Content-Type: text/plain Date: Sat, 20 Aug 2005 13:16:09 -0500 Message-Id: <1124561769.27881.200.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Content-Transfer-Encoding: 7bit Received-SPF: none (penguin.goodinassociates.com: 206.80.71.242 is neither permitted nor denied by domain of cs.earlham.edu) client-ip=206.80.71.242; envelope-from=jeremiah@cs.earlham.edu; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1034/Thu Aug 18 15:07:58 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.034 required=5 tests=[AWL=-0.016, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/297 X-Sequence-Number: 14044 I'm just watching gnome-system-monoitor. Which after careful consideration.....and looking at dstat means I'm on CRACK....GSM isn't showing cached memory usage....I asume that the cache memory usage is where data off of the disks would be cached...? memory output from dstat is this for a few seconds: ---procs--- ------memory-usage----- ---paging-- --disk/sda----disk/sdb- ----swap--- ----total-cpu-usage---- run blk new|_used _buff _cach _free|__in_ _out_|_read write:_read write|_used _free|usr sys idl wai hiq siq 0 0 0|1336M 10M 4603M 17M| 490B 833B|3823B 3503k:1607k 4285k| 160k 2048M| 4 1 89 7 0 0 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 0 : 0 464k| 160k 2048M| 25 0 75 0 0 0 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 0 : 0 0 | 160k 2048M| 25 0 75 0 0 0 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 48k: 0 0 | 160k 2048M| 25 0 75 0 0 0 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 0 : 0 0 | 160k 2048M| 25 0 75 0 0 0 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 132k: 0 0 | 160k 2048M| 25 0 75 0 0 0 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 36k: 0 0 | 160k 2048M| 25 0 75 0 0 0 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 0 : 0 0 | 160k 2048M| 25 0 75 0 0 0 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 12k: 0 0 | 160k 2048M| 25 0 75 0 0 0 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 0 : 0 0 | 160k 2048M| 25 0 75 0 0 0 2 0 0|1353M 10M 4585M 18M| 0 0 | 0 0 : 0 0 | 160k 2048M| 25 1 75 0 0 0 1 0 0|1321M 10M 4616M 19M| 0 0 | 0 0 : 0 0 | 160k 2048M| 18 8 74 0 0 0 1 0 0|1326M 10M 4614M 17M| 0 0 | 0 0 :4096B 0 | 160k 2048M| 16 10 74 1 0 0 1 0 0|1330M 10M 4609M 17M| 0 0 | 0 12k:4096B 0 | 160k 2048M| 17 9 74 0 0 0 0 1 0|1343M 10M 4596M 17M| 0 0 | 0 0 : 0 316M| 160k 2048M| 5 10 74 11 0 1 0 1 0|1339M 10M 4596M 21M| 0 0 | 0 0 : 0 0 | 160k 2048M| 0 0 74 25 0 1 0 2 0|1334M 10M 4596M 25M| 0 0 | 0 4096B: 0 0 | 160k 2048M| 0 0 54 44 0 1 1 0 0|1326M 10M 4596M 34M| 0 0 | 0 0 : 0 364k| 160k 2048M| 4 1 60 34 0 1 1 0 0|1290M 10M 4596M 70M| 0 0 | 0 12k: 0 0 | 160k 2048M| 24 1 75 0 0 0 1 0 0|1301M 10M 4596M 59M| 0 0 | 0 20k: 0 0 | 160k 2048M| 21 4 75 0 0 0 1 0 0|1312M 10M 4596M 48M| 0 0 | 0 0 : 0 0 | 160k 2048M| 22 4 75 0 0 0 1 0 0|1323M 10M 4596M 37M| 0 0 | 0 0 : 0 24k| 160k 2048M| 21 4 75 0 0 0 1 0 0|1334M 10M 4596M 25M| 0 0 | 0 0 : 0 56k| 160k 2048M| 21 4 75 0 0 0 On Fri, 2005-08-19 at 16:07 -0500, John A Meinel wrote: > Jeremiah Jahn wrote: > > Rebuild in progress with just ext3 on the raid array...will see if this > > helps the access times. If it doesn't I'll mess with the stripe size. I > > have REINDEXED, CLUSTERED, tablespaced and cached with 'cat table/index > > > >>/dev/null' none of this seems to have helped, or even increased my > > > > memory usage. argh! The only thing about this new system that I'm > > unfamiliar with is the array setup and LVM, which is why I think that's > > where the issue is. clustering and indexing as well as vacuum etc are > > things that I do and have been aware of for sometime. Perhaps slony is a > > factor, but I really don't see it causing problems on index read speed > > esp. when it's not running. > > > > thanx for your help, I really appreciate it. > > -jj- > > > > By the way, how are you measuring memory usage? Can you give the output > of that command, just to make sure you are reading it correctly. > > John > =:-> > -- Speak softly and carry a +6 two-handed sword. From pgsql-performance-owner@postgresql.org Sat Aug 20 15:31:31 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A324352E07 for ; Sat, 20 Aug 2005 15:31:23 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 04910-02 for ; Sat, 20 Aug 2005 18:31:17 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id 6750052D5A for ; Sat, 20 Aug 2005 15:31:15 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7KIVFhh017780; Sat, 20 Aug 2005 13:31:15 -0500 Subject: Re: extremly low memory usage From: Jeremiah Jahn To: Ron Cc: postgres performance In-Reply-To: <6.2.3.4.0.20050819233655.05c993a0@pop.earthlink.net> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> <43063196.9000001@arbash-meinel.com> <1124482288.27881.182.camel@bluejay.goodinassociates.com> <6.2.3.4.0.20050819233655.05c993a0@pop.earthlink.net> Content-Type: text/plain Date: Sat, 20 Aug 2005 13:31:15 -0500 Message-Id: <1124562675.27881.215.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Content-Transfer-Encoding: 7bit Received-SPF: none (penguin.goodinassociates.com: 206.80.71.242 is neither permitted nor denied by domain of cs.earlham.edu) client-ip=206.80.71.242; envelope-from=jeremiah@cs.earlham.edu; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1034/Thu Aug 18 15:07:58 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.034 required=5 tests=[AWL=-0.016, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/298 X-Sequence-Number: 14045 On Sat, 2005-08-20 at 11:59 -0400, Ron wrote: > At 04:11 PM 8/19/2005, Jeremiah Jahn wrote: > >On Fri, 2005-08-19 at 14:23 -0500, John A Meinel wrote: > > > Ron wrote: > > > > At 01:18 PM 8/19/2005, John A Meinel wrote: > > > > > > > >> Jeremiah Jahn wrote: > > > >> > Sorry about the formatting. > > > >> > > > > >> > On Thu, 2005-08-18 at 12:55 -0500, John Arbash Meinel wrote: > > > >> > > > > >> >>Jeremiah Jahn wrote: > > > >> >> > > > >> >> > > > >> > > > >> ... > > > >> > > > >> >>The expensive parts are the 4915 lookups into the litigant_details > > > >> (each > > > >> >>one takes approx 4ms for a total of ~20s). > > > >> >>And then you do it again on case_data (average 3ms each * 4906 loops = > > > >> >>~15s). > > > >> > > > > >> > Is there some way to avoid this? > > > >> > > > > >> > > > >> Well, in general, 3ms for a single lookup seems really long. Maybe your > > > >> index is bloated by not vacuuming often enough. Do you tend to get a lot > > > >> of updates to litigant_details? > > > > > > > > > > > > Given that the average access time for a 15Krpm HD is in the 5.5-6ms > > > > range (7.5-8ms for a 10Krpm HD), having an average of 3ms for a single > > > > lookup implies that ~1/2 (the 15Krpm case) or ~1/3 (the 10Krpm case) > > > > table accesses is requiring a seek. > > > > > >I think LVM may be a problem, since it also seems to break things up on > >the file system. My access time on the seek should be around 1/7th the > >15Krpm I believe since it's a 14 disk raid 10 array. And no other > >traffic at the moment. > > Oops. There's a misconception here. RAID arrays increase > _throughput_ AKA _bandwidth_ through parallel access to HDs. OTOH, > access time is _latency_, and that is not changed. Access time for a > RAID set is equal to that of the slowest access time, AKA highest > latency, HD in the RAID set. so I will max out at the 5.5-6ms rang for access time? > > > > Well, from what he has said, the total indexes are < 1GB and he has 6GB > > > of ram. So everything should fit. Not to mention he is only accessing > > > 5000/several million rows. > >I table spaced some of the indexes and they are around 211066880 bytes > >for the name_speed index and 149825330 for the lit_actor_speed index > >tables seem to be about a gig. > > Hmm. And you think you are only using 250MB out of your 6GB of > RAM? Something doesn't seem to add up here. From what's been > posted, I'd expect much more RAM to be in use. the cached memory usage is complete using up the rest of the memory. > > > > > > This implies a poor match between physical layout and access pattern. > > > > > > This seems to be the case. But since this is not the only query, it may > > > be that other access patterns are more important to optimize for. > > > > > > > > > > > If I understand correctly, the table should not be very fragmented given > > > > that this is a reasonably freshly loaded DB? That implies that the > > > > fields being looked up are not well sorted in the table compared to the > > > > query pattern. > > > > > > > > If the entire table could fit in RAM, this would be far less of a > > > > consideration. Failing that, the physical HD layout has to be improved > > > > or the query pattern has to be changed to reduce seeks. > > > > > > > > > > > > > > ... > > > > > > >> After CLUSTER, the current data will stay clustered, but new data will > > > >> not, so you have to continually CLUSTER, the same way that you might > > > >> VACUUM. *However*, IIRC CLUSTER grabs an Exclusive lock, so it is as > > > >> expensive as a VACUUM FULL. Be aware of this, but it might vastly > > > >> improve your performance, so it would be worth it. > > > > > > > > > > > > CLUSTER can be a very large maintenance overhead/problem if the table(s) > > > > in question actually need to be "continually" re CLUSTER ed. > > > > > > > > If there is no better solution available, then you do what you have to, > > > > but it feels like there should be a better answer here. > > > > > > > > Perhaps the DB schema needs examining to see if it matches up well with > > > > its real usage? > > > > > > > > Ron Peacetree > > > > > > > > > > I certainly agree that CLUSTER is expensive, and is an on-going > > > maintenance issue. If it is the normal access pattern, though, it may be > > > worth it. > > > >The query I've sent you is one of the most common I get just change the > >name. I handle about 180K of them a day mostly between 8 and 5. The > >clustering has never really been a problem. Like I said before I do it > >about once a week. I handle about 3000 update an hour consisting of > >about 1000-3000 statement per update. ie about 2.5 million updates per > >hour. In the last few months or so I've filtered these down to about > >400K update/delete/insert statements per hour. > > 2.5M updates per hour = ~695 updates per second. 400K per hour = > ~112 updates per sec. These should be well within the capabilities > of a RAID 10 subsystem based on 14 15Krpm HDs assuming a decent RAID > card. What is the exact HW of the RAID subsystem involved and how is > it configured? You shouldn't be having a performance problem AFAICT... dell perc4 with 14 drives and the each pair is raid 1 with spanning enabled across all of the pairs. It doesn't say raid 10...But it seem to be it. What else would you like to know? > > > > I also wonder, though, if his table is properly normalized. > > Which, as you mentioned, might lead to improved access patterns. > >The system is about as normalized as I can get it. In general the > >layout is the following: > >courts have cases, cases have litigant_details. Actors have > >identities and litigant_details. > > Hmmm. Can you tell us more about the actual schema, I may have an idea... In what format would you like it. What kind of things would you like to know..? I've probably missed a few things, but this is what running on the production box. There are no foreign keys. Cascading delete were far too slow. And having to determine the order of deletes was a pain in the but. CREATE TABLE actor ( actor_id character varying(50) NOT NULL, case_id character varying(50) DEFAULT '0'::character varying NOT NULL, court_ori character varying(18) NOT NULL, role_class_code character varying(50) NOT NULL ); CREATE TABLE identity ( identity_id character varying(50) NOT NULL, actor_id character varying(50) NOT NULL, case_id character varying(50) DEFAULT '0'::character varying NOT NULL, court_ori character varying(18) NOT NULL, identity_type character varying(10) NOT NULL, entity_type character varying(50), full_name character varying(60) NOT NULL, entity_acronym character varying(50), name_prefix character varying(50), first_name character varying(50), middle_name character varying(50), last_name character varying(50), name_suffix character varying(50), gender_code character varying(50), date_of_birth date, place_of_birth character varying(50), height character varying(50), height_unit character varying(50), weight character varying(50), weight_unit character varying(50), religion character varying(50), ethnicity character varying(50), citizenship_country character varying(50), hair_color character varying(50), eye_color character varying(50), scars_marks_tatto character varying(255), marital_status character varying(50) ); ALTER TABLE ONLY identity ALTER COLUMN full_name SET STATISTICS 1000; CREATE TABLE case_data ( case_id character varying(50) NOT NULL, court_ori character varying(18) NOT NULL, type_code character varying(50), subtype_code character varying(50), case_category character varying(50), case_title character varying(100), type_subtype_text character varying(255), case_year integer, extraction_datetime character varying(15) NOT NULL, update_date date NOT NULL, case_dom oid, data bytea ); CREATE TABLE litigant_details ( actor_id character varying(50) NOT NULL, case_id character varying(50) NOT NULL, court_ori character varying(18) NOT NULL, assigned_case_role character varying(50) NOT NULL, initial_file_date date, initial_close_date date, reopen_date date, reclose_date date, physical_file_location character varying(50), impound_litigant_data character varying(50), impound_litigant_minutes character varying(50), actor_type character varying(50) NOT NULL, conviction character varying(3) ); CREATE TABLE actor_identifier ( identity_id character varying(50) NOT NULL, actor_id character varying(50) NOT NULL, case_id character varying(50) DEFAULT '0'::character varying NOT NULL, court_ori character varying(18) NOT NULL, actor_identifier_type_code character varying(50) NOT NULL, actor_identifier_id character varying(50) NOT NULL ); CREATE TABLE actor_relationship ( litigant_actor_id character varying(50) NOT NULL, related_actor_id character varying(50) NOT NULL, case_id character varying(50) NOT NULL, court_ori character varying(18) NOT NULL, relationship_type character varying(50) NOT NULL ); CREATE INDEX lit_actor_speed ON litigant_details USING btree (actor_id); CREATE INDEX name_speed ON identity USING btree (full_name); ALTER TABLE identity CLUSTER ON name_speed; CREATE INDEX case_speed ON case_data USING btree (court_ori, case_id); ALTER TABLE case_data CLUSTER ON case_speed; ALTER TABLE ONLY actor ADD CONSTRAINT actor_pkey PRIMARY KEY (court_ori, case_id, actor_id); ALTER TABLE ONLY identity ADD CONSTRAINT identity_pkey PRIMARY KEY (court_ori, case_id, identity_id, actor_id); ALTER TABLE ONLY case_data ADD CONSTRAINT case_data_pkey PRIMARY KEY (court_ori, case_id); ALTER TABLE ONLY litigant_details ADD CONSTRAINT litigant_details_pkey PRIMARY KEY (actor_id, case_id, court_ori); > > > > > > > John > > > =:-> > >-- > >Speak softly and carry a +6 two-handed sword. > > Nah. A wand of 25th level automatic Magic Missile Fire ;-) > > Ron Peacetree > -- Speak softly and carry a +6 two-handed sword. From pgsql-performance-owner@postgresql.org Sat Aug 20 15:53:29 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 64B96528BE for ; Sat, 20 Aug 2005 15:53:27 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49344-01 for ; Sat, 20 Aug 2005 18:53:25 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id DBDFB52834 for ; Sat, 20 Aug 2005 15:53:24 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7KIrOrJ017902; Sat, 20 Aug 2005 13:53:24 -0500 Subject: Re: extremly low memory usage From: Jeremiah Jahn To: John A Meinel Cc: Jeff Trout , postgres performance In-Reply-To: <43064908.7010505@arbash-meinel.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <1124481386.27881.167.camel@bluejay.goodinassociates.com> <43064908.7010505@arbash-meinel.com> Content-Type: text/plain Date: Sat, 20 Aug 2005 13:53:23 -0500 Message-Id: <1124564003.27881.229.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Content-Transfer-Encoding: 7bit Received-SPF: none (penguin.goodinassociates.com: 206.80.71.242 is neither permitted nor denied by domain of cs.earlham.edu) client-ip=206.80.71.242; envelope-from=jeremiah@cs.earlham.edu; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1034/Thu Aug 18 15:07:58 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.035 required=5 tests=[AWL=-0.015, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/299 X-Sequence-Number: 14046 On Fri, 2005-08-19 at 16:03 -0500, John A Meinel wrote: > Jeremiah Jahn wrote: > > On Fri, 2005-08-19 at 12:18 -0500, John A Meinel wrote: > > > >>Jeremiah Jahn wrote: > >> > > > ... > > >> > >>Well, in general, 3ms for a single lookup seems really long. Maybe your > >>index is bloated by not vacuuming often enough. Do you tend to get a lot > >>of updates to litigant_details? > > > > I have vacuumed this already. I get lots of updates, but this data is > > mostly unchanging. > > > > > >>There are a couple possibilities at this point. First, you can REINDEX > >>the appropriate index, and see if that helps. However, if this is a test > >>box, it sounds like you just did a dump and reload, which wouldn't have > >>bloat in an index. > > > > > > I loaded it using slony > > I don't know that slony versus pg_dump/pg_restore really matters. The > big thing is that Updates wouldn't be trashing your index. > But if you are saying that you cluster once/wk your index can't be that > messed up anyway. (Unless CLUSTER messes up the non-clustered indexes, > but that would make cluster much less useful, so I would have guessed > this was not the case) > > > > > > >>Another possibility. Is this the column that you usually use when > >>pulling information out of litigant_details? If so, you can CLUSTER > >>litigant_details on the appropriate index. This will help things be > >>close together that should be, which decreases the index lookup costs. > > > > clustering on this right now. Most of the other things are already > > clustered. name and case_data > > Just as a reality check, they are clustered on the columns in question, > right? (I don't know if this column is a primary key or not, but any > index can be used for clustering). > > > > > > >>However, if this is not the common column, then you probably will slow > >>down whatever other accesses you may have on this table. > >> > >>After CLUSTER, the current data will stay clustered, but new data will > >>not, so you have to continually CLUSTER, the same way that you might > >>VACUUM. *However*, IIRC CLUSTER grabs an Exclusive lock, so it is as > >>expensive as a VACUUM FULL. Be aware of this, but it might vastly > >>improve your performance, so it would be worth it. > > > > I generally re-cluster once a week. > > > >>> > >>>>So there is no need for preloading your indexes on the identity table. > >>>>It is definitely not the bottleneck. > >>>> > >>>>So a few design bits, which may help your database. > >>>>Why is "actor_id" a text field instead of a number? > >>> > >>>This is simply due to the nature of the data. > >>> > >> > >>I'm just wondering if changing into a number, and using a number->name > >>lookup would be faster for you. It may not be. In general, I prefer to > >>use numbers for references. I may be over paranoid, but I know that some > >>locales are bad with string -> string comparisons. And since the data in > >>your database is stored as UNICODE, I'm not sure if it has to do any > >>translating or not. Again, something to consider, it may not make any > >>difference. > > > > I don't believe so. I initialze the DB as 'lang=C'. I used to have the > > problem where things were being inited as en_US. this would prevent any > > text based index from working. This doesn't seem to be the case here, so > > I'm not worried about it. > > > > Sorry, I think I was confusing you with someone else who posted SHOW ALL. > > > > > > >> > > ... > > > it's cached alright. I'm getting a read rate of about 150MB/sec. I would > > have thought is would be faster with my raid setup. I think I'm going to > > scrap the whole thing and get rid of LVM. I'll just do a straight ext3 > > system. Maybe that will help. Still trying to get suggestions for a > > stripe size. > > > > I don't think 150MB/s is out of the realm for a 14 drive array. > How fast is > time dd if=/dev/zero of=testfile bs=8192 count=1000000 time dd if=/dev/zero of=testfile bs=8192 count=1000000 1000000+0 records in 1000000+0 records out real 1m24.248s user 0m0.381s sys 0m33.028s > (That should create a 8GB file, which is too big to cache everything) > And then how fast is: > time dd if=testfile of=/dev/null bs=8192 count=1000000 time dd if=testfile of=/dev/null bs=8192 count=1000000 1000000+0 records in 1000000+0 records out real 0m54.139s user 0m0.326s sys 0m8.916s and on a second run: real 0m55.667s user 0m0.341s sys 0m9.013s > > That should give you a semi-decent way of measuring how fast the RAID > system is, since it should be too big to cache in ram. about 150MB/Sec. Is there no better way to make this go faster...? > > > > >>I can point you to REINDEX and CLUSTER, but if it is caching in ram, I > >>honestly can't say why the per loop would be that much slower. > >>Are both systems running the same postgres version? It sounds like it is > >>different (since you say something about switching to 8.0). > > > > These had little or no effect. > > The production machine is running 7.4 while the devel machine is running > > 8.0 > > > > Well, my concern is that maybe some portion of the 8.0 code actually > slowed things down for you. You could try reverting to 7.4 on the devel > box, though I think playing with upgrading to 8.1 might be more worthwhile. And the level of stability for 8.1? I started with 7.4 and it didn't really feel as fast as it should either. > > ... > > > > > this is a cached version. > > > > I assume that you mean this is the second run of the query. I can't > compare it too much, since this is "smith" rather than "jones". But this > one is 17s rather than the other one being 46s. > > And that includes having 8k rows instead of having 5k rows. > > Have you tried other values with disabled nested loops? Because this > query (at least in cached form) seems to be *way* faster than with > nested loops. > I know that you somehow managed to get 200s in your testing, but it > might just be that whatever needed to be loaded is now loaded, and you > would get better performance. > If this is true, it means you might need to tweak some settings, and > make sure your statistics are decent, so that postgres can actually pick > the optimal plan. > > > > >>copa=> explain analyze select full_name,identity_id,identity.case_id,court.id,date_of_birth,assigned_case_role,litigant_details.impound_litigant_data > >>copa-> from identity > >>copa-> join litigant_details on identity.actor_id = litigant_details.actor_id > >>copa-> join case_data on litigant_details.case_id = case_data.case_id and litigant_details.court_ori = case_data.court_ori > >>copa-> join court on identity.court_ori = court.id > >>copa-> where identity.court_ori = 'IL081025J' and full_name like 'SMITH%' order by full_name; > >> QUERY PLAN > >>------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- > >> Sort (cost=100502560.72..100502583.47 rows=9099 width=86) (actual time=17843.876..17849.401 rows=8094 loops=1) > >> Sort Key: identity.full_name > >> -> Merge Join (cost=100311378.72..100501962.40 rows=9099 width=86) (actual time=15195.816..17817.847 rows=8094 loops=1) > >> Merge Cond: ((("outer".court_ori)::text = "inner"."?column10?") AND (("outer".case_id)::text = "inner"."?column11?")) > >> -> Index Scan using case_speed on case_data (cost=0.00..170424.73 rows=3999943 width=26) (actual time=0.015..4540.525 rows=3018284 loops=1) > >> -> Sort (cost=100311378.72..100311400.82 rows=8839 width=112) (actual time=9594.985..9601.174 rows=8094 loops=1) > >> Sort Key: (litigant_details.court_ori)::text, (litigant_details.case_id)::text > >> -> Nested Loop (cost=100002491.43..100310799.34 rows=8839 width=112) (actual time=6892.755..9555.828 rows=8094 loops=1) > >> -> Seq Scan on court (cost=0.00..3.29 rows=1 width=12) (actual time=0.085..0.096 rows=1 loops=1) > >> Filter: ('IL081025J'::text = (id)::text) > > What I don't really understand is the next part. It seems to be doing an > index scan on 3.7M rows, and getting very decent performance (5s), and > then merging against a table which returns only 8k rows. > Why is it having to look through all of those rows? > I may be missing something, but this says it is able to do 600 index > lookups / millisecond. Which seems superfast. (Compared to your earlier > 4ms / lookup) > Makes me a little confused myself... > Something fishy is going on here. > > > >> -> Merge Join (cost=2491.43..310707.66 rows=8839 width=113) (actual time=6892.656..9519.680 rows=8094 loops=1) > >> Merge Cond: (("outer".actor_id)::text = "inner"."?column7?") > >> -> Index Scan using lit_actor_speed on litigant_details (cost=0.00..295722.00 rows=4956820 width=81) (actual time=0.027..5613.814 rows=3736703 loops=1) > >> -> Sort (cost=2491.43..2513.71 rows=8913 width=82) (actual time=116.071..122.272 rows=8100 loops=1) > >> Sort Key: (identity.actor_id)::text > >> -> Index Scan using name_speed on identity (cost=0.00..1906.66 rows=8913 width=82) (actual time=0.133..81.104 rows=8100 loops=1) > >> Index Cond: (((full_name)::text >= 'SMITH'::character varying) AND ((full_name)::text < 'SMITI'::character varying)) > >> Filter: (((court_ori)::text = 'IL081025J'::text) AND ((full_name)::text ~~ 'SMITH%'::text)) > >> Total runtime: 17859.917 ms > > > > > >>But really, you have worse index speed, and that needs to be figured out. > >> > >>John > >>=:-> > > I'm assuming your data is private (since it looks like legal stuff). > Unless maybe that makes it part of the public record. > Anyway, I'm not able to, but sometimes someone like Tom can profile > stuff to see what is going on. I've had tom on here before..:) not my devel box, but my production box a couple of years ago. > > I might just be messing up my ability to read the explain output. But > somehow things don't seem to be lining up with the cost of a single > index lookup. > On my crappy Celeron 450 box, an index lookup is 0.06ms once things are > cached in ram. > > John > =:-> > > -- Speak softly and carry a +6 two-handed sword. From pgsql-performance-owner@postgresql.org Sat Aug 20 18:02:55 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 573BE52CFE for ; Sat, 20 Aug 2005 18:02:02 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 00233-09 for ; Sat, 20 Aug 2005 21:01:59 +0000 (GMT) Received: from smtpauth04.mail.atl.earthlink.net (smtpauth04.mail.atl.earthlink.net [209.86.89.64]) by svr1.postgresql.org (Postfix) with ESMTP id 2AE6752C3F for ; Sat, 20 Aug 2005 18:01:58 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=bSSjAVD1BTl3ktwpPexOryQEADQutNOpxYBKGYwWyKtKneNsiAKqVr/1bgnMePXd; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:Cc:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth04.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E6aTV-0006Dy-TP; Sat, 20 Aug 2005 17:01:58 -0400 Message-Id: <6.2.3.4.0.20050820154422.06021150@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Sat, 20 Aug 2005 17:01:53 -0400 To: Jeremiah Jahn From: Ron Subject: Re: extremly low memory usage Cc: postgres performance In-Reply-To: <1124564003.27881.229.camel@bluejay.goodinassociates.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <1124481386.27881.167.camel@bluejay.goodinassociates.com> <43064908.7010505@arbash-meinel.com> <1124564003.27881.229.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bcba2b3f2007ee6fd7b8a17eb3b763a96c350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.499 required=5 tests=[AWL=0.125, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/300 X-Sequence-Number: 14047 At 02:53 PM 8/20/2005, Jeremiah Jahn wrote: >On Fri, 2005-08-19 at 16:03 -0500, John A Meinel wrote: > > Jeremiah Jahn wrote: > > > On Fri, 2005-08-19 at 12:18 -0500, John A Meinel wrote: > > > > > > > > > it's cached alright. I'm getting a read rate of about 150MB/sec. I would > > > have thought is would be faster with my raid setup. I think I'm going to > > > scrap the whole thing and get rid of LVM. I'll just do a straight ext3 > > > system. Maybe that will help. Still trying to get suggestions for a > > > stripe size. > > > > > > > I don't think 150MB/s is out of the realm for a 14 drive array. > > How fast is time dd if=/dev/zero of=testfile bs=8192 count=1000000 > > >time dd if=/dev/zero of=testfile bs=8192 count=1000000 >1000000+0 records in >1000000+0 records out > >real 1m24.248s >user 0m0.381s >sys 0m33.028s > > > > (That should create a 8GB file, which is too big to cache everything) > > And then how fast is: > > time dd if=testfile of=/dev/null bs=8192 count=1000000 > >time dd if=testfile of=/dev/null bs=8192 count=1000000 >1000000+0 records in >1000000+0 records out > >real 0m54.139s >user 0m0.326s >sys 0m8.916s > > >and on a second run: > >real 0m55.667s >user 0m0.341s >sys 0m9.013s > > > > > > That should give you a semi-decent way of measuring how fast the RAID > > system is, since it should be too big to cache in ram. > >about 150MB/Sec. Is there no better way to make this go faster...? Assuming these are U320 15Krpm 147GB HDs, a RAID 10 array of 14 of them doing raw sequential IO like this should be capable of at ~7*75MB/s= 525MB/s using Seagate Cheetah 15K.4's, ~7*79MB/s= 553MB/s if using Fujitsu MAU's, and ~7*86MB/s= 602MB/s if using Maxtor Atlas 15K II's to devices external to the RAID array. _IF_ the controller setup is high powered enough to keep that kind of IO rate up. This will require a controller or controllers providing dual channel U320 bandwidth externally and quad channel U320 bandwidth internally. IOW, it needs a controller or controllers talking 64b 133MHz PCI-X, reasonably fast DSP/CPU units, and probably a decent sized IO buffer as well. AFAICT, the Dell PERC4 controllers use various flavors of the LSI Logic MegaRAID controllers. What I don't know is which exact one yours is, nor do I know if it (or any of the MegaRAID controllers) are high powered enough. Talk to your HW supplier to make sure you have controllers adequate to your HD's. ...and yes, your average access time will be in the 5.5ms - 6ms range when doing a physical seek. Even with RAID, you want to minimize seeks and maximize sequential IO when accessing them. Best to not go to HD at all ;-) Hope this helps, Ron Peacetree From pgsql-performance-owner@postgresql.org Sat Aug 20 18:13:40 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2A95E52BC3 for ; Sat, 20 Aug 2005 18:12:15 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 19145-07 for ; Sat, 20 Aug 2005 21:12:12 +0000 (GMT) Received: from smtpauth01.mail.atl.earthlink.net (smtpauth01.mail.atl.earthlink.net [209.86.89.61]) by svr1.postgresql.org (Postfix) with ESMTP id 3AB4152996 for ; Sat, 20 Aug 2005 18:12:11 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=M9Lfp7ba3cPwkPIpq5bnsQXW7H2e1utk9NpNf3ftslAGrtOxHPcT4j80Oc1iUPff; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:Cc:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth01.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E6adP-0002X8-JQ; Sat, 20 Aug 2005 17:12:11 -0400 Message-Id: <6.2.3.4.0.20050820170952.05bcedb8@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Sat, 20 Aug 2005 17:12:07 -0400 To: Jeremiah Jahn From: Ron Subject: Re: extremly low memory usage Cc: postgres performance Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bcb32ade9d7d69f805dbe7f19d5ba55cbd350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.481 required=5 tests=[AWL=0.107, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/301 X-Sequence-Number: 14048 I'm reposting this because my mailer hiccuped when I sent it the first time. If this results in a double post, I apologize. At 02:53 PM 8/20/2005, Jeremiah Jahn wrote: >On Fri, 2005-08-19 at 16:03 -0500, John A Meinel wrote: > > Jeremiah Jahn wrote: > > > On Fri, 2005-08-19 at 12:18 -0500, John A Meinel wrote: > > > > > > > > > it's cached alright. I'm getting a read rate of about 150MB/sec. I would > > > have thought is would be faster with my raid setup. I think I'm going to > > > scrap the whole thing and get rid of LVM. I'll just do a straight ext3 > > > system. Maybe that will help. Still trying to get suggestions for a > > > stripe size. > > > > > > > I don't think 150MB/s is out of the realm for a 14 drive array. > > How fast is time dd if=/dev/zero of=testfile bs=8192 count=1000000 > > >time dd if=/dev/zero of=testfile bs=8192 count=1000000 >1000000+0 records in >1000000+0 records out > >real 1m24.248s >user 0m0.381s >sys 0m33.028s > > > > (That should create a 8GB file, which is too big to cache everything) > > And then how fast is: > > time dd if=testfile of=/dev/null bs=8192 count=1000000 > >time dd if=testfile of=/dev/null bs=8192 count=1000000 >1000000+0 records in >1000000+0 records out > >real 0m54.139s >user 0m0.326s >sys 0m8.916s > > >and on a second run: > >real 0m55.667s >user 0m0.341s >sys 0m9.013s > > > > > > That should give you a semi-decent way of measuring how fast the RAID > > system is, since it should be too big to cache in ram. > >about 150MB/Sec. Is there no better way to make this go faster...? Assuming these are U320 15Krpm 147GB HDs, a RAID 10 array of 14 of them doing raw sequential IO like this should be capable of at ~7*75MB/s= 525MB/s using Seagate Cheetah 15K.4's, ~7*79MB/s= 553MB/s if using Fujitsu MAU's, and ~7*86MB/s= 602MB/s if using Maxtor Atlas 15K II's to devices external to the RAID array. _IF_ the controller setup is high powered enough to keep that kind of IO rate up. This will require a controller or controllers providing dual channel U320 bandwidth externally and quad channel U320 bandwidth internally. IOW, it needs a controller or controllers talking 64b 133MHz PCI-X, reasonably fast DSP/CPU units, and probably a decent sized IO buffer as well. AFAICT, the Dell PERC4 controllers use various flavors of the LSI Logic MegaRAID controllers. What I don't know is which exact one yours is, nor do I know if it (or any of the MegaRAID controllers) are high powered enough. Talk to your HW supplier to make sure you have controllers adequate to your HD's. ...and yes, your average access time will be in the 5.5ms - 6ms range when doing a physical seek. Even with RAID, you want to minimize seeks and maximize sequential IO when accessing them. Best to not go to HD at all ;-) Hope this helps, Ron Peacetree From pgsql-performance-owner@postgresql.org Sat Aug 20 18:25:21 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0B97252B3E for ; Sat, 20 Aug 2005 18:23:31 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 61258-07 for ; Sat, 20 Aug 2005 21:23:29 +0000 (GMT) Received: from smtpauth01.mail.atl.earthlink.net (smtpauth01.mail.atl.earthlink.net [209.86.89.61]) by svr1.postgresql.org (Postfix) with ESMTP id 976FB52A88 for ; Sat, 20 Aug 2005 18:23:28 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=YuSmJZJhL5Sw4Ak1nBNwsCkWWigVf80zB8Vxx2k/7SeQvEhCUvfSO2tYc3naK9g2; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:Cc:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth01.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E6aoM-000847-P8; Sat, 20 Aug 2005 17:23:30 -0400 Message-Id: <6.2.3.4.0.20050820171521.05bd0a38@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Sat, 20 Aug 2005 17:23:25 -0400 To: Jeremiah Jahn From: Ron Subject: Re: extremly low memory usage Cc: postgres performance In-Reply-To: <1124561769.27881.200.camel@bluejay.goodinassociates.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> <43063196.9000001@arbash-meinel.com> <1124485299.27881.192.camel@bluejay.goodinassociates.com> <43064A06.8020102@arbash-meinel.com> <1124561769.27881.200.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bcd8796ffd4a735881ffc8faabeffd2923350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.466 required=5 tests=[AWL=0.092, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/302 X-Sequence-Number: 14049 At 02:16 PM 8/20/2005, Jeremiah Jahn wrote: >I'm just watching gnome-system-monoitor. Which after careful >consideration.....and looking at dstat means I'm on CRACK....GSM isn't >showing cached memory usage....I asume that the cache memory usage >is where data off of the disks would be cached...? > >memory output from dstat is this for a few seconds: > >---procs--- ------memory-usage----- ---paging-- >--disk/sda----disk/sdb- ----swap--- ----total-cpu-usage---- >run blk new|_used _buff _cach _free|__in_ _out_|_read write:_read >write|_used _free|usr sys idl wai hiq siq > 0 0 0|1336M 10M 4603M 17M| 490B 833B|3823B 3503k:1607k > 4285k| 160k 2048M| 4 1 89 7 0 0 > 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 0 > : 0 464k| 160k 2048M| 25 0 75 0 0 0 > > 1 0 0|1334M 10M 4596M 25M| 0 0 | 0 0 > : 0 56k| 160k 2048M| 21 4 75 0 0 0 Then the "low memory usage" was a chimera. Excellent! Given the evidence in this thread, IMO you should upgrade your box to 16GB of RAM ASAP. That should be enough to cache most, if not all, of the 10GB of the "hot" part of your DB; thereby dedicating your HD subsystem as much as possible to writes (which is unavoidable HD IO). As I've posted before, at $75-$150/GB, it's well worth the investment whenever you can prove it will help as we have here. Hope this helps, Ron Peacetree From pgsql-performance-owner@postgresql.org Sat Aug 20 23:29:24 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 92A1B52D9A for ; Sat, 20 Aug 2005 23:28:32 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 42777-05 for ; Sun, 21 Aug 2005 02:28:31 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id 9420652D96 for ; Sat, 20 Aug 2005 23:28:29 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050821022614m92009j2ufe>; Sun, 21 Aug 2005 02:28:33 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 59B9955FE1; Sat, 20 Aug 2005 21:26:13 -0500 (CDT) Received: from [192.168.1.11] (liliana.arbash-meinel.com [192.168.1.11]) by juju.arbash-meinel.com (Postfix) with ESMTP id 37E3455FAD; Sat, 20 Aug 2005 21:26:07 -0500 (CDT) Message-ID: <4307E63C.7050408@arbash-meinel.com> Date: Sat, 20 Aug 2005 21:26:04 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.2 (Windows/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Jeremiah Jahn Cc: Ron , postgres performance Subject: Re: extremly low memory usage References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> <43063196.9000001@arbash-meinel.com> <1124485299.27881.192.camel@bluejay.goodinassociates.com> <43064A06.8020102@arbash-meinel.com> <1124561769.27881.200.camel@bluejay.goodinassociates.com> In-Reply-To: <1124561769.27881.200.camel@bluejay.goodinassociates.com> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigB4CB00E9C4E22B57E9612252" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.042 required=5 tests=[AWL=-0.008, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/303 X-Sequence-Number: 14050 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigB4CB00E9C4E22B57E9612252 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Jeremiah Jahn wrote: > I'm just watching gnome-system-monoitor. Which after careful > consideration.....and looking at dstat means I'm on CRACK....GSM isn't > showing cached memory usage....I asume that the cache memory usage is > where data off of the disks would be cached...? > Well a simple "free" also tells you how much has been cached. I believe by reading the _cach line, it looks like you have 4.6G cached. So you are indeed using memory. I'm still concerned why it seems to be taking 3-4ms per index lookup, when things should already be cached in RAM. Now, I may be wrong about whether the indexes are cached, but I sure would expect them to be. What is the time for a cached query on your system (with normal nested loops)? (give the EXPLAIN ANALYZE for the *second* run, or maybe the fourth). I'm glad that we aren't seeing something weird with your kernel, at least. John =:-> > > > memory output from dstat is this for a few seconds: > > ---procs--- ------memory-usage----- ---paging-- --disk/sda----disk/sdb- ----swap--- ----total-cpu-usage---- > run blk new|_used _buff _cach _free|__in_ _out_|_read write:_read write|_used _free|usr sys idl wai hiq siq > 0 0 0|1336M 10M 4603M 17M| 490B 833B|3823B 3503k:1607k 4285k| 160k 2048M| 4 1 89 7 0 0 > 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 0 : 0 464k| 160k 2048M| 25 0 75 0 0 0 > 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 0 : 0 0 | 160k 2048M| 25 0 75 0 0 0 > 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 48k: 0 0 | 160k 2048M| 25 0 75 0 0 0 > 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 0 : 0 0 | 160k 2048M| 25 0 75 0 0 0 > 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 132k: 0 0 | 160k 2048M| 25 0 75 0 0 0 > 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 36k: 0 0 | 160k 2048M| 25 0 75 0 0 0 > 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 0 : 0 0 | 160k 2048M| 25 0 75 0 0 0 > 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 12k: 0 0 | 160k 2048M| 25 0 75 0 0 0 > 1 0 0|1337M 10M 4600M 18M| 0 0 | 0 0 : 0 0 | 160k 2048M| 25 0 75 0 0 0 > 2 0 0|1353M 10M 4585M 18M| 0 0 | 0 0 : 0 0 | 160k 2048M| 25 1 75 0 0 0 > 1 0 0|1321M 10M 4616M 19M| 0 0 | 0 0 : 0 0 | 160k 2048M| 18 8 74 0 0 0 > 1 0 0|1326M 10M 4614M 17M| 0 0 | 0 0 :4096B 0 | 160k 2048M| 16 10 74 1 0 0 > 1 0 0|1330M 10M 4609M 17M| 0 0 | 0 12k:4096B 0 | 160k 2048M| 17 9 74 0 0 0 > 0 1 0|1343M 10M 4596M 17M| 0 0 | 0 0 : 0 316M| 160k 2048M| 5 10 74 11 0 1 > 0 1 0|1339M 10M 4596M 21M| 0 0 | 0 0 : 0 0 | 160k 2048M| 0 0 74 25 0 1 > 0 2 0|1334M 10M 4596M 25M| 0 0 | 0 4096B: 0 0 | 160k 2048M| 0 0 54 44 0 1 > 1 0 0|1326M 10M 4596M 34M| 0 0 | 0 0 : 0 364k| 160k 2048M| 4 1 60 34 0 1 > 1 0 0|1290M 10M 4596M 70M| 0 0 | 0 12k: 0 0 | 160k 2048M| 24 1 75 0 0 0 > 1 0 0|1301M 10M 4596M 59M| 0 0 | 0 20k: 0 0 | 160k 2048M| 21 4 75 0 0 0 > 1 0 0|1312M 10M 4596M 48M| 0 0 | 0 0 : 0 0 | 160k 2048M| 22 4 75 0 0 0 > 1 0 0|1323M 10M 4596M 37M| 0 0 | 0 0 : 0 24k| 160k 2048M| 21 4 75 0 0 0 > 1 0 0|1334M 10M 4596M 25M| 0 0 | 0 0 : 0 56k| 160k 2048M| 21 4 75 0 0 0 > > > > On Fri, 2005-08-19 at 16:07 -0500, John A Meinel wrote: > >>Jeremiah Jahn wrote: >> >>>Rebuild in progress with just ext3 on the raid array...will see if this >>>helps the access times. If it doesn't I'll mess with the stripe size. I >>>have REINDEXED, CLUSTERED, tablespaced and cached with 'cat table/index >>> >>> >>>>/dev/null' none of this seems to have helped, or even increased my >>> >>>memory usage. argh! The only thing about this new system that I'm >>>unfamiliar with is the array setup and LVM, which is why I think that's >>>where the issue is. clustering and indexing as well as vacuum etc are >>>things that I do and have been aware of for sometime. Perhaps slony is a >>>factor, but I really don't see it causing problems on index read speed >>>esp. when it's not running. >>> >>>thanx for your help, I really appreciate it. >>>-jj- >>> >> >>By the way, how are you measuring memory usage? Can you give the output >>of that command, just to make sure you are reading it correctly. >> >>John >>=:-> >> --------------enigB4CB00E9C4E22B57E9612252 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDB+Y+JdeBCYSNAAMRAigBAKC1f3abT+2b+rCkhszkHOPwnY7cRQCfTg/K HNROGvXqnQyBg/Scof5BCao= =vL6A -----END PGP SIGNATURE----- --------------enigB4CB00E9C4E22B57E9612252-- From pgsql-performance-owner@postgresql.org Sat Aug 20 23:36:16 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 37F6352DEA for ; Sat, 20 Aug 2005 23:34:26 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 55941-03 for ; Sun, 21 Aug 2005 02:34:23 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id E2EAD52DC1 for ; Sat, 20 Aug 2005 23:34:21 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050821023210m9100g5klfe>; Sun, 21 Aug 2005 02:34:26 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 437A855FE1; Sat, 20 Aug 2005 21:32:10 -0500 (CDT) Received: from [192.168.1.11] (liliana.arbash-meinel.com [192.168.1.11]) by juju.arbash-meinel.com (Postfix) with ESMTP id 744C855FAD; Sat, 20 Aug 2005 21:32:04 -0500 (CDT) Message-ID: <4307E7A4.7090300@arbash-meinel.com> Date: Sat, 20 Aug 2005 21:32:04 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.2 (Windows/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Ron Cc: Jeremiah Jahn , postgres performance Subject: Re: extremly low memory usage References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <1124481386.27881.167.camel@bluejay.goodinassociates.com> <43064908.7010505@arbash-meinel.com> <1124564003.27881.229.camel@bluejay.goodinassociates.com> <6.2.3.4.0.20050820154422.06021150@pop.earthlink.net> In-Reply-To: <6.2.3.4.0.20050820154422.06021150@pop.earthlink.net> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig6CF0AF8208E4B7FC20509C12" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.042 required=5 tests=[AWL=-0.008, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/304 X-Sequence-Number: 14051 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig6CF0AF8208E4B7FC20509C12 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Ron wrote: > At 02:53 PM 8/20/2005, Jeremiah Jahn wrote: > >> On Fri, 2005-08-19 at 16:03 -0500, John A Meinel wrote: >> > Jeremiah Jahn wrote: >> > > On Fri, 2005-08-19 at 12:18 -0500, John A Meinel wrote: >> > > >> >> > >> > > it's cached alright. I'm getting a read rate of about 150MB/sec. I >> would >> > > have thought is would be faster with my raid setup. I think I'm >> going to >> > > scrap the whole thing and get rid of LVM. I'll just do a straight >> ext3 >> > > system. Maybe that will help. Still trying to get suggestions for a >> > > stripe size. >> > > Well, since you can get a read of the RAID at 150MB/s, that means that it is actual I/O speed. It may not be cached in RAM. Perhaps you could try the same test, only using say 1G, which should be cached. >> > >> > I don't think 150MB/s is out of the realm for a 14 drive array. >> > How fast is time dd if=/dev/zero of=testfile bs=8192 count=1000000 >> > >> time dd if=/dev/zero of=testfile bs=8192 count=1000000 >> 1000000+0 records in >> 1000000+0 records out >> >> real 1m24.248s >> user 0m0.381s >> sys 0m33.028s >> >> >> > (That should create a 8GB file, which is too big to cache everything) >> > And then how fast is: >> > time dd if=testfile of=/dev/null bs=8192 count=1000000 >> >> time dd if=testfile of=/dev/null bs=8192 count=1000000 >> 1000000+0 records in >> 1000000+0 records out >> >> real 0m54.139s >> user 0m0.326s >> sys 0m8.916s >> >> >> and on a second run: >> >> real 0m55.667s >> user 0m0.341s >> sys 0m9.013s >> >> >> > >> > That should give you a semi-decent way of measuring how fast the RAID >> > system is, since it should be too big to cache in ram. >> >> about 150MB/Sec. Is there no better way to make this go faster...? I'm actually curious about PCI bus saturation at this point. Old 32-bit 33MHz pci could only push 1Gbit = 100MB/s. Now, I'm guessing that this is a higher performance system. But I'm really surprised that your write speed is that close to your read speed. (100MB/s write, 150MB/s read). > > Assuming these are U320 15Krpm 147GB HDs, a RAID 10 array of 14 of them > doing raw sequential IO like this should be capable of at > ~7*75MB/s= 525MB/s using Seagate Cheetah 15K.4's, ~7*79MB/s= 553MB/s > if using Fujitsu MAU's, and ~7*86MB/s= 602MB/s if using Maxtor Atlas 15K > II's to devices external to the RAID array. I know I thought these were SATA drives, over 2 controllers. I could be completely wrong, though. > > _IF_ the controller setup is high powered enough to keep that kind of IO > rate up. This will require a controller or controllers providing dual > channel U320 bandwidth externally and quad channel U320 bandwidth > internally. IOW, it needs a controller or controllers talking 64b > 133MHz PCI-X, reasonably fast DSP/CPU units, and probably a decent sized > IO buffer as well. > > AFAICT, the Dell PERC4 controllers use various flavors of the LSI Logic > MegaRAID controllers. What I don't know is which exact one yours is, > nor do I know if it (or any of the MegaRAID controllers) are high > powered enough. > > Talk to your HW supplier to make sure you have controllers adequate to > your HD's. > > ...and yes, your average access time will be in the 5.5ms - 6ms range > when doing a physical seek. > Even with RAID, you want to minimize seeks and maximize sequential IO > when accessing them. > Best to not go to HD at all ;-) Well, certainly, if you can get more into RAM, you're always better off. For writing, a battery-backed write cache, and for reading lots of system RAM. > > Hope this helps, > Ron Peacetree > John =:-> --------------enig6CF0AF8208E4B7FC20509C12 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDB+ekJdeBCYSNAAMRAlbBAKCPJZ4Ox/4u9nFiD5u0fZT9zpTDpwCferGG CmQ+DrIWO1Rw86QSK9rrRfw= =PICd -----END PGP SIGNATURE----- --------------enig6CF0AF8208E4B7FC20509C12-- From pgsql-performance-owner@postgresql.org Sun Aug 21 00:49:29 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AA01852EF7 for ; Sun, 21 Aug 2005 00:45:34 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 30938-08 for ; Sun, 21 Aug 2005 03:45:30 +0000 (GMT) Received: from calvin.surfutopia.net (calvin.surfutopia.net [67.120.245.34]) by svr1.postgresql.org (Postfix) with ESMTP id 7BE4952DC4 for ; Sun, 21 Aug 2005 00:45:28 -0300 (ADT) Received: by calvin.surfutopia.net (Postfix, from userid 1000) id 32191F22A; Sat, 20 Aug 2005 20:48:41 -0700 (PDT) Date: Sat, 20 Aug 2005 20:48:41 -0700 From: John Mendenhall To: pgsql-performance list Subject: complex query performance assistance request Message-ID: <20050821034841.GA28968@calvin.surfutopia.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/305 X-Sequence-Number: 14052 I need to improve the performance for the following query. Soon after I reboot my server, the following query takes 20 seconds the first time I run it. When I run it after that, it takes approximately 2 seconds. I understand the caching taking place (at the os or db level, it doesn't matter here). Here are the results of the explain analyze run: ----- LOG: duration: 6259.632 ms statement: explain analyze SELECT c.id AS contact_id, sr.id AS sales_rep_id, LTRIM(RTRIM(sr.firstname || ' ' || sr.lastname)) AS sales_rep_name, p.id AS partner_id, p.company AS partner_company, coalesce(LTRIM(RTRIM(c.company)), LTRIM(RTRIM(c.firstname || ' ' || c.lastname))) AS contact_company, LTRIM(RTRIM(c.city || ' ' || c.state || ' ' || c.postalcode || ' ' || c.country)) AS contact_location, c.phone AS contact_phone, c.email AS contact_email, co.name AS contact_country, TO_CHAR(c.request_status_last_modified, 'mm/dd/yy hh12:mi pm') AS request_status_last_modified, TO_CHAR(c.request_status_last_modified, 'yyyymmddhh24miss') AS rqst_stat_last_mdfd_sortable, c.token_id, c.master_key_token AS token FROM sales_reps sr JOIN partners p ON (sr.id = p.sales_rep_id) JOIN contacts c ON (p.id = c.partner_id) JOIN countries co ON (LOWER(c.country) = LOWER(co.code)) JOIN partner_classification pc ON (p.classification_id = pc.id AND pc.classification != 'Sales Rep') WHERE c.lead_deleted IS NULL AND EXISTS ( SELECT lr.id FROM lead_requests lr, lead_request_status lrs WHERE c.id = lr.contact_id AND lr.status_id = lrs.id AND lrs.is_closed = 0 ) ORDER BY contact_company, contact_id; QUERY PLAN -------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------- Sort (cost=39093.16..39102.80 rows=3856 width=238) (actual time=6220.481..6221.188 rows=1071 loops=1) Sort Key: COALESCE(ltrim(rtrim((c.company)::text)), ltrim(rtrim((((c.firstname)::text || ' '::text) || (c.lastname)::text)))), c.id -> Merge Join (cost=38580.89..38863.48 rows=3856 width=238) (actual time=6015.751..6184.199 rows=1071 loops=1) Merge Cond: ("outer"."?column3?" = "inner"."?column19?") -> Sort (cost=14.00..14.61 rows=242 width=19) (actual time=9.250..9.500 rows=240 loops=1) Sort Key: lower((co.code)::text) -> Seq Scan on countries co (cost=0.00..4.42 rows=242 width=19) (actual time=0.132..4.498 rows=242 loops=1) -> Sort (cost=38566.89..38574.86 rows=3186 width=225) (actual time=6005.644..6006.954 rows=1071 loops=1) Sort Key: lower((c.country)::text) -> Merge Join (cost=75.65..38381.50 rows=3186 width=225) (actual time=58.086..5979.287 rows=1071 loops=1) Merge Cond: ("outer".partner_id = "inner".id) -> Index Scan using contacts_partner_id_idx on contacts c (cost=0.00..160907.39 rows=20106 width=171) (actual time=2.569..5816.985 rows=1547 loops=1) Filter: ((lead_deleted IS NULL) AND (subplan)) SubPlan -> Nested Loop (cost=1.16..6.56 rows=2 width=10) (actual time=0.119..0.119 rows=0 loops=40261) Join Filter: ("outer".status_id = "inner".id) -> Index Scan using lead_requests_contact_id_idx on lead_requests lr (cost=0.00..4.86 rows=3 width=20) (actual time=0.079..0.083 rows=0 loops=40261) Index Cond: ($0 = contact_id) -> Materialize (cost=1.16..1.24 rows=8 width=10) (actual time=0.002..0.011 rows=6 loops=12592) -> Seq Scan on lead_request_status lrs (cost=0.00..1.16 rows=8 width=10) (actual time=0.083..0.270 rows=7 loops=1) Filter: (is_closed = 0::numeric) -> Sort (cost=75.65..76.37 rows=290 width=64) (actual time=55.073..56.990 rows=1334 loops=1) Sort Key: p.id -> Merge Join (cost=59.24..63.79 rows=290 width=64) (actual time=31.720..41.096 rows=395 loops=1) Merge Cond: ("outer".id = "inner".sales_rep_id) -> Sort (cost=2.42..2.52 rows=39 width=31) (actual time=1.565..1.616 rows=39 loops=1) Sort Key: sr.id -> Seq Scan on sales_reps sr (cost=0.00..1.39 rows=39 width=31) (actual time=0.043..0.581 rows=39 loops=1) -> Sort (cost=56.82..57.55 rows=290 width=43) (actual time=29.921..30.310 rows=395 loops=1) Sort Key: p.sales_rep_id -> Nested Loop (cost=24.35..44.96 rows=290 width=43) (actual time=0.169..22.566 rows=395 loops=1) Join Filter: ("inner".classification_id = "outer".id) -> Seq Scan on partner_classification pc (cost=0.00..1.04 rows=2 width=10) (actual time=0.059..0.102 rows=2 loops=1) Filter: ((classification)::text <> 'Sales Rep'::text) -> Materialize (cost=24.35..28.70 rows=435 width=53) (actual time=0.023..5.880 rows=435 loops=2) -> Seq Scan on partners p (cost=0.00..24.35 rows=435 width=53) (actual time=0.034..8.937 rows=435 loops=1) Total runtime: 6225.791 ms (37 rows) ----- My first question is, what is the Materialize query plan element? It happens twice, and usually when I see it, my query is slow. My second and more important question is, does anyone have any ideas or suggestions as to how I can increase the speed for this query? Things I have already done are, modify the joins and conditions so it starts with smaller tables, thus the join set is smaller, modify the configuration of the server to ensure index scans are used as they should be, ran vacuumdb and analyze on the database. Thank you very much in advance for any pointers for additional places I can look. Thanks. JohnM -- John Mendenhall john@surfutopia.net surf utopia internet services From pgsql-performance-owner@postgresql.org Sun Aug 21 06:52:29 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 067211FEF168 for ; Sun, 21 Aug 2005 06:52:13 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 51611-01 for ; Sun, 21 Aug 2005 09:52:10 +0000 (GMT) Received: from fep01-app.kolumbus.fi (fep01-0.kolumbus.fi [193.229.0.41]) by svr1.postgresql.org (Postfix) with ESMTP id 302DF1FEF167 for ; Sun, 21 Aug 2005 06:52:09 -0300 (ADT) Received: from [192.168.1.101] (really [80.186.78.229]) by fep01-app.kolumbus.fi with ESMTP id <20050821095208.SQKB23558.fep01-app.kolumbus.fi@[192.168.1.101]>; Sun, 21 Aug 2005 12:52:08 +0300 Message-ID: <43084EC4.4070906@kolumbus.fi> Date: Sun, 21 Aug 2005 12:52:04 +0300 From: Marko Ristola User-Agent: Debian Thunderbird 1.0.2 (X11/20050331) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: Michael Stone , pgsql-performance@postgresql.org, jeremiah@cs.earlham.edu Subject: Re: extremly low memory usage References: <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <6.2.3.4.0.20050819140714.05c181e8@pop.earthlink.net> <43063196.9000001@arbash-meinel.com> <1124485299.27881.192.camel@bluejay.goodinassociates.com> <43071162.6050405@kolumbus.fi> <20050820122305.GN19080@mathom.us> <23044.1124548841@sss.pgh.pa.us> In-Reply-To: <23044.1124548841@sss.pgh.pa.us> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.124 required=5 tests=[AWL=0.074, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/306 X-Sequence-Number: 14053 I'm Sorry, that I wrote that the option would risk the LOG persistency with PostgreSQL. I should have asked instead, that how you have taken this into account. Tom Lane's email below convinces me, that you have taken the metadata only journalling into account and still fulfill the persistency of committed transactions. This means, that Ext3 with data=writeback is safe with PostgreSQL even with a hardware reset button. Metadata only journalling is faster, when it can be used. I didn't know, that any database can keep the database guarantees with the metadata only journalling option. I looked at your problem. One of the problems is that you need to keep the certain data cached in memory all the time. That could be solved by doing SELECT COUNT(*) from to_be_cached; as a cron job. It loads the whole table into the Linux Kernel memory cache. Marko Ristola Tom Lane wrote: >Right. I think the optimal setting for a Postgres data directory is >journaled metadata, non-journaled file content. Postgres can take care >of the data integrity for itself, but it does assume that the filesystem >stays structurally sane (eg, data blocks don't get reassigned to the >wrong file), so you need a filesystem guarantee about the metadata. > >WAL files are handled in a much more conservative way (created, filled >with zeroes, and fsync'd before we ever put any valuable data in 'em). >If you have WAL on its own drive then I think Mike's recommendation of >no filesystem journalling at all for that drive is probably OK. Or >you can do same as above (journal metadata only) if you want a little >extra protection. > >And of course all this reasoning depends on the assumption that the >drive tells the truth about write-completion. If the drive does write >caching it had better be able to complete all its accepted writes before >dying in a power failure. (Hence, battery-backed write cache is OK, any >other kind is evil.) > > regards, tom lane > >---------------------------(end of broadcast)--------------------------- >TIP 2: Don't 'kill -9' the postmaster > > From pgsql-performance-owner@postgresql.org Sun Aug 21 12:01:15 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C3C501FEE591 for ; Sun, 21 Aug 2005 11:54:38 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 98444-10 for ; Sun, 21 Aug 2005 14:54:29 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id 463361FEE588 for ; Sun, 21 Aug 2005 11:54:29 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7LEsQbd020366; Sun, 21 Aug 2005 09:54:27 -0500 Subject: Re: extremly low memory usage From: Jeremiah Jahn To: John A Meinel Cc: Ron , postgres performance In-Reply-To: <4307E7A4.7090300@arbash-meinel.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <1124481386.27881.167.camel@bluejay.goodinassociates.com> <43064908.7010505@arbash-meinel.com> <1124564003.27881.229.camel@bluejay.goodinassociates.com> <6.2.3.4.0.20050820154422.06021150@pop.earthlink.net> <4307E7A4.7090300@arbash-meinel.com> Content-Type: text/plain Date: Sun, 21 Aug 2005 09:54:26 -0500 Message-Id: <1124636066.27881.238.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Content-Transfer-Encoding: 7bit Received-SPF: none (penguin.goodinassociates.com: 206.80.71.242 is neither permitted nor denied by domain of cs.earlham.edu) client-ip=206.80.71.242; envelope-from=jeremiah@cs.earlham.edu; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1034/Thu Aug 18 15:07:58 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.01 required=5 tests=[AWL=0.010] X-Spam-Level: X-Archive-Number: 200508/307 X-Sequence-Number: 14054 On Sat, 2005-08-20 at 21:32 -0500, John A Meinel wrote: > Ron wrote: > > At 02:53 PM 8/20/2005, Jeremiah Jahn wrote: > > > >> On Fri, 2005-08-19 at 16:03 -0500, John A Meinel wrote: > >> > Jeremiah Jahn wrote: > >> > > On Fri, 2005-08-19 at 12:18 -0500, John A Meinel wrote: > >> > > > >> > >> > > >> > > it's cached alright. I'm getting a read rate of about 150MB/sec. I > >> would > >> > > have thought is would be faster with my raid setup. I think I'm > >> going to > >> > > scrap the whole thing and get rid of LVM. I'll just do a straight > >> ext3 > >> > > system. Maybe that will help. Still trying to get suggestions for a > >> > > stripe size. > >> > > > > Well, since you can get a read of the RAID at 150MB/s, that means that > it is actual I/O speed. It may not be cached in RAM. Perhaps you could > try the same test, only using say 1G, which should be cached. [root@io pgsql]# time dd if=/dev/zero of=testfile bs=1024 count=1000000 1000000+0 records in 1000000+0 records out real 0m8.885s user 0m0.299s sys 0m6.998s [root@io pgsql]# time dd of=/dev/null if=testfile bs=1024 count=1000000 1000000+0 records in 1000000+0 records out real 0m1.654s user 0m0.232s sys 0m1.415s > > >> > > >> > I don't think 150MB/s is out of the realm for a 14 drive array. > >> > How fast is time dd if=/dev/zero of=testfile bs=8192 count=1000000 > >> > > >> time dd if=/dev/zero of=testfile bs=8192 count=1000000 > >> 1000000+0 records in > >> 1000000+0 records out > >> > >> real 1m24.248s > >> user 0m0.381s > >> sys 0m33.028s > >> > >> > >> > (That should create a 8GB file, which is too big to cache everything) > >> > And then how fast is: > >> > time dd if=testfile of=/dev/null bs=8192 count=1000000 > >> > >> time dd if=testfile of=/dev/null bs=8192 count=1000000 > >> 1000000+0 records in > >> 1000000+0 records out > >> > >> real 0m54.139s > >> user 0m0.326s > >> sys 0m8.916s > >> > >> > >> and on a second run: > >> > >> real 0m55.667s > >> user 0m0.341s > >> sys 0m9.013s > >> > >> > >> > > >> > That should give you a semi-decent way of measuring how fast the RAID > >> > system is, since it should be too big to cache in ram. > >> > >> about 150MB/Sec. Is there no better way to make this go faster...? > > I'm actually curious about PCI bus saturation at this point. Old 32-bit > 33MHz pci could only push 1Gbit = 100MB/s. Now, I'm guessing that this > is a higher performance system. But I'm really surprised that your write > speed is that close to your read speed. (100MB/s write, 150MB/s read). The raid array I have is currently set up to use a single channel. But I have dual controllers In the array. And dual external slots on the card. The machine is brand new and has pci-e backplane. > > > > > Assuming these are U320 15Krpm 147GB HDs, a RAID 10 array of 14 of them > > doing raw sequential IO like this should be capable of at > > ~7*75MB/s= 525MB/s using Seagate Cheetah 15K.4's, ~7*79MB/s= 553MB/s BTW I'm using Seagate Cheetah 15K.4's > > if using Fujitsu MAU's, and ~7*86MB/s= 602MB/s if using Maxtor Atlas 15K > > II's to devices external to the RAID array. > > I know I thought these were SATA drives, over 2 controllers. I could be > completely wrong, though. > > > > > _IF_ the controller setup is high powered enough to keep that kind of IO > > rate up. This will require a controller or controllers providing dual > > channel U320 bandwidth externally and quad channel U320 bandwidth > > internally. IOW, it needs a controller or controllers talking 64b > > 133MHz PCI-X, reasonably fast DSP/CPU units, and probably a decent sized > > IO buffer as well. > > > > AFAICT, the Dell PERC4 controllers use various flavors of the LSI Logic > > MegaRAID controllers. What I don't know is which exact one yours is, > > nor do I know if it (or any of the MegaRAID controllers) are high > > powered enough. PERC4eDC-PCI Express, 128MB Cache, 2-External Channels > > > > Talk to your HW supplier to make sure you have controllers adequate to > > your HD's. > > > > ...and yes, your average access time will be in the 5.5ms - 6ms range > > when doing a physical seek. > > Even with RAID, you want to minimize seeks and maximize sequential IO > > when accessing them. > > Best to not go to HD at all ;-) > > Well, certainly, if you can get more into RAM, you're always better off. > For writing, a battery-backed write cache, and for reading lots of > system RAM. I'm not really worried about the writing, it's the reading the reading that needs to be faster. > > > > > Hope this helps, > > Ron Peacetree > > > > John > =:-> -- Speak softly and carry a +6 two-handed sword. From pgsql-performance-owner@postgresql.org Sun Aug 21 12:56:38 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3D93A1FEE930 for ; Sun, 21 Aug 2005 12:52:23 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 32559-06 for ; Sun, 21 Aug 2005 15:52:05 +0000 (GMT) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by svr1.postgresql.org (Postfix) with ESMTP id 9193C1FEE927 for ; Sun, 21 Aug 2005 12:51:57 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20050821155156m92009hsc4e>; Sun, 21 Aug 2005 15:51:56 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 0797855F80; Sun, 21 Aug 2005 10:51:56 -0500 (CDT) Received: from [192.168.1.11] (liliana.arbash-meinel.com [192.168.1.11]) by juju.arbash-meinel.com (Postfix) with ESMTP id E223F55F7A; Sun, 21 Aug 2005 10:51:52 -0500 (CDT) Message-ID: <4308A318.1040106@arbash-meinel.com> Date: Sun, 21 Aug 2005 10:51:52 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.2 (Windows/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Jeremiah Jahn Cc: Ron , postgres performance Subject: Re: extremly low memory usage References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <1124481386.27881.167.camel@bluejay.goodinassociates.com> <43064908.7010505@arbash-meinel.com> <1124564003.27881.229.camel@bluejay.goodinassociates.com> <6.2.3.4.0.20050820154422.06021150@pop.earthlink.net> <4307E7A4.7090300@arbash-meinel.com> <1124636066.27881.238.camel@bluejay.goodinassociates.com> In-Reply-To: <1124636066.27881.238.camel@bluejay.goodinassociates.com> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig8D5A08ABEE109B148706B224" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.042 required=5 tests=[AWL=-0.008, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/308 X-Sequence-Number: 14055 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig8D5A08ABEE109B148706B224 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Jeremiah Jahn wrote: > On Sat, 2005-08-20 at 21:32 -0500, John A Meinel wrote: > >>Ron wrote: >> >>>At 02:53 PM 8/20/2005, Jeremiah Jahn wrote: >> >>Well, since you can get a read of the RAID at 150MB/s, that means that >>it is actual I/O speed. It may not be cached in RAM. Perhaps you could >>try the same test, only using say 1G, which should be cached. > > > [root@io pgsql]# time dd if=/dev/zero of=testfile bs=1024 count=1000000 > 1000000+0 records in > 1000000+0 records out > > real 0m8.885s > user 0m0.299s > sys 0m6.998s > [root@io pgsql]# time dd of=/dev/null if=testfile bs=1024 count=1000000 > 1000000+0 records in > 1000000+0 records out > > real 0m1.654s > user 0m0.232s > sys 0m1.415s > The write time seems about the same (but you only have 128MB of write cache), but your read jumped up to 620MB/s. So you drives do seem to be giving you 150MB/s. > ... >>I'm actually curious about PCI bus saturation at this point. Old 32-bit >>33MHz pci could only push 1Gbit = 100MB/s. Now, I'm guessing that this >>is a higher performance system. But I'm really surprised that your write >>speed is that close to your read speed. (100MB/s write, 150MB/s read). > > > The raid array I have is currently set up to use a single channel. But I > have dual controllers In the array. And dual external slots on the card. > The machine is brand new and has pci-e backplane. > > > > >>>Assuming these are U320 15Krpm 147GB HDs, a RAID 10 array of 14 of them >>>doing raw sequential IO like this should be capable of at >>> ~7*75MB/s= 525MB/s using Seagate Cheetah 15K.4's, ~7*79MB/s= 553MB/s > > BTW I'm using Seagate Cheetah 15K.4's > Now, are the numbers that Ron is quoting in megabytes or megabits? I'm guessing he knows what he is talking about, and is doing megabytes. 80MB/s sustained seems rather high for a hard-disk. Though this page: http://www.storagereview.com/articles/200411/20041116ST3146754LW_2.html Does seem to agree with that statement. (Between 56 and 93MB/s) And since U320 is a 320MB/s bus, it doesn't seem like anything there should be saturating. So why the low performance???? >> >>>_IF_ the controller setup is high powered enough to keep that kind of IO >>>rate up. This will require a controller or controllers providing dual >>>channel U320 bandwidth externally and quad channel U320 bandwidth >>>internally. IOW, it needs a controller or controllers talking 64b >>>133MHz PCI-X, reasonably fast DSP/CPU units, and probably a decent sized >>>IO buffer as well. >>> >>>AFAICT, the Dell PERC4 controllers use various flavors of the LSI Logic >>>MegaRAID controllers. What I don't know is which exact one yours is, >>>nor do I know if it (or any of the MegaRAID controllers) are high >>>powered enough. > > > PERC4eDC-PCI Express, 128MB Cache, 2-External Channels Do you know which card it is? Does it look like this one: http://www.lsilogic.com/products/megaraid/megaraid_320_2e.html Judging by the 320 speed, and 2 external controllers, that is my guess. They at least claim a theoretical max of 2GB/s. Which makes you wonder why reading from RAM is only able to get throughput of 600MB/s. Did you run it multiple times? On my windows system, I get just under 550MB/s for what should be cached, copying from /dev/zero to /dev/null I get 2.4GB/s (though that might be a no-op). On a similar linux machine, I'm able to get 1200MB/s for a cached file. (And 3GB/s for a zero=>null copy). John =:-> > > >>>Talk to your HW supplier to make sure you have controllers adequate to >>>your HD's. >>> >>>...and yes, your average access time will be in the 5.5ms - 6ms range >>>when doing a physical seek. >>>Even with RAID, you want to minimize seeks and maximize sequential IO >>>when accessing them. >>>Best to not go to HD at all ;-) >> >>Well, certainly, if you can get more into RAM, you're always better off. >>For writing, a battery-backed write cache, and for reading lots of >>system RAM. > > > I'm not really worried about the writing, it's the reading the reading > that needs to be faster. > > >>>Hope this helps, >>>Ron Peacetree >>> >> >>John >>=:-> --------------enig8D5A08ABEE109B148706B224 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDCKMYJdeBCYSNAAMRAqnOAJ9D7H2TzoMHAgZYfZo8DfaWPusXfACfd69d lnoanPccpm+iqovJ7p5Vt3o= =45DT -----END PGP SIGNATURE----- --------------enig8D5A08ABEE109B148706B224-- From pgsql-performance-owner@postgresql.org Sun Aug 21 15:32:55 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id DC0F91FEE955 for ; Sun, 21 Aug 2005 15:32:41 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 89561-07 for ; Sun, 21 Aug 2005 18:32:37 +0000 (GMT) Received: from apate.telenet-ops.be (apate.telenet-ops.be [195.130.132.57]) by svr1.postgresql.org (Postfix) with ESMTP id 7DBA21FEE954 for ; Sun, 21 Aug 2005 15:32:36 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by apate.telenet-ops.be (Postfix) with SMTP id 5073138315 for ; Sun, 21 Aug 2005 20:32:34 +0200 (CEST) Received: from [10.0.1.2] (d5152B1F3.access.telenet.be [81.82.177.243]) by apate.telenet-ops.be (Postfix) with ESMTP id CCD5D382DB for ; Sun, 21 Aug 2005 20:32:33 +0200 (CEST) Mime-Version: 1.0 (Apple Message framework v622) To: pgsql-performance@postgresql.org Message-Id: Content-Type: multipart/mixed; boundary=Apple-Mail-1-313063916 From: Yves Vindevogel Subject: (Re)-indexing on updates Date: Sun, 21 Aug 2005 20:32:31 +0200 X-Mailer: Apple Mail (2.622) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/309 X-Sequence-Number: 14056 --Apple-Mail-1-313063916 Content-Type: multipart/alternative; boundary=Apple-Mail-2-313063917 --Apple-Mail-2-313063917 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=ISO-8859-1; format=flowed Hi, Say I have a table with column A, B, C, D A has a unique index on it (primary key) B and C have a normal index on it D has no index If I perform a query like update tbl set D =3D 'whatever' ; that should make no difference on the indexes on the other columns,=20 right ? Or is there some kind of mechanism that does create a sort of new=20 record, thus makes the indexes go wild. Met vriendelijke groeten, Bien =E0 vous, Kind regards, Yves Vindevogel Implements --Apple-Mail-2-313063917 Content-Transfer-Encoding: quoted-printable Content-Type: text/enriched; charset=ISO-8859-1 Hi, Say I have a table with column A, B, C, D A has a unique index on it (primary key) B and C have a normal index on it D has no index If I perform a query like update tbl set D =3D 'whatever' ; that should make no difference on the indexes on the other columns, right ? Or is there some kind of mechanism that does create a sort of new record, thus makes the indexes go wild. Met vriendelijke groeten, Bien =E0 vous, Kind regards, Yves Vindevogel Implements = --Apple-Mail-2-313063917-- --Apple-Mail-1-313063916 Content-Transfer-Encoding: base64 Content-Type: image/tiff; x-unix-mode=0666; name="Pasted Graphic 2.tiff" Content-Disposition: inline; filename="Pasted Graphic 2.tiff" TU0AKgAAFciAP6BP5/wWDQeEQmFQuGQ2HQ+FP5+v9rOh5P9IMBuP8zK9tP8fJNlP8QoNjP8FndhP 8MnyVjJHSMrqFnv9BL1vP9ett2v98PuJxChUOiUWGQOCUalUujOV4PZ/qBluR/mdVtJ/jVHst/ho 9MF/g08WABHJfv8AWa0WoAHFfWuzgg6sB/hY9ysUohjv8qqVqv9LMZxv9ouV40zEYmHUjFY2HPx+ RNtOl4Rhftt/phiRs6LC/iA5LewnJeWg46W22+02cAHOwa26bC0HPY7TZ7G2G9dv8EHFdP8Om9YP 8rKJov9UtFzv9Dzh/sRvu5/vZ9PzHdeF4zsUqgQRlOR5v9DMBwv8uqe/jZIVwNHu6B0/WAKHvWGr RAA1ra0G1c/vfgANrdraXrTQIAA6JW2TUrQNz/ja0oAjO4YAjSWS0DOWawjys4RkIk4Lj2sAWEUZ B/ieUJqH+PZdnAf5aG06R5nsfLtqW7UaoWZJwukwLyi4VRsn+F5HGYf8QJWBQ7JWAI5tYti2Dm1T bQHAq0De1Q2FwtA1Fof4AjVDIBDQV0vDQWK0DY/oADYWsrNQN81NOuDZtYOM6ycs4Cjq+Q9JWE5D pGJpOpqP5dMwW5sHUf53xlHCExupjqoIbh3Hu55vp6ch3Hof4wlMrAdEWYh/gmP1RtXK0DDi2rXy nVzXjpVs8NM1g1vuNU2gANEugGMBRwu4YADTNsvwyAA3v+0k6LRWLb2dBUpjk2I4NVOx/gYPC6BU QiwCgSsSlWZpxH+ahzsOaBzvCe7qoMpKlUghd1n2f5oHId5/lQaJ0n+RBf3GLpUGwf4iksrg0FVF ICzlXT9QBB0tWO/+FjlKVWtfijTNRBsvDbhrdP3NUAwY38KFfkc0PuNjRWLLz8ytiTUVdVlnjlA2 MWPAWRAAM8LAAMjhhuR5ir4ULjigT5pn+P5co2T6pn+YBup6dp6xoh6kHie95modNOFkah0H+Qhd pyKBNmcf68r2Bo9GGtdpwIAUotM1Q4QMNzUDZhuVZbLsvy7KkqZZCnAb4AFcQY1C3LW1Wb5lOa2t ZvUt78NNjcMN83wEOMDczvuTzXNQ3QEN01WRL2b8XBYEDuugQEFUYgkpIo5FjIJTmhfZkHG8J0nm fB/sifp/EQYLBioUKsBkRKVgWPLX7stA4VnaDa+tZlZel0j+gCMpWS8MRULQNeIDY/4zTIAIyFX0 GFybZnr+rZ8p2a1s6tUNfADXXI1Yb8jLX+MNQk+1Az7zWqwVa/FaJrHpvSLeAQOhdE/l7CSJwrAb BXDXH+PkyA/gUiPGaftASWW6JeDgnGAsCmLNvcYqlz5+FchnZMrqGbeE3JoYhAMAKbEvOGeigtaE CDZwJcgxtiML0uMgQYzllKt2GhtP/EaH6cgAm+fnCtVEVXFOKNU54AC1B/tyLfBx4YJBGwhAAG5m yUoGH/TgxlZ7nGXw3YdEpYTPWOvaSq/JVEX28rEDQsGKDoGPx+QYnE1URg2paZY4dy7EHTx9cc41 upqIfwkjSgJWyW0MgBdK9I1DGAAwnS8G8+4cGYyUWWbKMg/gSiOjQG+FMQ44qqlmgBNUSYYMohKW 2ULFZaR8NdMFucoz/w/ZcAAMzgH+j/AKGV8QAZkhmWNM0AJ/I5zCiJEGYjF1qoGmxI5MqwT7JVlG buLULJhPXlbK+WMt3HwHmIbF8xaA0n6mvPgN7EJhx9Vcax+hsX6wGZuldNDnTSgQDoloGohS6BYE 62gPQqyRhlFE2gIYkCRgdD4WcAUsnxvlYklKgUCn4upi29JkkpG/pbYgtKbr86TLOnbLCF066ZLP kGsI+7NY9rRjZHB+0ew4M5S0AKT4FA8l0BaIUk4XHkj/EYLwzAtVyj/GwRYf49B8D6H+OMdw9R/j DG8vcTgx1xhpFTBoHYilRgaDygQAobXCynVVSSVdRYTLJls5CcIaqX0zpw/Kms76c2DSnTsNNPaA TbpggqOSCg4MkQwcAPiBAzimJqI4Xg3R/itX0P8bY66xD7MiYkgg5h4vAF2NgdY/xNGCH+HQVMIQ WiCNVM2L6Ap4s3ShFixMubA2HnjcUulhab3GuIbGxVjLlxXZoawMKFgsCiaSOUd6nB6XbH+PofKN LtqcH7eMf5AyimQOtegf4/TIEFvYP8dyjTm2eV1KdmNgrlIKr/cOxF+LjmRldTZY88L/U6P/YtyF +b82PDAhkOQsyNkHu8jS76NB4YXH+PvDUG7v3lIFeu8mFbuj6q8PXE2HEaXqIEUEgxnh2T2YhD/B WBZcUunVjO59yMB3PsHc3BONLemxwaP8NwsTMEPvVeMieJKvXmHZk8f48h5GHyYP8e+V3g3tvMQs VY01FLCxjgTHk8b943yBTOVoJEiXJzPgae1zr+2HwYhnIuRzrjczwP8eY8zDmKy7l8NOYcx40zKW nNswizytg+kXHc/qS48x9o7QbNMhZ0yMdvPBG7wmNz/jB6WYs436uFmbHhbFrAFgjhl4Y/g5IvH+ 0EkYFw+NuAHqmoet9bWJwPnDHEfM55E0udjTNW7uZ+y9p6L+oNe3Bxtoa4lACzgQee2kQ5Jwo1RK QPQfK8xejaJ6FG6w/wEh1NZJ+L6ydoH+zfqTUK0dK7AzsY7Yem9jaA0Fu2wVf9eJObqW+TxuwbCS K4JEYVnh2O/vWQM5K+wtCehCCAP5YABybV035xCAJIJqnrTyFqB9H8e1Lu/OumM870MTp3MGn9J8 g5Zsxhk2ZURLfGmrBAEg5m/CISIf4jRekblaCgRoyS0BjWNZVw6uZ70hnsrkND/Fbn3kWWtmz1ID a5NxyLYJ1xujdI3iZTm9dkYy4/P9yFKIcFoDQsbtKW4nT2YbLrpr450V6zTGem7N+YzXVzDZ0PSl ddIb8GQVrlD9sQjy4FKHVy0ZDDeLHCPWuuD/68Y0Vg1LX8c0bQSBiAq9JrVv2pYwZkzq6mqxDi8y XJgBkjZGftMMdUgm5PKyCBnogBI72hYgZlgrD6WWgMzPZkxGdUzfIfIzsdb66PXr/Jxp+XwRGlOL GnuJm2b0l7vo/dJeDL4N7qwWN+qn5CqedNMATuzZmObaad1s4dBMl/6wmG/Pt1+oAAYk2hToiuQc 5FxyjyRoHqH0xYKE+Q8k+UIgH0KAH+HSHqXmGwHSIuEGF0I25eWE8A96/g2akwhIk8Puf25mpi3w /IeG/M8y0O10/WgM3cpUS8l0mS+ocOYgDQS0ASDUTOB6EaJWDkXyXwFyL+GqGkPKHQHMOkxIXmIO HAHAJyyuKgvGIIHgU2H+G0GuHKH+FwF6KwEAtAH+CUEcLOAiNC2QnKcOZSZCmO9ongNep2sA3Y2W nY/KwE9hDazc8w8Sug9mh6cAZ0/UAEC0FKH+AOCOEoJQBeD8H+A2A+DYYGBuD2H+D6DmFCH+F0Fq hDB+PKGcGaaSGsGmJyGOGGg0EuEgP0CqCOEOH+BIA+DQH+AgBSDrD+CAEcH+AICwFNA+pamupfDQ /G1E2asfDksJDesNBCv012x+5WDmQSseeiAaCGEiH+BEAmC0H+A8AcCyH+A4AXGqAqAOCsLqAQCw H+BMAuDIH+CJEWPMCpFKCaCEEASEBFESA2AaC2H+AoAOCqK6AVGqA8AYCvGcAiC9HmBlHZBglBDt F80LF7BNDdBHDg2U0JGI0lGEseWsAgB6EYH+BUAkC4H+BKAuDDGcArH+BCAnH+A8AeC6K7HjG7H4 AoANHtG1G4A1GxGkAfHkBAAjJOBCAqC/FMAuDAH+BOApJ8A0BYDy2aWtGEzI1G2dKRIUwDGDF9BP DooBISnVImB9IsBRJHI2AuDHK3HHI4DFK3K6BGArI8BAAlJ8A8AhJqAlJxJBK3LDLBK9LmBNKCH+ AwBZEZIGcjGM3zKVIRL7BEH8zVKfKojrKlMCQVIk2jKuH/KzH/I5K7MjLFMpMnI5HGBIAvG/I5J9 MvMrK5M/K7LrJ9LxL1DXL5KYp1L/MM80g2wABMzWAGjVDZIc/XDrIiNjKtKxK1MtNBN7MlN9HFJ5 M3J7NDONMnNHLvLzKNKnMS5dNPMAsGpgLKLoAQNswAH8EAGAXGB0EobQAwEEJOblOa1uzJIfNvF9 MXFVMbMfOPODOBOBMxM1K3M7PfPdNFLtNLOZOcfjIOgVPIfeAaD2VGBUEY6ECyFQL+awXWH+HCHa KgW+KoAqDsPu3Mt8rxMO+fOjPTNzMZN3MhPtN/M/PlOJPrPhPvHBPzOXL3KPKhP69Mv4SmYWWs4m QyDGFUGsH+GeHEOkHW204SIED6FmaSBqeZHmQ3Fikw7iPwZWkCSs40Yg0CwTQAxpPUAhPZN5RDS1 PjOHPpRRRFOTP1RZSo5W1skw6MDa6g5omYZW94AYDkN+BOD+NKCoE6SKHy1WA8EOL2Z8cuhGTU/f SlArAu95L2cmiO1xSrQ7PXQ/S/S1RJS9RFUlNBTDRXNPRavw3St25kTXBiQsADBfUFNO/fUC6K+A N+9fTI2SkOfG6YmY9MZWDEFS+0TInFUQSfPI1NQ9MdSzRPUnS5MzRLUdRPUqD02RRaeoseDdSaQn ScQi9G/kPu+e/o+E8UQUoGv/IXMKsOfqZu8O/eV2S89E7QZ7XG8w/edGQYxiTeoSB3FKBQAlRBV9 S3LnWDUjXpN7LCBOAnJ8AyBXKKV0Z7KPKO6i764rXI9wS8DJVq6TAw+hP5MFBJDi0Ofq0i86mQPv XMDRWlBi6QVy6iAYCUEuJIBDESBGA0DNI2AxLjRCAxS4BGAvH5I5J2BLZdZVV9ZtZqDKJIA+DUH+ AiB4ES2a4pDFDE+fVIkg84ZylyTUn7RdYjIZYhQy14fipY9SDTWhSjRg6UAEC+fZGWEfEMBCDbFM AtJ9MzI9LlM8BFZjJ5ZpNBMzLCBJLKH+A+A3ZSAiByEMmcC0FO7Q7eTVSW+ectBY9HQ2oFDTNXak Ns9fIa0hPPVUfqfq5i3OjohIAEC6fYAWB6ERFUBSDxLuAtZSA7JSA/JoH/LPH+BIAtK7bZG4BGAt JCApH+A+AhJPGnJOA0AfH+AiBIDoWuBzc6AICvD8PxUAWJWWj2sfWwsPP9cXWzKc/PGE0jPQvwiE 9kWsAeB+EVI2AaCpdQAbGjGnHkA4AbJPHhI0A2AZI0A6AdI1drG+A6AfG/HhGjfVI1fLfJGpfBGq BGAZHsAvKJA+j7eZencVNSlZGBelINchalStIqbTV7J5I8BDLtdrJPfLfSAZHkA2AXfPfXENJTfl JrK0BFdjMoBOArX7UsYhUxaledgRcZgVBLMTepVU17StSxXlMrZzbgAtLDgrdoAlGiBBK1dhZZLi AxK/OFORRVNNhbhu0PhhafgTW1gXilgbgRhzUbV/OOAwDPFMAxe/ZrPriXXnWJidP3hiuZgPiphk eGA8EUJGWEc3MNhtgdUXSvi5XxS1Z3Zhe/MzbfjNRRUrieVpeejtOg0OseDeNYQexAIEE4GcX2B4 EkRKAMDsNjXSTWcVVVjvi1jzh1WHM/j8AvkAAvkFlHkLjVjdjZF4pPk0nANKLKQIBCESL2D6F6XG XgvWqwHQMqEsGIRaL6RSAgD6bcQAbyNRBkLQDCWNhdNxkPj1V5h3i7N7lLlPlTmtUpjTTHak1tUE DHDyQMDWN+ggQIByEiJOEGOcGIG8Okq7CMXay2KMG8HaU4rcMGCIE8aSBcEWL2Ci4ccggbKXhxlD j3jPPdmxbdLnm3PxNJhZkO1CseseDUNKBIEMJWB8EqhCBqEuGgH+DiFiL+GXQcINAEIhl4McIkXc IICurWWuDulCNiDW85VVi3mplHmuH/j/oZM9odRTohkNNQnjYIc6QIAILmH+BgEOLOtabCUcIPpV qiIOHMKeOQGoJ6C+FaMwAuEJT474DYTfliJQrbgjmrj5OBoXkDobrTJ/LsArX+LQDI72nBlkJQLv C2FBpAEoGOMGGhqsvKH7nlqpqlnpsKKMHgJ+tgGQPKCmFIaSAwD8LOA4EGJWBiCdEFH1G/iPp0Av rXlRrboTdXLDdrI0BSBkD4H/n8JOAkD7o0EwhCECF+PKG6HceBsQRtsPtyOuq4/+22RcFYJGBkBK Ddp4AtI8A2AdHkA/H9uPZZrVp5lNp9OFMzK7iLJ9HhHlgrJ8BSA3FSEWECe+vNsVnkXdt4MVqnvR qiHmHkKgF+F0JqEMD8fECQB4D7uPI8AyAZGjubJ8BAAqClbqAng4AbfZdsKyBTuMDiDOE2H+FmFe JGHMU1vXwrvVwrwxCKagF3pADyDiFEH+CWCAEHJ+A3G4B+BmDmH+DQC7EEFGE4NKHYHUXvwxxqIj t3xtxyIaw0ImGwGus8HgUZx1yGavxxyJyPyRySKIKQFuFibQGOGEYFyVynypymKQCaB/xIDkDOE8 IQFSFEVGE+EsQIIWGEF4L+EwEeP6IOE4EoN2DOC0ErpCDLy6DYDAEyH+FiFU6EHyq6H+DqDUFAH+ G4GwOWIWw0IIEMD68HzsE0H+DgDIE4H+DaDD0cG8G4X2FgFS6EFGE2LAIWG0GsHMH+EoEWTaIOFn uEH+DKCzzkDjzoH+DJ1aH+FuFgbRsGIIEKD2fZyCrEIOGUGMI2E8EqNLz6XmDsDXEgDUC6EwH+DW C9zx1ZzkF4FuOOHUq0D8Dnb8DeDHwd0h0kEeEKQsKQCkCLFKDv2SIQD6DkFJENg2H/ygYFxkPCBu BQDkH+CwCXIsIOCiCJb2CQB0D+H+FEE0LoEYEGTOCgCGEKH/2uMrGfI8GEF7RyIWHwOoH+CEBptU EaEITOFCEyLOF0FoJr16aUDpb9fbH+GAF0RSz2KgCABjdACSB4ED3V3YH+C8CjbCFAEwLP006EG8 G2X35LbnK6B/5iH+GcGSs8FoFehCC0CYEavXAT5WRSDyDdxACKBtvwFwFkJqHEHAUVxeLACSB3HZ 2GNKEyEgP6EsEaNEKQCn3MH+DuDZEgIOEUEAZMDuDbxB7jFL5zbCDQC2Et2cC/zwIP8AH+BoBKDf 1iCxEEC4CfFgEqEYNFCeU4BWA2DSXIGgXGIWu+XmCQB35r2R0EEgEOQyIWEOD88GECDvVmCyCZIs DACpGaE0EkN+DP8GIQEGDyFUH/3LFL0Ue+HkHh18IMHPwoCqCRe54ILoCMBvEKC6Cl8CC18IIWFa FKL2DGCtEEIXE0MH2+H+EKD4fZ10fZ374Z7h7l7p7sIMEF9eRd6d1p1t4H4L/ohCDECsEkIQIAVC MiX+dTUoH++n0/X++4W/4hEHe7Xo/xQGTO/zQXEu/0Cd1S/0Yglg/3g73q/yWPUE/zsa1C/0IeVW /0GeFU/3G4Xa/0MfVY/1or2bQlbRUiiVpJpQ/y+U0jEYggzzOSkRUM/0KfJqfzop3+p1Aw3+53I7 n+UCJWak3m26H+OxYdX+Xiij6leVRY3+YKheaks1Yy3+VCPBEIe5zM5qOxbdICAADgEAAAMAAAAB ADgAAAEBAAMAAAABAEAAAAECAAMAAAAEAAAWdgEDAAMAAAABAAUAAAEGAAMAAAABAAIAAAERAAQA AAABAAAACAEVAAMAAAABAAQAAAEWAAQAAAABAAAAkgEXAAQAAAABAAAVwAEaAAUAAAABAAAWfgEb AAUAAAABAAAWhgEcAAMAAAABAAEAAAEoAAMAAAABAAIAAAFSAAMAAAABAAEAAAAAAAAACAAIAAgA CAAK/IAAACcQAAr8gAAAJxA= --Apple-Mail-1-313063916 Content-Type: multipart/alternative; boundary=Apple-Mail-3-313063918 --Apple-Mail-3-313063918 Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=US-ASCII; format=flowed Mail: yves.vindevogel@implements.be - Mobile: +32 (478) 80 82 91 Kempische Steenweg 206 - 3500 Hasselt - Tel-Fax: +32 (11) 43 55 76 Web: http://www.implements.be First they ignore you. Then they laugh at you. Then they fight you. Then you win. Mahatma Ghandi. --Apple-Mail-3-313063918 Content-Transfer-Encoding: 7bit Content-Type: text/enriched; charset=US-ASCII Mail: yves.vindevogel@implements.be - Mobile: +32 (478) 80 82 91 Kempische Steenweg 206 - 3500 Hasselt - Tel-Fax: +32 (11) 43 55 76 Web: http://www.implements.be First they ignore you. Then they laugh at you. Then they fight you. Then you win. Mahatma Ghandi. --Apple-Mail-3-313063918-- --Apple-Mail-1-313063916-- From pgsql-performance-owner@postgresql.org Sun Aug 21 17:37:53 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 19D19D729B for ; Sun, 21 Aug 2005 17:37:41 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 24581-05 for ; Sun, 21 Aug 2005 20:37:38 +0000 (GMT) Received: from mail28.sea5.speakeasy.net (mail28.sea5.speakeasy.net [69.17.117.30]) by svr1.postgresql.org (Postfix) with ESMTP id 9D6E71FEE071 for ; Sun, 21 Aug 2005 16:05:28 -0300 (ADT) Received: (qmail 5954 invoked from network); 21 Aug 2005 19:05:28 -0000 Received: from dsl081-060-184.sfo1.dsl.speakeasy.net (HELO noodles) ([64.81.60.184]) (envelope-sender ) by mail28.sea5.speakeasy.net (qmail-ldap-1.03) with RC4-MD5 encrypted SMTP for ; 21 Aug 2005 19:05:28 -0000 Subject: Re: (Re)-indexing on updates From: "Jeffrey W. Baker" To: Yves Vindevogel Cc: pgsql-performance@postgresql.org In-Reply-To: References: Content-Type: text/plain Date: Sun, 21 Aug 2005 12:06:05 -0700 Message-Id: <1124651165.5990.1.camel@noodles> Mime-Version: 1.0 X-Mailer: Evolution 2.2.1.1 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/312 X-Sequence-Number: 14059 On Sun, 2005-08-21 at 20:32 +0200, Yves Vindevogel wrote: > > > ______________________________________________________________________ > > Hi, > > Say I have a table with column A, B, C, D > A has a unique index on it (primary key) > B and C have a normal index on it > D has no index > > If I perform a query like update tbl set D = 'whatever' ; > that should make no difference on the indexes on the other columns, > right ? What postgresql does on update is to make a new record, so there will be two records in your table and two records in your index. You would need to vacuum the table to mark the space for the old record free, and you would need to reindex the table to shrink the index. > > Or is there some kind of mechanism that does create a sort of new > record, thus makes the indexes go wild. Yes. -jwb From pgsql-performance-owner@postgresql.org Sun Aug 21 16:36:25 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A788B1FEF0F4 for ; Sun, 21 Aug 2005 16:36:15 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 12767-04 for ; Sun, 21 Aug 2005 19:36:08 +0000 (GMT) Received: from asia.telenet-ops.be (asia.telenet-ops.be [195.130.132.59]) by svr1.postgresql.org (Postfix) with ESMTP id 737871FEF0E1 for ; Sun, 21 Aug 2005 16:36:07 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by asia.telenet-ops.be (Postfix) with SMTP id CD1272241EE for ; Sun, 21 Aug 2005 21:36:06 +0200 (MEST) Received: from [10.0.1.2] (d5152B1F3.access.telenet.be [81.82.177.243]) by asia.telenet-ops.be (Postfix) with ESMTP id C22652241C6 for ; Sun, 21 Aug 2005 21:36:05 +0200 (MEST) Mime-Version: 1.0 (Apple Message framework v622) To: pgsql-performance@postgresql.org Message-Id: <7d4e3af17fe3d48a741969cea5f462e3@implements.be> Content-Type: multipart/mixed; boundary=Apple-Mail-11-316875059 From: Yves Vindevogel Subject: Fwd: (Re)-indexing on updates Date: Sun, 21 Aug 2005 21:36:03 +0200 X-Mailer: Apple Mail (2.622) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/310 X-Sequence-Number: 14057 --Apple-Mail-11-316875059 Content-Type: multipart/alternative; boundary=Apple-Mail-12-316875060 --Apple-Mail-12-316875060 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=ISO-8859-1; delsp=yes; format=flowed I always forget that this goes to the writer itself and not to the =20 group. > > > Ok, this is a major setback in some of my procedures. > =46rom time to time, I must update one field in about 10% of the = records. > So this will take time. > > How can I work around that ? > > Some personal opinions ... > 1) Drop indexes, run update, create indexes, vacuum > 2) Move the field to another table and use joins ? I could delete the = =20 > records when needed and add them again > > > This mechanism, of inserting a new record and marking the old one, is =20= > that data kept somewhere where I can "see" it ? > I need for one app a trace of all my changes in the database. I have =20= > a set of triggers to do that for the moment on each table. > Could I use that mechanism somehow to avoid my triggers ? > Any documentation on that mechanism (hacker stuff like what tables are = =20 > used) ? > Any good books on stuff like this ? I love to read and know how the =20= > inside mechanics work. > > Tnx > > > > On 21 Aug 2005, at 21:06, Jeffrey W. Baker wrote: > >> On Sun, 2005-08-21 at 20:32 +0200, Yves Vindevogel wrote: >>> >>> >>> = _____________________________________________________________________=20 >>> _ >>> >>> Hi, >>> >>> Say I have a table with column A, B, C, D >>> A has a unique index on it (primary key) >>> B and C have a normal index on it >>> D has no index >>> >>> If I perform a query like update tbl set D =3D 'whatever' ; >>> that should make no difference on the indexes on the other columns, >>> right ? >> >> What postgresql does on update is to make a new record, so there will = =20 >> be >> two records in your table and two records in your index. You would =20= >> need >> to vacuum the table to mark the space for the old record free, and = you >> would need to reindex the table to shrink the index. >> >>> >>> Or is there some kind of mechanism that does create a sort of new >>> record, thus makes the indexes go wild. >> >> Yes. >> >> -jwb >> >> >> > Met vriendelijke groeten, > Bien =E0 vous, > Kind regards, > > Yves Vindevogel > Implements > --Apple-Mail-12-316875060 Content-Transfer-Encoding: quoted-printable Content-Type: text/enriched; charset=ISO-8859-1 I always forget that this goes to the writer itself and not to the group. Ok, this is a major setback in some of my procedures. =46rom time to time, I must update one field in about 10% of the = records. So this will take time. How can I work around that ? Some personal opinions ... 1) Drop indexes, run update, create indexes, vacuum 2) Move the field to another table and use joins ? I could delete the records when needed and add them again=20 This mechanism, of inserting a new record and marking the old one, is that data kept somewhere where I can "see" it ? I need for one app a trace of all my changes in the database. I have a set of triggers to do that for the moment on each table. Could I use that mechanism somehow to avoid my triggers ? Any documentation on that mechanism (hacker stuff like what tables are used) ? Any good books on stuff like this ? I love to read and know how the inside mechanics work. Tnx On 21 Aug 2005, at 21:06, Jeffrey W. Baker wrote: On Sun, 2005-08-21 at 20:32 +0200, Yves Vindevogel wrote: ______________________________________________________________________ Hi, Say I have a table with column A, B, C, D A has a unique index on it (primary key) B and C have a normal index on it D has no index If I perform a query like update tbl set D =3D 'whatever' ; that should make no difference on the indexes on the other columns, right ? What postgresql does on update is to make a new record, so there will be two records in your table and two records in your index. You would need to vacuum the table to mark the space for the old record free, and you would need to reindex the table to shrink the index. Or is there some kind of mechanism that does create a sort of new record, thus makes the indexes go wild. Yes. -jwb Met vriendelijke groeten, Bien =E0 vous, Kind regards, Yves Vindevogel Implements = --Apple-Mail-12-316875060-- --Apple-Mail-11-316875059 Content-Transfer-Encoding: base64 Content-Type: image/tiff; x-unix-mode=0666; name="Pasted Graphic 2.tiff" Content-Disposition: inline; filename="Pasted Graphic 2.tiff" TU0AKgAAFciAP6BP5/wWDQeEQmFQuGQ2HQ+FP5+v9rOh5P9IMBuP8zK9tP8fJNlP8QoNjP8FndhP 8MnyVjJHSMrqFnv9BL1vP9ett2v98PuJxChUOiUWGQOCUalUujOV4PZ/qBluR/mdVtJ/jVHst/ho 9MF/g08WABHJfv8AWa0WoAHFfWuzgg6sB/hY9ysUohjv8qqVqv9LMZxv9ouV40zEYmHUjFY2HPx+ RNtOl4Rhftt/phiRs6LC/iA5LewnJeWg46W22+02cAHOwa26bC0HPY7TZ7G2G9dv8EHFdP8Om9YP 8rKJov9UtFzv9Dzh/sRvu5/vZ9PzHdeF4zsUqgQRlOR5v9DMBwv8uqe/jZIVwNHu6B0/WAKHvWGr RAA1ra0G1c/vfgANrdraXrTQIAA6JW2TUrQNz/ja0oAjO4YAjSWS0DOWawjys4RkIk4Lj2sAWEUZ B/ieUJqH+PZdnAf5aG06R5nsfLtqW7UaoWZJwukwLyi4VRsn+F5HGYf8QJWBQ7JWAI5tYti2Dm1T bQHAq0De1Q2FwtA1Fof4AjVDIBDQV0vDQWK0DY/oADYWsrNQN81NOuDZtYOM6ycs4Cjq+Q9JWE5D pGJpOpqP5dMwW5sHUf53xlHCExupjqoIbh3Hu55vp6ch3Hof4wlMrAdEWYh/gmP1RtXK0DDi2rXy nVzXjpVs8NM1g1vuNU2gANEugGMBRwu4YADTNsvwyAA3v+0k6LRWLb2dBUpjk2I4NVOx/gYPC6BU QiwCgSsSlWZpxH+ahzsOaBzvCe7qoMpKlUghd1n2f5oHId5/lQaJ0n+RBf3GLpUGwf4iksrg0FVF ICzlXT9QBB0tWO/+FjlKVWtfijTNRBsvDbhrdP3NUAwY38KFfkc0PuNjRWLLz8ytiTUVdVlnjlA2 MWPAWRAAM8LAAMjhhuR5ir4ULjigT5pn+P5co2T6pn+YBup6dp6xoh6kHie95modNOFkah0H+Qhd pyKBNmcf68r2Bo9GGtdpwIAUotM1Q4QMNzUDZhuVZbLsvy7KkqZZCnAb4AFcQY1C3LW1Wb5lOa2t ZvUt78NNjcMN83wEOMDczvuTzXNQ3QEN01WRL2b8XBYEDuugQEFUYgkpIo5FjIJTmhfZkHG8J0nm fB/sifp/EQYLBioUKsBkRKVgWPLX7stA4VnaDa+tZlZel0j+gCMpWS8MRULQNeIDY/4zTIAIyFX0 GFybZnr+rZ8p2a1s6tUNfADXXI1Yb8jLX+MNQk+1Az7zWqwVa/FaJrHpvSLeAQOhdE/l7CSJwrAb BXDXH+PkyA/gUiPGaftASWW6JeDgnGAsCmLNvcYqlz5+FchnZMrqGbeE3JoYhAMAKbEvOGeigtaE CDZwJcgxtiML0uMgQYzllKt2GhtP/EaH6cgAm+fnCtVEVXFOKNU54AC1B/tyLfBx4YJBGwhAAG5m yUoGH/TgxlZ7nGXw3YdEpYTPWOvaSq/JVEX28rEDQsGKDoGPx+QYnE1URg2paZY4dy7EHTx9cc41 upqIfwkjSgJWyW0MgBdK9I1DGAAwnS8G8+4cGYyUWWbKMg/gSiOjQG+FMQ44qqlmgBNUSYYMohKW 2ULFZaR8NdMFucoz/w/ZcAAMzgH+j/AKGV8QAZkhmWNM0AJ/I5zCiJEGYjF1qoGmxI5MqwT7JVlG buLULJhPXlbK+WMt3HwHmIbF8xaA0n6mvPgN7EJhx9Vcax+hsX6wGZuldNDnTSgQDoloGohS6BYE 62gPQqyRhlFE2gIYkCRgdD4WcAUsnxvlYklKgUCn4upi29JkkpG/pbYgtKbr86TLOnbLCF066ZLP kGsI+7NY9rRjZHB+0ew4M5S0AKT4FA8l0BaIUk4XHkj/EYLwzAtVyj/GwRYf49B8D6H+OMdw9R/j DG8vcTgx1xhpFTBoHYilRgaDygQAobXCynVVSSVdRYTLJls5CcIaqX0zpw/Kms76c2DSnTsNNPaA TbpggqOSCg4MkQwcAPiBAzimJqI4Xg3R/itX0P8bY66xD7MiYkgg5h4vAF2NgdY/xNGCH+HQVMIQ WiCNVM2L6Ap4s3ShFixMubA2HnjcUulhab3GuIbGxVjLlxXZoawMKFgsCiaSOUd6nB6XbH+PofKN LtqcH7eMf5AyimQOtegf4/TIEFvYP8dyjTm2eV1KdmNgrlIKr/cOxF+LjmRldTZY88L/U6P/YtyF +b82PDAhkOQsyNkHu8jS76NB4YXH+PvDUG7v3lIFeu8mFbuj6q8PXE2HEaXqIEUEgxnh2T2YhD/B WBZcUunVjO59yMB3PsHc3BONLemxwaP8NwsTMEPvVeMieJKvXmHZk8f48h5GHyYP8e+V3g3tvMQs VY01FLCxjgTHk8b943yBTOVoJEiXJzPgae1zr+2HwYhnIuRzrjczwP8eY8zDmKy7l8NOYcx40zKW nNswizytg+kXHc/qS48x9o7QbNMhZ0yMdvPBG7wmNz/jB6WYs436uFmbHhbFrAFgjhl4Y/g5IvH+ 0EkYFw+NuAHqmoet9bWJwPnDHEfM55E0udjTNW7uZ+y9p6L+oNe3Bxtoa4lACzgQee2kQ5Jwo1RK QPQfK8xejaJ6FG6w/wEh1NZJ+L6ydoH+zfqTUK0dK7AzsY7Yem9jaA0Fu2wVf9eJObqW+TxuwbCS K4JEYVnh2O/vWQM5K+wtCehCCAP5YABybV035xCAJIJqnrTyFqB9H8e1Lu/OumM870MTp3MGn9J8 g5Zsxhk2ZURLfGmrBAEg5m/CISIf4jRekblaCgRoyS0BjWNZVw6uZ70hnsrkND/Fbn3kWWtmz1ID a5NxyLYJ1xujdI3iZTm9dkYy4/P9yFKIcFoDQsbtKW4nT2YbLrpr450V6zTGem7N+YzXVzDZ0PSl ddIb8GQVrlD9sQjy4FKHVy0ZDDeLHCPWuuD/68Y0Vg1LX8c0bQSBiAq9JrVv2pYwZkzq6mqxDi8y XJgBkjZGftMMdUgm5PKyCBnogBI72hYgZlgrD6WWgMzPZkxGdUzfIfIzsdb66PXr/Jxp+XwRGlOL GnuJm2b0l7vo/dJeDL4N7qwWN+qn5CqedNMATuzZmObaad1s4dBMl/6wmG/Pt1+oAAYk2hToiuQc 5FxyjyRoHqH0xYKE+Q8k+UIgH0KAH+HSHqXmGwHSIuEGF0I25eWE8A96/g2akwhIk8Puf25mpi3w /IeG/M8y0O10/WgM3cpUS8l0mS+ocOYgDQS0ASDUTOB6EaJWDkXyXwFyL+GqGkPKHQHMOkxIXmIO HAHAJyyuKgvGIIHgU2H+G0GuHKH+FwF6KwEAtAH+CUEcLOAiNC2QnKcOZSZCmO9ongNep2sA3Y2W nY/KwE9hDazc8w8Sug9mh6cAZ0/UAEC0FKH+AOCOEoJQBeD8H+A2A+DYYGBuD2H+D6DmFCH+F0Fq hDB+PKGcGaaSGsGmJyGOGGg0EuEgP0CqCOEOH+BIA+DQH+AgBSDrD+CAEcH+AICwFNA+pamupfDQ /G1E2asfDksJDesNBCv012x+5WDmQSseeiAaCGEiH+BEAmC0H+A8AcCyH+A4AXGqAqAOCsLqAQCw H+BMAuDIH+CJEWPMCpFKCaCEEASEBFESA2AaC2H+AoAOCqK6AVGqA8AYCvGcAiC9HmBlHZBglBDt F80LF7BNDdBHDg2U0JGI0lGEseWsAgB6EYH+BUAkC4H+BKAuDDGcArH+BCAnH+A8AeC6K7HjG7H4 AoANHtG1G4A1GxGkAfHkBAAjJOBCAqC/FMAuDAH+BOApJ8A0BYDy2aWtGEzI1G2dKRIUwDGDF9BP DooBISnVImB9IsBRJHI2AuDHK3HHI4DFK3K6BGArI8BAAlJ8A8AhJqAlJxJBK3LDLBK9LmBNKCH+ AwBZEZIGcjGM3zKVIRL7BEH8zVKfKojrKlMCQVIk2jKuH/KzH/I5K7MjLFMpMnI5HGBIAvG/I5J9 MvMrK5M/K7LrJ9LxL1DXL5KYp1L/MM80g2wABMzWAGjVDZIc/XDrIiNjKtKxK1MtNBN7MlN9HFJ5 M3J7NDONMnNHLvLzKNKnMS5dNPMAsGpgLKLoAQNswAH8EAGAXGB0EobQAwEEJOblOa1uzJIfNvF9 MXFVMbMfOPODOBOBMxM1K3M7PfPdNFLtNLOZOcfjIOgVPIfeAaD2VGBUEY6ECyFQL+awXWH+HCHa KgW+KoAqDsPu3Mt8rxMO+fOjPTNzMZN3MhPtN/M/PlOJPrPhPvHBPzOXL3KPKhP69Mv4SmYWWs4m QyDGFUGsH+GeHEOkHW204SIED6FmaSBqeZHmQ3Fikw7iPwZWkCSs40Yg0CwTQAxpPUAhPZN5RDS1 PjOHPpRRRFOTP1RZSo5W1skw6MDa6g5omYZW94AYDkN+BOD+NKCoE6SKHy1WA8EOL2Z8cuhGTU/f SlArAu95L2cmiO1xSrQ7PXQ/S/S1RJS9RFUlNBTDRXNPRavw3St25kTXBiQsADBfUFNO/fUC6K+A N+9fTI2SkOfG6YmY9MZWDEFS+0TInFUQSfPI1NQ9MdSzRPUnS5MzRLUdRPUqD02RRaeoseDdSaQn ScQi9G/kPu+e/o+E8UQUoGv/IXMKsOfqZu8O/eV2S89E7QZ7XG8w/edGQYxiTeoSB3FKBQAlRBV9 S3LnWDUjXpN7LCBOAnJ8AyBXKKV0Z7KPKO6i764rXI9wS8DJVq6TAw+hP5MFBJDi0Ofq0i86mQPv XMDRWlBi6QVy6iAYCUEuJIBDESBGA0DNI2AxLjRCAxS4BGAvH5I5J2BLZdZVV9ZtZqDKJIA+DUH+ AiB4ES2a4pDFDE+fVIkg84ZylyTUn7RdYjIZYhQy14fipY9SDTWhSjRg6UAEC+fZGWEfEMBCDbFM AtJ9MzI9LlM8BFZjJ5ZpNBMzLCBJLKH+A+A3ZSAiByEMmcC0FO7Q7eTVSW+ectBY9HQ2oFDTNXak Ns9fIa0hPPVUfqfq5i3OjohIAEC6fYAWB6ERFUBSDxLuAtZSA7JSA/JoH/LPH+BIAtK7bZG4BGAt JCApH+A+AhJPGnJOA0AfH+AiBIDoWuBzc6AICvD8PxUAWJWWj2sfWwsPP9cXWzKc/PGE0jPQvwiE 9kWsAeB+EVI2AaCpdQAbGjGnHkA4AbJPHhI0A2AZI0A6AdI1drG+A6AfG/HhGjfVI1fLfJGpfBGq BGAZHsAvKJA+j7eZencVNSlZGBelINchalStIqbTV7J5I8BDLtdrJPfLfSAZHkA2AXfPfXENJTfl JrK0BFdjMoBOArX7UsYhUxaledgRcZgVBLMTepVU17StSxXlMrZzbgAtLDgrdoAlGiBBK1dhZZLi AxK/OFORRVNNhbhu0PhhafgTW1gXilgbgRhzUbV/OOAwDPFMAxe/ZrPriXXnWJidP3hiuZgPiphk eGA8EUJGWEc3MNhtgdUXSvi5XxS1Z3Zhe/MzbfjNRRUrieVpeejtOg0OseDeNYQexAIEE4GcX2B4 EkRKAMDsNjXSTWcVVVjvi1jzh1WHM/j8AvkAAvkFlHkLjVjdjZF4pPk0nANKLKQIBCESL2D6F6XG XgvWqwHQMqEsGIRaL6RSAgD6bcQAbyNRBkLQDCWNhdNxkPj1V5h3i7N7lLlPlTmtUpjTTHak1tUE DHDyQMDWN+ggQIByEiJOEGOcGIG8Okq7CMXay2KMG8HaU4rcMGCIE8aSBcEWL2Ci4ccggbKXhxlD j3jPPdmxbdLnm3PxNJhZkO1CseseDUNKBIEMJWB8EqhCBqEuGgH+DiFiL+GXQcINAEIhl4McIkXc IICurWWuDulCNiDW85VVi3mplHmuH/j/oZM9odRTohkNNQnjYIc6QIAILmH+BgEOLOtabCUcIPpV qiIOHMKeOQGoJ6C+FaMwAuEJT474DYTfliJQrbgjmrj5OBoXkDobrTJ/LsArX+LQDI72nBlkJQLv C2FBpAEoGOMGGhqsvKH7nlqpqlnpsKKMHgJ+tgGQPKCmFIaSAwD8LOA4EGJWBiCdEFH1G/iPp0Av rXlRrboTdXLDdrI0BSBkD4H/n8JOAkD7o0EwhCECF+PKG6HceBsQRtsPtyOuq4/+22RcFYJGBkBK Ddp4AtI8A2AdHkA/H9uPZZrVp5lNp9OFMzK7iLJ9HhHlgrJ8BSA3FSEWECe+vNsVnkXdt4MVqnvR qiHmHkKgF+F0JqEMD8fECQB4D7uPI8AyAZGjubJ8BAAqClbqAng4AbfZdsKyBTuMDiDOE2H+FmFe JGHMU1vXwrvVwrwxCKagF3pADyDiFEH+CWCAEHJ+A3G4B+BmDmH+DQC7EEFGE4NKHYHUXvwxxqIj t3xtxyIaw0ImGwGus8HgUZx1yGavxxyJyPyRySKIKQFuFibQGOGEYFyVynypymKQCaB/xIDkDOE8 IQFSFEVGE+EsQIIWGEF4L+EwEeP6IOE4EoN2DOC0ErpCDLy6DYDAEyH+FiFU6EHyq6H+DqDUFAH+ G4GwOWIWw0IIEMD68HzsE0H+DgDIE4H+DaDD0cG8G4X2FgFS6EFGE2LAIWG0GsHMH+EoEWTaIOFn uEH+DKCzzkDjzoH+DJ1aH+FuFgbRsGIIEKD2fZyCrEIOGUGMI2E8EqNLz6XmDsDXEgDUC6EwH+DW C9zx1ZzkF4FuOOHUq0D8Dnb8DeDHwd0h0kEeEKQsKQCkCLFKDv2SIQD6DkFJENg2H/ygYFxkPCBu BQDkH+CwCXIsIOCiCJb2CQB0D+H+FEE0LoEYEGTOCgCGEKH/2uMrGfI8GEF7RyIWHwOoH+CEBptU EaEITOFCEyLOF0FoJr16aUDpb9fbH+GAF0RSz2KgCABjdACSB4ED3V3YH+C8CjbCFAEwLP006EG8 G2X35LbnK6B/5iH+GcGSs8FoFehCC0CYEavXAT5WRSDyDdxACKBtvwFwFkJqHEHAUVxeLACSB3HZ 2GNKEyEgP6EsEaNEKQCn3MH+DuDZEgIOEUEAZMDuDbxB7jFL5zbCDQC2Et2cC/zwIP8AH+BoBKDf 1iCxEEC4CfFgEqEYNFCeU4BWA2DSXIGgXGIWu+XmCQB35r2R0EEgEOQyIWEOD88GECDvVmCyCZIs DACpGaE0EkN+DP8GIQEGDyFUH/3LFL0Ue+HkHh18IMHPwoCqCRe54ILoCMBvEKC6Cl8CC18IIWFa FKL2DGCtEEIXE0MH2+H+EKD4fZ10fZ374Z7h7l7p7sIMEF9eRd6d1p1t4H4L/ohCDECsEkIQIAVC MiX+dTUoH++n0/X++4W/4hEHe7Xo/xQGTO/zQXEu/0Cd1S/0Yglg/3g73q/yWPUE/zsa1C/0IeVW /0GeFU/3G4Xa/0MfVY/1or2bQlbRUiiVpJpQ/y+U0jEYggzzOSkRUM/0KfJqfzop3+p1Aw3+53I7 n+UCJWak3m26H+OxYdX+Xiij6leVRY3+YKheaks1Yy3+VCPBEIe5zM5qOxbdICAADgEAAAMAAAAB ADgAAAEBAAMAAAABAEAAAAECAAMAAAAEAAAWdgEDAAMAAAABAAUAAAEGAAMAAAABAAIAAAERAAQA AAABAAAACAEVAAMAAAABAAQAAAEWAAQAAAABAAAAkgEXAAQAAAABAAAVwAEaAAUAAAABAAAWfgEb AAUAAAABAAAWhgEcAAMAAAABAAEAAAEoAAMAAAABAAIAAAFSAAMAAAABAAEAAAAAAAAACAAIAAgA CAAK/IAAACcQAAr8gAAAJxA= --Apple-Mail-11-316875059 Content-Type: multipart/alternative; boundary=Apple-Mail-13-316875061 --Apple-Mail-13-316875061 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=ISO-8859-1; format=flowed > > > Mail: yves.vindevogel@implements.be - Mobile: +32 (478) 80 82 91 > > Kempische Steenweg 206 - 3500 Hasselt - Tel-Fax: +32 (11) 43 55 76 > > Web: http://www.implements.be > > First they ignore you. Then they laugh at you. Then they fight you. =20= > Then you win. > Mahatma Ghandi. > Met vriendelijke groeten, Bien =E0 vous, Kind regards, Yves Vindevogel Implements --Apple-Mail-13-316875061 Content-Transfer-Encoding: quoted-printable Content-Type: text/enriched; charset=ISO-8859-1 Mail: yves.vindevogel@implements.be - Mobile: +32 (478) 80 82 91 Kempische Steenweg 206 - 3500 Hasselt - Tel-Fax: +32 (11) 43 55 76 Web: http://www.implements.be First they ignore you. Then they laugh at you. Then they fight you.=20 Then you win. Mahatma Ghandi. Met vriendelijke groeten, Bien =E0 vous, Kind regards, Yves Vindevogel Implements = --Apple-Mail-13-316875061-- --Apple-Mail-11-316875059 Content-Transfer-Encoding: base64 Content-Type: image/tiff; x-unix-mode=0666; name="Pasted Graphic 2.tiff" Content-Disposition: inline; filename="Pasted Graphic 2.tiff" TU0AKgAAFciAP6BP5/wWDQeEQmFQuGQ2HQ+FP5+v9rOh5P9IMBuP8zK9tP8fJNlP8QoNjP8FndhP 8MnyVjJHSMrqFnv9BL1vP9ett2v98PuJxChUOiUWGQOCUalUujOV4PZ/qBluR/mdVtJ/jVHst/ho 9MF/g08WABHJfv8AWa0WoAHFfWuzgg6sB/hY9ysUohjv8qqVqv9LMZxv9ouV40zEYmHUjFY2HPx+ RNtOl4Rhftt/phiRs6LC/iA5LewnJeWg46W22+02cAHOwa26bC0HPY7TZ7G2G9dv8EHFdP8Om9YP 8rKJov9UtFzv9Dzh/sRvu5/vZ9PzHdeF4zsUqgQRlOR5v9DMBwv8uqe/jZIVwNHu6B0/WAKHvWGr RAA1ra0G1c/vfgANrdraXrTQIAA6JW2TUrQNz/ja0oAjO4YAjSWS0DOWawjys4RkIk4Lj2sAWEUZ B/ieUJqH+PZdnAf5aG06R5nsfLtqW7UaoWZJwukwLyi4VRsn+F5HGYf8QJWBQ7JWAI5tYti2Dm1T bQHAq0De1Q2FwtA1Fof4AjVDIBDQV0vDQWK0DY/oADYWsrNQN81NOuDZtYOM6ycs4Cjq+Q9JWE5D pGJpOpqP5dMwW5sHUf53xlHCExupjqoIbh3Hu55vp6ch3Hof4wlMrAdEWYh/gmP1RtXK0DDi2rXy nVzXjpVs8NM1g1vuNU2gANEugGMBRwu4YADTNsvwyAA3v+0k6LRWLb2dBUpjk2I4NVOx/gYPC6BU QiwCgSsSlWZpxH+ahzsOaBzvCe7qoMpKlUghd1n2f5oHId5/lQaJ0n+RBf3GLpUGwf4iksrg0FVF ICzlXT9QBB0tWO/+FjlKVWtfijTNRBsvDbhrdP3NUAwY38KFfkc0PuNjRWLLz8ytiTUVdVlnjlA2 MWPAWRAAM8LAAMjhhuR5ir4ULjigT5pn+P5co2T6pn+YBup6dp6xoh6kHie95modNOFkah0H+Qhd pyKBNmcf68r2Bo9GGtdpwIAUotM1Q4QMNzUDZhuVZbLsvy7KkqZZCnAb4AFcQY1C3LW1Wb5lOa2t ZvUt78NNjcMN83wEOMDczvuTzXNQ3QEN01WRL2b8XBYEDuugQEFUYgkpIo5FjIJTmhfZkHG8J0nm fB/sifp/EQYLBioUKsBkRKVgWPLX7stA4VnaDa+tZlZel0j+gCMpWS8MRULQNeIDY/4zTIAIyFX0 GFybZnr+rZ8p2a1s6tUNfADXXI1Yb8jLX+MNQk+1Az7zWqwVa/FaJrHpvSLeAQOhdE/l7CSJwrAb BXDXH+PkyA/gUiPGaftASWW6JeDgnGAsCmLNvcYqlz5+FchnZMrqGbeE3JoYhAMAKbEvOGeigtaE CDZwJcgxtiML0uMgQYzllKt2GhtP/EaH6cgAm+fnCtVEVXFOKNU54AC1B/tyLfBx4YJBGwhAAG5m yUoGH/TgxlZ7nGXw3YdEpYTPWOvaSq/JVEX28rEDQsGKDoGPx+QYnE1URg2paZY4dy7EHTx9cc41 upqIfwkjSgJWyW0MgBdK9I1DGAAwnS8G8+4cGYyUWWbKMg/gSiOjQG+FMQ44qqlmgBNUSYYMohKW 2ULFZaR8NdMFucoz/w/ZcAAMzgH+j/AKGV8QAZkhmWNM0AJ/I5zCiJEGYjF1qoGmxI5MqwT7JVlG buLULJhPXlbK+WMt3HwHmIbF8xaA0n6mvPgN7EJhx9Vcax+hsX6wGZuldNDnTSgQDoloGohS6BYE 62gPQqyRhlFE2gIYkCRgdD4WcAUsnxvlYklKgUCn4upi29JkkpG/pbYgtKbr86TLOnbLCF066ZLP kGsI+7NY9rRjZHB+0ew4M5S0AKT4FA8l0BaIUk4XHkj/EYLwzAtVyj/GwRYf49B8D6H+OMdw9R/j DG8vcTgx1xhpFTBoHYilRgaDygQAobXCynVVSSVdRYTLJls5CcIaqX0zpw/Kms76c2DSnTsNNPaA TbpggqOSCg4MkQwcAPiBAzimJqI4Xg3R/itX0P8bY66xD7MiYkgg5h4vAF2NgdY/xNGCH+HQVMIQ WiCNVM2L6Ap4s3ShFixMubA2HnjcUulhab3GuIbGxVjLlxXZoawMKFgsCiaSOUd6nB6XbH+PofKN LtqcH7eMf5AyimQOtegf4/TIEFvYP8dyjTm2eV1KdmNgrlIKr/cOxF+LjmRldTZY88L/U6P/YtyF +b82PDAhkOQsyNkHu8jS76NB4YXH+PvDUG7v3lIFeu8mFbuj6q8PXE2HEaXqIEUEgxnh2T2YhD/B WBZcUunVjO59yMB3PsHc3BONLemxwaP8NwsTMEPvVeMieJKvXmHZk8f48h5GHyYP8e+V3g3tvMQs VY01FLCxjgTHk8b943yBTOVoJEiXJzPgae1zr+2HwYhnIuRzrjczwP8eY8zDmKy7l8NOYcx40zKW nNswizytg+kXHc/qS48x9o7QbNMhZ0yMdvPBG7wmNz/jB6WYs436uFmbHhbFrAFgjhl4Y/g5IvH+ 0EkYFw+NuAHqmoet9bWJwPnDHEfM55E0udjTNW7uZ+y9p6L+oNe3Bxtoa4lACzgQee2kQ5Jwo1RK QPQfK8xejaJ6FG6w/wEh1NZJ+L6ydoH+zfqTUK0dK7AzsY7Yem9jaA0Fu2wVf9eJObqW+TxuwbCS K4JEYVnh2O/vWQM5K+wtCehCCAP5YABybV035xCAJIJqnrTyFqB9H8e1Lu/OumM870MTp3MGn9J8 g5Zsxhk2ZURLfGmrBAEg5m/CISIf4jRekblaCgRoyS0BjWNZVw6uZ70hnsrkND/Fbn3kWWtmz1ID a5NxyLYJ1xujdI3iZTm9dkYy4/P9yFKIcFoDQsbtKW4nT2YbLrpr450V6zTGem7N+YzXVzDZ0PSl ddIb8GQVrlD9sQjy4FKHVy0ZDDeLHCPWuuD/68Y0Vg1LX8c0bQSBiAq9JrVv2pYwZkzq6mqxDi8y XJgBkjZGftMMdUgm5PKyCBnogBI72hYgZlgrD6WWgMzPZkxGdUzfIfIzsdb66PXr/Jxp+XwRGlOL GnuJm2b0l7vo/dJeDL4N7qwWN+qn5CqedNMATuzZmObaad1s4dBMl/6wmG/Pt1+oAAYk2hToiuQc 5FxyjyRoHqH0xYKE+Q8k+UIgH0KAH+HSHqXmGwHSIuEGF0I25eWE8A96/g2akwhIk8Puf25mpi3w /IeG/M8y0O10/WgM3cpUS8l0mS+ocOYgDQS0ASDUTOB6EaJWDkXyXwFyL+GqGkPKHQHMOkxIXmIO HAHAJyyuKgvGIIHgU2H+G0GuHKH+FwF6KwEAtAH+CUEcLOAiNC2QnKcOZSZCmO9ongNep2sA3Y2W nY/KwE9hDazc8w8Sug9mh6cAZ0/UAEC0FKH+AOCOEoJQBeD8H+A2A+DYYGBuD2H+D6DmFCH+F0Fq hDB+PKGcGaaSGsGmJyGOGGg0EuEgP0CqCOEOH+BIA+DQH+AgBSDrD+CAEcH+AICwFNA+pamupfDQ /G1E2asfDksJDesNBCv012x+5WDmQSseeiAaCGEiH+BEAmC0H+A8AcCyH+A4AXGqAqAOCsLqAQCw H+BMAuDIH+CJEWPMCpFKCaCEEASEBFESA2AaC2H+AoAOCqK6AVGqA8AYCvGcAiC9HmBlHZBglBDt F80LF7BNDdBHDg2U0JGI0lGEseWsAgB6EYH+BUAkC4H+BKAuDDGcArH+BCAnH+A8AeC6K7HjG7H4 AoANHtG1G4A1GxGkAfHkBAAjJOBCAqC/FMAuDAH+BOApJ8A0BYDy2aWtGEzI1G2dKRIUwDGDF9BP DooBISnVImB9IsBRJHI2AuDHK3HHI4DFK3K6BGArI8BAAlJ8A8AhJqAlJxJBK3LDLBK9LmBNKCH+ AwBZEZIGcjGM3zKVIRL7BEH8zVKfKojrKlMCQVIk2jKuH/KzH/I5K7MjLFMpMnI5HGBIAvG/I5J9 MvMrK5M/K7LrJ9LxL1DXL5KYp1L/MM80g2wABMzWAGjVDZIc/XDrIiNjKtKxK1MtNBN7MlN9HFJ5 M3J7NDONMnNHLvLzKNKnMS5dNPMAsGpgLKLoAQNswAH8EAGAXGB0EobQAwEEJOblOa1uzJIfNvF9 MXFVMbMfOPODOBOBMxM1K3M7PfPdNFLtNLOZOcfjIOgVPIfeAaD2VGBUEY6ECyFQL+awXWH+HCHa KgW+KoAqDsPu3Mt8rxMO+fOjPTNzMZN3MhPtN/M/PlOJPrPhPvHBPzOXL3KPKhP69Mv4SmYWWs4m QyDGFUGsH+GeHEOkHW204SIED6FmaSBqeZHmQ3Fikw7iPwZWkCSs40Yg0CwTQAxpPUAhPZN5RDS1 PjOHPpRRRFOTP1RZSo5W1skw6MDa6g5omYZW94AYDkN+BOD+NKCoE6SKHy1WA8EOL2Z8cuhGTU/f SlArAu95L2cmiO1xSrQ7PXQ/S/S1RJS9RFUlNBTDRXNPRavw3St25kTXBiQsADBfUFNO/fUC6K+A N+9fTI2SkOfG6YmY9MZWDEFS+0TInFUQSfPI1NQ9MdSzRPUnS5MzRLUdRPUqD02RRaeoseDdSaQn ScQi9G/kPu+e/o+E8UQUoGv/IXMKsOfqZu8O/eV2S89E7QZ7XG8w/edGQYxiTeoSB3FKBQAlRBV9 S3LnWDUjXpN7LCBOAnJ8AyBXKKV0Z7KPKO6i764rXI9wS8DJVq6TAw+hP5MFBJDi0Ofq0i86mQPv XMDRWlBi6QVy6iAYCUEuJIBDESBGA0DNI2AxLjRCAxS4BGAvH5I5J2BLZdZVV9ZtZqDKJIA+DUH+ AiB4ES2a4pDFDE+fVIkg84ZylyTUn7RdYjIZYhQy14fipY9SDTWhSjRg6UAEC+fZGWEfEMBCDbFM AtJ9MzI9LlM8BFZjJ5ZpNBMzLCBJLKH+A+A3ZSAiByEMmcC0FO7Q7eTVSW+ectBY9HQ2oFDTNXak Ns9fIa0hPPVUfqfq5i3OjohIAEC6fYAWB6ERFUBSDxLuAtZSA7JSA/JoH/LPH+BIAtK7bZG4BGAt JCApH+A+AhJPGnJOA0AfH+AiBIDoWuBzc6AICvD8PxUAWJWWj2sfWwsPP9cXWzKc/PGE0jPQvwiE 9kWsAeB+EVI2AaCpdQAbGjGnHkA4AbJPHhI0A2AZI0A6AdI1drG+A6AfG/HhGjfVI1fLfJGpfBGq BGAZHsAvKJA+j7eZencVNSlZGBelINchalStIqbTV7J5I8BDLtdrJPfLfSAZHkA2AXfPfXENJTfl JrK0BFdjMoBOArX7UsYhUxaledgRcZgVBLMTepVU17StSxXlMrZzbgAtLDgrdoAlGiBBK1dhZZLi AxK/OFORRVNNhbhu0PhhafgTW1gXilgbgRhzUbV/OOAwDPFMAxe/ZrPriXXnWJidP3hiuZgPiphk eGA8EUJGWEc3MNhtgdUXSvi5XxS1Z3Zhe/MzbfjNRRUrieVpeejtOg0OseDeNYQexAIEE4GcX2B4 EkRKAMDsNjXSTWcVVVjvi1jzh1WHM/j8AvkAAvkFlHkLjVjdjZF4pPk0nANKLKQIBCESL2D6F6XG XgvWqwHQMqEsGIRaL6RSAgD6bcQAbyNRBkLQDCWNhdNxkPj1V5h3i7N7lLlPlTmtUpjTTHak1tUE DHDyQMDWN+ggQIByEiJOEGOcGIG8Okq7CMXay2KMG8HaU4rcMGCIE8aSBcEWL2Ci4ccggbKXhxlD j3jPPdmxbdLnm3PxNJhZkO1CseseDUNKBIEMJWB8EqhCBqEuGgH+DiFiL+GXQcINAEIhl4McIkXc IICurWWuDulCNiDW85VVi3mplHmuH/j/oZM9odRTohkNNQnjYIc6QIAILmH+BgEOLOtabCUcIPpV qiIOHMKeOQGoJ6C+FaMwAuEJT474DYTfliJQrbgjmrj5OBoXkDobrTJ/LsArX+LQDI72nBlkJQLv C2FBpAEoGOMGGhqsvKH7nlqpqlnpsKKMHgJ+tgGQPKCmFIaSAwD8LOA4EGJWBiCdEFH1G/iPp0Av rXlRrboTdXLDdrI0BSBkD4H/n8JOAkD7o0EwhCECF+PKG6HceBsQRtsPtyOuq4/+22RcFYJGBkBK Ddp4AtI8A2AdHkA/H9uPZZrVp5lNp9OFMzK7iLJ9HhHlgrJ8BSA3FSEWECe+vNsVnkXdt4MVqnvR qiHmHkKgF+F0JqEMD8fECQB4D7uPI8AyAZGjubJ8BAAqClbqAng4AbfZdsKyBTuMDiDOE2H+FmFe JGHMU1vXwrvVwrwxCKagF3pADyDiFEH+CWCAEHJ+A3G4B+BmDmH+DQC7EEFGE4NKHYHUXvwxxqIj t3xtxyIaw0ImGwGus8HgUZx1yGavxxyJyPyRySKIKQFuFibQGOGEYFyVynypymKQCaB/xIDkDOE8 IQFSFEVGE+EsQIIWGEF4L+EwEeP6IOE4EoN2DOC0ErpCDLy6DYDAEyH+FiFU6EHyq6H+DqDUFAH+ G4GwOWIWw0IIEMD68HzsE0H+DgDIE4H+DaDD0cG8G4X2FgFS6EFGE2LAIWG0GsHMH+EoEWTaIOFn uEH+DKCzzkDjzoH+DJ1aH+FuFgbRsGIIEKD2fZyCrEIOGUGMI2E8EqNLz6XmDsDXEgDUC6EwH+DW C9zx1ZzkF4FuOOHUq0D8Dnb8DeDHwd0h0kEeEKQsKQCkCLFKDv2SIQD6DkFJENg2H/ygYFxkPCBu BQDkH+CwCXIsIOCiCJb2CQB0D+H+FEE0LoEYEGTOCgCGEKH/2uMrGfI8GEF7RyIWHwOoH+CEBptU EaEITOFCEyLOF0FoJr16aUDpb9fbH+GAF0RSz2KgCABjdACSB4ED3V3YH+C8CjbCFAEwLP006EG8 G2X35LbnK6B/5iH+GcGSs8FoFehCC0CYEavXAT5WRSDyDdxACKBtvwFwFkJqHEHAUVxeLACSB3HZ 2GNKEyEgP6EsEaNEKQCn3MH+DuDZEgIOEUEAZMDuDbxB7jFL5zbCDQC2Et2cC/zwIP8AH+BoBKDf 1iCxEEC4CfFgEqEYNFCeU4BWA2DSXIGgXGIWu+XmCQB35r2R0EEgEOQyIWEOD88GECDvVmCyCZIs DACpGaE0EkN+DP8GIQEGDyFUH/3LFL0Ue+HkHh18IMHPwoCqCRe54ILoCMBvEKC6Cl8CC18IIWFa FKL2DGCtEEIXE0MH2+H+EKD4fZ10fZ374Z7h7l7p7sIMEF9eRd6d1p1t4H4L/ohCDECsEkIQIAVC MiX+dTUoH++n0/X++4W/4hEHe7Xo/xQGTO/zQXEu/0Cd1S/0Yglg/3g73q/yWPUE/zsa1C/0IeVW /0GeFU/3G4Xa/0MfVY/1or2bQlbRUiiVpJpQ/y+U0jEYggzzOSkRUM/0KfJqfzop3+p1Aw3+53I7 n+UCJWak3m26H+OxYdX+Xiij6leVRY3+YKheaks1Yy3+VCPBEIe5zM5qOxbdICAADgEAAAMAAAAB ADgAAAEBAAMAAAABAEAAAAECAAMAAAAEAAAWdgEDAAMAAAABAAUAAAEGAAMAAAABAAIAAAERAAQA AAABAAAACAEVAAMAAAABAAQAAAEWAAQAAAABAAAAkgEXAAQAAAABAAAVwAEaAAUAAAABAAAWfgEb AAUAAAABAAAWhgEcAAMAAAABAAEAAAEoAAMAAAABAAIAAAFSAAMAAAABAAEAAAAAAAAACAAIAAgA CAAK/IAAACcQAAr8gAAAJxA= --Apple-Mail-11-316875059 Content-Type: multipart/alternative; boundary=Apple-Mail-14-316875063 --Apple-Mail-14-316875063 Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=US-ASCII; format=flowed Mail: yves.vindevogel@implements.be - Mobile: +32 (478) 80 82 91 Kempische Steenweg 206 - 3500 Hasselt - Tel-Fax: +32 (11) 43 55 76 Web: http://www.implements.be First they ignore you. Then they laugh at you. Then they fight you. Then you win. Mahatma Ghandi. --Apple-Mail-14-316875063 Content-Transfer-Encoding: 7bit Content-Type: text/enriched; charset=US-ASCII Mail: yves.vindevogel@implements.be - Mobile: +32 (478) 80 82 91 Kempische Steenweg 206 - 3500 Hasselt - Tel-Fax: +32 (11) 43 55 76 Web: http://www.implements.be First they ignore you. Then they laugh at you. Then they fight you. Then you win. Mahatma Ghandi. --Apple-Mail-14-316875063-- --Apple-Mail-11-316875059-- From pgsql-performance-owner@postgresql.org Sun Aug 21 17:00:13 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 21B891FEEC4B for ; Sun, 21 Aug 2005 16:59:51 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 20569-01 for ; Sun, 21 Aug 2005 19:59:47 +0000 (GMT) Received: from asia.telenet-ops.be (asia.telenet-ops.be [195.130.132.59]) by svr1.postgresql.org (Postfix) with ESMTP id 442951FEEC43 for ; Sun, 21 Aug 2005 16:59:16 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by asia.telenet-ops.be (Postfix) with SMTP id CD7FB224037 for ; Sun, 21 Aug 2005 21:59:15 +0200 (MEST) Received: from [10.0.1.2] (d5152B1F3.access.telenet.be [81.82.177.243]) by asia.telenet-ops.be (Postfix) with ESMTP id DDEE8224099 for ; Sun, 21 Aug 2005 21:59:14 +0200 (MEST) Mime-Version: 1.0 (Apple Message framework v622) To: pgsql-performance@postgresql.org Message-Id: Content-Type: multipart/mixed; boundary=Apple-Mail-18-318263928 From: Yves Vindevogel Subject: Fwd: (Re)-indexing on updates Date: Sun, 21 Aug 2005 21:59:11 +0200 X-Mailer: Apple Mail (2.622) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/311 X-Sequence-Number: 14058 --Apple-Mail-18-318263928 Content-Type: multipart/alternative; boundary=Apple-Mail-19-318263928 --Apple-Mail-19-318263928 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=ISO-8859-1; format=flowed My DB is quite simple. It holds data about printjobs that come from=20 the windows eventlog. The data is shown on a website. I have one main table: tblPrintjobs. We add some extra data to it. Like the applicationtype, based on rules=20= we define in other tables. When a rule changes, the updates take place (and take so long). Also, when new records are added, this takes place. For instance, rule 1 and rule 2 are changing positions in importance.=20 (1 was before 2, now 2 before 1) The records that hold reference to rule 1 are reset to null (one field) Rule 2 is assigned, then rule 1 is assigned. What I could do is also: delete all from tblRefRules where rule is 1 insert all from tblPrintjobs that are not yet in RefRules for Rule2,=20 then insert all for rule2 That would be a workaround for the MVCC. Not ? BTW: The good rule is: drop index, update, vacuum, create index ? I think I mistook the purpose of vacuum. If I index before the vacuum, my marked records will still be in the=20 index ? Even if all transactions are finished ? Begin forwarded message: > From: "Jeffrey W. Baker" > Date: Sun 21 Aug 2005 21:36:16 CEST > To: Yves Vindevogel > Subject: Re: [PERFORM] (Re)-indexing on updates > > On Sun, 2005-08-21 at 21:18 +0200, Yves Vindevogel wrote: >> >> >> = ______________________________________________________________________ >> >> Ok, this is a major setback in some of my procedures. >> =46rom time to time, I must update one field in about 10% of the >> records. >> So this will take time. >> >> How can I work around that ? >> >> Some personal opinions ... >> 1) Drop indexes, run update, create indexes, vacuum > > Drop index, update, vacuum, create index > > -or- > > update, vacuum, reindex > >> 2) Move the field to another table and use joins ? I could delete = the >> records when needed and add them again > > I'm not familiar with your application, but you could try it and tell=20= > us > if this works :) > >> >> This mechanism, of inserting a new record and marking the old one, is >> that data kept somewhere where I can "see" it ? > > This is MVCC: multi-version cuncurrency. The old record is kept=20 > because > there could be an old transaction that can still see it, and cannot = yet > see the updated record. And no other transaction can see your record > until you commit. The old row isn't removed until you vacuum. > >> I need for one app a trace of all my changes in the database. I have >> a set of triggers to do that for the moment on each table. >> Could I use that mechanism somehow to avoid my triggers ? >> Any documentation on that mechanism (hacker stuff like what tables = are >> used) ? > > You could search the postgresql documentation (or the web) for MVCC. > > Regards, > jwb > > > Met vriendelijke groeten, Bien =E0 vous, Kind regards, Yves Vindevogel Implements --Apple-Mail-19-318263928 Content-Transfer-Encoding: quoted-printable Content-Type: text/enriched; charset=ISO-8859-1 My DB is quite simple. It holds data about printjobs that come from the windows eventlog. The data is shown on a website. I have one main table: tblPrintjobs. We add some extra data to it. Like the applicationtype, based on rules we define in other tables. When a rule changes, the updates take place (and take so long). Also, when new records are added, this takes place. For instance, rule 1 and rule 2 are changing positions in importance. (1 was before 2, now 2 before 1) The records that hold reference to rule 1 are reset to null (one field) Rule 2 is assigned, then rule 1 is assigned. What I could do is also: delete all from tblRefRules where rule is 1 insert all from tblPrintjobs that are not yet in RefRules for Rule2, then insert all for rule2 That would be a workaround for the MVCC. Not ? BTW: The good rule is: drop index, update, vacuum, create index ? I think I mistook the purpose of vacuum. If I index before the vacuum, my marked records will still be in the index ? Even if all transactions are finished ? Begin forwarded message: 0000,0000,0000From: "Jeffrey W. Baker" < 0000,0000,0000Date: Sun 21 Aug 2005 21:36:16 CEST 0000,0000,0000To: Yves Vindevogel < 0000,0000,0000Subject: Re: [PERFORM] (Re)-indexing on updates On Sun, 2005-08-21 at 21:18 +0200, Yves Vindevogel wrote: ______________________________________________________________________ Ok, this is a major setback in some of my procedures. =46rom time to time, I must update one field in about 10% of the records. So this will take time. How can I work around that ? Some personal opinions ... 1) Drop indexes, run update, create indexes, vacuum Drop index, update, vacuum, create index -or- update, vacuum, reindex 2) Move the field to another table and use joins ? I could delete the records when needed and add them again=20 I'm not familiar with your application, but you could try it and tell us if this works :) This mechanism, of inserting a new record and marking the old one, is that data kept somewhere where I can "see" it ? This is MVCC: multi-version cuncurrency. The old record is kept because there could be an old transaction that can still see it, and cannot yet see the updated record. And no other transaction can see your record until you commit. The old row isn't removed until you vacuum. I need for one app a trace of all my changes in the database.=20= I have a set of triggers to do that for the moment on each table. Could I use that mechanism somehow to avoid my triggers ? Any documentation on that mechanism (hacker stuff like what tables are used) ? You could search the postgresql documentation (or the web) for MVCC. Regards, jwb Met vriendelijke groeten, Bien =E0 vous, Kind regards, Yves Vindevogel Implements = --Apple-Mail-19-318263928-- --Apple-Mail-18-318263928 Content-Transfer-Encoding: base64 Content-Type: image/tiff; x-unix-mode=0666; name="Pasted Graphic 2.tiff" Content-Disposition: inline; filename="Pasted Graphic 2.tiff" TU0AKgAAFciAP6BP5/wWDQeEQmFQuGQ2HQ+FP5+v9rOh5P9IMBuP8zK9tP8fJNlP8QoNjP8FndhP 8MnyVjJHSMrqFnv9BL1vP9ett2v98PuJxChUOiUWGQOCUalUujOV4PZ/qBluR/mdVtJ/jVHst/ho 9MF/g08WABHJfv8AWa0WoAHFfWuzgg6sB/hY9ysUohjv8qqVqv9LMZxv9ouV40zEYmHUjFY2HPx+ RNtOl4Rhftt/phiRs6LC/iA5LewnJeWg46W22+02cAHOwa26bC0HPY7TZ7G2G9dv8EHFdP8Om9YP 8rKJov9UtFzv9Dzh/sRvu5/vZ9PzHdeF4zsUqgQRlOR5v9DMBwv8uqe/jZIVwNHu6B0/WAKHvWGr RAA1ra0G1c/vfgANrdraXrTQIAA6JW2TUrQNz/ja0oAjO4YAjSWS0DOWawjys4RkIk4Lj2sAWEUZ B/ieUJqH+PZdnAf5aG06R5nsfLtqW7UaoWZJwukwLyi4VRsn+F5HGYf8QJWBQ7JWAI5tYti2Dm1T bQHAq0De1Q2FwtA1Fof4AjVDIBDQV0vDQWK0DY/oADYWsrNQN81NOuDZtYOM6ycs4Cjq+Q9JWE5D pGJpOpqP5dMwW5sHUf53xlHCExupjqoIbh3Hu55vp6ch3Hof4wlMrAdEWYh/gmP1RtXK0DDi2rXy nVzXjpVs8NM1g1vuNU2gANEugGMBRwu4YADTNsvwyAA3v+0k6LRWLb2dBUpjk2I4NVOx/gYPC6BU QiwCgSsSlWZpxH+ahzsOaBzvCe7qoMpKlUghd1n2f5oHId5/lQaJ0n+RBf3GLpUGwf4iksrg0FVF ICzlXT9QBB0tWO/+FjlKVWtfijTNRBsvDbhrdP3NUAwY38KFfkc0PuNjRWLLz8ytiTUVdVlnjlA2 MWPAWRAAM8LAAMjhhuR5ir4ULjigT5pn+P5co2T6pn+YBup6dp6xoh6kHie95modNOFkah0H+Qhd pyKBNmcf68r2Bo9GGtdpwIAUotM1Q4QMNzUDZhuVZbLsvy7KkqZZCnAb4AFcQY1C3LW1Wb5lOa2t ZvUt78NNjcMN83wEOMDczvuTzXNQ3QEN01WRL2b8XBYEDuugQEFUYgkpIo5FjIJTmhfZkHG8J0nm fB/sifp/EQYLBioUKsBkRKVgWPLX7stA4VnaDa+tZlZel0j+gCMpWS8MRULQNeIDY/4zTIAIyFX0 GFybZnr+rZ8p2a1s6tUNfADXXI1Yb8jLX+MNQk+1Az7zWqwVa/FaJrHpvSLeAQOhdE/l7CSJwrAb BXDXH+PkyA/gUiPGaftASWW6JeDgnGAsCmLNvcYqlz5+FchnZMrqGbeE3JoYhAMAKbEvOGeigtaE CDZwJcgxtiML0uMgQYzllKt2GhtP/EaH6cgAm+fnCtVEVXFOKNU54AC1B/tyLfBx4YJBGwhAAG5m yUoGH/TgxlZ7nGXw3YdEpYTPWOvaSq/JVEX28rEDQsGKDoGPx+QYnE1URg2paZY4dy7EHTx9cc41 upqIfwkjSgJWyW0MgBdK9I1DGAAwnS8G8+4cGYyUWWbKMg/gSiOjQG+FMQ44qqlmgBNUSYYMohKW 2ULFZaR8NdMFucoz/w/ZcAAMzgH+j/AKGV8QAZkhmWNM0AJ/I5zCiJEGYjF1qoGmxI5MqwT7JVlG buLULJhPXlbK+WMt3HwHmIbF8xaA0n6mvPgN7EJhx9Vcax+hsX6wGZuldNDnTSgQDoloGohS6BYE 62gPQqyRhlFE2gIYkCRgdD4WcAUsnxvlYklKgUCn4upi29JkkpG/pbYgtKbr86TLOnbLCF066ZLP kGsI+7NY9rRjZHB+0ew4M5S0AKT4FA8l0BaIUk4XHkj/EYLwzAtVyj/GwRYf49B8D6H+OMdw9R/j DG8vcTgx1xhpFTBoHYilRgaDygQAobXCynVVSSVdRYTLJls5CcIaqX0zpw/Kms76c2DSnTsNNPaA TbpggqOSCg4MkQwcAPiBAzimJqI4Xg3R/itX0P8bY66xD7MiYkgg5h4vAF2NgdY/xNGCH+HQVMIQ WiCNVM2L6Ap4s3ShFixMubA2HnjcUulhab3GuIbGxVjLlxXZoawMKFgsCiaSOUd6nB6XbH+PofKN LtqcH7eMf5AyimQOtegf4/TIEFvYP8dyjTm2eV1KdmNgrlIKr/cOxF+LjmRldTZY88L/U6P/YtyF +b82PDAhkOQsyNkHu8jS76NB4YXH+PvDUG7v3lIFeu8mFbuj6q8PXE2HEaXqIEUEgxnh2T2YhD/B WBZcUunVjO59yMB3PsHc3BONLemxwaP8NwsTMEPvVeMieJKvXmHZk8f48h5GHyYP8e+V3g3tvMQs VY01FLCxjgTHk8b943yBTOVoJEiXJzPgae1zr+2HwYhnIuRzrjczwP8eY8zDmKy7l8NOYcx40zKW nNswizytg+kXHc/qS48x9o7QbNMhZ0yMdvPBG7wmNz/jB6WYs436uFmbHhbFrAFgjhl4Y/g5IvH+ 0EkYFw+NuAHqmoet9bWJwPnDHEfM55E0udjTNW7uZ+y9p6L+oNe3Bxtoa4lACzgQee2kQ5Jwo1RK QPQfK8xejaJ6FG6w/wEh1NZJ+L6ydoH+zfqTUK0dK7AzsY7Yem9jaA0Fu2wVf9eJObqW+TxuwbCS K4JEYVnh2O/vWQM5K+wtCehCCAP5YABybV035xCAJIJqnrTyFqB9H8e1Lu/OumM870MTp3MGn9J8 g5Zsxhk2ZURLfGmrBAEg5m/CISIf4jRekblaCgRoyS0BjWNZVw6uZ70hnsrkND/Fbn3kWWtmz1ID a5NxyLYJ1xujdI3iZTm9dkYy4/P9yFKIcFoDQsbtKW4nT2YbLrpr450V6zTGem7N+YzXVzDZ0PSl ddIb8GQVrlD9sQjy4FKHVy0ZDDeLHCPWuuD/68Y0Vg1LX8c0bQSBiAq9JrVv2pYwZkzq6mqxDi8y XJgBkjZGftMMdUgm5PKyCBnogBI72hYgZlgrD6WWgMzPZkxGdUzfIfIzsdb66PXr/Jxp+XwRGlOL GnuJm2b0l7vo/dJeDL4N7qwWN+qn5CqedNMATuzZmObaad1s4dBMl/6wmG/Pt1+oAAYk2hToiuQc 5FxyjyRoHqH0xYKE+Q8k+UIgH0KAH+HSHqXmGwHSIuEGF0I25eWE8A96/g2akwhIk8Puf25mpi3w /IeG/M8y0O10/WgM3cpUS8l0mS+ocOYgDQS0ASDUTOB6EaJWDkXyXwFyL+GqGkPKHQHMOkxIXmIO HAHAJyyuKgvGIIHgU2H+G0GuHKH+FwF6KwEAtAH+CUEcLOAiNC2QnKcOZSZCmO9ongNep2sA3Y2W nY/KwE9hDazc8w8Sug9mh6cAZ0/UAEC0FKH+AOCOEoJQBeD8H+A2A+DYYGBuD2H+D6DmFCH+F0Fq hDB+PKGcGaaSGsGmJyGOGGg0EuEgP0CqCOEOH+BIA+DQH+AgBSDrD+CAEcH+AICwFNA+pamupfDQ /G1E2asfDksJDesNBCv012x+5WDmQSseeiAaCGEiH+BEAmC0H+A8AcCyH+A4AXGqAqAOCsLqAQCw H+BMAuDIH+CJEWPMCpFKCaCEEASEBFESA2AaC2H+AoAOCqK6AVGqA8AYCvGcAiC9HmBlHZBglBDt F80LF7BNDdBHDg2U0JGI0lGEseWsAgB6EYH+BUAkC4H+BKAuDDGcArH+BCAnH+A8AeC6K7HjG7H4 AoANHtG1G4A1GxGkAfHkBAAjJOBCAqC/FMAuDAH+BOApJ8A0BYDy2aWtGEzI1G2dKRIUwDGDF9BP DooBISnVImB9IsBRJHI2AuDHK3HHI4DFK3K6BGArI8BAAlJ8A8AhJqAlJxJBK3LDLBK9LmBNKCH+ AwBZEZIGcjGM3zKVIRL7BEH8zVKfKojrKlMCQVIk2jKuH/KzH/I5K7MjLFMpMnI5HGBIAvG/I5J9 MvMrK5M/K7LrJ9LxL1DXL5KYp1L/MM80g2wABMzWAGjVDZIc/XDrIiNjKtKxK1MtNBN7MlN9HFJ5 M3J7NDONMnNHLvLzKNKnMS5dNPMAsGpgLKLoAQNswAH8EAGAXGB0EobQAwEEJOblOa1uzJIfNvF9 MXFVMbMfOPODOBOBMxM1K3M7PfPdNFLtNLOZOcfjIOgVPIfeAaD2VGBUEY6ECyFQL+awXWH+HCHa KgW+KoAqDsPu3Mt8rxMO+fOjPTNzMZN3MhPtN/M/PlOJPrPhPvHBPzOXL3KPKhP69Mv4SmYWWs4m QyDGFUGsH+GeHEOkHW204SIED6FmaSBqeZHmQ3Fikw7iPwZWkCSs40Yg0CwTQAxpPUAhPZN5RDS1 PjOHPpRRRFOTP1RZSo5W1skw6MDa6g5omYZW94AYDkN+BOD+NKCoE6SKHy1WA8EOL2Z8cuhGTU/f SlArAu95L2cmiO1xSrQ7PXQ/S/S1RJS9RFUlNBTDRXNPRavw3St25kTXBiQsADBfUFNO/fUC6K+A N+9fTI2SkOfG6YmY9MZWDEFS+0TInFUQSfPI1NQ9MdSzRPUnS5MzRLUdRPUqD02RRaeoseDdSaQn ScQi9G/kPu+e/o+E8UQUoGv/IXMKsOfqZu8O/eV2S89E7QZ7XG8w/edGQYxiTeoSB3FKBQAlRBV9 S3LnWDUjXpN7LCBOAnJ8AyBXKKV0Z7KPKO6i764rXI9wS8DJVq6TAw+hP5MFBJDi0Ofq0i86mQPv XMDRWlBi6QVy6iAYCUEuJIBDESBGA0DNI2AxLjRCAxS4BGAvH5I5J2BLZdZVV9ZtZqDKJIA+DUH+ AiB4ES2a4pDFDE+fVIkg84ZylyTUn7RdYjIZYhQy14fipY9SDTWhSjRg6UAEC+fZGWEfEMBCDbFM AtJ9MzI9LlM8BFZjJ5ZpNBMzLCBJLKH+A+A3ZSAiByEMmcC0FO7Q7eTVSW+ectBY9HQ2oFDTNXak Ns9fIa0hPPVUfqfq5i3OjohIAEC6fYAWB6ERFUBSDxLuAtZSA7JSA/JoH/LPH+BIAtK7bZG4BGAt JCApH+A+AhJPGnJOA0AfH+AiBIDoWuBzc6AICvD8PxUAWJWWj2sfWwsPP9cXWzKc/PGE0jPQvwiE 9kWsAeB+EVI2AaCpdQAbGjGnHkA4AbJPHhI0A2AZI0A6AdI1drG+A6AfG/HhGjfVI1fLfJGpfBGq BGAZHsAvKJA+j7eZencVNSlZGBelINchalStIqbTV7J5I8BDLtdrJPfLfSAZHkA2AXfPfXENJTfl JrK0BFdjMoBOArX7UsYhUxaledgRcZgVBLMTepVU17StSxXlMrZzbgAtLDgrdoAlGiBBK1dhZZLi AxK/OFORRVNNhbhu0PhhafgTW1gXilgbgRhzUbV/OOAwDPFMAxe/ZrPriXXnWJidP3hiuZgPiphk eGA8EUJGWEc3MNhtgdUXSvi5XxS1Z3Zhe/MzbfjNRRUrieVpeejtOg0OseDeNYQexAIEE4GcX2B4 EkRKAMDsNjXSTWcVVVjvi1jzh1WHM/j8AvkAAvkFlHkLjVjdjZF4pPk0nANKLKQIBCESL2D6F6XG XgvWqwHQMqEsGIRaL6RSAgD6bcQAbyNRBkLQDCWNhdNxkPj1V5h3i7N7lLlPlTmtUpjTTHak1tUE DHDyQMDWN+ggQIByEiJOEGOcGIG8Okq7CMXay2KMG8HaU4rcMGCIE8aSBcEWL2Ci4ccggbKXhxlD j3jPPdmxbdLnm3PxNJhZkO1CseseDUNKBIEMJWB8EqhCBqEuGgH+DiFiL+GXQcINAEIhl4McIkXc IICurWWuDulCNiDW85VVi3mplHmuH/j/oZM9odRTohkNNQnjYIc6QIAILmH+BgEOLOtabCUcIPpV qiIOHMKeOQGoJ6C+FaMwAuEJT474DYTfliJQrbgjmrj5OBoXkDobrTJ/LsArX+LQDI72nBlkJQLv C2FBpAEoGOMGGhqsvKH7nlqpqlnpsKKMHgJ+tgGQPKCmFIaSAwD8LOA4EGJWBiCdEFH1G/iPp0Av rXlRrboTdXLDdrI0BSBkD4H/n8JOAkD7o0EwhCECF+PKG6HceBsQRtsPtyOuq4/+22RcFYJGBkBK Ddp4AtI8A2AdHkA/H9uPZZrVp5lNp9OFMzK7iLJ9HhHlgrJ8BSA3FSEWECe+vNsVnkXdt4MVqnvR qiHmHkKgF+F0JqEMD8fECQB4D7uPI8AyAZGjubJ8BAAqClbqAng4AbfZdsKyBTuMDiDOE2H+FmFe JGHMU1vXwrvVwrwxCKagF3pADyDiFEH+CWCAEHJ+A3G4B+BmDmH+DQC7EEFGE4NKHYHUXvwxxqIj t3xtxyIaw0ImGwGus8HgUZx1yGavxxyJyPyRySKIKQFuFibQGOGEYFyVynypymKQCaB/xIDkDOE8 IQFSFEVGE+EsQIIWGEF4L+EwEeP6IOE4EoN2DOC0ErpCDLy6DYDAEyH+FiFU6EHyq6H+DqDUFAH+ G4GwOWIWw0IIEMD68HzsE0H+DgDIE4H+DaDD0cG8G4X2FgFS6EFGE2LAIWG0GsHMH+EoEWTaIOFn uEH+DKCzzkDjzoH+DJ1aH+FuFgbRsGIIEKD2fZyCrEIOGUGMI2E8EqNLz6XmDsDXEgDUC6EwH+DW C9zx1ZzkF4FuOOHUq0D8Dnb8DeDHwd0h0kEeEKQsKQCkCLFKDv2SIQD6DkFJENg2H/ygYFxkPCBu BQDkH+CwCXIsIOCiCJb2CQB0D+H+FEE0LoEYEGTOCgCGEKH/2uMrGfI8GEF7RyIWHwOoH+CEBptU EaEITOFCEyLOF0FoJr16aUDpb9fbH+GAF0RSz2KgCABjdACSB4ED3V3YH+C8CjbCFAEwLP006EG8 G2X35LbnK6B/5iH+GcGSs8FoFehCC0CYEavXAT5WRSDyDdxACKBtvwFwFkJqHEHAUVxeLACSB3HZ 2GNKEyEgP6EsEaNEKQCn3MH+DuDZEgIOEUEAZMDuDbxB7jFL5zbCDQC2Et2cC/zwIP8AH+BoBKDf 1iCxEEC4CfFgEqEYNFCeU4BWA2DSXIGgXGIWu+XmCQB35r2R0EEgEOQyIWEOD88GECDvVmCyCZIs DACpGaE0EkN+DP8GIQEGDyFUH/3LFL0Ue+HkHh18IMHPwoCqCRe54ILoCMBvEKC6Cl8CC18IIWFa FKL2DGCtEEIXE0MH2+H+EKD4fZ10fZ374Z7h7l7p7sIMEF9eRd6d1p1t4H4L/ohCDECsEkIQIAVC MiX+dTUoH++n0/X++4W/4hEHe7Xo/xQGTO/zQXEu/0Cd1S/0Yglg/3g73q/yWPUE/zsa1C/0IeVW /0GeFU/3G4Xa/0MfVY/1or2bQlbRUiiVpJpQ/y+U0jEYggzzOSkRUM/0KfJqfzop3+p1Aw3+53I7 n+UCJWak3m26H+OxYdX+Xiij6leVRY3+YKheaks1Yy3+VCPBEIe5zM5qOxbdICAADgEAAAMAAAAB ADgAAAEBAAMAAAABAEAAAAECAAMAAAAEAAAWdgEDAAMAAAABAAUAAAEGAAMAAAABAAIAAAERAAQA AAABAAAACAEVAAMAAAABAAQAAAEWAAQAAAABAAAAkgEXAAQAAAABAAAVwAEaAAUAAAABAAAWfgEb AAUAAAABAAAWhgEcAAMAAAABAAEAAAEoAAMAAAABAAIAAAFSAAMAAAABAAEAAAAAAAAACAAIAAgA CAAK/IAAACcQAAr8gAAAJxA= --Apple-Mail-18-318263928 Content-Type: multipart/alternative; boundary=Apple-Mail-20-318263930 --Apple-Mail-20-318263930 Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=US-ASCII; format=flowed Mail: yves.vindevogel@implements.be - Mobile: +32 (478) 80 82 91 Kempische Steenweg 206 - 3500 Hasselt - Tel-Fax: +32 (11) 43 55 76 Web: http://www.implements.be First they ignore you. Then they laugh at you. Then they fight you. Then you win. Mahatma Ghandi. --Apple-Mail-20-318263930 Content-Transfer-Encoding: 7bit Content-Type: text/enriched; charset=US-ASCII Mail: yves.vindevogel@implements.be - Mobile: +32 (478) 80 82 91 Kempische Steenweg 206 - 3500 Hasselt - Tel-Fax: +32 (11) 43 55 76 Web: http://www.implements.be First they ignore you. Then they laugh at you. Then they fight you. Then you win. Mahatma Ghandi. --Apple-Mail-20-318263930-- --Apple-Mail-18-318263928-- From pgsql-performance-owner@postgresql.org Sun Aug 21 20:37:24 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 59EB4D76CF for ; Sun, 21 Aug 2005 20:35:55 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 92394-01 for ; Sun, 21 Aug 2005 23:35:49 +0000 (GMT) Received: from svr2.postgresql.org (svr2.postgresql.org [65.19.161.25]) by svr1.postgresql.org (Postfix) with ESMTP id A59F2D77B1 for ; Sun, 21 Aug 2005 20:33:33 -0300 (ADT) Received: from smtpauth09.mail.atl.earthlink.net (smtpauth09.mail.atl.earthlink.net [209.86.89.69]) by svr2.postgresql.org (Postfix) with ESMTP id 922BDF10A6 for ; Sun, 21 Aug 2005 21:13:46 +0100 (BST) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=nJ/bH4nQrkilcDUBB5f59PqXYtr/+F0IGnjnhXi9uGu4ef8rMGa4wqUIF7VBhAUC; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:Cc:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth09.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E6wGJ-0007a0-10; Sun, 21 Aug 2005 16:17:47 -0400 Message-Id: <6.2.3.4.0.20050821112849.01f79860@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Sun, 21 Aug 2005 16:13:17 -0400 To: Jeremiah Jahn From: Ron Subject: Re: extremly low memory usage Cc: postgres performance In-Reply-To: <1124636066.27881.238.camel@bluejay.goodinassociates.com> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <1124481386.27881.167.camel@bluejay.goodinassociates.com> <43064908.7010505@arbash-meinel.com> <1124564003.27881.229.camel@bluejay.goodinassociates.com> <6.2.3.4.0.20050820154422.06021150@pop.earthlink.net> <4307E7A4.7090300@arbash-meinel.com> <1124636066.27881.238.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bcd077a119f968650f22b316183a24595f350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.441 required=5 tests=[AWL=0.067, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/314 X-Sequence-Number: 14061 At 10:54 AM 8/21/2005, Jeremiah Jahn wrote: >On Sat, 2005-08-20 at 21:32 -0500, John A Meinel wrote: > > Ron wrote: > > > > Well, since you can get a read of the RAID at 150MB/s, that means that > > it is actual I/O speed. It may not be cached in RAM. Perhaps you could > > try the same test, only using say 1G, which should be cached. > >[root@io pgsql]# time dd if=/dev/zero of=testfile bs=1024 count=1000000 >1000000+0 records in >1000000+0 records out > >real 0m8.885s >user 0m0.299s >sys 0m6.998s This is abysmally slow. >[root@io pgsql]# time dd of=/dev/null if=testfile bs=1024 count=1000000 >1000000+0 records in >1000000+0 records out > >real 0m1.654s >user 0m0.232s >sys 0m1.415s This transfer rate is the only one out of the 4 you have posted that is in the vicinity of where it should be. >The raid array I have is currently set up to use a single channel. But I >have dual controllers in the array. And dual external slots on the card. >The machine is brand new and has pci-e backplane. > So you have 2 controllers each with 2 external slots? But you are currently only using 1 controller and only one external slot on that controller? > > > Assuming these are U320 15Krpm 147GB HDs, a RAID 10 array of 14 of them > > > doing raw sequential IO like this should be capable of at > > > ~7*75MB/s= 525MB/s using Seagate Cheetah 15K.4's >BTW I'm using Seagate Cheetah 15K.4's OK, now we have that nailed down. > > > AFAICT, the Dell PERC4 controllers use various flavors of the LSI Logic > > > MegaRAID controllers. What I don't know is which exact one yours is, > > > nor do I know if it (or any of the MegaRAID controllers) are high > > > powered enough. > >PERC4eDC-PCI Express, 128MB Cache, 2-External Channels Looks like they are using the LSI Logic MegaRAID SCSI 320-2E controller. IIUC, you have 2 of these, each with 2 external channels? The specs on these appear a bit strange. They are listed as being a PCI-Ex8 card, which means they should have a max bandwidth of 20Gb/s= 2GB/s, yet they are also listed as only supporting dual channel U320= 640MB/s when they could easily support quad channel U320= 1.28GB/s. Why bother building a PCI-Ex8 card when only a PCI-Ex4 card (which is a more standard physical format) would've been enough? Or if you are going to build a PCI-Ex8 card, why not support quad channel U320? This smells like there's a problem with LSI's design. The 128MB buffer also looks suspiciously small, and I do not see any upgrade path for it on LSI Logic's site. "Serious" RAID controllers from companies like Xyratex, Engino, and Dot-hill can have up to 1-2GB of buffer, and there's sound technical reasons for it. See if there's a buffer upgrade available or if you can get controllers that have larger buffer capabilities. Regardless of the above, each of these controllers should still be good for about 80-85% of 640MB/s, or ~510-540 MB/s apiece when doing raw sequential IO if you plug 3-4 fast enough HD's into each SCSI channel. Cheetah 15K.4's certainly are fast enough. Optimal setup is probably to split each RAID 1 pair so that one HD is on each of the SCSI channels, and then RAID 0 those pairs. That will also protect you from losing the entire disk subsystem if one of the SCSI channels dies. That 128MB of buffer cache may very well be too small to keep the IO rate up, and/or there may be a more subtle problem with the LSI card, and/or you may have a configuration problem, but _something(s)_ need fixing since you are only getting raw sequential IO of ~100-150MB/s when it should be above 500MB/s. This will make the most difference for initial reads (first time you load a table, first time you make a given query, etc) and for any writes. Your HW provider should be able to help you, even if some of the HW in question needs to be changed. You paid for a solution. As long as this stuff is performing at so much less then what it is supposed to, you have not received the solution you paid for. BTW, on the subject of RAID stripes IME the sweet spot tends to be in the 64KB to 256KB range (very large, very read heavy data mines can want larger RAID stripes.). Only experimentation will tell you what results in the best performance for your application. >I'm not really worried about the writing, it's the reading the reading >that needs to be faster. Initial reads are only going to be as fast as your HD subsystem, so there's a reason for making the HD subsystem faster even if all you care about is reads. In addition, I'll repeat my previous advice that upgrading to 16GB of RAM would be well worth it for you. Hope this helps, Ron Peacetree From pgsql-performance-owner@postgresql.org Sun Aug 21 19:59:21 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C4903D6D84 for ; Sun, 21 Aug 2005 19:59:19 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 74781-09 for ; Sun, 21 Aug 2005 22:59:18 +0000 (GMT) Received: from smtpauth01.mail.atl.earthlink.net (smtpauth01.mail.atl.earthlink.net [209.86.89.61]) by svr1.postgresql.org (Postfix) with ESMTP id 7A9E0D6D83 for ; Sun, 21 Aug 2005 19:59:16 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=BnlOVH8co24DmRuCFfRnR+J4eSD9eWQgiJZTqu3jZgFKaMtivQ9ZfzDfgQuI0hbD; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth01.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E6ymd-0002lZ-K7; Sun, 21 Aug 2005 18:59:19 -0400 Message-Id: <6.2.3.4.0.20050821185759.01f7c720@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Sun, 21 Aug 2005 18:59:15 -0400 To: Jeremiah Jahn , postgres performance From: Ron Subject: Re: extremly low memory usage Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bc830015ee01ac8153963bc8f9c0569fda350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.453 required=5 tests=[AWL=0.079, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/313 X-Sequence-Number: 14060 I'm resending this as it appears not to have made it to the list. At 10:54 AM 8/21/2005, Jeremiah Jahn wrote: >On Sat, 2005-08-20 at 21:32 -0500, John A Meinel wrote: > > Ron wrote: > > > > Well, since you can get a read of the RAID at 150MB/s, that means that > > it is actual I/O speed. It may not be cached in RAM. Perhaps you could > > try the same test, only using say 1G, which should be cached. > >[root@io pgsql]# time dd if=/dev/zero of=testfile bs=1024 count=1000000 >1000000+0 records in >1000000+0 records out > >real 0m8.885s >user 0m0.299s >sys 0m6.998s This is abysmally slow. >[root@io pgsql]# time dd of=/dev/null if=testfile bs=1024 count=1000000 >1000000+0 records in >1000000+0 records out > >real 0m1.654s >user 0m0.232s >sys 0m1.415s This transfer rate is the only one out of the 4 you have posted that is in the vicinity of where it should be. >The raid array I have is currently set up to use a single channel. But I >have dual controllers in the array. And dual external slots on the card. >The machine is brand new and has pci-e backplane. > So you have 2 controllers each with 2 external slots? But you are currently only using 1 controller and only one external slot on that controller? > > > Assuming these are U320 15Krpm 147GB HDs, a RAID 10 array of 14 of them > > > doing raw sequential IO like this should be capable of at > > > ~7*75MB/s= 525MB/s using Seagate Cheetah 15K.4's >BTW I'm using Seagate Cheetah 15K.4's OK, now we have that nailed down. > > > AFAICT, the Dell PERC4 controllers use various flavors of the LSI Logic > > > MegaRAID controllers. What I don't know is which exact one yours is, > > > nor do I know if it (or any of the MegaRAID controllers) are high > > > powered enough. > >PERC4eDC-PCI Express, 128MB Cache, 2-External Channels Looks like they are using the LSI Logic MegaRAID SCSI 320-2E controller. IIUC, you have 2 of these, each with 2 external channels? The specs on these appear a bit strange. They are listed as being a PCI-Ex8 card, which means they should have a max bandwidth of 20Gb/s= 2GB/s, yet they are also listed as only supporting dual channel U320= 640MB/s when they could easily support quad channel U320= 1.28GB/s. Why bother building a PCI-Ex8 card when only a PCI-Ex4 card (which is a more standard physical format) would've been enough? Or if you are going to build a PCI-Ex8 card, why not support quad channel U320? This smells like there's a problem with LSI's design. The 128MB buffer also looks suspiciously small, and I do not see any upgrade path for it on LSI Logic's site. "Serious" RAID controllers from companies like Xyratex, Engino, and Dot-hill can have up to 1-2GB of buffer, and there's sound technical reasons for it. See if there's a buffer upgrade available or if you can get controllers that have larger buffer capabilities. Regardless of the above, each of these controllers should still be good for about 80-85% of 640MB/s, or ~510-540 MB/s apiece when doing raw sequential IO if you plug 3-4 fast enough HD's into each SCSI channel. Cheetah 15K.4's certainly are fast enough. Optimal setup is probably to split each RAID 1 pair so that one HD is on each of the SCSI channels, and then RAID 0 those pairs. That will also protect you from losing the entire disk subsystem if one of the SCSI channels dies. That 128MB of buffer cache may very well be too small to keep the IO rate up, and/or there may be a more subtle problem with the LSI card, and/or you may have a configuration problem, but _something(s)_ need fixing since you are only getting raw sequential IO of ~100-150MB/s when it should be above 500MB/s. This will make the most difference for initial reads (first time you load a table, first time you make a given query, etc) and for any writes. Your HW provider should be able to help you, even if some of the HW in question needs to be changed. You paid for a solution. As long as this stuff is performing at so much less then what it is supposed to, you have not received the solution you paid for. BTW, on the subject of RAID stripes IME the sweet spot tends to be in the 64KB to 256KB range (very large, very read heavy data mines can want larger RAID stripes.). Only experimentation will tell you what results in the best performance for your application. >I'm not really worried about the writing, it's the reading the reading >that needs to be faster. Initial reads are only going to be as fast as your HD subsystem, so there's a reason for making the HD subsystem faster even if all you care about is reads. In addition, I'll repeat my previous advice that upgrading to 16GB of RAM would be well worth it for you. Hope this helps, Ron Peacetree From pgsql-performance-owner@postgresql.org Mon Aug 22 06:11:45 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6BEFAD6E91 for ; Mon, 22 Aug 2005 06:11:43 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 17642-07 for ; Mon, 22 Aug 2005 09:11:39 +0000 (GMT) Received: from apate.telenet-ops.be (apate.telenet-ops.be [195.130.132.57]) by svr1.postgresql.org (Postfix) with ESMTP id 628DCD6E7F for ; Mon, 22 Aug 2005 06:11:33 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by apate.telenet-ops.be (Postfix) with SMTP id 8462238189; Mon, 22 Aug 2005 11:11:26 +0200 (CEST) Received: from [10.0.1.2] (d5152B1F3.access.telenet.be [81.82.177.243]) by apate.telenet-ops.be (Postfix) with ESMTP id DE1783828E; Mon, 22 Aug 2005 11:11:25 +0200 (CEST) In-Reply-To: <1124651165.5990.1.camel@noodles> References: <1124651165.5990.1.camel@noodles> Mime-Version: 1.0 (Apple Message framework v622) Content-Type: multipart/mixed; boundary=Apple-Mail-46-365797590 Message-Id: <72e7fd0cfc36da6420ba0e92a390b561@implements.be> Cc: pgsql-performance@postgresql.org From: Yves Vindevogel Subject: Re: (Re)-indexing on updates Date: Mon, 22 Aug 2005 11:11:25 +0200 To: "Jeffrey W. Baker" X-Mailer: Apple Mail (2.622) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/315 X-Sequence-Number: 14062 --Apple-Mail-46-365797590 Content-Type: multipart/alternative; boundary=Apple-Mail-47-365797590 --Apple-Mail-47-365797590 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=ISO-8859-1; format=flowed The option with T1: A B C and T2 A D (to avoid the updates) works very well with a simple query Insert into T2 (A, D) select A, functionToGetD from T1 left join T2 on T1.A =3D T2.A where T2.A is null The above gives me the new records for those where D was not filled yet. Since they are all new records, I have no trouble with the MVCC On 21 Aug 2005, at 21:06, Jeffrey W. Baker wrote: > On Sun, 2005-08-21 at 20:32 +0200, Yves Vindevogel wrote: >> >> >> = ______________________________________________________________________ >> >> Hi, >> >> Say I have a table with column A, B, C, D >> A has a unique index on it (primary key) >> B and C have a normal index on it >> D has no index >> >> If I perform a query like update tbl set D =3D 'whatever' ; >> that should make no difference on the indexes on the other columns, >> right ? > > What postgresql does on update is to make a new record, so there will=20= > be > two records in your table and two records in your index. You would=20 > need > to vacuum the table to mark the space for the old record free, and you > would need to reindex the table to shrink the index. > >> >> Or is there some kind of mechanism that does create a sort of new >> record, thus makes the indexes go wild. > > Yes. > > -jwb > > > ---------------------------(end of=20 > broadcast)--------------------------- > TIP 9: In versions below 8.0, the planner will ignore your desire to > choose an index scan if your joining column's datatypes do not > match > > Met vriendelijke groeten, Bien =E0 vous, Kind regards, Yves Vindevogel Implements --Apple-Mail-47-365797590 Content-Transfer-Encoding: quoted-printable Content-Type: text/enriched; charset=ISO-8859-1 The option with=20 T1: A B C and T2 A D (to avoid the updates) works very well with a simple query Insert into T2 (A, D) select A, functionToGetD from T1 left join T2 on T1.A =3D T2.A=20 where T2.A is null The above gives me the new records for those where D was not filled yet. Since they are all new records, I have no trouble with the MVCC On 21 Aug 2005, at 21:06, Jeffrey W. Baker wrote: On Sun, 2005-08-21 at 20:32 +0200, Yves Vindevogel wrote: ______________________________________________________________________ Hi, Say I have a table with column A, B, C, D A has a unique index on it (primary key) B and C have a normal index on it D has no index If I perform a query like update tbl set D =3D 'whatever' ; that should make no difference on the indexes on the other columns, right ? What postgresql does on update is to make a new record, so there will be two records in your table and two records in your index. You would need to vacuum the table to mark the space for the old record free, and you would need to reindex the table to shrink the index. Or is there some kind of mechanism that does create a sort of new record, thus makes the indexes go wild. Yes. -jwb ---------------------------(end of broadcast)--------------------------- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match Met vriendelijke groeten, Bien =E0 vous, Kind regards, Yves Vindevogel Implements = --Apple-Mail-47-365797590-- --Apple-Mail-46-365797590 Content-Transfer-Encoding: base64 Content-Type: image/tiff; x-unix-mode=0666; name="Pasted Graphic 2.tiff" Content-Disposition: inline; filename="Pasted Graphic 2.tiff" TU0AKgAAFciAP6BP5/wWDQeEQmFQuGQ2HQ+FP5+v9rOh5P9IMBuP8zK9tP8fJNlP8QoNjP8FndhP 8MnyVjJHSMrqFnv9BL1vP9ett2v98PuJxChUOiUWGQOCUalUujOV4PZ/qBluR/mdVtJ/jVHst/ho 9MF/g08WABHJfv8AWa0WoAHFfWuzgg6sB/hY9ysUohjv8qqVqv9LMZxv9ouV40zEYmHUjFY2HPx+ RNtOl4Rhftt/phiRs6LC/iA5LewnJeWg46W22+02cAHOwa26bC0HPY7TZ7G2G9dv8EHFdP8Om9YP 8rKJov9UtFzv9Dzh/sRvu5/vZ9PzHdeF4zsUqgQRlOR5v9DMBwv8uqe/jZIVwNHu6B0/WAKHvWGr RAA1ra0G1c/vfgANrdraXrTQIAA6JW2TUrQNz/ja0oAjO4YAjSWS0DOWawjys4RkIk4Lj2sAWEUZ B/ieUJqH+PZdnAf5aG06R5nsfLtqW7UaoWZJwukwLyi4VRsn+F5HGYf8QJWBQ7JWAI5tYti2Dm1T bQHAq0De1Q2FwtA1Fof4AjVDIBDQV0vDQWK0DY/oADYWsrNQN81NOuDZtYOM6ycs4Cjq+Q9JWE5D pGJpOpqP5dMwW5sHUf53xlHCExupjqoIbh3Hu55vp6ch3Hof4wlMrAdEWYh/gmP1RtXK0DDi2rXy nVzXjpVs8NM1g1vuNU2gANEugGMBRwu4YADTNsvwyAA3v+0k6LRWLb2dBUpjk2I4NVOx/gYPC6BU QiwCgSsSlWZpxH+ahzsOaBzvCe7qoMpKlUghd1n2f5oHId5/lQaJ0n+RBf3GLpUGwf4iksrg0FVF ICzlXT9QBB0tWO/+FjlKVWtfijTNRBsvDbhrdP3NUAwY38KFfkc0PuNjRWLLz8ytiTUVdVlnjlA2 MWPAWRAAM8LAAMjhhuR5ir4ULjigT5pn+P5co2T6pn+YBup6dp6xoh6kHie95modNOFkah0H+Qhd pyKBNmcf68r2Bo9GGtdpwIAUotM1Q4QMNzUDZhuVZbLsvy7KkqZZCnAb4AFcQY1C3LW1Wb5lOa2t ZvUt78NNjcMN83wEOMDczvuTzXNQ3QEN01WRL2b8XBYEDuugQEFUYgkpIo5FjIJTmhfZkHG8J0nm fB/sifp/EQYLBioUKsBkRKVgWPLX7stA4VnaDa+tZlZel0j+gCMpWS8MRULQNeIDY/4zTIAIyFX0 GFybZnr+rZ8p2a1s6tUNfADXXI1Yb8jLX+MNQk+1Az7zWqwVa/FaJrHpvSLeAQOhdE/l7CSJwrAb BXDXH+PkyA/gUiPGaftASWW6JeDgnGAsCmLNvcYqlz5+FchnZMrqGbeE3JoYhAMAKbEvOGeigtaE CDZwJcgxtiML0uMgQYzllKt2GhtP/EaH6cgAm+fnCtVEVXFOKNU54AC1B/tyLfBx4YJBGwhAAG5m yUoGH/TgxlZ7nGXw3YdEpYTPWOvaSq/JVEX28rEDQsGKDoGPx+QYnE1URg2paZY4dy7EHTx9cc41 upqIfwkjSgJWyW0MgBdK9I1DGAAwnS8G8+4cGYyUWWbKMg/gSiOjQG+FMQ44qqlmgBNUSYYMohKW 2ULFZaR8NdMFucoz/w/ZcAAMzgH+j/AKGV8QAZkhmWNM0AJ/I5zCiJEGYjF1qoGmxI5MqwT7JVlG buLULJhPXlbK+WMt3HwHmIbF8xaA0n6mvPgN7EJhx9Vcax+hsX6wGZuldNDnTSgQDoloGohS6BYE 62gPQqyRhlFE2gIYkCRgdD4WcAUsnxvlYklKgUCn4upi29JkkpG/pbYgtKbr86TLOnbLCF066ZLP kGsI+7NY9rRjZHB+0ew4M5S0AKT4FA8l0BaIUk4XHkj/EYLwzAtVyj/GwRYf49B8D6H+OMdw9R/j DG8vcTgx1xhpFTBoHYilRgaDygQAobXCynVVSSVdRYTLJls5CcIaqX0zpw/Kms76c2DSnTsNNPaA TbpggqOSCg4MkQwcAPiBAzimJqI4Xg3R/itX0P8bY66xD7MiYkgg5h4vAF2NgdY/xNGCH+HQVMIQ WiCNVM2L6Ap4s3ShFixMubA2HnjcUulhab3GuIbGxVjLlxXZoawMKFgsCiaSOUd6nB6XbH+PofKN LtqcH7eMf5AyimQOtegf4/TIEFvYP8dyjTm2eV1KdmNgrlIKr/cOxF+LjmRldTZY88L/U6P/YtyF +b82PDAhkOQsyNkHu8jS76NB4YXH+PvDUG7v3lIFeu8mFbuj6q8PXE2HEaXqIEUEgxnh2T2YhD/B WBZcUunVjO59yMB3PsHc3BONLemxwaP8NwsTMEPvVeMieJKvXmHZk8f48h5GHyYP8e+V3g3tvMQs VY01FLCxjgTHk8b943yBTOVoJEiXJzPgae1zr+2HwYhnIuRzrjczwP8eY8zDmKy7l8NOYcx40zKW nNswizytg+kXHc/qS48x9o7QbNMhZ0yMdvPBG7wmNz/jB6WYs436uFmbHhbFrAFgjhl4Y/g5IvH+ 0EkYFw+NuAHqmoet9bWJwPnDHEfM55E0udjTNW7uZ+y9p6L+oNe3Bxtoa4lACzgQee2kQ5Jwo1RK QPQfK8xejaJ6FG6w/wEh1NZJ+L6ydoH+zfqTUK0dK7AzsY7Yem9jaA0Fu2wVf9eJObqW+TxuwbCS K4JEYVnh2O/vWQM5K+wtCehCCAP5YABybV035xCAJIJqnrTyFqB9H8e1Lu/OumM870MTp3MGn9J8 g5Zsxhk2ZURLfGmrBAEg5m/CISIf4jRekblaCgRoyS0BjWNZVw6uZ70hnsrkND/Fbn3kWWtmz1ID a5NxyLYJ1xujdI3iZTm9dkYy4/P9yFKIcFoDQsbtKW4nT2YbLrpr450V6zTGem7N+YzXVzDZ0PSl ddIb8GQVrlD9sQjy4FKHVy0ZDDeLHCPWuuD/68Y0Vg1LX8c0bQSBiAq9JrVv2pYwZkzq6mqxDi8y XJgBkjZGftMMdUgm5PKyCBnogBI72hYgZlgrD6WWgMzPZkxGdUzfIfIzsdb66PXr/Jxp+XwRGlOL GnuJm2b0l7vo/dJeDL4N7qwWN+qn5CqedNMATuzZmObaad1s4dBMl/6wmG/Pt1+oAAYk2hToiuQc 5FxyjyRoHqH0xYKE+Q8k+UIgH0KAH+HSHqXmGwHSIuEGF0I25eWE8A96/g2akwhIk8Puf25mpi3w /IeG/M8y0O10/WgM3cpUS8l0mS+ocOYgDQS0ASDUTOB6EaJWDkXyXwFyL+GqGkPKHQHMOkxIXmIO HAHAJyyuKgvGIIHgU2H+G0GuHKH+FwF6KwEAtAH+CUEcLOAiNC2QnKcOZSZCmO9ongNep2sA3Y2W nY/KwE9hDazc8w8Sug9mh6cAZ0/UAEC0FKH+AOCOEoJQBeD8H+A2A+DYYGBuD2H+D6DmFCH+F0Fq hDB+PKGcGaaSGsGmJyGOGGg0EuEgP0CqCOEOH+BIA+DQH+AgBSDrD+CAEcH+AICwFNA+pamupfDQ /G1E2asfDksJDesNBCv012x+5WDmQSseeiAaCGEiH+BEAmC0H+A8AcCyH+A4AXGqAqAOCsLqAQCw H+BMAuDIH+CJEWPMCpFKCaCEEASEBFESA2AaC2H+AoAOCqK6AVGqA8AYCvGcAiC9HmBlHZBglBDt F80LF7BNDdBHDg2U0JGI0lGEseWsAgB6EYH+BUAkC4H+BKAuDDGcArH+BCAnH+A8AeC6K7HjG7H4 AoANHtG1G4A1GxGkAfHkBAAjJOBCAqC/FMAuDAH+BOApJ8A0BYDy2aWtGEzI1G2dKRIUwDGDF9BP DooBISnVImB9IsBRJHI2AuDHK3HHI4DFK3K6BGArI8BAAlJ8A8AhJqAlJxJBK3LDLBK9LmBNKCH+ AwBZEZIGcjGM3zKVIRL7BEH8zVKfKojrKlMCQVIk2jKuH/KzH/I5K7MjLFMpMnI5HGBIAvG/I5J9 MvMrK5M/K7LrJ9LxL1DXL5KYp1L/MM80g2wABMzWAGjVDZIc/XDrIiNjKtKxK1MtNBN7MlN9HFJ5 M3J7NDONMnNHLvLzKNKnMS5dNPMAsGpgLKLoAQNswAH8EAGAXGB0EobQAwEEJOblOa1uzJIfNvF9 MXFVMbMfOPODOBOBMxM1K3M7PfPdNFLtNLOZOcfjIOgVPIfeAaD2VGBUEY6ECyFQL+awXWH+HCHa KgW+KoAqDsPu3Mt8rxMO+fOjPTNzMZN3MhPtN/M/PlOJPrPhPvHBPzOXL3KPKhP69Mv4SmYWWs4m QyDGFUGsH+GeHEOkHW204SIED6FmaSBqeZHmQ3Fikw7iPwZWkCSs40Yg0CwTQAxpPUAhPZN5RDS1 PjOHPpRRRFOTP1RZSo5W1skw6MDa6g5omYZW94AYDkN+BOD+NKCoE6SKHy1WA8EOL2Z8cuhGTU/f SlArAu95L2cmiO1xSrQ7PXQ/S/S1RJS9RFUlNBTDRXNPRavw3St25kTXBiQsADBfUFNO/fUC6K+A N+9fTI2SkOfG6YmY9MZWDEFS+0TInFUQSfPI1NQ9MdSzRPUnS5MzRLUdRPUqD02RRaeoseDdSaQn ScQi9G/kPu+e/o+E8UQUoGv/IXMKsOfqZu8O/eV2S89E7QZ7XG8w/edGQYxiTeoSB3FKBQAlRBV9 S3LnWDUjXpN7LCBOAnJ8AyBXKKV0Z7KPKO6i764rXI9wS8DJVq6TAw+hP5MFBJDi0Ofq0i86mQPv XMDRWlBi6QVy6iAYCUEuJIBDESBGA0DNI2AxLjRCAxS4BGAvH5I5J2BLZdZVV9ZtZqDKJIA+DUH+ AiB4ES2a4pDFDE+fVIkg84ZylyTUn7RdYjIZYhQy14fipY9SDTWhSjRg6UAEC+fZGWEfEMBCDbFM AtJ9MzI9LlM8BFZjJ5ZpNBMzLCBJLKH+A+A3ZSAiByEMmcC0FO7Q7eTVSW+ectBY9HQ2oFDTNXak Ns9fIa0hPPVUfqfq5i3OjohIAEC6fYAWB6ERFUBSDxLuAtZSA7JSA/JoH/LPH+BIAtK7bZG4BGAt JCApH+A+AhJPGnJOA0AfH+AiBIDoWuBzc6AICvD8PxUAWJWWj2sfWwsPP9cXWzKc/PGE0jPQvwiE 9kWsAeB+EVI2AaCpdQAbGjGnHkA4AbJPHhI0A2AZI0A6AdI1drG+A6AfG/HhGjfVI1fLfJGpfBGq BGAZHsAvKJA+j7eZencVNSlZGBelINchalStIqbTV7J5I8BDLtdrJPfLfSAZHkA2AXfPfXENJTfl JrK0BFdjMoBOArX7UsYhUxaledgRcZgVBLMTepVU17StSxXlMrZzbgAtLDgrdoAlGiBBK1dhZZLi AxK/OFORRVNNhbhu0PhhafgTW1gXilgbgRhzUbV/OOAwDPFMAxe/ZrPriXXnWJidP3hiuZgPiphk eGA8EUJGWEc3MNhtgdUXSvi5XxS1Z3Zhe/MzbfjNRRUrieVpeejtOg0OseDeNYQexAIEE4GcX2B4 EkRKAMDsNjXSTWcVVVjvi1jzh1WHM/j8AvkAAvkFlHkLjVjdjZF4pPk0nANKLKQIBCESL2D6F6XG XgvWqwHQMqEsGIRaL6RSAgD6bcQAbyNRBkLQDCWNhdNxkPj1V5h3i7N7lLlPlTmtUpjTTHak1tUE DHDyQMDWN+ggQIByEiJOEGOcGIG8Okq7CMXay2KMG8HaU4rcMGCIE8aSBcEWL2Ci4ccggbKXhxlD j3jPPdmxbdLnm3PxNJhZkO1CseseDUNKBIEMJWB8EqhCBqEuGgH+DiFiL+GXQcINAEIhl4McIkXc IICurWWuDulCNiDW85VVi3mplHmuH/j/oZM9odRTohkNNQnjYIc6QIAILmH+BgEOLOtabCUcIPpV qiIOHMKeOQGoJ6C+FaMwAuEJT474DYTfliJQrbgjmrj5OBoXkDobrTJ/LsArX+LQDI72nBlkJQLv C2FBpAEoGOMGGhqsvKH7nlqpqlnpsKKMHgJ+tgGQPKCmFIaSAwD8LOA4EGJWBiCdEFH1G/iPp0Av rXlRrboTdXLDdrI0BSBkD4H/n8JOAkD7o0EwhCECF+PKG6HceBsQRtsPtyOuq4/+22RcFYJGBkBK Ddp4AtI8A2AdHkA/H9uPZZrVp5lNp9OFMzK7iLJ9HhHlgrJ8BSA3FSEWECe+vNsVnkXdt4MVqnvR qiHmHkKgF+F0JqEMD8fECQB4D7uPI8AyAZGjubJ8BAAqClbqAng4AbfZdsKyBTuMDiDOE2H+FmFe JGHMU1vXwrvVwrwxCKagF3pADyDiFEH+CWCAEHJ+A3G4B+BmDmH+DQC7EEFGE4NKHYHUXvwxxqIj t3xtxyIaw0ImGwGus8HgUZx1yGavxxyJyPyRySKIKQFuFibQGOGEYFyVynypymKQCaB/xIDkDOE8 IQFSFEVGE+EsQIIWGEF4L+EwEeP6IOE4EoN2DOC0ErpCDLy6DYDAEyH+FiFU6EHyq6H+DqDUFAH+ G4GwOWIWw0IIEMD68HzsE0H+DgDIE4H+DaDD0cG8G4X2FgFS6EFGE2LAIWG0GsHMH+EoEWTaIOFn uEH+DKCzzkDjzoH+DJ1aH+FuFgbRsGIIEKD2fZyCrEIOGUGMI2E8EqNLz6XmDsDXEgDUC6EwH+DW C9zx1ZzkF4FuOOHUq0D8Dnb8DeDHwd0h0kEeEKQsKQCkCLFKDv2SIQD6DkFJENg2H/ygYFxkPCBu BQDkH+CwCXIsIOCiCJb2CQB0D+H+FEE0LoEYEGTOCgCGEKH/2uMrGfI8GEF7RyIWHwOoH+CEBptU EaEITOFCEyLOF0FoJr16aUDpb9fbH+GAF0RSz2KgCABjdACSB4ED3V3YH+C8CjbCFAEwLP006EG8 G2X35LbnK6B/5iH+GcGSs8FoFehCC0CYEavXAT5WRSDyDdxACKBtvwFwFkJqHEHAUVxeLACSB3HZ 2GNKEyEgP6EsEaNEKQCn3MH+DuDZEgIOEUEAZMDuDbxB7jFL5zbCDQC2Et2cC/zwIP8AH+BoBKDf 1iCxEEC4CfFgEqEYNFCeU4BWA2DSXIGgXGIWu+XmCQB35r2R0EEgEOQyIWEOD88GECDvVmCyCZIs DACpGaE0EkN+DP8GIQEGDyFUH/3LFL0Ue+HkHh18IMHPwoCqCRe54ILoCMBvEKC6Cl8CC18IIWFa FKL2DGCtEEIXE0MH2+H+EKD4fZ10fZ374Z7h7l7p7sIMEF9eRd6d1p1t4H4L/ohCDECsEkIQIAVC MiX+dTUoH++n0/X++4W/4hEHe7Xo/xQGTO/zQXEu/0Cd1S/0Yglg/3g73q/yWPUE/zsa1C/0IeVW /0GeFU/3G4Xa/0MfVY/1or2bQlbRUiiVpJpQ/y+U0jEYggzzOSkRUM/0KfJqfzop3+p1Aw3+53I7 n+UCJWak3m26H+OxYdX+Xiij6leVRY3+YKheaks1Yy3+VCPBEIe5zM5qOxbdICAADgEAAAMAAAAB ADgAAAEBAAMAAAABAEAAAAECAAMAAAAEAAAWdgEDAAMAAAABAAUAAAEGAAMAAAABAAIAAAERAAQA AAABAAAACAEVAAMAAAABAAQAAAEWAAQAAAABAAAAkgEXAAQAAAABAAAVwAEaAAUAAAABAAAWfgEb AAUAAAABAAAWhgEcAAMAAAABAAEAAAEoAAMAAAABAAIAAAFSAAMAAAABAAEAAAAAAAAACAAIAAgA CAAK/IAAACcQAAr8gAAAJxA= --Apple-Mail-46-365797590 Content-Type: multipart/alternative; boundary=Apple-Mail-48-365797592 --Apple-Mail-48-365797592 Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=US-ASCII; format=flowed Mail: yves.vindevogel@implements.be - Mobile: +32 (478) 80 82 91 Kempische Steenweg 206 - 3500 Hasselt - Tel-Fax: +32 (11) 43 55 76 Web: http://www.implements.be First they ignore you. Then they laugh at you. Then they fight you. Then you win. Mahatma Ghandi. --Apple-Mail-48-365797592 Content-Transfer-Encoding: 7bit Content-Type: text/enriched; charset=US-ASCII Mail: yves.vindevogel@implements.be - Mobile: +32 (478) 80 82 91 Kempische Steenweg 206 - 3500 Hasselt - Tel-Fax: +32 (11) 43 55 76 Web: http://www.implements.be First they ignore you. Then they laugh at you. Then they fight you. Then you win. Mahatma Ghandi. --Apple-Mail-48-365797592-- --Apple-Mail-46-365797590-- From pgsql-performance-owner@postgresql.org Mon Aug 22 06:29:24 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9ED92D6E50 for ; Mon, 22 Aug 2005 06:29:22 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 62122-01 for ; Mon, 22 Aug 2005 09:29:17 +0000 (GMT) Received: from smtp104.mail.sc5.yahoo.com (smtp104.mail.sc5.yahoo.com [66.163.169.223]) by svr1.postgresql.org (Postfix) with SMTP id B66DFD6E3E for ; Mon, 22 Aug 2005 06:29:15 -0300 (ADT) Received: (qmail 43782 invoked from network); 22 Aug 2005 09:29:13 -0000 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Received:User-Agent:Date:Subject:From:To:Message-ID:In-Reply-To:Mime-version:Content-type:Content-transfer-encoding; b=jWMuYO6I1MB8y7tsC64ummRCWCTYVb6bPFdocmn64Yt7n8YiKUuYoKLRzUe/w/pAvGSVLFHOeTZW9WYjMVO/J0rJb6WQ7Oy065wrHXd0EmnhkzjHl6GtEl30rnpnE+/LcaTnbTNzehC+rxMiYQbPCmrPx2pGE38ABniRrNjcOK4= ; Received: from unknown (HELO ?192.168.0.103?) (mcotner@66.56.49.200 with login) by smtp104.mail.sc5.yahoo.com with SMTP; 22 Aug 2005 09:29:13 -0000 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Mon, 22 Aug 2005 05:29:12 -0400 Subject: Re: sustained update load of 1-2k/sec From: Mark Cotner To: "J. Andrew Rogers" , Message-ID: In-Reply-To: Mime-version: 1.0 Content-type: text/plain; charset="US-ASCII" Content-transfer-encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.28 required=5 tests=[AWL=-0.094, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/316 X-Sequence-Number: 14063 Thanks again everyone for the excellent suggestions. I looked into IO::Reactor, but after a few hours of fiddling decided I was getting the kind of performance I wanted from using a slightly more than modest number of threads and decided(due to dev timelines) to come back to patching the SNMP libraries for Ruby to do async using Reactor later. I am unfortunately stuck with updates, but I think(with you're suggestions) I've made it work for me. MySQL = 1500 updates/sec PostgreSQL w/10k tx per commit using single thread = 1400 updates/sec Given the update heavy nature of this table I felt it was necessary to test during a vacuum. Turns out the hit wasn't that bad . . . PostgreSQL w/10k tx per commit using a single thread during a vacuum = 1300 updates/sec 100-200 updates/sec is a small price to pay for mature stored procedures, more stored procedure language options, acid compliance, mvcc, very few if any corrupt tables(get about 2 a week from MySQL on the 40 DBs I manage), more crash resistant db(crash about once a month on one of my 40 MySQL dbs), and replication that actually works for more than a day before quitting for no apparent reason ;) [/flame off] For those of you with Cox Communications cable modems look forward to better customer service and cable plant management. :) And if anyone's curious here's the app I'm rebuilding/updating http://www.mysql.com/customers/customer.php?id=16 We won runner up behind Saabre airline reservation system for MySQL app of the year. Needless to say they weren't too happy when they heard we might be switching DBs. 'njoy, Mark On 8/19/05 1:12 PM, "J. Andrew Rogers" wrote: > On 8/19/05 1:24 AM, "Mark Cotner" wrote: >> I'm currently working on an application that will poll >> thousands of cable modems per minute and I would like >> to use PostgreSQL to maintain state between polls of >> each device. This requires a very heavy amount of >> updates in place on a reasonably large table(100k-500k >> rows, ~7 columns mostly integers/bigint). Each row >> will be refreshed every 15 minutes, or at least that's >> how fast I can poll via SNMP. I hope I can tune the >> DB to keep up. >> >> The app is threaded and will likely have well over 100 >> concurrent db connections. Temp tables for storage >> aren't a preferred option since this is designed to be >> a shared nothing approach and I will likely have >> several polling processes. > > > Mark, > > We have PostgreSQL databases on modest hardware doing exactly what you are > attempting to (massive scalable SNMP monitoring system). The monitoring > volume for a single database server appears to exceed what you are trying to > do by a few orders of magnitude with no scaling or performance issues, so I > can state without reservation that PostgreSQL can easily handle your > application in theory. > > However, that is predicated on having a well-architected system that > minimizes resource contention and unnecessary blocking, and based on your > description you may be going about it a bit wrong. > > The biggest obvious bottleneck is the use of threads and massive > process-level parallelization. As others have pointed out, async queues are > your friends, as is partitioning the workload horizontally rather than > vertically through the app stack. A very scalable high-throughput engine > for SNMP polling only requires two or three threads handling different parts > of the workload to saturate the network, and by choosing what each thread > does carefully you can all but eliminate blocking when there is work to be > done. > > We only use a single database connection to insert all the data into > PostgreSQL, and that process/thread receives its data from a work queue. > Depending on how you design your system, you can batch many records in your > queue as a single transaction. In our case, we also use very few updates, > mostly just inserts, which is probably advantageous in terms of throughput > if you have the disk for it. The insert I/O load is easily handled, and our > disk array is a modest 10k SCSI rig. The only thing that really hammers the > server is when multiple reporting processes are running, which frequently > touch several million rows each (the database is much larger than the system > memory), and even this is manageable with clever database design. > > > In short, what you are trying to do is easily doable on PostgreSQL in > theory. However, restrictions on design choices may pose significant > hurdles. We did not start out with an ideal system either; it took a fair > amount of re-engineering to solve all the bottlenecks and problems that pop > up. > > Good luck, > > J. Andrew Rogers > jrogers@neopolitan.com > > > > ---------------------------(end of broadcast)--------------------------- > TIP 3: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faq From pgsql-performance-owner@postgresql.org Mon Aug 22 10:28:23 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id EB976D6EA7 for ; Mon, 22 Aug 2005 10:15:10 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 28144-01 for ; Mon, 22 Aug 2005 13:15:08 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id D71B9D6E91 for ; Mon, 22 Aug 2005 10:15:06 -0300 (ADT) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: Finding bottleneck Date: Mon, 22 Aug 2005 09:15:10 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3417DD15D@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Finding bottleneck Thread-Index: AcWkxr5EndVUITMSSs6KP/tRXihkJAAF32MAAI8Vy3A= From: "Merlin Moncure" To: "Tom Lane" Cc: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/317 X-Sequence-Number: 14064 > Bill of Materials Traversal ( ~ 62k records). >=20 > ISAM* pg 8.0 pg 8.1 devel delta 8.0->8.1 > running time 63 sec 90 secs 71 secs 21% > cpu load 17% 45% 32% 29% > loadsecs** 10.71 40.5 22.72 44% > recs/sec 984 688 873 > recs/loadsec 5882 1530 2728 >=20 > *ISAM is an anonymous commercial ISAM library in an optimized server > architecture (pg smokes the non-optimized flat file version). > **Loadsecs being seconds of CPU at 100% load. One thing that might interest you is that the penalty in 8.1 for stats_command_string=3Dtrue in this type of access pattern is very high: = I was experimenting to see if the new cpu efficiency gave me enough of a budget to start using this. This more than doubled the cpu load to around 70% with a runtime of 82 seconds. This is actually worse than 8.0 :(. This *might* be a somewhat win32 specific issue. I've had issues with the stats collector before. Anyways, the feature is a frill so it's not a big deal. Merlin From pgsql-performance-owner@postgresql.org Mon Aug 22 11:06:22 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 55327D6E69 for ; Mon, 22 Aug 2005 10:47:37 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 27253-01 for ; Mon, 22 Aug 2005 13:47:35 +0000 (GMT) Received: from hotmail.com (bay101-f3.bay101.hotmail.com [64.4.56.13]) by svr1.postgresql.org (Postfix) with ESMTP id 17564D6E5B for ; Mon, 22 Aug 2005 10:47:34 -0300 (ADT) Received: from mail pickup service by hotmail.com with Microsoft SMTPSVC; Mon, 22 Aug 2005 06:47:38 -0700 Message-ID: Received: from 64.4.56.200 by by101fd.bay101.hotmail.msn.com with HTTP; Mon, 22 Aug 2005 13:47:38 GMT X-Originating-IP: [64.4.56.200] X-Originating-Email: [jigneshk@hotmail.com] X-Sender: jigneshk@hotmail.com From: "Jignesh Shah" To: pgsql-performance@postgresql.org Subject: MemoryContextSwitchTo during table scan? Date: Mon, 22 Aug 2005 09:47:38 -0400 Mime-Version: 1.0 Content-Type: text/plain; format=flowed X-OriginalArrivalTime: 22 Aug 2005 13:47:38.0983 (UTC) FILETIME=[0F71B770:01C5A720] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=3.582 required=5 tests=[DNS_FROM_RFC_ABUSE=0.374, DNS_FROM_RFC_POST=1.376, MSGID_FROM_MTA_HEADER=0, RCVD_IN_BL_SPAMCOP_NET=1.832] X-Spam-Level: *** X-Archive-Number: 200508/318 X-Sequence-Number: 14065 Hello, I am running PostgreSQL 8.0.x on Solaris 10 AMD64. My Tablesize for this test is about 80G. When I run a query on a column which is not indexed, I get a full table scan query and that's what I am testing right now. (I am artificially creating that scenario to improve that corner case). Aparently I find that the full query is running much slower compared to what hardware can support and hence dug into DTrace to figure out where it is spending most of its time. Running a script (available on my blog) I find the following top 5 functions where it spends most time during a 10 second run of the script 
                                                    Time in (millisec)       
    Call Count
MemoryContextSwitchTo                   775                           106564
LockBuffer                                        707                        
    109367
LWLockAcquire                                 440                            
  58888
ExecEvalConst                                  418                           
    53282
ResourceOwnerRememberBuffer      400                              54684
TransactionIdFollowsOrEquals           392                              
53281
While the times look pretty small (0.775 second out of 10seconds which is about 7.75%), it still represents significant time since the table is pretty big and the entire scan takes about 30 minute (about 80G big table). Considering it is a single threaded single process scan all the hits of the function calls itself can delay the performance. MemoryContextSwitchTo and LockBuffer itself takes 15% of the total time of the query. I was expecting "read" to be the slowest part (biggest component) but it was way down in the 0.4% level. Now the question is why there are so many calls to MemoryContextSwitchTo in a single SELECT query command? Can it be minimized? Also is there any way to optimize LockBuffer? Is there anything else that can minimize the time spent in these calls itself? (Of course it is the first iteration but something else will be the bottleneck... but that's the goal). If there are any hackers interested in tackling this problem let me know. Thanks. Regards, Jignesh From pgsql-performance-owner@postgresql.org Mon Aug 22 11:50:04 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 52BCBD7019 for ; Mon, 22 Aug 2005 11:17:51 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 93421-03 for ; Mon, 22 Aug 2005 14:17:50 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 4B7BAD7016 for ; Mon, 22 Aug 2005 11:17:48 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7MEHrLp015639; Mon, 22 Aug 2005 10:17:53 -0400 (EDT) To: "Merlin Moncure" Cc: pgsql-performance@postgresql.org Subject: Re: Finding bottleneck In-reply-to: <6EE64EF3AB31D5448D0007DD34EEB3417DD15D@Herge.rcsinc.local> References: <6EE64EF3AB31D5448D0007DD34EEB3417DD15D@Herge.rcsinc.local> Comments: In-reply-to "Merlin Moncure" message dated "Mon, 22 Aug 2005 09:15:10 -0400" Date: Mon, 22 Aug 2005 10:17:53 -0400 Message-ID: <15638.1124720273@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/319 X-Sequence-Number: 14066 "Merlin Moncure" writes: > One thing that might interest you is that the penalty in 8.1 for > stats_command_string=true in this type of access pattern is very high: I > was experimenting to see if the new cpu efficiency gave me enough of a > budget to start using this. This more than doubled the cpu load to > around 70% with a runtime of 82 seconds. This is actually worse than > 8.0 :(. That seems quite peculiar; AFAICS the pgstat code shouldn't be any slower than before. At first I thought it might be because we'd increased PGSTAT_ACTIVITY_SIZE, but actually that happened before 8.0 release, so it shouldn't be a factor in this comparison. Can anyone else confirm a larger penalty for stats_command_string in HEAD than in 8.0? A self-contained test case would be nice too. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 22 13:18:10 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3BB59D6EC0 for ; Mon, 22 Aug 2005 11:42:44 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 60424-03 for ; Mon, 22 Aug 2005 14:42:42 +0000 (GMT) Received: from penguin.goodinassociates.com (unknown [206.80.71.229]) by svr1.postgresql.org (Postfix) with ESMTP id E10FED6E7E for ; Mon, 22 Aug 2005 11:42:41 -0300 (ADT) Received: from [206.80.71.242] ([206.80.71.242]) by penguin.goodinassociates.com (8.13.1/8.12.11) with ESMTP id j7MEgixx032272; Mon, 22 Aug 2005 09:42:44 -0500 Subject: Re: extremly low memory usage From: Jeremiah Jahn To: Ron Cc: postgres performance In-Reply-To: <6.2.3.4.0.20050821112849.01f79860@pop.earthlink.net> References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <1124481386.27881.167.camel@bluejay.goodinassociates.com> <43064908.7010505@arbash-meinel.com> <1124564003.27881.229.camel@bluejay.goodinassociates.com> <6.2.3.4.0.20050820154422.06021150@pop.earthlink.net> <4307E7A4.7090300@arbash-meinel.com> <1124636066.27881.238.camel@bluejay.goodinassociates.com> <6.2.3.4.0.20050821112849.01f79860@pop.earthlink.net> Content-Type: text/plain Date: Mon, 22 Aug 2005 09:42:43 -0500 Message-Id: <1124721763.27881.250.camel@bluejay.goodinassociates.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.2 (2.2.2-5) Content-Transfer-Encoding: 7bit Received-SPF: none (penguin.goodinassociates.com: 206.80.71.242 is neither permitted nor denied by domain of cs.earlham.edu) client-ip=206.80.71.242; envelope-from=jeremiah@cs.earlham.edu; helo=[206.80.71.242]; X-Virus-Scanned: ClamAV 0.86.1/1034/Thu Aug 18 15:07:58 2005 on penguin.goodinassociates.com X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.034 required=5 tests=[AWL=-0.016, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/321 X-Sequence-Number: 14068 On Sun, 2005-08-21 at 16:13 -0400, Ron wrote: > At 10:54 AM 8/21/2005, Jeremiah Jahn wrote: > >On Sat, 2005-08-20 at 21:32 -0500, John A Meinel wrote: > > > Ron wrote: > > > > > > Well, since you can get a read of the RAID at 150MB/s, that means that > > > it is actual I/O speed. It may not be cached in RAM. Perhaps you could > > > try the same test, only using say 1G, which should be cached. > > > >[root@io pgsql]# time dd if=/dev/zero of=testfile bs=1024 count=1000000 > >1000000+0 records in > >1000000+0 records out > > > >real 0m8.885s > >user 0m0.299s > >sys 0m6.998s > > This is abysmally slow. > > > >[root@io pgsql]# time dd of=/dev/null if=testfile bs=1024 count=1000000 > >1000000+0 records in > >1000000+0 records out > > > >real 0m1.654s > >user 0m0.232s > >sys 0m1.415s > > This transfer rate is the only one out of the 4 you have posted that > is in the vicinity of where it should be. > > > >The raid array I have is currently set up to use a single channel. But I > >have dual controllers in the array. And dual external slots on the card. > >The machine is brand new and has pci-e backplane. > > > So you have 2 controllers each with 2 external slots? But you are > currently only using 1 controller and only one external slot on that > controller? Sorry, no. I have one dual channel card in the system and two controllers on the array. Dell PowerVault 220S w/ PERC4eDC-PCI Express > > > > > > Assuming these are U320 15Krpm 147GB HDs, a RAID 10 array of 14 of them > > > > doing raw sequential IO like this should be capable of at > > > > ~7*75MB/s= 525MB/s using Seagate Cheetah 15K.4's > >BTW I'm using Seagate Cheetah 15K.4's > > OK, now we have that nailed down. > > > > > > AFAICT, the Dell PERC4 controllers use various flavors of the LSI Logic > > > > MegaRAID controllers. What I don't know is which exact one yours is, > > > > nor do I know if it (or any of the MegaRAID controllers) are high > > > > powered enough. > > > >PERC4eDC-PCI Express, 128MB Cache, 2-External Channels > > Looks like they are using the LSI Logic MegaRAID SCSI 320-2E > controller. IIUC, you have 2 of these, each with 2 external channels? > > The specs on these appear a bit strange. They are listed as being a > PCI-Ex8 card, which means they should have a max bandwidth of 20Gb/s= > 2GB/s, yet they are also listed as only supporting dual channel U320= > 640MB/s when they could easily support quad channel U320= > 1.28GB/s. Why bother building a PCI-Ex8 card when only a PCI-Ex4 > card (which is a more standard physical format) would've been > enough? Or if you are going to build a PCI-Ex8 card, why not support > quad channel U320? This smells like there's a problem with LSI's design. > > The 128MB buffer also looks suspiciously small, and I do not see any > upgrade path for it on LSI Logic's site. "Serious" RAID controllers > from companies like Xyratex, Engino, and Dot-hill can have up to > 1-2GB of buffer, and there's sound technical reasons for it. See if > there's a buffer upgrade available or if you can get controllers that > have larger buffer capabilities. > > Regardless of the above, each of these controllers should still be > good for about 80-85% of 640MB/s, or ~510-540 MB/s apiece when doing > raw sequential IO if you plug 3-4 fast enough HD's into each SCSI > channel. Cheetah 15K.4's certainly are fast enough. Optimal setup > is probably to split each RAID 1 pair so that one HD is on each of > the SCSI channels, and then RAID 0 those pairs. That will also > protect you from losing the entire disk subsystem if one of the SCSI > channels dies. I like this idea, but how exactly does one bond the two channels together? Won't this cause me to have both an /dev/sdb and an /dev/sdc? > > That 128MB of buffer cache may very well be too small to keep the IO > rate up, and/or there may be a more subtle problem with the LSI card, > and/or you may have a configuration problem, but _something(s)_ need > fixing since you are only getting raw sequential IO of ~100-150MB/s > when it should be above 500MB/s. It looks like there's a way to add more memory to it. > > This will make the most difference for initial reads (first time you > load a table, first time you make a given query, etc) and for any writes. > > Your HW provider should be able to help you, even if some of the HW > in question needs to be changed. You paid for a solution. As long > as this stuff is performing at so much less then what it is supposed > to, you have not received the solution you paid for. > > BTW, on the subject of RAID stripes IME the sweet spot tends to be in > the 64KB to 256KB range (very large, very read heavy data mines can > want larger RAID stripes.). Only experimentation will tell you what > results in the best performance for your application. I think I have them very small at the moment. > > > >I'm not really worried about the writing, it's the reading the reading > >that needs to be faster. > > Initial reads are only going to be as fast as your HD subsystem, so > there's a reason for making the HD subsystem faster even if all you > care about is reads. In addition, I'll repeat my previous advice > that upgrading to 16GB of RAM would be well worth it for you. 12GB is my max. I may run with it for a while and see. > > Hope this helps, > Ron Peacetree > > > > ---------------------------(end of broadcast)--------------------------- > TIP 9: In versions below 8.0, the planner will ignore your desire to > choose an index scan if your joining column's datatypes do not > match -- Speak softly and carry a +6 two-handed sword. From pgsql-performance-owner@postgresql.org Mon Aug 22 12:33:58 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A70C4D6EC0 for ; Mon, 22 Aug 2005 12:07:54 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83816-09 for ; Mon, 22 Aug 2005 15:07:50 +0000 (GMT) Received: from floppy.pyrenet.fr (news.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id 0B71AD6D71 for ; Mon, 22 Aug 2005 12:07:49 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id DCDB330B42; Mon, 22 Aug 2005 17:19:42 +0200 (MET DST) From: William Yu X-Newsgroups: pgsql.performance Subject: Re: extremly low memory usage Date: Mon, 22 Aug 2005 08:07:36 -0700 Organization: Hub.Org Networking Services Lines: 89 Message-ID: References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <1124481386.27881.167.camel@bluejay.goodinassociates.com> <43064908.7010505@arbash-meinel.com> <1124564003.27881.229.camel@bluejay.goodinassociates.com> <6.2.3.4.0.20050820154422.06021150@pop.earthlink.net> <4307E7A4.7090300@arbash-meinel.com> <1124636066.27881.238.camel@bluejay.goodinassociates.com> <6.2.3.4.0.20050821112849.01f79860@pop.earthlink.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Complaints-To: usenet@news.hub.org User-Agent: Mozilla Thunderbird 1.0 (X11/20041206) X-Accept-Language: en-us, en In-Reply-To: <6.2.3.4.0.20050821112849.01f79860@pop.earthlink.net> To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.017 required=5 tests=[AWL=0.017] X-Spam-Level: X-Archive-Number: 200508/320 X-Sequence-Number: 14067 Ron wrote: >> PERC4eDC-PCI Express, 128MB Cache, 2-External Channels > > Looks like they are using the LSI Logic MegaRAID SCSI 320-2E > controller. IIUC, you have 2 of these, each with 2 external channels? A lot of people have mentioned Dell's versions of the LSI cards can be WAY slower than the ones you buy from LSI. Why this is the case? Nobody knows for sure. Here's a guess on my part. A while back, I was doing some googling. And instead of typing "LSI MegaRAID xxx", I just typed "MegaRAID xxx". Going beyond the initial pages, I saw Tekram -- a company that supposedly produces their own controllers -- listing products with the exact model numbers and photos as cards from LSI and Areca. Seemed puzzling until I read a review about SATA RAID cards where it mentioned Tekram produces the Areca cards under their own name but using slower components to avoid competing at the highend with them. So what may be happening is that the logic circuitry on the Dell PERCs are the same as the source LSI cards, the speed of the RAID processor/RAM/internal buffers/etc is not as fast so Dell can shave off a few bucks for every server. That would mean while a true LSI card has the processing power to do the RAID calculates for X drives, the Dell version probably can only do X*0.6 drives or so. > The 128MB buffer also looks suspiciously small, and I do not see any > upgrade path for it on LSI Logic's site. "Serious" RAID controllers > from companies like Xyratex, Engino, and Dot-hill can have up to 1-2GB The card is upgradable. If you look at the pic of the card, it shows a SDRAM DIMM versus integrated RAM chips. I've also read reviews a while back comparing benchmarks of the 320-2 w/ 128K versus 512K onboard RAM. Their product literature is just nebulous on the RAM upgrade part. I'm sure if you opened up the PDF manuals, you could find the exact info > That 128MB of buffer cache may very well be too small to keep the IO > rate up, and/or there may be a more subtle problem with the LSI card, > and/or you may have a configuration problem, but _something(s)_ need > fixing since you are only getting raw sequential IO of ~100-150MB/s when > it should be above 500MB/s. I think it just might be the Dell hardware or the lack of 64-bit IOMMU on Xeon's. Here's my numbers on 320-1 w/ 128K paired up with Opterons compared to Jeremiah's. >> # time dd if=/dev/zero of=testfile bs=1024 count=1000000 >> 1000000+0 records in >> 1000000+0 records out >> >> real 0m8.885s >> user 0m0.299s >> sys 0m6.998s 2x15K RAID1 real 0m14.493s user 0m0.255s sys 0m11.712s 6x15K RAID10 (2x 320-1) real 0m9.986s user 0m0.200s sys 0m8.634s >> # time dd of=/dev/null if=testfile bs=1024 count=1000000 >> 1000000+0 records in >> 1000000+0 records out >> >> real 0m1.654s >> user 0m0.232s >> sys 0m1.415s 2x15K RAID1 real 0m3.383s user 0m0.176s sys 0m3.207s 6x15K RAID10 (2x 320-1) real 0m2.427s user 0m0.178s sys 0m2.250s If all 14 HDs are arranged in a RAID10 array, I'd say there's definitely something wrong with Jeremiah's hardware. From pgsql-performance-owner@postgresql.org Mon Aug 22 14:26:55 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 96BE9D6D71 for ; Mon, 22 Aug 2005 12:43:01 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 69655-02 for ; Mon, 22 Aug 2005 15:42:58 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 17292D70A5 for ; Mon, 22 Aug 2005 12:42:57 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7MFfeCY016260; Mon, 22 Aug 2005 11:41:40 -0400 (EDT) To: "Jignesh Shah" Cc: pgsql-performance@postgresql.org Subject: Re: MemoryContextSwitchTo during table scan? In-reply-to: References: Comments: In-reply-to "Jignesh Shah" message dated "Mon, 22 Aug 2005 09:47:38 -0400" Date: Mon, 22 Aug 2005 11:41:40 -0400 Message-ID: <16259.1124725300@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/322 X-Sequence-Number: 14069 "Jignesh Shah" writes: > Running a script (available on my blog) I find the following top 5 functions > where it spends most time during a 10 second run of the script It's pretty risky to draw conclusions from only 10 seconds' worth of gprof data --- that's only 1000 samples total at the common sampling rate of 100/sec. If there's one function eating 90% of the runtime, you'll find out, but you don't have enough data to believe that you know what is happening with resolution of a percent or so. I generally try to accumulate several minutes worth of CPU time in a gprof run. > MemoryContextSwitchTo and LockBuffer itself takes 15% of the total time of > the query. I was expecting "read" to be the slowest part (biggest component) > but it was way down in the 0.4% level. You do know that gprof counts only CPU time, and only user-space CPU time at that? read() isn't going to show up at all. It's fairly likely that your test case is I/O bound and that worrying about CPU efficiency for it is a waste of time anyway. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 22 15:19:58 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 96989D701C for ; Mon, 22 Aug 2005 14:18:34 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 77125-01 for ; Mon, 22 Aug 2005 17:18:30 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id 7FAB3D70E7 for ; Mon, 22 Aug 2005 14:18:29 -0300 (ADT) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: Finding bottleneck Date: Mon, 22 Aug 2005 13:18:28 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3417DD167@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Finding bottleneck Thread-Index: AcWnJEqjyZRHA1Q8SpqfQYHxLxmIfAAFyH0g From: "Merlin Moncure" To: "Tom Lane" Cc: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/324 X-Sequence-Number: 14071 > That seems quite peculiar; AFAICS the pgstat code shouldn't be any > slower than before. At first I thought it might be because we'd > increased PGSTAT_ACTIVITY_SIZE, but actually that happened before > 8.0 release, so it shouldn't be a factor in this comparison. Just FYI the last time I looked at stats was in the 8.0 beta period. =20 > Can anyone else confirm a larger penalty for stats_command_string in > HEAD than in 8.0? A self-contained test case would be nice too. looking into it. Merlin From pgsql-performance-owner@postgresql.org Mon Aug 22 15:03:23 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4DBC4D70A9 for ; Mon, 22 Aug 2005 14:46:48 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 15349-05 for ; Mon, 22 Aug 2005 17:46:46 +0000 (GMT) Received: from mail.mi8.com (d01gw04.mi8.com [63.240.6.44]) by svr1.postgresql.org (Postfix) with ESMTP id 2F32DD70A8 for ; Mon, 22 Aug 2005 14:46:44 -0300 (ADT) Received: from 172.16.1.118 by mail.mi8.com with ESMTP (- Welcome to Mi8 Corporation www.Mi8.com (D4)); Mon, 22 Aug 2005 13:46:35 -0400 X-Server-Uuid: C8FB4D43-1108-484A-A898-3CBCC7906230 Received: from MI8NYCMAIL06.Mi8.com ([172.16.1.174]) by d01smtp03.Mi8.com with Microsoft SMTPSVC(6.0.3790.1830); Mon, 22 Aug 2005 13:46:34 -0400 Received: from 67.103.45.218 ([67.103.45.218]) by MI8NYCMAIL06.Mi8.com ( [172.16.1.219]) via Exchange Front-End Server mi8owa.mi8.com ( [172.16.1.104]) with Microsoft Exchange Server HTTP-DAV ; Mon, 22 Aug 2005 13:46:33 -0500 User-Agent: Microsoft-Entourage/11.1.0.040913 Date: Mon, 22 Aug 2005 10:46:33 -0700 Subject: Re: MemoryContextSwitchTo during table scan? From: "Luke Lonergan" To: "Tom Lane" , "Jignesh Shah" Cc: pgsql-performance@postgresql.org Message-ID: In-Reply-To: <16259.1124725300@sss.pgh.pa.us> MIME-Version: 1.0 X-OriginalArrivalTime: 22 Aug 2005 17:46:34.0048 (UTC) FILETIME=[6FCF0C00:01C5A741] X-WSS-ID: 6F14D0F12TG9840475-01-01 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.575 required=5 tests=[AWL=-0.006, FORGED_RCVD_HELO=0.05, RCVD_NUMERIC_HELO=1.531] X-Spam-Level: * X-Archive-Number: 200508/323 X-Sequence-Number: 14070 Tom, On 8/22/05 8:41 AM, "Tom Lane" wrote: >> MemoryContextSwitchTo and LockBuffer itself takes 15% of the total time of >> the query. I was expecting "read" to be the slowest part (biggest component) >> but it was way down in the 0.4% level. > > You do know that gprof counts only CPU time, and only user-space CPU > time at that? read() isn't going to show up at all. It's fairly likely > that your test case is I/O bound and that worrying about CPU efficiency > for it is a waste of time anyway. He's running DTRACE, a CPU profiler that uses hardware performance registers, not gprof. BTW, if you statically link your app, you get profiling information for system calls with gprof. Jignesh has been analyzing PG for quite a while, there are definite issues with CPU consuming functions in the data path IMO. This result he reported is one of them on Solaris 10, and possibly on other platforms. We are limited to about 130MB/s of I/O throughput for sequential scan on platforms that can do 240MB/s of sequential 8k block I/O. Though I haven't profiled seqscan, I'd look at Jignesh's results carefully because they correlate with our experience. - Luke From pgsql-performance-owner@postgresql.org Mon Aug 22 15:58:33 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C4145D7231 for ; Mon, 22 Aug 2005 15:21:46 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 86567-06 for ; Mon, 22 Aug 2005 18:21:42 +0000 (GMT) Received: from calvin.surfutopia.net (calvin.surfutopia.net [67.120.245.34]) by svr1.postgresql.org (Postfix) with ESMTP id EF04FD727F for ; Mon, 22 Aug 2005 15:21:40 -0300 (ADT) Received: by calvin.surfutopia.net (Postfix, from userid 1000) id 83B08F237; Mon, 22 Aug 2005 11:21:38 -0700 (PDT) Date: Mon, 22 Aug 2005 11:21:38 -0700 From: John Mendenhall To: pgsql-performance list Subject: Re: complex query performance assistance request Message-ID: <20050822182138.GA30818@calvin.surfutopia.net> References: <20050821034841.GA28968@calvin.surfutopia.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20050821034841.GA28968@calvin.surfutopia.net> User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/325 X-Sequence-Number: 14072 On Sat, 20 Aug 2005, John Mendenhall wrote: > I need to improve the performance for the following > query. I have run the same query in the same database under different schemas. Each schema is pretty much the same tables and indices. One has an extra backup table and an extra index which are not used in either of the explain analyze plans. The first schema is a development schema, which I used to performance tune the server so everything was great. Here are the current results of the sql run in the development environment: ----- LOG: duration: 852.275 ms statement: explain analyze SELECT c.id AS contact_id, sr.id AS sales_rep_id, p.id AS partner_id, coalesce(LTRIM(RTRIM(c.company)), LTRIM(RTRIM(c.firstname || ' ' || c.lastname))) AS contact_company, co.name AS contact_country, c.master_key_token FROM sales_reps sr JOIN partners p ON (sr.id = p.sales_rep_id) JOIN contacts c ON (p.id = c.partner_id) JOIN countries co ON (LOWER(c.country) = LOWER(co.code)) JOIN partner_classification pc ON (p.classification_id = pc.id AND pc.classification != 'Sales Rep') WHERE c.lead_deleted IS NULL AND EXISTS ( SELECT lr.id FROM lead_requests lr, lead_request_status lrs WHERE c.id = lr.contact_id AND lr.status_id = lrs.id AND lrs.is_closed = 0 ) ORDER BY contact_company, contact_id QUERY PLAN -------------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------- Sort (cost=18238.25..18238.27 rows=11 width=102) (actual time=823.721..823.915 rows=247 loops=1) Sort Key: COALESCE(ltrim(rtrim((c.company)::text)), ltrim(rtrim((((c.firstname)::text || ' '::text) || (c.lastname)::text)))), c.id -> Hash Join (cost=18230.34..18238.06 rows=11 width=102) (actual time=808.042..818.427 rows=247 loops=1) Hash Cond: (lower(("outer".code)::text) = lower(("inner".country)::text)) -> Seq Scan on countries co (cost=0.00..4.42 rows=242 width=19) (actual time=0.032..1.208 rows=242 loops=1) -> Hash (cost=18230.31..18230.31 rows=9 width=95) (actual time=807.554..807.554 rows=0 loops=1) -> Merge Join (cost=18229.98..18230.31 rows=9 width=95) (actual time=794.413..804.855 rows=247 loops=1) Merge Cond: ("outer".sales_rep_id = "inner".id) -> Sort (cost=18227.56..18227.59 rows=9 width=95) (actual time=793.132..793.502 rows=250 loops=1) Sort Key: p.sales_rep_id -> Merge Join (cost=18227.26..18227.42 rows=9 width=95) (actual time=782.832..789.205 rows=250 loops=1) Merge Cond: ("outer".id = "inner".classification_id) -> Sort (cost=1.05..1.05 rows=2 width=10) (actual time=0.189..0.194 rows=2 loops=1) Sort Key: pc.id -> Seq Scan on partner_classification pc (cost=0.00..1.04 rows=2 width=10) (actual time=0.089..0.127 rows=2 loops=1) Filter: ((classification)::text <> 'Sales Rep'::text) -> Sort (cost=18226.21..18226.24 rows=13 width=105) (actual time=782.525..782.818 rows=251 loops=1) Sort Key: p.classification_id -> Merge Join (cost=0.00..18225.97 rows=13 width=105) (actual time=54.135..776.299 rows=449 loops=1) Merge Cond: ("outer".id = "inner".partner_id) -> Index Scan using partners_pkey on partners p (cost=0.00..30.80 rows=395 width=30) (actual time=0.073..6.873 rows=395 loops=1) -> Index Scan using contacts_partner_id_idx on contacts c (cost=0.00..130157.20 rows=93 width=85) (actual time=0.366..739.783 rows=453 loops=1) Filter: ((lead_deleted IS NULL) AND (subplan)) SubPlan -> Nested Loop (cost=0.00..6.75 rows=2 width=10) (actual time=0.103..0.103 rows=0 loops=5576) Join Filter: ("outer".status_id = "inner".id) -> Index Scan using lead_requests_contact_id_idx on lead_requests lr (cost=0.00..4.23 rows=2 width=20) (actual time=0.075..0.075 rows=0 loops=5576) Index Cond: ($0 = contact_id) -> Seq Scan on lead_request_status lrs (cost=0.00..1.16 rows=8 width=10) (actual time=0.028..0.098 rows=4 loops=522) Filter: (is_closed = 0::numeric) -> Sort (cost=2.42..2.52 rows=39 width=10) (actual time=1.183..1.569 rows=268 loops=1) Sort Key: sr.id -> Seq Scan on sales_reps sr (cost=0.00..1.39 rows=39 width=10) (actual time=0.056..0.353 rows=39 loops=1) Total runtime: 826.425 ms (34 rows) ----- Here is the current run in the production environment, which I need to figure out how to get to the performance level of the development environment: ----- LOG: duration: 6447.934 ms statement: explain analyze SELECT c.id AS contact_id, sr.id AS sales_rep_id, p.id AS partner_id, coalesce(LTRIM(RTRIM(c.company)), LTRIM(RTRIM(c.firstname || ' ' || c.lastname))) AS contact_company, co.name AS contact_country, c.master_key_token FROM sales_reps sr JOIN partners p ON (sr.id = p.sales_rep_id) JOIN contacts c ON (p.id = c.partner_id) JOIN countries co ON (LOWER(c.country) = LOWER(co.code)) JOIN partner_classification pc ON (p.classification_id = pc.id AND pc.classification != 'Sales Rep') WHERE c.lead_deleted IS NULL AND EXISTS ( SELECT lr.id FROM lead_requests lr, lead_request_status lrs WHERE c.id = lr.contact_id AND lr.status_id = lrs.id AND lrs.is_closed = 0 ) ORDER BY contact_company, contact_id QUERY PLAN -------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------- Sort (cost=40838.98..40849.08 rows=4042 width=102) (actual time=6418.732..6419.536 rows=1071 loops=1) Sort Key: COALESCE(ltrim(rtrim((c.company)::text)), ltrim(rtrim((((c.firstname)::text || ' '::text) || (c.lastname)::text)))), c.id -> Merge Join (cost=40442.25..40596.85 rows=4042 width=102) (actual time=6357.161..6389.616 rows=1071 loops=1) Merge Cond: ("outer"."?column3?" = "inner"."?column9?") -> Sort (cost=14.00..14.61 rows=242 width=19) (actual time=9.753..10.018 rows=240 loops=1) Sort Key: lower((co.code)::text) -> Seq Scan on countries co (cost=0.00..4.42 rows=242 width=19) (actual time=0.126..3.950 rows=242 loops=1) -> Sort (cost=40428.24..40436.59 rows=3340 width=95) (actual time=6347.154..6348.429 rows=1071 loops=1) Sort Key: lower((c.country)::text) -> Merge Join (cost=75.65..40232.76 rows=3340 width=95) (actual time=60.308..6331.266 rows=1071 loops=1) Merge Cond: ("outer".partner_id = "inner".id) -> Index Scan using contacts_partner_id_idx on contacts c (cost=0.00..161018.18 rows=20120 width=85) (actual time=2.769..6188.886 rows=1548 loops=1) Filter: ((lead_deleted IS NULL) AND (subplan)) SubPlan -> Nested Loop (cost=1.16..6.57 rows=2 width=10) (actual time=0.129..0.129 rows=0 loops=40262) Join Filter: ("outer".status_id = "inner".id) -> Index Scan using lead_requests_contact_id_idx on lead_requests lr (cost=0.00..4.86 rows=3 width=20) (actual time=0.086..0.092 rows=0 loops=40262) Index Cond: ($0 = contact_id) -> Materialize (cost=1.16..1.24 rows=8 width=10) (actual time=0.002..0.013 rows=6 loops=12593) -> Seq Scan on lead_request_status lrs (cost=0.00..1.16 rows=8 width=10) (actual time=0.078..0.243 rows=7 loops=1) Filter: (is_closed = 0::numeric) -> Sort (cost=75.65..76.37 rows=290 width=20) (actual time=57.243..59.574 rows=1334 loops=1) Sort Key: p.id -> Merge Join (cost=59.24..63.79 rows=290 width=20) (actual time=33.975..42.215 rows=395 loops=1) Merge Cond: ("outer".id = "inner".sales_rep_id) -> Sort (cost=2.42..2.52 rows=39 width=10) (actual time=1.206..1.285 rows=39 loops=1) Sort Key: sr.id -> Seq Scan on sales_reps sr (cost=0.00..1.39 rows=39 width=10) (actual time=0.028..0.365 rows=39 loops=1) -> Sort (cost=56.82..57.55 rows=290 width=20) (actual time=32.566..33.254 rows=395 loops=1) Sort Key: p.sales_rep_id -> Nested Loop (cost=24.35..44.96 rows=290 width=20) (actual time=0.158..25.227 rows=395 loops=1) Join Filter: ("inner".classification_id = "outer".id) -> Seq Scan on partner_classification pc (cost=0.00..1.04 rows=2 width=10) (actual time=0.050..0.096 rows=2 loops=1) Filter: ((classification)::text <> 'Sales Rep'::text) -> Materialize (cost=24.35..28.70 rows=435 width=30) (actual time=0.028..6.617 rows=435 loops=2) -> Seq Scan on partners p (cost=0.00..24.35 rows=435 width=30) (actual time=0.042..9.941 rows=435 loops=1) Total runtime: 6423.683 ms (37 rows) ----- The SQL is exactly the same. The issue is the query plan is different, and thus, not up to the performance we need. We have 256meg in the machine. Would it help if we threw some more memory in? Please let me know if you have *any* pointers as to the reason for the difference. Thank you very much in advance for any pointers or suggestions. JohnM -- John Mendenhall john@surfutopia.net surf utopia internet services From pgsql-performance-owner@postgresql.org Mon Aug 22 16:11:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AE64DD7253 for ; Mon, 22 Aug 2005 16:09:51 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 47884-09 for ; Mon, 22 Aug 2005 19:09:46 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id 0C922D70A8 for ; Mon, 22 Aug 2005 16:09:45 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [67.103.45.218] (account josh@agliodbs.com HELO [10.0.0.38]) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7907752; Mon, 22 Aug 2005 12:12:02 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: MemoryContextSwitchTo during table scan? Date: Mon, 22 Aug 2005 12:12:56 -0700 User-Agent: KMail/1.8 Cc: "Jignesh Shah" References: In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200508221212.57047.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[AWL=0.000] X-Spam-Level: X-Archive-Number: 200508/326 X-Sequence-Number: 14073 Jignesh, > Also is there any way to optimize LockBuffer? Yes, test on 8.1. The buffer manager was re-written for 8.1. You should see a decrease in both LockBuffer and context switch activity. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Mon Aug 22 18:13:03 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D0F62D7021 for ; Mon, 22 Aug 2005 17:15:22 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 97803-06 for ; Mon, 22 Aug 2005 20:15:18 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id C21C3D7761 for ; Mon, 22 Aug 2005 17:15:17 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7MKFICA020831; Mon, 22 Aug 2005 16:15:18 -0400 (EDT) To: John Mendenhall Cc: pgsql-performance list Subject: Re: complex query performance assistance request In-reply-to: <20050822182138.GA30818@calvin.surfutopia.net> References: <20050821034841.GA28968@calvin.surfutopia.net> <20050822182138.GA30818@calvin.surfutopia.net> Comments: In-reply-to John Mendenhall message dated "Mon, 22 Aug 2005 11:21:38 -0700" Date: Mon, 22 Aug 2005 16:15:18 -0400 Message-ID: <20830.1124741718@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/327 X-Sequence-Number: 14074 John Mendenhall writes: > The issue is the query plan is different, and thus, > not up to the performance we need. No, the issue is that you've got eight times as much data in the production server; so it's hardly surprising that it takes about eight times longer. The production query is spending most of its time on the subplan attached to the contacts table: > -> Index Scan using contacts_partner_id_idx on contacts c (cost=0.00..161018.18 rows=20120 width=85) (actual time=2.769..6188.886 rows=1548 loops=1) > Filter: ((lead_deleted IS NULL) AND (subplan)) > SubPlan > -> Nested Loop (cost=1.16..6.57 rows=2 width=10) (actual time=0.129..0.129 rows=0 loops=40262) 0.129 * 40262 = 5193.798, so about five seconds in the subplan and another one second in the indexscan proper. The problem is that the subplan (the EXISTS clause) is iterated for each of 40262 rows of contacts --- basically, every contacts row that has null lead_deleted. On the dev server the same scan shows these numbers: > -> Index Scan using contacts_partner_id_idx on contacts c (cost=0.00..130157.20 rows=93 width=85) (actual time=0.366..739.783 rows=453 loops=1) > Filter: ((lead_deleted IS NULL) AND (subplan)) > SubPlan > -> Nested Loop (cost=0.00..6.75 rows=2 width=10) (actual time=0.103..0.103 rows=0 loops=5576) Here the subplan is iterated only 5576 times for 574 total msec. It's still the bulk of the runtime though; the fact that the upper levels of the plan are a bit different has got little to do with where the time is going. I'd suggest trying to get rid of the EXISTS clause --- can you refactor that into something that joins at the top query level? Or, if this is 7.4 or later (and you should ALWAYS mention which version you are using in a performance question, because it matters), try to convert the EXISTS into an IN. "x IN (subselect)" is planned much better than "EXISTS(subselect-using-x)" these days. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 22 18:55:41 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 69F91D7025 for ; Mon, 22 Aug 2005 18:08:00 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 23027-05 for ; Mon, 22 Aug 2005 21:07:56 +0000 (GMT) Received: from calvin.surfutopia.net (calvin.surfutopia.net [67.120.245.34]) by svr1.postgresql.org (Postfix) with ESMTP id 458ABD7018 for ; Mon, 22 Aug 2005 18:07:54 -0300 (ADT) Received: by calvin.surfutopia.net (Postfix, from userid 1000) id 71C7CF23B; Mon, 22 Aug 2005 14:07:51 -0700 (PDT) Date: Mon, 22 Aug 2005 14:07:51 -0700 From: John Mendenhall To: Tom Lane Cc: pgsql-performance list Subject: Re: complex query performance assistance request Message-ID: <20050822210751.GA32479@calvin.surfutopia.net> References: <20050821034841.GA28968@calvin.surfutopia.net> <20050822182138.GA30818@calvin.surfutopia.net> <20830.1124741718@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20830.1124741718@sss.pgh.pa.us> User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/328 X-Sequence-Number: 14075 Tom, > No, the issue is that you've got eight times as much data in the > production server; so it's hardly surprising that it takes about > eight times longer. > > The production query is spending most of its time on the subplan > attached to the contacts table: > > > -> Index Scan using contacts_partner_id_idx on contacts c (cost=0.00..161018.18 rows=20120 width=85) (actual time=2.769..6188.886 rows=1548 loops=1) > > Filter: ((lead_deleted IS NULL) AND (subplan)) > > SubPlan > > -> Nested Loop (cost=1.16..6.57 rows=2 width=10) (actual time=0.129..0.129 rows=0 loops=40262) > > 0.129 * 40262 = 5193.798, so about five seconds in the subplan and > another one second in the indexscan proper. The problem is that the > subplan (the EXISTS clause) is iterated for each of 40262 rows of > contacts --- basically, every contacts row that has null lead_deleted. > > On the dev server the same scan shows these numbers: > > > -> Index Scan using contacts_partner_id_idx on contacts c (cost=0.00..130157.20 rows=93 width=85) (actual time=0.366..739.783 rows=453 loops=1) > > Filter: ((lead_deleted IS NULL) AND (subplan)) > > SubPlan > > -> Nested Loop (cost=0.00..6.75 rows=2 width=10) (actual time=0.103..0.103 rows=0 loops=5576) > > I'd suggest trying to get rid of the EXISTS clause --- can you refactor > that into something that joins at the top query level? > > Or, if this is 7.4 or later (and you should ALWAYS mention which version > you are using in a performance question, because it matters), try to > convert the EXISTS into an IN. "x IN (subselect)" is planned much better > than "EXISTS(subselect-using-x)" these days. We are using version 7.4.6. The number of contacts in the dev env is 37080. The number of contacts in the production env is 40307. The amount of data is statistically about the same. However, the number of lead_requests are much different. The dev env has 1438 lead_requests, the production env has 15554 lead_requests. Each contacts row can have multiple lead_requests, each lead_requests entry can have an open or closed status. We are trying to select the contacts with an open lead_request. Would it be best to attempt to rewrite it for IN? Or, should we try to tie it in with a join? I would probably need to GROUP so I can just get a count of those contacts with open lead_requests. Unless you know of a better way? Thanks for your assistance. This is helping a lot. BTW, what does the Materialize query plan element mean? Thanks again. JohnM -- John Mendenhall john@surfutopia.net surf utopia internet services From pgsql-performance-owner@postgresql.org Mon Aug 22 19:37:36 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BCE7FD6FEC for ; Mon, 22 Aug 2005 19:37:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 70495-07 for ; Mon, 22 Aug 2005 22:37:24 +0000 (GMT) Received: from hotmail.com (bay101-f22.bay101.hotmail.com [64.4.56.32]) by svr1.postgresql.org (Postfix) with ESMTP id 7E912D6D78 for ; Mon, 22 Aug 2005 19:37:22 -0300 (ADT) Received: from mail pickup service by hotmail.com with Microsoft SMTPSVC; Mon, 22 Aug 2005 15:37:21 -0700 Message-ID: Received: from 64.4.56.205 by by101fd.bay101.hotmail.msn.com with HTTP; Mon, 22 Aug 2005 22:37:21 GMT X-Originating-IP: [64.4.56.205] X-Originating-Email: [jigneshk@hotmail.com] X-Sender: jigneshk@hotmail.com In-Reply-To: <16259.1124725300@sss.pgh.pa.us> From: "Jignesh Shah" To: tgl@sss.pgh.pa.us Cc: pgsql-performance@postgresql.org Subject: Re: MemoryContextSwitchTo during table scan? Date: Mon, 22 Aug 2005 18:37:21 -0400 Mime-Version: 1.0 Content-Type: text/plain; format=flowed X-OriginalArrivalTime: 22 Aug 2005 22:37:21.0434 (UTC) FILETIME=[0F42DFA0:01C5A76A] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=2.228 required=5 tests=[AWL=0.478, DNS_FROM_RFC_ABUSE=0.374, DNS_FROM_RFC_POST=1.376, MSGID_FROM_MTA_HEADER=0] X-Spam-Level: ** X-Archive-Number: 200508/329 X-Sequence-Number: 14076 Hi Tom, Like I mentioned I am using DTrace on Solaris 10 x64 and not gprof. DTrace is not based on sampling but actual entry/exit point. Ofcourse my 10 second profile is just a sample that I can assure you is representative of the query since it is a very simple query that does simple table scan. (I am taken profiles at different times of the queries all giving similar outputs) In case of DTrace I am actually measuring "wall clock" for leaf functions. For more information on DTrace please refer to: http://docs.sun.com/app/docs/doc/817-6223/6mlkidlf1?a=view Regards, Jignesh ----Original Message Follows---- From: Tom Lane To: "Jignesh Shah" CC: pgsql-performance@postgresql.org Subject: Re: [PERFORM] MemoryContextSwitchTo during table scan? Date: Mon, 22 Aug 2005 11:41:40 -0400 "Jignesh Shah" writes: > Running a script (available on my blog) I find the following top 5 functions > where it spends most time during a 10 second run of the script It's pretty risky to draw conclusions from only 10 seconds' worth of gprof data --- that's only 1000 samples total at the common sampling rate of 100/sec. If there's one function eating 90% of the runtime, you'll find out, but you don't have enough data to believe that you know what is happening with resolution of a percent or so. I generally try to accumulate several minutes worth of CPU time in a gprof run. > MemoryContextSwitchTo and LockBuffer itself takes 15% of the total time of > the query. I was expecting "read" to be the slowest part (biggest component) > but it was way down in the 0.4% level. You do know that gprof counts only CPU time, and only user-space CPU time at that? read() isn't going to show up at all. It's fairly likely that your test case is I/O bound and that worrying about CPU efficiency for it is a waste of time anyway. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 22 21:10:49 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E749AD6D78 for ; Mon, 22 Aug 2005 19:48:24 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78049-09 for ; Mon, 22 Aug 2005 22:48:21 +0000 (GMT) Received: from mailbox.samurai.com (mailbox.samurai.com [205.207.28.82]) by svr1.postgresql.org (Postfix) with ESMTP id 55F7BD70B4 for ; Mon, 22 Aug 2005 19:48:19 -0300 (ADT) Received: from localhost (mailbox.samurai.com [205.207.28.82]) by mailbox.samurai.com (Postfix) with ESMTP id 592332395A6; Mon, 22 Aug 2005 18:48:22 -0400 (EDT) Received: from mailbox.samurai.com ([205.207.28.82]) by localhost (mailbox.samurai.com [205.207.28.82]) (amavisd-new, port 10024) with LMTP id 50302-01-8; Mon, 22 Aug 2005 18:48:20 -0400 (EDT) Received: from [192.168.0.105] (CPE3285f62f0d42-CM0011aec5ebbc.cpe.net.cable.rogers.com [69.199.116.160]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (No client certificate requested) by mailbox.samurai.com (Postfix) with ESMTP id B24AC2395B6; Mon, 22 Aug 2005 18:48:20 -0400 (EDT) Message-ID: <430A5632.10701@samurai.com> Date: Mon, 22 Aug 2005 18:48:18 -0400 From: Neil Conway User-Agent: Mozilla Thunderbird 1.0.2 (Macintosh/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Jignesh Shah Cc: pgsql-performance@postgresql.org Subject: Re: MemoryContextSwitchTo during table scan? References: In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at mailbox.samurai.com X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/330 X-Sequence-Number: 14077 Jignesh Shah wrote: > Now the question is why there are so many calls to MemoryContextSwitchTo > in a single SELECT query command? Can it be minimized? I agree with Tom -- if profiling indicates that MemoryContextSwitchTo() is the bottleneck, I would be suspicious that your profiling setup is misconfigured. MemoryContextSwitchTo() is essentially a function call, two pointer assignments, and a function return. Try rerunning the test with current sources -- MemoryContextSwitchTo() is now inlined when using GCC, which should just leave the assignments. -Neil From pgsql-performance-owner@postgresql.org Mon Aug 22 21:43:38 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0BCF2D7245 for ; Mon, 22 Aug 2005 21:43:34 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 56040-07 for ; Tue, 23 Aug 2005 00:43:33 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [67.100.216.10]) by svr1.postgresql.org (Postfix) with ESMTP id 87111D7201 for ; Mon, 22 Aug 2005 21:43:32 -0300 (ADT) Received: by flake.decibel.org (Postfix, from userid 1001) id 09AE91529B; Mon, 22 Aug 2005 19:43:35 -0500 (CDT) Date: Mon, 22 Aug 2005 19:43:35 -0500 From: "Jim C. Nasby" To: Sebastian Hennebrueder Cc: postgres performance Subject: Re: Looking for a large database for testing Message-ID: <20050823004335.GG17203@pervasive.com> References: <430195DC.70104@laliluna.de> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <430195DC.70104@laliluna.de> X-Operating-System: FreeBSD 4.11-RELEASE-p10 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/331 X-Sequence-Number: 14078 On Tue, Aug 16, 2005 at 09:29:32AM +0200, Sebastian Hennebrueder wrote: > Hello, > > I would like to test the performance of my Java/PostgreSQL applications > especially when making full text searches. > For this I am looking for a database with 50 to 300 MB having text fields. > e.g. A table with books with fields holding a comment, table of content > or example chapters > or what ever else. > > Does anybody have an idea where I can find a database like this or does > even have something like this? Most benchmarks (such as dbt* and pgbench) have data generators you could use. -- Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com Pervasive Software http://pervasive.com 512-569-9461 From pgsql-performance-owner@postgresql.org Mon Aug 22 21:48:59 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 820EBD7289 for ; Mon, 22 Aug 2005 21:48:56 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 72103-03 for ; Tue, 23 Aug 2005 00:48:54 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [67.100.216.10]) by svr1.postgresql.org (Postfix) with ESMTP id 70F57D6D78 for ; Mon, 22 Aug 2005 21:48:52 -0300 (ADT) Received: by flake.decibel.org (Postfix, from userid 1001) id 1242D152AE; Mon, 22 Aug 2005 19:48:56 -0500 (CDT) Date: Mon, 22 Aug 2005 19:48:56 -0500 From: "Jim C. Nasby" To: Alex Turner Cc: Ulrich Wisser , pgsql-performance@postgresql.org Subject: Re: Need for speed Message-ID: <20050823004856.GH17203@pervasive.com> References: <430208AE.7080100@relevanttraffic.se> <33c6269f050816105957ae5d6f@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <33c6269f050816105957ae5d6f@mail.gmail.com> X-Operating-System: FreeBSD 4.11-RELEASE-p10 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/332 X-Sequence-Number: 14079 RRS (http://rrs.decibel.org) might be of use in this case. On Tue, Aug 16, 2005 at 01:59:53PM -0400, Alex Turner wrote: > Are you calculating aggregates, and if so, how are you doing it (I ask > the question from experience of a similar application where I found > that my aggregating PGPLSQL triggers were bogging the system down, and > changed them so scheduled jobs instead). > > Alex Turner > NetEconomist > > On 8/16/05, Ulrich Wisser wrote: > > Hello, > > > > one of our services is click counting for on line advertising. We do > > this by importing Apache log files every five minutes. This results in a > > lot of insert and delete statements. At the same time our customers > > shall be able to do on line reporting. > > > > We have a box with > > Linux Fedora Core 3, Postgres 7.4.2 > > Intel(R) Pentium(R) 4 CPU 2.40GHz > > 2 scsi 76GB disks (15.000RPM, 2ms) > > > > I did put pg_xlog on another file system on other discs. > > > > Still when several users are on line the reporting gets very slow. > > Queries can take more then 2 min. > > > > I need some ideas how to improve performance in some orders of > > magnitude. I already thought of a box with the whole database on a ram > > disc. So really any idea is welcome. > > > > Ulrich > > > > > > > > -- > > Ulrich Wisser / System Developer > > > > RELEVANT TRAFFIC SWEDEN AB, Riddarg 17A, SE-114 57 Sthlm, Sweden > > Direct (+46)86789755 || Cell (+46)704467893 || Fax (+46)86789769 > > ________________________________________________________________ > > http://www.relevanttraffic.com > > > > ---------------------------(end of broadcast)--------------------------- > > TIP 1: if posting/reading through Usenet, please send an appropriate > > subscribe-nomail command to majordomo@postgresql.org so that your > > message can get through to the mailing list cleanly > > > > ---------------------------(end of broadcast)--------------------------- > TIP 5: don't forget to increase your free space map settings > -- Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com Pervasive Software http://pervasive.com 512-569-9461 From pgsql-performance-owner@postgresql.org Mon Aug 22 21:54:32 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2ED9ED6D8C for ; Mon, 22 Aug 2005 21:54:29 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 64584-09 for ; Tue, 23 Aug 2005 00:54:28 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 3B71CD6D78 for ; Mon, 22 Aug 2005 21:54:26 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7N0sUlM026870; Mon, 22 Aug 2005 20:54:31 -0400 (EDT) To: John Mendenhall Cc: pgsql-performance list Subject: Re: complex query performance assistance request In-reply-to: <20050822210751.GA32479@calvin.surfutopia.net> References: <20050821034841.GA28968@calvin.surfutopia.net> <20050822182138.GA30818@calvin.surfutopia.net> <20830.1124741718@sss.pgh.pa.us> <20050822210751.GA32479@calvin.surfutopia.net> Comments: In-reply-to John Mendenhall message dated "Mon, 22 Aug 2005 14:07:51 -0700" Date: Mon, 22 Aug 2005 20:54:30 -0400 Message-ID: <26869.1124758470@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/333 X-Sequence-Number: 14080 John Mendenhall writes: > Would it be best to attempt to rewrite it for IN? > Or, should we try to tie it in with a join? Couldn't say without a deeper understanding of what you're trying to accomplish. > BTW, what does the Materialize query plan element mean? Means "run the contained subplan once, and save the results aside in a buffer; on subsequent loops, just pass back the buffer contents instead of re-running the subplan". regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 22 22:19:29 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E8996D7201 for ; Mon, 22 Aug 2005 22:19:26 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 12645-09 for ; Tue, 23 Aug 2005 01:19:25 +0000 (GMT) Received: from smtpsig-4.ig.com.br (smtpsig-4.ig.com.br [200.226.132.141]) by svr1.postgresql.org (Postfix) with ESMTP id 4E9D1D6E31 for ; Mon, 22 Aug 2005 22:19:23 -0300 (ADT) Received: (qmail 17131 invoked from network); 23 Aug 2005 01:19:23 -0000 Received: from 200164168141.user.veloxzone.com.br (HELO [192.168.1.223]) (Usuario_autenticado:netsuporte@superig.com.br@[200.164.168.141]) (envelope-sender ) by smtpsig-4.ig.com.br (qmail-ldap-1.03) with SMTP for ; 23 Aug 2005 01:19:23 -0000 Message-ID: <430A78B2.8080803@superig.com.br> Date: Mon, 22 Aug 2005 22:15:30 -0300 From: Philip Pinkerton User-Agent: Debian Thunderbird 1.0.6 (X11/20050802) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance list Subject: pgbench X-Enigmail-Version: 0.86.1.0 X-Enigmail-Supports: pgp-inline, pgp-mime Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/334 X-Sequence-Number: 14081 I am looking for the latest pgbench and documentation. If someone know where I can locate them it would save a lot of search time. Thanks Philip Pinkerton TPC-C Benchmarks Sybase Independant Consultant Rio de Janeiro, RJ, Brazil 22031-010 From pgsql-performance-owner@postgresql.org Tue Aug 23 00:13:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BDA42D7766 for ; Tue, 23 Aug 2005 00:13:19 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 28422-02 for ; Tue, 23 Aug 2005 03:13:17 +0000 (GMT) Received: from unicorn.rentec.com (unicorn.rentec.com [216.223.240.9]) by svr1.postgresql.org (Postfix) with ESMTP id 55534D7764 for ; Tue, 23 Aug 2005 00:13:15 -0300 (ADT) Received: from wren.rentec.com (wren.rentec.com [192.5.35.106]) by unicorn.rentec.com (8.13.1/8.12.1) with ESMTP id j7N3CWWr015715 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO) for ; Mon, 22 Aug 2005 23:12:34 -0400 (EDT) X-Rentec: external Received: from [172.16.160.108] (stange-dhcp2.rentec.com [172.16.160.108]) by wren.rentec.com (8.13.1/8.12.1) with ESMTP id j7N3CV7g023672 for ; Mon, 22 Aug 2005 23:12:31 -0400 (EDT) Message-ID: <430A8F30.7050706@rentec.com> Date: Mon, 22 Aug 2005 22:51:28 -0400 From: Alan Stange Organization: Renaissance Technologies Corp. User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: unused item pointers? Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Logged: Logged by unicorn.rentec.com as j7N3CWWr015715 at Mon Aug 22 23:12:34 2005 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/335 X-Sequence-Number: 14082 Hello all, what are unused item pointers and how do I get rid of them? We have a fairly large table which is vacuumed daily and reindexed every weekend. NFO: vacuuming "public.tbltimeseries" INFO: index "idx_timeseries" now contains 26165807 row versions in 151713 pages DETAIL: 8610108 index row versions were removed. 58576 index pages have been deleted, 36223 are currently reusable. CPU 6.36s/18.46u sec elapsed 263.75 sec. INFO: "tbltimeseries": removed 8610108 row versions in 500766 pages DETAIL: CPU 37.07s/29.76u sec elapsed 826.82 sec. INFO: "tbltimeseries": found 8610108 removable, 26165807 nonremovable row versions in 5744789 pages DETAIL: 0 dead row versions cannot be removed yet. There were 235555635 unused item pointers. 0 pages are entirely empty. CPU 119.13s/61.09u sec elapsed 2854.22 sec. INFO: vacuuming "pg_toast.pg_toast_2361976783" INFO: index "pg_toast_2361976783_index" now contains 24749150 row versions in 108975 pages DETAIL: 5857243 index row versions were removed. 33592 index pages have been deleted, 16007 are currently reusable. CPU 4.15s/13.53u sec elapsed 78.56 sec. INFO: "pg_toast_2361976783": removed 5857243 row versions in 1125801 pages DETAIL: CPU 82.62s/69.48u sec elapsed 1571.43 sec. INFO: "pg_toast_2361976783": found 5857243 removable, 24749150 nonremovable row versions in 10791766 pages DETAIL: 0 dead row versions cannot be removed yet. There were 33395357 unused item pointers. 0 pages are entirely empty. CPU 235.46s/105.91u sec elapsed 4458.31 sec. INFO: "pg_toast_2361976783": truncated 10791766 to 10778290 pages DETAIL: CPU 0.21s/0.07u sec elapsed 7.09 sec. INFO: analyzing "public.tbltimeseries" INFO: "tbltimeseries": scanned 150000 of 5744789 pages, containing 691250 live rows and 0 dead rows; 150000 rows in sample, 26473903 estimated total rows as you can see we have 235M unused item pointers in the main table and a few 10's of millions more in other associated tables. Please note that the advice "vacuum more often" is a non-starter as the total time here is already about 3 hours and this is just one table. This is a fairly active table to which about 20M rows are added and removed daily. The free space map is set at 11M pages and just today we popped up over that amount in the vacuum output. I don't think this is an issue here though as the large number of unused item pointers has been present for a while. Thanks! -- Alan From pgsql-performance-owner@postgresql.org Sat Aug 27 02:04:09 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C1715D7240 for ; Tue, 23 Aug 2005 00:41:46 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 59464-06 for ; Tue, 23 Aug 2005 03:41:44 +0000 (GMT) Received: from tierw.net.avaya.com (tierw.net.avaya.com [198.152.13.100]) by svr1.postgresql.org (Postfix) with ESMTP id 90C3CD7210 for ; Tue, 23 Aug 2005 00:41:43 -0300 (ADT) Received: from tierw.net.avaya.com (localhost [127.0.0.1]) by tierw.net.avaya.com (Switch-3.1.2/Switch-3.1.0) with ESMTP id j7N3Sivl003840 for ; Mon, 22 Aug 2005 23:28:44 -0400 (EDT) Received: from au3010avexu1.global.avaya.com (h135-27-64-251.avaya.com [135.27.64.251]) by tierw.net.avaya.com (Switch-3.1.2/Switch-3.1.0) with ESMTP id j7N3Scvl003625 for ; Mon, 22 Aug 2005 23:28:39 -0400 (EDT) content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----_=_NextPart_001_01C5A794.8DEF9C8B" Subject: Need indexes on empty tables for good performance ? X-MimeOLE: Produced By Microsoft Exchange V6.0.6603.0 Date: Tue, 23 Aug 2005 13:41:32 +1000 Message-ID: <2773CAC687FD5F4689F526998C7E4E5F43F392@au3010avexu1.global.avaya.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: Need indexes on empty tables for good performance ? Thread-Index: AcWnlI2d7/66JAMCQSi8he+aZQ5/jg== From: "Lenard, Rohan (Rohan)" To: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.014 required=5 tests=[AWL=-0.014, HTML_60_70=0.027, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/470 X-Sequence-Number: 14217 This is a multi-part message in MIME format. ------_=_NextPart_001_01C5A794.8DEF9C8B Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable I've read that indexes aren't used for COUNT(*) and I've noticed (7.3.x) with EXPLAIN that indexes never seem to be used on empty tables - is there any reason to have indexes on empty tables, or will postgresql never use them. =20 This is not as silly as it sounds - with table inheritance you might have table children with the data and a parent that is empty. It'd be nice to make sure postgresql knows to never really look at the parent - especially is you don't know the names of all the children .. =20 Thoughts ? =20 thx, Rohan ------_=_NextPart_001_01C5A794.8DEF9C8B Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable
I've = read that=20 indexes aren't used for COUNT(*) and I've noticed (7.3.x) with EXPLAIN = that=20 indexes never seem to be used on empty tables - is there any reason to = have=20 indexes on empty tables, or will postgresql never use = them.
 
This = is not as silly=20 as it sounds - with table inheritance you might have table children with = the=20 data and a parent that is empty.  It'd be nice to make sure = postgresql=20 knows to never really look at the parent - especially is you don't know = the=20 names of all the children ..
 
Thoughts=20 ?
 
thx,
  = Rohan
------_=_NextPart_001_01C5A794.8DEF9C8B-- From pgsql-performance-owner@postgresql.org Tue Aug 23 01:42:37 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 86746D6D8F for ; Tue, 23 Aug 2005 01:42:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 66035-02 for ; Tue, 23 Aug 2005 04:42:14 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 08375D6D78 for ; Tue, 23 Aug 2005 01:42:09 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7N4g9RQ028372; Tue, 23 Aug 2005 00:42:09 -0400 (EDT) To: Alan Stange Cc: pgsql-performance@postgresql.org Subject: Re: unused item pointers? In-reply-to: <430A8F30.7050706@rentec.com> References: <430A8F30.7050706@rentec.com> Comments: In-reply-to Alan Stange message dated "Mon, 22 Aug 2005 22:51:28 -0400" Date: Tue, 23 Aug 2005 00:42:09 -0400 Message-ID: <28371.1124772129@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/336 X-Sequence-Number: 14083 Alan Stange writes: > INFO: "tbltimeseries": found 8610108 removable, 26165807 nonremovable > row versions in 5744789 pages > DETAIL: 0 dead row versions cannot be removed yet. > There were 235555635 unused item pointers. > 0 pages are entirely empty. The item pointers themselves are not very interesting --- at 4 bytes apiece, they're only accounting for 2% of the table space. However the fact that there are nearly 10x more unused than used ones suggests that this table is suffering a pretty serious bloat problem. Assuming constant-width rows in the table, that implies something like 90% of the space in the table is unused. (contrib/pgstattuple might be useful to confirm this estimate.) VACUUM FULL, or perhaps better CLUSTER, would get you out of that. And yes, you will need to vacuum more often afterwards if you want to keep the bloat under control. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 23 02:00:53 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8C663D7791 for ; Tue, 23 Aug 2005 02:00:46 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 73734-07 for ; Tue, 23 Aug 2005 05:00:36 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id 3DFA8D77D4 for ; Tue, 23 Aug 2005 02:00:34 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [63.195.55.98] (account josh@agliodbs.com HELO spooky) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7912501; Mon, 22 Aug 2005 22:02:50 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: pgbench Date: Mon, 22 Aug 2005 22:01:37 -0700 User-Agent: KMail/1.8 Cc: Philip Pinkerton References: <430A78B2.8080803@superig.com.br> In-Reply-To: <430A78B2.8080803@superig.com.br> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200508222201.38131.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.012 required=5 tests=[AWL=0.012] X-Spam-Level: X-Archive-Number: 200508/337 X-Sequence-Number: 14084 Phillip, > I am looking for the latest pgbench and documentation. Currently they are packaged with the PostgreSQL source code. However, if you're looking for a serious benchmark, may I suggest OSDL's DBT2? It's substantially similar to TPC-C. http://sourceforge.net/projects/osdldbt What's your interest in benchmarking PostgreSQL, BTW? -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Tue Aug 23 02:14:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 622CFD77FC for ; Tue, 23 Aug 2005 02:14:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 01445-07 for ; Tue, 23 Aug 2005 05:14:20 +0000 (GMT) Received: from mail27.sea5.speakeasy.net (mail27.sea5.speakeasy.net [69.17.117.29]) by svr1.postgresql.org (Postfix) with ESMTP id 714B2D77D9 for ; Tue, 23 Aug 2005 02:14:19 -0300 (ADT) Received: (qmail 18387 invoked from network); 23 Aug 2005 05:14:18 -0000 Received: from dsl081-060-184.sfo1.dsl.speakeasy.net (HELO noodles) ([64.81.60.184]) (envelope-sender ) by mail27.sea5.speakeasy.net (qmail-ldap-1.03) with RC4-MD5 encrypted SMTP for ; 23 Aug 2005 05:14:18 -0000 Subject: Re: unused item pointers? From: "Jeffrey W. Baker" To: Alan Stange Cc: pgsql-performance@postgresql.org In-Reply-To: <430A8F30.7050706@rentec.com> References: <430A8F30.7050706@rentec.com> Content-Type: text/plain Date: Mon, 22 Aug 2005 22:15:08 -0700 Message-Id: <1124774108.6978.7.camel@noodles> Mime-Version: 1.0 X-Mailer: Evolution 2.2.1.1 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/338 X-Sequence-Number: 14085 On Mon, 2005-08-22 at 22:51 -0400, Alan Stange wrote: > Hello all, > > what are unused item pointers and how do I get rid of them? > > We have a fairly large table which is vacuumed daily and reindexed every > weekend. > as you can see we have 235M unused item pointers in the main table and a > few 10's of millions more in other associated tables. > > Please note that the advice "vacuum more often" is a non-starter as the > total time here is already about 3 hours and this is just one table. > This is a fairly active table to which about 20M rows are added and > removed daily. That may be so, but the answer is still to VACUUM more often. Try the autovacuum. If it takes 3 hours with 90% wasted records, it would only take 20 minutes when running properly. You might be able to change your application to avoid generating so many dead rows. For example, test before insert so you don't make a dead tuple on duplicate insert. To repair this table, you can try VACUUM FULL but this is likely to take longer than you find reasonable. I would recommend dump and reload. -jwb From pgsql-performance-owner@postgresql.org Tue Aug 23 06:48:36 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 265D1D6EEF for ; Tue, 23 Aug 2005 06:48:35 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 88320-03 for ; Tue, 23 Aug 2005 09:48:33 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id AC976D6ED2 for ; Tue, 23 Aug 2005 06:48:30 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050823094829m9100g5ta8e>; Tue, 23 Aug 2005 09:48:29 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 4844C55FA6; Tue, 23 Aug 2005 04:48:29 -0500 (CDT) Received: from [10.10.0.94] (host35-174.pool82188.interbusiness.it [82.188.174.35]) by juju.arbash-meinel.com (Postfix) with ESMTP id D6EB055F9C; Tue, 23 Aug 2005 04:48:22 -0500 (CDT) Message-ID: <430AF0E4.3050205@arbash-meinel.com> Date: Tue, 23 Aug 2005 04:48:20 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Jeremiah Jahn Cc: Ron , postgres performance Subject: Re: extremly low memory usage References: <1124331108.27881.96.camel@bluejay.goodinassociates.com> <1124386761.27881.119.camel@bluejay.goodinassociates.com> <4304CB77.4070403@arbash-meinel.com> <1124470109.27881.152.camel@bluejay.goodinassociates.com> <43061472.2080201@arbash-meinel.com> <1124481386.27881.167.camel@bluejay.goodinassociates.com> <43064908.7010505@arbash-meinel.com> <1124564003.27881.229.camel@bluejay.goodinassociates.com> <6.2.3.4.0.20050820154422.06021150@pop.earthlink.net> <4307E7A4.7090300@arbash-meinel.com> <1124636066.27881.238.camel@bluejay.goodinassociates.com> <6.2.3.4.0.20050821112849.01f79860@pop.earthlink.net> <1124721763.27881.250.camel@bluejay.goodinassociates.com> In-Reply-To: <1124721763.27881.250.camel@bluejay.goodinassociates.com> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig70D3342FFE24F9008453FCB0" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 required=5 tests=[FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/339 X-Sequence-Number: 14086 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig70D3342FFE24F9008453FCB0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Jeremiah Jahn wrote: > On Sun, 2005-08-21 at 16:13 -0400, Ron wrote: > >>At 10:54 AM 8/21/2005, Jeremiah Jahn wrote: >> ... >>So you have 2 controllers each with 2 external slots? But you are >>currently only using 1 controller and only one external slot on that >>controller? > > > Sorry, no. I have one dual channel card in the system and two > controllers on the array. Dell PowerVault 220S w/ PERC4eDC-PCI Express > > ... >>Regardless of the above, each of these controllers should still be >>good for about 80-85% of 640MB/s, or ~510-540 MB/s apiece when doing >>raw sequential IO if you plug 3-4 fast enough HD's into each SCSI >>channel. Cheetah 15K.4's certainly are fast enough. Optimal setup >>is probably to split each RAID 1 pair so that one HD is on each of >>the SCSI channels, and then RAID 0 those pairs. That will also >>protect you from losing the entire disk subsystem if one of the SCSI >>channels dies. > > I like this idea, but how exactly does one bond the two channels > together? Won't this cause me to have both an /dev/sdb and an /dev/sdc? > Well, even if you did, you could always either use software raid, or lvm to turn it into a single volume. It also depends what the controller card bios would let you get away with. Some cards would let you setup 4 RAID1's (one drive from each channel), and then create a RAID0 of those pairs. Software raid should do this without any problem. And can even be done such that it can be grown in the future, as well as work across multiple cards (though the latter is supported by some cards as well). > > >>That 128MB of buffer cache may very well be too small to keep the IO >>rate up, and/or there may be a more subtle problem with the LSI card, >>and/or you may have a configuration problem, but _something(s)_ need >>fixing since you are only getting raw sequential IO of ~100-150MB/s >>when it should be above 500MB/s. > > > It looks like there's a way to add more memory to it. This memory probably helps more in writing than reading. If you are reading the same area over and over, it might end up being a little bit of extra cache for that (but it should already be cached in system RAM, so you don't really get anything). ... >>Initial reads are only going to be as fast as your HD subsystem, so >>there's a reason for making the HD subsystem faster even if all you >>care about is reads. In addition, I'll repeat my previous advice >>that upgrading to 16GB of RAM would be well worth it for you. > > > 12GB is my max. I may run with it for a while and see. If your working set truly is 10GB, then you can get a massive performance increase even at 12GB. If your working set is 10GB and you have 6GB of RAM, it probably is always swapping out what it just read for the new stuff, even though you will read that same thing again in a few seconds. So rather than just paying for the 4GB that can't be cached, you pay for the whole 10. John =:-> > > >>Hope this helps, >>Ron Peacetree >> >> >> >>---------------------------(end of broadcast)--------------------------- >>TIP 9: In versions below 8.0, the planner will ignore your desire to >> choose an index scan if your joining column's datatypes do not >> match --------------enig70D3342FFE24F9008453FCB0 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDCvDkJdeBCYSNAAMRAoHFAJ9BcZLQHppIRddenaUCQscU92321wCgpoYF vp7zmFoS9+22qOuILwy9U88= =lZuP -----END PGP SIGNATURE----- --------------enig70D3342FFE24F9008453FCB0-- From pgsql-performance-owner@postgresql.org Sat Aug 27 02:02:08 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 65065D7729 for ; Tue, 23 Aug 2005 09:14:32 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 09337-07 for ; Tue, 23 Aug 2005 12:14:30 +0000 (GMT) Received: from floppy.pyrenet.fr (floppy.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id 16991D7726 for ; Tue, 23 Aug 2005 09:14:28 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id 6CBDF30B42; Tue, 23 Aug 2005 14:26:38 +0200 (MET DST) From: "tobbe" X-Newsgroups: pgsql.performance Subject: Performance for relative large DB Date: 23 Aug 2005 05:14:24 -0700 Organization: http://groups.google.com Lines: 27 Message-ID: <1124799264.282119.167760@g47g2000cwa.googlegroups.com> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" X-Complaints-To: groups-abuse@google.com User-Agent: G2/0.2 Complaints-To: groups-abuse@google.com Injection-Info: g47g2000cwa.googlegroups.com; posting-host=212.209.39.154; posting-account=6PgMzAwAAAAyhWPJdwnKqBSRj5_iJFf6 To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/469 X-Sequence-Number: 14216 Hi. The company that I'm working for are surveying the djungle of DBMS since we are due to implement the next generation of our system. The companys buissnes is utilizing the DBMS to store data that are accessed trough the web at daytime (only SELECTs, sometimes with joins, etc). The data is a collection of bjects that are for sale. The data consists of basic text information about theese togheter with some group information, etc. The data is updated once every night. There are about 4 M posts in the database (one table) and is expected to grow with atleast 50% during a reasonable long time. How well would PostgreSQL fit our needs? We are using Pervasive SQL today and suspect that it is much to small. We have some problems with latency. Esp. when updating information, complicated conditions in selects and on concurrent usage. Best Regards Robert Bengtsson Project Manager From pgsql-performance-owner@postgresql.org Tue Aug 23 10:52:55 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 27A89D6F1D for ; Tue, 23 Aug 2005 10:52:54 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 53180-06 for ; Tue, 23 Aug 2005 13:52:51 +0000 (GMT) Received: from window.monsterlabs.com (window.monsterlabs.com [216.183.105.176]) by svr1.postgresql.org (Postfix) with SMTP id 9D320D6EC7 for ; Tue, 23 Aug 2005 10:52:49 -0300 (ADT) Received: (qmail 4699 invoked from network); 23 Aug 2005 13:52:51 -0000 Received: from 12-215-178-50.client.mchsi.com (HELO ?192.168.1.120?) (12.215.178.50) by 0 with SMTP; 23 Aug 2005 13:52:51 -0000 In-Reply-To: <430A78B2.8080803@superig.com.br> References: <430A78B2.8080803@superig.com.br> Mime-Version: 1.0 (Apple Message framework v734) Content-Type: text/plain; charset=WINDOWS-1252; delsp=yes; format=flowed Message-Id: Cc: pgsql-performance list Content-Transfer-Encoding: quoted-printable From: "Thomas F. O'Connell" Subject: Re: pgbench Date: Tue, 23 Aug 2005 08:52:46 -0500 To: Philip Pinkerton X-Mailer: Apple Mail (2.734) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/340 X-Sequence-Number: 14087 pgbench is located in the contrib directory of any source tarball, =20 along with a README that serves as documentation. -- Thomas F. O'Connell Co-Founder, Information Architect Sitening, LLC Strategic Open Source: Open Your i=99 http://www.sitening.com/ 110 30th Avenue North, Suite 6 Nashville, TN 37203-6320 615-469-5150 615-469-5151 (fax) On Aug 22, 2005, at 8:15 PM, Philip Pinkerton wrote: > I am looking for the latest pgbench and documentation. > > If someone know where I can locate them it would save a lot of =20 > search time. > > Thanks > > Philip Pinkerton > TPC-C Benchmarks Sybase > Independant Consultant > Rio de Janeiro, RJ, Brazil 22031-010 From pgsql-performance-owner@postgresql.org Tue Aug 23 11:50:27 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 92E7AD6F5A for ; Tue, 23 Aug 2005 11:50:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 63309-09 for ; Tue, 23 Aug 2005 14:50:22 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.201]) by svr1.postgresql.org (Postfix) with ESMTP id 9E0C9D6F14 for ; Tue, 23 Aug 2005 11:50:20 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i17so5174wra for ; Tue, 23 Aug 2005 07:50:26 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:user-agent:x-accept-language:mime-version:to:subject:content-type:content-transfer-encoding; b=CVW45rnpQE2pqFKdgY/vgpoQXsFk1EMiydRAtnvdJTy4s/vXQxjJ81xza4IrC5pj/cYnOlvb2647Xh4g3/rWuPzyZQw/RKTlm2gBBNH1PmrwSP3UCj+W/x8JKErpdeKDm9KCWvQFu+WIK+FIgHchYMsFJIUVj5fmQoBk+b4a/Fo= Received: by 10.54.27.6 with SMTP id a6mr3806653wra; Tue, 23 Aug 2005 07:50:26 -0700 (PDT) Received: from ?192.168.3.4? ([81.182.248.121]) by mx.gmail.com with ESMTP id g9sm11154106wra.2005.08.23.07.50.24; Tue, 23 Aug 2005 07:50:26 -0700 (PDT) Message-ID: <430B379A.7090204@gmail.com> Date: Tue, 23 Aug 2005 16:50:02 +0200 From: =?ISO-8859-2?Q?Sz=FBcs_G=E1bor?= User-Agent: Mozilla Thunderbird 1.0.6 (Windows/20050716) X-Accept-Language: hu-hu, hu, en-us, en MIME-Version: 1.0 To: "pgsql-performance@postgresql.org" Subject: fake condition causes far better plan Content-Type: text/plain; charset=ISO-8859-2; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.447 required=5 tests=[AWL=-0.047, RCVD_BY_IP=0.024, TO_ADDRESS_EQ_REAL=0.47] X-Spam-Level: X-Archive-Number: 200508/341 X-Sequence-Number: 14088 Dear Gurus, System: Debian "Woody" 2.4.28 Version: PostgreSQL 7.4.8 I have a join which causes a better hash if I provide a "trivial" condition: WHERE m.nap > '1900-01-01'::date This is a date field with a minimum of '2005-06-21'. However, if I omit this condition from the WHERE clause, I get a far worse plan. There's also something not quite right in the cost tuning part of the config file, but I *think* it shouldn't cause such a bad plan. Explain analyze times: With fake condition: 1104 msec Without it: 11653 msec Without, mergejoin disabled: 5776 msec For full query and plans, see below. The operator "!=@" is the nonequity operator extended so that it treats NULL as a one-element equivalence class, thus never returning NULL. (NULL !=@ NULL is false, NULL !=@ "anything else" is true) 1. What may be the cause that this "obvious" condition causes a far better hash plan than the one without it, even while mergejoin is disabled? 2. What may be the cause that the planner favors mergejoin to hashjoin? usually a sign of too high/too low random page cost, for example? I'm willing to provide config options if it helps. TIA, -- G. -------------- the query with fake condition (m.nap>=...) -------------- explain analyze SELECT DISTINCT mv.az, mv.vonalkod, mv.idopont, mv.muszakhely as mvhely, mv.muszaknap as mvnap, mv.muszakkod as mvmkod, m.hely, m.nap, m.muszakkod as mkod, m.tol, m.ig FROM muvelet_vonalkod mv left join olvaso_hely oh on (oh.olvaso_nev = mv.olvaso_nev and oh.tol <= mv.idopont and mv.idopont < oh.ig) left join muszak m on (oh.hely = m.hely and m.tol <= mv.idopont and mv.idopont < m.ig) , muvelet_vonalkod_ny ny where mv.az = ny.muvelet_vonalkod and ny.idopont >= now()-1 and m.nap >= '1900-01-01'::date and (mv.muszakhely!=@m.hely or mv.muszaknap!=@m.nap or mv.muszakkod!=@m.muszakkod); -------------- best plan with fake condition -------------- Unique (cost=6484.22..6826.73 rows=11417 width=75) (actual time=1103.870..1103.872 rows=1 loops=1) -> Sort (cost=6484.22..6512.76 rows=11417 width=75) (actual time=1103.867..1103.868 rows=1 loops=1) Sort Key: mv.az, mv.vonalkod, mv.idopont, mv.muszakhely, mv.muszaknap, mv.muszakkod, m.hely, m.nap, m.muszakkod, m.tol, m.ig -> Hash Join (cost=1169.78..5434.78 rows=11417 width=75) (actual time=1075.836..1103.835 rows=1 loops=1) Hash Cond: ("outer".hely = "inner".hely) Join Filter: (("inner".tol <= "outer".idopont) AND ("outer".idopont < "inner".ig) AND (CASE WHEN (("outer".muszakhely IS NULL) AND ("inner".hely IS NULL)) THEN false ELSE CASE WHEN (("outer".muszakhely IS NULL) AND ("inner".hely IS NOT NULL)) THEN true ELSE CASE WHEN (("outer".muszakhely IS NOT NULL) AND ("inner".hely IS NULL)) THEN true ELSE ("outer".muszakhely <> "inner".hely) END END END OR CASE WHEN (("outer".muszaknap IS NULL) AND ("inner".nap IS NULL)) THEN false ELSE CASE WHEN (("outer".muszaknap IS NULL) AND ("inner".nap IS NOT NULL)) THEN true ELSE CASE WHEN (("outer".muszaknap IS NOT NULL) AND ("inner".nap IS NULL)) THEN true ELSE ("outer".muszaknap <> "inner".nap) END END END OR CASE WHEN (("outer".muszakkod IS NULL) AND ("inner".muszakkod IS NULL)) THEN false ELSE CASE WHEN (("outer".muszakkod IS NULL) AND ("inner".muszakkod IS NOT NULL)) THEN true ELSE CASE WHEN (("outer".muszakkod IS NOT NULL) AND ("inner".muszakkod IS NULL)) THEN true ELSE ("outer".muszakkod <> "inner".muszakkod) END END END)) -> Hash Join (cost=1167.65..2860.48 rows=1370 width=51) (actual time=533.035..741.211 rows=3943 loops=1) Hash Cond: ("outer".muvelet_vonalkod = "inner".az) -> Index Scan using muvelet_vonalkod_ny_idopont on muvelet_vonalkod_ny ny (cost=0.00..1351.88 rows=24649 width=4) (actual time=0.161..10.735 rows=3943 loops=1) Index Cond: (idopont >= (now() - ('00:00:00'::interval + ('1 days'::text)::interval))) -> Hash (cost=1124.61..1124.61 rows=3618 width=51) (actual time=532.703..532.703 rows=0 loops=1) -> Nested Loop (cost=0.00..1124.61 rows=3618 width=51) (actual time=0.209..443.765 rows=61418 loops=1) -> Seq Scan on olvaso_hely oh (cost=0.00..1.01 rows=1 width=28) (actual time=0.031..0.036 rows=1 loops=1) -> Index Scan using muvelet_vonalkod_pk2 on muvelet_vonalkod mv (cost=0.00..1060.30 rows=3617 width=55) (actual time=0.162..244.158 rows=61418 loops=1) Index Cond: ((("outer".olvaso_nev)::text = (mv.olvaso_nev)::text) AND ("outer".tol <= mv.idopont) AND (mv.idopont < "outer".ig)) -> Hash (cost=1.94..1.94 rows=75 width=28) (actual time=0.333..0.333 rows=0 loops=1) -> Seq Scan on muszak m (cost=0.00..1.94 rows=75 width=28) (actual time=0.070..0.230 rows=73 loops=1) Filter: (nap >= '2001-01-01'::date) Total runtime: 1104.244 ms (19 rows) -------------- mergejoin disabled, no fake condition -------------- Unique (cost=256601.12..262763.39 rows=205409 width=75) (actual time=5776.476..5776.479 rows=1 loops=1) -> Sort (cost=256601.12..257114.64 rows=205409 width=75) (actual time=5776.472..5776.472 rows=1 loops=1) Sort Key: mv.az, mv.vonalkod, mv.idopont, mv.muszakhely, mv.muszaknap, mv.muszakkod, m.hely, m.nap, m.muszakkod, m.tol, m.ig -> Hash Join (cost=132547.25..228451.03 rows=205409 width=75) (actual time=5733.661..5776.428 rows=1 loops=1) Hash Cond: ("outer".muvelet_vonalkod = "inner".az) -> Index Scan using muvelet_vonalkod_ny_idopont on muvelet_vonalkod_ny ny (cost=0.00..1351.88 rows=24649 width=4) (actual time=0.179..8.578 rows=3940 loops=1) Index Cond: (idopont >= (now() - ('00:00:00'::interval + ('1 days'::text)::interval))) -> Hash (cost=124566.75..124566.75 rows=542600 width=75) (actual time=5697.192..5697.192 rows=0 loops=1) -> Hash Left Join (cost=2.95..124566.75 rows=542600 width=75) (actual time=33.430..5689.636 rows=484 loops=1) Hash Cond: ("outer".hely = "inner".hely) Join Filter: (("inner".tol <= "outer".idopont) AND ("outer".idopont < "inner".ig)) Filter: (CASE WHEN (("outer".muszakhely IS NULL) AND ("inner".hely IS NULL)) THEN false ELSE CASE WHEN (("outer".muszakhely IS NULL) AND ("inner".hely IS NOT NULL)) THEN true ELSE CASE WHEN (("outer".muszakhely IS NOT NULL) AND ("inner".hely IS NULL)) THEN true ELSE ("outer".muszakhely <> "inner".hely) END END END OR CASE WHEN (("outer".muszaknap IS NULL) AND ("inner".nap IS NULL)) THEN false ELSE CASE WHEN (("outer".muszaknap IS NULL) AND ("inner".nap IS NOT NULL)) THEN true ELSE CASE WHEN (("outer".muszaknap IS NOT NULL) AND ("inner".nap IS NULL)) THEN true ELSE ("outer".muszaknap <> "inner".nap) END END END OR CASE WHEN (("outer".muszakkod IS NULL) AND ("inner".muszakkod IS NULL)) THEN false ELSE CASE WHEN (("outer".muszakkod IS NULL) AND ("inner".muszakkod IS NOT NULL)) THEN true ELSE CASE WHEN (("outer".muszakkod IS NOT NULL) AND ("inner".muszakkod IS NULL)) THEN true ELSE ("outer".muszakkod <> "inner".muszakkod) END END END) -> Hash Left Join (cost=1.01..2317.03 rows=65112 width=51) (actual time=0.462..542.361 rows=61465 loops=1) Hash Cond: (("outer".olvaso_nev)::text = ("inner".olvaso_nev)::text) Join Filter: (("inner".tol <= "outer".idopont) AND ("outer".idopont < "inner".ig)) -> Seq Scan on muvelet_vonalkod mv (cost=0.00..1502.12 rows=65112 width=55) (actual time=0.028..123.649 rows=61465 loops=1) -> Hash (cost=1.01..1.01 rows=1 width=28) (actual time=0.045..0.045 rows=0 loops=1) -> Seq Scan on olvaso_hely oh (cost=0.00..1.01 rows=1 width=28) (actual time=0.031..0.033 rows=1 loops=1) -> Hash (cost=1.75..1.75 rows=75 width=28) (actual time=0.319..0.319 rows=0 loops=1) -> Seq Scan on muszak m (cost=0.00..1.75 rows=75 width=28) (actual time=0.067..0.215 rows=73 loops=1) Total runtime: 5776.778 ms (21 rows) -------------- mergejoin enabled, no fake condition -------------- Unique (cost=210234.71..216396.98 rows=205409 width=75) (actual time=11652.868..11652.870 rows=1 loops=1) -> Sort (cost=210234.71..210748.24 rows=205409 width=75) (actual time=11652.865..11652.865 rows=1 loops=1) Sort Key: mv.az, mv.vonalkod, mv.idopont, mv.muszakhely, mv.muszaknap, mv.muszakkod, m.hely, m.nap, m.muszakkod, m.tol, m.ig -> Merge Join (cost=3152.69..182084.63 rows=205409 width=75) (actual time=11408.433..11652.836 rows=1 loops=1) Merge Cond: ("outer".az = "inner".muvelet_vonalkod) -> Nested Loop Left Join (cost=2.76..174499.23 rows=542600 width=75) (actual time=1.506..11632.727 rows=484 loops=1) Join Filter: (("outer".hely = "inner".hely) AND ("inner".tol <= "outer".idopont) AND ("outer".idopont < "inner".ig)) Filter: (CASE WHEN (("outer".muszakhely IS NULL) AND ("inner".hely IS NULL)) THEN false ELSE CASE WHEN (("outer".muszakhely IS NULL) AND ("inner".hely IS NOT NULL)) THEN true ELSE CASE WHEN (("outer".muszakhely IS NOT NULL) AND ("inner".hely IS NULL)) THEN true ELSE ("outer".muszakhely <> "inner".hely) END END END OR CASE WHEN (("outer".muszaknap IS NULL) AND ("inner".nap IS NULL)) THEN false ELSE CASE WHEN (("outer".muszaknap IS NULL) AND ("inner".nap IS NOT NULL)) THEN true ELSE CASE WHEN (("outer".muszaknap IS NOT NULL) AND ("inner".nap IS NULL)) THEN true ELSE ("outer".muszaknap <> "inner".nap) END END END OR CASE WHEN (("outer".muszakkod IS NULL) AND ("inner".muszakkod IS NULL)) THEN false ELSE CASE WHEN (("outer".muszakkod IS NULL) AND ("inner".muszakkod IS NOT NULL)) THEN true ELSE CASE WHEN (("outer".muszakkod IS NOT NULL) AND ("inner".muszakkod IS NULL)) THEN true ELSE ("outer".muszakkod <> "inner".muszakkod) END END END) -> Nested Loop Left Join (cost=1.01..3578.48 rows=65112 width=51) (actual time=0.140..757.392 rows=61461 loops=1) Join Filter: ((("inner".olvaso_nev)::text = ("outer".olvaso_nev)::text) AND ("inner".tol <= "outer".idopont) AND ("outer".idopont < "inner".ig)) -> Index Scan using muvelet_vonalkod_pkey on muvelet_vonalkod mv (cost=0.00..1786.89 rows=65112 width=55) (actual time=0.103..144.516 rows=61461 loops=1) -> Materialize (cost=1.01..1.02 rows=1 width=28) (actual time=0.001..0.002 rows=1 loops=61461) -> Seq Scan on olvaso_hely oh (cost=0.00..1.01 rows=1 width=28) (actual time=0.005..0.007 rows=1 loops=1) -> Materialize (cost=1.75..2.50 rows=75 width=28) (actual time=0.001..0.054 rows=73 loops=61461) -> Seq Scan on muszak m (cost=0.00..1.75 rows=75 width=28) (actual time=0.012..0.179 rows=73 loops=1) -> Sort (cost=3149.93..3211.55 rows=24649 width=4) (actual time=15.420..17.108 rows=2356 loops=1) Sort Key: ny.muvelet_vonalkod -> Index Scan using muvelet_vonalkod_ny_idopont on muvelet_vonalkod_ny ny (cost=0.00..1351.88 rows=24649 width=4) (actual time=0.048..9.502 rows=3942 loops=1) Index Cond: (idopont >= (now() - ('00:00:00'::interval + ('1 days'::text)::interval))) Total runtime: 11653.429 ms (20 rows) From pgsql-performance-owner@postgresql.org Tue Aug 23 13:16:10 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B8F49D702D for ; Tue, 23 Aug 2005 13:04:09 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 66522-09 for ; Tue, 23 Aug 2005 16:04:07 +0000 (GMT) Received: from floppy.pyrenet.fr (floppy.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id A8164D702B for ; Tue, 23 Aug 2005 13:04:05 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id ACF4A30B42; Tue, 23 Aug 2005 18:16:13 +0200 (MET DST) From: Chris Browne X-Newsgroups: pgsql.performance Subject: Re: Performance for relative large DB Date: Tue, 23 Aug 2005 11:12:51 -0400 Organization: cbbrowne Computing Inc Lines: 51 Message-ID: <60wtmcptzg.fsf@dba2.int.libertyrms.com> References: <1124799264.282119.167760@g47g2000cwa.googlegroups.com> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 X-Complaints-To: usenet@news.hub.org User-Agent: Gnus/5.1007 (Gnus v5.10.7) XEmacs/21.4.17 (Jumbo Shrimp, linux) Cancel-Lock: sha1:RcRQTv8EtD3YOVl9Uf0wrl1M3nI= To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.116 required=5 tests=[AWL=0.116] X-Spam-Level: X-Archive-Number: 200508/343 X-Sequence-Number: 14090 "tobbe" writes: > The company that I'm working for are surveying the djungle of DBMS > since we are due to implement the next generation of our system. > > The companys buissnes is utilizing the DBMS to store data that are > accessed trough the web at daytime (only SELECTs, sometimes with joins, > etc). The data is a collection of bjects that are for sale. The data > consists of basic text information about theese togheter with some > group information, etc. > > The data is updated once every night. How much data is updated per night? The whole 4M "posts"? Or just some subset? > There are about 4 M posts in the database (one table) and is > expected to grow with atleast 50% during a reasonable long time. So you're expecting to have ~6M entries in the 'posts' table? > How well would PostgreSQL fit our needs? > > We are using Pervasive SQL today and suspect that it is much to small. > We have some problems with latency. Esp. when updating information, > complicated conditions in selects and on concurrent usage. If you're truly updating all 4M/6M rows each night, *that* would turn out to be something of a bottleneck, as every time you update a tuple, this creates a new copy, leaving the old one to be later cleaned away via VACUUM. That strikes me as unlikely: I expect instead that you update a few thousand or a few tens of thousands of entries per day, in which case the "vacuum pathology" won't be a problem. I wouldn't expect PostgreSQL to be "too small;" it can and does cope well with complex queries. And the use of MVCC allows there to be a relatively minimal amount of locking done even though there may be a lot of concurrent users, the particular merit there being that you can essentially eliminate most read locks. That is, you can get consistent reports without having to lock rows or tables. One table with millions of rows isn't that complex a scenario :-). -- output = ("cbbrowne" "@" "cbbrowne.com") http://cbbrowne.com/info/spiritual.html Appendium to the Rules of the Evil Overlord #1: "I will not build excessively integrated security-and-HVAC systems. They may be Really Cool, but are far too vulnerable to breakdowns." From pgsql-performance-owner@postgresql.org Tue Aug 23 12:18:04 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3E7D5D6F74 for ; Tue, 23 Aug 2005 12:14:33 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 45705-05 for ; Tue, 23 Aug 2005 15:14:30 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 5B1D7D6F63 for ; Tue, 23 Aug 2005 12:14:28 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7NFEKD3002639; Tue, 23 Aug 2005 11:14:22 -0400 (EDT) To: =?ISO-8859-2?Q?Sz=FBcs_G=E1bor?= Cc: "pgsql-performance@postgresql.org" Subject: Re: fake condition causes far better plan In-reply-to: <430B379A.7090204@gmail.com> References: <430B379A.7090204@gmail.com> Comments: In-reply-to =?ISO-8859-2?Q?Sz=FBcs_G=E1bor?= message dated "Tue, 23 Aug 2005 16:50:02 +0200" Date: Tue, 23 Aug 2005 11:14:19 -0400 Message-ID: <2638.1124810059@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/342 X-Sequence-Number: 14089 =?ISO-8859-2?Q?Sz=FBcs_G=E1bor?= writes: > [ bad query plan ] Most of the problem is here: > -> Index Scan using muvelet_vonalkod_ny_idopont on > muvelet_vonalkod_ny ny (cost=0.00..1351.88 rows=24649 width=4) (actual > time=0.161..10.735 rows=3943 loops=1) > Index Cond: (idopont >= (now() - > ('00:00:00'::interval + ('1 days'::text)::interval))) (BTW, you lied about the query, because this index condition doesn't match anything in the given query text.) Pre-8.0 releases aren't capable of making useful statistical estimates for conditions involving nonconstant subexpressions, so you get a badly-mistaken row count estimate that leads to a poor choice of plan. If you can't update to 8.0, the best answer is to do the date arithmetic on the client side. Another way is to create an allegedly-immutable function along the lines of "ago(interval) returns timestamptz" to hide the now() call --- this is dangerous but sometimes it's the easiest answer. See the archives. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 23 14:10:59 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E5AF5D6E2A for ; Tue, 23 Aug 2005 14:10:56 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 97915-03 for ; Tue, 23 Aug 2005 17:10:51 +0000 (GMT) Received: from web51309.mail.yahoo.com (web51309.mail.yahoo.com [206.190.38.175]) by svr1.postgresql.org (Postfix) with SMTP id 3BA7BD6F66 for ; Tue, 23 Aug 2005 14:10:50 -0300 (ADT) Received: (qmail 12114 invoked by uid 60001); 23 Aug 2005 17:10:45 -0000 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Message-ID:Received:Date:From:Subject:To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=33YFblMUk98I7sGSUJpwxrKEzFiJMgx0e6RcryMIhrQb46CucudvgCBTh90sT/zwlBG+fLp7/zFc1SeLrhO3lOJoLghE/4WG9B4j+iJ/CsoP5WHEBfkQTr6MoV0XkuX2EYuSwM6TrpkQzRBnRHXYGJQ5RGX39MBCVnckv61BLFM= ; Message-ID: <20050823171045.12112.qmail@web51309.mail.yahoo.com> Received: from [203.101.103.131] by web51309.mail.yahoo.com via HTTP; Tue, 23 Aug 2005 10:10:45 PDT Date: Tue, 23 Aug 2005 10:10:45 -0700 (PDT) From: gokulnathbabu manoharan Subject: Caching by Postgres To: pgsql-performance@postgresql.org MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.374 required=5 tests=[DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/344 X-Sequence-Number: 14091 Hi all, I like to know the caching policies of Postgresql. What parameter in the postgresql.conf affects the cache size used by the Postgresql? As far as I have searched my knowledge of the parameters are 1. shared_buffers - Sets the limit on the amount of shared memory used. If I take this is as the cache size then my performance should increase with the increase in the size of shared_buffers. But it seems it is not the case and my performance actually decreases with the increase in the shared_buffers. I have a RAM size of 32 GB. The table which I use more frequently has around 68 million rows. Can I cache this entire table in RAM? 2. work_mem - It is the amount of memory used by an operation. My guess is once the operation is complete this is freed and hence has nothing to do with the caching. 3. effective_cache_size - The parameter used by the query planner and has nothing to do with the actual caching. So kindly help me in pointing me to the correct parameter to set. It will be great if you can point me to the docs that explains the implementation of caching in Postgresql which will help me in understanding things much clearly. Thanks in advance. Gokul. __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com From pgsql-performance-owner@postgresql.org Tue Aug 23 14:26:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8F364D704F for ; Tue, 23 Aug 2005 14:26:22 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 01799-05 for ; Tue, 23 Aug 2005 17:26:18 +0000 (GMT) Received: from sccmmhc91.asp.att.net (sccmmhc91.asp.att.net [204.127.203.211]) by svr1.postgresql.org (Postfix) with ESMTP id 96673D6F66 for ; Tue, 23 Aug 2005 14:26:14 -0300 (ADT) Received: from juju.arbash-meinel.com ([12.214.18.81]) by sccmmhc91.asp.att.net (sccmmhc91) with ESMTP id <20050823172613m9100g70one>; Tue, 23 Aug 2005 17:26:13 +0000 Received: by juju.arbash-meinel.com (Postfix, from userid 505) id 75C5B55FA6; Tue, 23 Aug 2005 12:26:13 -0500 (CDT) Received: from [10.10.0.181] (unknown [85.20.238.130]) by juju.arbash-meinel.com (Postfix) with ESMTP id 9B9CC55F9C; Tue, 23 Aug 2005 12:26:01 -0500 (CDT) Message-ID: <430B5C27.9030007@arbash-meinel.com> Date: Tue, 23 Aug 2005 12:25:59 -0500 From: John A Meinel User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: gokulnathbabu manoharan Cc: pgsql-performance@postgresql.org Subject: Re: Caching by Postgres References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> In-Reply-To: <20050823171045.12112.qmail@web51309.mail.yahoo.com> X-Enigmail-Version: 0.92.0.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig6B9A2DFC2643B98ED6BD6B74" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 required=5 tests=[FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/345 X-Sequence-Number: 14092 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig6B9A2DFC2643B98ED6BD6B74 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit gokulnathbabu manoharan wrote: > Hi all, > > I like to know the caching policies of Postgresql. > What parameter in the postgresql.conf affects the > cache size used by the Postgresql? As far as I have > searched my knowledge of the parameters are In general, you don't. The OS handles caching based on file usage. So if you are using the files, the OS should cache them. Just like it does with any other program. > > 1. shared_buffers - Sets the limit on the amount of > shared memory used. If I take this is as the cache > size then my performance should increase with the > increase in the size of shared_buffers. But it seems > it is not the case and my performance actually > decreases with the increase in the shared_buffers. I > have a RAM size of 32 GB. The table which I use more > frequently has around 68 million rows. Can I cache > this entire table in RAM? There is a portion of this which is used for caching. But I believe before 8.1 there was code that went linearly through all of the shared_buffers and checked for dirty/clean pages. So there was a tradeoff that the bigger you make it, the longer that search goes. So you got diminishing returns, generally around 10k shared buffers. I think it is better in 8.1, but if the OS is going to cache it anyway (since it does), then having a Postgres cache is just wasting memory, and not letting cache as much. So I'm guessing that with 8.1 there would be 2 sweet spots. Low shared_buffers (<= 10k), and really high shared buffers (like all of available ram). But because postgres has been tuned for the former I would stick with it (I don't think shared_buffers can go >2GB, but that might just be work_mem/maintenance_work_mem). > > 2. work_mem - It is the amount of memory used by an > operation. My guess is once the operation is complete > this is freed and hence has nothing to do with the > caching. > > 3. effective_cache_size - The parameter used by the > query planner and has nothing to do with the actual > caching. This is important from a planner issue. Because the planner can then expect that the OS is doing its job and caching the tables, so index scans are cheaper than they would be otherwise. John =:-> > > So kindly help me in pointing me to the correct > parameter to set. > > It will be great if you can point me to the docs that > explains the implementation of caching in Postgresql > which will help me in understanding things much > clearly. > > Thanks in advance. > Gokul. > --------------enig6B9A2DFC2643B98ED6BD6B74 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (Darwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDC1woJdeBCYSNAAMRAo5oAKDK1553m7IjY7NssS5Ot/I7lDwKjQCfW21x 1B1z4H/JdL58uUMxS1ma3Ac= =QG/Q -----END PGP SIGNATURE----- --------------enig6B9A2DFC2643B98ED6BD6B74-- From pgsql-performance-owner@postgresql.org Tue Aug 23 14:41:46 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id DB981D6F66 for ; Tue, 23 Aug 2005 14:41:43 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 03546-07 for ; Tue, 23 Aug 2005 17:41:40 +0000 (GMT) Received: from wolff.to (wolff.to [66.93.197.194]) by svr1.postgresql.org (Postfix) with SMTP id 3CC02D6ED5 for ; Tue, 23 Aug 2005 14:41:39 -0300 (ADT) Received: (qmail 23428 invoked by uid 500); 23 Aug 2005 17:41:08 -0000 Date: Tue, 23 Aug 2005 12:41:08 -0500 From: Bruno Wolff III To: gokulnathbabu manoharan Cc: pgsql-performance@postgresql.org Subject: Re: Caching by Postgres Message-ID: <20050823174108.GD19586@wolff.to> Mail-Followup-To: Bruno Wolff III , gokulnathbabu manoharan , pgsql-performance@postgresql.org References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20050823171045.12112.qmail@web51309.mail.yahoo.com> User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/346 X-Sequence-Number: 14093 On Tue, Aug 23, 2005 at 10:10:45 -0700, gokulnathbabu manoharan wrote: > Hi all, > > I like to know the caching policies of Postgresql. > What parameter in the postgresql.conf affects the > cache size used by the Postgresql? As far as I have > searched my knowledge of the parameters are The main policy is to let the OS do most of the caching. > 1. shared_buffers - Sets the limit on the amount of > shared memory used. If I take this is as the cache > size then my performance should increase with the > increase in the size of shared_buffers. But it seems > it is not the case and my performance actually > decreases with the increase in the shared_buffers. I > have a RAM size of 32 GB. The table which I use more > frequently has around 68 million rows. Can I cache > this entire table in RAM? Using extermely large values for shared buffers is known to be a performance loss for Postgres. Some improvements were made for 8.0 and more for 8.1. The OS will cache frequently used data from files for you. So if you are using that table a lot and the rows aren't too wide, it should mostly be cached for you by the OS. > 2. work_mem - It is the amount of memory used by an > operation. My guess is once the operation is complete > this is freed and hence has nothing to do with the > caching. This is used for sorts and some other things. > 3. effective_cache_size - The parameter used by the > query planner and has nothing to do with the actual > caching. You are supposed to use this to give the planner an idea about how much space the OS will using for caching on behalf of Posgres. > So kindly help me in pointing me to the correct > parameter to set. > > It will be great if you can point me to the docs that > explains the implementation of caching in Postgresql > which will help me in understanding things much > clearly. You probably want to read the following: http://www.varlena.com/varlena/GeneralBits/Tidbits/perf.html From pgsql-performance-owner@postgresql.org Tue Aug 23 14:43:01 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CC667D6F5D for ; Tue, 23 Aug 2005 14:42:53 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 03180-03 for ; Tue, 23 Aug 2005 17:42:47 +0000 (GMT) Received: from frank.wiles.org (frank.wiles.org [24.124.39.75]) by svr1.postgresql.org (Postfix) with ESMTP id 2B176D6ED5 for ; Tue, 23 Aug 2005 14:42:46 -0300 (ADT) Received: from kungfu (frank.wiles.org [127.0.0.1]) by frank.wiles.org (8.13.1/8.13.1) with SMTP id j7NGpQiA008440; Tue, 23 Aug 2005 11:51:27 -0500 Date: Tue, 23 Aug 2005 12:43:23 -0500 From: Frank Wiles To: gokulnathbabu manoharan Cc: pgsql-performance@postgresql.org Subject: Re: Caching by Postgres Message-Id: <20050823124323.710a9dd3.frank@wiles.org> In-Reply-To: <20050823171045.12112.qmail@web51309.mail.yahoo.com> References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> X-Mailer: Sylpheed version 1.0.4 (GTK+ 1.2.10; i386-redhat-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/347 X-Sequence-Number: 14094 On Tue, 23 Aug 2005 10:10:45 -0700 (PDT) gokulnathbabu manoharan wrote: > Hi all, > > I like to know the caching policies of Postgresql. > What parameter in the postgresql.conf affects the > cache size used by the Postgresql? As far as I have > searched my knowledge of the parameters are > > 1. shared_buffers - Sets the limit on the amount of > shared memory used. If I take this is as the cache > size then my performance should increase with the > increase in the size of shared_buffers. But it seems > it is not the case and my performance actually > decreases with the increase in the shared_buffers. I > have a RAM size of 32 GB. The table which I use more > frequently has around 68 million rows. Can I cache > this entire table in RAM? increasing shared_buffers to a point helps, but after a certain threshold it can actually degree performance. > 2. work_mem - It is the amount of memory used by an > operation. My guess is once the operation is complete > this is freed and hence has nothing to do with the > caching. This is the amount of memory used for things like sorts and order bys on a per backend process basis. > 3. effective_cache_size - The parameter used by the > query planner and has nothing to do with the actual > caching. The instructs the query planner on how large the operating system's disk cache is. There isn't a built in cache, PostgreSQL relies on the operating system to cache the on disk information based on how often it is used. In most cases this is probably more accurate anyway. I wrote an article on PostgreSQL performance tuning that has links to several other related sites, you can find it here: http://www.revsys.com/writings/postgresql-performance.html --------------------------------- Frank Wiles http://www.wiles.org --------------------------------- From pgsql-performance-owner@postgresql.org Tue Aug 23 14:54:04 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7D9A1D70AB for ; Tue, 23 Aug 2005 14:54:03 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 06851-10 for ; Tue, 23 Aug 2005 17:53:57 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id DD2BCD6F9F for ; Tue, 23 Aug 2005 14:53:56 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [64.81.245.111] (account josh@agliodbs.com HELO temoku.sf.agliodbs.com) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7915661; Tue, 23 Aug 2005 10:56:14 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Caching by Postgres Date: Tue, 23 Aug 2005 10:57:09 -0700 User-Agent: KMail/1.8 Cc: John A Meinel , gokulnathbabu manoharan References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <430B5C27.9030007@arbash-meinel.com> In-Reply-To: <430B5C27.9030007@arbash-meinel.com> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200508231057.09885.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.048 required=5 tests=[AWL=-0.002, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/348 X-Sequence-Number: 14095 John, > So I'm guessing that with 8.1 there would be 2 sweet spots. Low > shared_buffers (<= 10k), and really high shared buffers (like all of > available ram). > But because postgres has been tuned for the former I would stick with it > (I don't think shared_buffers can go >2GB, but that might just be > work_mem/maintenance_work_mem). I'll be testing this as soon as we get some issues with the 64bit shared_buffer patch worked out. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Tue Aug 23 15:50:30 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9726DD710E for ; Tue, 23 Aug 2005 15:45:48 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 17023-08 for ; Tue, 23 Aug 2005 18:45:45 +0000 (GMT) Received: from nwkea-mail-1.sun.com (nwkea-mail-1.sun.com [192.18.42.13]) by svr1.postgresql.org (Postfix) with ESMTP id 3CDB7D7106 for ; Tue, 23 Aug 2005 15:45:44 -0300 (ADT) Received: from phys-bur-1 ([129.148.9.72]) by nwkea-mail-1.sun.com (8.12.10/8.12.9) with ESMTP id j7NIjj2B018077 for ; Tue, 23 Aug 2005 11:45:45 -0700 (PDT) Received: from conversion-daemon.bur-mail1.east.sun.com by bur-mail1.east.sun.com (iPlanet Messaging Server 5.2 HotFix 1.24 (built Dec 19 2003)) id <0ILO00M01UMBEL@bur-mail1.east.sun.com> (original mail from Donald.Courtney@Sun.COM) for pgsql-performance@postgresql.org; Tue, 23 Aug 2005 14:45:45 -0400 (EDT) Received: from [129.148.184.34] (gyama.East.Sun.COM [129.148.184.34]) by bur-mail1.east.sun.com (iPlanet Messaging Server 5.2 HotFix 1.24 (built Dec 19 2003)) with ESMTPA id <0ILO007CYUS82G@bur-mail1.east.sun.com>; Tue, 23 Aug 2005 14:45:45 -0400 (EDT) Date: Tue, 23 Aug 2005 14:41:39 -0400 From: Donald Courtney Subject: Re: Caching by Postgres In-reply-to: <20050823124323.710a9dd3.frank@wiles.org> To: pgsql-performance@postgresql.org Cc: Frank Wiles , gokulnathbabu manoharan Message-id: <430B6DE3.6040109@sun.com> Organization: Sun Microsystems MIME-version: 1.0 Content-type: text/plain; charset=ISO-8859-1; format=flowed Content-transfer-encoding: 7BIT X-Accept-Language: en-us, en User-Agent: Mozilla/5.0 (X11; U; SunOS i86pc; en-US; rv:1.7) Gecko/20041221 References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/349 X-Sequence-Number: 14096 I mean well with this comment - This whole issue of data caching is a troubling issue with postreSQL in that even if you ran postgreSQL on a 64 bit address space with larger number of CPUs you won't see much of a scale up and possibly even a drop. I am not alone in having the *expectation* that a database should have some cache size parameter and the option to skip the file system. If I use oracle, sybase, mysql and maxdb they all have the ability to size a data cache and move to 64 bits. Is this a crazy idea - that a project be started to get this adopted? Is it too big and structural to contemplate? From one who likes postgreSQL dc Frank Wiles wrote: >On Tue, 23 Aug 2005 10:10:45 -0700 (PDT) >gokulnathbabu manoharan wrote: > > > >>Hi all, >> >>I like to know the caching policies of Postgresql. >>What parameter in the postgresql.conf affects the >>cache size used by the Postgresql? As far as I have >>searched my knowledge of the parameters are >> >>1. shared_buffers - Sets the limit on the amount of >>shared memory used. If I take this is as the cache >>size then my performance should increase with the >>increase in the size of shared_buffers. But it seems >>it is not the case and my performance actually >>decreases with the increase in the shared_buffers. I >>have a RAM size of 32 GB. The table which I use more >>frequently has around 68 million rows. Can I cache >>this entire table in RAM? >> >> > > increasing shared_buffers to a point helps, but after > a certain threshold it can actually degree performance. > > > >>2. work_mem - It is the amount of memory used by an >>operation. My guess is once the operation is complete >>this is freed and hence has nothing to do with the >>caching. >> >> > > This is the amount of memory used for things like sorts and > order bys on a per backend process basis. > > > >>3. effective_cache_size - The parameter used by the >>query planner and has nothing to do with the actual >>caching. >> >> > > The instructs the query planner on how large the operating > system's disk cache is. There isn't a built in cache, PostgreSQL > relies on the operating system to cache the on disk information > based on how often it is used. In most cases this is probably > more accurate anyway. > > I wrote an article on PostgreSQL performance tuning that has > links to several other related sites, you can find it here: > > http://www.revsys.com/writings/postgresql-performance.html > > --------------------------------- > Frank Wiles > http://www.wiles.org > --------------------------------- > > >---------------------------(end of broadcast)--------------------------- >TIP 9: In versions below 8.0, the planner will ignore your desire to > choose an index scan if your joining column's datatypes do not > match > > From pgsql-performance-owner@postgresql.org Tue Aug 23 16:05:37 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A76B0D708A for ; Tue, 23 Aug 2005 16:05:35 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 34820-03 for ; Tue, 23 Aug 2005 19:05:31 +0000 (GMT) Received: from calvin.surfutopia.net (calvin.surfutopia.net [67.120.245.34]) by svr1.postgresql.org (Postfix) with ESMTP id 6E611D70AD for ; Tue, 23 Aug 2005 16:05:27 -0300 (ADT) Received: by calvin.surfutopia.net (Postfix, from userid 1000) id 8AA1FF22F; Tue, 23 Aug 2005 12:05:25 -0700 (PDT) Date: Tue, 23 Aug 2005 12:05:25 -0700 From: John Mendenhall To: Tom Lane Cc: pgsql-performance list Subject: Re: complex query performance assistance request Message-ID: <20050823190525.GA27623@calvin.surfutopia.net> References: <20050821034841.GA28968@calvin.surfutopia.net> <20050822182138.GA30818@calvin.surfutopia.net> <20830.1124741718@sss.pgh.pa.us> <20050822210751.GA32479@calvin.surfutopia.net> <26869.1124758470@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <26869.1124758470@sss.pgh.pa.us> User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/350 X-Sequence-Number: 14097 Tom, > > Would it be best to attempt to rewrite it for IN? > > Or, should we try to tie it in with a join? > > Couldn't say without a deeper understanding of what you're trying to > accomplish. Here are the results of each SQL rewrite. The first pass, I rewrote it as c.id IN (): ----- LOG: duration: 2669.682 ms statement: explain analyze SELECT c.id AS contact_id, sr.id AS sales_rep_id, p.id AS partner_id, coalesce(LTRIM(RTRIM(c.company)), LTRIM(RTRIM(c.firstname || ' ' || c.lastname))) AS contact_company, co.name AS contact_country, c.master_key_token FROM sales_reps sr JOIN partners p ON (sr.id = p.sales_rep_id) JOIN contacts c ON (p.id = c.partner_id) JOIN countries co ON (LOWER(c.country) = LOWER(co.code)) JOIN partner_classification pc ON (p.classification_id = pc.id AND pc.classification != 'Sales Rep') WHERE c.lead_deleted IS NULL AND c.id IN ( SELECT lr.contact_id FROM lead_requests lr, lead_request_status lrs WHERE lr.status_id = lrs.id AND lrs.is_closed = 0 ) ORDER BY contact_company, contact_id QUERY PLAN ------------------------------------------------------------------------------------------------------------------------------- Sort (cost=4413.35..4416.16 rows=1123 width=102) (actual time=2617.069..2617.719 rows=1071 loops=1) Sort Key: COALESCE(ltrim(rtrim((c.company)::text)), ltrim(rtrim((((c.firstname)::text || ' '::text) || (c.lastname)::text)))), c.id -> Merge Join (cost=4311.31..4356.45 rows=1123 width=102) (actual time=2549.717..2589.398 rows=1071 loops=1) Merge Cond: ("outer"."?column3?" = "inner"."?column9?") -> Sort (cost=14.00..14.61 rows=242 width=19) (actual time=9.765..9.966 rows=240 loops=1) Sort Key: lower((co.code)::text) -> Seq Scan on countries co (cost=0.00..4.42 rows=242 width=19) (actual time=0.142..5.118 rows=242 loops=1) -> Sort (cost=4297.31..4299.63 rows=928 width=95) (actual time=2539.685..2540.913 rows=1071 loops=1) Sort Key: lower((c.country)::text) -> Merge IN Join (cost=4163.02..4251.57 rows=928 width=95) (actual time=2377.539..2524.844 rows=1071 loops=1) Merge Cond: ("outer".id = "inner".contact_id) -> Sort (cost=1835.53..1851.27 rows=6296 width=95) (actual time=1843.866..1853.193 rows=6349 loops=1) Sort Key: c.id -> Merge Join (cost=75.65..1438.24 rows=6296 width=95) (actual time=51.713..1505.633 rows=6349 loops=1) Merge Cond: ("outer".partner_id = "inner".id) -> Index Scan using contacts_partner_id_idx on contacts c (cost=0.00..5303.84 rows=40243 width=85) (actual time=0.077..584.736 rows=40267 loops=1) Filter: (lead_deleted IS NULL) -> Sort (cost=75.65..76.37 rows=290 width=20) (actual time=51.508..62.288 rows=6462 loops=1) Sort Key: p.id -> Merge Join (cost=59.24..63.79 rows=290 width=20) (actual time=30.152..38.281 rows=395 loops=1) Merge Cond: ("outer".id = "inner".sales_rep_id) -> Sort (cost=2.42..2.52 rows=39 width=10) (actual time=1.390..1.505 rows=39 loops=1) Sort Key: sr.id -> Seq Scan on sales_reps sr (cost=0.00..1.39 rows=39 width=10) (actual time=0.026..0.380 rows=39 loops=1) -> Sort (cost=56.82..57.55 rows=290 width=20) (actual time=28.558..29.120 rows=395 loops=1) Sort Key: p.sales_rep_id -> Nested Loop (cost=24.35..44.96 rows=290 width=20) (actual time=0.191..21.408 rows=395 loops=1) Join Filter: ("inner".classification_id = "outer".id) -> Seq Scan on partner_classification pc (cost=0.00..1.04 rows=2 width=10) (actual time=0.068..0.121 rows=2 loops=1) Filter: ((classification)::text <> 'Sales Rep'::text) -> Materialize (cost=24.35..28.70 rows=435 width=30) (actual time=0.029..5.380 rows=435 loops=2) -> Seq Scan on partners p (cost=0.00..24.35 rows=435 width=30) (actual time=0.038..8.161 rows=435 loops=1) -> Sort (cost=2327.50..2351.43 rows=9573 width=11) (actual time=533.508..535.629 rows=1742 loops=1) Sort Key: lr.contact_id -> Merge Join (cost=1520.94..1694.49 rows=9573 width=11) (actual time=302.932..461.644 rows=1745 loops=1) Merge Cond: ("outer".id = "inner".status_id) -> Sort (cost=1.28..1.30 rows=8 width=10) (actual time=0.392..0.404 rows=7 loops=1) Sort Key: lrs.id -> Seq Scan on lead_request_status lrs (cost=0.00..1.16 rows=8 width=10) (actual time=0.117..0.280 rows=7 loops=1) Filter: (is_closed = 0::numeric) -> Sort (cost=1519.66..1558.55 rows=15556 width=21) (actual time=302.423..321.939 rows=15387 loops=1) Sort Key: lr.status_id -> Seq Scan on lead_requests lr (cost=0.00..436.56 rows=15556 width=21) (actual time=0.029..164.708 rows=15559 loops=1) Total runtime: 2632.987 ms (44 rows) ----- The second pass, I rewrote it to tie in with a JOIN, adding a DISTINCT at the top to get rid of the duplicates: ----- LOG: duration: 3285.645 ms statement: explain analyze SELECT DISTINCT c.id AS contact_id, sr.id AS sales_rep_id, p.id AS partner_id, coalesce(LTRIM(RTRIM(c.company)), LTRIM(RTRIM(c.firstname || ' ' || c.lastname))) AS contact_company, co.name AS contact_country, c.master_key_token FROM sales_reps sr JOIN partners p ON (sr.id = p.sales_rep_id) JOIN contacts c ON (p.id = c.partner_id) JOIN countries co ON (LOWER(c.country) = LOWER(co.code)) JOIN partner_classification pc ON (p.classification_id = pc.id AND pc.classification != 'Sales Rep') JOIN lead_requests lr ON (c.id = lr.contact_id) JOIN lead_request_status lrs ON (lr.status_id = lrs.id AND lrs.is_closed = 0) WHERE c.lead_deleted IS NULL ORDER BY contact_company, contact_id QUERY PLAN ------------------------------------------------------------------------------------------------------------------------------- Unique (cost=3039.78..3071.46 rows=1810 width=102) (actual time=3219.707..3228.637 rows=1071 loops=1) -> Sort (cost=3039.78..3044.31 rows=1810 width=102) (actual time=3219.695..3220.560 rows=1118 loops=1) Sort Key: COALESCE(ltrim(rtrim((c.company)::text)), ltrim(rtrim((((c.firstname)::text || ' '::text) || (c.lastname)::text)))), c.id, sr.id, p.id, co.name, c.master_key_token -> Merge Join (cost=2870.92..2941.85 rows=1810 width=102) (actual time=3156.788..3188.338 rows=1118 loops=1) Merge Cond: ("outer"."?column3?" = "inner"."?column9?") -> Sort (cost=14.00..14.61 rows=242 width=19) (actual time=9.196..9.445 rows=240 loops=1) Sort Key: lower((co.code)::text) -> Seq Scan on countries co (cost=0.00..4.42 rows=242 width=19) (actual time=0.128..3.914 rows=242 loops=1) -> Sort (cost=2856.92..2860.66 rows=1496 width=95) (actual time=3147.340..3148.477 rows=1118 loops=1) Sort Key: lower((c.country)::text) -> Merge Join (cost=2750.88..2778.03 rows=1496 width=95) (actual time=3008.933..3132.122 rows=1118 loops=1) Merge Cond: ("outer".id = "inner".status_id) -> Sort (cost=1.28..1.30 rows=8 width=10) (actual time=0.366..0.379 rows=7 loops=1) Sort Key: lrs.id -> Seq Scan on lead_request_status lrs (cost=0.00..1.16 rows=8 width=10) (actual time=0.094..0.254 rows=7 loops=1) Filter: (is_closed = 0::numeric) -> Sort (cost=2749.60..2755.67 rows=2430 width=105) (actual time=3008.396..3023.502 rows=9992 loops=1) Sort Key: lr.status_id -> Merge Join (cost=1835.53..2612.95 rows=2430 width=105) (actual time=1975.714..2912.632 rows=10089 loops=1) Merge Cond: ("outer".contact_id = "inner".id) -> Index Scan using lead_requests_contact_id_idx on lead_requests lr (cost=0.00..683.87 rows=15556 width=21) (actual time=0.073..247.148 rows=15556 loops=1) -> Sort (cost=1835.53..1851.27 rows=6296 width=95) (actual time=1975.273..1988.664 rows=10089 loops=1) Sort Key: c.id -> Merge Join (cost=75.65..1438.24 rows=6296 width=95) (actual time=56.107..1625.186 rows=6349 loops=1) Merge Cond: ("outer".partner_id = "inner".id) -> Index Scan using contacts_partner_id_idx on contacts c (cost=0.00..5303.84 rows=40243 width=85) (actual time=0.047..580.311 rows=40267 loops=1) Filter: (lead_deleted IS NULL) -> Sort (cost=75.65..76.37 rows=290 width=20) (actual time=55.935..65.502 rows=6462 loops=1) Sort Key: p.id -> Merge Join (cost=59.24..63.79 rows=290 width=20) (actual time=31.765..39.925 rows=395 loops=1) Merge Cond: ("outer".id = "inner".sales_rep_id) -> Sort (cost=2.42..2.52 rows=39 width=10) (actual time=1.072..1.117 rows=39 loops=1) Sort Key: sr.id -> Seq Scan on sales_reps sr (cost=0.00..1.39 rows=39 width=10) (actual time=0.022..0.312 rows=39 loops=1) -> Sort (cost=56.82..57.55 rows=290 width=20) (actual time=30.489..30.893 rows=395 loops=1) Sort Key: p.sales_rep_id -> Nested Loop (cost=24.35..44.96 rows=290 width=20) (actual time=0.159..23.356 rows=395 loops=1) Join Filter: ("inner".classification_id = "outer".id) -> Seq Scan on partner_classification pc (cost=0.00..1.04 rows=2 width=10) (actual time=0.047..0.086 rows=2 loops=1) Filter: ((classification)::text <> 'Sales Rep'::text) -> Materialize (cost=24.35..28.70 rows=435 width=30) (actual time=0.028..6.124 rows=435 loops=2) -> Seq Scan on partners p (cost=0.00..24.35 rows=435 width=30) (actual time=0.039..9.383 rows=435 loops=1) Total runtime: 3241.139 ms (43 rows) ----- The DISTINCT ON condition was about the same amount of time, statistically. Removing the DISTINCT entirely only gave a very slight improvement in performance. So, the bottom line is, unless there are other ideas to improve the performance, I will most likely rewrite our application to use the c.id IN () option. Thank you very much for your input and suggestions. JohnM -- John Mendenhall john@surfutopia.net surf utopia internet services From pgsql-performance-owner@postgresql.org Tue Aug 23 16:23:53 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D4D29D70F9 for ; Tue, 23 Aug 2005 16:23:51 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 45934-04 for ; Tue, 23 Aug 2005 19:23:47 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 3F429D70B1 for ; Tue, 23 Aug 2005 16:23:46 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7NJNipF004588; Tue, 23 Aug 2005 15:23:44 -0400 (EDT) To: Donald Courtney Cc: pgsql-performance@postgresql.org, Frank Wiles , gokulnathbabu manoharan Subject: Re: Caching by Postgres In-reply-to: <430B6DE3.6040109@sun.com> References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> Comments: In-reply-to Donald Courtney message dated "Tue, 23 Aug 2005 14:41:39 -0400" Date: Tue, 23 Aug 2005 15:23:43 -0400 Message-ID: <4587.1124825023@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/351 X-Sequence-Number: 14098 Donald Courtney writes: > I am not alone in having the *expectation* that a database should have > some cache size parameter and the option to skip the file system. If > I use oracle, sybase, mysql and maxdb they all have the ability to > size a data cache and move to 64 bits. And you're not alone in holding that opinion despite having no shred of evidence that it's worthwhile expanding the cache that far. However, since we've gotten tired of hearing this FUD over and over, 8.1 will have the ability to set shared_buffers as high as you want. I expect next we'll be hearing from people complaining that they set shared_buffers to use all of RAM and performance went into the tank ... regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 23 16:34:58 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8BF3CD7121 for ; Tue, 23 Aug 2005 16:34:55 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 61644-02 for ; Tue, 23 Aug 2005 19:34:51 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id 4F34AD711A for ; Tue, 23 Aug 2005 16:34:50 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [64.81.245.111] (account josh@agliodbs.com HELO temoku.sf.agliodbs.com) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7916342; Tue, 23 Aug 2005 12:37:09 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Caching by Postgres Date: Tue, 23 Aug 2005 12:38:04 -0700 User-Agent: KMail/1.8 Cc: Donald Courtney , Frank Wiles , gokulnathbabu manoharan References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> In-Reply-To: <430B6DE3.6040109@sun.com> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200508231238.04844.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.048 required=5 tests=[AWL=-0.002, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/352 X-Sequence-Number: 14099 Donald, > This whole issue of data caching is a troubling issue with postreSQL > in that even if you ran postgreSQL on a 64 bit address space > with larger number of CPUs you won't see much of a scale up > and possibly even a drop. Since when? Barring the context switch bug, you're not going to get a drop with more processors/more RAM. You may fail to get any gain, though. If your database is only 100MB in size, having 11G of cache space isn't going to help you much over having only 1G. > I am not alone in having the *expectation* > that a database should have some cache size parameter and > the option to skip the file system. Sure, because that's the conventional wisdom, as writ by Oracle. However, this comes with substantial code maintenance costs and portability limitations which have to be measured against any gain in performance. > If I use oracle, sybase, mysql > and maxdb they all have the ability to size a data cache and move > to 64 bits. And yet, we regularly outperform Sybase and MySQL on heavy OLTP loads on commodity x86 hardware. So apparently DB caching isn't everything. ;-) I'm not saying that it's not worth testing larger database caches -- even taking over most of RAM -- on high-speed systems. In fact, I'm working on doing that kind of test now. However, barring test results, we can't assume that taking over RAM and the FS cache would have a substantial performance benefit; that remains to be shown. The other thing is that we've had, and continue to have, low-hanging fruit which have a clear and measurable effect on performance and are fixable without bloating the PG code. Some of these issues (COPY path, context switching, locks, GiST concurrency, some aggregates) have been addressed in the 8.1 code; some remain to be addressed (sorts, disk spill, 64-bit sort mem, other aggregates, index-only access, etc.). Why tackle a huge, 250-hour project which could fail when a 20-hour patch is more likely to provide the same performance benefit? We have the same discussion (annually) about mmap. Using mmap *might* provide us with a huge performance boost. However, it would *definitely* require 300hours (or more) of programmer time to test properly, and might not benefit us at all. Of course, if *you* want to work on large database cache improvements, be my guest ... it's an open source project! Submit your patches! I'll be happy to test them. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Tue Aug 23 17:18:28 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B3DC9D7133 for ; Tue, 23 Aug 2005 17:13:14 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 74871-09 for ; Tue, 23 Aug 2005 20:13:10 +0000 (GMT) Received: from mail.envyfinancial.com (unknown [206.248.142.186]) by svr1.postgresql.org (Postfix) with ESMTP id AF829D7116 for ; Tue, 23 Aug 2005 17:13:08 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by mail.envyfinancial.com (Postfix) with ESMTP id 21B801D062; Tue, 23 Aug 2005 16:03:43 -0400 (EDT) Received: from mail.envyfinancial.com ([127.0.0.1]) by localhost (mark.mielke.cc [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 18870-05; Tue, 23 Aug 2005 16:03:42 -0400 (EDT) Received: by mail.envyfinancial.com (Postfix, from userid 500) id DD1241D069; Tue, 23 Aug 2005 16:03:42 -0400 (EDT) Date: Tue, 23 Aug 2005 16:03:42 -0400 From: mark@mark.mielke.cc To: Donald Courtney Cc: pgsql-performance@postgresql.org, Frank Wiles , gokulnathbabu manoharan Subject: Re: Caching by Postgres Message-ID: <20050823200342.GA18891@mark.mielke.cc> References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <430B6DE3.6040109@sun.com> User-Agent: Mutt/1.4.2.1i X-Virus-Scanned: amavisd-new at mail.envyfinancial.com X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.18 required=5 tests=[AWL=0.002, NO_REAL_NAME=0.178] X-Spam-Level: X-Archive-Number: 200508/353 X-Sequence-Number: 14100 On Tue, Aug 23, 2005 at 02:41:39PM -0400, Donald Courtney wrote: > I mean well with this comment - > This whole issue of data caching is a troubling issue with postreSQL > in that even if you ran postgreSQL on a 64 bit address space > with larger number of CPUs you won't see much of a scale up > and possibly even a drop. I am not alone in having the *expectation* > that a database should have some cache size parameter and > the option to skip the file system. If I use oracle, sybase, mysql > and maxdb they all have the ability to size a data cache and move > to 64 bits. > Is this a crazy idea - that a project be started to get this adopted? > Is it > too big and structural to contemplate? > From one who likes postgreSQL Hey Donald. :-) This is an operating system issue, not a PostgreSQL issue. If you have more physical memory than fits in 32-bit addresses, and your operating system isn't using this extra memory to cache files (or anything else), than your OS is what I would consider to be broken (or at the very least, not designed for a 64-bit host). The only questions that can be asked here is - 1) can PostgreSQL do a better job than the OS at best utilizing system RAM, and 2) if so, is the net gain worth the added complexity to PostgreSQL? I happen to think that yes, PostgreSQL can do a better job than most OS's, as it has better information to make decisions as to which pages are worth keeping, and which are not, but no, it isn't worth the effort until PostgreSQL developers start running out of things to do. Buy your 64-bit platforms - but if page caching is your concern, 1) ensure that you really have more physical memory than can fit in 32 bits, and 2) ensure that your operating system is comfortable caching data pages from files above the 32-bit mark. Cheers, mark -- mark@mielke.cc / markm@ncf.ca / markm@nortel.com __________________________ . . _ ._ . . .__ . . ._. .__ . . . .__ | Neighbourhood Coder |\/| |_| |_| |/ |_ |\/| | |_ | |/ |_ | | | | | | \ | \ |__ . | | .|. |__ |__ | \ |__ | Ottawa, Ontario, Canada One ring to rule them all, one ring to find them, one ring to bring them all and in the darkness bind them... http://mark.mielke.cc/ From pgsql-performance-owner@postgresql.org Tue Aug 23 17:37:24 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 49DD2D70FF for ; Tue, 23 Aug 2005 17:29:36 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91328-06 for ; Tue, 23 Aug 2005 20:29:34 +0000 (GMT) Received: from vms044pub.verizon.net (vms044pub.verizon.net [206.46.252.44]) by svr1.postgresql.org (Postfix) with ESMTP id D9389D6FD9 for ; Tue, 23 Aug 2005 17:29:32 -0300 (ADT) Received: from osgiliath.mathom.us ([151.200.14.55]) by vms044.mailsrvcs.net (Sun Java System Messaging Server 6.2 HotFix 0.04 (built Dec 24 2004)) with ESMTPA id <0ILO00DQIZL92XE1@vms044.mailsrvcs.net> for pgsql-performance@postgresql.org; Tue, 23 Aug 2005 15:29:34 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by osgiliath.mathom.us (Postfix) with ESMTP id 2B048602638; Tue, 23 Aug 2005 16:29:33 -0400 (EDT) Received: from osgiliath.mathom.us ([127.0.0.1]) by localhost (osgiliath [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 04490-02; Tue, 23 Aug 2005 16:29:32 -0400 (EDT) Received: by osgiliath.mathom.us (Postfix, from userid 1000) id D9B07600162; Tue, 23 Aug 2005 16:29:32 -0400 (EDT) Date: Tue, 23 Aug 2005 16:29:32 -0400 From: Michael Stone Subject: Re: Caching by Postgres In-reply-to: <200508231238.04844.josh@agliodbs.com> To: Josh Berkus Cc: pgsql-performance@postgresql.org Mail-Followup-To: Josh Berkus , pgsql-performance@postgresql.org Message-id: <20050823202932.GK8667@mathom.us> MIME-version: 1.0 Content-type: text/plain; charset=us-ascii; format=flowed Content-disposition: inline X-Pgp-Fingerprint: 53 FF 38 00 E7 DD 0A 9C 84 52 84 C5 EE DF 7C 88 X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mathom.us References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> <200508231238.04844.josh@agliodbs.com> User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 required=5 tests=[FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/354 X-Sequence-Number: 14101 On Tue, Aug 23, 2005 at 12:38:04PM -0700, Josh Berkus wrote: >which have a clear and measurable effect on performance and are fixable >without bloating the PG code. Some of these issues (COPY path, context >switching Does that include increasing the size of read/write blocks? I've noticed that with a large enough table it takes a while to do a sequential scan, even if it's cached; I wonder if the fact that it takes a million read(2) calls to get through an 8G table is part of that. Mike Stone From pgsql-performance-owner@postgresql.org Tue Aug 23 17:42:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 82FF7D6FD7 for ; Tue, 23 Aug 2005 17:42:23 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 05351-04 for ; Tue, 23 Aug 2005 20:42:18 +0000 (GMT) Received: from floppy.pyrenet.fr (floppy.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id 2637BD6FD5 for ; Tue, 23 Aug 2005 17:42:17 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id 7CE7130B42; Tue, 23 Aug 2005 22:54:30 +0200 (MET DST) From: Chris Browne X-Newsgroups: pgsql.performance Subject: Re: Caching by Postgres Date: Tue, 23 Aug 2005 16:41:33 -0400 Organization: cbbrowne Computing Inc Lines: 29 Message-ID: <60slx0perm.fsf@dba2.int.libertyrms.com> References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 X-Complaints-To: usenet@news.hub.org User-Agent: Gnus/5.1007 (Gnus v5.10.7) XEmacs/21.4.17 (Jumbo Shrimp, linux) Cancel-Lock: sha1:Z88FuvlZon01WvzFSgLpygDCH2g= To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.115 required=5 tests=[AWL=0.115] X-Spam-Level: X-Archive-Number: 200508/355 X-Sequence-Number: 14102 Donald.Courtney@Sun.COM (Donald Courtney) writes: > I mean well with this comment - > This whole issue of data caching is a troubling issue with postreSQL > in that even if you ran postgreSQL on a 64 bit address space > with larger number of CPUs you won't see much of a scale up > and possibly even a drop. I am not alone in having the *expectation* > that a database should have some cache size parameter and > the option to skip the file system. If I use oracle, sybase, mysql > and maxdb they all have the ability to size a data cache and move > to 64 bits. > > Is this a crazy idea - that a project be started to get this > adopted? Is it too big and structural to contemplate? This project amounts to "Implement Your Own Operating System," because it requires that the DBMS take over the things that operating systems normally do, like: a) Managing access to filesystems and b) Managing memory The world is already sufficiently filled up with numerous variations of Linux, BSD 4.4 Lite, and UNIX System V; I can't see justification for reinventing this wheel still again. -- (format nil "~S@~S" "cbbrowne" "acm.org") http://cbbrowne.com/info/multiplexor.html Rules of the Evil Overlord #196. "I will hire an expert marksman to stand by the entrance to my fortress. His job will be to shoot anyone who rides up to challenge me." From pgsql-performance-owner@postgresql.org Tue Aug 23 18:19:57 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id EEDABD7093 for ; Tue, 23 Aug 2005 18:19:53 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 28189-01 for ; Tue, 23 Aug 2005 21:19:51 +0000 (GMT) Received: from floppy.pyrenet.fr (floppy.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id 5F2C9D7087 for ; Tue, 23 Aug 2005 18:19:49 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id B5A6E30B42; Tue, 23 Aug 2005 23:32:03 +0200 (MET DST) From: Chris Browne X-Newsgroups: pgsql.performance Subject: Re: Caching by Postgres Date: Tue, 23 Aug 2005 17:17:42 -0400 Organization: cbbrowne Computing Inc Lines: 24 Message-ID: <60oe7opd3d.fsf@dba2.int.libertyrms.com> References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> <200508231238.04844.josh@agliodbs.com> <20050823202932.GK8667@mathom.us> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 X-Complaints-To: usenet@news.hub.org User-Agent: Gnus/5.1007 (Gnus v5.10.7) XEmacs/21.4.17 (Jumbo Shrimp, linux) Cancel-Lock: sha1:RQHjE2QF6M4vONlqAnZ1A7VO1ho= To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.114 required=5 tests=[AWL=0.114] X-Spam-Level: X-Archive-Number: 200508/356 X-Sequence-Number: 14103 mstone+postgres@mathom.us (Michael Stone) writes: > On Tue, Aug 23, 2005 at 12:38:04PM -0700, Josh Berkus wrote: >> which have a clear and measurable effect on performance and are >> fixable without bloating the PG code. Some of these issues (COPY >> path, context switching > > Does that include increasing the size of read/write blocks? I've > noticed that with a large enough table it takes a while to do a > sequential scan, even if it's cached; I wonder if the fact that it > takes a million read(2) calls to get through an 8G table is part of > that. But behind the scenes, the OS is still going to have to evaluate the "is this in cache?" question for each and every one of those pages. (Assuming the kernel's page size is 8K; if it's smaller, the number of evaluations will be even higher...) Grouping the read(2) calls together isn't going to have any impact on _that_ evaluation. -- let name="cbbrowne" and tld="ntlug.org" in name ^ "@" ^ tld;; http://www3.sympatico.ca/cbbrowne/finances.html "People who don't use computers are more sociable, reasonable, and ... less twisted" -- Arthur Norman From pgsql-performance-owner@postgresql.org Tue Aug 23 18:34:23 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0C4FFD7087 for ; Tue, 23 Aug 2005 18:29:08 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 30277-03 for ; Tue, 23 Aug 2005 21:29:02 +0000 (GMT) Received: from brmea-mail-3.sun.com (brmea-mail-3.Sun.COM [192.18.98.34]) by svr1.postgresql.org (Postfix) with ESMTP id B11A1D704D for ; Tue, 23 Aug 2005 18:29:00 -0300 (ADT) Received: from phys-bur-1 ([129.148.9.72]) by brmea-mail-3.sun.com (8.12.10/8.12.9) with ESMTP id j7NLT2WV012413 for ; Tue, 23 Aug 2005 15:29:02 -0600 (MDT) Received: from conversion-daemon.bur-mail1.east.sun.com by bur-mail1.east.sun.com (iPlanet Messaging Server 5.2 HotFix 1.24 (built Dec 19 2003)) id <0ILP00M012AL63@bur-mail1.east.sun.com> (original mail from J.K.Shah@Sun.COM) for pgsql-performance@postgresql.org; Tue, 23 Aug 2005 17:29:02 -0400 (EDT) Received: from bur-mail1.east.sun.com (phys-bur-1 [129.148.9.72]) by bur-mail1.east.sun.com (iPlanet Messaging Server 5.2 HotFix 1.24 (built Dec 19 2003)) with ESMTP id <0ILP00BW22CDYO@bur-mail1.east.sun.com>; Tue, 23 Aug 2005 17:29:01 -0400 (EDT) Received: from [129.148.168.2] by bur-mail1.east.sun.com (mshttpd); Tue, 23 Aug 2005 17:29:01 -0400 Date: Tue, 23 Aug 2005 17:29:01 -0400 From: Jignesh Shah Subject: Re: Read/Write block sizes (Was: Caching by Postgres) To: Michael Stone Cc: Josh Berkus , pgsql-performance@postgresql.org Message-id: MIME-version: 1.0 X-Mailer: iPlanet Messenger Express 5.2 HotFix 1.24 (built Dec 19 2003) Content-type: text/plain; charset=us-ascii Content-language: en Content-transfer-encoding: 7BIT Content-disposition: inline X-Accept-Language: en X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/357 X-Sequence-Number: 14104 > Does that include increasing the size of read/write blocks? I've > noticedthat with a large enough table it takes a while to do a > sequential scan, > even if it's cached; I wonder if the fact that it takes a million > read(2) calls to get through an 8G table is part of that. > Actually some of that readaheads,etc the OS does already if it does some sort of throttling/clubbing of reads/writes. But its not enough for such types of workloads. Here is what I think will help: * Support for different Blocksize TABLESPACE without recompiling the code.. (Atlease support for a different Blocksize for the whole database without recompiling the code) * Support for bigger sizes of WAL files instead of 16MB files WITHOUT recompiling the code.. Should be a tuneable if you ask me (with checkpoint_segments at 256.. you have too many 16MB files in the log directory) (This will help OLTP benchmarks more since now they don't spend time rotating log files) * Introduce a multiblock or extent tunable variable where you can define a multiple of 8K (or BlockSize tuneable) to read a bigger chunk and store it in the bufferpool.. (Maybe writes too) (Most devices now support upto 1MB chunks for reads and writes) *There should be a way to preallocate files for TABLES in TABLESPACES otherwise with multiple table writes in the same filesystem ends with fragmented files which causes poor "READS" from the files. * With 64bit 1GB file chunks is also moot.. Maybe it should be tuneable too like 100GB without recompiling the code. Why recompiling is bad? Most companies that will support Postgres will support their own binaries and they won't prefer different versions of binaries for different blocksizes, different WAL file sizes, etc... and hence more function using the same set of binaries is more desirable in enterprise environments Regards, Jignesh From pgsql-performance-owner@postgresql.org Tue Aug 23 19:00:06 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 87428D7247 for ; Tue, 23 Aug 2005 19:00:01 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 36467-07 for ; Tue, 23 Aug 2005 21:59:51 +0000 (GMT) Received: from floppy.pyrenet.fr (floppy.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id 69A21D70CA for ; Tue, 23 Aug 2005 18:59:49 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id C599730B42; Wed, 24 Aug 2005 00:12:04 +0200 (MET DST) From: William Yu X-Newsgroups: pgsql.performance Subject: Re: Caching by Postgres Date: Tue, 23 Aug 2005 14:59:42 -0700 Organization: Hub.Org Networking Services Lines: 27 Message-ID: References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Complaints-To: usenet@news.hub.org User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en In-Reply-To: <430B6DE3.6040109@sun.com> To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.016 required=5 tests=[AWL=0.016] X-Spam-Level: X-Archive-Number: 200508/358 X-Sequence-Number: 14105 Donald Courtney wrote: > in that even if you ran postgreSQL on a 64 bit address space > with larger number of CPUs you won't see much of a scale up > and possibly even a drop. I am not alone in having the *expectation* What's your basis for believing this is the case? Why would PostgreSQL's dependence on the OS's caching/filesystem limit scalability? I know when I went from 32bit to 64bit Linux, I got *HUGE* increases in performance using the same amount of memory. And when I went from 2x1P to 2xDC, my average cpu usage % dropped almost in half. > that a database should have some cache size parameter and > the option to skip the file system. If I use oracle, sybase, mysql > and maxdb they all have the ability to size a data cache and move > to 64 bits. Josh Berkus has already mentioned this as conventional wisdom as written by Oracle. This may also be legacy wisdom. Oracle/Sybase/etc has been around for a long time; it was probably a clear performance win way back when. Nowadays with how far open-source OS's have advanced, I'd take it with a grain of salt and do my own performance analysis. I suspect the big vendors wouldn't change their stance even if they knew it was no longer true due to the support hassles. My personal experience with PostgreSQL. Dropping shared buffers from 2GB to 750MB improved performance on my OLTP DB a good 25%. Going down from 750MB to 150MB was another +10%. From pgsql-performance-owner@postgresql.org Tue Aug 23 20:04:23 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E367FD722D for ; Tue, 23 Aug 2005 20:04:05 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 46512-09 for ; Tue, 23 Aug 2005 23:04:03 +0000 (GMT) Received: from floppy.pyrenet.fr (floppy.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id B147DD714E for ; Tue, 23 Aug 2005 20:04:01 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id A1EF330B42; Wed, 24 Aug 2005 01:16:16 +0200 (MET DST) From: Chris Browne X-Newsgroups: pgsql.performance Subject: Re: Read/Write block sizes Date: Tue, 23 Aug 2005 18:09:09 -0400 Organization: cbbrowne Computing Inc Lines: 75 Message-ID: <60k6icpapm.fsf@dba2.int.libertyrms.com> References: Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 X-Complaints-To: usenet@news.hub.org User-Agent: Gnus/5.1007 (Gnus v5.10.7) XEmacs/21.4.17 (Jumbo Shrimp, linux) Cancel-Lock: sha1:FglkrWkAo9E+UnKEe6okIBqwugA= To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.114 required=5 tests=[AWL=0.114] X-Spam-Level: X-Archive-Number: 200508/359 X-Sequence-Number: 14106 J.K.Shah@Sun.COM (Jignesh Shah) writes: >> Does that include increasing the size of read/write blocks? I've >> noticedthat with a large enough table it takes a while to do a >> sequential scan, even if it's cached; I wonder if the fact that it >> takes a million read(2) calls to get through an 8G table is part of >> that. > > Actually some of that readaheads,etc the OS does already if it does > some sort of throttling/clubbing of reads/writes. But its not enough > for such types of workloads. > > Here is what I think will help: > > * Support for different Blocksize TABLESPACE without recompiling the > code.. (Atlease support for a different Blocksize for the whole > database without recompiling the code) > > * Support for bigger sizes of WAL files instead of 16MB files > WITHOUT recompiling the code.. Should be a tuneable if you ask me > (with checkpoint_segments at 256.. you have too many 16MB files in > the log directory) (This will help OLTP benchmarks more since now > they don't spend time rotating log files) > > * Introduce a multiblock or extent tunable variable where you can > define a multiple of 8K (or BlockSize tuneable) to read a bigger > chunk and store it in the bufferpool.. (Maybe writes too) (Most > devices now support upto 1MB chunks for reads and writes) > > *There should be a way to preallocate files for TABLES in > TABLESPACES otherwise with multiple table writes in the same > filesystem ends with fragmented files which causes poor "READS" from > the files. > > * With 64bit 1GB file chunks is also moot.. Maybe it should be > tuneable too like 100GB without recompiling the code. > > Why recompiling is bad? Most companies that will support Postgres > will support their own binaries and they won't prefer different > versions of binaries for different blocksizes, different WAL file > sizes, etc... and hence more function using the same set of binaries > is more desirable in enterprise environments Every single one of these still begs the question of whether the changes will have a *material* impact on performance. What we have been finding, as RAID controllers get smarter, is that it is getting increasingly futile to try to attach knobs to 'disk stuff;' it is *way* more effective to add a few more spindles to an array than it is to fiddle with which disks are to be allocated to what database 'objects.' The above suggested 'knobs' are all going to add to complexity and it is NOT evident that any of them will forcibly help. I could be wrong; code contributions combined with Actual Benchmarking would be the actual proof of the merits of the ideas. But it also suggests another question, namely... Will these represent more worthwhile improvements to speed than working on other optimizations that are on the TODO list? If someone spends 100h working on one of these items, and gets a 2% performance improvement, that's almost certain to be less desirable than spending 50h on something else that gets a 4% improvement. And we might discover that memory management improvements in Linux 2.6.16 or FreeBSD 5.5 allow some OS kernels to provide some such improvements "for free" behind our backs without *any* need to write database code. :-) -- let name="cbbrowne" and tld="ntlug.org" in name ^ "@" ^ tld;; http://www3.sympatico.ca/cbbrowne/postgresql.html Wiener's Law of Libraries: There are no answers, only cross references. From pgsql-performance-owner@postgresql.org Tue Aug 23 20:18:37 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0A70AD71EC for ; Tue, 23 Aug 2005 20:18:34 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 54190-02 for ; Tue, 23 Aug 2005 23:18:31 +0000 (GMT) Received: from vms044pub.verizon.net (vms044pub.verizon.net [206.46.252.44]) by svr1.postgresql.org (Postfix) with ESMTP id 39959D70F9 for ; Tue, 23 Aug 2005 20:18:30 -0300 (ADT) Received: from osgiliath.mathom.us ([151.200.14.55]) by vms044.mailsrvcs.net (Sun Java System Messaging Server 6.2 HotFix 0.04 (built Dec 24 2004)) with ESMTPA id <0ILP00DO57523IO1@vms044.mailsrvcs.net> for pgsql-performance@postgresql.org; Tue, 23 Aug 2005 18:12:39 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by osgiliath.mathom.us (Postfix) with ESMTP id B1A87602638; Tue, 23 Aug 2005 19:12:38 -0400 (EDT) Received: from osgiliath.mathom.us ([127.0.0.1]) by localhost (osgiliath [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 07218-04-8; Tue, 23 Aug 2005 19:12:38 -0400 (EDT) Received: by osgiliath.mathom.us (Postfix, from userid 1000) id 86F56600162; Tue, 23 Aug 2005 19:12:38 -0400 (EDT) Date: Tue, 23 Aug 2005 19:12:38 -0400 From: Michael Stone Subject: Re: Read/Write block sizes (Was: Caching by Postgres) In-reply-to: To: Jignesh Shah Cc: pgsql-performance@postgresql.org Mail-Followup-To: Jignesh Shah , pgsql-performance@postgresql.org Message-id: <20050823231238.GL8667@mathom.us> MIME-version: 1.0 Content-type: text/plain; charset=us-ascii; format=flowed Content-disposition: inline X-Pgp-Fingerprint: 53 FF 38 00 E7 DD 0A 9C 84 52 84 C5 EE DF 7C 88 X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mathom.us References: User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.044 required=5 tests=[AWL=-0.006, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/360 X-Sequence-Number: 14107 On Tue, Aug 23, 2005 at 05:29:01PM -0400, Jignesh Shah wrote: >Actually some of that readaheads,etc the OS does already if it does >some sort of throttling/clubbing of reads/writes. Note that I specified the fully cached case--even with the workload in RAM the system still has to process a heck of a lot of read calls. >* Introduce a multiblock or extent tunable variable where you can >define a multiple of 8K (or BlockSize tuneable) to read a bigger chunk >and store it in the bufferpool.. (Maybe writes too) (Most devices now >support upto 1MB chunks for reads and writes) Yeah. The problem with relying on OS readahead is that the OS doesn't know whether you're doing a sequential scan or an index scan; if you have the OS agressively readahead you'll kill your seek performance. OTOH, if you don't do readaheads you'll kill your sequential scan performance. At the app level you know which makes sense for each operation. Mike Stone From pgsql-performance-owner@postgresql.org Tue Aug 23 20:24:33 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AED61D70DC for ; Tue, 23 Aug 2005 20:24:30 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 56228-03 for ; Tue, 23 Aug 2005 23:24:27 +0000 (GMT) Received: from vms048pub.verizon.net (vms048pub.verizon.net [206.46.252.48]) by svr1.postgresql.org (Postfix) with ESMTP id 56816D702B for ; Tue, 23 Aug 2005 20:24:26 -0300 (ADT) Received: from osgiliath.mathom.us ([151.200.14.55]) by vms048.mailsrvcs.net (Sun Java System Messaging Server 6.2 HotFix 0.04 (built Dec 24 2004)) with ESMTPA id <0ILP00F4C7OO1K72@vms048.mailsrvcs.net> for pgsql-performance@postgresql.org; Tue, 23 Aug 2005 18:24:25 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by osgiliath.mathom.us (Postfix) with ESMTP id A546E60067A; Tue, 23 Aug 2005 19:24:24 -0400 (EDT) Received: from osgiliath.mathom.us ([127.0.0.1]) by localhost (osgiliath [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 07726-01-7; Tue, 23 Aug 2005 19:24:24 -0400 (EDT) Received: by osgiliath.mathom.us (Postfix, from userid 1000) id 83C5D600162; Tue, 23 Aug 2005 19:24:24 -0400 (EDT) Date: Tue, 23 Aug 2005 19:24:24 -0400 From: Michael Stone Subject: Re: Read/Write block sizes In-reply-to: <60k6icpapm.fsf@dba2.int.libertyrms.com> To: Chris Browne Cc: pgsql-performance@postgresql.org Mail-Followup-To: Chris Browne , pgsql-performance@postgresql.org Message-id: <20050823232424.GM8667@mathom.us> MIME-version: 1.0 Content-type: text/plain; charset=us-ascii; format=flowed Content-disposition: inline X-Pgp-Fingerprint: 53 FF 38 00 E7 DD 0A 9C 84 52 84 C5 EE DF 7C 88 X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mathom.us References: <60k6icpapm.fsf@dba2.int.libertyrms.com> User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.042 required=5 tests=[AWL=-0.008, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/361 X-Sequence-Number: 14108 On Tue, Aug 23, 2005 at 06:09:09PM -0400, Chris Browne wrote: >What we have been finding, as RAID controllers get smarter, is that it >is getting increasingly futile to try to attach knobs to 'disk stuff;' >it is *way* more effective to add a few more spindles to an array than >it is to fiddle with which disks are to be allocated to what database >'objects.' That statement doesn't say anything about trying to maximize performance to or from a disk array. Yes, controllers are getting smarter--but they aren't omnicient. IME an I/O bound sequential table scan doesn't get data moving off the disk nearly as fast as say, a dd with a big ibs. Why? There's obviously a lot of factors at work, but one of those factors is that the raid controller can optimize "grab this meg" a lot more than it can optimize "grab this 8k". Mike Stone From pgsql-performance-owner@postgresql.org Tue Aug 23 20:29:50 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 79A8DD7217 for ; Tue, 23 Aug 2005 20:29:47 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 53111-09 for ; Tue, 23 Aug 2005 23:29:43 +0000 (GMT) Received: from boutiquenumerique.com (boutiquenumerique.com [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 70334D714A for ; Tue, 23 Aug 2005 20:29:41 -0300 (ADT) Received: (qmail 32007 invoked from network); 24 Aug 2005 01:30:07 +0200 Received: from unknown (HELO localhost) (boutiquenumerique-lists@192.168.0.4) by boutiquenumerique.com with SMTP; 24 Aug 2005 01:30:07 +0200 Date: Wed, 24 Aug 2005 01:29:42 +0200 To: "William Yu" , pgsql-performance@postgresql.org Subject: Re: Caching by Postgres References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> From: PFC Organization: =?iso-8859-15?Q?La_Boutique_Num=E9rique?= Content-Type: text/plain; format=flowed; delsp=yes; charset=iso-8859-15 MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Message-ID: In-Reply-To: User-Agent: Opera M2/8.0 (Linux, build 1095) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 required=5 tests=[AWL=0.005] X-Spam-Level: X-Archive-Number: 200508/362 X-Sequence-Number: 14109 > Josh Berkus has already mentioned this as conventional wisdom as written > by Oracle. This may also be legacy wisdom. Oracle/Sybase/etc has been > around for a long time; it was probably a clear performance win way back > when. Nowadays with how far open-source OS's have advanced, I'd take it > with a grain of salt and do my own performance analysis. I suspect the > big vendors wouldn't change their stance even if they knew it was no > longer true due to the support hassles. Reinvent a filesystem... that would be suicidal. Now, Hans Reiser has expressed interest on the ReiserFS list in tweaking his Reiser4 especially for Postgres. In his own words, he wants a "Killer app for reiser4". Reiser4 will offser transactional semantics via a special reiser4 syscall, so it might be possible, with a minimum of changes to postgres (ie maybe just another sync mode besides fsync, fdatasync et al) to use this. Other interesting details were exposed on the reiser list, too (ie. a transactional filesystems can give ACID guarantees to postgres without the need for fsync()). Very interesting. From pgsql-performance-owner@postgresql.org Tue Aug 23 20:36:25 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4C4B0D6FFB for ; Tue, 23 Aug 2005 20:36:23 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 56748-05 for ; Tue, 23 Aug 2005 23:36:20 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [67.100.216.10]) by svr1.postgresql.org (Postfix) with ESMTP id 6EF96D7157 for ; Tue, 23 Aug 2005 20:36:18 -0300 (ADT) Received: by flake.decibel.org (Postfix, from userid 1001) id DD23E15289; Tue, 23 Aug 2005 18:36:08 -0500 (CDT) Date: Tue, 23 Aug 2005 18:36:08 -0500 From: "Jim C. Nasby" To: Chris Browne Cc: pgsql-performance@postgresql.org Subject: Re: Read/Write block sizes Message-ID: <20050823233608.GA94425@pervasive.com> References: <60k6icpapm.fsf@dba2.int.libertyrms.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <60k6icpapm.fsf@dba2.int.libertyrms.com> X-Operating-System: FreeBSD 4.11-RELEASE-p10 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/363 X-Sequence-Number: 14110 On Tue, Aug 23, 2005 at 06:09:09PM -0400, Chris Browne wrote: > J.K.Shah@Sun.COM (Jignesh Shah) writes: > >> Does that include increasing the size of read/write blocks? I've > >> noticedthat with a large enough table it takes a while to do a > >> sequential scan, even if it's cached; I wonder if the fact that it > >> takes a million read(2) calls to get through an 8G table is part of > >> that. > > > > Actually some of that readaheads,etc the OS does already if it does > > some sort of throttling/clubbing of reads/writes. But its not enough > > for such types of workloads. > > > > Here is what I think will help: > > > > * Support for different Blocksize TABLESPACE without recompiling the > > code.. (Atlease support for a different Blocksize for the whole > > database without recompiling the code) > > > > * Support for bigger sizes of WAL files instead of 16MB files > > WITHOUT recompiling the code.. Should be a tuneable if you ask me > > (with checkpoint_segments at 256.. you have too many 16MB files in > > the log directory) (This will help OLTP benchmarks more since now > > they don't spend time rotating log files) > > > > * Introduce a multiblock or extent tunable variable where you can > > define a multiple of 8K (or BlockSize tuneable) to read a bigger > > chunk and store it in the bufferpool.. (Maybe writes too) (Most > > devices now support upto 1MB chunks for reads and writes) > > > > *There should be a way to preallocate files for TABLES in > > TABLESPACES otherwise with multiple table writes in the same > > filesystem ends with fragmented files which causes poor "READS" from > > the files. > > > > * With 64bit 1GB file chunks is also moot.. Maybe it should be > > tuneable too like 100GB without recompiling the code. > > > > Why recompiling is bad? Most companies that will support Postgres > > will support their own binaries and they won't prefer different > > versions of binaries for different blocksizes, different WAL file > > sizes, etc... and hence more function using the same set of binaries > > is more desirable in enterprise environments > > Every single one of these still begs the question of whether the > changes will have a *material* impact on performance. How many of these things are currently easy to change with a recompile? I should be able to start testing some of these ideas in the near future, if they only require minor code or configure changes. -- Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com Pervasive Software http://pervasive.com 512-569-9461 From pgsql-performance-owner@postgresql.org Tue Aug 23 20:44:56 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 63494D71EC for ; Tue, 23 Aug 2005 20:44:16 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 59901-03 for ; Tue, 23 Aug 2005 23:44:15 +0000 (GMT) Received: from gghcwest.com (adsl-71-128-90-172.dsl.pltn13.pacbell.net [71.128.90.172]) by svr1.postgresql.org (Postfix) with ESMTP id 80621D714E for ; Tue, 23 Aug 2005 20:44:13 -0300 (ADT) Received: from toonses.gghcwest.com (toonses.gghcwest.com [192.168.168.115]) by gghcwest.com (8.12.10/8.12.9) with ESMTP id j7NNi8md020032; Tue, 23 Aug 2005 16:44:09 -0700 Received: from jwb by toonses.gghcwest.com with local (Exim 4.50) id 1E7iR8-00044n-RC; Tue, 23 Aug 2005 16:44:10 -0700 Subject: Re: Read/Write block sizes (Was: Caching by Postgres) From: "Jeffrey W. Baker" To: Michael Stone Cc: Jignesh Shah , pgsql-performance@postgresql.org In-Reply-To: <20050823231238.GL8667@mathom.us> References: <20050823231238.GL8667@mathom.us> Content-Type: text/plain Content-Transfer-Encoding: 7bit Date: Tue, 23 Aug 2005 16:44:10 -0700 Message-Id: <1124840650.12932.1.camel@toonses.gghcwest.com> Mime-Version: 1.0 X-Mailer: Evolution 2.3.6.1 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=-0.714 required=5 tests=[AWL=-0.764, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/365 X-Sequence-Number: 14112 On Tue, 2005-08-23 at 19:12 -0400, Michael Stone wrote: > On Tue, Aug 23, 2005 at 05:29:01PM -0400, Jignesh Shah wrote: > >Actually some of that readaheads,etc the OS does already if it does > >some sort of throttling/clubbing of reads/writes. > > Note that I specified the fully cached case--even with the workload in > RAM the system still has to process a heck of a lot of read calls. > > >* Introduce a multiblock or extent tunable variable where you can > >define a multiple of 8K (or BlockSize tuneable) to read a bigger chunk > >and store it in the bufferpool.. (Maybe writes too) (Most devices now > >support upto 1MB chunks for reads and writes) > > Yeah. The problem with relying on OS readahead is that the OS doesn't > know whether you're doing a sequential scan or an index scan; if you > have the OS agressively readahead you'll kill your seek performance. > OTOH, if you don't do readaheads you'll kill your sequential scan > performance. At the app level you know which makes sense for each > operation. This is why we have MADVISE_RANDOM and MADVISE_SEQUENTIAL. -jwb From pgsql-performance-owner@postgresql.org Tue Aug 23 20:40:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 746D4D7261 for ; Tue, 23 Aug 2005 20:39:09 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 59110-01 for ; Tue, 23 Aug 2005 23:39:06 +0000 (GMT) Received: from mail.tecarta.com (66.238.115.135.ptr.us.xo.net [66.238.115.135]) by svr1.postgresql.org (Postfix) with ESMTP id 418FBD722D for ; Tue, 23 Aug 2005 20:39:05 -0300 (ADT) Received: from mail pickup service by mail.tecarta.com with Microsoft SMTPSVC; Tue, 23 Aug 2005 16:42:56 -0700 Received: from mail.tecarta.com ([192.168.160.2]) by mail.tecarta.com with Microsoft SMTPSVC(6.0.3790.0); Tue, 23 Aug 2005 16:42:52 -0700 Received: from barracuda.tecarta.com ([192.168.160.200]) by mail.tecarta.com (SAVSMTP 3.1.0.29) with SMTP id M2005082316425204244 for ; Tue, 23 Aug 2005 16:42:52 -0700 X-ASG-Debug-ID: 1124840343-17072-0-0 X-Barracuda-URL: http://192.168.160.200:8000/cgi-bin/mark.cgi Received: from mail1 (mail1.hq.corp [192.168.160.5]) by barracuda.tecarta.com (Spam Firewall) with SMTP id 435AC20362F7 for ; Tue, 23 Aug 2005 16:39:03 -0700 (PDT) Received: from amd64-laptop-spoe ([63.206.203.145]) by mail.tecarta.com with Microsoft SMTPSVC(6.0.3790.1830); Tue, 23 Aug 2005 16:42:49 -0700 X-ASG-Orig-Subj: Re: [PERFORM] Read/Write block sizes Subject: Re: Read/Write block sizes From: Steve Poe To: Chris Browne Cc: pgsql-performance@postgresql.org In-Reply-To: <60k6icpapm.fsf@dba2.int.libertyrms.com> References: <60k6icpapm.fsf@dba2.int.libertyrms.com> Content-Type: text/plain Date: Wed, 24 Aug 2005 01:25:43 +0000 Message-Id: <1124846743.12045.94.camel@amd64-laptop-spoe> Mime-Version: 1.0 X-Mailer: Evolution 2.2.3 Content-Transfer-Encoding: 7bit X-OriginalArrivalTime: 23 Aug 2005 23:42:49.0984 (UTC) FILETIME=[5F45C800:01C5A83C] X-Virus-Scanned: by Barracuda Spam Firewall at tecarta.com X-Barracuda-Spam-Score: 2.00 X-Barracuda-Spam-Status: No, SCORE=2.00 using global scores of TAG_LEVEL=4.0 QUARANTINE_LEVEL=1000.0 KILL_LEVEL=7.0 tests=BAYES_80 X-Barracuda-Spam-Report: Code version 3.02, rules version 3.0.3434 Rule breakdown below pts rule name description ---- ---------------------- -------------------------------------------------- 2.00 BAYES_80 BODY: Bayesian spam probability is 80 to 95% [score: 0.8081] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.205 required=5 tests=[AWL=0.155, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/364 X-Sequence-Number: 14111 Chris, Unless I am wrong, you're making the assumpting the amount of time spent and ROI is known. Maybe those who've been down this path know how to get that additional 2-4% in 30 minutes or less? While each person and business' performance gains (or not) could vary, someone spending the 50-100h to gain 2-4% over a course of a month for a 24x7 operation would seem worth the investment? I would assume that dbt2 with STP helps minimize the amount of hours someone has to invest to determine performance gains with configurable options? Steve Poe > If someone spends 100h working on one of these items, and gets a 2% > performance improvement, that's almost certain to be less desirable > than spending 50h on something else that gets a 4% improvement. > > And we might discover that memory management improvements in Linux > 2.6.16 or FreeBSD 5.5 allow some OS kernels to provide some such > improvements "for free" behind our backs without *any* need to write > database code. :-) From pgsql-performance-owner@postgresql.org Tue Aug 23 23:11:57 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 207F1D725F for ; Tue, 23 Aug 2005 23:10:58 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 02461-02 for ; Wed, 24 Aug 2005 02:10:56 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id D6914D71FC for ; Tue, 23 Aug 2005 23:10:49 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [63.195.55.98] (account josh@agliodbs.com HELO spooky) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7919635; Tue, 23 Aug 2005 19:13:11 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Caching by Postgres Date: Tue, 23 Aug 2005 19:11:58 -0700 User-Agent: KMail/1.8 Cc: PFC , "William Yu" References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-15" Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Message-Id: <200508231911.59100.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.011 required=5 tests=[AWL=0.011] X-Spam-Level: X-Archive-Number: 200508/366 X-Sequence-Number: 14113 PFC, > =A0=A0=A0=A0=A0=A0=A0=A0Now, Hans Reiser has expressed interest on the Re= iserFS list in > tweaking =A0 his Reiser4 especially for Postgres. In his own words, he wa= nts > a "Killer app for reiser4". Reiser4 will offser transactional semantics v= ia > a special reiser4 syscall, so it might be possible, with a minimum of > changes to postgres (ie maybe just another sync mode besides fsync, > fdatasync et al) to use this. Other interesting details were exposed on t= he > reiser list, too (ie. a transactional filesystems can give ACID guarantees > to postgres without the need for fsync()). Really? Cool, I'd like to see that. Could you follow up with Hans? Or g= ive=20 me his e-mail? =2D-=20 Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Tue Aug 23 23:26:23 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A14A2D70DC for ; Tue, 23 Aug 2005 23:25:58 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 09234-09 for ; Wed, 24 Aug 2005 02:25:56 +0000 (GMT) Received: from brmea-mail-4.sun.com (brmea-mail-4.Sun.COM [192.18.98.36]) by svr1.postgresql.org (Postfix) with ESMTP id C82D2D70CF for ; Tue, 23 Aug 2005 23:25:54 -0300 (ADT) Received: from phys-bur-1 ([129.148.9.72]) by brmea-mail-4.sun.com (8.12.10/8.12.9) with ESMTP id j7O2PwTa016676 for ; Tue, 23 Aug 2005 20:25:59 -0600 (MDT) Received: from conversion-daemon.bur-mail1.east.sun.com by bur-mail1.east.sun.com (iPlanet Messaging Server 5.2 HotFix 1.24 (built Dec 19 2003)) id <0ILP00H01G2OKS@bur-mail1.east.sun.com> (original mail from J.K.Shah@Sun.COM) for pgsql-performance@postgresql.org; Tue, 23 Aug 2005 22:25:58 -0400 (EDT) Received: from [129.150.32.236] (vpn-129-150-32-236.Central.Sun.COM [129.150.32.236]) by bur-mail1.east.sun.com (iPlanet Messaging Server 5.2 HotFix 1.24 (built Dec 19 2003)) with ESMTPA id <0ILP0092LG3910@bur-mail1.east.sun.com>; Tue, 23 Aug 2005 22:25:58 -0400 (EDT) Date: Tue, 23 Aug 2005 22:22:04 -0400 From: "Jignesh K. Shah" Subject: Re: Read/Write block sizes In-reply-to: <20050823233608.GA94425@pervasive.com> To: "Jim C. Nasby" Cc: Chris Browne , pgsql-performance@postgresql.org Message-id: <430BD9CC.3030306@sun.com> MIME-version: 1.0 Content-type: text/plain; charset=ISO-8859-1; format=flowed Content-transfer-encoding: 7BIT X-Accept-Language: en-us, en User-Agent: Mozilla Thunderbird 1.0.2 (Windows/20050317) References: <60k6icpapm.fsf@dba2.int.libertyrms.com> <20050823233608.GA94425@pervasive.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/367 X-Sequence-Number: 14114 Hi Jim, | How many of these things are currently easy to change with a recompile? | I should be able to start testing some of these ideas in the near | future, if they only require minor code or configure changes. The following * Data File Size 1GB * WAL File Size of 16MB * Block Size of 8K Are very easy to change with a recompile.. A Tunable will be greatly prefered as it will allow one binary for different tunings * MultiBlock read/write Is not available but will greatly help in reducing the number of system calls which will only increase as the size of the database increases if something is not done about i. * Pregrown files... maybe not important at this point since TABLESPACE can currently work around it a bit (Just need to create a different file system for each tablespace But if you really think hardware & OS is the answer for all small things...... I think we should now start to look on how to make Postgres Multi-threaded or multi-processed for each connection. With the influx of "Dual-Core" or "Multi-Core" being the fad.... Postgres can have the cutting edge if somehow exploiting cores is designed. Somebody mentioned that adding CPU to Postgres workload halved the average CPU usage... YEAH... PostgreSQL uses only 1 CPU per connection (assuming 100% usage) so if you add another CPU it is idle anyway and the system will report only 50% :-) BUT the importing to measure is.. whether the query time was cut down or not? ( No flames I am sure you were talking about multi-connection multi-user environment :-) ) But my point is then this approach is worth the ROI and the time and effort spent to solve this problem. I actually vote for a multi-threaded solution for each connection while still maintaining seperate process for each connections... This way the fundamental architecture of Postgres doesn't change, however a multi-threaded connection can then start to exploit different cores.. (Maybe have tunables for number of threads to read data files who knows.. If somebody is interested in actually working a design .. contact me and I will be interested in assisting this work. Regards, Jignesh Jim C. Nasby wrote: >On Tue, Aug 23, 2005 at 06:09:09PM -0400, Chris Browne wrote: > > >>J.K.Shah@Sun.COM (Jignesh Shah) writes: >> >> >>>>Does that include increasing the size of read/write blocks? I've >>>>noticedthat with a large enough table it takes a while to do a >>>>sequential scan, even if it's cached; I wonder if the fact that it >>>>takes a million read(2) calls to get through an 8G table is part of >>>>that. >>>> >>>> >>>Actually some of that readaheads,etc the OS does already if it does >>>some sort of throttling/clubbing of reads/writes. But its not enough >>>for such types of workloads. >>> >>>Here is what I think will help: >>> >>>* Support for different Blocksize TABLESPACE without recompiling the >>>code.. (Atlease support for a different Blocksize for the whole >>>database without recompiling the code) >>> >>>* Support for bigger sizes of WAL files instead of 16MB files >>>WITHOUT recompiling the code.. Should be a tuneable if you ask me >>>(with checkpoint_segments at 256.. you have too many 16MB files in >>>the log directory) (This will help OLTP benchmarks more since now >>>they don't spend time rotating log files) >>> >>>* Introduce a multiblock or extent tunable variable where you can >>>define a multiple of 8K (or BlockSize tuneable) to read a bigger >>>chunk and store it in the bufferpool.. (Maybe writes too) (Most >>>devices now support upto 1MB chunks for reads and writes) >>> >>>*There should be a way to preallocate files for TABLES in >>>TABLESPACES otherwise with multiple table writes in the same >>>filesystem ends with fragmented files which causes poor "READS" from >>>the files. >>> >>>* With 64bit 1GB file chunks is also moot.. Maybe it should be >>>tuneable too like 100GB without recompiling the code. >>> >>>Why recompiling is bad? Most companies that will support Postgres >>>will support their own binaries and they won't prefer different >>>versions of binaries for different blocksizes, different WAL file >>>sizes, etc... and hence more function using the same set of binaries >>>is more desirable in enterprise environments >>> >>> >>Every single one of these still begs the question of whether the >>changes will have a *material* impact on performance. >> >> From pgsql-performance-owner@postgresql.org Tue Aug 23 23:31:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D1DB9D71F9 for ; Tue, 23 Aug 2005 23:30:23 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 41174-03 for ; Wed, 24 Aug 2005 02:30:21 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id 1309DD70DC for ; Tue, 23 Aug 2005 23:30:20 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [63.195.55.98] (account josh@agliodbs.com HELO spooky) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7919862; Tue, 23 Aug 2005 19:32:42 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Read/Write block sizes Date: Tue, 23 Aug 2005 19:31:29 -0700 User-Agent: KMail/1.8 Cc: Steve Poe , Chris Browne References: <60k6icpapm.fsf@dba2.int.libertyrms.com> <1124846743.12045.94.camel@amd64-laptop-spoe> In-Reply-To: <1124846743.12045.94.camel@amd64-laptop-spoe> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200508231931.29682.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.011 required=5 tests=[AWL=0.011] X-Spam-Level: X-Archive-Number: 200508/368 X-Sequence-Number: 14115 Steve, > I would assume that dbt2 with STP helps minimize the amount of hours > someone has to invest to determine performance gains with configurable > options? Actually, these I/O operation issues show up mainly with DW workloads, so the STP isn't much use there. If I can ever get some of these machines back from the build people, I'd like to start testing some stuff. One issue with testing this is that currently PostgreSQL doesn't support block sizes above 128K. We've already done testing on that (well, Mark has) and the performance gains aren't even worth the hassle of remembering you're on a different block size (like, +4%). What the Sun people have done with other DB systems is show that substantial performance gains are possible on large databases (>100G) using block sizes of 1MB. I believe that's possible (and that it probably makes more of a difference on Solaris than on BSD) but we can't test it without some hackery first. -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Wed Aug 24 01:44:15 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 258C8D7267 for ; Wed, 24 Aug 2005 01:44:11 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83728-08 for ; Wed, 24 Aug 2005 04:44:07 +0000 (GMT) Received: from mail21.sea5.speakeasy.net (mail21.sea5.speakeasy.net [69.17.117.23]) by svr1.postgresql.org (Postfix) with ESMTP id 811B0D7012 for ; Wed, 24 Aug 2005 00:06:42 -0300 (ADT) Received: (qmail 23271 invoked from network); 24 Aug 2005 03:06:37 -0000 Received: from dsl081-060-184.sfo1.dsl.speakeasy.net (HELO noodles) ([64.81.60.184]) (envelope-sender ) by mail21.sea5.speakeasy.net (qmail-ldap-1.03) with RC4-MD5 encrypted SMTP for ; 24 Aug 2005 03:06:37 -0000 Subject: Re: Read/Write block sizes From: "Jeffrey W. Baker" To: Josh Berkus Cc: pgsql-performance@postgresql.org, Steve Poe , Chris Browne In-Reply-To: <200508231931.29682.josh@agliodbs.com> References: <60k6icpapm.fsf@dba2.int.libertyrms.com> <1124846743.12045.94.camel@amd64-laptop-spoe> <200508231931.29682.josh@agliodbs.com> Content-Type: text/plain Date: Tue, 23 Aug 2005 20:07:34 -0700 Message-Id: <1124852854.11147.5.camel@noodles> Mime-Version: 1.0 X-Mailer: Evolution 2.2.1.1 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/371 X-Sequence-Number: 14118 On Tue, 2005-08-23 at 19:31 -0700, Josh Berkus wrote: > Steve, > > > I would assume that dbt2 with STP helps minimize the amount of hours > > someone has to invest to determine performance gains with configurable > > options? > > Actually, these I/O operation issues show up mainly with DW workloads, so the > STP isn't much use there. If I can ever get some of these machines back > from the build people, I'd like to start testing some stuff. > > One issue with testing this is that currently PostgreSQL doesn't support block > sizes above 128K. We've already done testing on that (well, Mark has) and > the performance gains aren't even worth the hassle of remembering you're on a > different block size (like, +4%). > > What the Sun people have done with other DB systems is show that substantial > performance gains are possible on large databases (>100G) using block sizes > of 1MB. I believe that's possible (and that it probably makes more of a > difference on Solaris than on BSD) but we can't test it without some hackery > first. To get decent I/O you need 1MB fundamental units all the way down the stack. You need a filesystem that can take a 1MB write well, and you need an I/O scheduler that will keep it together, and you need a storage controller that can eat a 1MB request at once. Ideally you'd like an architecture with a 1MB page (Itanium has this, and AMD64 Linux will soon have this.) The Lustre people have done some work in this area, opening up the datapaths in the kernel so they can keep the hardware really working. They even modified the QLogic SCSI/FC driver so it supports such large transfers. Their work has shown that you can get significant perf boost on Linux just by thinking in terms of larger transfers. Unfortunately I'm really afraid that this conversation is about trees when the forest is the problem. PostgreSQL doesn't even have an async reader, which is the sort of thing that could double or triple its performance. You're talking about block sizes and such, but the kinds of improvements you can get there are in the tens of percents at most. -jwb From pgsql-performance-owner@postgresql.org Wed Aug 24 01:22:37 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 71784D70DC for ; Wed, 24 Aug 2005 01:22:33 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 31461-10 for ; Wed, 24 Aug 2005 04:22:30 +0000 (GMT) Received: from unicorn.rentec.com (unicorn.rentec.com [216.223.240.9]) by svr1.postgresql.org (Postfix) with ESMTP id C598DD710D for ; Wed, 24 Aug 2005 01:22:28 -0300 (ADT) Received: from wren.rentec.com (wren.rentec.com [192.5.35.106]) by unicorn.rentec.com (8.13.1/8.12.1) with ESMTP id j7O4LLYC017574 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Wed, 24 Aug 2005 00:21:22 -0400 (EDT) X-Rentec: external Received: from [172.16.160.108] (stange-dhcp2.rentec.com [172.16.160.108]) by wren.rentec.com (8.13.1/8.12.1) with ESMTP id j7O4LIla011044; Wed, 24 Aug 2005 00:21:18 -0400 (EDT) Message-ID: <430BF0DA.5010002@rentec.com> Date: Wed, 24 Aug 2005 00:00:26 -0400 From: Alan Stange Organization: Renaissance Technologies Corp. User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Josh Berkus Cc: pgsql-performance@postgresql.org, Steve Poe , Chris Browne Subject: Re: Read/Write block sizes References: <60k6icpapm.fsf@dba2.int.libertyrms.com> <1124846743.12045.94.camel@amd64-laptop-spoe> <200508231931.29682.josh@agliodbs.com> In-Reply-To: <200508231931.29682.josh@agliodbs.com> Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit X-Logged: Logged by unicorn.rentec.com as j7O4LLYC017574 at Wed Aug 24 00:21:22 2005 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/369 X-Sequence-Number: 14116 Josh Berkus wrote: >Steve, > > > >>I would assume that dbt2 with STP helps minimize the amount of hours >>someone has to invest to determine performance gains with configurable >>options? >> >> > >Actually, these I/O operation issues show up mainly with DW workloads, so the >STP isn't much use there. If I can ever get some of these machines back >from the build people, I'd like to start testing some stuff. > >One issue with testing this is that currently PostgreSQL doesn't support block >sizes above 128K. We've already done testing on that (well, Mark has) and >the performance gains aren't even worth the hassle of remembering you're on a >different block size (like, +4%). > > What size database was this on? >What the Sun people have done with other DB systems is show that substantial >performance gains are possible on large databases (>100G) using block sizes >of 1MB. I believe that's possible (and that it probably makes more of a >difference on Solaris than on BSD) but we can't test it without some hackery >first. > We're running on a 100+GB database, with long streams of 8KB reads with the occasional _llseek(). I've been thinking about running with a larger blocksize with the expectation that we'd see fewer system calls and a bit more throughput. read() calls are a very expensive way to get 8KB of memory (that we know is already resident) during scans. One has to trap into the kernel, do the usual process state accounting, find the block, copy the memory to userspace, return back from the kernel to user space reversing all the process accounting, pick out the bytes one needs, and repeat all over again. That's quite a few sacrificial cache lines for 8KB. Yeah, sure, Linux syscalls are fast, but they aren't that fast, and other operating systems (windows and solaris) have a bit more overhead on syscalls. Regarding large blocks sizes on Solaris: the Solaris folks can also use large memory pages and avoid a lot of the TLB overhead from the VM system. The various trapstat and cpustat commands can be quite interesting to look at when running any large application on a Solaris system. It should be noted that having a large shared memory segment can be a performance looser just from the standpoint of TLB thrashing. O(GB) memory access patterns can take a huge performance hit in user space with 4K pages compared to the kernel which would be mapping the "segmap" (in Solaris parlance) with 4MB pages. Anyway, I guess my point is that the balance between kernel managed vs. postgresql managed buffer isn't obvious at all. -- Alan From pgsql-performance-owner@postgresql.org Wed Aug 24 01:38:25 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A9573D7279 for ; Wed, 24 Aug 2005 01:38:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 82631-05 for ; Wed, 24 Aug 2005 04:38:15 +0000 (GMT) Received: from linuxworld.com.au (unknown [203.34.46.50]) by svr1.postgresql.org (Postfix) with ESMTP id BDD83D7116 for ; Wed, 24 Aug 2005 01:38:12 -0300 (ADT) Received: from linuxworld.com.au (IDENT:swm@localhost.localdomain [127.0.0.1]) by linuxworld.com.au (8.13.2/8.13.2) with ESMTP id j7O4c3Aj005242; Wed, 24 Aug 2005 14:38:04 +1000 Received: from localhost (swm@localhost) by linuxworld.com.au (8.13.2/8.13.2/Submit) with ESMTP id j7O4c2Ef005239; Wed, 24 Aug 2005 14:38:03 +1000 X-Authentication-Warning: linuxworld.com.au: swm owned process doing -bs Date: Wed, 24 Aug 2005 14:38:02 +1000 (EST) From: Gavin Sherry X-X-Sender: swm@linuxworld.com.au To: PFC Cc: William Yu , pgsql-performance@postgresql.org Subject: Re: Caching by Postgres In-Reply-To: Message-ID: References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.002 required=5 tests=[AWL=0.002] X-Spam-Level: X-Archive-Number: 200508/370 X-Sequence-Number: 14117 On Wed, 24 Aug 2005, PFC wrote: > > > Josh Berkus has already mentioned this as conventional wisdom as written > > by Oracle. This may also be legacy wisdom. Oracle/Sybase/etc has been > > around for a long time; it was probably a clear performance win way back > > when. Nowadays with how far open-source OS's have advanced, I'd take it > > with a grain of salt and do my own performance analysis. I suspect the > > big vendors wouldn't change their stance even if they knew it was no > > longer true due to the support hassles. > > Reinvent a filesystem... that would be suicidal. > > Now, Hans Reiser has expressed interest on the ReiserFS list in tweaking > his Reiser4 especially for Postgres. In his own words, he wants a "Killer > app for reiser4". Reiser4 will offser transactional semantics via a > special reiser4 syscall, so it might be possible, with a minimum of > changes to postgres (ie maybe just another sync mode besides fsync, > fdatasync et al) to use this. Other interesting details were exposed on > the reiser list, too (ie. a transactional filesystems can give ACID > guarantees to postgres without the need for fsync()). > > Very interesting. Ummm... I don't see anything here which will be a win for Postgres. The transactional semantics we're interested in are fairly complex: 1) Modifications to multiple objects can become visible to the system atomically 2) On error, a series of modifications which had been grouped together within a transaction can be rolled back 3) Using object version information, determine which version of which object is visible to a given session 4) Using version information and locking, detect and resolve read/write and write/write conflicts Now, I can see a file system offering (1) and (2). But a file system that can allow people to do (3) and (4) would require that we make *major* modifications to how postgresql is implemented. More over, it would be for no gain, since we've already written a system which can do it. A filesystem could, in theory, help us by providing an API which allows us to tell the file system either: the way we'd like it to read ahead, the fact that we don't want it to read ahead or the way we'd like it to cache (or not cache) data. The thing is, most OSes provide interfaces to do this already and we make only little use of them (I'm think of madv_sequential(), madv_random(), POSIX fadvise(), the various flags to open() which AIX, HPUX, Solaris provide). Gavin From pgsql-performance-owner@postgresql.org Wed Aug 24 02:07:41 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C7DA3D7012 for ; Wed, 24 Aug 2005 02:07:39 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 33799-04 for ; Wed, 24 Aug 2005 05:07:34 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id C1D4ED6DC2 for ; Wed, 24 Aug 2005 02:07:32 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7O57Mw2021393; Wed, 24 Aug 2005 01:07:22 -0400 (EDT) To: Gavin Sherry Cc: PFC , William Yu , pgsql-performance@postgresql.org Subject: Re: Caching by Postgres In-reply-to: References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> Comments: In-reply-to Gavin Sherry message dated "Wed, 24 Aug 2005 14:38:02 +1000" Date: Wed, 24 Aug 2005 01:07:22 -0400 Message-ID: <21392.1124860042@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/372 X-Sequence-Number: 14119 Gavin Sherry writes: > A filesystem could, in theory, help us by providing an API which allows us > to tell the file system either: the way we'd like it to read ahead, the > fact that we don't want it to read ahead or the way we'd like it to cache > (or not cache) data. The thing is, most OSes provide interfaces to do this > already and we make only little use of them (I'm think of > madv_sequential(), madv_random(), POSIX fadvise(), the various flags to > open() which AIX, HPUX, Solaris provide). Yeah ... the main reason we've not spent too much time on that sort of stuff is that *it's not portable*. And with all due respect to Hans, special tweaks for one filesystem are even less interesting than special tweaks for one OS. regards, tom lane From pgsql-performance-owner@postgresql.org Wed Aug 24 02:13:25 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E7963D7319 for ; Wed, 24 Aug 2005 02:10:30 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 33871-08 for ; Wed, 24 Aug 2005 05:10:28 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id D8EDAD731E for ; Wed, 24 Aug 2005 02:10:27 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7O5AKam021416; Wed, 24 Aug 2005 01:10:20 -0400 (EDT) To: "Jeffrey W. Baker" Cc: Josh Berkus , pgsql-performance@postgresql.org, Steve Poe , Chris Browne Subject: Re: Read/Write block sizes In-reply-to: <1124852854.11147.5.camel@noodles> References: <60k6icpapm.fsf@dba2.int.libertyrms.com> <1124846743.12045.94.camel@amd64-laptop-spoe> <200508231931.29682.josh@agliodbs.com> <1124852854.11147.5.camel@noodles> Comments: In-reply-to "Jeffrey W. Baker" message dated "Tue, 23 Aug 2005 20:07:34 -0700" Date: Wed, 24 Aug 2005 01:10:20 -0400 Message-ID: <21415.1124860220@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/373 X-Sequence-Number: 14120 "Jeffrey W. Baker" writes: > To get decent I/O you need 1MB fundamental units all the way down the > stack. It would also be a good idea to have an application that isn't likely to change a single bit in a 1MB range and then expect you to record that change. This pretty much lets Postgres out of the picture. regards, tom lane From pgsql-performance-owner@postgresql.org Wed Aug 24 02:20:40 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id F0D24D6F0F for ; Wed, 24 Aug 2005 02:20:39 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 50726-08 for ; Wed, 24 Aug 2005 05:20:35 +0000 (GMT) Received: from serv-0.esphion.com (serv-0.esphion.com [202.55.97.148]) by svr1.postgresql.org (Postfix) with ESMTP id 7B240D6EA5 for ; Wed, 24 Aug 2005 02:20:29 -0300 (ADT) Received: from conker.alb-nz.esphion.com ([10.0.1.137] helo=conker.esphion.com) by serv-0.esphion.com with smtp (Exim 4.50) id 1E7ngW-0000K9-1g; Wed, 24 Aug 2005 17:20:24 +1200 Received: (nullmailer pid 1553 invoked by uid 10001); Wed, 24 Aug 2005 05:20:24 -0000 Date: Wed, 24 Aug 2005 17:20:23 +1200 From: Guy Thornley To: "Jeffrey W. Baker" Cc: Josh Berkus , pgsql-performance@postgresql.org, Steve Poe , Chris Browne Subject: Re: Read/Write block sizes Message-ID: <20050824052023.GB31806@conker.esphion.com> Reply-To: Guy Thornley References: <60k6icpapm.fsf@dba2.int.libertyrms.com> <1124846743.12045.94.camel@amd64-laptop-spoe> <200508231931.29682.josh@agliodbs.com> <1124852854.11147.5.camel@noodles> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1124852854.11147.5.camel@noodles> User-Agent: Mutt/1.5.6+20040907i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/374 X-Sequence-Number: 14121 > Unfortunately I'm really afraid that this conversation is about trees > when the forest is the problem. PostgreSQL doesn't even have an async > reader, which is the sort of thing that could double or triple its > performance. You're talking about block sizes and such, but the kinds > of improvements you can get there are in the tens of percents at most. Not 100% sure, but I'm fairly cirtain we were seeing significant performance degradation by too much _scheduled_ I/O activity ie: too much work being submitted to the kernel, due to excessive parallelism already!! The classic example of this is a seqscan being interleved by a index scan, and the disks end up doing nothing but seek activity Out of all the stuff talked about on this thread so far, only tweaking the block size (and the madvise() stuff) makes any real-world sense, as its the only thing talked about that increases the _work_per_seek_. As for the async IO, sure you might think 'oh async IO would be so cool!!' and I did, once, too. But then I sat down and _thought_ about it, and decided well, no, actually, theres _very_ few areas it could actually help, and in most cases it just make it easier to drive your box into lseek() induced IO collapse. Dont forget that already in postgres, you have a process per connection, and all the processes take care of their own I/O. Somebody mentioned having threaded backends too, but the only benefit would be reduced memory footprint (a backend consumes 1-2MB of RAM, which is almost enough to be a concern for largish systems with a lot of backends) but personally I _know_ the complixities introduced through threading are usually not worth it. IMVVHO (naive experience) what is needed is a complete architecture change (probably infeasible and only useful as a thought experiment), where: * a network I/O process deals with client connections * a limited pool of worker processes deal with statements (perhaps related to number of spindles somehow) so when a client issues a statement, the net-IO process simply forwards the connection state to a worker process and says 'deal with this'. (Clearly the state object needs to contain all user and transaction state the connection is involved in). - Guy Thornley From pgsql-performance-owner@postgresql.org Wed Aug 24 02:24:21 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 55C5DD710D for ; Wed, 24 Aug 2005 02:24:20 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 64207-08 for ; Wed, 24 Aug 2005 05:24:15 +0000 (GMT) Received: from linuxworld.com.au (unknown [203.34.46.50]) by svr1.postgresql.org (Postfix) with ESMTP id BFD1BD7093 for ; Wed, 24 Aug 2005 02:24:13 -0300 (ADT) Received: from linuxworld.com.au (IDENT:swm@localhost.localdomain [127.0.0.1]) by linuxworld.com.au (8.13.2/8.13.2) with ESMTP id j7O5O7Ja005621; Wed, 24 Aug 2005 15:24:07 +1000 Received: from localhost (swm@localhost) by linuxworld.com.au (8.13.2/8.13.2/Submit) with ESMTP id j7O5O75h005618; Wed, 24 Aug 2005 15:24:07 +1000 X-Authentication-Warning: linuxworld.com.au: swm owned process doing -bs Date: Wed, 24 Aug 2005 15:24:07 +1000 (EST) From: Gavin Sherry X-X-Sender: swm@linuxworld.com.au To: Tom Lane Cc: PFC , William Yu , pgsql-performance@postgresql.org Subject: Re: Caching by Postgres In-Reply-To: <21392.1124860042@sss.pgh.pa.us> Message-ID: References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> <21392.1124860042@sss.pgh.pa.us> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.002 required=5 tests=[AWL=0.002] X-Spam-Level: X-Archive-Number: 200508/375 X-Sequence-Number: 14122 On Wed, 24 Aug 2005, Tom Lane wrote: > Gavin Sherry writes: > > A filesystem could, in theory, help us by providing an API which allows us > > to tell the file system either: the way we'd like it to read ahead, the > > fact that we don't want it to read ahead or the way we'd like it to cache > > (or not cache) data. The thing is, most OSes provide interfaces to do this > > already and we make only little use of them (I'm think of > > madv_sequential(), madv_random(), POSIX fadvise(), the various flags to > > open() which AIX, HPUX, Solaris provide). > > Yeah ... the main reason we've not spent too much time on that sort of > stuff is that *it's not portable*. And with all due respect to Hans, > special tweaks for one filesystem are even less interesting than special > tweaks for one OS. Right. As an aside, it seems to me that if there is merit in all this low level interaction with the file system (not to mention the other platform specific microoptimisations which come up regularly on the lists) then the companies currently producing niche commercial releases of PostgreSQL should be taking advantage of them: if it increases performance, then there's a reason to buy as opposed to just downloading the OSS version. Gavin From pgsql-performance-owner@postgresql.org Wed Aug 24 02:24:29 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AFA38D7215 for ; Wed, 24 Aug 2005 02:24:28 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 77847-05 for ; Wed, 24 Aug 2005 05:24:25 +0000 (GMT) Received: from mail28.sea5.speakeasy.net (mail28.sea5.speakeasy.net [69.17.117.30]) by svr1.postgresql.org (Postfix) with ESMTP id 1A990D7268 for ; Wed, 24 Aug 2005 02:24:23 -0300 (ADT) Received: (qmail 5444 invoked from network); 24 Aug 2005 05:24:23 -0000 Received: from dsl081-060-184.sfo1.dsl.speakeasy.net (HELO noodles) ([64.81.60.184]) (envelope-sender ) by mail28.sea5.speakeasy.net (qmail-ldap-1.03) with RC4-MD5 encrypted SMTP for ; 24 Aug 2005 05:24:23 -0000 Subject: Re: Read/Write block sizes From: "Jeffrey W. Baker" To: Guy Thornley Cc: Josh Berkus , pgsql-performance@postgresql.org, Steve Poe , Chris Browne In-Reply-To: <20050824052023.GB31806@conker.esphion.com> References: <60k6icpapm.fsf@dba2.int.libertyrms.com> <1124846743.12045.94.camel@amd64-laptop-spoe> <200508231931.29682.josh@agliodbs.com> <1124852854.11147.5.camel@noodles> <20050824052023.GB31806@conker.esphion.com> Content-Type: text/plain Date: Tue, 23 Aug 2005 22:25:21 -0700 Message-Id: <1124861121.11270.1.camel@noodles> Mime-Version: 1.0 X-Mailer: Evolution 2.2.1.1 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/376 X-Sequence-Number: 14123 On Wed, 2005-08-24 at 17:20 +1200, Guy Thornley wrote: > As for the async IO, sure you might think 'oh async IO would be so cool!!' > and I did, once, too. But then I sat down and _thought_ about it, and > decided well, no, actually, theres _very_ few areas it could actually help, > and in most cases it just make it easier to drive your box into lseek() > induced IO collapse. > > Dont forget that already in postgres, you have a process per connection, and > all the processes take care of their own I/O. That's the problem. Instead you want 1 or 4 or 10 i/o slaves coordinating the I/O of all the backends optimally. For instance, with synchronous scanning. -jwb From pgsql-performance-owner@postgresql.org Wed Aug 24 02:57:33 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D91E0D725B for ; Wed, 24 Aug 2005 02:57:31 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 22522-05 for ; Wed, 24 Aug 2005 05:57:29 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 7F235D7233 for ; Wed, 24 Aug 2005 02:57:26 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7O5ui65021701; Wed, 24 Aug 2005 01:56:44 -0400 (EDT) To: "Jeffrey W. Baker" Cc: Guy Thornley , Josh Berkus , pgsql-performance@postgresql.org, Steve Poe , Chris Browne Subject: Re: Read/Write block sizes In-reply-to: <1124861121.11270.1.camel@noodles> References: <60k6icpapm.fsf@dba2.int.libertyrms.com> <1124846743.12045.94.camel@amd64-laptop-spoe> <200508231931.29682.josh@agliodbs.com> <1124852854.11147.5.camel@noodles> <20050824052023.GB31806@conker.esphion.com> <1124861121.11270.1.camel@noodles> Comments: In-reply-to "Jeffrey W. Baker" message dated "Tue, 23 Aug 2005 22:25:21 -0700" Date: Wed, 24 Aug 2005 01:56:44 -0400 Message-ID: <21700.1124863004@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/377 X-Sequence-Number: 14124 "Jeffrey W. Baker" writes: > On Wed, 2005-08-24 at 17:20 +1200, Guy Thornley wrote: >> Dont forget that already in postgres, you have a process per connection, and >> all the processes take care of their own I/O. > That's the problem. Instead you want 1 or 4 or 10 i/o slaves > coordinating the I/O of all the backends optimally. For instance, with > synchronous scanning. And why exactly are we going to do a better job of I/O scheduling than the OS itself can do? There's a fairly basic disconnect in viewpoint involved here. The old-school viewpoint (as embodied in Oracle and a few other DBMSes) is that the OS is too stupid to be worth anything, and the DB should bypass the OS to the greatest extent possible, doing its own caching, disk space layout, I/O scheduling, yadda yadda. That might have been defensible twenty-odd years ago when Oracle was designed. Postgres prefers to lay off to the OS anything that the OS can do well --- and that definitely includes caching and I/O scheduling. There are a whole lot of smart people working on those problems at the OS level. Maybe we could make marginal improvements on their results after spending a lot of effort reinventing the wheel ... but our time will be repaid much more if we work at levels that the OS cannot have knowledge of, such as join planning and data statistics. There are some things we could do to reduce the impedance between us and the OS --- for instance, the upthread criticism that a seqscan asks the OS for only 8K at a time is fair enough. But that doesn't translate to a conclusion that we should schedule the I/O instead of the OS. regards, tom lane From pgsql-performance-owner@postgresql.org Wed Aug 24 03:22:06 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 02CB1D6FBF for ; Wed, 24 Aug 2005 03:22:05 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 84569-05 for ; Wed, 24 Aug 2005 06:22:00 +0000 (GMT) Received: from mail28.sea5.speakeasy.net (mail28.sea5.speakeasy.net [69.17.117.30]) by svr1.postgresql.org (Postfix) with ESMTP id 1D59FD6E30 for ; Wed, 24 Aug 2005 03:21:54 -0300 (ADT) Received: (qmail 10794 invoked from network); 24 Aug 2005 06:21:54 -0000 Received: from dsl081-060-184.sfo1.dsl.speakeasy.net (HELO noodles) ([64.81.60.184]) (envelope-sender ) by mail28.sea5.speakeasy.net (qmail-ldap-1.03) with RC4-MD5 encrypted SMTP for ; 24 Aug 2005 06:21:54 -0000 Subject: Re: Read/Write block sizes From: "Jeffrey W. Baker" To: Tom Lane Cc: Guy Thornley , Josh Berkus , pgsql-performance@postgresql.org, Steve Poe , Chris Browne In-Reply-To: <21700.1124863004@sss.pgh.pa.us> References: <60k6icpapm.fsf@dba2.int.libertyrms.com> <1124846743.12045.94.camel@amd64-laptop-spoe> <200508231931.29682.josh@agliodbs.com> <1124852854.11147.5.camel@noodles> <20050824052023.GB31806@conker.esphion.com> <1124861121.11270.1.camel@noodles> <21700.1124863004@sss.pgh.pa.us> Content-Type: text/plain Date: Tue, 23 Aug 2005 23:22:52 -0700 Message-Id: <1124864572.11270.16.camel@noodles> Mime-Version: 1.0 X-Mailer: Evolution 2.2.1.1 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/378 X-Sequence-Number: 14125 On Wed, 2005-08-24 at 01:56 -0400, Tom Lane wrote: > "Jeffrey W. Baker" writes: > > On Wed, 2005-08-24 at 17:20 +1200, Guy Thornley wrote: > >> Dont forget that already in postgres, you have a process per connection, and > >> all the processes take care of their own I/O. > > > That's the problem. Instead you want 1 or 4 or 10 i/o slaves > > coordinating the I/O of all the backends optimally. For instance, with > > synchronous scanning. > > And why exactly are we going to do a better job of I/O scheduling than > the OS itself can do? ... > There are some things we could do to reduce the impedance between us and > the OS --- for instance, the upthread criticism that a seqscan asks the > OS for only 8K at a time is fair enough. But that doesn't translate > to a conclusion that we should schedule the I/O instead of the OS. Synchronous scanning is a fairly huge and obvious win. If you have two processes 180 degrees out-of-phase in a linear read, neither process is going to get anywhere near the throughput they would get from a single scan. I think you're being deliberately obtuse with regards to file I/O and the operating system. The OS isn't magical. It has to strike a balance between a reasonable read latency and a reasonable throughput. As far as the kernel is concerned, a busy postgresql server is indistinguishable from 100 unrelated activities. All backends will be served equally, even if in this case "equally" means "quite badly all around." An I/O slave process could be a big win in Postgres for many kinds of reads. Instead of opening and reading files the backends would connect to the I/O slave and request the file be read. If a scan of that file were already underway, the new backends would be attached. Otherwise a new scan would commence. In either case, the slave process can issue (sometimes non-dependant) reads well ahead of the needs of the backend. You may think the OS can do this for you but it can't. On postgres knows that it needs the whole file from beginning to end. The OS can only guess. Ask me sometime about my replacement for GNU sort. It uses the same sorting algorithm, but it's an order of magnitude faster due to better I/O strategy. Someday, in my infinite spare time, I hope to demonstrate that kind of improvement with a patch to pg. -jwb From pgsql-performance-owner@postgresql.org Sat Aug 27 02:11:19 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 59A72D728E for ; Wed, 24 Aug 2005 03:25:20 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 70091-09 for ; Wed, 24 Aug 2005 06:25:14 +0000 (GMT) Received: from floppy.pyrenet.fr (floppy.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id EA6BBD72AC for ; Wed, 24 Aug 2005 03:25:12 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id 3991630B42; Wed, 24 Aug 2005 08:37:28 +0200 (MET DST) From: "tobbe" X-Newsgroups: pgsql.performance Subject: Re: Performance for relative large DB Date: 23 Aug 2005 23:25:02 -0700 Organization: http://groups.google.com Lines: 27 Message-ID: <1124864702.822522.64410@g49g2000cwa.googlegroups.com> References: <1124799264.282119.167760@g47g2000cwa.googlegroups.com> <60wtmcptzg.fsf@dba2.int.libertyrms.com> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" X-Complaints-To: groups-abuse@google.com User-Agent: G2/0.2 Complaints-To: groups-abuse@google.com Injection-Info: g49g2000cwa.googlegroups.com; posting-host=212.209.39.154; posting-account=6PgMzAwAAAAyhWPJdwnKqBSRj5_iJFf6 To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/471 X-Sequence-Number: 14218 Hi Chris. Thanks for the answer. Sorry that i was a bit unclear. 1) We update around 20.000 posts per night. 2) What i meant was that we suspect that the DBMS called PervasiveSQL that we are using today is much to small. That's why we're looking for alternatives. Today we base our solution much on using querry-specific tables created at night, so instead of doing querrys direct on the "post" table (with 4-6M rows) at daytime, we have the data pre-aligned in several much smaller tables. This is just to make the current DBMS coop with our amount of data. What I am particulary interested in is if we can expect to run all our select querrys directly from the "post" table with PostgreSQL. 3) How well does postgres work with load balancing environments. Is it built-in? Best Regards Robert Bengtsson Project Manager From pgsql-performance-owner@postgresql.org Wed Aug 24 07:35:15 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4B856D71FA for ; Wed, 24 Aug 2005 07:35:12 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 87077-08 for ; Wed, 24 Aug 2005 10:35:09 +0000 (GMT) Received: from boutiquenumerique.com (boutiquenumerique.com [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 7CACFD7164 for ; Wed, 24 Aug 2005 07:35:07 -0300 (ADT) Received: (qmail 7628 invoked from network); 24 Aug 2005 12:35:33 +0200 Received: from unknown (HELO localhost) (boutiquenumerique-lists@192.168.0.4) by boutiquenumerique.com with SMTP; 24 Aug 2005 12:35:33 +0200 Date: Wed, 24 Aug 2005 12:35:08 +0200 To: "Tom Lane" , "Jeffrey W. Baker" Subject: Re: Read/Write block sizes Cc: "Guy Thornley" , "Josh Berkus" , pgsql-performance@postgresql.org, "Steve Poe" , "Chris Browne" References: <60k6icpapm.fsf@dba2.int.libertyrms.com> <1124846743.12045.94.camel@amd64-laptop-spoe> <200508231931.29682.josh@agliodbs.com> <1124852854.11147.5.camel@noodles> <20050824052023.GB31806@conker.esphion.com> <1124861121.11270.1.camel@noodles> <21700.1124863004@sss.pgh.pa.us> From: PFC Organization: =?iso-8859-15?Q?La_Boutique_Num=E9rique?= Content-Type: text/plain; format=flowed; delsp=yes; charset=iso-8859-15 MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Message-ID: In-Reply-To: <21700.1124863004@sss.pgh.pa.us> User-Agent: Opera M2/8.0 (Linux, build 1095) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 required=5 tests=[AWL=0.005] X-Spam-Level: X-Archive-Number: 200508/379 X-Sequence-Number: 14126 > of effort reinventing the wheel ... but our time will be repaid much > more if we work at levels that the OS cannot have knowledge of, such as > join planning and data statistics. Considering a global budget of man-hours which is the best ? 1- Spend it on reimplementing half of VFS in postgres, half of Windows in postgres, half of FreeBSD in postgres, half of Solaris in Postgres, only to discover you gain a meagre speed increase and a million and a half bugs, 2- Spending 5% of that time lowering the impedance between the OS and Postgres, and another 5% annoying Kernel people and helping them tweaking stuff for database use, and the rest on useful features that give useful speedups, like bitmap indexes, skip scans, and other features that enhance power and usability ? If you're Oracle and have almost unlimited resources, maybe. But even Microsoft opted for option 2 : they implemented ReadFileGather and WriteFileScatter to lower the syscall overhead and that's it. And point 2 will benefit to many other apps, wether 1 would benefit only postgres, and then only in certain cases. I do believe there is something ineresting to uncover with reiser4 though (it definitely fits point 2). I'm happy that the pg team chose point 2 and that new versions keep coming with new features at an unbelievable rate these times. Do you guys sleep ? From pgsql-performance-owner@postgresql.org Wed Aug 24 10:27:02 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 051E9D7243 for ; Wed, 24 Aug 2005 10:25:18 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 18184-07 for ; Wed, 24 Aug 2005 13:25:15 +0000 (GMT) Received: from brmea-mail-3.sun.com (brmea-mail-3.Sun.COM [192.18.98.34]) by svr1.postgresql.org (Postfix) with ESMTP id 1FB65D7166 for ; Wed, 24 Aug 2005 10:25:14 -0300 (ADT) Received: from phys-bur-1 ([129.148.9.72]) by brmea-mail-3.sun.com (8.12.10/8.12.9) with ESMTP id j7ODPIWR027571 for ; Wed, 24 Aug 2005 07:25:18 -0600 (MDT) Received: from conversion-daemon.bur-mail1.east.sun.com by bur-mail1.east.sun.com (iPlanet Messaging Server 5.2 HotFix 1.24 (built Dec 19 2003)) id <0ILQ00F01AIXYF@bur-mail1.east.sun.com> (original mail from Donald.Courtney@Sun.COM) for pgsql-performance@postgresql.org; Wed, 24 Aug 2005 09:25:18 -0400 (EDT) Received: from [129.148.184.34] (gyama.East.Sun.COM [129.148.184.34]) by bur-mail1.east.sun.com (iPlanet Messaging Server 5.2 HotFix 1.24 (built Dec 19 2003)) with ESMTPA id <0ILQ009QTAM6OB@bur-mail1.east.sun.com>; Wed, 24 Aug 2005 09:25:18 -0400 (EDT) Date: Wed, 24 Aug 2005 09:21:12 -0400 From: Donald Courtney Subject: Re: Caching by Postgres In-reply-to: To: William Yu Cc: pgsql-performance@postgresql.org Message-id: <430C7448.7030103@sun.com> Organization: Sun Microsystems MIME-version: 1.0 Content-type: text/plain; charset=ISO-8859-1; format=flowed Content-transfer-encoding: 7BIT X-Accept-Language: en-us, en User-Agent: Mozilla/5.0 (X11; U; SunOS i86pc; en-US; rv:1.7) Gecko/20041221 References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/380 X-Sequence-Number: 14127 Great discussion and illuminating for those of us who are still learning the subtleties of postGres. William To be clear - I built postgreSQL 8.1 64K bit on solaris 10 a few months ago and side by side with the 32 bit postgreSQL build saw no improvement. In fact the 64 bit result was slightly lower. I used the *same 64 bit S10 OS* for both versions. I think your experience makes sense since your change was from 32 to 64 bit Linux. From my experiment I am surmising that there will not be any file/os/buffer-cache scale up effect on the same OS with postgreSQL 64. I was testing on a 4 core system in both cases. William Yu wrote: > Donald Courtney wrote: > >> in that even if you ran postgreSQL on a 64 bit address space >> with larger number of CPUs you won't see much of a scale up >> and possibly even a drop. I am not alone in having the *expectation* > > > What's your basis for believing this is the case? Why would > PostgreSQL's dependence on the OS's caching/filesystem limit > scalability? I know when I went from 32bit to 64bit Linux, I got > *HUGE* increases in performance using the same amount of memory. And > when I went from 2x1P to 2xDC, my average cpu usage % dropped almost > in half. > >> that a database should have some cache size parameter and >> the option to skip the file system. If I use oracle, sybase, mysql >> and maxdb they all have the ability to size a data cache and move >> to 64 bits. > > > Josh Berkus has already mentioned this as conventional wisdom as > written by Oracle. This may also be legacy wisdom. Oracle/Sybase/etc > has been around for a long time; it was probably a clear performance > win way back when. Nowadays with how far open-source OS's have > advanced, I'd take it with a grain of salt and do my own performance > analysis. I suspect the big vendors wouldn't change their stance even > if they knew it was no longer true due to the support hassles. > > My personal experience with PostgreSQL. Dropping shared buffers from > 2GB to 750MB improved performance on my OLTP DB a good 25%. Going down > from 750MB to 150MB was another +10%. > > ---------------------------(end of broadcast)--------------------------- > TIP 1: if posting/reading through Usenet, please send an appropriate > subscribe-nomail command to majordomo@postgresql.org so that your > message can get through to the mailing list cleanly From pgsql-performance-owner@postgresql.org Wed Aug 24 10:52:34 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6B3C7D6D8A for ; Wed, 24 Aug 2005 10:52:33 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 00146-01 for ; Wed, 24 Aug 2005 13:52:28 +0000 (GMT) Received: from candle.pha.pa.us (candle.pha.pa.us [64.139.89.126]) by svr1.postgresql.org (Postfix) with ESMTP id A3B3CD6E53 for ; Wed, 24 Aug 2005 10:52:26 -0300 (ADT) Received: (from pgman@localhost) by candle.pha.pa.us (8.11.6/8.11.6) id j7ODqRU08395; Wed, 24 Aug 2005 09:52:27 -0400 (EDT) From: Bruce Momjian Message-Id: <200508241352.j7ODqRU08395@candle.pha.pa.us> Subject: Re: Read/Write block sizes In-Reply-To: <430BD9CC.3030306@sun.com> To: "Jignesh K. Shah" Date: Wed, 24 Aug 2005 09:52:27 -0400 (EDT) Cc: "Jim C. Nasby" , Chris Browne , pgsql-performance@postgresql.org X-Mailer: ELM [version 2.4ME+ PL121 (25)] MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=US-ASCII X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.009 required=5 tests=[AWL=0.009] X-Spam-Level: X-Archive-Number: 200508/381 X-Sequence-Number: 14128 This thread covers several performance ideas. First is the idea that more parameters should be configurable. While this seems like a noble goal, we try to make parameters auto-tuning, or if users have to configure it, the parameter should be useful for a significant number of users. In the commercial software world, if you can convince your boss that a feature/knob is useful, it usually gets into the product. Unfortunately, this leads to the golden doorknob on a shack, where some features are out of sync with the rest of the product in terms of usefulness and utility. With open source, if a feature can not be auto-tuned, or has significant overhead, the features has to be implemented and then proven to be a benefit. In terms of adding async I/O, threading, and other things, it might make sense to explore how these could be implemented in a way that fits the above criteria. --------------------------------------------------------------------------- Jignesh K. Shah wrote: > Hi Jim, > > | How many of these things are currently easy to change with a recompile? > | I should be able to start testing some of these ideas in the near > | future, if they only require minor code or configure changes. > > > The following > * Data File Size 1GB > * WAL File Size of 16MB > * Block Size of 8K > > Are very easy to change with a recompile.. A Tunable will be greatly > prefered as it will allow one binary for different tunings > > * MultiBlock read/write > > Is not available but will greatly help in reducing the number of system > calls which will only increase as the size of the database increases if > something is not done about i. > > * Pregrown files... maybe not important at this point since TABLESPACE > can currently work around it a bit (Just need to create a different file > system for each tablespace > > But if you really think hardware & OS is the answer for all small > things...... I think we should now start to look on how to make Postgres > Multi-threaded or multi-processed for each connection. With the influx > of "Dual-Core" or "Multi-Core" being the fad.... Postgres can have the > cutting edge if somehow exploiting cores is designed. > > Somebody mentioned that adding CPU to Postgres workload halved the > average CPU usage... > YEAH... PostgreSQL uses only 1 CPU per connection (assuming 100% > usage) so if you add another CPU it is idle anyway and the system will > report only 50% :-) BUT the importing to measure is.. whether the query > time was cut down or not? ( No flames I am sure you were talking about > multi-connection multi-user environment :-) ) But my point is then this > approach is worth the ROI and the time and effort spent to solve this > problem. > > I actually vote for a multi-threaded solution for each connection while > still maintaining seperate process for each connections... This way the > fundamental architecture of Postgres doesn't change, however a > multi-threaded connection can then start to exploit different cores.. > (Maybe have tunables for number of threads to read data files who > knows.. If somebody is interested in actually working a design .. > contact me and I will be interested in assisting this work. > > Regards, > Jignesh > > > Jim C. Nasby wrote: > > >On Tue, Aug 23, 2005 at 06:09:09PM -0400, Chris Browne wrote: > > > > > >>J.K.Shah@Sun.COM (Jignesh Shah) writes: > >> > >> > >>>>Does that include increasing the size of read/write blocks? I've > >>>>noticedthat with a large enough table it takes a while to do a > >>>>sequential scan, even if it's cached; I wonder if the fact that it > >>>>takes a million read(2) calls to get through an 8G table is part of > >>>>that. > >>>> > >>>> > >>>Actually some of that readaheads,etc the OS does already if it does > >>>some sort of throttling/clubbing of reads/writes. But its not enough > >>>for such types of workloads. > >>> > >>>Here is what I think will help: > >>> > >>>* Support for different Blocksize TABLESPACE without recompiling the > >>>code.. (Atlease support for a different Blocksize for the whole > >>>database without recompiling the code) > >>> > >>>* Support for bigger sizes of WAL files instead of 16MB files > >>>WITHOUT recompiling the code.. Should be a tuneable if you ask me > >>>(with checkpoint_segments at 256.. you have too many 16MB files in > >>>the log directory) (This will help OLTP benchmarks more since now > >>>they don't spend time rotating log files) > >>> > >>>* Introduce a multiblock or extent tunable variable where you can > >>>define a multiple of 8K (or BlockSize tuneable) to read a bigger > >>>chunk and store it in the bufferpool.. (Maybe writes too) (Most > >>>devices now support upto 1MB chunks for reads and writes) > >>> > >>>*There should be a way to preallocate files for TABLES in > >>>TABLESPACES otherwise with multiple table writes in the same > >>>filesystem ends with fragmented files which causes poor "READS" from > >>>the files. > >>> > >>>* With 64bit 1GB file chunks is also moot.. Maybe it should be > >>>tuneable too like 100GB without recompiling the code. > >>> > >>>Why recompiling is bad? Most companies that will support Postgres > >>>will support their own binaries and they won't prefer different > >>>versions of binaries for different blocksizes, different WAL file > >>>sizes, etc... and hence more function using the same set of binaries > >>>is more desirable in enterprise environments > >>> > >>> > >>Every single one of these still begs the question of whether the > >>changes will have a *material* impact on performance. > >> > >> > > > ---------------------------(end of broadcast)--------------------------- > TIP 3: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faq > -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073 From pgsql-performance-owner@postgresql.org Wed Aug 24 12:37:06 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4C099D7826 for ; Wed, 24 Aug 2005 11:51:40 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83934-08 for ; Wed, 24 Aug 2005 14:51:36 +0000 (GMT) Received: from postfix.vectorx.com.br (fire.vectorx.com.br [200.247.35.133]) by svr1.postgresql.org (Postfix) with ESMTP id 75DECD77FD for ; Wed, 24 Aug 2005 11:51:34 -0300 (ADT) Received: from localhost (mail [127.0.0.1]) by postfix.vectorx.com.br (Postfix) with ESMTP id F30CDE06F for ; Wed, 24 Aug 2005 11:47:02 -0400 (EDT) Received: from postfix.vectorx.com.br ([127.0.0.1]) by localhost (vectorx.com.br [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 05812-08 for ; Wed, 24 Aug 2005 11:46:58 -0400 (EDT) Received: from [192.168.16.7] (unknown [192.168.16.7]) by postfix.vectorx.com.br (Postfix) with ESMTP id D1B34E04F for ; Wed, 24 Aug 2005 11:46:56 -0400 (EDT) Message-ID: <430C8779.9080500@vectorx.com.br> Date: Wed, 24 Aug 2005 11:43:05 -0300 From: Alexandre Barros User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.11) Gecko/20050731 X-Accept-Language: pt-br, en-us, en MIME-Version: 1.0 To: postgres performance Subject: performance drop on RAID5 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: amavisd-new at vectorx.com.br X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.025 required=5 tests=[AWL=0.025] X-Spam-Level: X-Archive-Number: 200508/383 X-Sequence-Number: 14130 Hello, i have a pg-8.0.3 running on Linux kernel 2.6.8, CPU Sempron 2600+, 1Gb RAM on IDE HD ( which could be called a "heavy desktop" ), measuring this performance with pgbench ( found on /contrib ) it gave me an average ( after several runs ) of 170 transactions per second; for the sake of experimentation ( actually, i'm scared this IDE drive could fail at any time, hence i'm looking for an alternative, more "robust", machine ), i've installed on an aging Compaq Proliant server ( freshly compiled SMP kernel 2.6.12.5 with preemption ), dual Pentium III Xeon 500Mhz, 512Mb RAM, (older) SCSI-2 80pin drives, and re-tested, when the database was on a single SCSI drive, pgbench gave me an average of 90 transactions per second, but, and that scared me most, when the database was on a RAID-5 array ( four 9Gb disks, using linux software RAID mdadm and LVM2, with the default filesystem cluster size of 32Kb ), the performance dropped to about 55 transactions per second. Despite the amount of RAM difference, none machine seems to be swapping. All filesystems ( on both machines ) are Reiserfs. Both pg-8.0.3 were compiled with CFLAGS -O3 and -mtune for their respective architectures... and "gmake -j2" on the server. Both machines have an original ( except by the pg and the kernel ) Mandrake 10.1 install. I've googled a little, and maybe the cluster size might be one problem, but despite that, the performance dropping when running on "server-class" hardware with RAID-5 SCSI-2 drives was way above my most delirious expectations... i need some help to figure out what is **so** wrong... i wouldn't be so stunned if the newer machine was ( say ) twice faster than the older server, but over three times faster is disturbing. the postgresql.conf of both machines is here: max_connections = 50 shared_buffers = 1000 # min 16, at least max_connections*2, 8KB each debug_print_parse = false debug_print_rewritten = false debug_print_plan = false debug_pretty_print = false log_statement = 'all' log_parser_stats = false log_planner_stats = false log_executor_stats = false log_statement_stats = false lc_messages = 'en_US' # locale for system error message strings lc_monetary = 'en_US' # locale for monetary formatting lc_numeric = 'en_US' # locale for number formatting lc_time = 'en_US' # locale for time formatting many thanks in advance ! From pgsql-performance-owner@postgresql.org Wed Aug 24 12:32:32 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B2F43D7842 for ; Wed, 24 Aug 2005 11:55:16 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91406-09 for ; Wed, 24 Aug 2005 14:55:10 +0000 (GMT) Received: from ns.snowman.net (ns.snowman.net [66.92.160.21]) by svr1.postgresql.org (Postfix) with ESMTP id 97D37D7805 for ; Wed, 24 Aug 2005 11:55:08 -0300 (ADT) Received: by ns.snowman.net (Postfix, from userid 1000) id C2A2817B11; Wed, 24 Aug 2005 10:55:33 -0400 (EDT) Date: Wed, 24 Aug 2005 10:55:33 -0400 From: Stephen Frost To: Donald Courtney Cc: William Yu , pgsql-performance@postgresql.org Subject: Re: Caching by Postgres Message-ID: <20050824145533.GN6026@ns.snowman.net> Mail-Followup-To: Donald Courtney , William Yu , pgsql-performance@postgresql.org References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> <430C7448.7030103@sun.com> Mime-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="BU7+kJFeeDlNltZg" Content-Disposition: inline In-Reply-To: <430C7448.7030103@sun.com> X-Editor: Vim http://www.vim.org/ X-Info: http://www.snowman.net X-Operating-System: Linux/2.4.24ns.3.0 (i686) X-Uptime: 10:48:33 up 74 days, 7:06, 6 users, load average: 0.05, 0.13, 0.09 User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.004 required=5 tests=[AWL=0.004] X-Spam-Level: X-Archive-Number: 200508/382 X-Sequence-Number: 14129 --BU7+kJFeeDlNltZg Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable * Donald Courtney (Donald.Courtney@sun.com) wrote: > To be clear - > I built postgreSQL 8.1 64K bit on solaris 10 a few months ago > and side by side with the 32 bit postgreSQL build saw no improvement.=20 > In fact the 64 bit result was slightly lower. That makes some sense actually. It really depends on what you're doing alot of the time. On a Sparc system you're not likely to get much of a speed improvment by going to 64bit (unless, maybe, you're doing lots of intensive 64bit math ops). You'll have larger pointers and whatnot though. > I used the *same 64 bit S10 OS* for both versions. I think your > experience makes sense since your change was from 32 to 64 bit Linux. 32bit to 64bit Linux on a Sparc platform really shouldn't affect performance all that much (I'd expect it to be similar to 32bit to 64bit under Solaris actually, at least in terms of the performance difference). 32bit to 64bit Linux on an amd64 platform is another matter entirely though, but not because of the number of bits involved. Under amd64, 32bit is limited to 32bit on i386 which has a limited number of registers and whatnot. Under amd64/64bit you get more registers (and I think some other niceities) which will improve performance. That's not a 32bit vs. 64bit thing, that's i386 vs. native amd64. It's really mainly an oddity of the platform. On a mips system I'd expect the same kind of performance difference between 32bit and 64bit as you'd see on a sparc platform. Enjoy, Stephen --BU7+kJFeeDlNltZg Content-Type: application/pgp-signature; name="signature.asc" Content-Description: Digital signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (GNU/Linux) iD8DBQFDDIplrzgMPqB3kigRAgdZAJ9qbWNbxrPc6bSy7foIPir/sPv0RwCfWhVT 9IKq1E6qwm6VW6vw3arTAhc= =pqT1 -----END PGP SIGNATURE----- --BU7+kJFeeDlNltZg-- From pgsql-performance-owner@postgresql.org Wed Aug 24 13:18:25 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CB0D9D6FBB for ; Wed, 24 Aug 2005 12:36:42 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 73999-08 for ; Wed, 24 Aug 2005 15:36:40 +0000 (GMT) Received: from boutiquenumerique.com (boutiquenumerique.com [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 5810DD6E00 for ; Wed, 24 Aug 2005 12:36:38 -0300 (ADT) Received: (qmail 21623 invoked from network); 24 Aug 2005 17:37:01 +0200 Received: from unknown (HELO localhost) (boutiquenumerique-lists@192.168.0.4) by boutiquenumerique.com with SMTP; 24 Aug 2005 17:37:01 +0200 To: "Josh Berkus" , pgsql-performance@postgresql.org Cc: "William Yu" Subject: Re: Caching by Postgres References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <200508231911.59100.josh@agliodbs.com> Message-ID: From: PFC Organization: =?iso-8859-15?Q?La_Boutique_Num=E9rique?= Content-Type: text/plain; format=flowed; delsp=yes; charset=iso-8859-15 MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Date: Wed, 24 Aug 2005 17:36:35 +0200 In-Reply-To: <200508231911.59100.josh@agliodbs.com> User-Agent: Opera M2/8.0 (Linux, build 1095) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/385 X-Sequence-Number: 14132 > Really? Cool, I'd like to see that. Could you follow up with Hans? > Or give > me his e-mail? You can subscribe to the Reiser mailinglist on namesys.com or : reiser@namesys.com From pgsql-performance-owner@postgresql.org Wed Aug 24 12:54:20 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B64B7D726E for ; Wed, 24 Aug 2005 12:52:12 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 11324-06 for ; Wed, 24 Aug 2005 15:52:07 +0000 (GMT) Received: from frank.wiles.org (frank.wiles.org [24.124.39.75]) by svr1.postgresql.org (Postfix) with ESMTP id 6A1BFD7166 for ; Wed, 24 Aug 2005 12:52:06 -0300 (ADT) Received: from kungfu (frank.wiles.org [127.0.0.1]) by frank.wiles.org (8.13.1/8.13.1) with SMTP id j7OF0ljd015554; Wed, 24 Aug 2005 10:00:47 -0500 Date: Wed, 24 Aug 2005 10:52:43 -0500 From: Frank Wiles To: Alexandre Barros Cc: pgsql-performance@postgresql.org Subject: Re: performance drop on RAID5 Message-Id: <20050824105243.067f0ae5.frank@wiles.org> In-Reply-To: <430C8779.9080500@vectorx.com.br> References: <430C8779.9080500@vectorx.com.br> X-Mailer: Sylpheed version 1.0.4 (GTK+ 1.2.10; i386-redhat-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/384 X-Sequence-Number: 14131 On Wed, 24 Aug 2005 11:43:05 -0300 Alexandre Barros wrote: > I've googled a little, and maybe the cluster size might be one > problem, but despite that, the performance dropping when running on > "server-class" hardware with RAID-5 SCSI-2 drives was way above my > most delirious expectations... i need some help to figure out what is > **so** wrong... RAID-5 isn't great for databases in general. What would be better would be to mirror the disks to redundancy or do RAID 1+0. You could probably also increase your shared_buffers some, but that alone most likely won't make up your speed difference. --------------------------------- Frank Wiles http://www.wiles.org --------------------------------- From pgsql-performance-owner@postgresql.org Wed Aug 24 14:04:00 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 11E93D7266 for ; Wed, 24 Aug 2005 13:02:57 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 42195-07 for ; Wed, 24 Aug 2005 16:02:53 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id 0BFEED70BF for ; Wed, 24 Aug 2005 13:02:51 -0300 (ADT) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: performance drop on RAID5 Date: Wed, 24 Aug 2005 12:02:51 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3417DD1A3@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] performance drop on RAID5 Thread-Index: AcWowgLuh0xdZeDaSZuHRbYhqWtA8gAAgmlw From: "Merlin Moncure" To: "Alexandre Barros" Cc: "postgres performance" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/392 X-Sequence-Number: 14139 > Hello, > i have a pg-8.0.3 running on Linux kernel 2.6.8, CPU Sempron 2600+, > 1Gb RAM on IDE HD ( which could be called a "heavy desktop" ), measuring > this performance with pgbench ( found on /contrib ) it gave me an > average ( after several runs ) of 170 transactions per second; 170 tps is not plausible no a single platter IDE disk without using write caching of some kind. For a 7200 rpm drive any result much over 100 tps is a little suspicious. (my 10k sata raptor can do about 120). =20 > for the sake of experimentation ( actually, i'm scared this IDE drive > could fail at any time, hence i'm looking for an alternative, more > "robust", machine ), i've installed on an aging Compaq Proliant server ( > freshly compiled SMP kernel 2.6.12.5 with preemption ), dual Pentium > III Xeon 500Mhz, 512Mb RAM, (older) SCSI-2 80pin drives, and re-tested, > when the database was on a single SCSI drive, pgbench gave me an average > of 90 transactions per second, but, and that scared me most, when the > database was on a RAID-5 array ( four 9Gb disks, using linux software > RAID mdadm and LVM2, with the default filesystem cluster size of 32Kb ), > the performance dropped to about 55 transactions per second. Is natural to see a slight to moderate drop in write performance moving to RAID 5. The only raid levels that are faster than single disk levels for writing are the ones with '0' in it or caching raid controllers. Even for 0+1, expect modest gains in tps vs. single disk if not using write caching. Merlin From pgsql-performance-owner@postgresql.org Wed Aug 24 13:43:30 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 878DCD77F7 for ; Wed, 24 Aug 2005 13:36:19 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 22547-04 for ; Wed, 24 Aug 2005 16:36:14 +0000 (GMT) Received: from floppy.pyrenet.fr (floppy.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id 3C60CD7336 for ; Wed, 24 Aug 2005 13:36:13 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id 3693A30B42; Wed, 24 Aug 2005 18:48:30 +0200 (MET DST) From: Chris Browne X-Newsgroups: pgsql.performance Subject: Re: Read/Write block sizes Date: Wed, 24 Aug 2005 12:12:22 -0400 Organization: cbbrowne Computing Inc Lines: 63 Message-ID: <60br3npb4p.fsf@dba2.int.libertyrms.com> References: <60k6icpapm.fsf@dba2.int.libertyrms.com> <1124846743.12045.94.camel@amd64-laptop-spoe> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 X-Complaints-To: usenet@news.hub.org User-Agent: Gnus/5.1007 (Gnus v5.10.7) XEmacs/21.4.17 (Jumbo Shrimp, linux) Cancel-Lock: sha1:mK8J5RkNPPi1vcmSWCFrL193fzI= To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.114 required=5 tests=[AWL=0.114] X-Spam-Level: X-Archive-Number: 200508/387 X-Sequence-Number: 14134 spoe@sfnet.cc (Steve Poe) writes: > Chris, > > Unless I am wrong, you're making the assumpting the amount of time spent > and ROI is known. Maybe those who've been down this path know how to get > that additional 2-4% in 30 minutes or less? > > While each person and business' performance gains (or not) could vary, > someone spending the 50-100h to gain 2-4% over a course of a month for a > 24x7 operation would seem worth the investment? What we *do* know is that adding these "knobs" would involve a significant amount of effort, as the values are widely used throughout the database engine. Making them dynamic (e.g. - so they could be tuned on a tablespace-by-tablespace basis) would undoubtedly require rather a lot of development effort. They are definitely NOT 30 minute changes. Moreover, knowing how to adjust them is almost certainly also NOT a 30 minute configuration change; significant benchmarking effort for the individual application is almost sure to be needed. It's not much different from the reason why PostgreSQL doesn't use threading... The problem with using threading is that introducing it to the code base would require a pretty enormous amount of effort (I'll bet multiple person-years), and it wouldn't provide *any* benefit until you get rather a long ways down the road. Everyone involved in development seems to me to have a reasonably keen understanding as to what the potential benefits of threading are; the value is that there fall out plenty of opportunities to parallelize the evaluation of portions of queries. Alas, it wouldn't be until *after* all the effort goes in that we would get any idea as to what kinds of speedups this would provide. In effect, there has to be a year invested in *breaking* PostgreSQL (because this would initially break a lot, since thread programming is a really tough skill) where you don't actually see any benefits. > I would assume that dbt2 with STP helps minimize the amount of hours > someone has to invest to determine performance gains with > configurable options? That's going to help in constructing a "default" knob value. And if we find an "optimal default," that encourages sticking with the current approach, of using #define to apply that value... >> If someone spends 100h working on one of these items, and gets a 2% >> performance improvement, that's almost certain to be less desirable >> than spending 50h on something else that gets a 4% improvement. >> >> And we might discover that memory management improvements in Linux >> 2.6.16 or FreeBSD 5.5 allow some OS kernels to provide some such >> improvements "for free" behind our backs without *any* need to write >> database code. :-) -- let name="cbbrowne" and tld="ntlug.org" in String.concat "@" [name;tld];; http://www.ntlug.org/~cbbrowne/lisp.html "For those of you who are into writing programs that are as obscure and complicated as possible, there are opportunities for... real fun here" -- Arthur Norman From pgsql-performance-owner@postgresql.org Wed Aug 24 13:54:11 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B4302D7844 for ; Wed, 24 Aug 2005 13:16:29 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 84853-02 for ; Wed, 24 Aug 2005 16:16:25 +0000 (GMT) Received: from hosting.commandprompt.com (128.commandprompt.com [207.173.200.128]) by svr1.postgresql.org (Postfix) with ESMTP id 9CE39D7803 for ; Wed, 24 Aug 2005 13:16:22 -0300 (ADT) Received: from [192.168.1.100] (clbb-248.saw.net [64.146.135.248]) (authenticated bits=0) by hosting.commandprompt.com (8.13.4/8.13.4) with ESMTP id j7OGEGBe025056; Wed, 24 Aug 2005 09:14:17 -0700 Message-ID: <430C9DA6.2080801@commandprompt.com> Date: Wed, 24 Aug 2005 09:17:42 -0700 From: "Joshua D. Drake" User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050727) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Alexandre Barros Cc: postgres performance Subject: Re: performance drop on RAID5 References: <430C8779.9080500@vectorx.com.br> In-Reply-To: <430C8779.9080500@vectorx.com.br> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Greylist: Sender succeded SMTP AUTH authentication, not delayed by milter-greylist-1.6 (hosting.commandprompt.com [192.168.1.101]); Wed, 24 Aug 2005 09:14:18 -0700 (PDT) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.022 required=5 tests=[AWL=0.022] X-Spam-Level: X-Archive-Number: 200508/391 X-Sequence-Number: 14138 Alexandre Barros wrote: > Hello, > i have a pg-8.0.3 running on Linux kernel 2.6.8, CPU Sempron > 2600+, 1Gb RAM on IDE HD ( which could be called a "heavy desktop" ), > measuring this performance with pgbench ( found on /contrib ) it gave > me an average ( after several runs ) of 170 transactions per second; That is going to be because IDE drives LIE about write times because of the large cache. > for the sake of experimentation ( actually, i'm scared this IDE drive > could fail at any time, hence i'm looking for an alternative, more > "robust", machine ), i've installed on an aging Compaq Proliant server > ( freshly compiled SMP kernel 2.6.12.5 with preemption ), dual > Pentium III Xeon 500Mhz, 512Mb RAM, (older) SCSI-2 80pin drives, and > re-tested, when the database was on a single SCSI drive, pgbench gave > me an average of 90 transactions per second, but, and that scared me > most, when the database was on a RAID-5 array ( four 9Gb disks, using > linux software RAID mdadm and LVM2, with the default filesystem > cluster size of 32Kb ), the performance dropped to about 55 > transactions per second. That seems more reasonable and probably truthful. I would be curious what type of performance you would get with the exact same setup EXCEPT remove LVM2. Just have the software RAID. In fact, since you have 4 drives you could do RAID 10. > > i wouldn't be so stunned if the newer machine was ( say ) twice faster > than the older server, but over three times faster is disturbing. > > the postgresql.conf of both machines is here: > > max_connections = 50 > shared_buffers = 1000 # min 16, at least max_connections*2, > 8KB each You should look at the annotated conf: http://www.powerpostgresql.com/Downloads/annotated_conf_80.html Sincerely, Joshua D. Drake > debug_print_parse = false > debug_print_rewritten = false > debug_print_plan = false > debug_pretty_print = false > log_statement = 'all' > log_parser_stats = false > log_planner_stats = false > log_executor_stats = false > log_statement_stats = false > lc_messages = 'en_US' # locale for system error message strings > lc_monetary = 'en_US' # locale for monetary formatting > lc_numeric = 'en_US' # locale for number formatting > lc_time = 'en_US' # locale for time formatting > > many thanks in advance ! > > > ---------------------------(end of broadcast)--------------------------- > TIP 3: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faq From pgsql-performance-owner@postgresql.org Wed Aug 24 13:50:24 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6592AD768A for ; Wed, 24 Aug 2005 13:25:38 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 76142-06 for ; Wed, 24 Aug 2005 16:25:36 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id 22105D72C6 for ; Wed, 24 Aug 2005 13:25:35 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [63.195.55.98] (account josh@agliodbs.com HELO spooky) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7926794; Wed, 24 Aug 2005 09:27:52 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: Tom Lane Subject: Re: Read/Write block sizes Date: Wed, 24 Aug 2005 09:26:39 -0700 User-Agent: KMail/1.8 Cc: "Jeffrey W. Baker" , pgsql-performance@postgresql.org, Steve Poe , Chris Browne References: <1124852854.11147.5.camel@noodles> <21415.1124860220@sss.pgh.pa.us> In-Reply-To: <21415.1124860220@sss.pgh.pa.us> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200508240926.39885.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.011 required=5 tests=[AWL=0.011] X-Spam-Level: X-Archive-Number: 200508/390 X-Sequence-Number: 14137 Tom, Gavin, > > To get decent I/O you need 1MB fundamental units all the way down the > > stack. > > It would also be a good idea to have an application that isn't likely > to change a single bit in a 1MB range and then expect you to record > that change. This pretty much lets Postgres out of the picture. We're looking at this pretty much just for data warehousing, where you constantly have gigabytes of data which don't change from month to month or even year to year. I agree that it would *not* be an optimization for OLTP systems. Which is why a build-time option would be fine. > Ummm... I don't see anything here which will be a win for Postgres. The > transactional semantics we're interested in are fairly complex: > > 1) Modifications to multiple objects can become visible to the system > atomically > 2) On error, a series of modifications which had been grouped together > within a transaction can be rolled back > 3) Using object version information, determine which version of which > object is visible to a given session > 4) Using version information and locking, detect and resolve read/write > and write/write conflicts I wasn't thinking of database transactions. I was thinking specifically of using Reiser4 transactions (and other transactional filesytems) to do things like eliminate the need for full page writes in the WAL. Filesystems are low-level things which should take care of low-level needs, like making sure an 8K page got written to disk even in the event of a system failure. -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Wed Aug 24 13:32:41 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4A838D7849 for ; Wed, 24 Aug 2005 13:32:30 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 05121-07 for ; Wed, 24 Aug 2005 16:32:26 +0000 (GMT) Received: from olive.qinip.net (olive.qinip.net [62.100.30.40]) by svr1.postgresql.org (Postfix) with ESMTP id 3AA99D7858 for ; Wed, 24 Aug 2005 13:32:25 -0300 (ADT) Received: from [10.0.0.2] (h8441139206.dsl.speedlinq.nl [84.41.139.206]) by olive.qinip.net (Postfix) with ESMTP id 81325180FB; Wed, 24 Aug 2005 18:32:17 +0200 (MEST) Message-ID: <430CA115.8030006@tweakers.net> Date: Wed, 24 Aug 2005 18:32:21 +0200 From: Arjen van der Meijden User-Agent: Mozilla Thunderbird 1.0RC1 (Windows/20041201) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Alexandre Barros Cc: postgres performance Subject: Re: performance drop on RAID5 References: <430C8779.9080500@vectorx.com.br> In-Reply-To: <430C8779.9080500@vectorx.com.br> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/386 X-Sequence-Number: 14133 On 24-8-2005 16:43, Alexandre Barros wrote: > Hello, > i have a pg-8.0.3 running on Linux kernel 2.6.8, CPU Sempron 2600+, > 1Gb RAM on IDE HD ( which could be called a "heavy desktop" ), measuring > this performance with pgbench ( found on /contrib ) it gave me an > average ( after several runs ) of 170 transactions per second; Nowadays you can call that a "light desktop", although the amount of RAM is a bit more than normal. ;) > for the sake of experimentation ( actually, i'm scared this IDE drive > could fail at any time, hence i'm looking for an alternative, more > "robust", machine ), i've installed on an aging Compaq Proliant server ( > freshly compiled SMP kernel 2.6.12.5 with preemption ), dual Pentium Preemption is afaik counter-productive for a server. > III Xeon 500Mhz, 512Mb RAM, (older) SCSI-2 80pin drives, and re-tested, > when the database was on a single SCSI drive, pgbench gave me an average > of 90 transactions per second, but, and that scared me most, when the > database was on a RAID-5 array ( four 9Gb disks, using linux software > RAID mdadm and LVM2, with the default filesystem cluster size of 32Kb ), > the performance dropped to about 55 transactions per second. The default disk io scheduler of the 2.6-series is designed for disks or controllers that have no command queueing (like most standaard IDE-disks). Try changing your default "anticipatory" scheduler on the test-device to "deadline" or "cfq" (see the two *-iosched.txt files in /usr/src/linux/Documentation/block/ for more information). Changing is simple with a 2.6.11+ kernel, just do "echo 'deadline' > /sys/block/*devicename*/queue/scheduler" at runtime. > Despite the amount of RAM difference, none machine seems to be swapping. But there is a 512MB extra amount of file-cache. Which can make a significant difference. > All filesystems ( on both machines ) are Reiserfs. > Both pg-8.0.3 were compiled with CFLAGS -O3 and -mtune for their > respective architectures... and "gmake -j2" on the server. > Both machines have an original ( except by the pg and the kernel ) > Mandrake 10.1 install. > > I've googled a little, and maybe the cluster size might be one problem, > but despite that, the performance dropping when running on > "server-class" hardware with RAID-5 SCSI-2 drives was way above my most > delirious expectations... i need some help to figure out what is **so** > wrong... Did you consider you're overestimating the raid's performance and usage? If the benchmark was mostly run from the memory, you're not going to see much gain in performance from a faster disk. But even worse is that for sequential reads and writes, the performance of current (large) IDE drives is very good. It may actually outperform your RAID on that one. Random access will probably still be slower, but may not be that much slower. And if the database resides in memory, that doesn't matter much anyway. > i wouldn't be so stunned if the newer machine was ( say ) twice faster > than the older server, but over three times faster is disturbing. I'm actually not surprised. Old scsi disks are not faster than new ones anymore, although they still may be a bit faster on random access issues or under (very) high load. Especially if: - you only ran it with 1 client - the database mostly or entirely fits in the desktop's memory - the database did not fit entirely in the server's memory. Even worse would be if the database does fit entirely in the desktop's memory, but not in the server's! Please don't forget your server probably has much slower memory-access, it will likely have 133Mhz SDR Ram instead of your current DDR2700 orso. The latter is much faster (in theory more than twice). Your desktop cpu will very likely, even when multiple processes exist, be faster especially with the faster memory accesses. The Xeon's probably only beat it on the amount of cache. So please check if pgbench actually makes much use of the disk, if it does check how large the test databases will be, etc, etc. Btw, if you'd prefer to use your desktop, but are afraid of the IDE-drive dying on you, buy a "server class" SATA disk. Most manufacturers have those, Western Digital even has "scsi like" sata disks (the Raptor drives), they generally have 3 to 5 years warranty and higher class components. Best regards, Arjen From pgsql-performance-owner@postgresql.org Wed Aug 24 13:43:53 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8F8B0D77F4 for ; Wed, 24 Aug 2005 13:36:18 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 22537-05 for ; Wed, 24 Aug 2005 16:36:14 +0000 (GMT) Received: from floppy.pyrenet.fr (floppy.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id 3BDDED732F for ; Wed, 24 Aug 2005 13:36:11 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id D295B30B43; Wed, 24 Aug 2005 18:48:30 +0200 (MET DST) From: Chris Browne X-Newsgroups: pgsql.performance Subject: Re: Performance for relative large DB Date: Wed, 24 Aug 2005 12:34:51 -0400 Organization: cbbrowne Computing Inc Lines: 80 Message-ID: <607jebpa38.fsf@dba2.int.libertyrms.com> References: <1124799264.282119.167760@g47g2000cwa.googlegroups.com> <60wtmcptzg.fsf@dba2.int.libertyrms.com> <1124864702.822522.64410@g49g2000cwa.googlegroups.com> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 X-Complaints-To: usenet@news.hub.org User-Agent: Gnus/5.1007 (Gnus v5.10.7) XEmacs/21.4.17 (Jumbo Shrimp, linux) Cancel-Lock: sha1:dfPIkB8Ya3TN84LEbPwiDeFwbS4= To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.115 required=5 tests=[AWL=0.115] X-Spam-Level: X-Archive-Number: 200508/388 X-Sequence-Number: 14135 "tobbe" writes: > Hi Chris. > > Thanks for the answer. > Sorry that i was a bit unclear. > > 1) We update around 20.000 posts per night. No surprise there; I would have been surprised to see 100/nite or 6M/nite... > 2) What i meant was that we suspect that the DBMS called PervasiveSQL > that we are using today is much to small. That's why we're looking for > alternatives. > > Today we base our solution much on using querry-specific tables created > at night, so instead of doing querrys direct on the "post" table (with > 4-6M rows) at daytime, we have the data pre-aligned in several much > smaller tables. This is just to make the current DBMS coop with our > amount of data. > > What I am particulary interested in is if we can expect to run all our > select querrys directly from the "post" table with PostgreSQL. Given a decent set of indices, I'd expect that to work OK... Whether 4M or 6M rows, that's pretty moderate in size. If there are specific states that rows are in which are "of interest," then you can get big wins out of having partial indices... Consider... create index partial_post_status on posts where status in ('Active', 'Pending', 'Locked'); -- When processing of postings are completely finished, they wind up with 'Closed' status We have some 'stateful' tables in our environment where the interesting states are 'P' (where work is "pending") and 'C' (where all the work has been completed and the records are never of interest again except as ancient history); the partial index "where status = 'P'" winds up being incredibly helpful. It's worth your while to dump data out from Pervasive and load it into a PostgreSQL instance and to do some typical sorts of queries on the PostgreSQL side. Do "EXPLAIN ANALYZE [some select statement];" and you'll get a feel for how PostgreSQL is running the queries. Fiddling with indices to see how that affects things will also be a big help. You may find there are columns with large cardinalities (quite a lot of unique values) where you want to improve the stats analysis via... alter posts alter column [whatever] set statistics 100; -- Default is 10 bins analyze posts; -- then run ANALYZE to update statistics > 3) How well does postgres work with load balancing environments. Is > it built-in? Load balancing means too many things. Can you be more specific about what you consider it to mean? For Internet registry operations, we use replication (Slony-I) to create replicas used to take particular sorts of load off the "master" systems. But you might be referring to something else... For instance, connection pools, whether implemented inside applications (everyone doing Java has one or more favorite Java connection pool implementations) or in web servers (Apache has a DB connection pool manager) or in an outside application (pgpool, a C-based connection pool manager) are also sometimes used for load balancing. -- (reverse (concatenate 'string "gro.mca" "@" "enworbbc")) http://www3.sympatico.ca/cbbrowne/postgresql.html In case you weren't aware, "ad homineum" is not latin for "the user of this technique is a fine debater." -- Thomas F. Burdick From pgsql-performance-owner@postgresql.org Wed Aug 24 13:47:03 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 23B19D78B7 for ; Wed, 24 Aug 2005 13:35:08 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 92951-07 for ; Wed, 24 Aug 2005 16:35:03 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.200]) by svr1.postgresql.org (Postfix) with ESMTP id 37124D7897 for ; Wed, 24 Aug 2005 13:35:02 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i22so132067wra for ; Wed, 24 Aug 2005 09:35:01 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=b0N2msU1qLDiAm9q/iyXDo63t1Y8GDD9kVYQX8Fl2+PfWQHNHaiIpnVAY8fe0f34IBLM6/MBL3TLn3xHMh6cjnh2pIbdpxCrkPoiBb/G9CI6PV55k/6MgBmMLig9+2xwWBfncoEoNwTBUJBAQYstyQ5PCiJhp0FakRYSkY06ww4= Received: by 10.54.111.11 with SMTP id j11mr839076wrc; Wed, 24 Aug 2005 09:35:01 -0700 (PDT) Received: by 10.54.96.7 with HTTP; Wed, 24 Aug 2005 09:35:01 -0700 (PDT) Message-ID: <37d451f70508240935315add63@mail.gmail.com> Date: Wed, 24 Aug 2005 11:35:01 -0500 From: Rosser Schwarz To: Alexandre Barros Subject: Re: performance drop on RAID5 Cc: postgres performance In-Reply-To: <430C8779.9080500@vectorx.com.br> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <430C8779.9080500@vectorx.com.br> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.089 required=5 tests=[AWL=0.065, RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/389 X-Sequence-Number: 14136 On 8/24/05, Alexandre Barros wrote: > i wouldn't be so stunned if the newer machine was ( say ) twice faster > than the older server, but over three times faster is disturbing. RAID5 on so few spindles is a known losing case for PostgreSQL. You'd be far, far better off doing a pair of RAID1 sets or a single RAID10 set. /rls --=20 :wq From pgsql-performance-owner@postgresql.org Wed Aug 24 14:36:17 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 085BFD78F8 for ; Wed, 24 Aug 2005 13:57:00 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 62000-06 for ; Wed, 24 Aug 2005 16:56:55 +0000 (GMT) Received: from zproxy.gmail.com (zproxy.gmail.com [64.233.162.193]) by svr1.postgresql.org (Postfix) with ESMTP id 50CC7D7833 for ; Wed, 24 Aug 2005 13:56:54 -0300 (ADT) Received: by zproxy.gmail.com with SMTP id 8so85104nzo for ; Wed, 24 Aug 2005 09:56:54 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:mime-version:content-type:content-transfer-encoding:content-disposition; b=gMKjZvIOlGpCuVmiCnaTfbNwP5PX+4gVkP14bo3lpZWWoblAYpM2CGLD2oKtKPVnod7P8oJ5yQKGbwT9Ox6fVAzsbPreHYq2h0xvoWnhuzCJJUDLPmXZexHGZ8ZOe4PM4KDNeum/g7sSxewke70xbUT9H3KKGy6qiexSEevpxek= Received: by 10.36.74.13 with SMTP id w13mr3206721nza; Wed, 24 Aug 2005 09:56:54 -0700 (PDT) Received: by 10.36.24.2 with HTTP; Wed, 24 Aug 2005 09:56:54 -0700 (PDT) Message-ID: <1d219a6f05082409563775a8a7@mail.gmail.com> Date: Wed, 24 Aug 2005 12:56:54 -0400 From: Chris Hoover To: pgsql-performance@postgresql.org Subject: Some ideas for comment Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/393 X-Sequence-Number: 14140 Ok, there is always a lot of talk about tuning PostgreSQL on linux and how PostgreSQL uses the linux kernel cache to cache the tables and indexes. My question is, is there anyway to see what files linux is caching at this moment? My reasoning behind this question is: I have several database systems each with 1 PostgreSQL cluster.=20 However, each cluster has a large number of identical databases on it. Since there can be a great amount of size disparity between the databases, I am wondering if some of slowness we see might be caused by the kernel cache having to reload a lot of data that keeps getting swapped out. (most systems have at least 100GB of data/indexes on them with 8 or 12GB ram). If this is the case, what sort of symptoms would you expect? To help mitigate this potential, I have been researching the following, and am thinking of proposing it to management. Any comments would be appreciated. 1. Implement a partition type layout using views and rules - This will allow me to have one table in each view with the "active" data, and the inactive data stored by year in other tables. So I would have the following (for each major table): Table View as select * from active_table union all=20 select * from table_2005 union all select * from table_2004 etc. Each table would have identical indexes, however only the "active_table" would have data that is actively being worked. The rules and a nightly job can keep the data correctly stored. I am thinking that with this setup, the active table indexes should almost always be in memory. And, if they do happen to be pushed out, they are much smaller than the indexes I have today (where all data is in one table), so they should load faster with less i/o pressure. From the testing I have done so far, I believe I can implement this system with out having to ask for developer time. This is a "Good Thing". Also, the database is not normalized and is very ugly, by using the view to partition and abstract the actual data, I will be in a better position to start normalizing some of the tables w/o developer time (once again, a "Good Thing") 2. I am also thinking of recommending we collapse all databases in a cluster into one "mega" database. I can then use schema's and views to control database access and ensure that no customer can see another customers data. This would mean that there are only one set of indexes being loaded into the cache. While they would be larger, I think in combination with the partition from idea 1, we would be ahead of the ball game.=20 Since there would only be one set of indexes, everyone would be sharing them so they should always be in memory. I don't have real numbers to give you, but we know that our systems are hurting i/o wise and we are growing by about 2GB+ per week (net).=20 We actually grow by about 5GB/week/server. However, when I run my weekly maintenance of vacuum full, reindex, and the vacuum analyze, we end up getting about 3GB back. Unfortunately, I do not have the i/o bandwidth to vacuum during the day as it causes major slowdowns on our system. Each night, I do run a vacuum analyze across all db's to try and help. I also have my fsm parameters set high (8000000 fsm pages, and 5000 fsm relations) to try and compensate. I believe this is only hurting us as any queries that choose to tablescan are only getting slower and slower. Also, obviously, our indexes are continually growing. The partitioning should help as the actual number of records being worked on each table is a very small percentage ( a site may have 1 million records, but only load and work a few thousand each day). The archive tables would be doing the most growing while the active tables should stay small. Most of the queries that are tablescanning can not be fixed as the database clusters have been initialized with a non-C locale and won't use indexes on our queries that are using like with a wild card. Right now, we are still on 7.3.4. However, these ideas would be implemented as part of an upgrade to 8.x (plus, we'll initialize the new clusters with a C locale). Anyway, I hope this makes since, and any comments, ideas, and/or suggestions would be appreciated. Thanks, Chris From pgsql-performance-owner@postgresql.org Wed Aug 24 15:04:42 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CC8ECD769F for ; Wed, 24 Aug 2005 14:39:26 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 12039-09 for ; Wed, 24 Aug 2005 17:39:21 +0000 (GMT) Received: from mail.envyfinancial.com (unknown [206.248.142.186]) by svr1.postgresql.org (Postfix) with ESMTP id 7AE3BD78AC for ; Wed, 24 Aug 2005 14:39:17 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by mail.envyfinancial.com (Postfix) with ESMTP id 603DD1B7F8; Wed, 24 Aug 2005 13:30:56 -0400 (EDT) Received: from mail.envyfinancial.com ([127.0.0.1]) by localhost (mark.mielke.cc [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 12034-02; Wed, 24 Aug 2005 13:30:56 -0400 (EDT) Received: by mail.envyfinancial.com (Postfix, from userid 500) id 2108F1B802; Wed, 24 Aug 2005 13:30:56 -0400 (EDT) Date: Wed, 24 Aug 2005 13:30:56 -0400 From: mark@mark.mielke.cc To: Donald Courtney Cc: William Yu , pgsql-performance@postgresql.org Subject: Re: Caching by Postgres Message-ID: <20050824173055.GA11788@mark.mielke.cc> References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> <430C7448.7030103@sun.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <430C7448.7030103@sun.com> User-Agent: Mutt/1.4.2.1i X-Virus-Scanned: amavisd-new at mail.envyfinancial.com X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.18 required=5 tests=[AWL=0.002, NO_REAL_NAME=0.178] X-Spam-Level: X-Archive-Number: 200508/394 X-Sequence-Number: 14141 On Wed, Aug 24, 2005 at 09:21:12AM -0400, Donald Courtney wrote: > I built postgreSQL 8.1 64K bit on solaris 10 a few months ago > and side by side with the 32 bit postgreSQL build saw no improvement. > In fact the 64 bit result was slightly lower. I've had this sort of argument with a friend of mine who works at a retail computer sales company who always tries to pitch 64-bit platforms to me (I don't have one yet). There are a few issues in here that are hard to properly detach to allow for a fair comparison. The first, to always remember - is that the move from 64-bits to 32-bits doesn't come for free. In a real 64-bit system with a 64-bit operating system, and 64-bit applications, pointers are now double their 32-bit size. This means more bytes to copy around memory, and in an extreme case, has the potential to approach halfing both the memory latency to access many such pointers from RAM, and half the effective amount of RAM. In real world cases, not everything is a pointer, so this sort of performance degradation is doubtful - but it is something to keep in mind. In response to this, it appears that, at least on the Intel/AMD side of things, they've increased the bandwidth on the motherboard, and allowed for faster memory to be connected to the motherboard. They've increased the complexity of the chip, to allow 64-bit register operations to be equivalent in speed to 32-bit register operations. I have no idea what else they've done... :-) So, it may be difficult to properly compare a 32-bit system to a 64-bit system. Even if the Ghz on the chip appears equal, it isn't the same chip, and unless it is the exact same make, product and version of the motherboard, it may not be a fair compairson. Turning support for 32-bit on or off, and using a kernel that is only 32-bit may give good comparisons - but with the above explanation, I would expect the 32-bit application + kernel to out-perform the 64-bit application. So then we move on to what 64-bit is really useful for. Obviously, there is the arithmetic. If you were previously doing 64-bit arithmetic through software, you will notice an immediate speed improvement when doing it through hardware instead. If you have a program that is scanning memory in any way, it may benefit from 64-bit instructions (for example - copying data 64-bit words at a time instead of 32-bit words at a time). PostgreSQL might benefit slightly from either of these, slightly balancing the performance degradation of using more memory to store the pointers, and more memory bandwidth the access the pointers. The real benefit of 64-bit is address space. From the kernel perspective, it means that more programs, or bigger programs can run at once. From the application perspective, it means your application can use more than 32-bits of address space. For programs that make extensive use of mmap(), this can be a necessity. They are mapping very large files into their own address space. This isn't a performance boost, as much as it is a 'you can't do it', if the files mmap()'ed at the same time, will not fit within 32-bits of address space. This also becomes, potentially, a performance degradation, as the system is now having to manage applications that have very large page tables. Page faults may become expensive. PostgreSQL uses read(), instead of mmap(), and uses <2 Gbyte files. PostgreSQL doesn't require the additional address space for normal operation. If, however, you happen to have a very large amount of physical memory - more memory than is supported by a 32-bit system, but is supported by your 64-bit system, then the operating system should be able to use this additional physical memory to cache file system data pages, which will benefit PostgreSQL if used with tables that are larger than the memory supported by your 32-bit system, and which have queries which require more pages than the memory supported by your 32-bit system to be frequently accessed. If you have a huge database, with many clients accessing the data, this would be a definate yes. With anything less, it is a maybe, or a probably not. I've been looking at switching to 64-bit, mostly to benefit from the better motherboard bandwidth, and just to play around. I'm not expecting to require the 64-bit instructions. Hope this helps, mark -- mark@mielke.cc / markm@ncf.ca / markm@nortel.com __________________________ . . _ ._ . . .__ . . ._. .__ . . . .__ | Neighbourhood Coder |\/| |_| |_| |/ |_ |\/| | |_ | |/ |_ | | | | | | \ | \ |__ . | | .|. |__ |__ | \ |__ | Ottawa, Ontario, Canada One ring to rule them all, one ring to find them, one ring to bring them all and in the darkness bind them... http://mark.mielke.cc/ From pgsql-performance-owner@postgresql.org Wed Aug 24 16:25:03 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A5E50D72B3 for ; Wed, 24 Aug 2005 15:47:22 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 61138-09 for ; Wed, 24 Aug 2005 18:47:20 +0000 (GMT) Received: from unicorn.rentec.com (unicorn.rentec.com [216.223.240.9]) by svr1.postgresql.org (Postfix) with ESMTP id B18E1D79B7 for ; Wed, 24 Aug 2005 15:47:18 -0300 (ADT) Received: from wren.rentec.com (wren.rentec.com [192.5.35.106]) by unicorn.rentec.com (8.13.1/8.12.1) with ESMTP id j7OIl9GK020673 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Wed, 24 Aug 2005 14:47:10 -0400 (EDT) X-Rentec: external Received: from [172.26.132.145] (hoopoe.rentec.com [172.26.132.145]) by wren.rentec.com (8.13.1/8.12.1) with ESMTP id j7OIl9ds022058; Wed, 24 Aug 2005 14:47:09 -0400 (EDT) Message-ID: <430CC0AD.5040400@rentec.com> Date: Wed, 24 Aug 2005 14:47:09 -0400 From: Alan Stange Reply-To: stange@rentec.com Organization: Renaissance Technologies Corp. User-Agent: Mozilla Thunderbird 1.0+ (X11/20050712) MIME-Version: 1.0 To: mark@mark.mielke.cc Cc: Donald Courtney , William Yu , pgsql-performance@postgresql.org Subject: Re: Caching by Postgres References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> <430C7448.7030103@sun.com> <20050824173055.GA11788@mark.mielke.cc> In-Reply-To: <20050824173055.GA11788@mark.mielke.cc> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Logged: Logged by unicorn.rentec.com as j7OIl9GK020673 at Wed Aug 24 14:47:10 2005 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.02 required=5 tests=[AWL=0.020] X-Spam-Level: X-Archive-Number: 200508/397 X-Sequence-Number: 14144 mark@mark.mielke.cc wrote: > So then we move on to what 64-bit is really useful for. Obviously, > there is the arithmetic. If you were previously doing 64-bit > arithmetic through software, you will notice an immediate speed > improvement when doing it through hardware instead. If you have > a program that is scanning memory in any way, it may benefit from > 64-bit instructions (for example - copying data 64-bit words at > a time instead of 32-bit words at a time). PostgreSQL might benefit > slightly from either of these, slightly balancing the performance > degradation of using more memory to store the pointers, and more > memory bandwidth the access the pointers. > At least on Sparc processors, v8 and newer, any double precision math (including longs) is performed with a single instruction, just like for a 32 bit datum. Loads and stores of 8 byte datums are also handled via a single instruction. The urban myth that 64bit math is different/better on a 64 bit processor is just that; yes, some lower end processors would emulate/trap those instructions but that an implementation detail, not architecture. I believe that this is all true for other RISC processors as well. The 64bit API on UltraSparcs does bring along some extra FP registers IIRC. > If, however, you happen to have a very large amount of physical memory > - more memory than is supported by a 32-bit system, but is supported > by your 64-bit system, then the operating system should be able to use > this additional physical memory to cache file system data pages, which > will benefit PostgreSQL if used with tables that are larger than the > memory supported by your 32-bit system, and which have queries which > require more pages than the memory supported by your 32-bit system to > be frequently accessed. If you have a huge database, with many clients > accessing the data, this would be a definate yes. With anything less, > it is a maybe, or a probably not. > Solaris, at least, provided support for far more than 4GB of physical memory on 32 bit kernels. A newer 64 bit kernel might be more efficient, but that's just because the time was taken to support large page sizes and more efficient data structures. It's nothing intrinsic to a 32 vs 64 bit kernel. -- Alan From pgsql-performance-owner@postgresql.org Wed Aug 24 16:06:23 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6B701D7903 for ; Wed, 24 Aug 2005 16:01:14 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 94524-03 for ; Wed, 24 Aug 2005 19:01:08 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id 11354D7771 for ; Wed, 24 Aug 2005 16:01:08 -0300 (ADT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.13.1/8.13.1) with ESMTP id j7OJ18BN005228 for ; Wed, 24 Aug 2005 19:01:08 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.13.1/8.13.1/Submit) id j7OIsmFU093991 for pgsql-performance@postgresql.org; Wed, 24 Aug 2005 18:54:48 GMT (envelope-from news) From: William Yu X-Newsgroups: pgsql.performance Subject: Re: Caching by Postgres Date: Wed, 24 Aug 2005 11:54:42 -0700 Organization: Hub.Org Networking Services Lines: 21 Message-ID: References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> <430C7448.7030103@sun.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Complaints-To: usenet@news.hub.org User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en In-Reply-To: <430C7448.7030103@sun.com> To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=-2.641 required=5 tests=[ALL_TRUSTED=-2.82, AWL=0.179] X-Spam-Level: X-Archive-Number: 200508/395 X-Sequence-Number: 14142 Donald Courtney wrote: > I built postgreSQL 8.1 64K bit on solaris 10 a few months ago > and side by side with the 32 bit postgreSQL build saw no improvement. In > fact the 64 bit result was slightly lower. I'm not surprised 32-bit binaries running on a 64-bit OS would be faster than 64-bit/64-bit. 64-bit isn't some magical wand you wave and it's all ok. Programs compiled as 64-bit will only run faster if (1) you need 64-bit address space and you've been using ugly hacks like PAE to get access to memory > 2GB or (2) you need native 64-bit data types and you've been using ugly hacks to piece 32-bit ints together (example, encryption/compression). In most cases, 64-bit will run slightly slower due to extra overhead of using larger datatypes. Since PostgreSQL hands off the majority of memory management/data caching to the OS, only the OS needs to be 64-bit to reap the benefits of better memory management. Since Postgres *ALREADY* reaps the 64-bit benefit, I'm not sure how the argument moving caching/mm/fs into Postgres would apply. Yes there's the point about possibly implementing better/smarter/more appropriate caching algorithms but that has nothing to do with 64-bit. From pgsql-performance-owner@postgresql.org Wed Aug 24 16:17:02 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A4050D72B3 for ; Wed, 24 Aug 2005 15:59:18 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 75965-08 for ; Wed, 24 Aug 2005 18:59:14 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id C0D51D7A01 for ; Wed, 24 Aug 2005 15:59:12 -0300 (ADT) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: Some ideas for comment Date: Wed, 24 Aug 2005 14:59:12 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3417DD1AB@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Some ideas for comment Thread-Index: AcWo0rYU7GIt7hM2Qe2fSAtg0O/haAACZp4Q From: "Merlin Moncure" To: "Chris Hoover" Cc: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/396 X-Sequence-Number: 14143 > Ok, there is always a lot of talk about tuning PostgreSQL on linux and > how PostgreSQL uses the linux kernel cache to cache the tables and > indexes. [...] >=20 > 1. Implement a partition type layout using views and rules - This > will allow me to have one table in each view with the "active" data, > and the inactive data stored by year in other tables. >=20 > So I would have the following (for each major table): >=20 > Table View as > select * from active_table > union all > select * from table_2005 > union all > select * from table_2004 > etc. Linux does a pretty good job of deciding what to cache. I don't think this will help much. You can always look at partial indexes too. > 2. I am also thinking of recommending we collapse all databases in a > cluster into one "mega" database. I can then use schema's and views > to control database access and ensure that no customer can see another > customers data. hm. keep in mind views are tightly bound to the tables they are created with (views can't 'float' over tables in different schemas). pl/pgsql functions can, though. This is a more efficient use of server resources, IMO, but not a windfall. =20 > This would mean that there are only one set of indexes being loaded > into the cache. While they would be larger, I think in combination > with the partition from idea 1, we would be ahead of the ball game. > Since there would only be one set of indexes, everyone would be > sharing them so they should always be in memory. I would strongly consider adding more memory :). =20 > I don't have real numbers to give you, but we know that our systems > are hurting i/o wise and we are growing by about 2GB+ per week (net). > We actually grow by about 5GB/week/server. However, when I run my > weekly maintenance of vacuum full, reindex, and the vacuum analyze, we > end up getting about 3GB back. Unfortunately, I do not have the i/o > bandwidth to vacuum during the day as it causes major slowdowns on our > system. Each night, I do run a vacuum analyze across all db's to try > and help. I also have my fsm parameters set high (8000000 fsm pages, > and 5000 fsm relations) to try and compensate. Generally, you can reduce data turnover for the same workload by normalizing your database. IOW, try and make your database more efficient in the way it stores data. > Right now, we are still on 7.3.4. However, these ideas would be > implemented as part of an upgrade to 8.x (plus, we'll initialize the > new clusters with a C locale). yes, do this! Merlin From pgsql-performance-owner@postgresql.org Wed Aug 24 17:08:27 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B6BA5D7803 for ; Wed, 24 Aug 2005 16:43:01 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 54356-08 for ; Wed, 24 Aug 2005 19:42:59 +0000 (GMT) Received: from mail.envyfinancial.com (unknown [206.248.142.186]) by svr1.postgresql.org (Postfix) with ESMTP id 1B354D6F72 for ; Wed, 24 Aug 2005 16:42:58 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by mail.envyfinancial.com (Postfix) with ESMTP id 9A9911BB55; Wed, 24 Aug 2005 15:34:41 -0400 (EDT) Received: from mail.envyfinancial.com ([127.0.0.1]) by localhost (mark.mielke.cc [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 13987-01; Wed, 24 Aug 2005 15:34:41 -0400 (EDT) Received: by mail.envyfinancial.com (Postfix, from userid 500) id 2C9791BB58; Wed, 24 Aug 2005 15:34:41 -0400 (EDT) Date: Wed, 24 Aug 2005 15:34:41 -0400 From: mark@mark.mielke.cc To: Alan Stange Cc: Donald Courtney , William Yu , pgsql-performance@postgresql.org Subject: Re: Caching by Postgres Message-ID: <20050824193441.GA13772@mark.mielke.cc> References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> <430C7448.7030103@sun.com> <20050824173055.GA11788@mark.mielke.cc> <430CC0AD.5040400@rentec.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <430CC0AD.5040400@rentec.com> User-Agent: Mutt/1.4.2.1i X-Virus-Scanned: amavisd-new at mail.envyfinancial.com X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.22 required=5 tests=[AWL=0.042, NO_REAL_NAME=0.178] X-Spam-Level: X-Archive-Number: 200508/400 X-Sequence-Number: 14147 On Wed, Aug 24, 2005 at 02:47:09PM -0400, Alan Stange wrote: > At least on Sparc processors, v8 and newer, any double precision math > (including longs) is performed with a single instruction, just like for > a 32 bit datum. Loads and stores of 8 byte datums are also handled via > a single instruction. The urban myth that 64bit math is > different/better on a 64 bit processor is just that; yes, some lower > end processors would emulate/trap those instructions but that an > implementation detail, not architecture. It isn't an urban myth that 64-bit math on a 64-bit processor is faster, at least if done using registers. It definately is faster. It may be an urban myth, though, that most applications perform a sufficient amount of 64-bit arithmetic to warrant the upgrade. The benefit may be lost in the noise for an application such as PostgreSQL. It takes, effectively, infinately longer to access a disk page, than to increment a 64-bit integer in software. For the lower end processors that emulate/trap these instructions, they are being performed in software, along with the overhead of a trap, and are therefore not a single instruction any more. We are coming at this from different sides (which is good - perspective is always good :-) ). From the Intel/AMD side of things, ALL non 64-bit platforms are 'lower end processors', and don't emulate/trap the instructions as they didn't exist (at least not yet - who knows what clever and sufficiently motivated people will do :-) ). > >If, however, you happen to have a very large amount of physical memory > >- more memory than is supported by a 32-bit system, but is supported > >by your 64-bit system, then the operating system should be able to use > >this additional physical memory to cache file system data pages, which > >will benefit PostgreSQL if used with tables that are larger than the > >memory supported by your 32-bit system, and which have queries which > >require more pages than the memory supported by your 32-bit system to > >be frequently accessed. If you have a huge database, with many clients > >accessing the data, this would be a definate yes. With anything less, > >it is a maybe, or a probably not. > Solaris, at least, provided support for far more than 4GB of physical > memory on 32 bit kernels. A newer 64 bit kernel might be more > efficient, but that's just because the time was taken to support large > page sizes and more efficient data structures. It's nothing intrinsic > to a 32 vs 64 bit kernel. Hehe. That's why I was so careful to qualify my statements. :-) But yeah, I agree. It's a lot of hype, for not much gain (and some loss, depending on what it is being used for). I only want one because they're built better, and because I want to play around. Cheers, mark -- mark@mielke.cc / markm@ncf.ca / markm@nortel.com __________________________ . . _ ._ . . .__ . . ._. .__ . . . .__ | Neighbourhood Coder |\/| |_| |_| |/ |_ |\/| | |_ | |/ |_ | | | | | | \ | \ |__ . | | .|. |__ |__ | \ |__ | Ottawa, Ontario, Canada One ring to rule them all, one ring to find them, one ring to bring them all and in the darkness bind them... http://mark.mielke.cc/ From pgsql-performance-owner@postgresql.org Wed Aug 24 16:51:46 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 143BFD7832 for ; Wed, 24 Aug 2005 16:51:42 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 75508-09 for ; Wed, 24 Aug 2005 19:51:40 +0000 (GMT) Received: from mx2.surnet.cl (mx2.surnet.cl [216.155.73.181]) by svr1.postgresql.org (Postfix) with ESMTP id C71C9D77E5 for ; Wed, 24 Aug 2005 16:51:37 -0300 (ADT) Received: from unknown (HELO cluster.surnet.cl) (216.155.73.164) by mx2.surnet.cl with ESMTP; 24 Aug 2005 16:51:16 -0400 X-IronPort-AV: i="3.96,138,1122868800"; d="scan'208"; a="2531923:sNHT94544510" Received: from alvh.no-ip.org (216.155.78.140) by cluster.surnet.cl (7.0.043) (authenticated as alvherre@surnet.cl) id 42B3EF600094A550; Wed, 24 Aug 2005 15:51:38 -0400 Received: by alvh.no-ip.org (Postfix, from userid 1000) id 2E120C2DC2A; Wed, 24 Aug 2005 15:51:43 -0400 (CLT) Date: Wed, 24 Aug 2005 15:51:43 -0400 From: Alvaro Herrera To: Chris Hoover Cc: pgsql-performance@postgresql.org Subject: Re: Some ideas for comment Message-ID: <20050824195143.GE17609@surnet.cl> Mail-Followup-To: Chris Hoover , pgsql-performance@postgresql.org References: <1d219a6f05082409563775a8a7@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1d219a6f05082409563775a8a7@mail.gmail.com> User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=2.069 required=5 tests=[AWL=-0.707, DNS_FROM_RFC_ABUSE=0.374, DNS_FROM_RFC_POST=1.376, RCVD_IN_NJABL_PROXY=1.026] X-Spam-Level: ** X-Archive-Number: 200508/398 X-Sequence-Number: 14145 On Wed, Aug 24, 2005 at 12:56:54PM -0400, Chris Hoover wrote: > I don't have real numbers to give you, but we know that our systems > are hurting i/o wise and we are growing by about 2GB+ per week (net). > We actually grow by about 5GB/week/server. However, when I run my > weekly maintenance of vacuum full, reindex, and the vacuum analyze, we > end up getting about 3GB back. Unfortunately, I do not have the i/o > bandwidth to vacuum during the day as it causes major slowdowns on our > system. Each night, I do run a vacuum analyze across all db's to try > and help. I also have my fsm parameters set high (8000000 fsm pages, > and 5000 fsm relations) to try and compensate. [...] > Right now, we are still on 7.3.4. However, these ideas would be > implemented as part of an upgrade to 8.x (plus, we'll initialize the > new clusters with a C locale). If you were on a newer version, I'd suggest that you use the cost-based vacuum delay, and vacuum at least some of the tables more often. This way you can reduce the continual growth of the data files without affecting day-to-day performance, because you allow the VACUUM-inflicted I/O to be interleaved by normal query execution. Sadly (for you), I think the cost-based vacuum delay feature was only introduced in 8.0. -- Alvaro Herrera () Officer Krupke, what are we to do? Gee, officer Krupke, Krup you! (West Side Story, "Gee, Officer Krupke") From pgsql-performance-owner@postgresql.org Wed Aug 24 17:03:49 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C457FD7974 for ; Wed, 24 Aug 2005 17:00:51 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91731-06 for ; Wed, 24 Aug 2005 20:00:49 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id CA1E3D76B7 for ; Wed, 24 Aug 2005 17:00:46 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7OK0jjW008482; Wed, 24 Aug 2005 16:00:45 -0400 (EDT) To: "Merlin Moncure" Cc: "Chris Hoover" , pgsql-performance@postgresql.org Subject: Re: Some ideas for comment In-reply-to: <6EE64EF3AB31D5448D0007DD34EEB3417DD1AB@Herge.rcsinc.local> References: <6EE64EF3AB31D5448D0007DD34EEB3417DD1AB@Herge.rcsinc.local> Comments: In-reply-to "Merlin Moncure" message dated "Wed, 24 Aug 2005 14:59:12 -0400" Date: Wed, 24 Aug 2005 16:00:45 -0400 Message-ID: <8481.1124913645@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/399 X-Sequence-Number: 14146 "Merlin Moncure" writes: >> Right now, we are still on 7.3.4. However, these ideas would be >> implemented as part of an upgrade to 8.x (plus, we'll initialize the >> new clusters with a C locale). > yes, do this! Moving from 7.3 to 8.0 is alone likely to give you a noticeable performance boost. regards, tom lane From pgsql-performance-owner@postgresql.org Wed Aug 24 18:37:54 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 91BC8D79DE for ; Wed, 24 Aug 2005 17:03:49 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 85114-06 for ; Wed, 24 Aug 2005 20:03:47 +0000 (GMT) Received: from boutiquenumerique.com (boutiquenumerique.com [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 0EBB4D79CD for ; Wed, 24 Aug 2005 17:03:45 -0300 (ADT) Received: (qmail 1694 invoked from network); 24 Aug 2005 22:04:08 +0200 Received: from unknown (HELO localhost) (boutiquenumerique-lists@192.168.0.4) by boutiquenumerique.com with SMTP; 24 Aug 2005 22:04:08 +0200 Date: Wed, 24 Aug 2005 22:03:43 +0200 To: stange@rentec.com, mark@mark.mielke.cc Subject: Re: Caching by Postgres Cc: "Donald Courtney" , "William Yu" , pgsql-performance@postgresql.org References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> <430C7448.7030103@sun.com> <20050824173055.GA11788@mark.mielke.cc> <430CC0AD.5040400@rentec.com> From: PFC Organization: =?iso-8859-15?Q?La_Boutique_Num=E9rique?= Content-Type: text/plain; format=flowed; delsp=yes; charset=iso-8859-15 MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Message-ID: In-Reply-To: <430CC0AD.5040400@rentec.com> User-Agent: Opera M2/8.0 (Linux, build 1095) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/405 X-Sequence-Number: 14152 > At least on Sparc processors, v8 and newer, any double precision math > (including longs) is performed with a single instruction, just like for > a 32 bit datum. Loads and stores of 8 byte datums are also handled via > a single instruction. The urban myth that 64bit math is > different/better on a 64 bit processor is just that; yes, some lower > end processors would emulate/trap those instructions but that an > implementation detail, not architecture. I believe that this is all > true for other RISC processors as well. > > The 64bit API on UltraSparcs does bring along some extra FP registers > IIRC. It's very different on x86. 64-bit x86 like the Opteron has more registers, which are very scarce on the base x86 (8 I think). This alone is very important. There are other factors as well. > Solaris, at least, provided support for far more than 4GB of physical > memory on 32 bit kernels. A newer 64 bit kernel might be more > efficient, but that's just because the time was taken to support large > page sizes and more efficient data structures. It's nothing intrinsic > to a 32 vs 64 bit kernel. Well, on a large working set, a processor which can directly address more than 4GB of memory will be a lot faster than one which can't, and has to play with the MMU and paging units ! From pgsql-performance-owner@postgresql.org Wed Aug 24 18:28:25 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2A707D77CC for ; Wed, 24 Aug 2005 17:10:48 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 05573-01 for ; Wed, 24 Aug 2005 20:10:44 +0000 (GMT) Received: from ausimss.pervasive.com (ausimss.pervasive.com [66.45.103.246]) by svr1.postgresql.org (Postfix) with ESMTP id 7C761D76B7 for ; Wed, 24 Aug 2005 17:10:39 -0300 (ADT) Received: from ausbayes2.aus.pervasive.com ([172.16.8.8]) by ausimss.pervasive.com with InterScan Messaging Security Suite; Wed, 24 Aug 2005 15:10:38 -0500 Received: from ausbayes1.aus.pervasive.com ([172.16.8.6]) by ausbayes2.aus.pervasive.com with Microsoft SMTPSVC(5.0.2195.6713); Wed, 24 Aug 2005 15:10:38 -0500 Received: from ausmailowa.aus.pervasive.com ([172.16.4.8]) by ausbayes1.aus.pervasive.com with Microsoft SMTPSVC(5.0.2195.6713); Wed, 24 Aug 2005 15:10:38 -0500 Received: from ausmail2k2.aus.pervasive.com ([172.16.4.9]) by ausmailowa.aus.pervasive.com with Microsoft SMTPSVC(6.0.3790.1830); Wed, 24 Aug 2005 15:10:37 -0500 X-MimeOLE: Produced By Microsoft Exchange V6.0.6249.0 content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Subject: Re: Read/Write block sizes Date: Wed, 24 Aug 2005 15:10:37 -0500 Message-ID: <072BDB2B234F3840B0AC03411084C9AF869911@ausmail2k2.aus.pervasive.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: Read/Write block sizes Thread-Index: AcWo5+SGCG6LxAh9SkKyyvvry40Dug== From: "Lance Obermeyer" To: "Bruce Momjian" , "Jignesh K. Shah" Cc: "Jim Nasby" , "Chris Browne" , X-OriginalArrivalTime: 24 Aug 2005 20:10:37.0777 (UTC) FILETIME=[E4B2EC10:01C5A8E7] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/404 X-Sequence-Number: 14151 Since Bruce referred to the "corporate software world" I'll chime in... It has been a while since adding knobs and dials has been considered a = good idea. Customers are almost always bad at tuning their systems, = which decreases customer satisfaction. While many people assume the = corporate types don't care, that is actually far from the truth. Well = run commercial software companies regularly commission (expensive) = customer satisfaction surveys. These numbers are the second most = important numbers in all of the enterprise, trailing only revenue in = importance. Results are sliced and diced in every way imaginable. The commercial world is trying to auto-tune their systems just as much. = Examples are the work that many of the big boys are doing towards = "autonomic" computing. While it is driven by naked self interest of = wanting to sell version upgrades, those efforts increase customer = satisfaction and decrease support costs. Works well for everyone... -----Original Message----- From: Bruce Momjian [mailto:pgman@candle.pha.pa.us] Sent: Wednesday, August 24, 2005 8:52 AM To: Jignesh K. Shah Cc: Jim Nasby; Chris Browne; pgsql-performance@postgresql.org Subject: Re: Read/Write block sizes This thread covers several performance ideas. First is the idea that more parameters should be configurable. While this seems like a noble goal, we try to make parameters auto-tuning, or if users have to configure it, the parameter should be useful for a significant number of users. In the commercial software world, if you can convince your boss that a feature/knob is useful, it usually gets into the product.=20 Unfortunately, this leads to the golden doorknob on a shack, where some features are out of sync with the rest of the product in terms of usefulness and utility. With open source, if a feature can not be auto-tuned, or has significant overhead, the features has to be implemented and then proven to be a benefit. In terms of adding async I/O, threading, and other things, it might make sense to explore how these could be implemented in a way that fits the above criteria. -------------------------------------------------------------------------= -- Jignesh K. Shah wrote: > Hi Jim, >=20 > | How many of these things are currently easy to change with a = recompile? > | I should be able to start testing some of these ideas in the near > | future, if they only require minor code or configure changes. >=20 >=20 > The following > * Data File Size 1GB > * WAL File Size of 16MB > * Block Size of 8K >=20 > Are very easy to change with a recompile.. A Tunable will be greatly=20 > prefered as it will allow one binary for different tunings >=20 > * MultiBlock read/write >=20 > Is not available but will greatly help in reducing the number of = system=20 > calls which will only increase as the size of the database increases = if=20 > something is not done about i. >=20 > * Pregrown files... maybe not important at this point since TABLESPACE = > can currently work around it a bit (Just need to create a different = file=20 > system for each tablespace >=20 > But if you really think hardware & OS is the answer for all small=20 > things...... I think we should now start to look on how to make = Postgres=20 > Multi-threaded or multi-processed for each connection. With the influx = > of "Dual-Core" or "Multi-Core" being the fad.... Postgres can have = the=20 > cutting edge if somehow exploiting cores is designed. >=20 > Somebody mentioned that adding CPU to Postgres workload halved the=20 > average CPU usage... > YEAH... PostgreSQL uses only 1 CPU per connection (assuming 100%=20 > usage) so if you add another CPU it is idle anyway and the system = will=20 > report only 50% :-) BUT the importing to measure is.. whether the = query=20 > time was cut down or not? ( No flames I am sure you were talking about = > multi-connection multi-user environment :-) ) But my point is then = this=20 > approach is worth the ROI and the time and effort spent to solve this=20 > problem. >=20 > I actually vote for a multi-threaded solution for each connection = while=20 > still maintaining seperate process for each connections... This way = the=20 > fundamental architecture of Postgres doesn't change, however a=20 > multi-threaded connection can then start to exploit different cores..=20 > (Maybe have tunables for number of threads to read data files who=20 > knows.. If somebody is interested in actually working a design ..=20 > contact me and I will be interested in assisting this work. >=20 > Regards, > Jignesh >=20 >=20 > Jim C. Nasby wrote: >=20 > >On Tue, Aug 23, 2005 at 06:09:09PM -0400, Chris Browne wrote: > > =20 > > > >>J.K.Shah@Sun.COM (Jignesh Shah) writes: > >> =20 > >> > >>>>Does that include increasing the size of read/write blocks? I've > >>>>noticedthat with a large enough table it takes a while to do a > >>>>sequential scan, even if it's cached; I wonder if the fact that it > >>>>takes a million read(2) calls to get through an 8G table is part = of > >>>>that. > >>>> =20 > >>>> > >>>Actually some of that readaheads,etc the OS does already if it does > >>>some sort of throttling/clubbing of reads/writes. But its not = enough > >>>for such types of workloads. > >>> > >>>Here is what I think will help: > >>> > >>>* Support for different Blocksize TABLESPACE without recompiling = the > >>>code.. (Atlease support for a different Blocksize for the whole > >>>database without recompiling the code) > >>> > >>>* Support for bigger sizes of WAL files instead of 16MB files > >>>WITHOUT recompiling the code.. Should be a tuneable if you ask me > >>>(with checkpoint_segments at 256.. you have too many 16MB files in > >>>the log directory) (This will help OLTP benchmarks more since now > >>>they don't spend time rotating log files) > >>> > >>>* Introduce a multiblock or extent tunable variable where you can > >>>define a multiple of 8K (or BlockSize tuneable) to read a bigger > >>>chunk and store it in the bufferpool.. (Maybe writes too) (Most > >>>devices now support upto 1MB chunks for reads and writes) > >>> > >>>*There should be a way to preallocate files for TABLES in > >>>TABLESPACES otherwise with multiple table writes in the same > >>>filesystem ends with fragmented files which causes poor "READS" = from > >>>the files. > >>> > >>>* With 64bit 1GB file chunks is also moot.. Maybe it should be > >>>tuneable too like 100GB without recompiling the code. > >>> > >>>Why recompiling is bad? Most companies that will support Postgres > >>>will support their own binaries and they won't prefer different > >>>versions of binaries for different blocksizes, different WAL file > >>>sizes, etc... and hence more function using the same set of = binaries > >>>is more desirable in enterprise environments > >>> =20 > >>> > >>Every single one of these still begs the question of whether the > >>changes will have a *material* impact on performance. > >> =20 > >> >=20 >=20 > ---------------------------(end of = broadcast)--------------------------- > TIP 3: Have you checked our extensive FAQ? >=20 > http://www.postgresql.org/docs/faq >=20 --=20 Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania = 19073 From pgsql-performance-owner@postgresql.org Wed Aug 24 18:19:02 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 61578D79A1 for ; Wed, 24 Aug 2005 17:20:42 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 04768-07 for ; Wed, 24 Aug 2005 20:20:37 +0000 (GMT) Received: from brmea-mail-3.sun.com (brmea-mail-3.Sun.COM [192.18.98.34]) by svr1.postgresql.org (Postfix) with ESMTP id EE41DD78D3 for ; Wed, 24 Aug 2005 17:20:33 -0300 (ADT) Received: from phys-bur-1 ([129.148.9.72]) by brmea-mail-3.sun.com (8.12.10/8.12.9) with ESMTP id j7OKKYWZ000041 for ; Wed, 24 Aug 2005 14:20:35 -0600 (MDT) Received: from conversion-daemon.bur-mail1.east.sun.com by bur-mail1.east.sun.com (iPlanet Messaging Server 5.2 HotFix 1.24 (built Dec 19 2003)) id <0ILQ00F01TU0ZC@bur-mail1.east.sun.com> (original mail from J.K.Shah@Sun.COM) for pgsql-performance@postgresql.org; Wed, 24 Aug 2005 16:20:34 -0400 (EDT) Received: from bur-mail1.east.sun.com (phys-bur-1 [129.148.9.72]) by bur-mail1.east.sun.com (iPlanet Messaging Server 5.2 HotFix 1.24 (built Dec 19 2003)) with ESMTP id <0ILQ00E43TUAD8@bur-mail1.east.sun.com>; Wed, 24 Aug 2005 16:20:34 -0400 (EDT) Received: from [129.148.168.2] by bur-mail1.east.sun.com (mshttpd); Wed, 24 Aug 2005 16:20:34 -0400 Date: Wed, 24 Aug 2005 16:20:34 -0400 From: Jignesh Shah Subject: Re: Read/Write block sizes To: Lance Obermeyer Cc: Bruce Momjian , Jim Nasby , Chris Browne , pgsql-performance@postgresql.org Message-id: MIME-version: 1.0 X-Mailer: iPlanet Messenger Express 5.2 HotFix 1.24 (built Dec 19 2003) Content-type: text/plain; charset=us-ascii Content-language: en Content-transfer-encoding: 7BIT Content-disposition: inline X-Accept-Language: en X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/403 X-Sequence-Number: 14150 Agreed!!! But the knowledge to Auto-tune your application comes from years of understanding of how users are using the so-called "knobs".. But if the "knobs" are not there in the first place.. how do you know what people are using? The "so-called" big boys are also using their knowledge base of what works for the customer in their autonomic self healers and its based on the experience of all the settings possible and based on service requests on what had failed that they get the knowledge about avoiding what fails and tuning what works. Remember "recompiling" is a risk with upteem number of variables which not every production release engineer is happy about. Its easy to change back the knob to the previous value rather than trying to figure out how do I get my old binaries back. -Jignesh ----- Original Message ----- From: Lance Obermeyer Date: Wednesday, August 24, 2005 4:10 pm Subject: RE: Read/Write block sizes > Since Bruce referred to the "corporate software world" I'll chime > in... > It has been a while since adding knobs and dials has been > considered a good idea. Customers are almost always bad at tuning > their systems, which decreases customer satisfaction. While many > people assume the corporate types don't care, that is actually far > from the truth. Well run commercial software companies regularly > commission (expensive) customer satisfaction surveys. These > numbers are the second most important numbers in all of the > enterprise, trailing only revenue in importance. Results are > sliced and diced in every way imaginable. > > The commercial world is trying to auto-tune their systems just as > much. Examples are the work that many of the big boys are doing > towards "autonomic" computing. While it is driven by naked self > interest of wanting to sell version upgrades, those efforts > increase customer satisfaction and decrease support costs. Works > well for everyone... > > > > -----Original Message----- > From: Bruce Momjian [pgman@candle.pha.pa.us] > Sent: Wednesday, August 24, 2005 8:52 AM > To: Jignesh K. Shah > Cc: Jim Nasby; Chris Browne; pgsql-performance@postgresql.org > Subject: Re: Read/Write block sizes > > > > This thread covers several performance ideas. First is the idea that > more parameters should be configurable. While this seems like a > noblegoal, we try to make parameters auto-tuning, or if users have to > configure it, the parameter should be useful for a significant > number of > users. > > In the commercial software world, if you can convince your boss > that a > feature/knob is useful, it usually gets into the product. > Unfortunately, this leads to the golden doorknob on a shack, where > somefeatures are out of sync with the rest of the product in terms of > usefulness and utility. With open source, if a feature can not be > auto-tuned, or has significant overhead, the features has to be > implemented and then proven to be a benefit. > > In terms of adding async I/O, threading, and other things, it might > makesense to explore how these could be implemented in a way that > fits the > above criteria. > > -------------------------------------------------------------------- > ------- > > Jignesh K. Shah wrote: > > Hi Jim, > > > > | How many of these things are currently easy to change with a > recompile?> | I should be able to start testing some of these ideas > in the near > > | future, if they only require minor code or configure changes. > > > > > > The following > > * Data File Size 1GB > > * WAL File Size of 16MB > > * Block Size of 8K > > > > Are very easy to change with a recompile.. A Tunable will be > greatly > > prefered as it will allow one binary for different tunings > > > > * MultiBlock read/write > > > > Is not available but will greatly help in reducing the number of > system > > calls which will only increase as the size of the database > increases if > > something is not done about i. > > > > * Pregrown files... maybe not important at this point since > TABLESPACE > > can currently work around it a bit (Just need to create a > different file > > system for each tablespace > > > > But if you really think hardware & OS is the answer for all > small > > things...... I think we should now start to look on how to make > Postgres > > Multi-threaded or multi-processed for each connection. With the > influx > > of "Dual-Core" or "Multi-Core" being the fad.... Postgres can > have the > > cutting edge if somehow exploiting cores is designed. > > > > Somebody mentioned that adding CPU to Postgres workload halved > the > > average CPU usage... > > YEAH... PostgreSQL uses only 1 CPU per connection (assuming 100% > > usage) so if you add another CPU it is idle anyway and the > system will > > report only 50% :-) BUT the importing to measure is.. whether > the query > > time was cut down or not? ( No flames I am sure you were talking > about > > multi-connection multi-user environment :-) ) But my point is > then this > > approach is worth the ROI and the time and effort spent to solve > this > > problem. > > > > I actually vote for a multi-threaded solution for each connection > while > > still maintaining seperate process for each connections... This > way the > > fundamental architecture of Postgres doesn't change, however a > > multi-threaded connection can then start to exploit different > cores.. > > (Maybe have tunables for number of threads to read data files who > > knows.. If somebody is interested in actually working a design .. > > contact me and I will be interested in assisting this work. > > > > Regards, > > Jignesh > > > > > > Jim C. Nasby wrote: > > > > >On Tue, Aug 23, 2005 at 06:09:09PM -0400, Chris Browne wrote: > > > > > > > > >>J.K.Shah@Sun.COM (Jignesh Shah) writes: > > >> > > >> > > >>>>Does that include increasing the size of read/write blocks? I've > > >>>>noticedthat with a large enough table it takes a while to do a > > >>>>sequential scan, even if it's cached; I wonder if the fact > that it > > >>>>takes a million read(2) calls to get through an 8G table is > part of > > >>>>that. > > >>>> > > >>>> > > >>>Actually some of that readaheads,etc the OS does already if it > does> >>>some sort of throttling/clubbing of reads/writes. But its > not enough > > >>>for such types of workloads. > > >>> > > >>>Here is what I think will help: > > >>> > > >>>* Support for different Blocksize TABLESPACE without > recompiling the > > >>>code.. (Atlease support for a different Blocksize for the whole > > >>>database without recompiling the code) > > >>> > > >>>* Support for bigger sizes of WAL files instead of 16MB files > > >>>WITHOUT recompiling the code.. Should be a tuneable if you ask me > > >>>(with checkpoint_segments at 256.. you have too many 16MB > files in > > >>>the log directory) (This will help OLTP benchmarks more since now > > >>>they don't spend time rotating log files) > > >>> > > >>>* Introduce a multiblock or extent tunable variable where you can > > >>>define a multiple of 8K (or BlockSize tuneable) to read a bigger > > >>>chunk and store it in the bufferpool.. (Maybe writes too) (Most > > >>>devices now support upto 1MB chunks for reads and writes) > > >>> > > >>>*There should be a way to preallocate files for TABLES in > > >>>TABLESPACES otherwise with multiple table writes in the same > > >>>filesystem ends with fragmented files which causes poor > "READS" from > > >>>the files. > > >>> > > >>>* With 64bit 1GB file chunks is also moot.. Maybe it should be > > >>>tuneable too like 100GB without recompiling the code. > > >>> > > >>>Why recompiling is bad? Most companies that will support Postgres > > >>>will support their own binaries and they won't prefer different > > >>>versions of binaries for different blocksizes, different WAL file > > >>>sizes, etc... and hence more function using the same set of > binaries> >>>is more desirable in enterprise environments > > >>> > > >>> > > >>Every single one of these still begs the question of whether the > > >>changes will have a *material* impact on performance. > > >> > > >> > > > > > > ---------------------------(end of broadcast)--------------------- > ------ > > TIP 3: Have you checked our extensive FAQ? > > > > http://www.postgresql.org/docs/faq > > > > -- > Bruce Momjian | http://candle.pha.pa.us > pgman@candle.pha.pa.us | (610) 359-1001 > + If your life is a hard drive, | 13 Roberts Road > + Christ can be your backup. | Newtown Square, > Pennsylvania 19073 > From pgsql-performance-owner@postgresql.org Wed Aug 24 18:14:29 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BCB61D7A51 for ; Wed, 24 Aug 2005 17:26:44 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 04229-10 for ; Wed, 24 Aug 2005 20:26:41 +0000 (GMT) Received: from zproxy.gmail.com (zproxy.gmail.com [64.233.162.202]) by svr1.postgresql.org (Postfix) with ESMTP id 767A6D7A44 for ; Wed, 24 Aug 2005 17:26:39 -0300 (ADT) Received: by zproxy.gmail.com with SMTP id 8so111568nzo for ; Wed, 24 Aug 2005 13:26:41 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=E90xCPGHHNzOfIFscQm1vVxK7ta4PedI01UeMIfAugZMm4fFzPG+raPwF0G47oi5/9+0L0HTrPw9P+9OcPdziYgBoOHGCuXBWd58o+MjpwgGN/W71sKQyq2nQZXDSErZIayi3hhTaEr/TkfUFMtAF58yXzTXdBaO0pkREy5vtLo= Received: by 10.36.10.20 with SMTP id 20mr3304159nzj; Wed, 24 Aug 2005 13:26:40 -0700 (PDT) Received: by 10.36.24.2 with HTTP; Wed, 24 Aug 2005 13:26:40 -0700 (PDT) Message-ID: <1d219a6f05082413265e872134@mail.gmail.com> Date: Wed, 24 Aug 2005 16:26:40 -0400 From: Chris Hoover To: Merlin Moncure Subject: Re: Some ideas for comment Cc: pgsql-performance@postgresql.org In-Reply-To: <6EE64EF3AB31D5448D0007DD34EEB3417DD1AB@Herge.rcsinc.local> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <6EE64EF3AB31D5448D0007DD34EEB3417DD1AB@Herge.rcsinc.local> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[AWL=-0.000, RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/402 X-Sequence-Number: 14149 On 8/24/05, Merlin Moncure wrote: > > Ok, there is always a lot of talk about tuning PostgreSQL on linux and > > how PostgreSQL uses the linux kernel cache to cache the tables and > > indexes. > [...] > > > > 1. Implement a partition type layout using views and rules - This > > will allow me to have one table in each view with the "active" data, > > and the inactive data stored by year in other tables. > > > > So I would have the following (for each major table): > > > > Table View as > > select * from active_table > > union all > > select * from table_2005 > > union all > > select * from table_2004 > > etc. >=20 > Linux does a pretty good job of deciding what to cache. I don't think > this will help much. You can always look at partial indexes too. >=20 Yes, but won't this help create the need to store less? If I have 1,000.000 rows in a table, but only 4,000 are active, if I move those 4 to another table and link the tables via a view, should that not help keep the 9,996,000 rows out of the kernel cache (the majority of the time at least)? This would mean I have more room for other objects and hopefully less turn over in the cache, and less disk i/o. Yes? [...] > I would strongly consider adding more memory :). Unfortunately, it looks like 12GB is all our Dell servers can handle. :( >=20 > > I don't have real numbers to give you, but we know that our systems > > are hurting i/o wise and we are growing by about 2GB+ per week (net). > > We actually grow by about 5GB/week/server. However, when I run my > > weekly maintenance of vacuum full, reindex, and the vacuum analyze, we > > end up getting about 3GB back. Unfortunately, I do not have the i/o > > bandwidth to vacuum during the day as it causes major slowdowns on our > > system. Each night, I do run a vacuum analyze across all db's to try > > and help. I also have my fsm parameters set high (8000000 fsm pages, > > and 5000 fsm relations) to try and compensate. >=20 > Generally, you can reduce data turnover for the same workload by > normalizing your database. IOW, try and make your database more > efficient in the way it stores data. >=20 That's the ultimate goal, but this database structure was developed and released into production before I started work here. I'm trying to slowly change it into a better db, but it is a slow process.=20 Normalization does not make it at the top of the priority list, unfortunately. > > Right now, we are still on 7.3.4. However, these ideas would be > > implemented as part of an upgrade to 8.x (plus, we'll initialize the > > new clusters with a C locale). > > > 2. I am also thinking of recommending we collapse all databases in a > > cluster into one "mega" database. I can then use schema's and views > > to control database access and ensure that no customer can see another > > customers data. >=20 > hm. keep in mind views are tightly bound to the tables they are created > with (views can't 'float' over tables in different schemas). pl/pgsql > functions can, though. This is a more efficient use of server > resources, IMO, but not a windfall. This I know. Each schema would have to have a "custom" set of views replacing the tables with the view programmed to only return that customers data. I was thinking all of the tables in schema my_tables and the views all querying the tables stored in the my_tables schema. I would add an identifying column to each table so that I can differentiate the data. Chris From pgsql-performance-owner@postgresql.org Wed Aug 24 17:52:04 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8D0CCD7A1E for ; Wed, 24 Aug 2005 17:43:53 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 82006-08 for ; Wed, 24 Aug 2005 20:43:50 +0000 (GMT) Received: from zproxy.gmail.com (zproxy.gmail.com [64.233.162.196]) by svr1.postgresql.org (Postfix) with ESMTP id 45064D7A12 for ; Wed, 24 Aug 2005 17:43:49 -0300 (ADT) Received: by zproxy.gmail.com with SMTP id 8so113823nzo for ; Wed, 24 Aug 2005 13:43:51 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:mime-version:content-type:content-transfer-encoding:content-disposition; b=O11vR1UMsMxWbs+ugX+la4+iLi+numSsTJc+ZIQzTnMVSBGm4MEcQBuUmrFBBuXdXVhuhMyJv5Dy+5UWwWpLIIqiiMsp1Wo4qsl2+MOmTWk8WWMdc7S66kGLqc/5o22zPTi18hOg6bh62XS8MG9nqasi1lrdpHFlm/d6N/Uq9rs= Received: by 10.37.2.4 with SMTP id e4mr3022nzi; Wed, 24 Aug 2005 13:43:51 -0700 (PDT) Received: by 10.36.158.7 with HTTP; Wed, 24 Aug 2005 13:43:51 -0700 (PDT) Message-ID: Date: Wed, 24 Aug 2005 14:43:51 -0600 From: Mark Fox To: pgsql-performance@postgresql.org Subject: Performance indexing of a simple query Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.177 required=5 tests=[AWL=0.153, RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/401 X-Sequence-Number: 14148 I have a table called 'jobs' with several million rows, and the only columns that are important to this discussion are 'start_time' and 'completion_time'. The sort of queries I want to execute (among others) are like: SELECT * FROM jobs WHERE completion_time > SOMEDATE AND start_time < SOMEDATE; In plain english: All the jobs that were running at SOMEDATE. The result of the query is on the order of 500 rows. I've got seperate indexes on 'start_time' and 'completion_time'. Now, if SOMEDATE is such that the number of rows with completion_time > SOMEDATE is small (say 10s of thousands), the query uses index scans and executes quickly. If not, the query uses sequential scans and is unacceptably slow (a couple of minutes). I've used EXPLAIN and EXPLAIN ANALYZE to confirm this. This makes perfect sense to me. I've played with some of the memory settings for PostgreSQL, but none has had a significant impact. Any ideas on how to structure the query or add/change indexes in such a way to improve its performance? In desperation, I tried using a subquery, but unsurprisingly it made no (positive) difference. I feel like there might be a way of using an index on both 'completion_time' and 'start_time', but can't put a temporal lobe on the details. Mark From pgsql-performance-owner@postgresql.org Wed Aug 24 19:57:22 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 54ED0D78DF for ; Wed, 24 Aug 2005 17:57:49 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 95843-01 for ; Wed, 24 Aug 2005 20:57:46 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [67.100.216.10]) by svr1.postgresql.org (Postfix) with ESMTP id AF010D7832 for ; Wed, 24 Aug 2005 17:57:44 -0300 (ADT) Received: by flake.decibel.org (Postfix, from userid 1001) id B8E5C15255; Wed, 24 Aug 2005 15:57:44 -0500 (CDT) Date: Wed, 24 Aug 2005 15:57:44 -0500 From: "Jim C. Nasby" To: Chris Browne Cc: pgsql-performance@postgresql.org Subject: Re: Read/Write block sizes Message-ID: <20050824205744.GX96732@pervasive.com> References: <60k6icpapm.fsf@dba2.int.libertyrms.com> <1124846743.12045.94.camel@amd64-laptop-spoe> <60br3npb4p.fsf@dba2.int.libertyrms.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <60br3npb4p.fsf@dba2.int.libertyrms.com> X-Operating-System: FreeBSD 4.11-RELEASE-p10 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.01 required=5 tests=[AWL=0.010] X-Spam-Level: X-Archive-Number: 200508/409 X-Sequence-Number: 14156 On Wed, Aug 24, 2005 at 12:12:22PM -0400, Chris Browne wrote: > Everyone involved in development seems to me to have a reasonably keen > understanding as to what the potential benefits of threading are; the > value is that there fall out plenty of opportunities to parallelize > the evaluation of portions of queries. Alas, it wouldn't be until > *after* all the effort goes in that we would get any idea as to what > kinds of speedups this would provide. My understanding is that the original suggestion was to use threads within individual backends to allow for parallel query execution, not swiching to a completely thread-based model. In any case, there are other ways to enable parallelism without using threads, such as handing actual query execution off to a set of processes. -- Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com Pervasive Software http://pervasive.com 512-569-9461 From pgsql-performance-owner@postgresql.org Wed Aug 24 19:45:31 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E9B7FD7A59 for ; Wed, 24 Aug 2005 18:10:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 64704-10 for ; Wed, 24 Aug 2005 21:10:24 +0000 (GMT) Received: from unicorn.rentec.com (unicorn.rentec.com [216.223.240.9]) by svr1.postgresql.org (Postfix) with ESMTP id F39C2D7A19 for ; Wed, 24 Aug 2005 18:10:23 -0300 (ADT) Received: from ram.rentec.com (mailhost [192.5.35.66]) by unicorn.rentec.com (8.13.1/8.12.1) with ESMTP id j7OL959w025968 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Wed, 24 Aug 2005 17:09:06 -0400 (EDT) X-Rentec: external Received: from [172.26.132.145] (hoopoe.rentec.com [172.26.132.145]) by ram.rentec.com (8.13.1/8.12.1) with ESMTP id j7OL949a019445; Wed, 24 Aug 2005 17:09:05 -0400 (EDT) Message-ID: <430CE1F0.3080205@rentec.com> Date: Wed, 24 Aug 2005 17:09:04 -0400 From: Alan Stange Reply-To: stange@rentec.com Organization: Renaissance Technologies Corp. User-Agent: Mozilla Thunderbird 1.0+ (X11/20050712) MIME-Version: 1.0 To: mark@mark.mielke.cc Cc: Donald Courtney , pgsql-performance@postgresql.org Subject: Re: Caching by Postgres References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> <430C7448.7030103@sun.com> <20050824173055.GA11788@mark.mielke.cc> <430CC0AD.5040400@rentec.com> <20050824193441.GA13772@mark.mielke.cc> In-Reply-To: <20050824193441.GA13772@mark.mielke.cc> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Logged: Logged by unicorn.rentec.com as j7OL959w025968 at Wed Aug 24 17:09:06 2005 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.019 required=5 tests=[AWL=0.019] X-Spam-Level: X-Archive-Number: 200508/407 X-Sequence-Number: 14154 mark@mark.mielke.cc wrote: > On Wed, Aug 24, 2005 at 02:47:09PM -0400, Alan Stange wrote: > >> At least on Sparc processors, v8 and newer, any double precision math >> (including longs) is performed with a single instruction, just like for >> a 32 bit datum. Loads and stores of 8 byte datums are also handled via >> a single instruction. The urban myth that 64bit math is >> different/better on a 64 bit processor is just that; yes, some lower >> end processors would emulate/trap those instructions but that an >> implementation detail, not architecture. >> > > It isn't an urban myth that 64-bit math on a 64-bit processor is > faster, at least if done using registers. It definately is faster. > The older 32bit RISC processors do have 64 bit registers, ALUs and datapaths, and they are marketed toward high end scientific computing, and you're claiming that such a processor is slower than one which has the addition of 64 bit pointers added to it? As an example, an UltraSparc running a 32 bit kernel+application will have the same double precision floating point performance as one running a 64bit kernel+application (except for the additional FP registers in the 64bit API). For a function like daxpy, it's the exact same hardware running the exact same instructions! So why do you think the performance would be different? I believe the IBM Power processors also upped everything to double precision internally because of some details of the "multiply-add fused" instructions. It's been a few years since I taught H&P to CS undergrads, but I'm fairly sure the details are all the same for MIPS processors as well. -- Alan From pgsql-performance-owner@postgresql.org Wed Aug 24 19:49:36 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 95FB0D79B8 for ; Wed, 24 Aug 2005 18:09:19 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 94582-07 for ; Wed, 24 Aug 2005 21:09:16 +0000 (GMT) Received: from vms042pub.verizon.net (vms042pub.verizon.net [206.46.252.42]) by svr1.postgresql.org (Postfix) with ESMTP id C551AD78C3 for ; Wed, 24 Aug 2005 18:09:12 -0300 (ADT) Received: from osgiliath.mathom.us ([151.200.14.55]) by vms042.mailsrvcs.net (Sun Java System Messaging Server 6.2 HotFix 0.04 (built Dec 24 2004)) with ESMTPA id <0ILQ009HTW3950E1@vms042.mailsrvcs.net> for pgsql-performance@postgresql.org; Wed, 24 Aug 2005 16:09:11 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by osgiliath.mathom.us (Postfix) with ESMTP id 861F060263C for ; Wed, 24 Aug 2005 17:09:09 -0400 (EDT) Received: from osgiliath.mathom.us ([127.0.0.1]) by localhost (osgiliath [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 29379-02-8 for ; Wed, 24 Aug 2005 17:09:09 -0400 (EDT) Received: by osgiliath.mathom.us (Postfix, from userid 1000) id 6A627602638; Wed, 24 Aug 2005 17:09:09 -0400 (EDT) Date: Wed, 24 Aug 2005 17:09:09 -0400 From: Michael Stone Subject: Re: Caching by Postgres In-reply-to: <20050824193441.GA13772@mark.mielke.cc> To: pgsql-performance@postgresql.org Mail-Followup-To: pgsql-performance@postgresql.org Message-id: <20050824210909.GN8667@mathom.us> MIME-version: 1.0 Content-type: text/plain; charset=us-ascii; format=flowed Content-disposition: inline X-Pgp-Fingerprint: 53 FF 38 00 E7 DD 0A 9C 84 52 84 C5 EE DF 7C 88 X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mathom.us References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> <430C7448.7030103@sun.com> <20050824173055.GA11788@mark.mielke.cc> <430CC0AD.5040400@rentec.com> <20050824193441.GA13772@mark.mielke.cc> User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.041 required=5 tests=[AWL=-0.009, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/408 X-Sequence-Number: 14155 On Wed, Aug 24, 2005 at 03:34:41PM -0400, mark@mark.mielke.cc wrote: >It isn't an urban myth that 64-bit math on a 64-bit processor is >faster, at least if done using registers. It definately is faster. >It may be an urban myth, though, that most applications perform >a sufficient amount of 64-bit arithmetic to warrant the upgrade. The mjor problem is that the definition of "64bit processor" is fuzzy. The major slowdown of "64bitness" is the necessity of carting around 64 bit pointers. It's not, however, necessary to do 64bit pointers to get 64bit registers & fast 64 bit ops. E.g., sgi has "n32" & "n64" abi's which can access exactly the same instruction set & registers, the difference between them is the size of pointers and whether a "long" is the same as a "long long". Any discussion of "64 bit processors" is doomed from the start because people tend to start making implementation assumptions on top of an already vague concept. Current & future discussions are tinged by the fact that amd64 *doubles* the number of registers in 64 bit mode, potentially providing a major speedup--but one that doesn't really have anything to do with being "64bit". Pretty much any discussion of 64 bit mode really needs to be a discussion of a particular abi on a particular processor; talking about "64 bit processors" abstractly is a waste of time. Mike Stone From pgsql-performance-owner@postgresql.org Wed Aug 24 19:21:49 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6321BD6FF5 for ; Wed, 24 Aug 2005 18:22:41 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 75415-02 for ; Wed, 24 Aug 2005 21:22:34 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [67.100.216.10]) by svr1.postgresql.org (Postfix) with ESMTP id 9A75CD7A64 for ; Wed, 24 Aug 2005 18:22:32 -0300 (ADT) Received: by flake.decibel.org (Postfix, from userid 1001) id DD02A1527C; Wed, 24 Aug 2005 16:22:34 -0500 (CDT) Date: Wed, 24 Aug 2005 16:22:34 -0500 From: "Jim C. Nasby" To: Mark Fox Cc: pgsql-performance@postgresql.org Subject: Re: Performance indexing of a simple query Message-ID: <20050824212234.GZ96732@pervasive.com> References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: X-Operating-System: FreeBSD 4.11-RELEASE-p10 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.009 required=5 tests=[AWL=0.009] X-Spam-Level: X-Archive-Number: 200508/406 X-Sequence-Number: 14153 Try CREATE INDEX start_complete ON jobs( start_time, completion_time ); Try also completion_time, start_time. One might work better than the other. Or, depending on your data, you might want to keep both. In 8.1 you'll be able to do bitmap-based index combination, which might allow making use of the seperate indexes. On Wed, Aug 24, 2005 at 02:43:51PM -0600, Mark Fox wrote: > I have a table called 'jobs' with several million rows, and the only > columns that are important to this discussion are 'start_time' and > 'completion_time'. > > The sort of queries I want to execute (among others) are like: > > SELECT * FROM jobs > WHERE completion_time > SOMEDATE AND start_time < SOMEDATE; > > In plain english: All the jobs that were running at SOMEDATE. The > result of the query is on the order of 500 rows. > > I've got seperate indexes on 'start_time' and 'completion_time'. > > Now, if SOMEDATE is such that the number of rows with completion_time > > SOMEDATE is small (say 10s of thousands), the query uses index scans > and executes quickly. If not, the query uses sequential scans and is > unacceptably slow (a couple of minutes). I've used EXPLAIN and > EXPLAIN ANALYZE to confirm this. This makes perfect sense to me. > > I've played with some of the memory settings for PostgreSQL, but none > has had a significant impact. > > Any ideas on how to structure the query or add/change indexes in such > a way to improve its performance? In desperation, I tried using a > subquery, but unsurprisingly it made no (positive) difference. I feel > like there might be a way of using an index on both 'completion_time' > and 'start_time', but can't put a temporal lobe on the details. > > > Mark > > ---------------------------(end of broadcast)--------------------------- > TIP 4: Have you searched our list archives? > > http://archives.postgresql.org > -- Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com Pervasive Software http://pervasive.com 512-569-9461 From pgsql-performance-owner@postgresql.org Wed Aug 24 21:31:45 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7DF3CD7AA1 for ; Wed, 24 Aug 2005 20:42:01 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 34651-02 for ; Wed, 24 Aug 2005 23:42:00 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 651FFD78D3 for ; Wed, 24 Aug 2005 20:41:58 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7ONg0oU016555; Wed, 24 Aug 2005 19:42:00 -0400 (EDT) To: Mark Fox Cc: pgsql-performance@postgresql.org Subject: Re: Performance indexing of a simple query In-reply-to: References: Comments: In-reply-to Mark Fox message dated "Wed, 24 Aug 2005 14:43:51 -0600" Date: Wed, 24 Aug 2005 19:42:00 -0400 Message-ID: <16554.1124926920@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/412 X-Sequence-Number: 14159 Mark Fox writes: > The sort of queries I want to execute (among others) are like: > SELECT * FROM jobs > WHERE completion_time > SOMEDATE AND start_time < SOMEDATE; > In plain english: All the jobs that were running at SOMEDATE. AFAIK there is no good way to do this with btree indexes; the problem is that it's fundamentally a 2-dimensional query and btrees are 1-dimensional. There are various hacks you can try if you're willing to constrain the problem (eg, if you can assume some not-very-large maximum on the running time of jobs) but in full generality btrees are just the Wrong Thing. So what you want to look at is a non-btree index, ie, rtree or gist. For example, the contrib/seg data type could pretty directly be adapted to solve this problem, since it can index searches for overlapping line segments. The main drawback of these index types in existing releases is that they are bad on concurrent updates and don't have WAL support. Both those things are (allegedly) fixed for GIST in 8.1 ... are you interested in trying out 8.1beta? regards, tom lane From pgsql-performance-owner@postgresql.org Wed Aug 24 21:30:32 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D3DF8D7B01 for ; Wed, 24 Aug 2005 21:22:59 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 47444-03 for ; Thu, 25 Aug 2005 00:22:54 +0000 (GMT) Received: from mail.envyfinancial.com (unknown [206.248.142.186]) by svr1.postgresql.org (Postfix) with ESMTP id 5F924D7B1A for ; Wed, 24 Aug 2005 21:22:52 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by mail.envyfinancial.com (Postfix) with ESMTP id 5DDB01BD2D; Wed, 24 Aug 2005 20:13:20 -0400 (EDT) Received: from mail.envyfinancial.com ([127.0.0.1]) by localhost (mark.mielke.cc [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 18226-05; Wed, 24 Aug 2005 20:13:20 -0400 (EDT) Received: by mail.envyfinancial.com (Postfix, from userid 500) id 235881BD2E; Wed, 24 Aug 2005 20:13:20 -0400 (EDT) Date: Wed, 24 Aug 2005 20:13:20 -0400 From: mark@mark.mielke.cc To: Alan Stange Cc: Donald Courtney , pgsql-performance@postgresql.org Subject: Re: Caching by Postgres Message-ID: <20050825001320.GA18300@mark.mielke.cc> References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> <430C7448.7030103@sun.com> <20050824173055.GA11788@mark.mielke.cc> <430CC0AD.5040400@rentec.com> <20050824193441.GA13772@mark.mielke.cc> <430CE1F0.3080205@rentec.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <430CE1F0.3080205@rentec.com> User-Agent: Mutt/1.4.2.1i X-Virus-Scanned: amavisd-new at mail.envyfinancial.com X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.215 required=5 tests=[AWL=0.037, NO_REAL_NAME=0.178] X-Spam-Level: X-Archive-Number: 200508/410 X-Sequence-Number: 14157 On Wed, Aug 24, 2005 at 05:09:04PM -0400, Alan Stange wrote: > The older 32bit RISC processors do have 64 bit registers, ALUs and > datapaths, and they are marketed toward high end scientific computing, > and you're claiming that such a processor is slower than one which has > the addition of 64 bit pointers added to it? No. I'm claiming that you are talking about a hybrid 64/32 processor, and that this isn't fair to declare that 64-bit arithmetic units don't provide benefit for 64-bit math. :-) > As an example, an UltraSparc running a 32 bit kernel+application will > have the same double precision floating point performance as one > running a 64bit kernel+application (except for the additional FP > registers in the 64bit API). For a function like daxpy, it's the exact > same hardware running the exact same instructions! So why do you think > the performance would be different? Double precision floating point isn't 64-bit integer arithmetic. I think this is all a little besides the point. If you point is that the SPARC was designed well - I agree with you. I won't agree that a SPARC with 64-bit registers should be considered a 32-bit machine. The AMD 64-bit machines come in two forms as well - the ones that allow you to use the 64-bit integer registers (not floating point! those are already 80-bit!), and the ones that allow you to address more memory. I wouldn't consider either to be a 32-bit CPU, although they will allow 32-bit applications to run fine. > I believe the IBM Power processors also upped everything to double > precision internally because of some details of the "multiply-add fused" > instructions. It's been a few years since I taught H&P to CS > undergrads, but I'm fairly sure the details are all the same for MIPS > processors as well. Smart design, that obscures the difference - but doesn't make the difference a myth. If it's already there, then it's already there, and we can't talk as if it isn't. Cheers, mark -- mark@mielke.cc / markm@ncf.ca / markm@nortel.com __________________________ . . _ ._ . . .__ . . ._. .__ . . . .__ | Neighbourhood Coder |\/| |_| |_| |/ |_ |\/| | |_ | |/ |_ | | | | | | \ | \ |__ . | | .|. |__ |__ | \ |__ | Ottawa, Ontario, Canada One ring to rule them all, one ring to find them, one ring to bring them all and in the darkness bind them... http://mark.mielke.cc/ From pgsql-performance-owner@postgresql.org Wed Aug 24 21:31:18 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 13567D7A97 for ; Wed, 24 Aug 2005 21:17:01 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 46913-02 for ; Thu, 25 Aug 2005 00:16:58 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [67.100.216.10]) by svr1.postgresql.org (Postfix) with ESMTP id 347AFD76D1 for ; Wed, 24 Aug 2005 21:16:57 -0300 (ADT) Received: by flake.decibel.org (Postfix, from userid 1001) id 599071529A; Wed, 24 Aug 2005 19:17:01 -0500 (CDT) Date: Wed, 24 Aug 2005 19:17:01 -0500 From: "Jim C. Nasby" To: pgsql-performance@postgresql.org Subject: RAID arrays (and vendors) Message-ID: <20050825001701.GD56081@pervasive.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline X-Operating-System: FreeBSD 4.11-RELEASE-p10 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.009 required=5 tests=[AWL=0.009] X-Spam-Level: X-Archive-Number: 200508/411 X-Sequence-Number: 14158 I'm looking for an external RAID array (with external controller); either ~8 15kRPM SCSI drives or something with more SATA drives. This will be used in a test environment and could get moved between machines, so I'd like something with it's own RAID controller. Support for a broad range of OSes is important. Can anyone recommend hardware as well as vendors? Feel free to reply off-list. -- Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com Pervasive Software http://pervasive.com 512-569-9461 From pgsql-performance-owner@postgresql.org Wed Aug 24 22:35:49 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 23645D730C for ; Wed, 24 Aug 2005 21:30:44 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49604-02 for ; Thu, 25 Aug 2005 00:30:38 +0000 (GMT) Received: from mail.envyfinancial.com (unknown [206.248.142.186]) by svr1.postgresql.org (Postfix) with ESMTP id 3E831D7300 for ; Wed, 24 Aug 2005 21:30:37 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by mail.envyfinancial.com (Postfix) with ESMTP id 1BCCF1B6AA for ; Wed, 24 Aug 2005 20:21:06 -0400 (EDT) Received: from mail.envyfinancial.com ([127.0.0.1]) by localhost (mark.mielke.cc [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 18249-04 for ; Wed, 24 Aug 2005 20:21:05 -0400 (EDT) Received: by mail.envyfinancial.com (Postfix, from userid 500) id C58761B8F0; Wed, 24 Aug 2005 20:21:05 -0400 (EDT) Date: Wed, 24 Aug 2005 20:21:05 -0400 From: mark@mark.mielke.cc To: pgsql-performance@postgresql.org Subject: Re: Caching by Postgres Message-ID: <20050825002105.GB18300@mark.mielke.cc> References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> <430C7448.7030103@sun.com> <20050824173055.GA11788@mark.mielke.cc> <430CC0AD.5040400@rentec.com> <20050824193441.GA13772@mark.mielke.cc> <20050824210909.GN8667@mathom.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20050824210909.GN8667@mathom.us> User-Agent: Mutt/1.4.2.1i X-Virus-Scanned: amavisd-new at mail.envyfinancial.com X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.214 required=5 tests=[AWL=0.036, NO_REAL_NAME=0.178] X-Spam-Level: X-Archive-Number: 200508/413 X-Sequence-Number: 14160 On Wed, Aug 24, 2005 at 05:09:09PM -0400, Michael Stone wrote: > On Wed, Aug 24, 2005 at 03:34:41PM -0400, mark@mark.mielke.cc wrote: > >It isn't an urban myth that 64-bit math on a 64-bit processor is > >faster, at least if done using registers. It definately is faster. > >It may be an urban myth, though, that most applications perform > >a sufficient amount of 64-bit arithmetic to warrant the upgrade. > The mjor problem is that the definition of "64bit processor" is fuzzy. > The major slowdown of "64bitness" is the necessity of carting around > 64 bit pointers. It's not, however, necessary to do 64bit pointers to > get 64bit registers & fast 64 bit ops. E.g., sgi has "n32" & "n64" abi's > which can access exactly the same instruction set & registers, the > difference between them is the size of pointers and whether a "long" is > the same as a "long long". Any discussion of "64 bit processors" is > doomed from the start because people tend to start making implementation > assumptions on top of an already vague concept. Current & future > discussions are tinged by the fact that amd64 *doubles* the number > of registers in 64 bit mode, potentially providing a major speedup--but > one that doesn't really have anything to do with being "64bit". > Pretty much any discussion of 64 bit mode really needs to be a > discussion of a particular abi on a particular processor; talking about > "64 bit processors" abstractly is a waste of time. Agree. :-) As this very thread has shown! Hehe... There is no way the manufacturers would release two machines, side by side that could easily show that the 64-bit version is slower for regular application loads. They added these other things specifically to mask this... :-) Cheers, mark -- mark@mielke.cc / markm@ncf.ca / markm@nortel.com __________________________ . . _ ._ . . .__ . . ._. .__ . . . .__ | Neighbourhood Coder |\/| |_| |_| |/ |_ |\/| | |_ | |/ |_ | | | | | | \ | \ |__ . | | .|. |__ |__ | \ |__ | Ottawa, Ontario, Canada One ring to rule them all, one ring to find them, one ring to bring them all and in the darkness bind them... http://mark.mielke.cc/ From pgsql-performance-owner@postgresql.org Thu Aug 25 04:08:42 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9F8CBD7B5B for ; Thu, 25 Aug 2005 04:08:30 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 63249-02 for ; Thu, 25 Aug 2005 07:08:28 +0000 (GMT) Received: from thor.netera.se (thor.netera.se [85.112.172.11]) by svr1.postgresql.org (Postfix) with ESMTP id 33C35D7B09 for ; Thu, 25 Aug 2005 04:08:23 -0300 (ADT) Received: from [192.168.2.240] (1-1-1-41a.o.sth.bostream.se [81.26.246.14]) by thor.netera.se (Postfix) with ESMTP id 3BFD613AC04F for ; Thu, 25 Aug 2005 09:08:19 +0200 (CEST) Message-ID: <430D6EED.10603@relevanttraffic.se> Date: Thu, 25 Aug 2005 09:10:37 +0200 From: Ulrich Wisser Organization: Relevant Traffic AB User-Agent: Mozilla Thunderbird 1.0.6 (Windows/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Need for speed 2 References: <6.2.3.4.0.20050817140116.05d317b0@pop.earthlink.net> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.001 required=5 tests=[AWL=0.001] X-Spam-Level: X-Archive-Number: 200508/414 X-Sequence-Number: 14161 Hello, I realize I need to be much more specific. Here is a more detailed description of my hardware and system design. Pentium 4 2.4GHz Memory 4x DIMM DDR 1GB PC3200 400MHZ CAS3, KVR Motherboard chipset 'I865G', two IDE channels on board 2x SEAGATE BARRACUDA 7200.7 80GB 7200RPM ATA/100 (software raid 1, system, swap, pg_xlog) ADAPTEC SCSI RAID 2100S ULTRA160 32MB 1-CHANNEL 2x SEAGATE CHEETAH 15K.3 73GB ULTRA320 68-PIN WIDE (raid 1, /var/lib/pgsql) Database size on disc is 22GB. (without pg_xlog) Please find my postgresql.conf below. Putting pg_xlog on the IDE drives gave about 10% performance improvement. Would faster disks give more performance? What my application does: Every five minutes a new logfile will be imported. Depending on the source of the request it will be imported in one of three "raw click" tables. (data from two months back, to be able to verify customer complains) For reporting I have a set of tables. These contain data from the last two years. My app deletes all entries from today and reinserts updated data calculated from the raw data tables. The queries contain no joins only aggregates. I have several indexes to speed different kinds of queries. My problems occur when one users does a report that contains to much old data. In that case all cache mechanisms will fail and disc io is the limiting factor. If one query contains so much data, that a full table scan is needed, I do not care if it takes two minutes to answer. But all other queries with less data (at the same time) still have to be fast. I can not stop users doing that kind of reporting. :( I need more speed in orders of magnitude. Will more disks / more memory do that trick? Money is of course a limiting factor but it doesn't have to be real cheap. Ulrich # ----------------------------- # PostgreSQL configuration file # ----------------------------- #--------------------------------------------------------------------------- # CONNECTIONS AND AUTHENTICATION #--------------------------------------------------------------------------- # - Connection Settings - tcpip_socket = true max_connections = 100 # note: increasing max_connections costs about 500 bytes of shared # memory per connection slot, in addition to costs from shared_buffers # and max_locks_per_transaction. #superuser_reserved_connections = 2 #port = 5432 #unix_socket_directory = '' #unix_socket_group = '' #unix_socket_permissions = 0777 # octal #virtual_host = '' # what interface to listen on; defaults to any #rendezvous_name = '' # defaults to the computer name # - Security & Authentication - #authentication_timeout = 60 # 1-600, in seconds #ssl = false #password_encryption = true #krb_server_keyfile = '' #db_user_namespace = false #--------------------------------------------------------------------------- # RESOURCE USAGE (except WAL) #--------------------------------------------------------------------------- # - Memory - shared_buffers = 20000 # min 16, at least max_connections*2, 8KB each sort_mem = 4096 # min 64, size in KB vacuum_mem = 8192 # min 1024, size in KB # - Free Space Map - max_fsm_pages = 200000 # min max_fsm_relations*16, 6 bytes each max_fsm_relations = 10000 # min 100, ~50 bytes each # - Kernel Resource Usage - #max_files_per_process = 1000 # min 25 #preload_libraries = '' #--------------------------------------------------------------------------- # WRITE AHEAD LOG #--------------------------------------------------------------------------- # - Settings - fsync = false # turns forced synchronization on or off #wal_sync_method = fsync # the default varies across platforms: # fsync, fdatasync, open_sync, or open_datasync wal_buffers = 128 # min 4, 8KB each # - Checkpoints - checkpoint_segments = 16 # in logfile segments, min 1, 16MB each #checkpoint_timeout = 300 # range 30-3600, in seconds #checkpoint_warning = 30 # 0 is off, in seconds #commit_delay = 0 # range 0-100000, in microseconds #commit_siblings = 5 # range 1-1000 #--------------------------------------------------------------------------- # QUERY TUNING #--------------------------------------------------------------------------- # - Planner Method Enabling - #enable_hashagg = true #enable_hashjoin = true #enable_indexscan = true #enable_mergejoin = true #enable_nestloop = true #enable_seqscan = true #enable_sort = true #enable_tidscan = true # - Planner Cost Constants - #effective_cache_size = 1000 # typically 8KB each #random_page_cost = 4 # units are one sequential page fetch cost #cpu_tuple_cost = 0.01 # (same) #cpu_index_tuple_cost = 0.001 # (same) #cpu_operator_cost = 0.0025 # (same) # - Genetic Query Optimizer - #geqo = true #geqo_threshold = 11 #geqo_effort = 1 #geqo_generations = 0 #geqo_pool_size = 0 # default based on tables in statement, # range 128-1024 #geqo_selection_bias = 2.0 # range 1.5-2.0 # - Other Planner Options - #default_statistics_target = 10 # range 1-1000 #from_collapse_limit = 8 #join_collapse_limit = 8 # 1 disables collapsing of explicit JOINs #--------------------------------------------------------------------------- # ERROR REPORTING AND LOGGING #--------------------------------------------------------------------------- # - Syslog - syslog = 2 # range 0-2; 0=stdout; 1=both; 2=syslog syslog_facility = 'LOCAL0' syslog_ident = 'postgres' # - When to Log - client_min_messages = info # Values, in order of decreasing detail: # debug5, debug4, debug3, debug2, debug1, # log, info, notice, warning, error log_min_messages = info # Values, in order of decreasing detail: # debug5, debug4, debug3, debug2, debug1, # info, notice, warning, error, log, fatal, # panic log_error_verbosity = verbose # terse, default, or verbose messages log_min_error_statement = info # Values in order of increasing severity: # debug5, debug4, debug3, debug2, debug1, # info, notice, warning, error, panic(off) log_min_duration_statement = 1000 # Log all statements whose # execution time exceeds the value, in # milliseconds. Zero prints all queries. # Minus-one disables. silent_mode = false # DO NOT USE without Syslog! # - What to Log - #debug_print_parse = false #debug_print_rewritten = false #debug_print_plan = false #debug_pretty_print = false log_connections = true #log_duration = false #log_pid = false #log_statement = false #log_timestamp = false #log_hostname = false #log_source_port = false #--------------------------------------------------------------------------- # RUNTIME STATISTICS #--------------------------------------------------------------------------- # - Statistics Monitoring - #log_parser_stats = false #log_planner_stats = false #log_executor_stats = false #log_statement_stats = false # - Query/Index Statistics Collector - #stats_start_collector = true #stats_command_string = false #stats_block_level = false #stats_row_level = false #stats_reset_on_server_start = true #--------------------------------------------------------------------------- # CLIENT CONNECTION DEFAULTS #--------------------------------------------------------------------------- # - Statement Behavior - #search_path = '$user,public' # schema names #check_function_bodies = true #default_transaction_isolation = 'read committed' #default_transaction_read_only = false #statement_timeout = 0 # 0 is disabled, in milliseconds # - Locale and Formatting - #datestyle = 'iso, mdy' #timezone = unknown # actually, defaults to TZ environment setting #australian_timezones = false #extra_float_digits = 0 # min -15, max 2 #client_encoding = sql_ascii # actually, defaults to database encoding # These settings are initialized by initdb -- they may be changed lc_messages = 'en_US' # locale for system error message strings lc_monetary = 'en_US' # locale for monetary formatting lc_numeric = 'en_US' # locale for number formatting lc_time = 'en_US' # locale for time formatting # - Other Defaults - #explain_pretty_print = true #dynamic_library_path = '$libdir' #max_expr_depth = 10000 # min 10 #--------------------------------------------------------------------------- # LOCK MANAGEMENT #--------------------------------------------------------------------------- #deadlock_timeout = 1000 # in milliseconds #max_locks_per_transaction = 64 # min 10, ~260*max_connections bytes each #--------------------------------------------------------------------------- # VERSION/PLATFORM COMPATIBILITY #--------------------------------------------------------------------------- # - Previous Postgres Versions - #add_missing_from = true #regex_flavor = advanced # advanced, extended, or basic #sql_inheritance = true # - Other Platforms & Clients - #transform_null_equals = false From pgsql-performance-owner@postgresql.org Thu Aug 25 04:18:32 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5530BD7A57 for ; Thu, 25 Aug 2005 04:18:31 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 62383-06 for ; Thu, 25 Aug 2005 07:18:29 +0000 (GMT) Received: from smtp1.asco.de (smtp1.asco.de [217.13.70.154]) by svr1.postgresql.org (Postfix) with ESMTP id 57407D78AD for ; Thu, 25 Aug 2005 04:18:24 -0300 (ADT) Received: from [192.168.1.72] (pitr.asco.de [192.168.1.72]) (envelope-sender: ) (authenticated j_schicke CRAM-MD5 bits=0) by smtp1.asco.de (8.13.4/8.13.4/Debian-3) with ESMTP id j7P7IMvM032388 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Thu, 25 Aug 2005 09:18:22 +0200 Date: Thu, 25 Aug 2005 09:18:22 +0200 From: Jens-Wolfhard Schicke Reply-To: Jens-Wolfhard Schicke To: Chris Hoover Cc: pgsql-performance@postgresql.org Subject: Re: Some ideas for comment Message-ID: <489BA31DB80A2298E8BC1C52@[192.168.1.72]> In-Reply-To: <1d219a6f05082413265e872134@mail.gmail.com> References: <6EE64EF3AB31D5448D0007DD34EEB3417DD1AB@Herge.rcsinc.local> <1d219a6f05082413265e872134@mail.gmail.com> X-Mailer: Mulberry/3.1.6 (Linux/x86) MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1; format=flowed Content-Transfer-Encoding: quoted-printable Content-Disposition: inline X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/415 X-Sequence-Number: 14162 --On Mittwoch, August 24, 2005 16:26:40 -0400 Chris Hoover=20 wrote: > On 8/24/05, Merlin Moncure wrote: >> Linux does a pretty good job of deciding what to cache. I don't think >> this will help much. You can always look at partial indexes too. >> > Yes, but won't this help create the need to store less? If I have > 1,000.000 rows in a table, but only 4,000 are active, if I move those > 4 to another table and link the tables via a view, should that not > help keep the 9,996,000 rows out of the kernel cache (the majority of > the time at least)? The kernel caches per page, not per file. It is likely linux only caches=20 those pages which contain active rows, as long as no statement does a=20 seq-scan on that table. To optimize the thing, you could consider to cluster by some index which=20 sorts by the "activity" of the rows first. That way pages with active rows=20 are likely to contain more than only 1 active row and so the cache is=20 utilized better. Cluster is rather slow however and tables need to be reclustered from time=20 to time. Mit freundlichem Gru=DF Jens Schicke --=20 Jens Schicke j.schicke@asco.de asco GmbH http://www.asco.de Mittelweg 7 Tel 0531/3906-127 38106 Braunschweig Fax 0531/3906-400 From pgsql-performance-owner@postgresql.org Sat Aug 27 01:58:38 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AD00AD7F18 for ; Thu, 25 Aug 2005 09:28:11 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 55518-10 for ; Thu, 25 Aug 2005 12:28:10 +0000 (GMT) Received: from floppy.pyrenet.fr (floppy.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id 2A817D7F16 for ; Thu, 25 Aug 2005 09:28:08 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id E839130B42; Thu, 25 Aug 2005 14:40:41 +0200 (MET DST) From: Thomas Ganss X-Newsgroups: pgsql.performance Subject: Re: Caching by Postgres Date: Thu, 25 Aug 2005 14:23:49 +0100 Organization: T-Online Lines: 18 Message-ID: References: <20050823171045.12112.qmail@web51309.mail.yahoo.com> <20050823124323.710a9dd3.frank@wiles.org> <430B6DE3.6040109@sun.com> <430C7448.7030103@sun.com> <20050824173055.GA11788@mark.mielke.cc> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Complaints-To: usenet-abuse@t-online.de X-ID: Eqk63QZT8eXFIO1vXj4+shE4YiLPu+9v2BbtCVQlwLs4QKHH+vvogW User-Agent: Mozilla Thunderbird 1.0 (Windows/20041206) X-Accept-Language: de-DE, de, en-us, en In-Reply-To: <20050824173055.GA11788@mark.mielke.cc> To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=2.174 required=5 tests=[FORGED_YAHOO_RCVD=2.174] X-Spam-Level: ** X-Archive-Number: 200508/468 X-Sequence-Number: 14215 > The first, to always remember - is that the move from 64-bits to > 32-bits doesn't come for free. In a real 64-bit system with a > 64-bit operating system, and 64-bit applications, pointers are > now double their 32-bit size. This means more bytes to copy around > memory, and in an extreme case, has the potential to approach > halfing both the memory latency to access many such pointers from > RAM, and half the effective amount of RAM. In real world cases, > not everything is a pointer, so this sort of performance degradation > is doubtful - but it is something to keep in mind. > In addition to the above it lessens the effects of the CPU cache, so be sure to take the larger cached versions if you have structures needing to fit into the cache... my 0.02 EUR thomas From pgsql-performance-owner@postgresql.org Thu Aug 25 10:46:38 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 13A39D8034 for ; Thu, 25 Aug 2005 10:46:36 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 81557-05 for ; Thu, 25 Aug 2005 13:46:33 +0000 (GMT) Received: from frank.wiles.org (frank.wiles.org [24.124.39.75]) by svr1.postgresql.org (Postfix) with ESMTP id 7E9EDD6F4D for ; Thu, 25 Aug 2005 10:46:29 -0300 (ADT) Received: from kungfu (frank.wiles.org [127.0.0.1]) by frank.wiles.org (8.13.1/8.13.1) with SMTP id j7PDkVHA018413; Thu, 25 Aug 2005 08:46:32 -0500 Date: Thu, 25 Aug 2005 08:47:11 -0500 From: Frank Wiles To: Ulrich Wisser Cc: pgsql-performance@postgresql.org Subject: Re: Need for speed 2 Message-Id: <20050825084711.2b08f485.frank@wiles.org> In-Reply-To: <430D6EED.10603@relevanttraffic.se> References: <6.2.3.4.0.20050817140116.05d317b0@pop.earthlink.net> <430D6EED.10603@relevanttraffic.se> X-Mailer: Sylpheed version 1.0.4 (GTK+ 1.2.10; i386-redhat-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.02 required=5 tests=[AWL=0.020] X-Spam-Level: X-Archive-Number: 200508/416 X-Sequence-Number: 14163 On Thu, 25 Aug 2005 09:10:37 +0200 Ulrich Wisser wrote: > Pentium 4 2.4GHz > Memory 4x DIMM DDR 1GB PC3200 400MHZ CAS3, KVR > Motherboard chipset 'I865G', two IDE channels on board > 2x SEAGATE BARRACUDA 7200.7 80GB 7200RPM ATA/100 > (software raid 1, system, swap, pg_xlog) > ADAPTEC SCSI RAID 2100S ULTRA160 32MB 1-CHANNEL > 2x SEAGATE CHEETAH 15K.3 73GB ULTRA320 68-PIN WIDE > (raid 1, /var/lib/pgsql) > > Database size on disc is 22GB. (without pg_xlog) > > Please find my postgresql.conf below. > > Putting pg_xlog on the IDE drives gave about 10% performance > improvement. Would faster disks give more performance? Faster as in RPM on your pg_xlog partition probably won't make much of a difference. However, if you can get a drive with better overall write performance then it would be a benefit. Another thing to consider on this setup is whether or not you're hitting swap often and/or logging to that same IDE RAID set. For optimal insertion benefit you want the heads of your disks to essentially be only used for pg_xlog. If you're having to jump around the disk in the following manner: write to pg_xlog read from swap write syslog data write to pg_xlog ... ... You probably aren't getting anywhere near the benefit you could. One thing you could easily try is to break your IDE RAID set and put OS/swap on one disk and pg_xlog on the other. > If one query contains so much data, that a full table scan is needed, > I do not care if it takes two minutes to answer. But all other > queries with less data (at the same time) still have to be fast. > > I can not stop users doing that kind of reporting. :( > > I need more speed in orders of magnitude. Will more disks / more > memory do that trick? More disk and more memory always helps out. Since you say these queries are mostly on not-often-used data I would lean toward more disks in your SCSI RAID-1 setup than maxing out available RAM based on the size of your database. --------------------------------- Frank Wiles http://www.wiles.org --------------------------------- From pgsql-performance-owner@postgresql.org Thu Aug 25 16:28:11 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 889FDD6E77 for ; Thu, 25 Aug 2005 16:28:08 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 74262-02 for ; Thu, 25 Aug 2005 19:28:04 +0000 (GMT) Received: from mail.tecarta.com (66.238.115.135.ptr.us.xo.net [66.238.115.135]) by svr1.postgresql.org (Postfix) with ESMTP id F0A02D6E67 for ; Thu, 25 Aug 2005 16:28:03 -0300 (ADT) Received: from mail pickup service by mail.tecarta.com with Microsoft SMTPSVC; Thu, 25 Aug 2005 12:31:54 -0700 Received: from mail.tecarta.com ([192.168.160.2]) by mail.tecarta.com with Microsoft SMTPSVC(6.0.3790.0); Thu, 25 Aug 2005 12:31:48 -0700 Received: from barracuda.tecarta.com ([192.168.160.200]) by mail.tecarta.com (SAVSMTP 3.1.0.29) with SMTP id M2005082512314710024 for ; Thu, 25 Aug 2005 12:31:47 -0700 X-ASG-Debug-ID: 1124998078-3388-3-0 X-Barracuda-URL: http://192.168.160.200:8000/cgi-bin/mark.cgi Received: from mail1 (mail1.hq.corp [192.168.160.5]) by barracuda.tecarta.com (Spam Firewall) with SMTP id 1BAB120206D1 for ; Thu, 25 Aug 2005 12:27:58 -0700 (PDT) Received: from [10.0.0.250] ([63.193.127.22]) by mail.tecarta.com with Microsoft SMTPSVC(6.0.3790.1830); Thu, 25 Aug 2005 12:31:44 -0700 X-ASG-Orig-Subj: Re: [PERFORM] What *_mem to increase when running CLUSTER Subject: Re: What *_mem to increase when running CLUSTER From: Steve Poe To: andrew@pillette.com Cc: pgsql-performance@postgresql.org In-Reply-To: <430E1AF6.9030803@pillette.com> References: <6EE64EF3AB31D5448D0007DD34EEB3417DD1D2@Herge.rcsinc.local> <430E1AF6.9030803@pillette.com> Content-Type: text/plain Date: Thu, 25 Aug 2005 13:56:11 +0000 Message-Id: <1124978171.11102.12.camel@amd64-laptop-spoe> Mime-Version: 1.0 X-Mailer: Evolution 2.2.3 Content-Transfer-Encoding: 7bit X-OriginalArrivalTime: 25 Aug 2005 19:31:44.0159 (UTC) FILETIME=[A02AD2F0:01C5A9AB] X-Virus-Scanned: by Barracuda Spam Firewall at tecarta.com X-Barracuda-Spam-Score: 2.00 X-Barracuda-Spam-Status: No, SCORE=2.00 using global scores of TAG_LEVEL=4.0 QUARANTINE_LEVEL=1000.0 KILL_LEVEL=7.0 tests=BAYES_80 X-Barracuda-Spam-Report: Code version 3.02, rules version 3.0.3477 Rule breakdown below pts rule name description ---- ---------------------- -------------------------------------------------- 2.00 BAYES_80 BODY: Bayesian spam probability is 80 to 95% [score: 0.9434] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 required=5 tests=[FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/421 X-Sequence-Number: 14168 Andrew, On Thu, 2005-08-25 at 12:24 -0700, Andrew Lazarus wrote: > Should I temporarily increase sort_mem, vacuum_mem, neither, or both > when doing a CLUSTER on a large (100 million row) table where as many as > half of the tuples are deadwood from UPDATEs or DELETEs? I have large > batch (10 million row) inserts, updates, and deletes so I'm not sure > frequent vacuuming would help. You may need to experiment with both. What version of Postgres? What is the size of your database? How many concurrent users? If you're seeing half of the tuples are dead, I look at checking your max_fsm_pages and max_fsm_relations after a full vacuum analyze before doing too much with sort mem. Your mileage may vary. Best of luck. Steve Poe > > ---------------------------(end of broadcast)--------------------------- > TIP 5: don't forget to increase your free space map settings From pgsql-performance-owner@postgresql.org Thu Aug 25 12:16:49 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2DECED6E6D for ; Thu, 25 Aug 2005 12:16:46 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 07269-04 for ; Thu, 25 Aug 2005 15:16:43 +0000 (GMT) Received: from smtpauth09.mail.atl.earthlink.net (smtpauth09.mail.atl.earthlink.net [209.86.89.69]) by svr1.postgresql.org (Postfix) with ESMTP id 99EE5D8314 for ; Thu, 25 Aug 2005 12:16:42 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=nQG0LCn/5XfBlwGCQzJJzfwrtVZ/g6NZ3Kr/a8Eh7v9YGRHY0ZQgB1TLw/I+mRnw; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth09.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E8JT6-00057l-5P; Thu, 25 Aug 2005 11:16:40 -0400 Message-Id: <6.2.3.4.0.20050825094630.05e6b160@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Thu, 25 Aug 2005 11:16:33 -0400 To: Ulrich Wisser , pgsql-performance@postgresql.org From: Ron Subject: Re: Need for speed 2 In-Reply-To: <430D6EED.10603@relevanttraffic.se> References: <6.2.3.4.0.20050817140116.05d317b0@pop.earthlink.net> <430D6EED.10603@relevanttraffic.se> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bcbc02d0e36cde4f55982224a72df9e4b3350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.431 required=5 tests=[AWL=0.057, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/417 X-Sequence-Number: 14164 At 03:10 AM 8/25/2005, Ulrich Wisser wrote: >I realize I need to be much more specific. Here is a more detailed >description of my hardware and system design. > > >Pentium 4 2.4GHz >Memory 4x DIMM DDR 1GB PC3200 400MHZ CAS3, KVR >Motherboard chipset 'I865G', two IDE channels on board First suggestion: Get better server HW. AMD Opteron based dual processor board is the current best in terms of price/performance ratio, _particularly_ for DB applications like the one you have described. Such mainboards cost ~$400-$500. RAM will cost about $75-$150/GB. Opteron 2xx are ~$200-$700 apiece. So a 2P AMD system can be had for as little as ~$850 + the cost of the RAM you need. In the worst case where you need 24GB of RAM (~$3600), the total comes in at ~$4450. As you can see from the numbers, buying only what RAM you actually need can save you a great deal on money. Given what little you said about how much of your DB is frequently accessed, I'd suggest buying a server based around the 2P 16 DIMM slot IWill DK88 mainboard (Tyan has announced a 16 DIMM slot mainboard, but I do not think it is actually being sold yet.). Then fill it with the minimum amount of RAM that will allow the "working set" of the DB to be cached in RAM. In the worst case where DB access is essentially uniform and essentially random, you will need 24GB of RAM to hold the 22GB DB + OS + etc. That worst case is _rare_. Usually DB's have a working set that is smaller than the entire DB. You want to keep that working set in RAM. If you can't identify the working set, buy enough RAM to hold the entire DB. In particular, you want to make sure that any frequently accessed read only tables or indexes are kept in RAM. The "read only" part is very important. Tables (and their indexes) that are frequently written to _have_ to access HD. Therefore you get much less out of having them in RAM. Read only tables and their indexes can be loaded into tmpfs at boot time thereby keeping out of the way of the file system buffer cache. tmpfs does not save data if the host goes down so it is very important that you ONLY use this trick with read only tables. The other half of the trick is to make sure that the file system buffer cache does _not_ cache whatever you have loaded into tmpfs. >2x SEAGATE BARRACUDA 7200.7 80GB 7200RPM ATA/100 >(software raid 1, system, swap, pg_xlog) >ADAPTEC SCSI RAID 2100S ULTRA160 32MB 1-CHANNEL >2x SEAGATE CHEETAH 15K.3 73GB ULTRA320 68-PIN WIDE >(raid 1, /var/lib/pgsql) Second suggestion: you need a MUCH better IO subsystem. In fact, given that you have described this system as being primarily OLTP like, this is more important that the above server HW. Best would be to upgrade everything, but if you are strapped for cash, upgrade the IO subsystem first. You need many more spindles and a decent RAID card or cards. You want 15Krpm (best) or 10Krpm HDs. As long as all of the HD's are at least 10Krpm, more spindles is more important than faster spindles. If it's a choice between more 10Krpm discs or fewer 15Krpm discs, buy the 10Krpm discs. Get the spindle count as high as you RAID cards can handle. Whatever RAID cards you get should have as much battery backed write buffer as possible. In the commodity market, presently the highest performance RAID cards I know of, and the ones that support the largest battery backed write buffer, are made by Areca. >Database size on disc is 22GB. (without pg_xlog) Find out what the working set, ie the most frequently accessed portion, of this 22GB is and you will know how much RAM is worth having. 4GB is definitely too little! >Please find my postgresql.conf below. Third suggestion: make sure you are running a 2.6 based kernel and at least PG 8.0.3. Helping beta test PG 8.1 might be an option for you as well. >Putting pg_xlog on the IDE drives gave about 10% performance >improvement. Would faster disks give more performance? > >What my application does: > >Every five minutes a new logfile will be imported. Depending on the >source of the request it will be imported in one of three "raw click" >tables. (data from two months back, to be able to verify customer >complains) For reporting I have a set of tables. These contain data >from the last two years. My app deletes all entries from today and >reinserts updated data calculated from the raw data tables. The raw data tables seem to be read only? If so, you should buy enough RAM to load them into tmpfs at boot time and have them be completely RAM resident in addition to having enough RAM for the OS to cache an appropriate amount of the rest of the DB. >The queries contain no joins only aggregates. I have several indexes >to speed different kinds of queries. > >My problems occur when one users does a report that contains too >much old data. In that case all cache mechanisms will fail and disc >io is the limiting factor. > >If one query contains so much data, that a full table scan is >needed, I do not care if it takes two minutes to answer. But all >other queries with less data (at the same time) still have to be fast. HDs can only do one thing at once. If they are in the middle of a full table scan, everything else that requires HD access is going to wait until it is done. At some point, looking at your DB schema and queries will be worth it for optimization purposes. Right now, you HW is so underpowered compared to the demands you are placing on it that there's little point to SW tuning. >I can not stop users doing that kind of reporting. :( > >I need more speed in orders of magnitude. Will more disks / more >memory do that trick? If you do the right things with them ;) >Money is of course a limiting factor but it doesn't have to be real cheap. > >Ulrich > > > > > ># ----------------------------- ># PostgreSQL configuration file ># ----------------------------- >#--------------------------------------------------------------------------- ># CONNECTIONS AND AUTHENTICATION >#--------------------------------------------------------------------------- > ># - Connection Settings - > >tcpip_socket = true >max_connections = 100 > # note: increasing max_connections costs about 500 bytes of shared > # memory per connection slot, in addition to costs from > shared_buffers > # and max_locks_per_transaction. >#superuser_reserved_connections = 2 >#port = 5432 >#unix_socket_directory = '' >#unix_socket_group = '' >#unix_socket_permissions = 0777 # octal >#virtual_host = '' # what interface to listen on; defaults to any >#rendezvous_name = '' # defaults to the computer name > ># - Security & Authentication - > >#authentication_timeout = 60 # 1-600, in seconds >#ssl = false >#password_encryption = true >#krb_server_keyfile = '' >#db_user_namespace = false > > >#--------------------------------------------------------------------------- ># RESOURCE USAGE (except WAL) >#--------------------------------------------------------------------------- > ># - Memory - > >shared_buffers = 20000 # min 16, at least max_connections*2, 8KB each >sort_mem = 4096 # min 64, size in KB 4MB seems small. Find out how much memory you usually need for a sort, and how many sorts you are usually doing at once to set this to a sane size. >vacuum_mem = 8192 # min 1024, size in KB > ># - Free Space Map - > >max_fsm_pages = 200000 # min max_fsm_relations*16, 6 bytes each >max_fsm_relations = 10000 # min 100, ~50 bytes each > ># - Kernel Resource Usage - > >#max_files_per_process = 1000 # min 25 >#preload_libraries = '' > > >#--------------------------------------------------------------------------- ># WRITE AHEAD LOG >#--------------------------------------------------------------------------- > ># - Settings - > >fsync = false # turns forced synchronization on or off >#wal_sync_method = fsync # the default varies across platforms: > # fsync, fdatasync, open_sync, or I hope you have a battery backed write buffer! >open_datasync >wal_buffers = 128 # min 4, 8KB each There might be a better value for you to use. I'll hold off on looking at the rest of this... ># - Checkpoints - > >checkpoint_segments = 16 # in logfile segments, min 1, 16MB each >#checkpoint_timeout = 300 # range 30-3600, in seconds >#checkpoint_warning = 30 # 0 is off, in seconds >#commit_delay = 0 # range 0-100000, in microseconds >#commit_siblings = 5 # range 1-1000 > > >#--------------------------------------------------------------------------- ># QUERY TUNING >#--------------------------------------------------------------------------- > ># - Planner Method Enabling - > >#enable_hashagg = true >#enable_hashjoin = true >#enable_indexscan = true >#enable_mergejoin = true >#enable_nestloop = true >#enable_seqscan = true >#enable_sort = true >#enable_tidscan = true > ># - Planner Cost Constants - > >#effective_cache_size = 1000 # typically 8KB each >#random_page_cost = 4 # units are one sequential page fetch cost >#cpu_tuple_cost = 0.01 # (same) >#cpu_index_tuple_cost = 0.001 # (same) >#cpu_operator_cost = 0.0025 # (same) > ># - Genetic Query Optimizer - > >#geqo = true >#geqo_threshold = 11 >#geqo_effort = 1 >#geqo_generations = 0 >#geqo_pool_size = 0 # default based on tables in statement, > # range 128-1024 >#geqo_selection_bias = 2.0 # range 1.5-2.0 > ># - Other Planner Options - > >#default_statistics_target = 10 # range 1-1000 >#from_collapse_limit = 8 >#join_collapse_limit = 8 # 1 disables collapsing of explicit JOINs > > >#--------------------------------------------------------------------------- ># ERROR REPORTING AND LOGGING >#--------------------------------------------------------------------------- > ># - Syslog - > >syslog = 2 # range 0-2; 0=stdout; 1=both; 2=syslog >syslog_facility = 'LOCAL0' >syslog_ident = 'postgres' > ># - When to Log - > >client_min_messages = info # Values, in order of decreasing detail: > # debug5, debug4, debug3, debug2, debug1, > # log, info, notice, warning, error > >log_min_messages = info # Values, in order of decreasing detail: > # debug5, debug4, debug3, debug2, debug1, > # info, notice, warning, error, log, >fatal, > # panic > >log_error_verbosity = verbose # terse, default, or verbose messages > >log_min_error_statement = info # Values in order of increasing severity: > # debug5, debug4, debug3, debug2, >debug1, > # info, notice, warning, error, >panic(off) > >log_min_duration_statement = 1000 # Log all statements whose > # execution time exceeds the value, in > # milliseconds. Zero prints all queries. > # Minus-one disables. > >silent_mode = false # DO NOT USE without Syslog! > ># - What to Log - > >#debug_print_parse = false >#debug_print_rewritten = false >#debug_print_plan = false >#debug_pretty_print = false >log_connections = true >#log_duration = false >#log_pid = false >#log_statement = false >#log_timestamp = false >#log_hostname = false >#log_source_port = false > > >#--------------------------------------------------------------------------- ># RUNTIME STATISTICS >#--------------------------------------------------------------------------- > ># - Statistics Monitoring - > >#log_parser_stats = false >#log_planner_stats = false >#log_executor_stats = false >#log_statement_stats = false > ># - Query/Index Statistics Collector - > >#stats_start_collector = true >#stats_command_string = false >#stats_block_level = false >#stats_row_level = false >#stats_reset_on_server_start = true > > >#--------------------------------------------------------------------------- ># CLIENT CONNECTION DEFAULTS >#--------------------------------------------------------------------------- > ># - Statement Behavior - > >#search_path = '$user,public' # schema names >#check_function_bodies = true >#default_transaction_isolation = 'read committed' >#default_transaction_read_only = false >#statement_timeout = 0 # 0 is disabled, in milliseconds > ># - Locale and Formatting - > >#datestyle = 'iso, mdy' >#timezone = unknown # actually, defaults to TZ environment >setting >#australian_timezones = false >#extra_float_digits = 0 # min -15, max 2 >#client_encoding = sql_ascii # actually, defaults to database encoding > ># These settings are initialized by initdb -- they may be changed >lc_messages = 'en_US' # locale for system error message strings >lc_monetary = 'en_US' # locale for monetary formatting >lc_numeric = 'en_US' # locale for number formatting >lc_time = 'en_US' # locale for time formatting > ># - Other Defaults - > >#explain_pretty_print = true >#dynamic_library_path = '$libdir' >#max_expr_depth = 10000 # min 10 > > >#--------------------------------------------------------------------------- ># LOCK MANAGEMENT >#--------------------------------------------------------------------------- > >#deadlock_timeout = 1000 # in milliseconds >#max_locks_per_transaction = 64 # min 10, ~260*max_connections bytes each > > >#--------------------------------------------------------------------------- ># VERSION/PLATFORM COMPATIBILITY >#--------------------------------------------------------------------------- > ># - Previous Postgres Versions - > >#add_missing_from = true >#regex_flavor = advanced # advanced, extended, or basic >#sql_inheritance = true > ># - Other Platforms & Clients - > >#transform_null_equals = false > > > >---------------------------(end of broadcast)--------------------------- >TIP 2: Don't 'kill -9' the postmaster From pgsql-performance-owner@postgresql.org Thu Aug 25 14:30:56 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8AA64D6E67 for ; Thu, 25 Aug 2005 14:30:52 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 46485-02 for ; Thu, 25 Aug 2005 17:30:49 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id 8913AD6DF8 for ; Thu, 25 Aug 2005 14:30:48 -0300 (ADT) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: Need for speed 2 Date: Thu, 25 Aug 2005 13:30:48 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3417DD1D2@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Need for speed 2 Thread-Index: AcWpRE73bCrrIw0sQFWp+cqC7CzJygAVHOOw From: "Merlin Moncure" To: "Ulrich Wisser" Cc: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/418 X-Sequence-Number: 14165 > Putting pg_xlog on the IDE drives gave about 10% performance > improvement. Would faster disks give more performance? >=20 > What my application does: >=20 > Every five minutes a new logfile will be imported. Depending on the > source of the request it will be imported in one of three "raw click" > tables. (data from two months back, to be able to verify customer > complains) > For reporting I have a set of tables. These contain data from the last > two years. My app deletes all entries from today and reinserts updated > data calculated from the raw data tables. >=20 > The queries contain no joins only aggregates. I have several indexes to > speed different kinds of queries. >=20 > My problems occur when one users does a report that contains to much old > data. In that case all cache mechanisms will fail and disc io is the > limiting factor. It seems like you are pushing limit of what server can handle. This means: 1. expensive server upgrade. or=20 2. make software more efficient. Since you sound I/O bound, you can tackle 1. by a. adding more memory or b. increasing i/o throughput. =20 Unfortunately, you already have a pretty decent server (for x86) so 1. means 64 bit platform and 2. means more expensive hard drives. The archives is full of information about this... Is your data well normalized? You can do tricks like: if table has fields a,b,c,d,e,f with a is primary key, and d,e,f not frequently queried or missing, move d,e,f to seprate table. well normalized structures are always more cache efficient. Do you have lots of repeating and/or empty data values in your tables? Make your indexes and data as small as possible to reduce pressure on the cache, here are just a few tricks: 1. use int2/int4 instead of numeric 2. know when to use char and varchar=20 3. use functional indexes to reduce index expression complexity. This can give extreme benefits if you can, for example, reduce double field index to Boolean. Merlin From pgsql-performance-owner@postgresql.org Thu Aug 25 16:00:43 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5B15CD768F for ; Thu, 25 Aug 2005 16:00:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 67926-03 for ; Thu, 25 Aug 2005 19:00:23 +0000 (GMT) Received: from tbmail.tradebot.com (Tradebot-Systems-1096753.cust-rtr.swbell.net [68.90.170.2]) by svr1.postgresql.org (Postfix) with ESMTP id 48D44D768E for ; Thu, 25 Aug 2005 16:00:22 -0300 (ADT) Received: from 192.168.200.29 ([192.168.200.29]) by tbmail.tradebot.com ([192.168.1.50]) with Microsoft Exchange Server HTTP-DAV ; Thu, 25 Aug 2005 19:00:22 +0000 Received: from krb06 by TBMAIL; 25 Aug 2005 14:00:22 -0500 Subject: Re: Need for speed 2 From: Kelly Burkhart To: Ron Cc: pgsql-performance@postgresql.org In-Reply-To: <6.2.3.4.0.20050825094630.05e6b160@pop.earthlink.net> References: <6.2.3.4.0.20050817140116.05d317b0@pop.earthlink.net> <430D6EED.10603@relevanttraffic.se> <6.2.3.4.0.20050825094630.05e6b160@pop.earthlink.net> Content-Type: text/plain Content-Transfer-Encoding: 7bit Date: Thu, 25 Aug 2005 14:00:22 -0500 Message-Id: <1124996422.6969.47.camel@krb06.tradebot.com> Mime-Version: 1.0 X-Mailer: Evolution 2.2.1 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 required=5 tests=[FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/419 X-Sequence-Number: 14166 On Thu, 2005-08-25 at 11:16 -0400, Ron wrote: > ># - Settings - > > > >fsync = false # turns forced synchronization on or off > >#wal_sync_method = fsync # the default varies across platforms: > > # fsync, fdatasync, open_sync, or > > I hope you have a battery backed write buffer! Battery backed write buffer will do nothing here, because the OS is taking it's sweet time flushing to the controller's battery backed write buffer! Isn't the reason for batter backed controller cache to make fsync()s fast? -K From pgsql-performance-owner@postgresql.org Thu Aug 25 16:13:22 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B310CD710A for ; Thu, 25 Aug 2005 16:13:19 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 71816-03 for ; Thu, 25 Aug 2005 19:13:17 +0000 (GMT) Received: from pillette.com (adsl-67-119-5-202.dsl.snfc21.pacbell.net [67.119.5.202]) by svr1.postgresql.org (Postfix) with ESMTP id 8C304D7142 for ; Thu, 25 Aug 2005 16:13:15 -0300 (ADT) Received: from [192.168.1.213] (dhcp213.pillette.com [192.168.1.213] (may be forged)) by pillette.com (8.11.6/8.11.6) with ESMTP id j7PJDFu21089 for ; Thu, 25 Aug 2005 12:13:15 -0700 Message-ID: <430E1AF6.9030803@pillette.com> Date: Thu, 25 Aug 2005 12:24:38 -0700 From: Andrew Lazarus Reply-To: andrew@pillette.com Organization: Pillette Investment Management User-Agent: Mozilla Thunderbird 1.0.6 (Windows/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: What *_mem to increase when running CLUSTER References: <6EE64EF3AB31D5448D0007DD34EEB3417DD1D2@Herge.rcsinc.local> In-Reply-To: <6EE64EF3AB31D5448D0007DD34EEB3417DD1D2@Herge.rcsinc.local> Content-Type: multipart/mixed; boundary="------------050003000907070802050103" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.112 required=5 tests=[AWL=0.062, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/420 X-Sequence-Number: 14167 This is a multi-part message in MIME format. --------------050003000907070802050103 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Should I temporarily increase sort_mem, vacuum_mem, neither, or both when doing a CLUSTER on a large (100 million row) table where as many as half of the tuples are deadwood from UPDATEs or DELETEs? I have large batch (10 million row) inserts, updates, and deletes so I'm not sure frequent vacuuming would help. --------------050003000907070802050103 Content-Type: text/x-vcard; charset=utf-8; name="andrew.vcf" Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="andrew.vcf" begin:vcard fn:Andrew Lazarus n:Lazarus;Andrew org:Pillette Investment Management;Research and Development adr;dom:;;3028 Fillmore;San Francisco;CA;94123 email;internet:andrew@pillette.com title:Director tel;work:800-366-0688 tel;fax:415-440-4093 url:http://www.pillette.com version:2.1 end:vcard --------------050003000907070802050103-- From pgsql-performance-owner@postgresql.org Thu Aug 25 16:42:01 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 939B9D7244 for ; Thu, 25 Aug 2005 16:41:48 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 77678-02 for ; Thu, 25 Aug 2005 19:41:44 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id B3F06D71EA for ; Thu, 25 Aug 2005 16:41:43 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [64.81.245.111] (account josh@agliodbs.com HELO temoku.sf.agliodbs.com) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7939593; Thu, 25 Aug 2005 12:44:00 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Read/Write block sizes Date: Thu, 25 Aug 2005 12:45:03 -0700 User-Agent: KMail/1.8 Cc: "Jeffrey W. Baker" , Tom Lane , Guy Thornley , Steve Poe , Chris Browne References: <21700.1124863004@sss.pgh.pa.us> <1124864572.11270.16.camel@noodles> In-Reply-To: <1124864572.11270.16.camel@noodles> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Message-Id: <200508251245.03473.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.048 required=5 tests=[AWL=-0.002, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/422 X-Sequence-Number: 14169 Jeff, > Ask me sometime about my replacement for GNU sort. =C2=A0It uses the same > sorting algorithm, but it's an order of magnitude faster due to better > I/O strategy. =C2=A0Someday, in my infinite spare time, I hope to demonst= rate > that kind of improvement with a patch to pg. Since we desperately need some improvements in sort performance, I do hope= =20 you follow up on this. =2D-=20 =2D-Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Thu Aug 25 17:19:55 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8546FD76C2 for ; Thu, 25 Aug 2005 17:19:52 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83907-08 for ; Thu, 25 Aug 2005 20:19:51 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 68704D7682 for ; Thu, 25 Aug 2005 17:19:49 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7PKJovh014787; Thu, 25 Aug 2005 16:19:50 -0400 (EDT) To: andrew@pillette.com Cc: pgsql-performance@postgresql.org Subject: Re: What *_mem to increase when running CLUSTER In-reply-to: <430E1AF6.9030803@pillette.com> References: <6EE64EF3AB31D5448D0007DD34EEB3417DD1D2@Herge.rcsinc.local> <430E1AF6.9030803@pillette.com> Comments: In-reply-to Andrew Lazarus message dated "Thu, 25 Aug 2005 12:24:38 -0700" Date: Thu, 25 Aug 2005 16:19:50 -0400 Message-ID: <14786.1125001190@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/423 X-Sequence-Number: 14170 Andrew Lazarus writes: > Should I temporarily increase sort_mem, vacuum_mem, neither, or both > when doing a CLUSTER on a large (100 million row) table The only part of that job that can use lots of memory is the index rebuilds. In recent PG versions maintenance_work_mem is the thing to increase for an index build; previously sort_mem controlled it. I forget when the changeover was; maybe 8.0. regards, tom lane From pgsql-performance-owner@postgresql.org Thu Aug 25 17:26:47 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id DE9F9D76D6 for ; Thu, 25 Aug 2005 17:26:42 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 86419-09 for ; Thu, 25 Aug 2005 20:26:38 +0000 (GMT) Received: from smtpauth01.mail.atl.earthlink.net (smtpauth01.mail.atl.earthlink.net [209.86.89.61]) by svr1.postgresql.org (Postfix) with ESMTP id D68C0D76C2 for ; Thu, 25 Aug 2005 17:26:37 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=NqoHMEt4GyPZWq99tKj9cQgjBtyrABeNbkE4ue8UXNpLJgAWAOQV0437FNyDgsTW; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:In-Reply-To:References:Mime-Version:Content-Type:Content-Transfer-Encoding:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth01.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E8OJ4-0003ZP-LX; Thu, 25 Aug 2005 16:26:38 -0400 Message-Id: <6.2.3.4.0.20050825162308.05dbf298@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Thu, 25 Aug 2005 16:26:34 -0400 To: josh@agliodbs.com, pgsql-performance@postgresql.org From: Ron Subject: Re: Read/Write block sizes In-Reply-To: <200508251245.03473.josh@agliodbs.com> References: <21700.1124863004@sss.pgh.pa.us> <1124864572.11270.16.camel@noodles> <200508251245.03473.josh@agliodbs.com> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1"; format=flowed Content-Transfer-Encoding: quoted-printable X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bcaa3d4fb93db5250a4c66d5258db04eb7350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.442 required=5 tests=[AWL=0.068, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/424 X-Sequence-Number: 14171 At 03:45 PM 8/25/2005, Josh Berkus wrote: >Jeff, > > > Ask me sometime about my replacement for GNU sort. =C2 It uses the same > > sorting algorithm, but it's an order of magnitude faster due to better > > I/O strategy. =C2 Someday, in my infinite spare time, I hope to= demonstrate > > that kind of improvement with a patch to pg. > >Since we desperately need some improvements in sort performance, I do hope >you follow up on this. > >-- >--Josh I'll generalize that. IMO we desperately need=20 any and all improvements in IO performance. Even=20 more so than we need improvements in sorting or sorting IO performance. Ron From pgsql-performance-owner@postgresql.org Thu Aug 25 17:49:50 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BAA0AD7712 for ; Thu, 25 Aug 2005 17:49:39 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 93231-01 for ; Thu, 25 Aug 2005 20:49:37 +0000 (GMT) Received: from floppy.pyrenet.fr (floppy.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id 819F9D76C8 for ; Thu, 25 Aug 2005 17:49:35 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id 2124E31F50; Thu, 25 Aug 2005 23:02:09 +0200 (MET DST) From: Chris Browne X-Newsgroups: pgsql.performance Subject: Re: Read/Write block sizes Date: Thu, 25 Aug 2005 16:49:26 -0400 Organization: cbbrowne Computing Inc Lines: 41 Message-ID: <604q9doi7d.fsf@dba2.int.libertyrms.com> References: <21700.1124863004@sss.pgh.pa.us> <1124864572.11270.16.camel@noodles> <200508251245.03473.josh@agliodbs.com> <6.2.3.4.0.20050825162308.05dbf298@pop.earthlink.net> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Complaints-To: usenet@news.hub.org User-Agent: Gnus/5.1007 (Gnus v5.10.7) XEmacs/21.4.17 (Jumbo Shrimp, linux) Cancel-Lock: sha1:ySqOwaJz/b3BL3ibyrYi6KhtleU= To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.109 required=5 tests=[AWL=0.109] X-Spam-Level: X-Archive-Number: 200508/425 X-Sequence-Number: 14172 rjpeace@earthlink.net (Ron) writes: > At 03:45 PM 8/25/2005, Josh Berkus wrote: >> > Ask me sometime about my replacement for GNU sort. � It uses the >> > same sorting algorithm, but it's an order of magnitude faster due >> > to better I/O strategy. � Someday, in my infinite spare time, I >> > hope to demonstrate that kind of improvement with a patch to pg. >> >>Since we desperately need some improvements in sort performance, I >>do hope you follow up on this. > > I'll generalize that. IMO we desperately need any and all > improvements in IO performance. Even more so than we need > improvements in sorting or sorting IO performance. That's frankly a step backwards. Feel free to "specialise" that instead. A patch that improves some specific aspect of performance is a thousand times better than any sort of "desperate desire for any and all improvements in I/O performance." The latter is unlikely to provide any usable result. The "specialized patch" is also pointedly better in that a *confidently submitted* patch is likely to be way better than any sort of "desperate clutching at whatever may come to hand." Far too often, I see people trying to address performance problems via the "desperate clutching at whatever seems near to hand," and that generally turns out very badly as a particular result of the whole "desperate clutching" part. If you can get a sort improvement submitted, that's a concrete improvement... -- select 'cbbrowne' || '@' || 'ntlug.org'; http://www3.sympatico.ca/cbbrowne/lisp.html Appendium to the Rules of the Evil Overlord #1: "I will not build excessively integrated security-and-HVAC systems. They may be Really Cool, but are far too vulnerable to breakdowns." From pgsql-performance-owner@postgresql.org Thu Aug 25 20:47:37 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0BCCED78D0 for ; Thu, 25 Aug 2005 20:46:57 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 37751-09 for ; Thu, 25 Aug 2005 23:46:54 +0000 (GMT) Received: from smtpauth05.mail.atl.earthlink.net (smtpauth05.mail.atl.earthlink.net [209.86.89.65]) by svr1.postgresql.org (Postfix) with ESMTP id AB486D789C for ; Thu, 25 Aug 2005 20:46:52 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=tGouZISx5acGbjd110zxz8F1ASDWQ2nto6247G5c+DVO9G9Q8xetSVyLymjV1k4x; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:In-Reply-To:References:Mime-Version:Content-Type:Content-Transfer-Encoding:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth05.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E8RQu-0002ky-6U; Thu, 25 Aug 2005 19:46:56 -0400 Message-Id: <6.2.3.4.0.20050825185441.05e63ef0@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Thu, 25 Aug 2005 19:46:51 -0400 To: Chris Browne , pgsql-performance@postgresql.org From: Ron Subject: Re: Read/Write block sizes In-Reply-To: <604q9doi7d.fsf@dba2.int.libertyrms.com> References: <21700.1124863004@sss.pgh.pa.us> <1124864572.11270.16.camel@noodles> <200508251245.03473.josh@agliodbs.com> <6.2.3.4.0.20050825162308.05dbf298@pop.earthlink.net> <604q9doi7d.fsf@dba2.int.libertyrms.com> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1"; format=flowed Content-Transfer-Encoding: quoted-printable X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bc2469d0c2931e1f8d42fb3ee955c9969c350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.433 required=5 tests=[AWL=0.059, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/426 X-Sequence-Number: 14173 At 04:49 PM 8/25/2005, Chris Browne wrote: >rjpeace@earthlink.net (Ron) writes: > > At 03:45 PM 8/25/2005, Josh Berkus wrote: > >> > Ask me sometime about my replacement for GNU sort. =C2 It uses the > >> > same sorting algorithm, but it's an order of magnitude faster due > >> > to better I/O strategy. =C2 Someday, in my infinite spare time, I > >> > hope to demonstrate that kind of improvement with a patch to pg. > >> > >>Since we desperately need some improvements in sort performance, I > >>do hope you follow up on this. > > > > I'll generalize that. IMO we desperately need any and all > > improvements in IO performance. Even more so than we need > > improvements in sorting or sorting IO performance. > >That's frankly a step backwards. Feel free to "specialise" that instead. We can agree to disagree, I'm cool with that. I'm well aware that a Systems Approach to SW=20 Architecture is not always popular in the Open=20 Source world. Nonetheless, my POV is that if we=20 want to be taken seriously and beat "the big=20 boys", we have to do everything smarter and=20 faster, as well as cheaper, than they do. You=20 are not likely to be able to do that consistently=20 without using some of the "icky" stuff one is=20 required to study as part of formal training in=20 the Comp Sci and SW Engineering fields. >A patch that improves some specific aspect of=20 >performance is a thousand times better than any=20 >sort of "desperate desire for any and >all improvements in I/O performance." minor twisting of my words: substituting "desire"=20 for "need". The need is provable. Just put "the=20 big 5" (SQL Server, Oracle, DB2, mySQL, and=20 PostgreSQL) into some realistic benches to see that. Major twisting of my words: the apparent=20 implication by you that I don't appreciate=20 improvements in the IO behavior of specific=20 things like sorting as much as I'd appreciate=20 more "general" IO performance=20 improvements. Performance optimization is best=20 done as an iterative improvement process that=20 starts with measuring where the need is greatest,=20 then improving that greatest need by the most you=20 can, then repeating the whole cycle. _Every_=20 improvement in such a process is a specific=20 improvement, even if the improvement is a=20 decision to re-architect the entire product to=20 solve the current biggest issue. Improving=20 sorting IO is cool. OTOH, if pg's biggest IO=20 problems are elsewhere, then the amount of=20 overall benefit we will get from improving=20 sorting IO is going to be minimized until we=20 improve the bigger problem(s). Amdahl's Law. >The "specialized patch" is also pointedly better=20 >in that a *confidently submitted* patch is=20 >likely to be way better than any sort of=20 >"desperate clutching at whatever may come to hand." Another distortion of my statement and POV. I=20 never suggested nor implied any sort of=20 "desperate clutching...". We have _measurable_=20 IO issues that need to be addressed in order for=20 pg to be a better competitor in the=20 marketplace. Just as we do with sorting performance. >Far too often, I see people trying to address=20 >performance problems via the "desperate=20 >clutching at whatever seems near to hand," and that >generally turns out very badly as a particular=20 >result of the whole "desperate clutching" part. > >If you can get a sort improvement submitted, that's a concrete= improvement... As I said, I'm all in favor of concrete,=20 measurable improvement. I do not think I ever=20 stated I was in favor of anything else. You evidently are mildly ranting because you've=20 seen some examples of poor SW Engineering=20 Discipline/Practice by people with perhaps=20 inadequate skills for the issues they were trying=20 to address. We all have. "90% of everything is=20 Jreck (eg of too low a quality)." OTOH, I do not think I've given you any reason to=20 think I lack such Clue, nor do I think my post was advocating such= thrashing. My post was intended to say that we need an=20 Overall Systems Approach to pg optimization=20 rather than just applying what compiler writer's=20 call "peephole optimizations" to pg. No more, no less. I apologize if I somehow misled you, Ron Peacetree From pgsql-performance-owner@postgresql.org Thu Aug 25 21:27:31 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 00651D70D1 for ; Thu, 25 Aug 2005 21:27:29 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 52378-03 for ; Fri, 26 Aug 2005 00:27:26 +0000 (GMT) Received: from mail.nordicbet.com (mail.nordicbet.com [193.69.167.130]) by svr1.postgresql.org (Postfix) with ESMTP id 3E647D703F for ; Thu, 25 Aug 2005 21:27:24 -0300 (ADT) Received: from [80.203.125.83] (helo=tobias.nordicbet.invalid) (Authenticated Sender=tobias@nordicbet.com) by mail.nordicbet.com with esmtpa (Exim 4.50 #1 (Debian)) id 1E8S1Y-00017T-Mf for ; Fri, 26 Aug 2005 02:24:59 +0200 Received: by tobias.nordicbet.invalid (Postfix, from userid 500) id 938E6DECFD; Fri, 26 Aug 2005 02:27:09 +0200 (CEST) Date: Fri, 26 Aug 2005 02:27:09 +0200 From: Tobias Brox To: pgsql-performance@postgresql.org Subject: Limit + group + join Message-ID: <20050826002709.GK10328@tobias.lan> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Organization: Group Nordicbet User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/427 X-Sequence-Number: 14174 Consider this setup - which is a gross simplification of parts of our production system ;-) create table c (id integer primary key); create table b (id integer primary key, c_id integer); create index b_on_c on b(c_id) insert into c (select ... lots of IDs ...); insert into b (select id, id from c); /* keep it simple :-) */ Now, I'm just interessted in some few rows. All those gives good plans: explain select c.id from c order by c.id limit 1; explain select c.id from c group by c.id order by c.id limit 1; explain select c.id from c join b on c_id=c.id order by c.id limit 1; ... BUT ... combining join, group and limit makes havoc: explain select c.id from c join b on c_id=c.id group by c.id order by c.id desc limit 5; QUERY PLAN ------------------------------------------------------------------------------------- Limit (cost=3809.65..3809.67 rows=5 width=4) -> Group (cost=3809.65..3940.59 rows=26187 width=4) -> Sort (cost=3809.65..3875.12 rows=26188 width=4) Sort Key: c.id -> Hash Join (cost=559.34..1887.89 rows=26188 width=4) Hash Cond: ("outer".id = "inner".c_id) -> Seq Scan on c (cost=0.00..403.87 rows=26187 width=4) -> Hash (cost=403.87..403.87 rows=26187 width=4) -> Seq Scan on b (cost=0.00..403.87 rows=26187 width=4) (9 rows) I get the same behaviour on pg 7.4.7 and pg 8.0.2. Of course, I can probably use subqueries instead of join - though, I would have wished the planner could do better ;-) -- Notice of Confidentiality: This information may be confidential, and blah-blah-blah - so please keep your eyes closed. Please delete and destroy this email. Failure to comply will cause my lawyer to yawn. From pgsql-performance-owner@postgresql.org Thu Aug 25 22:56:31 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A5519D7908 for ; Thu, 25 Aug 2005 22:55:44 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 81172-04 for ; Fri, 26 Aug 2005 01:55:43 +0000 (GMT) Received: from mail27.sea5.speakeasy.net (mail27.sea5.speakeasy.net [69.17.117.29]) by svr1.postgresql.org (Postfix) with ESMTP id EF9C9D7911 for ; Thu, 25 Aug 2005 22:55:42 -0300 (ADT) Received: (qmail 25321 invoked from network); 26 Aug 2005 01:55:46 -0000 Received: from dsl081-060-184.sfo1.dsl.speakeasy.net (HELO noodles) ([64.81.60.184]) (envelope-sender ) by mail27.sea5.speakeasy.net (qmail-ldap-1.03) with RC4-MD5 encrypted SMTP for ; 26 Aug 2005 01:55:46 -0000 Subject: Re: Limit + group + join From: "Jeffrey W. Baker" To: Tobias Brox Cc: pgsql-performance@postgresql.org In-Reply-To: <20050826002709.GK10328@tobias.lan> References: <20050826002709.GK10328@tobias.lan> Content-Type: text/plain Date: Thu, 25 Aug 2005 18:56:59 -0700 Message-Id: <1125021419.16451.0.camel@noodles> Mime-Version: 1.0 X-Mailer: Evolution 2.2.1.1 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.043 required=5 tests=[AWL=0.043] X-Spam-Level: X-Archive-Number: 200508/428 X-Sequence-Number: 14175 On Fri, 2005-08-26 at 02:27 +0200, Tobias Brox wrote: > Consider this setup - which is a gross simplification of parts of our > production system ;-) > > create table c (id integer primary key); > create table b (id integer primary key, c_id integer); > create index b_on_c on b(c_id) > > insert into c (select ... lots of IDs ...); > insert into b (select id, id from c); /* keep it simple :-) */ > > Now, I'm just interessted in some few rows. > > All those gives good plans: > > explain select c.id from c order by c.id limit 1; > explain select c.id from c group by c.id order by c.id limit 1; > explain select c.id from c join b on c_id=c.id order by c.id limit 1; > > ... BUT ... combining join, group and limit makes havoc: > > explain select c.id from c join b on c_id=c.id group by c.id order by c.id > desc limit 5; Where's b in this join clause? It looks like a cartesian product to me. -jwb From pgsql-performance-owner@postgresql.org Thu Aug 25 23:07:44 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1331BD6E53 for ; Thu, 25 Aug 2005 23:07:30 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 86805-03 for ; Fri, 26 Aug 2005 02:06:52 +0000 (GMT) Received: from mail.nordicbet.com (mail.nordicbet.com [193.69.167.130]) by svr1.postgresql.org (Postfix) with ESMTP id 9569AD791D for ; Thu, 25 Aug 2005 23:06:49 -0300 (ADT) Received: from [80.203.125.83] (helo=tobias.nordicbet.invalid) (Authenticated Sender=tobias@nordicbet.com) by mail.nordicbet.com with esmtpa (Exim 4.50 #1 (Debian)) id 1E8TZl-0006IO-Ly; Fri, 26 Aug 2005 04:04:23 +0200 Received: by tobias.nordicbet.invalid (Postfix, from userid 500) id 90C47DECFD; Fri, 26 Aug 2005 04:06:35 +0200 (CEST) Date: Fri, 26 Aug 2005 04:06:35 +0200 From: Tobias Brox To: "Jeffrey W. Baker" Cc: Tobias Brox , pgsql-performance@postgresql.org Subject: Re: Limit + group + join Message-ID: <20050826020635.GM10328@tobias.lan> References: <20050826002709.GK10328@tobias.lan> <1125021419.16451.0.camel@noodles> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1125021419.16451.0.camel@noodles> Organization: Group Nordicbet User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/429 X-Sequence-Number: 14176 [Jeffrey W. Baker - Thu at 06:56:59PM -0700] > > explain select c.id from c join b on c_id=c.id group by c.id order by c.id > > desc limit 5; > > Where's b in this join clause? "join b on c_id=c.id" It just a funny way of writing: select c.id from c,b where c_id=c.id group by c.id order by c.id desc limit 5; > It looks like a cartesian product to me. No. The query will return exactly the same as the simplest query: select c.id from c order by c.id desc limit 5; As said, this is a gross oversimplification of the production envorinment. In the production environment, I really need to use both join, group and limit. I tested a bit with subqueries, it was not a good solution (selecting really a lot of rows and aggregates from many of the tables). The next idea is to hack it up by manually finding out where the "limit" will cut, and place a restriction in the where-part of the query. -- Notice of Confidentiality: This information may be confidential, and blah-blah-blah - so please keep your eyes closed. Please delete and destroy this email. Failure to comply will cause my lawyer to yawn. From pgsql-performance-owner@postgresql.org Thu Aug 25 23:33:57 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CFE0CD7930 for ; Thu, 25 Aug 2005 23:30:14 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 96597-05 for ; Fri, 26 Aug 2005 02:30:11 +0000 (GMT) Received: from mail23.sea5.speakeasy.net (mail23.sea5.speakeasy.net [69.17.117.25]) by svr1.postgresql.org (Postfix) with ESMTP id 0611BD78E3 for ; Thu, 25 Aug 2005 23:30:04 -0300 (ADT) Received: (qmail 12130 invoked from network); 26 Aug 2005 02:30:07 -0000 Received: from dsl081-060-184.sfo1.dsl.speakeasy.net (HELO noodles) ([64.81.60.184]) (envelope-sender ) by mail23.sea5.speakeasy.net (qmail-ldap-1.03) with RC4-MD5 encrypted SMTP for ; 26 Aug 2005 02:30:07 -0000 Subject: Re: Limit + group + join From: "Jeffrey W. Baker" To: Tobias Brox Cc: pgsql-performance@postgresql.org In-Reply-To: <1125021419.16451.0.camel@noodles> References: <20050826002709.GK10328@tobias.lan> <1125021419.16451.0.camel@noodles> Content-Type: text/plain Date: Thu, 25 Aug 2005 19:31:20 -0700 Message-Id: <1125023480.16451.2.camel@noodles> Mime-Version: 1.0 X-Mailer: Evolution 2.2.1.1 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.038 required=5 tests=[AWL=0.038] X-Spam-Level: X-Archive-Number: 200508/430 X-Sequence-Number: 14177 On Thu, 2005-08-25 at 18:56 -0700, Jeffrey W. Baker wrote: > On Fri, 2005-08-26 at 02:27 +0200, Tobias Brox wrote: > > Consider this setup - which is a gross simplification of parts of our > > production system ;-) > > > > create table c (id integer primary key); > > create table b (id integer primary key, c_id integer); > > create index b_on_c on b(c_id) > > > > insert into c (select ... lots of IDs ...); > > insert into b (select id, id from c); /* keep it simple :-) */ > > > > Now, I'm just interessted in some few rows. > > > > All those gives good plans: > > > > explain select c.id from c order by c.id limit 1; > > explain select c.id from c group by c.id order by c.id limit 1; > > explain select c.id from c join b on c_id=c.id order by c.id limit 1; > > > > ... BUT ... combining join, group and limit makes havoc: > > > > explain select c.id from c join b on c_id=c.id group by c.id order by c.id > > desc limit 5; > > Where's b in this join clause? It looks like a cartesian product to me. Nevermind. I read c_id as c.id. -jwb From pgsql-performance-owner@postgresql.org Fri Aug 26 00:44:12 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id F411FD6FC7 for ; Fri, 26 Aug 2005 00:17:53 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 17829-03 for ; Fri, 26 Aug 2005 03:17:52 +0000 (GMT) Received: from linda-3.paradise.net.nz (bm-3a.paradise.net.nz [202.0.58.22]) by svr1.postgresql.org (Postfix) with ESMTP id 08BFAD6FDD for ; Fri, 26 Aug 2005 00:17:51 -0300 (ADT) Received: from smtp-1.paradise.net.nz (smtp-1a.paradise.net.nz [202.0.32.194]) by linda-3.paradise.net.nz (Paradise.net.nz) with ESMTP id <0ILT00KK77TO3C@linda-3.paradise.net.nz> for pgsql-performance@postgresql.org; Fri, 26 Aug 2005 15:17:50 +1200 (NZST) Received: from [192.168.1.11] (218-101-14-16.paradise.net.nz [218.101.14.16]) by smtp-1.paradise.net.nz (Postfix) with ESMTP id 9C26E831B2; Fri, 26 Aug 2005 15:01:03 +1200 (NZST) Date: Fri, 26 Aug 2005 15:01:01 +1200 From: Mark Kirkwood Subject: Re: Limit + group + join In-reply-to: <20050826002709.GK10328@tobias.lan> To: Tobias Brox Cc: pgsql-performance@postgresql.org Message-id: <430E85ED.60109@paradise.net.nz> MIME-version: 1.0 Content-type: text/plain; format=flowed; charset=ISO-8859-1 Content-transfer-encoding: 7bit X-Accept-Language: en-us, en User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050726) References: <20050826002709.GK10328@tobias.lan> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.653 required=5 tests=[AWL=-0.373, RCVD_IN_NJABL_PROXY=1.026] X-Spam-Level: X-Archive-Number: 200508/431 X-Sequence-Number: 14178 Tobias, Interesting example: The 'desc' seems to be the guy triggering the sort, e.g: explain select c.id from c join b on c_id=c.id group by c.id order by c.id limit 5; QUERY PLAN ----------------------------------------------------------------------------------------- Limit (cost=0.00..0.28 rows=5 width=4) -> Group (cost=0.00..4476.00 rows=80000 width=4) -> Merge Join (cost=0.00..4276.00 rows=80000 width=4) Merge Cond: ("outer".id = "inner".c_id) -> Index Scan using c_pkey on c (cost=0.00..1518.00 rows=80000 width=4) -> Index Scan using b_on_c on b (cost=0.00..1558.00 rows=80000 width=4) (6 rows) Whereas with it back in again: explain select c.id from c join b on c_id=c.id group by c.id order by c.id desc limit 5; QUERY PLAN -------------------------------------------------------------------------------------- Limit (cost=10741.08..10741.11 rows=5 width=4) -> Group (cost=10741.08..11141.08 rows=80000 width=4) -> Sort (cost=10741.08..10941.08 rows=80000 width=4) Sort Key: c.id -> Hash Join (cost=1393.00..4226.00 rows=80000 width=4) Hash Cond: ("outer".c_id = "inner".id) -> Seq Scan on b (cost=0.00..1233.00 rows=80000 width=4) -> Hash (cost=1193.00..1193.00 rows=80000 width=4) -> Seq Scan on c (cost=0.00..1193.00 rows=80000 width=4) (9 rows) However being a bit brutal: set enable_mergejoin=false; set enable_hashjoin=false; explain select c.id from c join b on c_id=c.id group by c.id order by c.id desc limit 5; QUERY PLAN -------------------------------------------------------------------------------------------------- Limit (cost=0.00..15.24 rows=5 width=4) -> Group (cost=0.00..243798.00 rows=80000 width=4) -> Nested Loop (cost=0.00..243598.00 rows=80000 width=4) -> Index Scan Backward using c_pkey on c (cost=0.00..1518.00 rows=80000 width=4) -> Index Scan using b_on_c on b (cost=0.00..3.01 rows=1 width=4) Index Cond: (b.c_id = "outer".id) (6 rows) What is interesting is why this plan is being rejected... Cheers Mark Tobias Brox wrote: > Consider this setup - which is a gross simplification of parts of our > production system ;-) > > create table c (id integer primary key); > create table b (id integer primary key, c_id integer); > create index b_on_c on b(c_id) > > insert into c (select ... lots of IDs ...); > insert into b (select id, id from c); /* keep it simple :-) */ > > Now, I'm just interessted in some few rows. > > All those gives good plans: > > explain select c.id from c order by c.id limit 1; > explain select c.id from c group by c.id order by c.id limit 1; > explain select c.id from c join b on c_id=c.id order by c.id limit 1; > > ... BUT ... combining join, group and limit makes havoc: > > explain select c.id from c join b on c_id=c.id group by c.id order by c.id > desc limit 5; > From pgsql-performance-owner@postgresql.org Fri Aug 26 03:08:44 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5A222D6F75 for ; Fri, 26 Aug 2005 03:08:43 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 60615-09 for ; Fri, 26 Aug 2005 06:08:41 +0000 (GMT) Received: from smtp-gw.fnbs.net.my (smtp-gw.fnbs.net.my [202.9.108.191]) by svr1.postgresql.org (Postfix) with ESMTP id 0AF9FD6E40 for ; Fri, 26 Aug 2005 03:08:39 -0300 (ADT) Received: from mail-std.fnbs.net.my (smtp-std.fnbs.net.my [202.9.108.197]) by fnsrvlx9.fnbs.net.my (SMTP Mailer) with ESMTP id 786E152AB7 for ; Fri, 26 Aug 2005 14:08:37 +0800 (MYT) Received: from Beh (unverified [203.106.54.162]) by mail-std.fnbs.net.my (SurgeMail 3.0c2) with ESMTP id 18087424 for ; Fri, 26 Aug 2005 14:08:37 +0800 MYT Message-ID: <000b01c5aa04$a1ba0100$a279640a@Beh> From: "Chun Yit(Chronos)" To: Subject: postmaster memory keep going up???? Date: Fri, 26 Aug 2005 14:08:51 +0800 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0008_01C5AA47.AF9CDCA0" X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 6.00.2900.2180 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.077 required=5 tests=[AWL=-0.019, HTML_50_60=0.095, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/432 X-Sequence-Number: 14179 This is a multi-part message in MIME format. ------=_NextPart_000_0008_01C5AA47.AF9CDCA0 Content-Type: text/plain; charset="gb2312" Content-Transfer-Encoding: quoted-printable >I have a pl/pgsql function that using temp table to perform searching = logic, >we have one server running on 512MB, Red Hat 9.0, postgresql-7.4.5. >the problem is the pl/pgsql function that i created will increase = postmaster memory when calling to function >become more frequent, i did a test by getting out all the logic inside = the function and what left only >create temporary table and drop the temporary table statement (at the = end if this function), i monitor the %mem for postmaster >using linux command, ps -eo pid,comm,user,%mem | grep postmaster. >when i start the postmaster, the %mem show only 2.0 something, but = after i run the function for more that 1000 time, then >the %mem will go up until 10.0 something. >my question is,it is postmaster have memory leaking problem? >hope someone can give me some help and best is how to identify the = problem it is come from postgresql? > >thanks >regards >ivan ------=_NextPart_000_0008_01C5AA47.AF9CDCA0 Content-Type: text/html; charset="gb2312" Content-Transfer-Encoding: quoted-printable
>I have a pl/pgsql function that = using temp=20 table to perform searching logic,
>we have one=20 server running on 512MB, Red Hat 9.0, postgresql-7.4.5.
>the problem is the pl/pgsql = function that i=20 created will increase postmaster memory when calling to = function
>become more frequent, i did a test = by getting=20 out all the logic inside the function and what left only
>create temporary table and drop the = temporary=20 table statement (at the end if this function), i monitor the %mem for=20 postmaster
>using linux command, ps -eo = pid,comm,user,%mem=20 | grep postmaster.
>when i start the postmaster, the = %mem show only=20 2.0 something, but after i run the function for more that 1000 time,=20 then
>the %mem will go up until 10.0=20 something.
>my question is,it is postmaster = have memory=20 leaking problem?
>hope someone can give me some help = and best is=20 how to identify the problem it is come from postgresql?
>
>thanks
>regards
>ivan
------=_NextPart_000_0008_01C5AA47.AF9CDCA0-- From pgsql-performance-owner@postgresql.org Fri Aug 26 03:21:16 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 94794D7859 for ; Fri, 26 Aug 2005 03:21:14 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 67689-06 for ; Fri, 26 Aug 2005 06:21:11 +0000 (GMT) Received: from mail.nordicbet.com (mail.nordicbet.com [193.69.167.130]) by svr1.postgresql.org (Postfix) with ESMTP id 60568D7838 for ; Fri, 26 Aug 2005 03:21:09 -0300 (ADT) Received: from [80.203.125.83] (helo=tobias.nordicbet.invalid) (Authenticated Sender=tobias@nordicbet.com) by mail.nordicbet.com with esmtpa (Exim 4.50 #1 (Debian)) id 1E8XXp-0006yn-GO; Fri, 26 Aug 2005 08:18:39 +0200 Received: by tobias.nordicbet.invalid (Postfix, from userid 500) id 8FB69DECFD; Fri, 26 Aug 2005 08:20:51 +0200 (CEST) Date: Fri, 26 Aug 2005 08:20:51 +0200 From: Tobias Brox To: Mark Kirkwood Cc: Tobias Brox , pgsql-performance@postgresql.org Subject: Re: Limit + group + join Message-ID: <20050826062051.GN10328@tobias.lan> References: <20050826002709.GK10328@tobias.lan> <430E85ED.60109@paradise.net.nz> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <430E85ED.60109@paradise.net.nz> Organization: Group Nordicbet User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/433 X-Sequence-Number: 14180 [Mark Kirkwood - Fri at 03:01:01PM +1200] > Tobias, > Interesting example: > > The 'desc' seems to be the guy triggering the sort, e.g: Oh; really an accident that I didn't notice myself, I was actually going to remove all instances of "desc" in my simplification, but seems like I forgot. > However being a bit brutal: > > set enable_mergejoin=false; > set enable_hashjoin=false; :-) maybe I can use that in production. I'll check. -- Notice of Confidentiality: This information may be confidential, and blah-blah-blah - so please keep your eyes closed. Please delete and destroy this email. Failure to comply will cause my lawyer to yawn. From pgsql-performance-owner@postgresql.org Fri Aug 26 05:35:36 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C8D8ED7C5F for ; Fri, 26 Aug 2005 05:35:35 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 03987-08 for ; Fri, 26 Aug 2005 08:35:34 +0000 (GMT) Received: from mail.metronet.co.uk (mail.metronet.co.uk [213.162.97.75]) by svr1.postgresql.org (Postfix) with ESMTP id 124CFD7C70 for ; Fri, 26 Aug 2005 05:35:31 -0300 (ADT) Received: from mainbox.archonet.com (84-51-143-99.archon037.adsl.metronet.co.uk [84.51.143.99]) by smtp.metronet.co.uk (MetroNet Mail) with ESMTP id 2A54B417C49; Fri, 26 Aug 2005 09:35:27 +0100 (BST) Received: from localhost (localhost.localdomain [127.0.0.1]) by mainbox.archonet.com (Postfix) with ESMTP id 45ABC15EDB; Fri, 26 Aug 2005 09:34:34 +0100 (BST) Received: from mainbox.archonet.com ([127.0.0.1]) by localhost (mainbox [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 31733-01; Fri, 26 Aug 2005 09:34:32 +0100 (BST) Received: from [192.168.1.17] (client17.office.archonet.com [192.168.1.17]) by mainbox.archonet.com (Postfix) with ESMTP id E8FFE15ED9; Fri, 26 Aug 2005 09:34:31 +0100 (BST) Message-ID: <430ED417.2050601@archonet.com> Date: Fri, 26 Aug 2005 09:34:31 +0100 From: Richard Huxton User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: "Chun Yit(Chronos)" Cc: pgsql-performance@postgresql.org Subject: Re: postmaster memory keep going up???? References: <000b01c5aa04$a1ba0100$a279640a@Beh> In-Reply-To: <000b01c5aa04$a1ba0100$a279640a@Beh> Content-Type: text/plain; charset=GB2312 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/434 X-Sequence-Number: 14181 Chun Yit(Chronos) wrote: >>I have a pl/pgsql function that using temp table to perform searching logic, >>my question is,it is postmaster have memory leaking problem? First step - upgrade to the latest 7.4.x release. Second step - read the "release notes" section of the manuals for 7.4.x and 8.0.x and see what it says about memory and plpgsql. -- Richard Huxton Archonet Ltd From pgsql-performance-owner@postgresql.org Fri Aug 26 08:09:35 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D215FD6FC6 for ; Fri, 26 Aug 2005 08:09:34 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 42421-07 for ; Fri, 26 Aug 2005 11:09:32 +0000 (GMT) Received: from gollum.cambrium.nl (mx1.cambrium.nl [217.19.16.130]) by svr1.postgresql.org (Postfix) with SMTP id 7B3B4D6D7B for ; Fri, 26 Aug 2005 08:09:28 -0300 (ADT) Received: (qmail 7482 invoked from network); 26 Aug 2005 11:09:28 -0000 Received: from office.tweakers.net (HELO ?10.0.0.157?) (84.245.2.46) by gollum.cambrium.nl with SMTP; 26 Aug 2005 11:09:28 -0000 Message-ID: <430EF844.70905@tweakers.net> Date: Fri, 26 Aug 2005 13:08:52 +0200 From: Arjen van der Meijden User-Agent: Mozilla Thunderbird 1.0+ (Windows/20050711) MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Inefficient queryplan for query with intersectable subselects/joins Content-Type: multipart/mixed; boundary="------------050607070007070009010202" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 required=5 tests=[FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/435 X-Sequence-Number: 14182 This is a multi-part message in MIME format. --------------050607070007070009010202 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Hi list, I'm writing an application that will aggregate records with a few million records into averages/sums/minimums etc grouped per day. Clients can add filters and do lots of customization on what they want to see. And I've to translate that to one or more queries. Basically, I append each filter as either an extra and-in-the-where or joined with the clauses as ON-clause. The application now uses 8.1devel but I think the basic plans are similar to 8.0. At least for this first query. I noticed a query taking over 25 seconds to execute: SELECT "make a timestamp" grouper, chart_2.Prijs as field_2_0 FROM pwprijs as chart_2 JOIN pwprodukten t_0 ON chart_2.ProduktID = t_0.ID AND t_0.Cat2 IN (SELECT 545 UNION SELECT ID FROM cat WHERE ParentID = 545) JOIN pwprijs t_1 ON chart_2.ProduktID = t_1.ProduktID AND t_1.LeverancierID = 938 AND t_1.recordtimestamp >= "last timestamp" WHERE chart_2.Prijs > 0 It yields quite a long plan, so I've send that as an attachment along. Basically it combines two tables against an original to fetch "all prices (of all suppliers) for products of a certain category that are sold by a certain supplier". I was wondering how rewriting it to subselects would improve performance, but that wasn't a very clear winner. It shaved of about 5 seconds. So I took the subselects and used INTERSECT to unite them and have only one IN-clause in the query. That made it go down to around 13 seconds. I noticed it was doing a seq scan on the largest table to get the "Prijs > 0"-condition. But since there are only 947 of the 7692207 with prijs = 0 and none with < 0, it shouldn't be the main thing to look for. Dropping the clause made a minor improvement in performance for the queries. But disabling sequential scans allowed an improvement to only 660 ms compared to the 13 seconds earlier! Row-estimates seem to be quite a bit off, so I already set the statistics target to 1000 and re-analyzed. Btw, adding the prijs-clause again would make it choose another index and thus resulted in much longer operation. The final query, only taking 650ms, would be: SELECT "make a timestamp" as grouper, chart_2.Prijs as field_2_0 FROM pwprijs as chart_2 WHERE chart_2.ProduktID IN (SELECT ID FROM pwprodukten WHERE Cat2 IN (SELECT 545 UNION SELECT ID FROM cat WHERE ParentID = 545) INTERSECT SELECT produktid FROM pwprijs WHERE LeverancierID = 938 AND recordtimestamp >= "last timestamp") So I'm wondering: how can I make postgres decide to use the (correct) index without having to disable seq scans and how can I still add the prijs-clause without dropping the index for it (since it should be used for other filters). And for ease of use in my application I'd prefer to use the first query or the version with two seperate IN-clauses. Is that possible? I left all the configuration-stuff to the defaults since changing values didn't seem to impact much. Apart from the buffers and effective cache, increasing those made the performance worse. Best regards, Arjen --------------050607070007070009010202 Content-Type: text/plain; name="query-plans.txt" Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="query-plans.txt" Table "public.pwprijs" Column | Type | Modifiers -----------------+---------+----------- produktid | integer | not null leverancierid | integer | not null prijs | real | not null Time | bigint | not null recordtimestamp | bigint | not null Indexes: "pwprijs_levid_idx" btree (leverancierid), tablespace "raptor" "pwprijs_levid_pid_idx" btree (leverancierid, produktid), tablespace "raptor" "pwprijs_levid_rects_idx" btree (leverancierid, recordtimestamp), tablespace "raptor" "pwprijs_produktid_timestamp_idx" btree (produktid, recordtimestamp), tablespace "raptor" "pwprijs_rec_levid_pid" btree (recordtimestamp, leverancierid, produktid), tablespace "raptor" "pwprijs_recordtimestamp_idx" btree (recordtimestamp), tablespace "raptor" "pwprijs_recordtimestamp_produktid_prijs_idx" btree (recordtimestamp, produktid, prijs), tablespace "raptor" Tablespace: "raptor" Table "public.pwprodukten" Column | Type | Modifiers -----------------+-----------------------+------------------------------ id | integer | not null naam | character varying(80) | not null cat2 | smallint | not null recordtimestamp | bigint | not null Indexes: "pwprodukten_pkey" PRIMARY KEY, btree (id) "pwprodukten_cat_idx" btree (cat2) SELECT CAST('epoch'::timestamp + (chart_2.RecordTimestamp - (chart_2.RecordTimestamp % 86400)) * interval '1 second' as date) as grouper, chart_2.Prijs as field_2_0 FROM pwprijs as chart_2 JOIN pwprodukten t_0 ON chart_2.ProduktID = t_0.ID AND t_0.Cat2 IN (SELECT 545 UNION SELECT ID FROM cat WHERE ParentID = 545) JOIN pwprijs t_1 ON chart_2.ProduktID = t_1.ProduktID AND t_1.LeverancierID = 938 AND t_1.recordtimestamp >= (SELECT max_date - 60 FROM last_dates WHERE table_name = 'pricetracker') WHERE chart_2.Prijs > 0 Hash Join (cost=72062.34..940100.37 rows=22901407 width=12) (actual time=746.520..29655.748 rows=58065 loops=1) Hash Cond: ("outer".produktid = "inner".id) InitPlan -> Seq Scan on last_dates (cost=0.00..1.06 rows=1 width=8) (actual time=0.041..0.045 rows=1 loops=1) Filter: ((table_name)::text = 'pricetracker'::text) -> Seq Scan on pwprijs chart_2 (cost=0.00..156721.59 rows=7691800 width=16) (actual time=8.457..14713.939 rows=7691260 loops=1) Filter: (prijs > 0::double precision) -> Hash (cost=71769.47..71769.47 rows=42319 width=8) (actual time=606.996..606.996 rows=103 loops=1) -> Hash Join (cost=4329.24..71769.47 rows=42319 width=8) (actual time=599.771..606.782 rows=103 loops=1) Hash Cond: ("outer".produktid = "inner".id) -> Bitmap Heap Scan on pwprijs t_1 (cost=787.84..65786.28 rows=84640 width=4) (actual time=36.049..40.618 rows=4405 loops=1) Recheck Cond: ((leverancierid = 938) AND (recordtimestamp >= $1)) -> Bitmap Index Scan on pwprijs_levid_rects_idx (cost=0.00..787.84 rows=84640 width=0) (actual time=36.004..36.004 rows=4405 loops=1) Index Cond: ((leverancierid = 938) AND (recordtimestamp >= $1)) -> Hash (cost=3399.01..3399.01 rows=22156 width=4) (actual time=561.313..561.313 rows=1458 loops=1) -> Seq Scan on pwprodukten t_0 (cost=10.45..3399.01 rows=22156 width=4) (actual time=60.957..559.595 rows=1458 loops=1) Filter: (subplan) SubPlan -> Materialize (cost=10.45..10.53 rows=8 width=2) (actual time=0.002..0.006 rows=8 loops=44313) -> Unique (cost=10.40..10.44 rows=8 width=2) (actual time=44.831..44.846 rows=8 loops=1) -> Sort (cost=10.40..10.42 rows=8 width=2) (actual time=44.829..44.834 rows=8 loops=1) Sort Key: "?column?" -> Append (cost=0.00..10.28 rows=8 width=2) (actual time=0.006..44.754 rows=8 loops=1) -> Subquery Scan "*SELECT* 1" (cost=0.00..0.02 rows=1 width=0) (actual time=0.004..0.006 rows=1 loops=1) -> Result (cost=0.00..0.01 rows=1 width=0) (actual time=0.002..0.003 rows=1 loops=1) -> Subquery Scan "*SELECT* 2" (cost=2.02..10.26 rows=7 width=2) (actual time=44.723..44.738 rows=7 loops=1) -> Bitmap Heap Scan on cat (cost=2.02..10.19 rows=7 width=2) (actual time=44.719..44.726 rows=7 loops=1) Recheck Cond: (parentid = 545) -> Bitmap Index Scan on cat_parentid_idx (cost=0.00..2.02 rows=7 width=0) (actual time=44.704..44.704 rows=7 loops=1) Index Cond: (parentid = 545) Total runtime: 29688.736 ms SELECT CAST('epoch'::timestamp + (chart_2.RecordTimestamp - (chart_2.RecordTimestamp % 86400)) * interval '1 second' as date) as grouper, chart_2.Prijs as field_2_0 FROM pwprijs as chart_2 WHERE chart_2.ProduktID IN (SELECT ID FROM pwprodukten WHERE Cat2 IN (SELECT 545 UNION SELECT ID FROM cat WHERE ParentID = 545)) AND chart_2.ProduktID IN (SELECT produktid FROM pwprijs WHERE LeverancierID = 938 AND recordtimestamp >= (SELECT max_date - 60 FROM last_dates WHERE table_name = 'pricetracker')) AND chart_2.Prijs > 0 Nested Loop IN Join (cost=2850.25..230556.55 rows=1699 width=12) (actual time=482.862..24419.848 rows=58065 loops=1) InitPlan -> Seq Scan on last_dates (cost=0.00..1.06 rows=1 width=8) (actual time=0.022..0.025 rows=1 loops=1) Filter: ((table_name)::text = 'pricetracker'::text) -> Hash IN Join (cost=2849.18..200109.24 rows=207947 width=20) (actual time=360.511..15450.592 rows=190487 loops=1) Hash Cond: ("outer".produktid = "inner".id) -> Seq Scan on pwprijs chart_2 (cost=0.00..156721.59 rows=7691800 width=16) (actual time=11.587..10137.642 rows=7691260 loops=1) Filter: (prijs > 0::double precision) -> Hash (cost=2846.19..2846.19 rows=1198 width=4) (actual time=244.576..244.576 rows=1458 loops=1) -> Nested Loop (cost=12.93..2846.19 rows=1198 width=4) (actual time=48.646..243.239 rows=1458 loops=1) -> Unique (cost=10.40..10.44 rows=8 width=2) (actual time=19.552..19.580 rows=8 loops=1) -> Sort (cost=10.40..10.42 rows=8 width=2) (actual time=19.550..19.559 rows=8 loops=1) Sort Key: "?column?" -> Append (cost=0.00..10.28 rows=8 width=2) (actual time=0.005..19.532 rows=8 loops=1) -> Subquery Scan "*SELECT* 1" (cost=0.00..0.02 rows=1 width=0) (actual time=0.005..0.007 rows=1 loops=1) -> Result (cost=0.00..0.01 rows=1 width=0) (actual time=0.003..0.003 rows=1 loops=1) -> Subquery Scan "*SELECT* 2" (cost=2.02..10.26 rows=7 width=2) (actual time=19.501..19.518 rows=7 loops=1) -> Bitmap Heap Scan on cat (cost=2.02..10.19 rows=7 width=2) (actual time=19.499..19.506 rows=7 loops=1) Recheck Cond: (parentid = 545) -> Bitmap Index Scan on cat_parentid_idx (cost=0.00..2.02 rows=7 width=0) (actual time=19.493..19.493 rows=7 loops=1) Index Cond: (parentid = 545) -> Bitmap Heap Scan on pwprodukten (cost=2.52..352.58 rows=150 width=6) (actual time=3.717..27.732 rows=182 loops=8) Recheck Cond: (pwprodukten.cat2 = "outer"."?column?") -> Bitmap Index Scan on pwprodukten_cat_idx (cost=0.00..2.52 rows=150 width=0) (actual time=1.510..1.510 rows=182 loops=8) Index Cond: (pwprodukten.cat2 = "outer"."?column?") -> Index Scan using pwprijs_levid_pid_idx on pwprijs (cost=0.00..34.16 rows=3 width=4) (actual time=0.044..0.044 rows=0 loops=190487) Index Cond: ((pwprijs.leverancierid = 938) AND ("outer".produktid = pwprijs.produktid)) Filter: (recordtimestamp >= $0) Total runtime: 24466.560 ms SELECT CAST('epoch'::timestamp + (chart_2.RecordTimestamp - (chart_2.RecordTimestamp % 86400)) * interval '1 second' as date) as grouper, chart_2.Prijs as field_2_0 FROM pwprijs as chart_2 WHERE chart_2.ProduktID IN (SELECT ID FROM pwprodukten WHERE Cat2 IN (SELECT 545 UNION SELECT ID FROM cat WHERE ParentID = 545) INTERSECT SELECT produktid FROM pwprijs WHERE LeverancierID = 938 AND recordtimestamp >= (SELECT max_date - 60 FROM last_dates WHERE table_name = 'pricetracker')) AND chart_2.Prijs > 0 Hash Join (cost=77791.44..457846.83 rows=2010732 width=12) (actual time=409.838..13202.566 rows=58065 loops=1) Hash Cond: ("outer".produktid = "inner".id) -> Seq Scan on pwprijs chart_2 (cost=0.00..156721.59 rows=7691800 width=16) (actual time=7.896..8253.585 rows=7691260 loops=1) Filter: (prijs > 0::double precision) -> Hash (cost=77769.98..77769.98 rows=8584 width=4) (actual time=325.008..325.008 rows=103 loops=1) -> SetOp Intersect (cost=77254.95..77684.14 rows=8584 width=4) (actual time=317.384..324.932 rows=103 loops=1) -> Sort (cost=77254.95..77469.55 rows=85838 width=4) (actual time=317.361..320.519 rows=5863 loops=1) Sort Key: id -> Append (cost=12.93..69491.92 rows=85838 width=4) (actual time=38.765..308.837 rows=5863 loops=1) -> Subquery Scan "*SELECT* 1" (cost=12.93..2858.17 rows=1198 width=4) (actual time=38.763..256.795 rows=1458 loops=1) -> Nested Loop (cost=12.93..2846.19 rows=1198 width=4) (actual time=38.761..254.989 rows=1458 loops=1) -> Unique (cost=10.40..10.44 rows=8 width=2) (actual time=8.994..9.027 rows=8 loops=1) -> Sort (cost=10.40..10.42 rows=8 width=2) (actual time=8.993..9.000 rows=8 loops=1) Sort Key: "?column?" -> Append (cost=0.00..10.28 rows=8 width=2) (actual time=0.003..8.978 rows=8 loops=1) -> Subquery Scan "*SELECT* 1" (cost=0.00..0.02 rows=1 width=0) (actual time=0.003..0.005 rows=1 loops=1) -> Result (cost=0.00..0.01 rows=1 width=0) (actual time=0.001..0.001 rows=1 loops=1) -> Subquery Scan "*SELECT* 2" (cost=2.02..10.26 rows=7 width=2) (actual time=8.949..8.963 rows=7 loops=1) -> Bitmap Heap Scan on cat (cost=2.02..10.19 rows=7 width=2) (actual time=8.945..8.952 rows=7 loops=1) Recheck Cond: (parentid = 545) -> Bitmap Index Scan on cat_parentid_idx (cost=0.00..2.02 rows=7 width=0) (actual time=8.940..8.940 rows=7 loops=1) Index Cond: (parentid = 545) -> Bitmap Heap Scan on pwprodukten (cost=2.52..352.58 rows=150 width=6) (actual time=3.847..30.522 rows=182 loops=8) Recheck Cond: (pwprodukten.cat2 = "outer"."?column?") -> Bitmap Index Scan on pwprodukten_cat_idx (cost=0.00..2.52 rows=150 width=0) (actual time=1.583..1.583 rows=182 loops=8) Index Cond: (pwprodukten.cat2 = "outer"."?column?") -> Subquery Scan "*SELECT* 2" (cost=788.90..66633.75 rows=84640 width=4) (actual time=36.653..46.266 rows=4405 loops=1) -> Bitmap Heap Scan on pwprijs (cost=788.90..65787.35 rows=84640 width=4) (actual time=36.651..41.326 rows=4405 loops=1) Recheck Cond: ((leverancierid = 938) AND (recordtimestamp >= $0)) InitPlan -> Seq Scan on last_dates (cost=0.00..1.06 rows=1 width=8) (actual time=0.025..0.029 rows=1 loops=1) Filter: ((table_name)::text = 'pricetracker'::text) -> Bitmap Index Scan on pwprijs_levid_rects_idx (cost=0.00..787.84 rows=84640 width=0) (actual time=36.614..36.614 rows=4405 loops=1) Index Cond: ((leverancierid = 938) AND (recordtimestamp >= $0)) Total runtime: 13234.570 ms set enable_seqscan = false; without prijs-clause: Nested Loop (cost=77299.32..7858118.21 rows=2013780 width=12) (actual time=33.964..634.359 rows=58065 loops=1) -> SetOp Intersect (cost=77296.51..77727.66 rows=8623 width=4) (actual time=33.632..42.356 rows=103 loops=1) -> Sort (cost=77296.51..77512.08 rows=86231 width=4) (actual time=33.616..36.979 rows=5863 loops=1) Sort Key: id -> Append (cost=12.93..69495.92 rows=86231 width=4) (actual time=0.214..25.380 rows=5863 loops=1) -> Subquery Scan "*SELECT* 1" (cost=12.93..2858.17 rows=1198 width=4) (actual time=0.213..8.779 rows=1458 loops=1) -> Nested Loop (cost=12.93..2846.19 rows=1198 width=4) (actual time=0.211..7.065 rows=1458 loops=1) -> Unique (cost=10.40..10.44 rows=8 width=2) (actual time=0.095..0.122 rows=8 loops=1) -> Sort (cost=10.40..10.42 rows=8 width=2) (actual time=0.093..0.099 rows=8 loops=1) Sort Key: "?column?" -> Append (cost=0.00..10.28 rows=8 width=2) (actual time=0.003..0.079 rows=8 loops=1) -> Subquery Scan "*SELECT* 1" (cost=0.00..0.02 rows=1 width=0) (actual time=0.003..0.005 rows=1 loops=1) -> Result (cost=0.00..0.01 rows=1 width=0) (actual time=0.001..0.001 rows=1 loops=1) -> Subquery Scan "*SELECT* 2" (cost=2.02..10.26 rows=7 width=2) (actual time=0.050..0.065 rows=7 loops=1) -> Bitmap Heap Scan on cat (cost=2.02..10.19 rows=7 width=2) (actual time=0.048..0.054 rows=7 loops=1) Recheck Cond: (parentid = 545) -> Bitmap Index Scan on cat_parentid_idx (cost=0.00..2.02 rows=7 width=0) (actual time=0.040..0.040 rows=7 loops=1) Index Cond: (parentid = 545) -> Bitmap Heap Scan on pwprodukten (cost=2.52..352.58 rows=150 width=6) (actual time=0.076..0.658 rows=182 loops=8) Recheck Cond: (pwprodukten.cat2 = "outer"."?column?") -> Bitmap Index Scan on pwprodukten_cat_idx (cost=0.00..2.52 rows=150 width=0) (actual time=0.056..0.056 rows=182 loops=8) Index Cond: (pwprodukten.cat2 = "outer"."?column?") -> Subquery Scan "*SELECT* 2" (cost=794.53..66637.75 rows=85033 width=4) (actual time=1.064..10.785 rows=4405 loops=1) -> Bitmap Heap Scan on pwprijs (cost=794.53..65787.42 rows=85033 width=4) (actual time=1.062..5.792 rows=4405 loops=1) Recheck Cond: ((leverancierid = 938) AND (recordtimestamp >= $0)) InitPlan -> Index Scan using last_dates_pkey on last_dates (cost=0.00..3.33 rows=1 width=8) (actual time=0.042..0.044 rows=1 loops=1) Index Cond: ((table_name)::text = 'pricetracker'::text) -> Bitmap Index Scan on pwprijs_levid_rects_idx (cost=0.00..791.20 rows=85033 width=0) (actual time=1.027..1.027 rows=4405 loops=1) Index Cond: ((leverancierid = 938) AND (recordtimestamp >= $0)) -> Bitmap Heap Scan on pwprijs chart_2 (cost=2.82..895.85 rows=234 width=16) (actual time=0.363..3.980 rows=564 loops=103) Recheck Cond: (chart_2.produktid = "outer".id) -> Bitmap Index Scan on pwprijs_produktid_timestamp_idx (cost=0.00..2.82 rows=234 width=0) (actual time=0.206..0.206 rows=564 loops=103) Index Cond: (chart_2.produktid = "outer".id) Total runtime: 665.335 ms with prijs-clause: Hash Join (cost=134379.17..514503.02 rows=2013674 width=12) (actual time=3089.037..15063.409 rows=58065 loops=1) Hash Cond: ("outer".produktid = "inner".id) -> Bitmap Heap Scan on pwprijs chart_2 (cost=56543.72..213260.22 rows=7691800 width=16) (actual time=2972.869..10299.608 rows=7691260 loops=1) Recheck Cond: (prijs > 0::double precision) -> Bitmap Index Scan on pwprijs_recordtimestamp_produktid_prijs_idx (cost=0.00..56543.72 rows=7691800 width=0) (actual time=2962.018..2962.018 rows=7691260 loops=1) Index Cond: (prijs > 0::double precision) -> Hash (cost=77813.89..77813.89 rows=8623 width=4) (actual time=41.373..41.373 rows=103 loops=1) -> SetOp Intersect (cost=77296.51..77727.66 rows=8623 width=4) (actual time=33.770..41.297 rows=103 loops=1) -> Sort (cost=77296.51..77512.08 rows=86231 width=4) (actual time=33.756..36.971 rows=5863 loops=1) Sort Key: id -> Append (cost=12.93..69495.92 rows=86231 width=4) (actual time=0.247..25.379 rows=5863 loops=1) -> Subquery Scan "*SELECT* 1" (cost=12.93..2858.17 rows=1198 width=4) (actual time=0.246..8.816 rows=1458 loops=1) -> Nested Loop (cost=12.93..2846.19 rows=1198 width=4) (actual time=0.243..7.092 rows=1458 loops=1) -> Unique (cost=10.40..10.44 rows=8 width=2) (actual time=0.110..0.135 rows=8 loops=1) -> Sort (cost=10.40..10.42 rows=8 width=2) (actual time=0.108..0.115 rows=8 loops=1) Sort Key: "?column?" -> Append (cost=0.00..10.28 rows=8 width=2) (actual time=0.004..0.092 rows=8 loops=1) -> Subquery Scan "*SELECT* 1" (cost=0.00..0.02 rows=1 width=0) (actual time=0.004..0.006 rows=1 loops=1) -> Result (cost=0.00..0.01 rows=1 width=0) (actual time=0.002..0.003 rows=1 loops=1) -> Subquery Scan "*SELECT* 2" (cost=2.02..10.26 rows=7 width=2) (actual time=0.062..0.077 rows=7 loops=1) -> Bitmap Heap Scan on cat (cost=2.02..10.19 rows=7 width=2) (actual time=0.059..0.066 rows=7 loops=1) Recheck Cond: (parentid = 545) -> Bitmap Index Scan on cat_parentid_idx (cost=0.00..2.02 rows=7 width=0) (actual time=0.042..0.042 rows=7 loops=1) Index Cond: (parentid = 545) -> Bitmap Heap Scan on pwprodukten (cost=2.52..352.58 rows=150 width=6) (actual time=0.078..0.658 rows=182 loops=8) Recheck Cond: (pwprodukten.cat2 = "outer"."?column?") -> Bitmap Index Scan on pwprodukten_cat_idx (cost=0.00..2.52 rows=150 width=0) (actual time=0.058..0.058 rows=182 loops=8) Index Cond: (pwprodukten.cat2 = "outer"."?column?") -> Subquery Scan "*SELECT* 2" (cost=794.53..66637.75 rows=85033 width=4) (actual time=1.075..10.724 rows=4405 loops=1) -> Bitmap Heap Scan on pwprijs (cost=794.53..65787.42 rows=85033 width=4) (actual time=1.072..5.846 rows=4405 loops=1) Recheck Cond: ((leverancierid = 938) AND (recordtimestamp >= $0)) InitPlan -> Index Scan using last_dates_pkey on last_dates (cost=0.00..3.33 rows=1 width=8) (actual time=0.045..0.047 rows=1 loops=1) Index Cond: ((table_name)::text = 'pricetracker'::text) -> Bitmap Index Scan on pwprijs_levid_rects_idx (cost=0.00..791.20 rows=85033 width=0) (actual time=1.037..1.037 rows=4405 loops=1) Index Cond: ((leverancierid = 938) AND (recordtimestamp >= $0)) Total runtime: 15095.402 ms --------------050607070007070009010202-- From pgsql-performance-owner@postgresql.org Fri Aug 26 10:10:24 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 839FFD7066 for ; Fri, 26 Aug 2005 10:10:23 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 81936-01 for ; Fri, 26 Aug 2005 13:10:18 +0000 (GMT) Received: from mail.metronet.co.uk (mail.metronet.co.uk [213.162.97.75]) by svr1.postgresql.org (Postfix) with ESMTP id 314B1D6FEF for ; Fri, 26 Aug 2005 10:10:16 -0300 (ADT) Received: from mainbox.archonet.com (84-51-143-99.archon037.adsl.metronet.co.uk [84.51.143.99]) by smtp.metronet.co.uk (MetroNet Mail) with ESMTP id B87C440DCD2; Fri, 26 Aug 2005 14:10:11 +0100 (BST) Received: from localhost (localhost.localdomain [127.0.0.1]) by mainbox.archonet.com (Postfix) with ESMTP id 991FC15EDA; Fri, 26 Aug 2005 14:05:56 +0100 (BST) Received: from mainbox.archonet.com ([127.0.0.1]) by localhost (mainbox [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 06572-01; Fri, 26 Aug 2005 14:05:53 +0100 (BST) Received: from [192.168.1.17] (client17.office.archonet.com [192.168.1.17]) by mainbox.archonet.com (Postfix) with ESMTP id 0497D15EDE; Fri, 26 Aug 2005 14:05:24 +0100 (BST) Message-ID: <430F1393.5040704@archonet.com> Date: Fri, 26 Aug 2005 14:05:23 +0100 From: Richard Huxton User-Agent: Mozilla Thunderbird 1.0.2 (X11/20050317) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Arjen van der Meijden Cc: pgsql-performance@postgresql.org Subject: Re: Inefficient queryplan for query with intersectable References: <430EF844.70905@tweakers.net> In-Reply-To: <430EF844.70905@tweakers.net> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/436 X-Sequence-Number: 14183 Arjen van der Meijden wrote: > > I left all the configuration-stuff to the defaults since changing values > didn't seem to impact much. Apart from the buffers and effective cache, > increasing those made the performance worse. I've not looked at the rest of your problem in detail, but using the default configuration values is certainly not going to help things. In particular effective_cache is supposed to tell PG how much memory your OS is using to cache data. Read this through and make sure your configuration settings are sane, then it might be worthwhile looking in detail at this particular query. http://www.powerpostgresql.com/PerfList -- Richard Huxton Archonet Ltd From pgsql-performance-owner@postgresql.org Fri Aug 26 11:18:36 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8CCE2D705A for ; Fri, 26 Aug 2005 11:18:34 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91898-09 for ; Fri, 26 Aug 2005 14:18:33 +0000 (GMT) Received: from megazone.bigpanda.com (megazone.bigpanda.com [64.147.171.210]) by svr1.postgresql.org (Postfix) with ESMTP id DA05FD6F0C for ; Fri, 26 Aug 2005 11:18:32 -0300 (ADT) Received: by megazone.bigpanda.com (Postfix, from userid 1001) id F334435485; Fri, 26 Aug 2005 07:18:34 -0700 (PDT) Received: from localhost (localhost [127.0.0.1]) by megazone.bigpanda.com (Postfix) with ESMTP id DAAC63547A; Fri, 26 Aug 2005 07:18:34 -0700 (PDT) Date: Fri, 26 Aug 2005 07:18:30 -0700 (PDT) From: Stephan Szabo To: Mark Kirkwood Cc: Tobias Brox , pgsql-performance@postgresql.org Subject: Re: Limit + group + join In-Reply-To: <430E85ED.60109@paradise.net.nz> Message-ID: <20050826071621.P12124@megazone.bigpanda.com> References: <20050826002709.GK10328@tobias.lan> <430E85ED.60109@paradise.net.nz> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/437 X-Sequence-Number: 14184 On Fri, 26 Aug 2005, Mark Kirkwood wrote: > However being a bit brutal: > > set enable_mergejoin=false; > set enable_hashjoin=false; > > explain select c.id from c join b on c_id=c.id group by c.id order by > c.id desc limit 5; > QUERY PLAN > > -------------------------------------------------------------------------------------------------- > Limit (cost=0.00..15.24 rows=5 width=4) > -> Group (cost=0.00..243798.00 rows=80000 width=4) > -> Nested Loop (cost=0.00..243598.00 rows=80000 width=4) > -> Index Scan Backward using c_pkey on c > (cost=0.00..1518.00 rows=80000 width=4) > -> Index Scan using b_on_c on b (cost=0.00..3.01 > rows=1 width=4) > Index Cond: (b.c_id = "outer".id) > (6 rows) > > What is interesting is why this plan is being rejected... Well, it expects 80000 probles into b_on_c to be more expensive than the hash join and sort. I wonder what explain analyze shows for the original and the version with the enables changed. From pgsql-performance-owner@postgresql.org Fri Aug 26 11:49:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5B116D79F9 for ; Fri, 26 Aug 2005 11:45:14 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 08460-01 for ; Fri, 26 Aug 2005 14:45:11 +0000 (GMT) Received: from unicorn.rentec.com (unicorn.rentec.com [216.223.240.9]) by svr1.postgresql.org (Postfix) with ESMTP id 64683D79C0 for ; Fri, 26 Aug 2005 11:45:06 -0300 (ADT) Received: from wren.rentec.com (wren.rentec.com [192.5.35.106]) by unicorn.rentec.com (8.13.1/8.12.1) with ESMTP id j7QEj72t013241 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO) for ; Fri, 26 Aug 2005 10:45:09 -0400 (EDT) X-Rentec: external Received: from [172.26.132.145] (hoopoe.rentec.com [172.26.132.145]) by wren.rentec.com (8.13.1/8.12.1) with ESMTP id j7QEj7wf029455 for ; Fri, 26 Aug 2005 10:45:07 -0400 (EDT) Message-ID: <430F2AF3.1080704@rentec.com> Date: Fri, 26 Aug 2005 10:45:07 -0400 From: Alan Stange Reply-To: stange@rentec.com Organization: Renaissance Technologies Corp. User-Agent: Mozilla Thunderbird 1.0+ (X11/20050712) MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: difference in plan between 8.0 and 8.1? Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Logged: Logged by unicorn.rentec.com as j7QEj72t013241 at Fri Aug 26 10:45:09 2005 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.018 required=5 tests=[AWL=0.018] X-Spam-Level: X-Archive-Number: 200508/438 X-Sequence-Number: 14185 Hello all, I was hoping someone could explain the plan for a statement. We have a table with a column of longs being used as an index. The query plan in 8.0 was like this: # explain select distinct timeseriesid from tbltimeseries where timeseriesid > 0 order by timeseriesid; SET QUERY PLAN ------------------------------------------------------------------------------------------------------- Unique (cost=0.00..15065908.60 rows=10854026 width=8) -> Index Scan using idx_timeseris on tbltimeseries (cost=0.00..15038773.53 rows=10854026 width=8) Index Cond: (timeseriesid > 0) (3 rows) In 8.1, (using the same database after a dump+restore+vacuum+analyze) I get the following: # explain select distinct timeseriesid from tbltimeseries where timeseriesid > 0 order by timeseriesid; QUERY PLAN --------------------------------------------------------------------------------------------------- Unique (cost=2717137.08..2771407.21 rows=10854026 width=8) -> Sort (cost=2717137.08..2744272.14 rows=10854026 width=8) Sort Key: timeseriesid -> Bitmap Heap Scan on tbltimeseries (cost=48714.09..1331000.42 rows=10854026 width=8) Recheck Cond: (timeseriesid > 0) -> Bitmap Index Scan on idx_timeseris (cost=0.00..48714.09 rows=10854026 width=0) Index Cond: (timeseriesid > 0) (7 rows) I'm hoping someone can explain the new query plan (as I'm not sure I understand what it is doing). Thanks! -- Alan From pgsql-performance-owner@postgresql.org Fri Aug 26 12:03:16 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7A7E7D6F0C for ; Fri, 26 Aug 2005 11:49:21 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 08438-05 for ; Fri, 26 Aug 2005 14:49:19 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id 2FA62D7B54 for ; Fri, 26 Aug 2005 11:49:18 -0300 (ADT) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: Limit + group + join Date: Fri, 26 Aug 2005 10:49:23 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3417DD1F1@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Limit + group + join Thread-Index: AcWqBqTCoYy6uEvNSTuYW78DFcX66gARl6pQ From: "Merlin Moncure" To: "Tobias Brox" Cc: , "Mark Kirkwood" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/439 X-Sequence-Number: 14186 Mark Kirkwood > > The 'desc' seems to be the guy triggering the sort, e.g: >=20 > Oh; really an accident that I didn't notice myself, I was actually going > to > remove all instances of "desc" in my simplification, but seems like I > forgot. If desc is the problem you can push the query into a subquery without sorting and sort the result. This is called an inline view. Sometimes you can pull a couple of tricks to force the view to materialize before it is sorted. aka=20 select q.* from ( some_complex_query=09 ) q order by ...; Merlin From pgsql-performance-owner@postgresql.org Fri Aug 26 12:46:10 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6DE24D79CA for ; Fri, 26 Aug 2005 12:12:52 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 13630-09 for ; Fri, 26 Aug 2005 15:12:50 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id 6443DD7971 for ; Fri, 26 Aug 2005 12:12:49 -0300 (ADT) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: difference in plan between 8.0 and 8.1? Date: Fri, 26 Aug 2005 11:12:48 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3417DD1F3@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] difference in plan between 8.0 and 8.1? Thread-Index: AcWqTbsvXjeYODI0SE2PjPG4tWTjAgAAk4pg From: "Merlin Moncure" To: Cc: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.057 required=5 tests=[AWL=0.007, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/443 X-Sequence-Number: 14190 > Hello all, >=20 > I was hoping someone could explain the plan for a statement. >=20 > We have a table with a column of longs being used as an index. The > query plan in 8.0 was like this: >=20 > # explain select distinct timeseriesid from tbltimeseries where > timeseriesid > 0 order by timeseriesid; I had the same problem. You probably already have seq scan turned off, or the server would be using that. You may have to turn bitmap off or rework you query such that the server will use the index. (between?). Anyways, distinct is code word for 'bad performance' :). Consider laying out tables such that it not necessary, for example set up table with RI link. Then you can do this in zero time. Good luck! Merlin From pgsql-performance-owner@postgresql.org Fri Aug 26 12:41:17 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id F1227D7BB8 for ; Fri, 26 Aug 2005 12:16:18 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 14515-10 for ; Fri, 26 Aug 2005 15:16:15 +0000 (GMT) Received: from tigger.fuhr.org (tigger.fuhr.org [63.214.45.158]) by svr1.postgresql.org (Postfix) with ESMTP id A4C36D7A74 for ; Fri, 26 Aug 2005 12:16:13 -0300 (ADT) Received: from winnie.fuhr.org (winnie.fuhr.org [10.1.0.1]) by tigger.fuhr.org (8.13.3/8.13.3) with ESMTP id j7QFG8WJ019211 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK); Fri, 26 Aug 2005 09:16:11 -0600 (MDT) (envelope-from mfuhr@winnie.fuhr.org) Received: from winnie.fuhr.org (localhost [127.0.0.1]) by winnie.fuhr.org (8.13.3/8.13.3) with ESMTP id j7QFG8kP016969; Fri, 26 Aug 2005 09:16:08 -0600 (MDT) (envelope-from mfuhr@winnie.fuhr.org) Received: (from mfuhr@localhost) by winnie.fuhr.org (8.13.3/8.13.3/Submit) id j7QFG8Td016968; Fri, 26 Aug 2005 09:16:08 -0600 (MDT) (envelope-from mfuhr) Date: Fri, 26 Aug 2005 09:16:08 -0600 From: Michael Fuhr To: Alan Stange Cc: pgsql-performance@postgresql.org Subject: Re: difference in plan between 8.0 and 8.1? Message-ID: <20050826151608.GA16945@winnie.fuhr.org> References: <430F2AF3.1080704@rentec.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <430F2AF3.1080704@rentec.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 required=5 tests=[AWL=0.005] X-Spam-Level: X-Archive-Number: 200508/441 X-Sequence-Number: 14188 On Fri, Aug 26, 2005 at 10:45:07AM -0400, Alan Stange wrote: > -> Bitmap Heap Scan on tbltimeseries (cost=48714.09..1331000.42 rows=10854026 width=8) > Recheck Cond: (timeseriesid > 0) > -> Bitmap Index Scan on idx_timeseris (cost=0.00..48714.09 rows=10854026 width=0) > Index Cond: (timeseriesid > 0) > > I'm hoping someone can explain the new query plan (as I'm not sure I > understand what it is doing). Search for "bitmap" in the 8.1 Release Notes: http://developer.postgresql.org/docs/postgres/release.html#RELEASE-8-1 You could probably find more detailed discussion in the pgsql-hackers archives. -- Michael Fuhr From pgsql-performance-owner@postgresql.org Fri Aug 26 12:40:33 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CBCDED79BC for ; Fri, 26 Aug 2005 12:17:02 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 16234-03 for ; Fri, 26 Aug 2005 15:16:59 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id A3115D7108 for ; Fri, 26 Aug 2005 12:16:55 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7QFGtGH009249; Fri, 26 Aug 2005 11:16:55 -0400 (EDT) To: stange@rentec.com Cc: pgsql-performance@postgresql.org Subject: Re: difference in plan between 8.0 and 8.1? In-reply-to: <430F2AF3.1080704@rentec.com> References: <430F2AF3.1080704@rentec.com> Comments: In-reply-to Alan Stange message dated "Fri, 26 Aug 2005 10:45:07 -0400" Date: Fri, 26 Aug 2005 11:16:55 -0400 Message-ID: <9248.1125069415@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/440 X-Sequence-Number: 14187 Alan Stange writes: > Unique (cost=2717137.08..2771407.21 rows=10854026 width=8) > -> Sort (cost=2717137.08..2744272.14 rows=10854026 width=8) > Sort Key: timeseriesid > -> Bitmap Heap Scan on tbltimeseries > (cost=48714.09..1331000.42 rows=10854026 width=8) > Recheck Cond: (timeseriesid > 0) > -> Bitmap Index Scan on idx_timeseris > (cost=0.00..48714.09 rows=10854026 width=0) > Index Cond: (timeseriesid > 0) > (7 rows) > I'm hoping someone can explain the new query plan (as I'm not sure I > understand what it is doing). The index scan is reading the index to find out which heap tuple IDs (TIDs) the index says meet the condition. It returns a bitmap of the tuple locations (actually, an array of per-page bitmaps). The heap scan goes and fetches the tuples from the table, working in TID order to avoid re-reading the same page many times, as can happen for ordinary index scans. Since the result isn't sorted, we have to do a sort to get it into the correct order for the Unique step. Because it avoids random access to the heap, this plan can be a lot faster than a regular index scan. I'm not sure at all that 8.1 is doing good relative cost estimation yet, though. It would be interesting to see EXPLAIN ANALYZE results for both ways. (You can use enable_bitmapscan and enable_indexscan to force the planner to pick the plan it thinks is slower.) regards, tom lane From pgsql-performance-owner@postgresql.org Fri Aug 26 12:43:22 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8BD77D7822 for ; Fri, 26 Aug 2005 12:14:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 15622-03 for ; Fri, 26 Aug 2005 15:14:15 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.199]) by svr1.postgresql.org (Postfix) with ESMTP id 9C018D7BAE for ; Fri, 26 Aug 2005 12:14:11 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id 55so157146wri for ; Fri, 26 Aug 2005 08:14:11 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:date:from:to:subject:message-id:reply-to:mime-version:content-type:content-disposition:user-agent; b=OjPL+4xlPi1NInDxjEedtv92kRT1jQBt4djtwtrAESZNIn3CP44vEAJI7HxPmqNRpZBcBSLHZJZpvjaym7lyVFRkoVYEWW1dSE7hss3by6+Ezv03zrN1ypGeqnRhe4AaZEpLjL2SCege2lyS98y/BS0x2uBKMKuWJEeWmEGIXMM= Received: by 10.54.43.24 with SMTP id q24mr3448803wrq; Fri, 26 Aug 2005 08:14:09 -0700 (PDT) Received: from localhost ( [61.246.59.42]) by mx.gmail.com with ESMTP id 43sm2981017wri.2005.08.26.08.14.04; Fri, 26 Aug 2005 08:14:09 -0700 (PDT) Date: Fri, 26 Aug 2005 20:54:09 +0530 From: Ligesh To: pgsql-performance@postgresql.org Subject: Sending a select to multiple servers. Message-ID: <20050826152409.GA22060@lxlabs.com> Reply-To: gxlists@gmail.com Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/442 X-Sequence-Number: 14189 I would like to know if the following kind of database client exists: I need a 'select' query to be sent to say 10 db servers simultaneously in parallel (using threading), the results should be re-sorted and returned. For example I have a query: 'select * from table where parent_clname = 'parent' order by name limit 10'. Now this query has to be sent to 10 servers, and the maximum number of results would be 100. Now this 100 result set has to be re-sorted, out of which 90 has to be discarded, and the 10 has to be returned. Does such a solution exist now. To me this appears to be in entirety of what should constitute a database cluster. Only the search needs to be done on all the servers simultaneously at the low level. Once you get the results, the writing can be determined by the upper level logic (which can even be in a scripting language). But the search across many servers has to be done using proper threading, and the re-sorting also needs to be done fast. Thanks a lot in advance. -- :: Ligesh :: http://ligesh.com From pgsql-performance-owner@postgresql.org Fri Aug 26 12:49:36 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 29751D79CA for ; Fri, 26 Aug 2005 12:47:00 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 21828-05 for ; Fri, 26 Aug 2005 15:46:57 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id AADCFD79BC for ; Fri, 26 Aug 2005 12:46:55 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7QFkX2B009502; Fri, 26 Aug 2005 11:46:33 -0400 (EDT) To: Mark Kirkwood Cc: Tobias Brox , pgsql-performance@postgresql.org Subject: Re: Limit + group + join In-reply-to: <430E85ED.60109@paradise.net.nz> References: <20050826002709.GK10328@tobias.lan> <430E85ED.60109@paradise.net.nz> Comments: In-reply-to Mark Kirkwood message dated "Fri, 26 Aug 2005 15:01:01 +1200" Date: Fri, 26 Aug 2005 11:46:33 -0400 Message-ID: <9501.1125071193@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/444 X-Sequence-Number: 14191 Mark Kirkwood writes: > What is interesting is why this plan is being rejected... Which PG version are you using exactly? That mistake looks like an artifact of the 8.0 "fuzzy plan cost" patch, which we fixed recently: http://archives.postgresql.org/pgsql-committers/2005-07/msg00474.php But Tobias wasn't happy with 7.4 either, so I'm not sure that the fuzzy cost issue explains his results. As far as the "desc" point goes, the problem is that mergejoins aren't capable of dealing with backward sort order, so a merge plan isn't considered for that case (or at least, it would have to have a sort after it, which pretty much defeats the point for a fast-start plan). I have some ideas about fixing this but it won't happen before 8.2. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Aug 26 13:06:21 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 150FBD7B30 for ; Fri, 26 Aug 2005 13:04:27 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 25173-10 for ; Fri, 26 Aug 2005 16:04:22 +0000 (GMT) Received: from frank.wiles.org (frank.wiles.org [24.124.39.75]) by svr1.postgresql.org (Postfix) with ESMTP id 7B0A8D7B06 for ; Fri, 26 Aug 2005 13:04:21 -0300 (ADT) Received: from kungfu (frank.wiles.org [127.0.0.1]) by frank.wiles.org (8.13.1/8.13.1) with SMTP id j7QG4HPW004397; Fri, 26 Aug 2005 11:04:17 -0500 Date: Fri, 26 Aug 2005 11:04:59 -0500 From: Frank Wiles To: gxlists@gmail.com Cc: pgsql-performance@postgresql.org Subject: Re: Sending a select to multiple servers. Message-Id: <20050826110459.03cce4d9.frank@wiles.org> In-Reply-To: <20050826152409.GA22060@lxlabs.com> References: <20050826152409.GA22060@lxlabs.com> X-Mailer: Sylpheed version 1.0.4 (GTK+ 1.2.10; i386-redhat-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.019 required=5 tests=[AWL=0.019] X-Spam-Level: X-Archive-Number: 200508/445 X-Sequence-Number: 14192 On Fri, 26 Aug 2005 20:54:09 +0530 Ligesh wrote: > > I would like to know if the following kind of database client exists: > I need a 'select' query to be sent to say 10 db servers > simultaneously in parallel (using threading), the results should be > re-sorted and returned. For example I have a query: 'select * from > table where parent_clname = 'parent' order by name limit 10'. Now > this query has to be sent to 10 servers, and the maximum number of > results would be 100. Now this 100 result set has to be re-sorted, > out of which 90 has to be discarded, and the 10 has to be returned. > > Does such a solution exist now. To me this appears to be in entirety > of what should constitute a database cluster. Only the search needs > to be done on all the servers simultaneously at the low level. Once > you get the results, the writing can be determined by the upper level > logic (which can even be in a scripting language). But the search > across many servers has to be done using proper threading, and the > re-sorting also needs to be done fast. This is typically handled by the application layer, not a standard client. Mostly because every situation is different, you may have 10 servers and need 10 rows of results, others may need something entirely different. This isn't really a "cluster" either. In a clustered environment you would send the one query to any of the 10 servers and it would return the proper results. But like I said this type of application is fairly trivial to write in most scripting or higher level languages. --------------------------------- Frank Wiles http://www.wiles.org --------------------------------- From pgsql-performance-owner@postgresql.org Fri Aug 26 13:12:48 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CCF36D79D2 for ; Fri, 26 Aug 2005 13:11:22 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 29863-02 for ; Fri, 26 Aug 2005 16:11:21 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id AD0D7D79BC for ; Fri, 26 Aug 2005 13:11:19 -0300 (ADT) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: Sending a select to multiple servers. Date: Fri, 26 Aug 2005 12:11:18 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3417DD1F8@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Sending a select to multiple servers. Thread-Index: AcWqVXd3ATuMkienQpK/MSZ5pFjdQAAAmLrw From: "Merlin Moncure" To: Cc: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.056 required=5 tests=[AWL=0.006, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/446 X-Sequence-Number: 14193 > Does such a solution exist now. To me this appears to be in entirety of > what should constitute a database cluster. Only the search needs to be > done on all the servers simultaneously at the low level. Once you get the > results, the writing can be determined by the upper level logic (which can > even be in a scripting language). But the search across many servers has > to be done using proper threading, and the re-sorting also needs to be > done fast. Well the fastest way would be to write a libpq wrapper, personally I would choose C++ for extreme performance. STL bring super fast sorting to the table and will make dealing with ExecParams/ExecPrepared a little bit easier. To make available from scripting languages you need to make C wrappers for interface functions and build in a shared library. You could use any of a number of high level scripting languages but performance will not be as good. YMMV. Antother interesting take on this problem would be to use dblink contrib. module. Check that out and see if it can meet your needs. Merlin From pgsql-performance-owner@postgresql.org Fri Aug 26 13:28:08 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B9DDDD791C for ; Fri, 26 Aug 2005 13:28:06 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 32381-04 for ; Fri, 26 Aug 2005 16:28:04 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [67.100.216.10]) by svr1.postgresql.org (Postfix) with ESMTP id BC865D7745 for ; Fri, 26 Aug 2005 13:28:02 -0300 (ADT) Received: by flake.decibel.org (Postfix, from userid 1001) id 7133B1529A; Fri, 26 Aug 2005 11:28:02 -0500 (CDT) Date: Fri, 26 Aug 2005 11:28:02 -0500 From: "Jim C. Nasby" To: Tom Lane Cc: Mark Fox , pgsql-performance@postgresql.org Subject: Re: Performance indexing of a simple query Message-ID: <20050826162801.GF11282@pervasive.com> References: <16554.1124926920@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <16554.1124926920@sss.pgh.pa.us> X-Operating-System: FreeBSD 4.11-RELEASE-p10 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.012 required=5 tests=[AWL=0.012] X-Spam-Level: X-Archive-Number: 200508/447 X-Sequence-Number: 14194 On Wed, Aug 24, 2005 at 07:42:00PM -0400, Tom Lane wrote: > Mark Fox writes: > > The sort of queries I want to execute (among others) are like: > > SELECT * FROM jobs > > WHERE completion_time > SOMEDATE AND start_time < SOMEDATE; > > In plain english: All the jobs that were running at SOMEDATE. Uh, the plain english and the SQL don't match. That query will find every job that was NOT running at the time you said. > AFAIK there is no good way to do this with btree indexes; the problem > is that it's fundamentally a 2-dimensional query and btrees are > 1-dimensional. There are various hacks you can try if you're willing > to constrain the problem (eg, if you can assume some not-very-large > maximum on the running time of jobs) but in full generality btrees are > just the Wrong Thing. Ignoring the SQL and doing what the author actually wanted, wouldn't a bitmap combination of indexes work here? Or with an index on (start_time, completion_time), start an index scan at start_time = SOMEDATE and only include rows where completion_time < SOMEDATE. Of course if SOMEDATE is near the beginning of the table that wouldn't help. -- Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com Pervasive Software http://pervasive.com 512-569-9461 From pgsql-performance-owner@postgresql.org Fri Aug 26 15:12:54 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 02277D7738 for ; Fri, 26 Aug 2005 15:12:53 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 52406-10 for ; Fri, 26 Aug 2005 18:12:51 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 8B5EBD6FEF for ; Fri, 26 Aug 2005 13:42:45 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7QGgf4t009942; Fri, 26 Aug 2005 12:42:41 -0400 (EDT) To: "Jim C. Nasby" Cc: Mark Fox , pgsql-performance@postgresql.org Subject: Re: Performance indexing of a simple query In-reply-to: <20050826162801.GF11282@pervasive.com> References: <16554.1124926920@sss.pgh.pa.us> <20050826162801.GF11282@pervasive.com> Comments: In-reply-to "Jim C. Nasby" message dated "Fri, 26 Aug 2005 11:28:02 -0500" Date: Fri, 26 Aug 2005 12:42:41 -0400 Message-ID: <9941.1125074561@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/448 X-Sequence-Number: 14195 "Jim C. Nasby" writes: > Uh, the plain english and the SQL don't match. That query will find > every job that was NOT running at the time you said. No, I think it was right. But anyway it was just an example. > On Wed, Aug 24, 2005 at 07:42:00PM -0400, Tom Lane wrote: >> AFAIK there is no good way to do this with btree indexes; the problem >> is that it's fundamentally a 2-dimensional query and btrees are >> 1-dimensional. There are various hacks you can try if you're willing >> to constrain the problem (eg, if you can assume some not-very-large >> maximum on the running time of jobs) but in full generality btrees are >> just the Wrong Thing. > Ignoring the SQL and doing what the author actually wanted, wouldn't a > bitmap combination of indexes work here? > Or with an index on (start_time, completion_time), start an index scan > at start_time = SOMEDATE and only include rows where completion_time < > SOMEDATE. Of course if SOMEDATE is near the beginning of the table that > wouldn't help. The trouble with either of those is that you have to scan very large fractions of the index (if not indeed *all* of it) in order to get your answer; certainly you hit much more of the index than just the region containing matching rows. Btree just doesn't have a good way to answer this type of query. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Aug 26 15:58:24 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 286B8D7A57 for ; Fri, 26 Aug 2005 15:58:23 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 64213-09 for ; Fri, 26 Aug 2005 18:58:20 +0000 (GMT) Received: from sccrmhc11.comcast.net (sccrmhc11.comcast.net [204.127.202.55]) by svr1.postgresql.org (Postfix) with ESMTP id 080C2D799D for ; Fri, 26 Aug 2005 15:58:18 -0300 (ADT) Received: from jefftrout.com ([24.147.120.205]) by comcast.net (sccrmhc11) with SMTP id <2005082618581801100ljhn4e>; Fri, 26 Aug 2005 18:58:19 +0000 Received: (qmail 87348 invoked from network); 26 Aug 2005 18:59:14 -0000 Received: from c-24-147-120-205.hsd1.ma.comcast.net (HELO ?192.168.0.106?) (24.147.120.205) by 192.168.0.109 with SMTP; 26 Aug 2005 18:59:14 -0000 Mime-Version: 1.0 (Apple Message framework v733) Content-Transfer-Encoding: 7bit Message-Id: Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed To: postgres performance From: Jeff Trout Subject: OSX & Performance Date: Fri, 26 Aug 2005 14:58:17 -0400 X-Mailer: Apple Mail (2.733) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/449 X-Sequence-Number: 14196 Well folks, I've been trying to track down why this Athlon 2800 (2.1ghz) has been handing my 2.5ghz G5 its cake. I have a query that (makes no io - the dataset can live in ram easily) takes about 700ms on the athlon and about 10 seconds on the G5. Tracking ti down a bit timestamp_cmp_internal (The btree was made of a timestamp & and int) was taking a large amount of time - specifically all the calls it makes to isnan(x). 14.1% in __isnand (which is the libSystem function & guts, which according to the darwin source copies the double to memory and accesses it as 2 ints looking for a specific pattern). (For reference, the other top functions are _bt_checkkeys at 30%, FunctionCall2 at 15.8% , _bt_step at 9% and _bt_first at 7%) . Talking to some of the mac super guru's on irc they said the problem is how the Mach-O ABI works, basically you get kicked in the nuts for accessing global or static data (like those constants __isnand uses). (You can read http://www.unsanity.org/archives/000044.php for a touch of info on it). I think given the function-call-rich arch of PG may make its performance on OSX always lower than other counterparts. Especially things like that __isnand. I'm going to be doing a couple experiments: 1. making an inline version of isnan to see how that improves performance 2. Trying it out on linux ppc to see how it runs. It may be worth noting these in the docs or faq somewhere. Also, two things to note, one of which is quite important: On tiger (10.4) PG compiles with NO OPTIMIZATION. Probably a template file needs to be updated. Panther seems to compile with -O2 though. If you want to profile PG on Tiger do not use gprof - it seems to be broken. I get func call #s, but no timing data. Instead you can do something even better - compile PG normally and attach to it with Shark (Comes with the CHUD tools) and check out its profile. Quite slick actually :) I'll keep people updated on my progress, but I just wanted to get these issues out in the air. -- Jeff Trout http://www.jefftrout.com/ http://www.stuarthamm.net/ From pgsql-performance-owner@postgresql.org Fri Aug 26 16:02:33 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9BE3AD7C75 for ; Fri, 26 Aug 2005 16:02:06 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 67688-01 for ; Fri, 26 Aug 2005 19:02:02 +0000 (GMT) Received: from unicorn.rentec.com (unicorn.rentec.com [216.223.240.9]) by svr1.postgresql.org (Postfix) with ESMTP id 79ADAD7C19 for ; Fri, 26 Aug 2005 16:02:00 -0300 (ADT) Received: from wren.rentec.com (wren.rentec.com [192.5.35.106]) by unicorn.rentec.com (8.13.1/8.12.1) with ESMTP id j7QJ1tEv022223 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Fri, 26 Aug 2005 15:01:58 -0400 (EDT) X-Rentec: external Received: from [172.26.132.145] (hoopoe.rentec.com [172.26.132.145]) by wren.rentec.com (8.13.1/8.12.1) with ESMTP id j7QJ1tPg017724; Fri, 26 Aug 2005 15:01:55 -0400 (EDT) Message-ID: <430F6723.5050505@rentec.com> Date: Fri, 26 Aug 2005 15:01:55 -0400 From: Alan Stange Reply-To: stange@rentec.com Organization: Renaissance Technologies Corp. User-Agent: Mozilla Thunderbird 1.0+ (X11/20050712) MIME-Version: 1.0 To: Tom Lane Cc: pgsql-performance@postgresql.org Subject: Re: difference in plan between 8.0 and 8.1? References: <430F2AF3.1080704@rentec.com> <9248.1125069415@sss.pgh.pa.us> In-Reply-To: <9248.1125069415@sss.pgh.pa.us> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Logged: Logged by unicorn.rentec.com as j7QJ1tEv022223 at Fri Aug 26 15:01:58 2005 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.017 required=5 tests=[AWL=0.017] X-Spam-Level: X-Archive-Number: 200508/450 X-Sequence-Number: 14197 Tom Lane wrote: > Alan Stange writes: > >> Unique (cost=2717137.08..2771407.21 rows=10854026 width=8) >> -> Sort (cost=2717137.08..2744272.14 rows=10854026 width=8) >> Sort Key: timeseriesid >> -> Bitmap Heap Scan on tbltimeseries >> (cost=48714.09..1331000.42 rows=10854026 width=8) >> Recheck Cond: (timeseriesid > 0) >> -> Bitmap Index Scan on idx_timeseris >> (cost=0.00..48714.09 rows=10854026 width=0) >> Index Cond: (timeseriesid > 0) >> (7 rows) >> > > >> I'm hoping someone can explain the new query plan (as I'm not sure I >> understand what it is doing). >> > > The index scan is reading the index to find out which heap tuple IDs > (TIDs) the index says meet the condition. It returns a bitmap of the > tuple locations (actually, an array of per-page bitmaps). The heap > scan goes and fetches the tuples from the table, working in TID order > to avoid re-reading the same page many times, as can happen for ordinary > index scans. Since the result isn't sorted, we have to do a sort to get > it into the correct order for the Unique step. > > Because it avoids random access to the heap, this plan can be a lot > faster than a regular index scan. I'm not sure at all that 8.1 is > doing good relative cost estimation yet, though. It would be > interesting to see EXPLAIN ANALYZE results for both ways. (You can > use enable_bitmapscan and enable_indexscan to force the planner to pick > the plan it thinks is slower.) Just to be clear. The index is on the timeseriesid column. Also, We usually have the where clause with some non-zero number. Anyway, here's the basic query, with variations added on belowe: fiasco=# explain analyze select timeseriesid from tbltimeseries where timeseriesid > 0; QUERY PLAN ------------------------------------------------------------------------------------------------------------------------------------------------ Bitmap Heap Scan on tbltimeseries (cost=48906.82..1332935.19 rows=10905949 width=8) (actual time=16476.337..787480.979 rows=10907853 loops=1) Recheck Cond: (timeseriesid > 0) -> Bitmap Index Scan on idx_timeseris (cost=0.00..48906.82 rows=10905949 width=0) (actual time=16443.585..16443.585 rows=10907853 loops=1) Index Cond: (timeseriesid > 0) Total runtime: 791340.341 ms (5 rows) Now add the order: fiasco=# explain analyze select timeseriesid from tbltimeseries where timeseriesid > 0 order by timeseriesid; QUERY PLAN ------------------------------------------------------------------------------------------------------------------------------------------------------ Sort (cost=2726087.93..2753352.81 rows=10905949 width=8) (actual time=821090.666..826353.054 rows=10913868 loops=1) Sort Key: timeseriesid -> Bitmap Heap Scan on tbltimeseries (cost=48912.82..1332941.19 rows=10905949 width=8) (actual time=16353.921..757075.349 rows=10913868 loops=1) Recheck Cond: (timeseriesid > 0) -> Bitmap Index Scan on idx_timeseris (cost=0.00..48912.82 rows=10905949 width=0) (actual time=16335.239..16335.239 rows=10913868 loops=1) Index Cond: (timeseriesid > 0) Total runtime: 830829.145 ms (7 rows) and the distinct: fiasco=# explain analyze select distinct timeseriesid from tbltimeseries where timeseriesid > 0 order by timeseriesid; QUERY PLAN ------------------------------------------------------------------------------------------------------------------------------------------------------------ Unique (cost=2726087.93..2780617.68 rows=10905949 width=8) (actual time=816938.970..831119.423 rows=10913868 loops=1) -> Sort (cost=2726087.93..2753352.81 rows=10905949 width=8) (actual time=816938.967..822298.802 rows=10913868 loops=1) Sort Key: timeseriesid -> Bitmap Heap Scan on tbltimeseries (cost=48912.82..1332941.19 rows=10905949 width=8) (actual time=15866.736..752851.006 rows=10913868 loops=1) Recheck Cond: (timeseriesid > 0) -> Bitmap Index Scan on idx_timeseris (cost=0.00..48912.82 rows=10905949 width=0) (actual time=15852.652..15852.652 rows=10913868 loops=1) Index Cond: (timeseriesid > 0) Total runtime: 835558.312 ms (8 rows) Now the usual query from 8.0: fiasco=# set enable_bitmapscan=false; explain analyze select distinct timeseriesid from tbltimeseries where timeseriesid > 0 order by timeseriesid; SET QUERY PLAN -------------------------------------------------------------------------------------------------------------------------------------------------------------- Unique (cost=0.00..14971276.10 rows=10905949 width=8) (actual time=24.930..999645.638 rows=10913868 loops=1) -> Index Scan using idx_timeseris on tbltimeseries (cost=0.00..14944011.22 rows=10905949 width=8) (actual time=24.926..989117.882 rows=10913868 loops=1) Index Cond: (timeseriesid > 0) Total runtime: 1003549.067 ms (4 rows) And now a sequential scan of the table itself: fiasco=# set enable_bitmapscan=false; set enable_indexscan=false; explain analyze select distinct timeseriesid from tbltimeseries where timeseriesid > 0 order by timeseriesid; SET SET QUERY PLAN ------------------------------------------------------------------------------------------------------------------------------------------------------ Unique (cost=102677188.75..102731718.49 rows=10905949 width=8) (actual time=956783.989..971036.657 rows=10919883 loops=1) -> Sort (cost=102677188.75..102704453.62 rows=10905949 width=8) (actual time=956783.985..962115.616 rows=10919883 loops=1) Sort Key: timeseriesid -> Seq Scan on tbltimeseries (cost=100000000.00..101284042.00 rows=10905949 width=8) (actual time=7.314..893267.030 rows=10919883 loops=1) Filter: (timeseriesid > 0) Total runtime: 975393.678 ms (6 rows) For us, the query is best served by the index scan as the ordering comes for free and results can be streamed to a client immediately. So, while the whole query is a bit slower, the client can begin processing the results immediately. The client has three threads which stream in two sets of id's and emit delete statements in smaller batches. It can be done as one statement, but on our production system that statement can run for 10 hours and delete 20M rows...which conflicts with the vacuum process. This version can be throttled, stopped and restarted at any time and no work is lost compared to a single long running query. From pgsql-performance-owner@postgresql.org Fri Aug 26 17:15:29 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 09F8ED7CC7 for ; Fri, 26 Aug 2005 17:14:20 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83212-01 for ; Fri, 26 Aug 2005 20:14:18 +0000 (GMT) Received: from zproxy.gmail.com (zproxy.gmail.com [64.233.162.201]) by svr1.postgresql.org (Postfix) with ESMTP id C1342D7CC5 for ; Fri, 26 Aug 2005 17:14:17 -0300 (ADT) Received: by zproxy.gmail.com with SMTP id 8so371206nzo for ; Fri, 26 Aug 2005 13:14:18 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:mime-version:content-type:content-transfer-encoding:content-disposition; b=r2aY7OCC0toXKqqtLEzpZOItKZN991cIl7ZU/NdOkrAhE/+1zU3VrgCDWXtE8WwPoIHb20B8sfr6rRWwO7u7ZqEvi5M5HvWRfaQBFchUpY5tR4v6siKNVxFhdrL+97L2ZEtXOpLPRsh87AAMyVAesOoW2JodGXcNm2KcFJisPJY= Received: by 10.36.250.44 with SMTP id x44mr300895nzh; Fri, 26 Aug 2005 13:14:18 -0700 (PDT) Received: by 10.36.24.2 with HTTP; Fri, 26 Aug 2005 13:14:18 -0700 (PDT) Message-ID: <1d219a6f0508261314516fdb0@mail.gmail.com> Date: Fri, 26 Aug 2005 16:14:18 -0400 From: Chris Hoover To: pgsql-performance@postgresql.org Subject: How does the planner execute unions? Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/451 X-Sequence-Number: 14198 Hopefully a quick question. In 7.3.4, how does the planner execute a query with union alls in it? Does it execute the unions serially, or does it launch a "thread" for each union (or maybe something else entirely). Thanks, Chris Here is an explain from the view I'm thinking about, how does postgres run this query? hmd=3D# explain select count(1) from clmhdr where hdr_user_id =3D 'user_id'= ; QUERY PLAN ---------------------------------------------------------------------------= ------------------------------------------------------------- Aggregate (cost=3D42.48..42.48 rows=3D1 width=3D924) -> Subquery Scan clmhdr (cost=3D0.00..42.41 rows=3D30 width=3D924) -> Append (cost=3D0.00..42.41 rows=3D30 width=3D924) -> Subquery Scan "*SELECT* 1" (cost=3D0.00..7.07 rows=3D5 width=3D924) -> Index Scan using clmhdr_live_hdr_user_id_hdr_clm_status_idx on clmhdr_live=20 (cost=3D0.00..7.07 rows=3D5 width=3D924) Index Cond: (hdr_user_id =3D 'user_id'::character varying) -> Subquery Scan "*SELECT* 2" (cost=3D0.00..7.07 rows=3D5 width=3D924) -> Index Scan using clmhdr_2003_hdr_user_id_hdr_clm_status_idx on clmhdr_2003=20 (cost=3D0.00..7.07 rows=3D5 width=3D924) Index Cond: (hdr_user_id =3D 'user_id'::character varying) -> Subquery Scan "*SELECT* 3" (cost=3D0.00..7.07 rows=3D5 width=3D924) -> Index Scan using clmhdr_2004_hdr_user_id_hdr_clm_status_idx on clmhdr_2004=20 (cost=3D0.00..7.07 rows=3D5 width=3D924) Index Cond: (hdr_user_id =3D 'user_id'::character varying) -> Subquery Scan "*SELECT* 4" (cost=3D0.00..7.07 rows=3D5 width=3D924) -> Index Scan using clmhdr_2005_hdr_user_id_hdr_clm_status_idx on clmhdr_2005=20 (cost=3D0.00..7.07 rows=3D5 width=3D924) Index Cond: (hdr_user_id =3D 'user_id'::character varying) -> Subquery Scan "*SELECT* 5" (cost=3D0.00..7.07 rows=3D5 width=3D924) -> Index Scan using clmhdr_2006_hdr_user_id_hdr_clm_status_idx on clmhdr_2006=20 (cost=3D0.00..7.07 rows=3D5 width=3D924) Index Cond: (hdr_user_id =3D 'user_id'::character varying) -> Subquery Scan "*SELECT* 6" (cost=3D0.00..7.07 rows=3D5 width=3D924) -> Index Scan using clmhdr_2007_hdr_user_id_hdr_clm_status_idx on clmhdr_2007=20 (cost=3D0.00..7.07 rows=3D5 width=3D924) Index Cond: (hdr_user_id =3D 'user_id'::character varying) (21 rows) hmd=3D# From pgsql-performance-owner@postgresql.org Fri Aug 26 17:46:34 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id ECB61D6FEB for ; Fri, 26 Aug 2005 17:46:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 86282-07 for ; Fri, 26 Aug 2005 20:46:16 +0000 (GMT) Received: from wolff.to (wolff.to [66.93.197.194]) by svr1.postgresql.org (Postfix) with SMTP id CA87ED7CF0 for ; Fri, 26 Aug 2005 17:46:15 -0300 (ADT) Received: (qmail 11667 invoked by uid 500); 26 Aug 2005 20:45:54 -0000 Date: Fri, 26 Aug 2005 15:45:54 -0500 From: Bruno Wolff III To: Chris Hoover Cc: pgsql-performance@postgresql.org Subject: Re: How does the planner execute unions? Message-ID: <20050826204554.GB11078@wolff.to> Mail-Followup-To: Bruno Wolff III , Chris Hoover , pgsql-performance@postgresql.org References: <1d219a6f0508261314516fdb0@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1d219a6f0508261314516fdb0@mail.gmail.com> User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/452 X-Sequence-Number: 14199 On Fri, Aug 26, 2005 at 16:14:18 -0400, Chris Hoover wrote: > Hopefully a quick question. > > In 7.3.4, how does the planner execute a query with union alls in it? > > Does it execute the unions serially, or does it launch a "thread" for > each union (or maybe something else entirely). Postgres doesn't have parallel execution of parts of queries. So it is going to do one part followed by the other part. From pgsql-performance-owner@postgresql.org Fri Aug 26 18:19:19 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D91D5D7A18 for ; Fri, 26 Aug 2005 18:16:11 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 95746-03 for ; Fri, 26 Aug 2005 21:16:09 +0000 (GMT) Received: from olive.qinip.net (olive.qinip.net [62.100.30.40]) by svr1.postgresql.org (Postfix) with ESMTP id F19F4D79C5 for ; Fri, 26 Aug 2005 18:16:08 -0300 (ADT) Received: from [10.0.0.2] (h8441139206.dsl.speedlinq.nl [84.41.139.206]) by olive.qinip.net (Postfix) with ESMTP id 77DC918153; Fri, 26 Aug 2005 23:16:06 +0200 (MEST) Message-ID: <430F8699.3030400@tweakers.net> Date: Fri, 26 Aug 2005 23:16:09 +0200 From: Arjen van der Meijden User-Agent: Mozilla Thunderbird 1.0RC1 (Windows/20041201) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Richard Huxton Cc: pgsql-performance@postgresql.org Subject: Re: Inefficient queryplan for query with intersectable References: <430EF844.70905@tweakers.net> <430F1393.5040704@archonet.com> In-Reply-To: <430F1393.5040704@archonet.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/453 X-Sequence-Number: 14200 On 26-8-2005 15:05, Richard Huxton wrote: > Arjen van der Meijden wrote: > >> >> I left all the configuration-stuff to the defaults since changing >> values didn't seem to impact much. Apart from the buffers and >> effective cache, increasing those made the performance worse. > > > I've not looked at the rest of your problem in detail, but using the > default configuration values is certainly not going to help things. In > particular effective_cache is supposed to tell PG how much memory your > OS is using to cache data. > > Read this through and make sure your configuration settings are sane, > then it might be worthwhile looking in detail at this particular query. > http://www.powerpostgresql.com/PerfList Thanks for the advice. But as said, I tried such things. Adjusting shared buffers to 5000, 10000 or 15000 made minor improvements. But adjusting the effective_cache was indeed very dramatic, to make matters worse! Changing the random_page_cost to 2.0 also made it choose another plan, but still not the fast plan. The machine has 1GB of memory, so I tested for effective cache size 10000, 20000, 40000, 60000 and 80000. The "600ms"-plan I'm talking about will not come up with an effective cache set to 60000 or above and for the lower values there was no improvement in performance over that already very fast plan. As said, it chooses sequential scans or "the wrong index plans" over a perfectly good plan that is just not selected when the parameters are "too well tuned" or sequential scanning of the table is allowed. So I'm still looking for a way to get it to use the fast plan, instead of the slower ones that appear when I tend to tune the database... From pgsql-performance-owner@postgresql.org Fri Aug 26 19:04:33 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A3504D7D2A for ; Fri, 26 Aug 2005 19:04:31 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 05914-03 for ; Fri, 26 Aug 2005 22:04:27 +0000 (GMT) Received: from kartal.liqia.com (unknown [65.254.45.184]) by svr1.postgresql.org (Postfix) with ESMTP id 3B875D7D30 for ; Fri, 26 Aug 2005 19:04:24 -0300 (ADT) Received: from [81.215.232.82] (helo=[10.0.0.201]) by kartal.liqia.com with esmtpa (Exim 4.50 #1 (Debian)) id 1E8mJC-0007WZ-R2 for ; Sat, 27 Aug 2005 01:04:25 +0300 Subject: Weird performance drop after VACUUM From: =?ISO-8859-1?Q?=DCmit_=D6ztosun?= To: pgsql-performance@postgresql.org Content-Type: multipart/mixed; boundary="=-MRyV2tcUtXWirnwXOIJ6" Date: Sat, 27 Aug 2005 01:04:19 +0300 Message-Id: <1125093859.9329.30.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.2.1.1 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/454 X-Sequence-Number: 14201 --=-MRyV2tcUtXWirnwXOIJ6 Content-Type: text/plain Content-Transfer-Encoding: 7bit Hello, We are using PostgreSQL for our business application. Recently, during testing of our application with large volumes of data, we faced a weird problem. Our query performance dropped *dramatically* after "VACUUM FULL ANALYZE" command. We have encountered a similar problem listed on mailing list archives, but the submitter solved his problem by rewriting his query, which is unfortunatelly very hard for us. I am attaching two EXPLAIN ANALYZE outputs, first one is just before the VACUUM FULL ANALYZE command and the other is the one after. Also attached is the SQL query, which is simplified to clearify the problem. In the example query time increases from 1.8 second to > 4.0 secons. The difference for the complete query is much bigger, query time increases from 7.8 seconds to > 110 seconds. Any help is appreciated, we were unable to identify what causes the query planner to choose a different/poor performing plan. Notes: Our production platform is Ubuntu Linux Hoary on i386, PostgreSQL 8.0.3, compiled from sources. Same tests were carried on Windows XP Professional and PostgreSQL 8.0.1 with similar results. The queries use little IO, high CPU. The largest table involved in the sample query has about 10000 rows. Indexes are used intensively, some tables use > 4 indexes. Best regards, Umit Oztosun --=-MRyV2tcUtXWirnwXOIJ6 Content-Disposition: attachment; filename=query.sql Content-Type: text/x-sql; name=query.sql; charset=UTF-8 Content-Transfer-Encoding: 7bit SELECT * FROM ( SELECT COALESCE ( (SELECT COALESCE (sum(irskal.anamiktar), 0) * (SELECT birim.fiyat2 * (SELECT kur1 FROM sis_doviz_kuru kur WHERE birim._key_sis_doviz2 = kur._key_sis_doviz ORDER BY tarih desc LIMIT 1) FROM scf_stokkart_birimleri birim WHERE _key_scf_stokkart = stok._key AND anabirim = '1' ) FROM scf_irsaliye irs, scf_irsaliye_kalemi irskal WHERE irskal._key_kalemturu = stok._key AND irskal._key_scf_irsaliye = irs._key AND irs.karsifirma = 'KENDI' AND (irs.turu='MAI' OR irs.turu='KGI' OR irs.turu='PS' OR irs.turu='TS' OR irs.turu='KC' OR irs.turu='KCO') AND ( irs._key_sis_depo_dest = '$$$$0000003l$1$$' OR irs._key_sis_depo_dest = '$$$$00000048$1$$' OR irs._key_sis_depo_dest = '$$$$0000004b$1$$' OR irs._key_sis_depo_dest = '$$$$0000004d$1$$' ) AND ((irskal._key LIKE '0000%' OR irskal._key LIKE '0101%' OR irskal._key LIKE '$$%')) AND irs.tarih <= '2005-08-26' ), 0 ) as arti_fiili_irs_karsifirma, stok.* FROM scf_stokkart stok ) AS _SWT WHERE (_key LIKE '00%' OR _key LIKE '01%' OR _key LIKE '$$%') ORDER BY _key desc --=-MRyV2tcUtXWirnwXOIJ6 Content-Disposition: attachment; filename=before_vacuum.txt Content-Type: text/plain; name=before_vacuum.txt; charset=UTF-8 Content-Transfer-Encoding: 7bit Before VACUUM FULL ANALYZE - Short Query --------------------------------------- Sort (cost=9094.31..9094.40 rows=37 width=817) (actual time=1852.799..1877.738 rows=10000 loops=1) Sort Key: stok._key -> Seq Scan on scf_stokkart stok (cost=0.00..9093.34 rows=37 width=817) (actual time=8.670..1575.586 rows=10000 loops=1) Filter: (((_key)::text ~~ '00%'::text) OR ((_key)::text ~~ '01%'::text) OR ((_key)::text ~~ '$$%'::text)) SubPlan -> Aggregate (cost=237.29..237.29 rows=1 width=16) (actual time=0.136..0.138 rows=1 loops=10000) InitPlan -> Index Scan using scf_stokkart_birimleri_key_scf_stokkart_idx on scf_stokkart_birimleri birim (cost=0.00..209.59 rows=1 width=58) (actual time=0.088..0.093 rows=1 loops=10000) Index Cond: ((_key_scf_stokkart)::text = ($1)::text) Filter: (anabirim = '1'::bpchar) SubPlan -> Limit (cost=9.31..9.31 rows=1 width=17) (actual time=0.046..0.048 rows=1 loops=10000) -> Sort (cost=9.31..9.31 rows=2 width=17) (actual time=0.041..0.041 rows=1 loops=10000) Sort Key: tarih -> Index Scan using sis_doviz_kuru_key_sis_doviz_idx on sis_doviz_kuru kur (cost=0.00..9.30 rows=2 width=17) (actual time=0.018..0.029 rows=2 loops=10000) Index Cond: (($0)::text = (_key_sis_doviz)::text) -> Nested Loop (cost=0.00..27.69 rows=1 width=16) (actual time=0.033..0.033 rows=0 loops=10000) -> Index Scan using scf_irsaliye_kalemi_key_kalemturu_idx on scf_irsaliye_kalemi irskal (cost=0.00..21.75 rows=1 width=58) (actual time=0.017..0.020 rows=0 loops=10000) Index Cond: ((_key_kalemturu)::text = ($1)::text) Filter: (((_key)::text ~~ '0000%'::text) OR ((_key)::text ~~ '0101%'::text) OR ((_key)::text ~~ '$$%'::text)) -> Index Scan using scf_irsaliye_pkey on scf_irsaliye irs (cost=0.00..5.94 rows=1 width=42) (actual time=0.021..0.021 rows=0 loops=3000) Index Cond: (("outer"._key_scf_irsaliye)::text = (irs._key)::text) Filter: (((karsifirma)::text = 'KENDI'::text) AND (((turu)::text = 'MAI'::text) OR ((turu)::text = 'KGI'::text) OR ((turu)::text = 'PS'::text) OR ((turu)::text = 'TS'::text) OR ((turu)::text = 'KC'::text) OR ((turu)::text = 'KCO'::text)) AND (((_key_sis_depo_dest)::text = '$$$$0000003l$1$$'::text) OR ((_key_sis_depo_dest)::text = '$$$$00000048$1$$'::text) OR ((_key_sis_depo_dest)::text = '$$$$0000004b$1$$'::text) OR ((_key_sis_depo_dest)::text = '$$$$0000004d$1$$'::text)) AND (tarih <= '2005-08-26'::date)) Total runtime: 1899.533 ms --=-MRyV2tcUtXWirnwXOIJ6 Content-Disposition: attachment; filename=after_vacuum.txt Content-Type: text/plain; name=after_vacuum.txt; charset=UTF-8 Content-Transfer-Encoding: 7bit After VACUUM FULL ANALYZE - Short Query --------------------------------------- Index Scan Backward using scf_stokkart_pkey on scf_stokkart stok (cost=0.00..392045.63 rows=9998 width=166) (actual time=0.661..4431.568 rows=10000 loops=1) Filter: (((_key)::text ~~ '00%'::text) OR ((_key)::text ~~ '01%'::text) OR ((_key)::text ~~ '$$%'::text)) SubPlan -> Aggregate (cost=39.16..39.16 rows=1 width=10) (actual time=0.416..0.418 rows=1 loops=10000) InitPlan -> Index Scan using scf_stokkart_birimleri_key_scf_stokkart_idx on scf_stokkart_birimleri birim (cost=0.00..5.25 rows=2 width=28) (actual time=0.101..0.105 rows=1 loops=10000) Index Cond: ((_key_scf_stokkart)::text = ($1)::text) Filter: (anabirim = '1'::bpchar) SubPlan -> Limit (cost=1.08..1.09 rows=1 width=15) (actual time=0.048..0.050 rows=1 loops=10000) -> Sort (cost=1.08..1.09 rows=2 width=15) (actual time=0.043..0.043 rows=1 loops=10000) Sort Key: tarih -> Seq Scan on sis_doviz_kuru kur (cost=0.00..1.07 rows=2 width=15) (actual time=0.009..0.026 rows=2 loops=10000) Filter: (($0)::text = (_key_sis_doviz)::text) -> Nested Loop (cost=0.00..33.90 rows=1 width=10) (actual time=0.295..0.295 rows=0 loops=10000) -> Seq Scan on scf_irsaliye irs (cost=0.00..30.00 rows=1 width=20) (actual time=0.290..0.290 rows=0 loops=10000) Filter: (((karsifirma)::text = 'KENDI'::text) AND (((turu)::text = 'MAI'::text) OR ((turu)::text = 'KGI'::text) OR ((turu)::text = 'PS'::text) OR ((turu)::text = 'TS'::text) OR ((turu)::text = 'KC'::text) OR ((turu)::text = 'KCO'::text)) AND (((_key_sis_depo_dest)::text = '$$$$0000003l$1$$'::text) OR ((_key_sis_depo_dest)::text = '$$$$00000048$1$$'::text) OR ((_key_sis_depo_dest)::text = '$$$$0000004b$1$$'::text) OR ((_key_sis_depo_dest)::text = '$$$$0000004d$1$$'::text)) AND (tarih <= '2005-08-26'::date)) -> Index Scan using scf_irsaliye_kalemi_key_scf_irsaliye_idx on scf_irsaliye_kalemi irskal (cost=0.00..3.89 rows=1 width=30) (never executed) Index Cond: ((irskal._key_scf_irsaliye)::text = ("outer"._key)::text) Filter: (((_key_kalemturu)::text = ($1)::text) AND (((_key)::text ~~ '0000%'::text) OR ((_key)::text ~~ '0101%'::text) OR ((_key)::text ~~ '$$%'::text))) Total runtime: 4456.895 ms --=-MRyV2tcUtXWirnwXOIJ6-- From pgsql-performance-owner@postgresql.org Fri Aug 26 19:35:44 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id F3F58D7D2F for ; Fri, 26 Aug 2005 19:35:42 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 11841-08 for ; Fri, 26 Aug 2005 22:35:40 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.197]) by svr1.postgresql.org (Postfix) with ESMTP id 1B611D7D2B for ; Fri, 26 Aug 2005 19:35:39 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i21so295550wra for ; Fri, 26 Aug 2005 15:35:42 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:date:from:to:cc:subject:message-id:reply-to:references:mime-version:content-type:content-disposition:in-reply-to:user-agent; b=nl1D+2AeOgBzCkmkbbMvMQyFQScsF8O6c1VvG724bHpQxQjFFyte9/f50BgPdIhDxgb0ADXKclrs2NBt8x+wENPW+VZG6vLrOoBtz5aXWqzte7+QuFb5ShgyMauPKzT0uDSI64l3Z4+FcGcefmtrT24fgxu2iOyi/qi0Dqf8sv4= Received: by 10.54.32.62 with SMTP id f62mr1536730wrf; Fri, 26 Aug 2005 15:35:42 -0700 (PDT) Received: from localhost ( [61.246.59.42]) by mx.gmail.com with ESMTP id 27sm5524422wrl.2005.08.26.15.35.37; Fri, 26 Aug 2005 15:35:42 -0700 (PDT) Date: Sat, 27 Aug 2005 04:15:40 +0530 From: Ligesh To: Frank Wiles Cc: pgsql-performance@postgresql.org Subject: Re: Sending a select to multiple servers. Message-ID: <20050826224540.GA24476@lxlabs.com> Reply-To: gxlists@gmail.com References: <20050826152409.GA22060@lxlabs.com> <20050826110459.03cce4d9.frank@wiles.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20050826110459.03cce4d9.frank@wiles.org> User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[AWL=-0.000, RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/455 X-Sequence-Number: 14202 On Fri, Aug 26, 2005 at 11:04:59AM -0500, Frank Wiles wrote: > On Fri, 26 Aug 2005 20:54:09 +0530 > This is typically handled by the application layer, not a standard > client. Mostly because every situation is different, you may have > 10 servers and need 10 rows of results, others may need something > entirely different. > > This isn't really a "cluster" either. In a clustered environment > you would send the one query to any of the 10 servers and it would > return the proper results. > > But like I said this type of application is fairly trivial to write > in most scripting or higher level languages. > The cluster logic is sort of implemented by this client library. If you write this at higher level the scalability becomes an issue. For 10 servers it is alright. If you want to retrieve 100 rows, and there are 100 servers, you will have a total result set of 10,000, which should be re-sorted and the 9900 of them should be droped, and only the 100 should be returned. Anyway, what I want to know is, if there is such a functionality offered by some C library. Thanks. -- :: Ligesh :: http://ligesh.com From pgsql-performance-owner@postgresql.org Fri Aug 26 19:50:39 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 17CD4D7CD1 for ; Fri, 26 Aug 2005 19:41:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 13471-05 for ; Fri, 26 Aug 2005 22:41:22 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.204]) by svr1.postgresql.org (Postfix) with ESMTP id 06A90D736E for ; Fri, 26 Aug 2005 19:41:20 -0300 (ADT) Received: by wproxy.gmail.com with SMTP id i11so44827wra for ; Fri, 26 Aug 2005 15:41:21 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:date:from:to:cc:subject:message-id:reply-to:references:mime-version:content-type:content-disposition:in-reply-to:user-agent; b=t/gbpOpgY+8BuMOZ9ius3PSF8+sLSzR/UuEsecLttNe3+LjI0mSaN1IOzUGFvoT9x8OR935DS5Ohit8byYELAZNDB548HM520IFxJAbDK5or/1zYV8z09Vvd7Yn9cSJbD9e+kvmSg6Vpf4O4LLdgPHVVdR8bXcfthOkVRdw9iRg= Received: by 10.54.14.70 with SMTP id 70mr3994410wrn; Fri, 26 Aug 2005 15:41:21 -0700 (PDT) Received: from localhost ( [61.246.59.42]) by mx.gmail.com with ESMTP id 12sm15089wrl.2005.08.26.15.41.17; Fri, 26 Aug 2005 15:41:21 -0700 (PDT) Date: Sat, 27 Aug 2005 04:21:23 +0530 From: Ligesh To: Frank Wiles Cc: pgsql-performance@postgresql.org Subject: Re: Sending a select to multiple servers. Message-ID: <20050826225123.GB24476@lxlabs.com> Reply-To: gxlists@gmail.com References: <20050826152409.GA22060@lxlabs.com> <20050826110459.03cce4d9.frank@wiles.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20050826110459.03cce4d9.frank@wiles.org> User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/456 X-Sequence-Number: 14203 On Fri, Aug 26, 2005 at 11:04:59AM -0500, Frank Wiles wrote: > On Fri, 26 Aug 2005 20:54:09 +0530 > Ligesh wrote: > Mostly because every situation is different, you may have > 10 servers and need 10 rows of results, others may need something > entirely different. > No. I have say 'm' number of servers, and I need 'n' rows. To get the results, you need to run the query against all the 'm' servers, which will return 'm x n' results, then you have to re-sort it and drop the 'm x n - n' rows and return only the 'n'. So this is like retrieving the 'n' rows amongst ALL the servers, that satisfy your search criteria. Once you retrieve the data, you will know which server each row belongs to, and you can do the writes yourself at the higher level. Thanks. -- :: Ligesh :: http://ligesh.com From pgsql-performance-owner@postgresql.org Fri Aug 26 19:55:19 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 43902D7D5E for ; Fri, 26 Aug 2005 19:52:28 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 16560-03 for ; Fri, 26 Aug 2005 22:52:24 +0000 (GMT) Received: from web35212.mail.mud.yahoo.com (web35212.mail.mud.yahoo.com [66.163.179.91]) by svr1.postgresql.org (Postfix) with SMTP id DE5E8D7D50 for ; Fri, 26 Aug 2005 19:52:23 -0300 (ADT) Received: (qmail 27639 invoked by uid 60001); 26 Aug 2005 22:52:24 -0000 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Message-ID:Received:Date:From:Subject:To:In-Reply-To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=l+6JkIlfeYWsHytkmJ8V5JSIZErK3VS/NqPQeWyrVqWb0JFLfgEI2Pm93KO3OrgTE+mVXDQmsuz1t/lmtKjL+DHLnmh5dXWdz0FSl2DNtX8Xp6CFDfsGX3y4HT6WJWOgLkj9S/jzZ0lk7B5D4g76gNfWUKyn6lALvycHynXk7w0= ; Message-ID: <20050826225224.27637.qmail@web35212.mail.mud.yahoo.com> Received: from [65.218.233.130] by web35212.mail.mud.yahoo.com via HTTP; Fri, 26 Aug 2005 15:52:24 PDT Date: Fri, 26 Aug 2005 15:52:24 -0700 (PDT) From: asif ali Subject: Re: Weird performance drop after VACUUM To: pgsql-performance@postgresql.org In-Reply-To: <1125093859.9329.30.camel@localhost.localdomain> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.374 required=5 tests=[DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/457 X-Sequence-Number: 14204 Hi, I have the same issue. After doing "VACCUME ANALYZE" performance of the query dropped. Here is the query explain select * from conversion_table c where c.conversion_date BETWEEN '2005-06-07' and '2005-08-17' Before "VACCUME ANALYZE" "Index Scan using conversion_table_pk on keyword_conversion_table c (cost=0.00..18599.25 rows=4986 width=95)" " Index Cond: ((conversion_date >= '2005-06-07'::date) AND (conversion_date <= '2005-08-17'::date))" After "VACCUME ANALYZE" "Seq Scan on conversion_table c (cost=0.00..29990.83 rows=1094820 width=66)" " Filter: ((conversion_date >= '2005-06-07'::date) AND (conversion_date <= '2005-08-17'::date))" I dont know why system is doing "Seq scan" now. Thanks asif ali --- �mit �ztosun wrote: > Hello, > > We are using PostgreSQL for our business > application. Recently, during > testing of our application with large volumes of > data, we faced a weird > problem. Our query performance dropped > *dramatically* after "VACUUM FULL > ANALYZE" command. We have encountered a similar > problem listed on > mailing list archives, but the submitter solved his > problem by rewriting > his query, which is unfortunatelly very hard for us. > > I am attaching two EXPLAIN ANALYZE outputs, first > one is just before the > VACUUM FULL ANALYZE command and the other is the one > after. Also > attached is the SQL query, which is simplified to > clearify the problem. > In the example query time increases from 1.8 second > to > 4.0 secons. The > difference for the complete query is much bigger, > query time increases > from 7.8 seconds to > 110 seconds. > > Any help is appreciated, we were unable to identify > what causes the > query planner to choose a different/poor performing > plan. > > Notes: > Our production platform is Ubuntu Linux Hoary on > i386, PostgreSQL 8.0.3, > compiled from sources. Same tests were carried on > Windows XP > Professional and PostgreSQL 8.0.1 with similar > results. The queries use > little IO, high CPU. The largest table involved in > the sample query has > about 10000 rows. Indexes are used intensively, some > tables use > 4 > indexes. > > Best regards, > Umit Oztosun > > > SELECT * FROM ( > SELECT > COALESCE ( > (SELECT COALESCE (sum(irskal.anamiktar), > 0) > * (SELECT > birim.fiyat2 * (SELECT kur1 > FROM > sis_doviz_kuru kur > WHERE > birim._key_sis_doviz2 = kur._key_sis_doviz > ORDER BY tarih > desc > LIMIT 1) > FROM scf_stokkart_birimleri > birim > WHERE _key_scf_stokkart = > stok._key > AND anabirim = '1' > ) > FROM scf_irsaliye irs, > scf_irsaliye_kalemi irskal > WHERE irskal._key_kalemturu = > stok._key > AND irskal._key_scf_irsaliye = > irs._key > AND irs.karsifirma = 'KENDI' > AND (irs.turu='MAI' OR > irs.turu='KGI' OR irs.turu='PS' OR irs.turu='TS' OR > irs.turu='KC' OR irs.turu='KCO') > AND ( irs._key_sis_depo_dest = > '$$$$0000003l$1$$' OR irs._key_sis_depo_dest = > '$$$$00000048$1$$' OR irs._key_sis_depo_dest = > '$$$$0000004b$1$$' OR irs._key_sis_depo_dest = > '$$$$0000004d$1$$' ) > AND ((irskal._key LIKE '0000%' OR > irskal._key LIKE '0101%' OR irskal._key LIKE '$$%')) > AND irs.tarih <= '2005-08-26' > ), 0 > ) as arti_fiili_irs_karsifirma, > stok.* > FROM scf_stokkart stok > ) AS _SWT WHERE (_key LIKE '00%' OR _key LIKE '01%' > OR _key LIKE '$$%') ORDER BY _key desc > > Before VACUUM FULL ANALYZE - Short Query > --------------------------------------- > Sort (cost=9094.31..9094.40 rows=37 width=817) > (actual time=1852.799..1877.738 rows=10000 loops=1) > Sort Key: stok._key > -> Seq Scan on scf_stokkart stok > (cost=0.00..9093.34 rows=37 width=817) (actual > time=8.670..1575.586 rows=10000 loops=1) > Filter: (((_key)::text ~~ '00%'::text) OR > ((_key)::text ~~ '01%'::text) OR ((_key)::text ~~ > '$$%'::text)) > SubPlan > -> Aggregate (cost=237.29..237.29 rows=1 > width=16) (actual time=0.136..0.138 rows=1 > loops=10000) > InitPlan > -> Index Scan using > scf_stokkart_birimleri_key_scf_stokkart_idx on > scf_stokkart_birimleri birim (cost=0.00..209.59 > rows=1 width=58) (actual time=0.088..0.093 rows=1 > loops=10000) > Index Cond: > ((_key_scf_stokkart)::text = ($1)::text) > Filter: (anabirim = > '1'::bpchar) > SubPlan > -> Limit > (cost=9.31..9.31 rows=1 width=17) (actual > time=0.046..0.048 rows=1 loops=10000) > -> Sort > (cost=9.31..9.31 rows=2 width=17) (actual > time=0.041..0.041 rows=1 loops=10000) > Sort Key: > tarih > -> Index Scan > using sis_doviz_kuru_key_sis_doviz_idx on > sis_doviz_kuru kur (cost=0.00..9.30 rows=2 > width=17) (actual time=0.018..0.029 rows=2 > loops=10000) > Index > Cond: (($0)::text = (_key_sis_doviz)::text) > -> Nested Loop (cost=0.00..27.69 > rows=1 width=16) (actual time=0.033..0.033 rows=0 > loops=10000) > -> Index Scan using > scf_irsaliye_kalemi_key_kalemturu_idx on > scf_irsaliye_kalemi irskal (cost=0.00..21.75 rows=1 > width=58) (actual time=0.017..0.020 rows=0 > loops=10000) > Index Cond: > ((_key_kalemturu)::text = ($1)::text) > Filter: (((_key)::text > ~~ '0000%'::text) OR ((_key)::text ~~ '0101%'::text) > OR ((_key)::text ~~ '$$%'::text)) > -> Index Scan using > scf_irsaliye_pkey on scf_irsaliye irs > (cost=0.00..5.94 rows=1 width=42) (actual > time=0.021..0.021 rows=0 loops=3000) > Index Cond: > (("outer"._key_scf_irsaliye)::text = > (irs._key)::text) > Filter: > (((karsifirma)::text = 'KENDI'::text) AND > (((turu)::text = 'MAI'::text) OR ((turu)::text = > 'KGI'::text) OR ((turu)::text = 'PS'::text) OR > ((turu)::text = 'TS'::text) OR ((turu)::text = > 'KC'::text) OR ((turu)::text = 'KCO'::text)) AND > (((_key_sis_depo_dest)::text = > '$$$$0000003l$1$$'::text) OR > ((_key_sis_depo_dest)::text = > '$$$$00000048$1$$'::text) OR > ((_key_sis_depo_dest)::text = > '$$$$0000004b$1$$'::text) OR > ((_key_sis_depo_dest)::text = > '$$$$0000004d$1$$'::text)) AND (tarih <= > '2005-08-26'::date)) > Total runtime: 1899.533 ms > > After VACUUM FULL ANALYZE - Short Query > --------------------------------------- > Index Scan Backward using scf_stokkart_pkey on > scf_stokkart stok (cost=0.00..392045.63 rows=9998 > width=166) (actual time=0.661..4431.568 rows=10000 > loops=1) > Filter: (((_key)::text ~~ '00%'::text) OR > ((_key)::text ~~ '01%'::text) OR ((_key)::text ~~ > '$$%'::text)) > SubPlan > -> Aggregate (cost=39.16..39.16 rows=1 > width=10) (actual time=0.416..0.418 rows=1 > loops=10000) > InitPlan > -> Index Scan using > scf_stokkart_birimleri_key_scf_stokkart_idx on > scf_stokkart_birimleri birim (cost=0.00..5.25 > rows=2 width=28) (actual time=0.101..0.105 rows=1 > loops=10000) > Index Cond: > ((_key_scf_stokkart)::text = ($1)::text) > Filter: (anabirim = '1'::bpchar) > SubPlan > -> Limit (cost=1.08..1.09 > rows=1 width=15) (actual time=0.048..0.050 rows=1 > loops=10000) > -> Sort (cost=1.08..1.09 > rows=2 width=15) (actual time=0.043..0.043 rows=1 > loops=10000) > Sort Key: tarih > -> Seq Scan on > sis_doviz_kuru kur (cost=0.00..1.07 rows=2 > width=15) (actual time=0.009..0.026 rows=2 > loops=10000) > Filter: > (($0)::text = (_key_sis_doviz)::text) > -> Nested Loop (cost=0.00..33.90 rows=1 > width=10) (actual time=0.295..0.295 rows=0 > loops=10000) > -> Seq Scan on scf_irsaliye irs > (cost=0.00..30.00 rows=1 width=20) (actual > time=0.290..0.290 rows=0 loops=10000) > Filter: (((karsifirma)::text = > 'KENDI'::text) AND (((turu)::text = 'MAI'::text) OR > ((turu)::text = 'KGI'::text) OR ((turu)::text = > 'PS'::text) OR ((turu)::text = 'TS'::text) OR > ((turu)::text = 'KC'::text) OR ((turu)::text = > 'KCO'::text)) AND (((_key_sis_depo_dest)::text = > '$$$$0000003l$1$$'::text) OR > ((_key_sis_depo_dest)::text = > '$$$$00000048$1$$'::text) OR > ((_key_sis_depo_dest)::text = > '$$$$0000004b$1$$'::text) OR > ((_key_sis_depo_dest)::text = > '$$$$0000004d$1$$'::text)) AND (tarih <= > '2005-08-26'::date)) > -> Index Scan using > scf_irsaliye_kalemi_key_scf_irsaliye_idx on > scf_irsaliye_kalemi irskal (cost=0.00..3.89 rows=1 > width=30) (never executed) > Index Cond: > ((irskal._key_scf_irsaliye)::text = > ("outer"._key)::text) > Filter: > (((_key_kalemturu)::text = ($1)::text) AND > (((_key)::text ~~ '0000%'::text) OR ((_key)::text ~~ > '0101%'::text) OR ((_key)::text ~~ '$$%'::text))) > Total runtime: 4456.895 ms > > > ---------------------------(end of > broadcast)--------------------------- > TIP 1: if posting/reading through Usenet, please > send an appropriate > subscribe-nomail command to > majordomo@postgresql.org so that your > message can get through to the mailing list > cleanly > __________________________________ Yahoo! Mail for Mobile Take Yahoo! Mail with you! Check email on your mobile phone. http://mobile.yahoo.com/learn/mail From pgsql-performance-owner@postgresql.org Fri Aug 26 20:07:08 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C8214D7D3B for ; Fri, 26 Aug 2005 19:56:18 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 15707-09 for ; Fri, 26 Aug 2005 22:56:17 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id B994AD7B46 for ; Fri, 26 Aug 2005 19:56:16 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7QMuGjX026437; Fri, 26 Aug 2005 18:56:16 -0400 (EDT) To: Arjen van der Meijden Cc: Richard Huxton , pgsql-performance@postgresql.org Subject: Re: Inefficient queryplan for query with intersectable In-reply-to: <430F8699.3030400@tweakers.net> References: <430EF844.70905@tweakers.net> <430F1393.5040704@archonet.com> <430F8699.3030400@tweakers.net> Comments: In-reply-to Arjen van der Meijden message dated "Fri, 26 Aug 2005 23:16:09 +0200" Date: Fri, 26 Aug 2005 18:56:16 -0400 Message-ID: <26436.1125096976@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/458 X-Sequence-Number: 14205 Arjen van der Meijden writes: > As said, it chooses sequential scans or "the wrong index plans" over a > perfectly good plan that is just not selected when the parameters are > "too well tuned" or sequential scanning of the table is allowed. I think some part of the problem comes from using inconsistent datatypes. For instance, it seems very odd that the thing is not using a hash or something to handle t_0.Cat2 IN (SELECT 545 UNION SELECT ID FROM cat WHERE ParentID = 545) seeing that it correctly guesses there are only going to be about 8 rows in the union. Part of the reason is that cat2 is smallint, whereas the output of the union must be at least int, maybe wider depending on the datatype of cat.id (which you did not show us); so the comparison isn't hashable. Even a smallint vs int comparison would be mergejoinable, though, so I'm really wondering what cat.id is. Another big part of the problem comes from poor result size estimation. I'm not sure you can eliminate that entirely given the multiple conditions on different columns (which'd require cross-column statistics to really do well, which we do not have). But you could avoid constructs like WHERE ... t_1.recordtimestamp >= (SELECT max_date - 60 FROM last_dates WHERE table_name = 'pricetracker') The planner is basically going to throw up its hands and make a default guess on the selectivity of this; it's not smart enough to decide that the sub-select probably represents a constant. What I'd do with this is to define a function marked STABLE for the sub-select result, perhaps something like create function get_last_date(tabname text, offsetdays int) returns timestamp as $$ SELECT max_date - $2 FROM last_dates WHERE table_name = $1 $$ language sql strict stable; (I'm guessing as to datatypes and the amount of parameterization you need.) Then write the query like WHERE ... t_1.recordtimestamp >= get_last_date('pricetracker', 60) In this formulation the planner will be able to make a reasonable guess about how many rows will match ... at least if your statistics are up to date ... regards, tom lane From pgsql-performance-owner@postgresql.org Fri Aug 26 20:15:01 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E5ECAD7CF3 for ; Fri, 26 Aug 2005 20:14:02 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 21118-05 for ; Fri, 26 Aug 2005 23:14:01 +0000 (GMT) Received: from zhonka1.zhonka.net (zhonka1.zhonka.net [66.228.195.5]) by svr1.postgresql.org (Postfix) with ESMTP id 16071D7CD9 for ; Fri, 26 Aug 2005 20:14:00 -0300 (ADT) Received: from wolf.pjkh.com ([66.228.196.74]) by zhonka1.zhonka.net (Post.Office MTA v3.5.3 release 223 ID# 0-58414U4500L450S0V35) with ESMTP id net; Fri, 26 Aug 2005 16:14:03 -0700 Received: from localhost (localhost [127.0.0.1]) by wolf.pjkh.com (Postfix) with ESMTP id 8CDA8584D; Fri, 26 Aug 2005 16:13:56 -0700 (PDT) Received: from wolf.pjkh.com ([127.0.0.1]) by localhost (wolf.pjkh.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 21985-09; Fri, 26 Aug 2005 16:13:56 -0700 (PDT) Received: by wolf.pjkh.com (Postfix, from userid 1000) id 550AF584B; Fri, 26 Aug 2005 16:13:56 -0700 (PDT) Received: from localhost (localhost [127.0.0.1]) by wolf.pjkh.com (Postfix) with ESMTP id 50FD05773; Fri, 26 Aug 2005 16:13:56 -0700 (PDT) Date: Fri, 26 Aug 2005 16:13:56 -0700 (PDT) From: Philip Hallstrom To: asif ali Cc: pgsql-performance@postgresql.org Subject: Re: Weird performance drop after VACUUM In-Reply-To: <20050826225224.27637.qmail@web35212.mail.mud.yahoo.com> Message-ID: <20050826161119.P22700@wolf.pjkh.com> References: <20050826225224.27637.qmail@web35212.mail.mud.yahoo.com> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed X-Virus-Scanned: by amavisd-new at pjkh.com X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=-0.056 required=5 tests=[AWL=-0.056] X-Spam-Level: X-Archive-Number: 200508/459 X-Sequence-Number: 14206 > Hi, > I have the same issue. After doing "VACCUME ANALYZE" > performance of the query dropped. > > Here is the query > explain select * from conversion_table c where > c.conversion_date BETWEEN '2005-06-07' and > '2005-08-17' > > Before "VACCUME ANALYZE" > > "Index Scan using conversion_table_pk on > keyword_conversion_table c (cost=0.00..18599.25 > rows=4986 width=95)" > " Index Cond: ((conversion_date >= > '2005-06-07'::date) AND (conversion_date <= > '2005-08-17'::date))" > > > After "VACCUME ANALYZE" > > > "Seq Scan on conversion_table c (cost=0.00..29990.83 > rows=1094820 width=66)" > " Filter: ((conversion_date >= '2005-06-07'::date) > AND (conversion_date <= '2005-08-17'::date))" > > > I dont know why system is doing "Seq scan" now. I could be wrong as I'm definitely no expert on reading the output of EXPLAIN, but it seems to say that prior to VACUUM it was expecting to retrieve 4986 rows and afterwards expecting to retrieve 1094820 rows. Which is a pretty big difference. So maybe the statistics were just really really off prior to vacuuming and once it did vacuum it realized there would be a lot more matches and since there were a lot more matches the planner decided to do a seq scan since it would be quicker overall... Maybe? Seems I've heard Tom Lane say something to that affect, although much more eloquently :-) -philip From pgsql-performance-owner@postgresql.org Fri Aug 26 20:28:18 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 13C73D7686 for ; Fri, 26 Aug 2005 20:26:46 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 22405-09 for ; Fri, 26 Aug 2005 23:26:43 +0000 (GMT) Received: from tigger.fuhr.org (tigger.fuhr.org [63.214.45.158]) by svr1.postgresql.org (Postfix) with ESMTP id 3E944D736E for ; Fri, 26 Aug 2005 20:26:41 -0300 (ADT) Received: from winnie.fuhr.org (winnie.fuhr.org [10.1.0.1]) by tigger.fuhr.org (8.13.3/8.13.3) with ESMTP id j7QNQgCm019593 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK); Fri, 26 Aug 2005 17:26:44 -0600 (MDT) (envelope-from mfuhr@winnie.fuhr.org) Received: from winnie.fuhr.org (localhost [127.0.0.1]) by winnie.fuhr.org (8.13.3/8.13.3) with ESMTP id j7QNQfrB019672; Fri, 26 Aug 2005 17:26:41 -0600 (MDT) (envelope-from mfuhr@winnie.fuhr.org) Received: (from mfuhr@localhost) by winnie.fuhr.org (8.13.3/8.13.3/Submit) id j7QNQfUM019671; Fri, 26 Aug 2005 17:26:41 -0600 (MDT) (envelope-from mfuhr) Date: Fri, 26 Aug 2005 17:26:41 -0600 From: Michael Fuhr To: asif ali Cc: pgsql-performance@postgresql.org Subject: Re: Weird performance drop after VACUUM Message-ID: <20050826232641.GA19583@winnie.fuhr.org> References: <1125093859.9329.30.camel@localhost.localdomain> <20050826225224.27637.qmail@web35212.mail.mud.yahoo.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20050826225224.27637.qmail@web35212.mail.mud.yahoo.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 required=5 tests=[AWL=0.005] X-Spam-Level: X-Archive-Number: 200508/460 X-Sequence-Number: 14207 On Fri, Aug 26, 2005 at 03:52:24PM -0700, asif ali wrote: > I have the same issue. After doing "VACCUME ANALYZE" > performance of the query dropped. Your EXPLAIN output doesn't show the actual query times -- could you post the EXPLAIN ANALYZE output? That'll also show how accurate the planner's row count estimates are. > Before "VACCUME ANALYZE" > > "Index Scan using conversion_table_pk on > keyword_conversion_table c (cost=0.00..18599.25 > rows=4986 width=95)" > " Index Cond: ((conversion_date >= > '2005-06-07'::date) AND (conversion_date <= > '2005-08-17'::date))" > > After "VACCUME ANALYZE" > > "Seq Scan on conversion_table c (cost=0.00..29990.83 > rows=1094820 width=66)" > " Filter: ((conversion_date >= '2005-06-07'::date) > AND (conversion_date <= '2005-08-17'::date))" > > I dont know why system is doing "Seq scan" now. Notice the row count estimates: 4986 in the "before" query and 1094820 in the "after" query. In the latter, the planner thinks it has to fetch so much of the table that a sequential scan would be faster than an index scan. You can see whether that guess is correct by disabling enable_seqscan to force an index scan. It might be useful to see the output of the following: SET enable_seqscan TO on; SET enable_indexscan TO off; EXPLAIN ANALYZE SELECT ...; SET enable_seqscan TO off; SET enable_indexscan TO on; EXPLAIN ANALYZE SELECT ...; You might also experiment with planner variables like effective_cache_size and random_page_cost to see how changing them affects the query plan. However, be careful of tuning the system based on one query: make sure adjustments result in reasonable plans for many different queries. -- Michael Fuhr From pgsql-performance-owner@postgresql.org Fri Aug 26 20:33:25 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 92B71D7CF5 for ; Fri, 26 Aug 2005 20:31:12 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 23093-09 for ; Fri, 26 Aug 2005 23:31:11 +0000 (GMT) Received: from linda-4.paradise.net.nz (bm-4a.paradise.net.nz [202.0.58.23]) by svr1.postgresql.org (Postfix) with ESMTP id 05398D7CD9 for ; Fri, 26 Aug 2005 20:31:10 -0300 (ADT) Received: from smtp-1.paradise.net.nz (smtp-1a.paradise.net.nz [202.0.32.194]) by linda-4.paradise.net.nz (Paradise.net.nz) with ESMTP id <0ILU00KG2RWEZ8@linda-4.paradise.net.nz> for pgsql-performance@postgresql.org; Sat, 27 Aug 2005 11:29:02 +1200 (NZST) Received: from [192.168.1.11] (218-101-14-104.paradise.net.nz [218.101.14.104]) by smtp-1.paradise.net.nz (Postfix) with ESMTP id 5F34E82B87; Sat, 27 Aug 2005 11:29:02 +1200 (NZST) Date: Sat, 27 Aug 2005 11:29:01 +1200 From: Mark Kirkwood Subject: Re: Limit + group + join In-reply-to: <9501.1125071193@sss.pgh.pa.us> To: Tom Lane Cc: Tobias Brox , pgsql-performance@postgresql.org Message-id: <430FA5BD.2070407@paradise.net.nz> MIME-version: 1.0 Content-type: text/plain; format=flowed; charset=ISO-8859-1 Content-transfer-encoding: 7bit X-Accept-Language: en-us, en User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050726) References: <20050826002709.GK10328@tobias.lan> <430E85ED.60109@paradise.net.nz> <9501.1125071193@sss.pgh.pa.us> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.147 required=5 tests=[AWL=0.147] X-Spam-Level: X-Archive-Number: 200508/462 X-Sequence-Number: 14209 Tom Lane wrote: > Mark Kirkwood writes: > >>What is interesting is why this plan is being rejected... > > > Which PG version are you using exactly? That mistake looks like an > artifact of the 8.0 "fuzzy plan cost" patch, which we fixed recently: > http://archives.postgresql.org/pgsql-committers/2005-07/msg00474.php > Right on - 8.0.3 (I might look at how CVS tip handles this, could be interesting). > But Tobias wasn't happy with 7.4 either, so I'm not sure that the fuzzy > cost issue explains his results. > > As far as the "desc" point goes, the problem is that mergejoins aren't > capable of dealing with backward sort order, so a merge plan isn't > considered for that case (or at least, it would have to have a sort > after it, which pretty much defeats the point for a fast-start plan). > I have some ideas about fixing this but it won't happen before 8.2. That doesn't explain why the nested loop is being kicked tho', or have I missed something obvious? - it's been known to happen :-)... Cheers Mark From pgsql-performance-owner@postgresql.org Fri Aug 26 20:32:57 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4B5A8D7A5E for ; Fri, 26 Aug 2005 20:31:51 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 24101-04 for ; Fri, 26 Aug 2005 23:31:50 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 993EFD7705 for ; Fri, 26 Aug 2005 20:31:49 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7QNVpGB026746; Fri, 26 Aug 2005 19:31:51 -0400 (EDT) To: =?ISO-8859-1?Q?=DCmit_=D6ztosun?= Cc: pgsql-performance@postgresql.org Subject: Re: Weird performance drop after VACUUM In-reply-to: <1125093859.9329.30.camel@localhost.localdomain> References: <1125093859.9329.30.camel@localhost.localdomain> Comments: In-reply-to =?ISO-8859-1?Q?=DCmit_=D6ztosun?= message dated "Sat, 27 Aug 2005 01:04:19 +0300" Date: Fri, 26 Aug 2005 19:31:51 -0400 Message-ID: <26745.1125099111@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/461 X-Sequence-Number: 14208 =?ISO-8859-1?Q?=DCmit_=D6ztosun?= writes: > We are using PostgreSQL for our business application. Recently, during > testing of our application with large volumes of data, we faced a weird > problem. Our query performance dropped *dramatically* after "VACUUM FULL > ANALYZE" command. I think the problem is that the planner is underestimating the cost of evaluating this complicated filter condition: > -> Seq Scan on scf_irsaliye irs (cost=0.00..30.00 rows=1 width=20) (actual time=0.290..0.290 rows=0 loops=10000) > Filter: (((karsifirma)::text = 'KENDI'::text) AND (((turu)::text = 'MAI'::text) OR ((turu)::text = 'KGI'::text) OR ((turu)::text = 'PS'::text) OR ((turu)::text = 'TS'::text) OR ((turu)::text = 'KC'::text) OR ((turu)::text = 'KCO'::text)) AND (((_key_sis_depo_dest)::text = '$$$$0000003l$1$$'::text) OR ((_key_sis_depo_dest)::text = '$$$$00000048$1$$'::text) OR ((_key_sis_depo_dest)::text = '$$$$0000004b$1$$'::text) OR ((_key_sis_depo_dest)::text = '$$$$0000004d$1$$'::text)) AND (tarih <= '2005-08-26'::date)) While you could attack that by raising the cpu_operator_cost parameter, it would also be worth inquiring *why* the condition is so expensive to evaluate. I am suspicious that you are running the database in a locale in which strcoll() is really slow. Can you run it in C locale instead, or do you really need locale-aware behavior? Can you switch to a different database encoding? (A single-byte encoding such as Latin1 might be faster than UTF8, for example.) Another possibility is to take a hard look at whether you can't simplify the filter condition, but that'd require more knowledge of your application than I have. Or you could just play with the order of the filter conditions ... for example, the date condition at the end is probably far cheaper to test than the text comparisons, so if that's fairly selective it'd be worth putting it first. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Aug 26 20:48:19 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3D852D7D75 for ; Fri, 26 Aug 2005 20:48:18 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 22863-10 for ; Fri, 26 Aug 2005 23:48:17 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 623C5D7D72 for ; Fri, 26 Aug 2005 20:48:15 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7QNm5Zx026875; Fri, 26 Aug 2005 19:48:05 -0400 (EDT) To: Mark Kirkwood Cc: Tobias Brox , pgsql-performance@postgresql.org Subject: Re: Limit + group + join In-reply-to: <430FA5BD.2070407@paradise.net.nz> References: <20050826002709.GK10328@tobias.lan> <430E85ED.60109@paradise.net.nz> <9501.1125071193@sss.pgh.pa.us> <430FA5BD.2070407@paradise.net.nz> Comments: In-reply-to Mark Kirkwood message dated "Sat, 27 Aug 2005 11:29:01 +1200" Date: Fri, 26 Aug 2005 19:48:05 -0400 Message-ID: <26874.1125100085@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/463 X-Sequence-Number: 14210 Mark Kirkwood writes: > That doesn't explain why the nested loop is being kicked tho', No, but I think the fuzzy-cost bug does. There are two different issues here. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Aug 26 21:10:53 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id EDDE7D7705 for ; Fri, 26 Aug 2005 21:10:51 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 31418-10 for ; Sat, 27 Aug 2005 00:10:47 +0000 (GMT) Received: from web35210.mail.mud.yahoo.com (web35210.mail.mud.yahoo.com [66.163.179.89]) by svr1.postgresql.org (Postfix) with SMTP id BBA3ED736E for ; Fri, 26 Aug 2005 21:10:45 -0300 (ADT) Received: (qmail 58961 invoked by uid 60001); 27 Aug 2005 00:10:49 -0000 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Message-ID:Received:Date:From:Subject:To:Cc:In-Reply-To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=P7dsyrW1AuZ5qo9mOCrfFeNwY6D2KAGORCVLijqlE6PtMA/xMq3bUKzpNz5ni2nEwdF+Zhok+ZsDNQe3UwjVxWzgtFl08TkSaoW75j+zrjr7o3YHsJ84PYmWpLtmOg7dsEzGJoy/FKw/PuztpReEZQWXWN3VY69cGeLkmWUntyM= ; Message-ID: <20050827001049.58959.qmail@web35210.mail.mud.yahoo.com> Received: from [65.218.233.130] by web35210.mail.mud.yahoo.com via HTTP; Fri, 26 Aug 2005 17:10:49 PDT Date: Fri, 26 Aug 2005 17:10:49 -0700 (PDT) From: asif ali Subject: Re: Weird performance drop after VACUUM To: Michael Fuhr Cc: pgsql-performance@postgresql.org In-Reply-To: <20050826232641.GA19583@winnie.fuhr.org> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.28 required=5 tests=[AWL=-0.093, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/464 X-Sequence-Number: 14211 Thanks Michael For your reply. Here is performance on the database on which i did VACUUM ANALYZE explain analyze select keyword_id ,sum(daily_impressions) as daily_impressions ,sum(daily_actions)as daily_actions from conversion_table c where c.conversion_date BETWEEN '2005-06-07' and '2005-08-17' group by keyword_Id "GroupAggregate (cost=195623.66..206672.52 rows=20132 width=16) (actual time=8205.283..10139.369 rows=55291 loops=1)" " -> Sort (cost=195623.66..198360.71 rows=1094820 width=16) (actual time=8205.114..9029.501 rows=863883 loops=1)" " Sort Key: keyword_id" " -> Seq Scan on keyword_conversion_table c (cost=0.00..29990.83 rows=1094820 width=16) (actual time=0.057..1422.319 rows=863883 loops=1)" " Filter: ((conversion_date >= '2005-06-07'::date) AND (conversion_date <= '2005-08-17'::date))" "Total runtime: 14683.617 ms" Now see if am changing the query and commenting one column. explain analyze select keyword_id ,sum(daily_impressions) as daily_impressions -- ,sum(daily_actions)as daily_actions from conversion_table c where c.conversion_date BETWEEN '2005-06-07' and '2005-08-17' group by keyword_Id "HashAggregate (cost=27373.51..27373.52 rows=2 width=16) (actual time=3030.386..3127.073 rows=55717 loops=1)" " -> Seq Scan on conversion_table c (cost=0.00..27336.12 rows=4986 width=16) (actual time=0.050..1357.164 rows=885493 loops=1)" " Filter: ((conversion_date >= '2005-06-07'::date) AND (conversion_date <= '2005-08-17'::date))" "Total runtime: 3159.162 ms" I noticed "GroupAggregate" changes to "HashAggregate" and performance from 14 sec to 3 sec. On the other hand I have another database which I did not do "VACUUM ANALYZE" working fine. explain analyze select keyword_id ,sum(daily_impressions) as daily_impressions ,sum(daily_actions)as daily_actions from conversion_table c where c.conversion_date BETWEEN '2005-06-07' and '2005-08-17' group by keyword_Id "HashAggregate (cost=27373.51..27373.52 rows=2 width=16) (actual time=3024.289..3120.324 rows=55717 loops=1)" " -> Seq Scan on conversion_table c (cost=0.00..27336.12 rows=4986 width=16) (actual time=0.047..1352.212 rows=885493 loops=1)" " Filter: ((conversion_date >= '2005-06-07'::date) AND (conversion_date <= '2005-08-17'::date))" "Total runtime: 3152.437 ms" I am new to postgres. Thanks in advance. asif ali --- Michael Fuhr wrote: > On Fri, Aug 26, 2005 at 03:52:24PM -0700, asif ali > wrote: > > I have the same issue. After doing "VACCUME > ANALYZE" > > performance of the query dropped. > > Your EXPLAIN output doesn't show the actual query > times -- could > you post the EXPLAIN ANALYZE output? That'll also > show how accurate > the planner's row count estimates are. > > > Before "VACCUME ANALYZE" > > > > "Index Scan using conversion_table_pk on > > keyword_conversion_table c (cost=0.00..18599.25 > > rows=4986 width=95)" > > " Index Cond: ((conversion_date >= > > '2005-06-07'::date) AND (conversion_date <= > > '2005-08-17'::date))" > > > > After "VACCUME ANALYZE" > > > > "Seq Scan on conversion_table c > (cost=0.00..29990.83 > > rows=1094820 width=66)" > > " Filter: ((conversion_date >= > '2005-06-07'::date) > > AND (conversion_date <= '2005-08-17'::date))" > > > > I dont know why system is doing "Seq scan" now. > > Notice the row count estimates: 4986 in the "before" > query and > 1094820 in the "after" query. In the latter, the > planner thinks > it has to fetch so much of the table that a > sequential scan would > be faster than an index scan. You can see whether > that guess is > correct by disabling enable_seqscan to force an > index scan. It > might be useful to see the output of the following: > > SET enable_seqscan TO on; > SET enable_indexscan TO off; > EXPLAIN ANALYZE SELECT ...; > > SET enable_seqscan TO off; > SET enable_indexscan TO on; > EXPLAIN ANALYZE SELECT ...; > > You might also experiment with planner variables > like effective_cache_size > and random_page_cost to see how changing them > affects the query > plan. However, be careful of tuning the system > based on one query: > make sure adjustments result in reasonable plans for > many different > queries. > > -- > Michael Fuhr > > ---------------------------(end of > broadcast)--------------------------- > TIP 4: Have you searched our list archives? > > http://archives.postgresql.org > ____________________________________________________ Start your day with Yahoo! - make it your home page http://www.yahoo.com/r/hs From pgsql-performance-owner@postgresql.org Fri Aug 26 22:41:31 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6D7CFD7E0D for ; Fri, 26 Aug 2005 22:41:30 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 53392-02 for ; Sat, 27 Aug 2005 01:41:28 +0000 (GMT) Received: from tigger.fuhr.org (tigger.fuhr.org [63.214.45.158]) by svr1.postgresql.org (Postfix) with ESMTP id 61445D7E0F for ; Fri, 26 Aug 2005 22:41:26 -0300 (ADT) Received: from winnie.fuhr.org (winnie.fuhr.org [10.1.0.1]) by tigger.fuhr.org (8.13.3/8.13.3) with ESMTP id j7R1fRE1019691 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK); Fri, 26 Aug 2005 19:41:29 -0600 (MDT) (envelope-from mfuhr@winnie.fuhr.org) Received: from winnie.fuhr.org (localhost [127.0.0.1]) by winnie.fuhr.org (8.13.3/8.13.3) with ESMTP id j7R1fRG4033726; Fri, 26 Aug 2005 19:41:27 -0600 (MDT) (envelope-from mfuhr@winnie.fuhr.org) Received: (from mfuhr@localhost) by winnie.fuhr.org (8.13.3/8.13.3/Submit) id j7R1fQEn033725; Fri, 26 Aug 2005 19:41:26 -0600 (MDT) (envelope-from mfuhr) Date: Fri, 26 Aug 2005 19:41:26 -0600 From: Michael Fuhr To: asif ali Cc: pgsql-performance@postgresql.org Subject: Re: Weird performance drop after VACUUM Message-ID: <20050827014126.GA33637@winnie.fuhr.org> References: <20050826232641.GA19583@winnie.fuhr.org> <20050827001049.58959.qmail@web35210.mail.mud.yahoo.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20050827001049.58959.qmail@web35210.mail.mud.yahoo.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 required=5 tests=[AWL=0.005] X-Spam-Level: X-Archive-Number: 200508/465 X-Sequence-Number: 14212 On Fri, Aug 26, 2005 at 05:10:49PM -0700, asif ali wrote: > "GroupAggregate (cost=195623.66..206672.52 rows=20132 > width=16) (actual time=8205.283..10139.369 rows=55291 > loops=1)" > " -> Sort (cost=195623.66..198360.71 rows=1094820 > width=16) (actual time=8205.114..9029.501 rows=863883 > loops=1)" > " Sort Key: keyword_id" > " -> Seq Scan on keyword_conversion_table c > (cost=0.00..29990.83 rows=1094820 width=16) (actual > time=0.057..1422.319 rows=863883 loops=1)" > " Filter: ((conversion_date >= > '2005-06-07'::date) AND (conversion_date <= > '2005-08-17'::date))" > "Total runtime: 14683.617 ms" What are your effective_cache_size and work_mem (8.x) or sort_mem (7.x) settings? How much RAM does the machine have? If you have enough memory then raising those variables should result in better plans; you might also want to experiment with random_page_cost. Be careful not to set work_mem/sort_mem too high, though. See "Run-time Configuration" in the "Server Run-time Environment" chapter of the documentation for more information about these variables. -- Michael Fuhr From pgsql-performance-owner@postgresql.org Fri Aug 26 23:55:23 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9506CD7D7D for ; Fri, 26 Aug 2005 23:55:22 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 66784-06 for ; Sat, 27 Aug 2005 02:55:21 +0000 (GMT) Received: from linda-5.paradise.net.nz (bm-5a.paradise.net.nz [202.0.58.24]) by svr1.postgresql.org (Postfix) with ESMTP id 84A0AD7D5F for ; Fri, 26 Aug 2005 23:55:19 -0300 (ADT) Received: from smtp-1.paradise.net.nz (smtp-1a.paradise.net.nz [202.0.32.194]) by linda-5.paradise.net.nz (Paradise.net.nz) with ESMTP id <0ILV00LFU1GBSJ@linda-5.paradise.net.nz> for pgsql-performance@postgresql.org; Sat, 27 Aug 2005 14:55:24 +1200 (NZST) Received: from [192.168.1.11] (218-101-14-104.paradise.net.nz [218.101.14.104]) by smtp-1.paradise.net.nz (Postfix) with ESMTP id 7411A82B27; Sat, 27 Aug 2005 14:55:23 +1200 (NZST) Date: Sat, 27 Aug 2005 14:55:22 +1200 From: Mark Kirkwood Subject: Re: Limit + group + join In-reply-to: <9501.1125071193@sss.pgh.pa.us> To: Tom Lane Cc: Tobias Brox , pgsql-performance@postgresql.org Message-id: <430FD61A.3020902@paradise.net.nz> MIME-version: 1.0 Content-type: text/plain; format=flowed; charset=ISO-8859-1 Content-transfer-encoding: 7bit X-Accept-Language: en-us, en User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050726) References: <20050826002709.GK10328@tobias.lan> <430E85ED.60109@paradise.net.nz> <9501.1125071193@sss.pgh.pa.us> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.145 required=5 tests=[AWL=0.145] X-Spam-Level: X-Archive-Number: 200508/466 X-Sequence-Number: 14213 Interestingly enough, 7.4.8 and 8.1devel-2005-08-23 all behave the same as 8.0.3 for me (tables freshly ANALYZEd): joinlimit=# SELECT version(); version ------------------------------------------------------------------------------------------------- PostgreSQL 7.4.8 on i386-unknown-freebsd5.4, compiled by GCC gcc (GCC) 3.4.2 [FreeBSD] 20040728 (1 row) joinlimit=# EXPLAIN SELECT c.id FROM c JOIN b ON c_id=c.id GROUP BY c.id ORDER BY c.id DESC LIMIT 5; QUERY PLAN ----------------------------------------------------------------------------------------------- Limit (cost=10591.36..10591.39 rows=5 width=4) -> Group (cost=10591.36..10992.02 rows=80131 width=4) -> Sort (cost=10591.36..10791.69 rows=80131 width=4) Sort Key: c.id -> Merge Join (cost=0.00..4064.66 rows=80131 width=4) Merge Cond: ("outer".id = "inner".c_id) -> Index Scan using c_pkey on c (cost=0.00..1411.31 rows=80131 width=4) -> Index Scan using b_on_c on b (cost=0.00..1451.72 rows=80172 width=4) (8 rows) joinlimit=# EXPLAIN SELECT c.id FROM c JOIN b ON c_id=c.id GROUP BY c.id ORDER BY c.id LIMIT 5; QUERY PLAN ----------------------------------------------------------------------------------------- Limit (cost=0.00..0.27 rows=5 width=4) -> Group (cost=0.00..4264.99 rows=80131 width=4) -> Merge Join (cost=0.00..4064.66 rows=80131 width=4) Merge Cond: ("outer".id = "inner".c_id) -> Index Scan using c_pkey on c (cost=0.00..1411.31 rows=80131 width=4) -> Index Scan using b_on_c on b (cost=0.00..1451.72 rows=80172 width=4) (6 rows) joinlimit=# SELECT version(); version ---------------------------------------------------------------------------------------------------- PostgreSQL 8.1devel on i386-unknown-freebsd5.4, compiled by GCC gcc (GCC) 3.4.2 [FreeBSD] 20040728 (1 row) joinlimit=# EXPLAIN SELECT c.id FROM c JOIN b ON c_id=c.id GROUP BY c.id ORDER BY c.id DESC LIMIT 5; QUERY PLAN ----------------------------------------------------------------------------------------------- Limit (cost=10654.53..10654.55 rows=5 width=4) -> Group (cost=10654.53..11054.53 rows=80000 width=4) -> Sort (cost=10654.53..10854.53 rows=80000 width=4) Sort Key: c.id -> Merge Join (cost=0.00..4139.44 rows=80000 width=4) Merge Cond: ("outer".id = "inner".c_id) -> Index Scan using c_pkey on c (cost=0.00..1450.00 rows=80000 width=4) -> Index Scan using b_on_c on b (cost=0.00..1490.00 rows=80000 width=4) (8 rows) joinlimit=# EXPLAIN SELECT c.id FROM c JOIN b ON c_id=c.id GROUP BY c.id ORDER BY c.id LIMIT 5; QUERY PLAN ----------------------------------------------------------------------------------------- Limit (cost=0.00..0.27 rows=5 width=4) -> Group (cost=0.00..4339.44 rows=80000 width=4) -> Merge Join (cost=0.00..4139.44 rows=80000 width=4) Merge Cond: ("outer".id = "inner".c_id) -> Index Scan using c_pkey on c (cost=0.00..1450.00 rows=80000 width=4) -> Index Scan using b_on_c on b (cost=0.00..1490.00 rows=80000 width=4) (6 rows) The non default server params of relevance are: shared_buffers = 12000 effective_cache_size = 100000 work_mem/sort_mem = 20480 I did wonder if the highish sort_mem might be a factor, but no, with it set to 1024 I get the same behaviour (just higher sort cost estimates). Cheers Mark Tom Lane wrote: > > > Which PG version are you using exactly? That mistake looks like an > artifact of the 8.0 "fuzzy plan cost" patch, which we fixed recently: > http://archives.postgresql.org/pgsql-committers/2005-07/msg00474.php > > But Tobias wasn't happy with 7.4 either, so I'm not sure that the fuzzy > cost issue explains his results. > > As far as the "desc" point goes, the problem is that mergejoins aren't > capable of dealing with backward sort order, so a merge plan isn't > considered for that case (or at least, it would have to have a sort > after it, which pretty much defeats the point for a fast-start plan). > I have some ideas about fixing this but it won't happen before 8.2. > From pgsql-performance-owner@postgresql.org Sat Aug 27 00:03:50 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 551F7D7E0D for ; Sat, 27 Aug 2005 00:03:50 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 98219-03 for ; Sat, 27 Aug 2005 03:03:48 +0000 (GMT) Received: from stark.xeocode.com (stark.xeocode.com [216.58.44.227]) by svr1.postgresql.org (Postfix) with ESMTP id 67074D7E08 for ; Sat, 27 Aug 2005 00:03:47 -0300 (ADT) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1E8qyg-0006zX-00; Fri, 26 Aug 2005 23:03:30 -0400 To: Tom Lane Cc: Mark Kirkwood , Tobias Brox , pgsql-performance@postgresql.org Subject: Re: Limit + group + join References: <20050826002709.GK10328@tobias.lan> <430E85ED.60109@paradise.net.nz> <9501.1125071193@sss.pgh.pa.us> In-Reply-To: <9501.1125071193@sss.pgh.pa.us> From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 26 Aug 2005 23:03:30 -0400 Message-ID: <87mzn484jh.fsf@stark.xeocode.com> Lines: 14 User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.3 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 required=5 tests=[AWL=0.005] X-Spam-Level: X-Archive-Number: 200508/467 X-Sequence-Number: 14214 Tom Lane writes: > As far as the "desc" point goes, the problem is that mergejoins aren't > capable of dealing with backward sort order, so a merge plan isn't > considered for that case (or at least, it would have to have a sort > after it, which pretty much defeats the point for a fast-start plan). > I have some ideas about fixing this but it won't happen before 8.2. Of course in this case assuming "id" is an integer column you can just sort by -id instead. -- greg From pgsql-performance-owner@postgresql.org Sat Aug 27 03:20:17 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1C83FD7E91 for ; Sat, 27 Aug 2005 03:20:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 44540-03 for ; Sat, 27 Aug 2005 06:20:13 +0000 (GMT) Received: from wolff.to (wolff.to [66.93.197.194]) by svr1.postgresql.org (Postfix) with SMTP id 5B517D7E62 for ; Sat, 27 Aug 2005 03:20:11 -0300 (ADT) Received: (qmail 28325 invoked by uid 500); 27 Aug 2005 06:19:49 -0000 Date: Sat, 27 Aug 2005 01:19:49 -0500 From: Bruno Wolff III To: "Lenard, Rohan (Rohan)" Cc: pgsql-performance@postgresql.org Subject: Re: Need indexes on empty tables for good performance ? Message-ID: <20050827061949.GB27248@wolff.to> Mail-Followup-To: Bruno Wolff III , "Lenard, Rohan (Rohan)" , pgsql-performance@postgresql.org References: <2773CAC687FD5F4689F526998C7E4E5F43F392@au3010avexu1.global.avaya.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <2773CAC687FD5F4689F526998C7E4E5F43F392@au3010avexu1.global.avaya.com> User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/472 X-Sequence-Number: 14219 On Tue, Aug 23, 2005 at 13:41:32 +1000, "Lenard, Rohan (Rohan)" wrote: > I've read that indexes aren't used for COUNT(*) and I've noticed (7.3.x) > with EXPLAIN that indexes never seem to be used on empty tables - is > there any reason to have indexes on empty tables, or will postgresql > never use them. count will use indexes if appropiate. The counts themselves are NOT in the indexes, so counts of significant fractions of a table (in particular of the whole table) won't benefit from indexes. You aren't going to get query speed ups by putting indexes on empty tables. However, they may be required if you have unique or primary keys declared in the table. You may want them to enforce some kinds of constraints. From pgsql-performance-owner@postgresql.org Sat Aug 27 03:30:16 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C9F73D7EC7 for ; Sat, 27 Aug 2005 03:30:13 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 45327-07 for ; Sat, 27 Aug 2005 06:30:10 +0000 (GMT) Received: from mail.travelamericas.com (unknown [206.130.134.147]) by svr1.postgresql.org (Postfix) with SMTP id 9A9ADD7EC6 for ; Sat, 27 Aug 2005 03:30:08 -0300 (ADT) Received: (qmail 12839 invoked from network); 27 Aug 2005 06:30:08 -0000 Received: from unknown (HELO ?10.0.0.253?) (10.0.0.253) by verkiel.travelamericas.com with SMTP; 27 Aug 2005 06:30:08 -0000 Message-ID: <43100870.8070406@travelamericas.com> Date: Fri, 26 Aug 2005 23:30:08 -0700 From: Chris Travers User-Agent: Mozilla Thunderbird 1.0 (X11/20041206) X-Accept-Language: en-us, en MIME-Version: 1.0 To: tobbe Cc: pgsql-performance@postgresql.org Subject: Re: Performance for relative large DB References: <1124799264.282119.167760@g47g2000cwa.googlegroups.com> <60wtmcptzg.fsf@dba2.int.libertyrms.com> <1124864702.822522.64410@g49g2000cwa.googlegroups.com> In-Reply-To: <1124864702.822522.64410@g49g2000cwa.googlegroups.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.042 required=5 tests=[AWL=-0.008, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/473 X-Sequence-Number: 14220 tobbe wrote: >Hi Chris. > >Thanks for the answer. >Sorry that i was a bit unclear. > >1) We update around 20.000 posts per night. > >2) What i meant was that we suspect that the DBMS called PervasiveSQL >that we are using today is much to small. That's why we're looking for >alternatives. > >Today we base our solution much on using querry-specific tables created >at night, so instead of doing querrys direct on the "post" table (with >4-6M rows) at daytime, we have the data pre-aligned in several much >smaller tables. This is just to make the current DBMS coop with our >amount of data. > >What I am particulary interested in is if we can expect to run all our >select querrys directly from the "post" table with PostgreSQL. > > 20k transactions per day? Doesn't seem too bad. That amounts to how many transactions per second during peak times? Personally I don't think it will be a problem, but you might want to clarify what sort of load you are expecting during its peak time. >3) How well does postgres work with load balancing environments. Is it >built-in? > > There is no load balancing "built in." You would need to use Slony-I and possibly Pg-Pool for that. I don't know about Pg-Pool, but Slony-I was written in large part by member(s?) of the core development team so even if it is not "built in" it is not as if it is a team of outsiders who wrote it. If you need something proprietary, there are similar solutions with replication built in which are based on PostgreSQL and licensed under proprietary licenses. Best Wishes, Chris Travers Metatron Technology Consulting From pgsql-performance-owner@postgresql.org Tue Aug 30 13:03:08 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8595FD7EB2 for ; Sat, 27 Aug 2005 03:34:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 45167-06 for ; Sat, 27 Aug 2005 06:34:24 +0000 (GMT) Received: from mail.travelamericas.com (unknown [206.130.134.147]) by svr1.postgresql.org (Postfix) with SMTP id 69056D7ED8 for ; Sat, 27 Aug 2005 03:34:22 -0300 (ADT) Received: (qmail 12882 invoked from network); 27 Aug 2005 06:34:23 -0000 Received: from unknown (HELO ?10.0.0.253?) (10.0.0.253) by verkiel.travelamericas.com with SMTP; 27 Aug 2005 06:34:23 -0000 Message-ID: <4310096F.3020805@metatrontech.com> Date: Fri, 26 Aug 2005 23:34:23 -0700 From: Chris Travers User-Agent: Mozilla Thunderbird 1.0 (X11/20041206) X-Accept-Language: en-us, en MIME-Version: 1.0 To: "Lenard, Rohan (Rohan)" Cc: pgsql-performance@postgresql.org Subject: Re: Need indexes on empty tables for good performance ? References: <2773CAC687FD5F4689F526998C7E4E5F43F392@au3010avexu1.global.avaya.com> In-Reply-To: <2773CAC687FD5F4689F526998C7E4E5F43F392@au3010avexu1.global.avaya.com> Content-Type: multipart/mixed; boundary="------------060901010708010604050506" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 required=5 tests=[AWL=-0.000, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/519 X-Sequence-Number: 14266 This is a multi-part message in MIME format. --------------060901010708010604050506 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Lenard, Rohan (Rohan) wrote: > I've read that indexes aren't used for COUNT(*) and I've noticed > (7.3.x) with EXPLAIN that indexes never seem to be used on empty > tables - is there any reason to have indexes on empty tables, or will > postgresql never use them. You could add a row, vacuum analyze, delete the row, etc.... Then you are fine until you vacuum analyze again ;-) This is a feature designed to prevent really bad plans when you are loading tables with data. However, you are right. It can create bad plans sometimes. Any chance one can eventually come up with a way to tell the planner that an empty table is expected not to grow? Otherwise, I can see nightmares in a data warehouse environment where you have an empty parent table... Best Wishes, Chris Travers Metatron Technology Consulting --------------060901010708010604050506 Content-Type: text/x-vcard; charset=utf-8; name="chris.vcf" Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="chris.vcf" begin:vcard fn:Chris Travers n:Travers;Chris email;internet:chris@metatrontech.com x-mozilla-html:FALSE version:2.1 end:vcard --------------060901010708010604050506-- From pgsql-performance-owner@postgresql.org Sat Aug 27 06:31:22 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A2CA9D6E06 for ; Sat, 27 Aug 2005 06:31:21 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83280-08 for ; Sat, 27 Aug 2005 09:31:18 +0000 (GMT) Received: from kartal.liqia.com (unknown [65.254.45.184]) by svr1.postgresql.org (Postfix) with ESMTP id C035BD6D88 for ; Sat, 27 Aug 2005 06:31:17 -0300 (ADT) Received: from [81.215.232.82] (helo=[10.0.0.201]) by kartal.liqia.com with esmtpa (Exim 4.50 #1 (Debian)) id 1E8x1v-00018K-Tq; Sat, 27 Aug 2005 12:31:17 +0300 Subject: Re: Weird performance drop after VACUUM From: Umit Oztosun To: Tom Lane Cc: pgsql-performance@postgresql.org In-Reply-To: <26745.1125099111@sss.pgh.pa.us> References: <1125093859.9329.30.camel@localhost.localdomain> <26745.1125099111@sss.pgh.pa.us> Content-Type: text/plain Date: Sat, 27 Aug 2005 12:31:13 +0300 Message-Id: <1125135073.8001.21.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.2.1.1 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/474 X-Sequence-Number: 14221 On Cum, 2005-08-26 at 19:31 -0400, Tom Lane wrote: > I think the problem is that the planner is underestimating the cost of > evaluating this complicated filter condition: > > > -> Seq Scan on scf_irsaliye irs (cost=0.00..30.00 rows=1 width=20) (actual time=0.290..0.290 rows=0 loops=10000) > > Filter: (((karsifirma)::text = 'KENDI'::text) AND (((turu)::text = 'MAI'::text) OR ((turu)::text = 'KGI'::text) OR ((turu)::text = 'PS'::text) OR ((turu)::text = 'TS'::text) OR ((turu)::text = 'KC'::text) OR ((turu)::text = 'KCO'::text)) AND (((_key_sis_depo_dest)::text = '$$$$0000003l$1$$'::text) OR ((_key_sis_depo_dest)::text = '$$$$00000048$1$$'::text) OR ((_key_sis_depo_dest)::text = '$$$$0000004b$1$$'::text) OR ((_key_sis_depo_dest)::text = '$$$$0000004d$1$$'::text)) AND (tarih <= '2005-08-26'::date)) > > While you could attack that by raising the cpu_operator_cost parameter, > it would also be worth inquiring *why* the condition is so expensive to > evaluate. I am suspicious that you are running the database in a locale > in which strcoll() is really slow. Can you run it in C locale instead, > or do you really need locale-aware behavior? Can you switch to a > different database encoding? (A single-byte encoding such as Latin1 > might be faster than UTF8, for example.) Yes, you are perfectly right. We are using UTF8 and tr_TR.UTF8 locale. However, I tried the same tests with latin1 and C locale, it is surely faster, but not dramatically. i.e.: Before Vacuum After Vacuum UTF8 and tr_TR.UTF8: ~8 s ~110 s latin1 and C: ~7 s ~65 s I also played with cpu_operator_cost parameter and it dramatically reduced query times, but not to the level before vacuum: Before Vacuum After Vacuum UTF8 and tr_TR.UTF8: ~8 s ~11 s latin1 and C: ~7 s ~9 s These values are much better but I really wonder if I can reach the performance levels before vacuum. I am also worried about the side-effects that may be caused by the non-default cpu_operator_cost parameter. > Another possibility is to take a hard look at whether you can't simplify > the filter condition, but that'd require more knowledge of your > application than I have. Yes that is another option, we are even considering schema changes to use less character types, but these are really costly and error-prone operations at the moment. > Or you could just play with the order of the filter conditions ... for > example, the date condition at the end is probably far cheaper to test > than the text comparisons, so if that's fairly selective it'd be worth > putting it first. We are experimenting on this. Thanks your help! Best Regards, Umit Oztosun From pgsql-performance-owner@postgresql.org Sat Aug 27 07:19:55 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 348DCD79B1 for ; Sat, 27 Aug 2005 07:19:54 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 96278-01 for ; Sat, 27 Aug 2005 10:19:48 +0000 (GMT) Received: from cassarossa.samfundet.no (cassarossa.samfundet.no [129.241.93.19]) by svr1.postgresql.org (Postfix) with ESMTP id C3539D7078 for ; Sat, 27 Aug 2005 07:19:46 -0300 (ADT) Received: from trofast.sesse.net ([129.241.93.32]) by cassarossa.samfundet.no with esmtp (Exim 4.50) id 1E8xms-000247-A3 for pgsql-performance@postgresql.org; Sat, 27 Aug 2005 12:19:47 +0200 Received: from sesse by trofast.sesse.net with local (Exim 3.36 #1 (Debian)) id 1E8xmr-0004vm-00 for ; Sat, 27 Aug 2005 12:19:45 +0200 Date: Sat, 27 Aug 2005 12:19:45 +0200 From: "Steinar H. Gunderson" To: pgsql-performance@postgresql.org Subject: Re: Weird performance drop after VACUUM Message-ID: <20050827101945.GA18155@uio.no> Mail-Followup-To: pgsql-performance@postgresql.org References: <1125093859.9329.30.camel@localhost.localdomain> <26745.1125099111@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <26745.1125099111@sss.pgh.pa.us> X-Operating-System: Linux 2.6.11.8 on a i686 X-Message-Flag: Outlook? --> http://www.mozilla.org/products/thunderbird/ User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.015 required=5 tests=[AWL=0.015] X-Spam-Level: X-Archive-Number: 200508/475 X-Sequence-Number: 14222 On Fri, Aug 26, 2005 at 07:31:51PM -0400, Tom Lane wrote: > Or you could just play with the order of the filter conditions ... for > example, the date condition at the end is probably far cheaper to test > than the text comparisons, so if that's fairly selective it'd be worth > putting it first. That's an interesting approach -- could the planner do such things itself? /* Steinar */ -- Homepage: http://www.sesse.net/ From pgsql-performance-owner@postgresql.org Sat Aug 27 07:50:25 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1990ED6E1F for ; Sat, 27 Aug 2005 07:50:24 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 00832-04 for ; Sat, 27 Aug 2005 10:50:21 +0000 (GMT) Received: from olive.qinip.net (olive.qinip.net [62.100.30.40]) by svr1.postgresql.org (Postfix) with ESMTP id 37377D6E11 for ; Sat, 27 Aug 2005 07:50:20 -0300 (ADT) Received: from [10.0.0.2] (h8441139206.dsl.speedlinq.nl [84.41.139.206]) by olive.qinip.net (Postfix) with ESMTP id D811C18156; Sat, 27 Aug 2005 12:50:19 +0200 (MEST) Message-ID: <4310456F.8040303@tweakers.net> Date: Sat, 27 Aug 2005 12:50:23 +0200 From: Arjen van der Meijden User-Agent: Mozilla Thunderbird 1.0RC1 (Windows/20041201) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: pgsql-performance@postgresql.org Subject: Re: Inefficient queryplan for query with intersectable References: <430EF844.70905@tweakers.net> <430F1393.5040704@archonet.com> <430F8699.3030400@tweakers.net> <26436.1125096976@sss.pgh.pa.us> In-Reply-To: <26436.1125096976@sss.pgh.pa.us> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/476 X-Sequence-Number: 14223 On 27-8-2005 0:56, Tom Lane wrote: > Arjen van der Meijden writes: > >>As said, it chooses sequential scans or "the wrong index plans" over a >>perfectly good plan that is just not selected when the parameters are >>"too well tuned" or sequential scanning of the table is allowed. > > > I think some part of the problem comes from using inconsistent > datatypes. For instance, it seems very odd that the thing is not > using a hash or something to handle > > t_0.Cat2 IN (SELECT 545 UNION SELECT ID FROM cat WHERE ParentID = 545) > > seeing that it correctly guesses there are only going to be about 8 rows > in the union. Part of the reason is that cat2 is smallint, whereas the > output of the union must be at least int, maybe wider depending on the > datatype of cat.id (which you did not show us); so the comparison isn't > hashable. Even a smallint vs int comparison would be mergejoinable, > though, so I'm really wondering what cat.id is. cat.id is a smallint. I replaced that subquery with these two: t_0.Cat2 IN (SELECT '545'::smallint UNION SELECT ID FROM cat WHERE ParentID = '545'::smallint) t_0.Cat2 IN (SELECT '545' UNION SELECT ID FROM cat WHERE ParentID = '545') But appareantly there is a bug in the explain mechanism of the 8.1devel I'm using (I downloaded a nightly 25 august somewhere in the morning (CEST)), since it returned: ERROR: bogus varno: 9 So I can't see whether the plan changed, execution times didn't change much. I also replaced the subselect with the result of that query (like ('545', '546', ...) ) but that didn't seem to make much difference in the execution time as well. The plan did change of course, it used a BitmapOr of 8 Bitmap Index Scans over the pwprodukten. By the way, as far as I know, this is the only datatype mismatch in the query. > Another big part of the problem comes from poor result size estimation. > I'm not sure you can eliminate that entirely given the multiple > conditions on different columns (which'd require cross-column statistics > to really do well, which we do not have). But you could avoid > constructs like > > WHERE ... t_1.recordtimestamp >= > (SELECT max_date - 60 FROM last_dates WHERE table_name = 'pricetracker') > > The planner is basically going to throw up its hands and make a default > guess on the selectivity of this; it's not smart enough to decide that > the sub-select probably represents a constant. What I'd do with this > is to define a function marked STABLE for the sub-select result, perhaps > something like [...] > need.) Then write the query like > > WHERE ... t_1.recordtimestamp >= get_last_date('pricetracker', 60) > > In this formulation the planner will be able to make a reasonable guess > about how many rows will match ... at least if your statistics are up > to date ... I tried such a function and also tried replacing it with the fixed outcome of that suquery itself. Although it has a considerable more accurate estimate of the rows returned, it doesn't seem to impact the basic plan much. It does make the sub-query itself use another index (the one on the recordtimestamp alone, rather than the combined index on leverancierid and recordtimestamp). With that changed subquery it estimates about 4173 rows over 4405 real rows. Actually with the adjusted or original query, it seems to favor the hash join over a nested loop, but the rest of the plan (for the subqueries) seems to be exactly the same. Here is the first part of the explain analyze when it can do any trick it wants: Hash Join (cost=7367.43..186630.19 rows=132426 width=12) (actual time=191.726..11072.025 rows=58065 loops=1) Hash Cond: ("outer".produktid = "inner".id) -> Seq Scan on pwprijs chart_2 (cost=0.00..137491.07 rows=7692207 width=16) (actual time=0.018..6267.744 rows=7692207 loops=1) -> Hash (cost=7366.02..7366.02 rows=565 width=4) (actual time=123.265..123.265 rows=103 loops=1) -> SetOp Intersect (cost=7332.10..7360.37 rows=565 width=4) (actual time=115.760..123.192 rows=103 loops=1) [snip] And here is the first (and last) part when I disable hash joins or seq scans: Nested Loop (cost=7334.92..517159.39 rows=132426 width=12) (actual time=111.905..512.575 rows=58065 loops=1) -> SetOp Intersect (cost=7332.10..7360.37 rows=565 width=4) (actual time=111.588..120.035 rows=103 loops=1) [snip] -> Bitmap Heap Scan on pwprijs chart_2 (cost=2.82..895.85 rows=234 width=16) (actual time=0.344..2.149 rows=564 loops=103) Recheck Cond: (chart_2.produktid = "outer".id) -> Bitmap Index Scan on pwprijs_produktid_idx (cost=0.00..2.82 rows=234 width=0) (actual time=0.189..0.189 rows=564 loops=103) Index Cond: (chart_2.produktid = "outer".id) Is a nested loop normally so much (3x) more costly than a hash join? Or is it just this query that gets estimated wronly? Best regards, Arjen From pgsql-performance-owner@postgresql.org Sat Aug 27 11:27:01 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4B947D7D0A for ; Sat, 27 Aug 2005 11:26:59 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 45407-09 for ; Sat, 27 Aug 2005 14:26:58 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 96F8DD7D04 for ; Sat, 27 Aug 2005 11:26:56 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7RER19s001117; Sat, 27 Aug 2005 10:27:01 -0400 (EDT) To: Arjen van der Meijden Cc: pgsql-performance@postgresql.org Subject: Re: Inefficient queryplan for query with intersectable In-reply-to: <4310456F.8040303@tweakers.net> References: <430EF844.70905@tweakers.net> <430F1393.5040704@archonet.com> <430F8699.3030400@tweakers.net> <26436.1125096976@sss.pgh.pa.us> <4310456F.8040303@tweakers.net> Comments: In-reply-to Arjen van der Meijden message dated "Sat, 27 Aug 2005 12:50:23 +0200" Date: Sat, 27 Aug 2005 10:27:01 -0400 Message-ID: <1116.1125152821@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/477 X-Sequence-Number: 14224 Arjen van der Meijden writes: > But appareantly there is a bug in the explain mechanism of the 8.1devel > I'm using (I downloaded a nightly 25 august somewhere in the morning > (CEST)), since it returned: > ERROR: bogus varno: 9 Yeah, someone else sent in a test case for this failure (or at least one with a similar symptom) yesterday. I'll try to fix it today. > Is a nested loop normally so much (3x) more costly than a hash join? Or > is it just this query that gets estimated wronly? There's been some discussion that we are overestimating the cost of nestloops in general, because we don't take into account that successive scans of the inner relation are likely to find many pages already in cache from the earlier scans. So far no one's come up with a good cost model to use for this, though. regards, tom lane From pgsql-performance-owner@postgresql.org Sat Aug 27 13:42:53 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 021EFD7D64 for ; Sat, 27 Aug 2005 13:42:53 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 70429-05 for ; Sat, 27 Aug 2005 16:42:49 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id C4982D7DD6 for ; Sat, 27 Aug 2005 12:05:08 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7RF51gh001405; Sat, 27 Aug 2005 11:05:01 -0400 (EDT) To: "Steinar H. Gunderson" Cc: pgsql-performance@postgresql.org Subject: Re: Weird performance drop after VACUUM In-reply-to: <20050827101945.GA18155@uio.no> References: <1125093859.9329.30.camel@localhost.localdomain> <26745.1125099111@sss.pgh.pa.us> <20050827101945.GA18155@uio.no> Comments: In-reply-to "Steinar H. Gunderson" message dated "Sat, 27 Aug 2005 12:19:45 +0200" Date: Sat, 27 Aug 2005 11:05:01 -0400 Message-ID: <1404.1125155101@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/481 X-Sequence-Number: 14228 "Steinar H. Gunderson" writes: > On Fri, Aug 26, 2005 at 07:31:51PM -0400, Tom Lane wrote: >> Or you could just play with the order of the filter conditions ... for >> example, the date condition at the end is probably far cheaper to test >> than the text comparisons, so if that's fairly selective it'd be worth >> putting it first. > That's an interesting approach -- could the planner do such things itself? It could, but it doesn't really have enough information. We don't currently have any model that some operators are more expensive than others. IIRC the only sort of reordering the current code will do in a filter condition list is to push clauses involving sub-SELECTs to the end. regards, tom lane From pgsql-performance-owner@postgresql.org Sat Aug 27 12:11:00 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 37942D6E2C for ; Sat, 27 Aug 2005 12:10:59 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 53989-06 for ; Sat, 27 Aug 2005 15:10:56 +0000 (GMT) Received: from smtpauth09.mail.atl.earthlink.net (smtpauth09.mail.atl.earthlink.net [209.86.89.69]) by svr1.postgresql.org (Postfix) with ESMTP id 25AA8D7C5B for ; Sat, 27 Aug 2005 12:10:55 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=ZYgVoeWoDS6VPD3ahdSv1VFqkTXz8uxTbZDthNscQr1xuo4+Tsi7qyWoWQ8HahTo; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth09.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1E92Kc-0006nQ-HO; Sat, 27 Aug 2005 11:10:54 -0400 Message-Id: <6.2.3.4.0.20050827104222.05e36150@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Sat, 27 Aug 2005 11:07:37 -0400 To: Tom Lane , pgsql-performance@postgresql.org From: Ron Subject: Re: Inefficient queryplan for query with In-Reply-To: <1116.1125152821@sss.pgh.pa.us> References: <430EF844.70905@tweakers.net> <430F1393.5040704@archonet.com> <430F8699.3030400@tweakers.net> <26436.1125096976@sss.pgh.pa.us> <4310456F.8040303@tweakers.net> <1116.1125152821@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bc714bbc61b37f6eea4fc0fe41ee64c9e8350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.425 required=5 tests=[AWL=0.051, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/478 X-Sequence-Number: 14225 At 10:27 AM 8/27/2005, Tom Lane wrote: >Arjen van der Meijden writes: > > But appareantly there is a bug in the explain mechanism of the 8.1devel > > I'm using (I downloaded a nightly 25 august somewhere in the morning > > (CEST)), since it returned: > > ERROR: bogus varno: 9 > >Yeah, someone else sent in a test case for this failure (or at least one >with a similar symptom) yesterday. I'll try to fix it today. > > > Is a nested loop normally so much (3x) more costly than a hash join? Or > > is it just this query that gets estimated wronly? > >There's been some discussion that we are overestimating the cost of >nestloops in general, because we don't take into account that successive >scans of the inner relation are likely to find many pages already in >cache from the earlier scans. So far no one's come up with a good cost >model to use for this, though. > > regards, tom lane It certainly seems common in the EXPLAIN ANALYZE output I see that the (estimated) cost of Nested Loop is far higher than the actual time measured. What happened when someone tried the naive approach of telling the planner to estimate the cost of a nested loop based on fitting whatever entities are involved in the nested loop in RAM as much as possible? When there are multiple such mappings, use whichever one results in the lowest cost for the NL in question. Clearly, this should lead to an underestimate of the cost of the constant of operation involved, but since nested loops have the only polynomial growth function of the planner's choices, NL's should still have a decent chance of being more expensive than other choices under most circumstances. In addition, if those costs are based on actual measurements of how long it takes to do such scans then the estimated cost has a decent chance of being fairly accurate under such circumstances. It might not work well, but it seems like a reasonable first attempt at a solution? Ron Peacetree From pgsql-performance-owner@postgresql.org Sat Aug 27 12:26:13 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9A8FFD790E for ; Sat, 27 Aug 2005 12:26:12 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 58452-02 for ; Sat, 27 Aug 2005 15:26:09 +0000 (GMT) Received: from cassarossa.samfundet.no (cassarossa.samfundet.no [129.241.93.19]) by svr1.postgresql.org (Postfix) with ESMTP id 67424D795F for ; Sat, 27 Aug 2005 12:26:06 -0300 (ADT) Received: from trofast.sesse.net ([129.241.93.32]) by cassarossa.samfundet.no with esmtp (Exim 4.50) id 1E92ZI-00065v-Hm for pgsql-performance@postgresql.org; Sat, 27 Aug 2005 17:26:04 +0200 Received: from sesse by trofast.sesse.net with local (Exim 3.36 #1 (Debian)) id 1E92ZH-0006Pr-00 for ; Sat, 27 Aug 2005 17:26:03 +0200 Date: Sat, 27 Aug 2005 17:26:03 +0200 From: "Steinar H. Gunderson" To: pgsql-performance@postgresql.org Subject: Re: Weird performance drop after VACUUM Message-ID: <20050827152603.GF17097@uio.no> Mail-Followup-To: pgsql-performance@postgresql.org References: <1125093859.9329.30.camel@localhost.localdomain> <26745.1125099111@sss.pgh.pa.us> <20050827101945.GA18155@uio.no> <1404.1125155101@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <1404.1125155101@sss.pgh.pa.us> X-Operating-System: Linux 2.6.11.8 on a i686 X-Message-Flag: Outlook? --> http://www.mozilla.org/products/thunderbird/ User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.015 required=5 tests=[AWL=0.015] X-Spam-Level: X-Archive-Number: 200508/479 X-Sequence-Number: 14226 On Sat, Aug 27, 2005 at 11:05:01AM -0400, Tom Lane wrote: > It could, but it doesn't really have enough information. We don't > currently have any model that some operators are more expensive than > others. IIRC the only sort of reordering the current code will do > in a filter condition list is to push clauses involving sub-SELECTs > to the end. I was more thinking along the lines of reordering "a AND/OR b" to "b AND/OR a" if b has lower selectivity than a. /* Steinar */ -- Homepage: http://www.sesse.net/ From pgsql-performance-owner@postgresql.org Sat Aug 27 12:40:48 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 84733D7C46 for ; Sat, 27 Aug 2005 12:40:47 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 61700-02 for ; Sat, 27 Aug 2005 15:40:46 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id A3B65D795F for ; Sat, 27 Aug 2005 12:40:45 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7RFeg2O001639; Sat, 27 Aug 2005 11:40:42 -0400 (EDT) To: "Steinar H. Gunderson" Cc: pgsql-performance@postgresql.org Subject: Re: Weird performance drop after VACUUM In-reply-to: <20050827152603.GF17097@uio.no> References: <1125093859.9329.30.camel@localhost.localdomain> <26745.1125099111@sss.pgh.pa.us> <20050827101945.GA18155@uio.no> <1404.1125155101@sss.pgh.pa.us> <20050827152603.GF17097@uio.no> Comments: In-reply-to "Steinar H. Gunderson" message dated "Sat, 27 Aug 2005 17:26:03 +0200" Date: Sat, 27 Aug 2005 11:40:42 -0400 Message-ID: <1638.1125157242@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/480 X-Sequence-Number: 14227 "Steinar H. Gunderson" writes: > On Sat, Aug 27, 2005 at 11:05:01AM -0400, Tom Lane wrote: >> It could, but it doesn't really have enough information. We don't >> currently have any model that some operators are more expensive than >> others. IIRC the only sort of reordering the current code will do >> in a filter condition list is to push clauses involving sub-SELECTs >> to the end. > I was more thinking along the lines of reordering "a AND/OR b" to "b AND/OR > a" if b has lower selectivity than a. Yeah, but if b is considerably more expensive to evaluate than a, that could still be a net loss. To do it correctly you really need to trade off cost of evaluation against selectivity, and the planner currently only knows something about the latter (and all too often, not enough :-(). I'd like to do this someday, but until we get some cost info in there I think it'd be a mistake to do much re-ordering of conditions. Currently the SQL programmer can determine what happens by writing his query carefully --- if we reorder based on selectivity only, we could make things worse, and there'd be no way to override it. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 30 13:04:34 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 54FE5D79A5 for ; Sat, 27 Aug 2005 13:03:09 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 65600-04 for ; Sat, 27 Aug 2005 16:03:06 +0000 (GMT) Received: from mailserv.pune.wibhu.com (unknown [210.212.168.94]) by svr1.postgresql.org (Postfix) with ESMTP id 142E4D6F81 for ; Sat, 27 Aug 2005 13:03:03 -0300 (ADT) Received: from mailserv.pune.wibhu.com (localhost.localdomain [127.0.0.1]) by mailserv.pune.wibhu.com (8.12.8/8.12.5) with ESMTP id j7RFwwrw018599 for ; Sat, 27 Aug 2005 21:28:59 +0530 Received: (from apache@localhost) by mailserv.pune.wibhu.com (8.12.8/8.12.8/Submit) id j7RFwvQj018597; Sat, 27 Aug 2005 21:28:57 +0530 X-Authentication-Warning: mailserv.pune.wibhu.com: apache set sender to akshay@airtightnetworks.net using -f Received: from 192.168.1.175 (SquirrelMail authenticated user akshay) by mailserv with HTTP; Sat, 27 Aug 2005 21:28:57 +0530 (IST) Message-ID: <2527.192.168.1.175.1125158337.squirrel@mailserv> Date: Sat, 27 Aug 2005 21:28:57 +0530 (IST) Subject: Observation about db response time From: To: X-Priority: 3 Importance: Normal X-Mailer: SquirrelMail (version 1.2.10) MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.178 required=5 tests=[NO_REAL_NAME=0.178] X-Spam-Level: X-Archive-Number: 200508/521 X-Sequence-Number: 14268 Hello Friends, We were having a database in pgsql7.4. The database was responding very slowly even after full vacuum (select count(*) from some_table_having_18000_records was taking 18 Sec). We took a backup of that db and restored it back. Now the same db on same PC is responding fast (same query is taking 18 ms). But we can't do the same as a solution of slow response. Do anybody has faced similar problem? Is this due to any internal problem of pgsql? Is there any clue to fasteen the database? Regards, akshay From pgsql-performance-owner@postgresql.org Sat Aug 27 19:18:06 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 80560D7F1B for ; Sat, 27 Aug 2005 19:18:05 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 63588-03 for ; Sat, 27 Aug 2005 22:18:03 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 206EED7F0A for ; Sat, 27 Aug 2005 19:18:01 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7RMHpaX010923; Sat, 27 Aug 2005 18:17:52 -0400 (EDT) To: Mark Kirkwood Cc: Tobias Brox , pgsql-performance@postgresql.org Subject: Re: Limit + group + join In-reply-to: <430FD61A.3020902@paradise.net.nz> References: <20050826002709.GK10328@tobias.lan> <430E85ED.60109@paradise.net.nz> <9501.1125071193@sss.pgh.pa.us> <430FD61A.3020902@paradise.net.nz> Comments: In-reply-to Mark Kirkwood message dated "Sat, 27 Aug 2005 14:55:22 +1200" Date: Sat, 27 Aug 2005 18:17:51 -0400 Message-ID: <10922.1125181071@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/482 X-Sequence-Number: 14229 Mark Kirkwood writes: > joinlimit=# EXPLAIN SELECT c.id FROM c JOIN b ON c_id=c.id GROUP BY > c.id ORDER BY c.id DESC LIMIT 5; > [ fails to pick an available index-scan-backward plan ] I looked into this and found that indeed the desirable join plan was getting generated, but it wasn't picked because query_planner didn't have an accurate idea of how much of the join needed to be scanned to satisfy the GROUP BY step. I've committed some changes that hopefully will let 8.1 be smarter about GROUP BY ... LIMIT queries. regards, tom lane From pgsql-performance-owner@postgresql.org Sat Aug 27 21:49:02 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A79B8D7FB3 for ; Sat, 27 Aug 2005 21:49:01 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 98036-04 for ; Sun, 28 Aug 2005 00:49:00 +0000 (GMT) Received: from linda-3.paradise.net.nz (bm-3a.paradise.net.nz [202.0.58.22]) by svr1.postgresql.org (Postfix) with ESMTP id 8EF57D7F8B for ; Sat, 27 Aug 2005 21:48:58 -0300 (ADT) Received: from smtp-3.paradise.net.nz (smtp-3a.paradise.net.nz [202.0.32.196]) by linda-3.paradise.net.nz (Paradise.net.nz) with ESMTP id <0ILW00E5IQ9P82@linda-3.paradise.net.nz> for pgsql-performance@postgresql.org; Sun, 28 Aug 2005 12:49:02 +1200 (NZST) Received: from [192.168.1.11] (218-101-45-94.paradise.net.nz [218.101.45.94]) by smtp-3.paradise.net.nz (Postfix) with ESMTP id 467C7AE927; Sun, 28 Aug 2005 12:49:01 +1200 (NZST) Date: Sun, 28 Aug 2005 12:48:59 +1200 From: Mark Kirkwood Subject: Re: Limit + group + join In-reply-to: <10922.1125181071@sss.pgh.pa.us> To: Tom Lane Cc: Tobias Brox , pgsql-performance@postgresql.org Message-id: <431109FB.6000505@paradise.net.nz> MIME-version: 1.0 Content-type: text/plain; format=flowed; charset=ISO-8859-1 Content-transfer-encoding: 7bit X-Accept-Language: en-us, en User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050726) References: <20050826002709.GK10328@tobias.lan> <430E85ED.60109@paradise.net.nz> <9501.1125071193@sss.pgh.pa.us> <430FD61A.3020902@paradise.net.nz> <10922.1125181071@sss.pgh.pa.us> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.142 required=5 tests=[AWL=0.142] X-Spam-Level: X-Archive-Number: 200508/483 X-Sequence-Number: 14230 Tom Lane wrote: > > I looked into this and found that indeed the desirable join plan was > getting generated, but it wasn't picked because query_planner didn't > have an accurate idea of how much of the join needed to be scanned to > satisfy the GROUP BY step. I've committed some changes that hopefully > will let 8.1 be smarter about GROUP BY ... LIMIT queries. > Very nice :-) joinlimit=# EXPLAIN SELECT c.id FROM c JOIN b ON c_id=c.id GROUP BY c.id ORDER BY c.id DESC LIMIT 5; QUERY PLAN -------------------------------------------------------------------------------------------------- Limit (cost=0.00..15.23 rows=5 width=4) -> Group (cost=0.00..243730.00 rows=80000 width=4) -> Nested Loop (cost=0.00..243530.00 rows=80000 width=4) -> Index Scan Backward using c_pkey on c (cost=0.00..1450.00 rows=80000 width=4) -> Index Scan using b_on_c on b (cost=0.00..3.01 rows=1 width=4) Index Cond: (b.c_id = "outer".id) (6 rows) This is 8.1devel from today. regards Mark From pgsql-performance-owner@postgresql.org Sat Aug 27 22:43:03 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9C467D7F1B for ; Sat, 27 Aug 2005 22:43:01 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 10635-05 for ; Sun, 28 Aug 2005 01:42:58 +0000 (GMT) Received: from mail.nordicbet.com (mail.nordicbet.com [193.69.167.130]) by svr1.postgresql.org (Postfix) with ESMTP id E7ADDD78B5 for ; Sat, 27 Aug 2005 22:42:55 -0300 (ADT) Received: from [80.203.125.83] (helo=tobias.nordicbet.invalid) (Authenticated Sender=tobias@nordicbet.com) by mail.nordicbet.com with esmtpa (Exim 4.50 #1 (Debian)) id 1E9C9g-0005f1-Ao; Sun, 28 Aug 2005 03:40:24 +0200 Received: by tobias.nordicbet.invalid (Postfix, from userid 500) id 25EA0E169B; Sun, 28 Aug 2005 03:42:41 +0200 (CEST) Date: Sun, 28 Aug 2005 03:42:40 +0200 From: Tobias Brox To: Mark Kirkwood Cc: Tom Lane , Tobias Brox , pgsql-performance@postgresql.org Subject: Re: Limit + group + join Message-ID: <20050828014240.GB5736@tobias.lan> References: <20050826002709.GK10328@tobias.lan> <430E85ED.60109@paradise.net.nz> <9501.1125071193@sss.pgh.pa.us> <430FD61A.3020902@paradise.net.nz> <10922.1125181071@sss.pgh.pa.us> <431109FB.6000505@paradise.net.nz> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <431109FB.6000505@paradise.net.nz> Organization: Group Nordicbet User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/484 X-Sequence-Number: 14231 [Tom Lane] > I looked into this and (...) I've committed some changes that hopefully will > let 8.1 be smarter about GROUP BY ... LIMIT queries. [Mark Kirkwood] > Very nice :-) (...) > This is 8.1devel from today. Splendid :-) Unfortunately we will not be upgrading for some monthes still, but anyway I'm happy. This provides yet another good argument for upgrading sooner. I'm also happy to see such a perfect match: - A problem that can be reduced from beeing complex and production-specific, to simple and easily reproducible. - Enthusiastic people testing it and pinpointing even more precisely what conditions will cause the condition - Programmers actually fixing the issue - Testers verifying that it was fixed Long live postgresql! :-) -- Notice of Confidentiality: This email is sent unencrypted over the network, and may be stored on several email servers; it can be read by third parties as easy as a postcard. Do not rely on email for confidential information. From pgsql-performance-owner@postgresql.org Sun Aug 28 11:44:38 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id DD7CCD7DB1 for ; Sun, 28 Aug 2005 11:44:37 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 97976-07 for ; Sun, 28 Aug 2005 14:44:32 +0000 (GMT) Received: from olive.qinip.net (olive.qinip.net [62.100.30.40]) by svr1.postgresql.org (Postfix) with ESMTP id 2C78DD7D9F for ; Sun, 28 Aug 2005 11:44:31 -0300 (ADT) Received: from [10.0.0.2] (h8441139206.dsl.speedlinq.nl [84.41.139.206]) by olive.qinip.net (Postfix) with ESMTP id 5B420180DA; Sun, 28 Aug 2005 16:44:33 +0200 (MEST) Message-ID: <4311CDD5.8060508@tweakers.net> Date: Sun, 28 Aug 2005 16:44:37 +0200 From: Arjen van der Meijden User-Agent: Mozilla Thunderbird 1.0RC1 (Windows/20041201) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: pgsql-performance@postgresql.org Subject: Re: Inefficient queryplan for query with intersectable References: <430EF844.70905@tweakers.net> <430F1393.5040704@archonet.com> <430F8699.3030400@tweakers.net> <26436.1125096976@sss.pgh.pa.us> <4310456F.8040303@tweakers.net> <1116.1125152821@sss.pgh.pa.us> In-Reply-To: <1116.1125152821@sss.pgh.pa.us> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/485 X-Sequence-Number: 14232 On 27-8-2005 16:27, Tom Lane wrote: > Arjen van der Meijden writes: > >>Is a nested loop normally so much (3x) more costly than a hash join? Or >>is it just this query that gets estimated wronly? > > There's been some discussion that we are overestimating the cost of > nestloops in general, because we don't take into account that successive > scans of the inner relation are likely to find many pages already in > cache from the earlier scans. So far no one's come up with a good cost > model to use for this, though. Ah, that explains. I take it, there already is an estimation for the cost of "the amount of pages that will be loaded for this operation". For indexed lookups this will probably be something like "the amount of expected pages to fetch * the random page cost"? And appareantly for the nested loop its something like "iterations * amount of pages per iteration * random page cost" ? The naive approach seems to me, is to just calculate the probable amount of pages to fetch from disk rather than from cache. In this case there are 7692207 rows in 60569 pages and on average 234 rows per product (per nested loop) in the estimation. It estimates that it'll have to do 565 iterations. In worst case for the first 234 rows, no pages are already cached and the rows are all in a seperate page. So thats 234 pages to fetch. In the second iteration, you know already 234 pages are fetched and that's about 0.386% of the total pages. So the expected amount of pages for the next 234 pages expected to be in cache is 234 * 0.00386 = 1. After that you'll have 234 + 233 pages in cache, etc, etc. Following that approach, the 565th iteration only has to pull in about 27 new pages in the worst case of all records being perfectly scattered over the pages, not 234. Of course this has to be adjusted for the amount of available buffers and cache and the expected amount of pages to fetch for the iterations, which may be less than 234. When a fetch of a random page costs 4 and one from cache 0.01, there is quite a large difference: 565 * (234 * 4) = 530535 vs 215864,93 Actually the likeliness of a page being in cache is a bit higher, since the expectation increases for each newly fetched page, not for batches of 234. I didn't use that in my calculation here. Anyway, this is probably been thought over already and there may be many flaws in it. If not, please think it over. Best regards, Arjen From pgsql-performance-owner@postgresql.org Sun Aug 28 17:00:52 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B4DB0D80BF for ; Sun, 28 Aug 2005 17:00:51 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 70984-10 for ; Sun, 28 Aug 2005 20:00:50 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 1CA10D80A3 for ; Sun, 28 Aug 2005 17:00:47 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7SK0nVu026477; Sun, 28 Aug 2005 16:00:49 -0400 (EDT) To: Jeff Trout Cc: postgres performance Subject: Re: OSX & Performance In-reply-to: References: Comments: In-reply-to Jeff Trout message dated "Fri, 26 Aug 2005 14:58:17 -0400" Date: Sun, 28 Aug 2005 16:00:49 -0400 Message-ID: <26476.1125259249@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/486 X-Sequence-Number: 14233 Jeff Trout writes: > Tracking ti down a bit timestamp_cmp_internal (The btree was made of > a timestamp & and int) was taking a large amount of time - > specifically all the calls it makes to isnan(x). 14.1% in __isnand Hmm, can you provide a test case for other people to poke at? > Also, two things to note, one of which is quite important: On tiger > (10.4) PG compiles with NO OPTIMIZATION. Probably a template file > needs to be updated. > Panther seems to compile with -O2 though. I see -O2 when building PG (CVS tip) on a fully up-to-date 10.4.2 machine. Maybe something odd in your environment, like a preset CFLAGS setting? regards, tom lane From pgsql-performance-owner@postgresql.org Sun Aug 28 17:41:35 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7864DD7F54 for ; Sun, 28 Aug 2005 17:41:34 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 84281-02 for ; Sun, 28 Aug 2005 20:41:28 +0000 (GMT) Received: from Apachihuilliztli.mtu.ru (apachihuilliztli.mtu.ru [195.34.32.124]) by svr1.postgresql.org (Postfix) with ESMTP id 40413D7E3D for ; Sun, 28 Aug 2005 17:41:26 -0300 (ADT) Received: from umail.ru (umail.mtu.ru [195.34.32.101]) by Apachihuilliztli.mtu.ru (Postfix) with ESMTP id 8A08C20A6F6 for ; Mon, 29 Aug 2005 00:41:25 +0400 (MSD) (envelope-from ilia@obnovlenie.ru) Received: from [85.140.125.21] (HELO ilia) by umail.ru (CommuniGate Pro SMTP 4.2b6) with ESMTP id 558490720 for pgsql-performance@postgresql.org; Mon, 29 Aug 2005 00:41:24 +0400 From: "Ilia Kantor" To: Subject: intarray is broken ? (8.1b1) Date: Mon, 29 Aug 2005 00:41:18 +0400 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0017_01C5AC32.5E9A9B20" X-Mailer: Microsoft Office Outlook, Build 11.0.5510 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 Thread-Index: AcWsENc/4YuF/s7KSymYIzUiy/y9EA== Message-ID: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.192 required=5 tests=[AWL=0.141, FORGED_RCVD_HELO=0.05, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/487 X-Sequence-Number: 14234 This is a multi-part message in MIME format. ------=_NextPart_000_0017_01C5AC32.5E9A9B20 Content-Type: text/plain; charset="koi8-r" Content-Transfer-Encoding: 7bit I tried to use intarray on 8.1 . It seems to give same estimates for anything I ask: explain analyze select * from objects_hier where tg && array[10001] explain analyze select * from objects_hier where tg && array[0] explain analyze select * from objects_hier where tg @ array[10001] explain analyze select * from objects_hier where tg ~ array[0] Some of queries cover whole table, some cover none, but all give same estimated number of rows: Bitmap Heap Scan on objects_hier (cost=2.10..102.75 rows=30 width=337) (actual time=0.028..0.028 rows=0 loops=1) Recheck Cond: (tg && '{0}'::integer[]) -> Bitmap Index Scan on gistbla2 (cost=0.00..2.10 rows= !! 30 !! width=0) (actual time=0.024..0.024 rows=0 loops=1) Index Cond: (tg && '{0}'::integer[]) See the number of estimated rows is 30 is all cases. But actually it varies from whole table (30000 rows) to 0. Looks like GIST indexes for intarray give no statistic at all. It makes them much much less useless than they could be.. Because planner can't plan them well and makes horrible mistakes. For example, puts nested loops in order when for each of 30k rows it makes an index scan within 5 rows => that leads to 30k nested scans, while it should for each of 5 rows perform single index scan among those 30k. Yes, I have all necessary indexes on tables. And yes, I run VACUUM FULL ANALYZE just before the tests. The lack of estimation is not documented anywhere so I just hope this is a bug and can be fixed fast :-) ------=_NextPart_000_0017_01C5AC32.5E9A9B20 Content-Type: text/html; charset="koi8-r" Content-Transfer-Encoding: quoted-printable

I tried to use intarray on 8.1 . It seems to = give same estimates for anything I ask:

 

explain analyze select * from objects_hier = where tg && array[10001]

explain analyze select * from objects_hier = where tg && array[0]

explain analyze select * from objects_hier = where tg @ array[10001]

explain analyze select * from objects_hier = where tg ~ array[0]

 

Some of queries cover whole table, some cover = none, but all give same estimated number of rows:

 

Bitmap Heap Scan on objects_hier=9A = (cost=3D2.10..102.75 rows=3D30 width=3D337) (actual time=3D0.028..0.028 rows=3D0 = loops=3D1)

=9A=9A Recheck Cond: (tg && = '{0}'::integer[])

=9A=9A ->=9A Bitmap Index Scan on = gistbla2=9A (cost=3D0.00..2.10 rows=3D !! 30 !! width=3D0) (actual = time=3D0.024..0.024 rows=3D0 loops=3D1)

=9A=9A=9A=9A=9A=9A=9A=9A Index Cond: (tg = && '{0}'::integer[])

 

See the number of estimated rows is 30 is all = cases.

But actually it varies from whole table (30000 = rows) to 0.

 

Looks like GIST indexes for intarray give no statistic at all.

 

It makes them much much less useless than they = could be.. Because planner can’t plan them well and makes horrible = mistakes.

 

For example, puts nested loops in order when = for each of 30k rows it makes an index scan within 5 rows =3D> that leads to = 30k nested scans, while it should for each of 5 rows perform single index scan = among those 30k.

 

 

Yes, I have all necessary indexes on = tables.

And yes, I run VACUUM FULL ANALYZE just before = the tests.

 

The lack of estimation is not documented = anywhere so I just hope this is a bug and can be fixed fast J

 

 

------=_NextPart_000_0017_01C5AC32.5E9A9B20-- From pgsql-performance-owner@postgresql.org Sun Aug 28 17:48:52 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4E00DD809C for ; Sun, 28 Aug 2005 17:48:51 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 86573-01 for ; Sun, 28 Aug 2005 20:48:47 +0000 (GMT) Received: from Apachihuilliztli.mtu.ru (apachihuilliztli.mtu.ru [195.34.32.124]) by svr1.postgresql.org (Postfix) with ESMTP id 22A3AD8100 for ; Sun, 28 Aug 2005 17:48:45 -0300 (ADT) Received: from umail.ru (umail.mtu.ru [195.34.32.101]) by Apachihuilliztli.mtu.ru (Postfix) with ESMTP id A985820A626 for ; Mon, 29 Aug 2005 00:48:47 +0400 (MSD) (envelope-from ilia@obnovlenie.ru) Received: from [85.140.125.21] (HELO ilia) by umail.ru (CommuniGate Pro SMTP 4.2b6) with ESMTP id 558491990 for pgsql-performance@postgresql.org; Mon, 29 Aug 2005 00:48:47 +0400 From: "Ilia Kantor" To: Subject: Bitmap scan when it is not needed Date: Mon, 29 Aug 2005 00:48:40 +0400 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_001E_01C5AC33.6672B430" X-Mailer: Microsoft Office Outlook, Build 11.0.5510 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 Thread-Index: AcWsEd8vVV8k6GFwS/uc4O6Py5kyLw== Message-ID: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.156 required=5 tests=[AWL=0.105, FORGED_RCVD_HELO=0.05, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/488 X-Sequence-Number: 14235 This is a multi-part message in MIME format. ------=_NextPart_000_001E_01C5AC33.6672B430 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit explain analyze select * from objects_hier where tg && array[0] and id <10000; Bitmap Heap Scan on objects_hier (cost=4.79..8.80 rows=1 width=337) (actual time=0.110..0.110 rows=0 loops=1) Recheck Cond: ((tg && '{0}'::integer[]) AND (id < 10000)) -> BitmapAnd (cost=4.79..4.79 rows=1 width=0) (actual time=0.106..0.106 rows=0 loops=1) -> Bitmap Index Scan on gistbla2 (cost=0.00..2.10 rows=30 width=0) (actual time=0.042..0.042 rows=0 loops=1) Index Cond: (tg && '{0}'::integer[]) -> Bitmap Index Scan on ohid (cost=0.00..2.44 rows=1240 width=0) (actual time=0.058..0.058 rows=1255 loops=1) Index Cond: (id < 10000) I see, Bitmap is going to AND my indexes.. It read one with less number of rows estimated the first (right).. It found 0 records at gistbla2 index. Then why is it reading ohid ? Maybe a quickfix is possible for cases when 0 records is found to stop reading other AND elements.. ------=_NextPart_000_001E_01C5AC33.6672B430 Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable

explain analyze select * from objects_hier = where tg && array[0] and id <10000;

 

Bitmap Heap Scan on objects_hier  (cost=3D4.79..8.80 rows=3D1 width=3D337) (actual time=3D0.110..0.110 = rows=3D0 loops=3D1)

   Recheck Cond: ((tg && '{0}'::integer[]) AND (id < 10000))

   ->  BitmapAnd  (cost=3D4.79..4.79 rows=3D1 width=3D0) (actual time=3D0.106..0.106 = rows=3D0 loops=3D1)

        = ; ->  Bitmap Index Scan on gistbla2  (cost=3D0.00..2.10 = rows=3D30 width=3D0) (actual time=3D0.042..0.042 rows=3D0 = loops=3D1)

        = ;       Index Cond: (tg && = '{0}'::integer[])

        = ; ->  Bitmap Index Scan on ohid  (cost=3D0.00..2.44 = rows=3D1240 width=3D0) (actual time=3D0.058..0.058 rows=3D1255 = loops=3D1)

        = ;       Index Cond: (id < 10000)

 

I see, Bitmap is going to AND my indexes.. It = read one with less number of rows estimated the first = (right)..

It found 0 records at gistbla2 = index…

 

Then why is it reading ohid ?  =

 

Maybe a quickfix is possible for cases when 0 = records is found to stop reading other AND = elements..

 

------=_NextPart_000_001E_01C5AC33.6672B430-- From pgsql-performance-owner@postgresql.org Sun Aug 28 17:59:42 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3C06DD809C for ; Sun, 28 Aug 2005 17:59:42 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 88311-05 for ; Sun, 28 Aug 2005 20:59:40 +0000 (GMT) Received: from Apachihuilliztli.mtu.ru (apachihuilliztli.mtu.ru [195.34.32.124]) by svr1.postgresql.org (Postfix) with ESMTP id A8059D7F54 for ; Sun, 28 Aug 2005 17:59:38 -0300 (ADT) Received: from umail.ru (umail.mtu.ru [195.34.32.101]) by Apachihuilliztli.mtu.ru (Postfix) with ESMTP id 5ECC520A61B for ; Mon, 29 Aug 2005 00:59:40 +0400 (MSD) (envelope-from ilia@obnovlenie.ru) Received: from [85.140.125.21] (HELO ilia) by umail.ru (CommuniGate Pro SMTP 4.2b6) with ESMTP id 558493797 for pgsql-performance@postgresql.org; Mon, 29 Aug 2005 00:59:39 +0400 From: "Ilia Kantor" To: Subject: Planner improvement suggestion Date: Mon, 29 Aug 2005 00:59:33 +0400 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0022_01C5AC34.EB419DB0" X-Mailer: Microsoft Office Outlook, Build 11.0.5510 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 Thread-Index: AcWsE2QHh/PeY0pPRG29Y8huSDcQOw== Message-ID: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.135 required=5 tests=[AWL=0.084, FORGED_RCVD_HELO=0.05, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/489 X-Sequence-Number: 14236 This is a multi-part message in MIME format. ------=_NextPart_000_0022_01C5AC34.EB419DB0 Content-Type: text/plain; charset="koi8-r" Content-Transfer-Encoding: 7bit I have a query: SELECT oh.id FROM objects_hier oh where oh.id < 2000 (!) and oh.id in ( SELECT id as id FROM objects_access oa WHERE oa.master IN (1,2,10001) AND oa.id < 2000 (!) ) The sense of the query is simple: I choose ids from objects_hier where access has necessary masters. The problem is: I have duplicate conditions for id here. They are marked with '!'. I just can't remove any of them, because planner needs to estimate both outer and inner selects to calculate the order Of nested loop or choose a join. If I remove one of duplicate conditions - planner can't estimate well. It's obvious that condition on oh.id can be put inside or outside "oh.id in ( .. )" statement with same result. So I just suggest that the planner should take this into account and "propagate" the condition outside or inside for planning if needed. P.S Is there a way to fix this particular query? Usually oh.id condition is not like <2000, but an inner join. ------=_NextPart_000_0022_01C5AC34.EB419DB0 Content-Type: text/html; charset="koi8-r" Content-Transfer-Encoding: quoted-printable

I have a query:

 

SELECT =
oh.id
FROM =
objects_hier oh
where
oh.id =
< 2000 (!)
and
oh.id =
in (
=9A=9A=9A=9A=9A SELECT id as id FROM =
objects_access oa
=9A=9A=9A=9A=9A =
WHERE
=9A=9A=9A=9A=9A oa.master IN =
(1,2,10001)
=9A=9A=9A=9A=9A =
AND
=9A=9A=9A=9A=9A oa.id < 2000 =
(!)

)

 

The sense of the query is simple: I choose = ids from objects_hier where access has necessary = masters.

 

The problem is: I have duplicate conditions = for id here. They are marked with ‘!’.

 

I just can’t remove any of them, = because planner needs to estimate both outer and inner selects to calculate the = order

Of nested loop or choose a join. If I remove = one of duplicate conditions – planner can’t estimate = well.

 

It’s obvious that condition on oh.id = can be put inside or outside “oh.id in ( .. )” statement with same = result.

 

So I just suggest that the planner should = take this into account and “propagate” the condition outside or inside = for planning if needed.

 

P.S

Is there a way to fix this particular query? = Usually oh.id condition is not like <2000, but an inner = join.

 

 

 

 

 

 

------=_NextPart_000_0022_01C5AC34.EB419DB0-- From pgsql-performance-owner@postgresql.org Sun Aug 28 18:25:10 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 36AE9D712A for ; Sun, 28 Aug 2005 18:25:09 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 95360-01 for ; Sun, 28 Aug 2005 21:25:05 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 1B9C6D6E11 for ; Sun, 28 Aug 2005 18:25:04 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7SLOvH5026889; Sun, 28 Aug 2005 17:24:57 -0400 (EDT) To: "Ilia Kantor" Cc: pgsql-performance@postgresql.org Subject: Re: intarray is broken ? (8.1b1) In-reply-to: References: Comments: In-reply-to "Ilia Kantor" message dated "Mon, 29 Aug 2005 00:41:18 +0400" Date: Sun, 28 Aug 2005 17:24:57 -0400 Message-ID: <26888.1125264297@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/490 X-Sequence-Number: 14237 "Ilia Kantor" writes: > Looks like GIST indexes for intarray give no statistic at all. Feel free to contribute some stats routines that aren't stubs ... regards, tom lane From pgsql-performance-owner@postgresql.org Sun Aug 28 18:30:51 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 550B8D8097 for ; Sun, 28 Aug 2005 18:30:51 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 94778-06 for ; Sun, 28 Aug 2005 21:30:50 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 9A0B7D809A for ; Sun, 28 Aug 2005 18:30:49 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7SLUmHF026928; Sun, 28 Aug 2005 17:30:50 -0400 (EDT) To: "Ilia Kantor" Cc: pgsql-performance@postgresql.org Subject: Re: Planner improvement suggestion In-reply-to: References: Comments: In-reply-to "Ilia Kantor" message dated "Mon, 29 Aug 2005 00:59:33 +0400" Date: Sun, 28 Aug 2005 17:30:48 -0400 Message-ID: <26927.1125264648@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/491 X-Sequence-Number: 14238 "Ilia Kantor" writes: > So I just suggest that the planner should take this into account and > "propagate" the condition outside or inside for planning if needed. I believe it does this already for equality conditions, but not for inequalities. regards, tom lane From pgsql-performance-owner@postgresql.org Sun Aug 28 19:49:11 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A733BD6ECD for ; Sun, 28 Aug 2005 19:49:10 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 16343-03 for ; Sun, 28 Aug 2005 22:49:08 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 346E8D6EA0 for ; Sun, 28 Aug 2005 19:49:07 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7SMn7MZ028245; Sun, 28 Aug 2005 18:49:08 -0400 (EDT) To: "Ilia Kantor" Cc: pgsql-performance@postgresql.org Subject: Re: Bitmap scan when it is not needed In-reply-to: References: Comments: In-reply-to "Ilia Kantor" message dated "Mon, 29 Aug 2005 00:48:40 +0400" Date: Sun, 28 Aug 2005 18:49:07 -0400 Message-ID: <28244.1125269347@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/492 X-Sequence-Number: 14239 "Ilia Kantor" writes: > Maybe a quickfix is possible for cases when 0 records is found to stop > reading other AND elements.. Not sure how useful this will be in practice (since the planner tends not to bother ANDing unselective indexes at all), but it's easy enough to do ... so I did it. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 30 13:03:11 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E8F6ED7AA4 for ; Mon, 29 Aug 2005 10:42:51 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 75114-10 for ; Mon, 29 Aug 2005 13:42:49 +0000 (GMT) Received: from iad2.emailsrvr.com (iad2.emailsrvr.com [207.97.227.212]) by svr1.postgresql.org (Postfix) with ESMTP id 271C1D79F5 for ; Mon, 29 Aug 2005 10:42:46 -0300 (ADT) Received: from [10.10.20.22] (unknown [216.94.157.147]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (No client certificate requested) (Authenticated sender: remyb@medrium.com) by relay3.r3.iad.emailsrvr.com (SMTP Server) with ESMTP id 55FC344C2A3; Mon, 29 Aug 2005 09:42:39 -0400 (EDT) Mime-Version: 1.0 (Apple Message framework v622) Message-Id: Content-Type: multipart/alternative; boundary=Apple-Mail-1-986878859 To: pgsql-performance@postgresql.org Subject: High load and iowait but no disk access From: =?ISO-8859-1?Q?R=E9my_Beaumont?= Date: Mon, 29 Aug 2005 09:42:46 -0400 X-Mailer: Apple Mail (2.622) X-Virus-Scanned: OK X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/520 X-Sequence-Number: 14267 --Apple-Mail-1-986878859 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=ISO-8859-1; format=flowed We have been trying to pinpoint what originally seem to be a I/O=20 bottleneck but which now seems to be an issue with either Postgresql or=20= RHES 3. We have the following test environment on which we can reproduce the=20 problem: 1) Test System A Dell 6650 Quad Xeon Pentium 4 8 Gig of RAM OS: RHES 3 update 2 Storage: NetApp FAS270 connected using an FC card using 10 disks 2) Test System B Dell Dual Xeon Pentium III 2 Gig o RAM OS: RHES 3 update 2 Storage: NetApp FAS920 connected using an FC card using 28 disks Our Database size is around 30G. The behavior we see is that when running queries that do random reads=20 on disk, IOWAIT goes over 80% and actual disk IO falls to a crawl at a=20= throughput bellow 3000kB/s (We usually average 40000 kB/s to 80000 kB/s=20= on sequential read operations on the netapps) The stats of the NetApp do confirm that it is sitting idle. Doing an=20 strace on the Postgresql process shows that is it doing seeks and=20 reads. So my question is where is this iowait time spent ? Is there a way to pinpoint the problem in more details ? We are able to reproduce this behavior with Postgresql 7.4.8 and 8.0.3 I have included the output of top,vmstat,strace and systat from the=20 Netapp from System B while running a single query that generates this=20 behavior. R=E9my top output: 06:27:28 up 5 days, 16:59, 6 users, load average: 1.04, 1.30, 1.01 72 processes: 71 sleeping, 1 running, 0 zombie, 0 stopped CPU states: cpu user nice system irq softirq iowait idle total 2.7% 0.0% 1.0% 0.1% 0.2% 46.0% = 49.5% cpu00 0.2% 0.0% 0.2% 0.0% 0.2% 2.2% = 97.2% cpu01 5.3% 0.0% 1.9% 0.3% 0.3% 89.8% = 1.9% Mem: 2061696k av, 2043936k used, 17760k free, 0k shrd, =20 3916k buff 1566332k actv, 296648k in_d, 30504k in_c Swap: 16771584k av, 21552k used, 16750032k free =20 1933772k cached PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU=20 COMMAND 30960 postgres 15 0 13424 10M 9908 D 2.7 0.5 2:00 1=20 postmaster 30538 root 15 0 1080 764 524 S 0.7 0.0 0:43 0 sshd 1 root 15 0 496 456 436 S 0.0 0.0 0:08 0 init 2 root RT 0 0 0 0 SW 0.0 0.0 0:00 0=20 migration/0 3 root RT 0 0 0 0 SW 0.0 0.0 0:00 1=20 migration/1 4 root 15 0 0 0 0 SW 0.0 0.0 0:01 0=20 keventd 5 root 34 19 0 0 0 SWN 0.0 0.0 0:00 0=20 ksoftirqd/0 6 root 34 19 0 0 0 SWN 0.0 0.0 0:00 1=20 ksoftirqd/1 9 root 15 0 0 0 0 SW 0.0 0.0 0:24 1=20 bdflush 7 root 15 0 0 0 0 SW 0.0 0.0 6:53 1 = kswapd 8 root 15 0 0 0 0 SW 0.0 0.0 8:44 1 = kscand 10 root 15 0 0 0 0 SW 0.0 0.0 0:13 0=20 kupdated 11 root 25 0 0 0 0 SW 0.0 0.0 0:00 0=20 mdrecoveryd 17 root 15 0 0 0 0 SW 0.0 0.0 0:00 0=20 ahc_dv_0 vmstat output procs memory swap io system =20= cpu r b swpd free buff cache si so bi bo in cs us=20= sy id wa 0 1 21552 17796 4872 1931928 2 3 3 1 27 6 2 =20= 1 7 3 0 1 21552 18044 4880 1931652 0 0 1652 0 397 512 1 =20= 2 50 47 0 1 21552 17976 4896 1931664 0 0 2468 0 407 552 2 =20= 2 50 47 1 0 21552 17984 4896 1931608 0 0 2124 0 418 538 3 =20= 3 48 46 0 1 21552 18028 4900 1931536 0 0 1592 0 385 509 1 =20= 3 50 46 0 1 21552 18040 4916 1931488 0 0 1620 820 419 581 2 =20= 2 50 46 0 1 21552 17968 4916 1931536 0 4 1708 4 402 554 3 =20= 1 50 46 1 1 21552 18052 4916 1931388 0 0 1772 0 409 531 3 =20= 1 49 47 0 1 21552 17912 4924 1931492 0 0 1772 0 408 565 3 =20= 1 48 48 0 1 21552 17932 4932 1931440 0 4 1356 4 391 545 5 =20= 0 49 46 0 1 21552 18320 4944 1931016 0 4 1500 840 414 571 1 =20= 1 48 50 0 1 21552 17872 4944 1931440 0 0 2116 0 392 496 1 =20= 5 46 48 0 1 21552 18060 4944 1931232 0 0 2232 0 423 597 1 =20= 2 48 49 1 1 21552 17684 4944 1931584 0 0 1752 0 395 537 1 =20= 1 50 48 0 1 21552 18000 4944 1931240 0 0 1576 0 401 549 0 =20= 1 50 49 NetApp stats: CPU NFS CIFS HTTP Total Net kB/s Disk kB/s Tape kB/s=20 Cache Cache CP CP Disk DAFS FCP iSCSI FCP kB/s in out read write read write =20= age hit time ty util in out 2% 0 0 0 139 0 0 2788 0 0 0 =20= 3 96% 0% - 15% 0 139 0 3 2277 2% 0 0 0 144 0 0 2504 0 0 0 =20= 3 96% 0% - 18% 0 144 0 3 2150 2% 0 0 0 130 0 0 2212 0 0 0 =20= 3 96% 0% - 13% 0 130 0 3 1879 3% 0 0 0 169 0 0 2937 80 0 0 =20= 3 96% 0% - 13% 0 169 0 4 2718 2% 0 0 0 139 0 0 2448 0 0 0 =20= 3 96% 0% - 12% 0 139 0 3 2096 2% 0 0 0 137 0 0 2116 0 0 0 =20= 3 96% 0% - 10% 0 137 0 3 1892 3% 0 0 0 107 0 0 2660 812 0 0 =20= 3 96% 24% T 20% 0 107 0 3 1739 2% 0 0 0 118 0 0 1788 0 0 0 =20= 3 96% 0% - 13% 0 118 0 3 1608 2% 0 0 0 136 0 0 2228 0 0 0 =20= 3 96% 0% - 11% 0 136 0 3 2018 2% 0 0 0 119 0 0 1940 0 0 0 =20= 3 96% 0% - 13% 0 119 0 3 1998 2% 0 0 0 136 0 0 2175 0 0 0 =20= 3 96% 0% - 14% 0 136 0 3 1929 2% 0 0 0 133 0 0 1924 0 0 0 =20= 3 96% 0% - 19% 0 133 0 3 2292 2% 0 0 0 115 0 0 2044 0 0 0 =20= 3 96% 0% - 11% 0 115 0 3 1682 2% 0 0 0 134 0 0 2256 0 0 0 =20= 3 96% 0% - 12% 0 134 0 3 2096 2% 0 0 0 112 0 0 2184 0 0 0 =20= 3 96% 0% - 12% 0 112 0 3 1633 2% 0 0 0 163 0 0 2348 0 0 0 =20= 3 96% 0% - 13% 0 163 0 4 2421 2% 0 0 0 120 0 0 2056 184 0 0 =20= 3 96% 8% T 14% 0 120 0 3 1703 strace output: read(55, "\4\0\0\0\10fm}\1\0\0\0p\0\264\0\0 \2 \230\236\320\0020"...,=20 8192) =3D 8192 _llseek(55, 857997312, [857997312], SEEK_SET) =3D 0 read(55, "\4\0\0\0\\\315\321|\1\0\0\0p\0\354\0\0 \2 \250\236\260"...,=20 8192) =3D 8192 _llseek(55, 883220480, [883220480], SEEK_SET) =3D 0 read(55, "\4\0\0\0T\17a~\1\0\0\0p\0\20\1\0 \2 \270\236\220\2D\235"...,=20= 8192) =3D 8192 _llseek(55, 858005504, [858005504], SEEK_SET) =3D 0 read(55, "\4\0\0\0\300\356\321|\1\0\0\0p\0\330\0\0 \2 \260\236\240"...,=20= 8192) =3D 8192 _llseek(55, 857964544, [857964544], SEEK_SET) =3D 0 read(55, "\4\0\0\0lH\321|\1\0\0\0p\0<\1\0 \2 \300\236\200\2p\235"...,=20 8192) =3D 8192 _llseek(55, 857956352, [857956352], SEEK_SET) =3D 0 read(55, "\4\0\0\0l\'\321|\1\0\0\0p\0\320\0\0 \2 \260\236\240\2\\"...,=20= 8192) =3D 8192 _llseek(55, 910802944, [910802944], SEEK_SET) =3D 0 read(55, "\4\0\0\0\10}\25\200\1\0\0\0l\0\274\1\0 \2 \250\236\260"...,=20 8192) =3D 8192 _llseek(55, 857948160, [857948160], SEEK_SET) =3D 0 read(55, "\4\0\0\0\370\5\321|\1\0\0\0p\0\350\0\0 \2 \230\236\320"...,=20 8192) =3D 8192 _llseek(56, 80371712, [80371712], SEEK_SET) =3D 0 read(56, "\4\0\0\0Lf \217\1\0\0\0p\0\f\1\0 \2 \250\236\260\2T\235"...,=20= 8192) =3D 8192 read(102, "\2\0\34\0001\236\0\0\1\0\0\0\t\0\0\00020670\0\0\0B\6\0"...,=20= 8192) =3D 8192 _llseek(55, 857939968, [857939968], SEEK_SET) =3D 0 read(55, "\4\0\0\0\244\344\320|\1\0\0\0l\0\230\1\0 \2 \244\236\270"...,=20= 8192) =3D 8192 _llseek(55, 857923584, [857923584], SEEK_SET) =3D 0 read(55, "\4\0\0\0\224\242\320|\1\0\0\0p\0|\0\0 \2=20 \234\236\310\002"..., 8192) =3D 8192 _llseek(55, 57270272, [57270272], SEEK_SET) =3D 0 read(55, "\4\0\0\0\3204FK\1\0\0\0t\0\340\0\0 \2=20 \310\236j\2\214\235"..., 8192) =3D 8192 _llseek(55, 870727680, [870727680], SEEK_SET) =3D 0 read(55, "\4\0\0\0x>\233}\1\0\0\0p\0@\1\0 \2 \250\236\260\2X\235"...,=20 8192) =3D 8192 _llseek(55, 1014734848, [1014734848], SEEK_SET) =3D 0 read(55, "\4\0\0\0\34\354\201\206\1\0\0\0p\0p\0\0 \2 \264\236\230"...,=20= 8192) =3D 8192 _llseek(55, 857874432, [857874432], SEEK_SET) =3D 0 read(55, "\4\0\0\0\214\331\317|\1\0\0\0l\0\324\1\0 \2 \224\236\330"...,=20= 8192) =3D 8192 _llseek(55, 760872960, [760872960], SEEK_SET) =3D 0 read(55, "\4\0\0\0\30\257\321v\1\0\0\0p\0\230\0\0 \2 \234\236\310"...,=20= 8192) =3D 8192 _llseek(55, 891715584, [891715584], SEEK_SET) =3D 0 read(55, "\4\0\0\0\370\220\347~\1\0\0\0p\0P\1\0 \2 \230\236\320\2"...,=20= 8192) =3D 8192 _llseek(55, 857858048, [857858048], SEEK_SET) =3D 0 read(55, "\4\0\0\0\250\227\317|\1\0\0\0p\0\264\0\0 \2 \254\236\250"...,=20= 8192) =3D 8192 _llseek(55, 666910720, [666910720], SEEK_SET) =3D 0 read(55, "\4\0\0\0x\206\3q\1\0\0\0p\0004\1\0 \2=20 \254\236\242\2P\235"..., 8192) =3D 8192 _llseek(55, 857841664, [857841664], SEEK_SET) =3D 0 read(55, "\4\0\0\0dT\317|\1\0\0\0p\0\224\0\0 \2 \214\236\350\2\30"...,=20= 8192) =3D 8192 --Apple-Mail-1-986878859 Content-Transfer-Encoding: quoted-printable Content-Type: text/enriched; charset=ISO-8859-1 CourierWe have been trying to pinpoint what originally seem to be a I/O bottleneck but which now seems to be an issue with either Postgresql or RHES 3. We have the following test environment on which we can reproduce the problem: 1) Test System A Dell 6650 Quad Xeon Pentium 4 8 Gig of RAM OS: RHES 3 update 2 Storage: NetApp FAS270 connected using an FC card using 10 disks 2) Test System B Dell Dual Xeon Pentium III 2 Gig o RAM OS: RHES 3 update 2 Storage: NetApp FAS920 connected using an FC card using 28 disks Our Database size is around 30G.=20 The behavior we see is that when running queries that do random reads on disk, IOWAIT goes over 80% and actual disk IO falls to a crawl at a throughput bellow 3000kB/s (We usually average 40000 kB/s to 80000 kB/s on sequential read operations on the netapps) The stats of the NetApp do confirm that it is sitting idle. Doing an strace on the Postgresql process shows that is it doing seeks and reads. So my question is where is this iowait time spent ? Is there a way to pinpoint the problem in more details ? We are able to reproduce this behavior with Postgresql 7.4.8 and 8.0.3 I have included the output of top,vmstat,strace and systat from the Netapp from System B while running a single query that generates this behavior. R=E9my top output: 06:27:28 up 5 days, 16:59, 6 users, load average: 1.04, 1.30, 1.01 72 processes: 71 sleeping, 1 running, 0 zombie, 0 stopped CPU states: cpu user nice system irq softirq iowait =20 idle total 2.7% 0.0% 1.0% 0.1% 0.2% 46.0% =20 49.5% cpu00 0.2% 0.0% 0.2% 0.0% 0.2% 2.2% =20 97.2% cpu01 5.3% 0.0% 1.9% 0.3% 0.3% 89.8% =20 1.9% Mem: 2061696k av, 2043936k used, 17760k free, 0k shrd, =20 3916k buff 1566332k actv, 296648k in_d, 30504k in_c Swap: 16771584k av, 21552k used, 16750032k free =20 1933772k cached PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND 30960 postgres 15 0 13424 10M 9908 D 2.7 0.5 2:00 1 postmaster 30538 root 15 0 1080 764 524 S 0.7 0.0 0:43 0 sshd 1 root 15 0 496 456 436 S 0.0 0.0 0:08 0 init 2 root RT 0 0 0 0 SW 0.0 0.0 0:00 0 migration/0 3 root RT 0 0 0 0 SW 0.0 0.0 0:00 1 migration/1 4 root 15 0 0 0 0 SW 0.0 0.0 0:01 0 keventd 5 root 34 19 0 0 0 SWN 0.0 0.0 0:00 0 ksoftirqd/0 6 root 34 19 0 0 0 SWN 0.0 0.0 0:00 1 ksoftirqd/1 9 root 15 0 0 0 0 SW 0.0 0.0 0:24 1 bdflush 7 root 15 0 0 0 0 SW 0.0 0.0 6:53 1 kswapd 8 root 15 0 0 0 0 SW 0.0 0.0 8:44 1 kscand 10 root 15 0 0 0 0 SW 0.0 0.0 0:13 0 kupdated 11 root 25 0 0 0 0 SW 0.0 0.0 0:00 0 mdrecoveryd 17 root 15 0 0 0 0 SW 0.0 0.0 0:00 0 ahc_dv_0 vmstat output=20 procs memory swap io system = =20 cpu r b swpd free buff cache si so bi bo in cs us sy id wa 0 1 21552 17796 4872 1931928 2 3 3 1 27 6 2=20= 1 7 3 0 1 21552 18044 4880 1931652 0 0 1652 0 397 512 1=20= 2 50 47 0 1 21552 17976 4896 1931664 0 0 2468 0 407 552 2=20= 2 50 47 1 0 21552 17984 4896 1931608 0 0 2124 0 418 538 3=20= 3 48 46 0 1 21552 18028 4900 1931536 0 0 1592 0 385 509 1=20= 3 50 46 0 1 21552 18040 4916 1931488 0 0 1620 820 419 581 2=20= 2 50 46 0 1 21552 17968 4916 1931536 0 4 1708 4 402 554 3=20= 1 50 46 1 1 21552 18052 4916 1931388 0 0 1772 0 409 531 3=20= 1 49 47 0 1 21552 17912 4924 1931492 0 0 1772 0 408 565 3=20= 1 48 48 0 1 21552 17932 4932 1931440 0 4 1356 4 391 545 5=20= 0 49 46 0 1 21552 18320 4944 1931016 0 4 1500 840 414 571 1=20= 1 48 50 0 1 21552 17872 4944 1931440 0 0 2116 0 392 496 1=20= 5 46 48 0 1 21552 18060 4944 1931232 0 0 2232 0 423 597 1=20= 2 48 49 1 1 21552 17684 4944 1931584 0 0 1752 0 395 537 1=20= 1 50 48 0 1 21552 18000 4944 1931240 0 0 1576 0 401 549 0=20= 1 50 49 NetApp stats: CPU NFS CIFS HTTP Total Net kB/s Disk kB/s Tape kB/s Cache Cache CP CP Disk DAFS FCP iSCSI FCP kB/s in out read write read write =20 age hit time ty util in out 2% 0 0 0 139 0 0 2788 0 0 0 =20= 3 96% 0% - 15% 0 139 0 3 2277 2% 0 0 0 144 0 0 2504 0 0 0 =20= 3 96% 0% - 18% 0 144 0 3 2150 2% 0 0 0 130 0 0 2212 0 0 0 =20= 3 96% 0% - 13% 0 130 0 3 1879 3% 0 0 0 169 0 0 2937 80 0 0 =20= 3 96% 0% - 13% 0 169 0 4 2718 2% 0 0 0 139 0 0 2448 0 0 0 =20= 3 96% 0% - 12% 0 139 0 3 2096 2% 0 0 0 137 0 0 2116 0 0 0 =20= 3 96% 0% - 10% 0 137 0 3 1892 3% 0 0 0 107 0 0 2660 812 0 0 =20= 3 96% 24% T 20% 0 107 0 3 1739 2% 0 0 0 118 0 0 1788 0 0 0 =20= 3 96% 0% - 13% 0 118 0 3 1608 2% 0 0 0 136 0 0 2228 0 0 0 =20= 3 96% 0% - 11% 0 136 0 3 2018 2% 0 0 0 119 0 0 1940 0 0 0 =20= 3 96% 0% - 13% 0 119 0 3 1998 2% 0 0 0 136 0 0 2175 0 0 0 =20= 3 96% 0% - 14% 0 136 0 3 1929 2% 0 0 0 133 0 0 1924 0 0 0 =20= 3 96% 0% - 19% 0 133 0 3 2292 2% 0 0 0 115 0 0 2044 0 0 0 =20= 3 96% 0% - 11% 0 115 0 3 1682 2% 0 0 0 134 0 0 2256 0 0 0 =20= 3 96% 0% - 12% 0 134 0 3 2096 2% 0 0 0 112 0 0 2184 0 0 0 =20= 3 96% 0% - 12% 0 112 0 3 1633 2% 0 0 0 163 0 0 2348 0 0 0 =20= 3 96% 0% - 13% 0 163 0 4 2421 2% 0 0 0 120 0 0 2056 184 0 0 =20= 3 96% 8% T 14% 0 120 0 3 1703 strace output: read(55, "\4\0\0\0\10fm}\1\0\0\0p\0\264\0\0 \2 \230\236\320\0020"..., 8192) =3D 8192 _llseek(55, 857997312, [857997312], SEEK_SET) =3D 0 read(55, "\4\0\0\0\\\315\321|\1\0\0\0p\0\354\0\0 \2 \250\236\260"..., 8192) =3D 8192 _llseek(55, 883220480, [883220480], SEEK_SET) =3D 0 read(55, "\4\0\0\0T\17a~\1\0\0\0p\0\20\1\0 \2 \270\236\220\2D\235"..., 8192) =3D 8192 _llseek(55, 858005504, [858005504], SEEK_SET) =3D 0 read(55, "\4\0\0\0\300\356\321|\1\0\0\0p\0\330\0\0 \2 \260\236\240"..., 8192) =3D 8192 _llseek(55, 857964544, [857964544], SEEK_SET) =3D 0 read(55, "\4\0\0\0lH\321|\1\0\0\0p\0<<\1\0 \2 \300\236\200\2p\235"..., 8192) =3D 8192 _llseek(55, 857956352, [857956352], SEEK_SET) =3D 0 read(55, "\4\0\0\0l\'\321|\1\0\0\0p\0\320\0\0 \2 \260\236\240\2\\"..., 8192) =3D 8192 _llseek(55, 910802944, [910802944], SEEK_SET) =3D 0 read(55, "\4\0\0\0\10}\25\200\1\0\0\0l\0\274\1\0 \2 \250\236\260"..., 8192) =3D 8192 _llseek(55, 857948160, [857948160], SEEK_SET) =3D 0 read(55, "\4\0\0\0\370\5\321|\1\0\0\0p\0\350\0\0 \2 \230\236\320"..., 8192) =3D 8192 _llseek(56, 80371712, [80371712], SEEK_SET) =3D 0 read(56, "\4\0\0\0Lf \217\1\0\0\0p\0\f\1\0 \2 \250\236\260\2T\235"..., 8192) =3D 8192 read(102, "\2\0\34\0001\236\0\0\1\0\0\0\t\0\0\00020670\0\0\0B\6\0"..., 8192) =3D 8192 _llseek(55, 857939968, [857939968], SEEK_SET) =3D 0 read(55, "\4\0\0\0\244\344\320|\1\0\0\0l\0\230\1\0 \2 \244\236\270"..., 8192) =3D 8192 _llseek(55, 857923584, [857923584], SEEK_SET) =3D 0 read(55, "\4\0\0\0\224\242\320|\1\0\0\0p\0|\0\0 \2 \234\236\310\002"..., 8192) =3D 8192 _llseek(55, 57270272, [57270272], SEEK_SET) =3D 0 read(55, "\4\0\0\0\3204FK\1\0\0\0t\0\340\0\0 \2 \310\236j\2\214\235"..., 8192) =3D 8192 _llseek(55, 870727680, [870727680], SEEK_SET) =3D 0 read(55, "\4\0\0\0x>\233}\1\0\0\0p\0@\1\0 \2 \250\236\260\2X\235"..., 8192) =3D 8192 _llseek(55, 1014734848, [1014734848], SEEK_SET) =3D 0 read(55, "\4\0\0\0\34\354\201\206\1\0\0\0p\0p\0\0 \2 \264\236\230"..., 8192) =3D 8192 _llseek(55, 857874432, [857874432], SEEK_SET) =3D 0 read(55, "\4\0\0\0\214\331\317|\1\0\0\0l\0\324\1\0 \2 \224\236\330"..., 8192) =3D 8192 _llseek(55, 760872960, [760872960], SEEK_SET) =3D 0 read(55, "\4\0\0\0\30\257\321v\1\0\0\0p\0\230\0\0 \2 \234\236\310"..., 8192) =3D 8192 _llseek(55, 891715584, [891715584], SEEK_SET) =3D 0 read(55, "\4\0\0\0\370\220\347~\1\0\0\0p\0P\1\0 \2 \230\236\320\2"..., 8192) =3D 8192 _llseek(55, 857858048, [857858048], SEEK_SET) =3D 0 read(55, "\4\0\0\0\250\227\317|\1\0\0\0p\0\264\0\0 \2 \254\236\250"..., 8192) =3D 8192 _llseek(55, 666910720, [666910720], SEEK_SET) =3D 0 read(55, "\4\0\0\0x\206\3q\1\0\0\0p\0004\1\0 \2 \254\236\242\2P\235"..., 8192) =3D 8192 _llseek(55, 857841664, [857841664], SEEK_SET) =3D 0 read(55, "\4\0\0\0dT\317|\1\0\0\0p\0\224\0\0 \2 \214\236\350\2\30"..., 8192) =3D 8192 =20 = --Apple-Mail-1-986878859-- From pgsql-performance-owner@postgresql.org Mon Aug 29 13:23:29 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9BDB2D77ED for ; Mon, 29 Aug 2005 13:23:27 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 25009-08 for ; Mon, 29 Aug 2005 16:23:23 +0000 (GMT) Received: from web53211.mail.yahoo.com (web53211.mail.yahoo.com [206.190.49.81]) by svr1.postgresql.org (Postfix) with SMTP id 437D4D6F5F for ; Mon, 29 Aug 2005 13:23:21 -0300 (ADT) Received: (qmail 9578 invoked by uid 60001); 29 Aug 2005 16:23:20 -0000 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com.br; h=Message-ID:Received:Date:From:Subject:To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=YcxVTyik85IbDygyRR7Gk8lqBcJA+DY9kbbIgpkykgzdqVJnXq/k4JOPyhNFgyXWRIYggUTk8sOtWw/bTTlKA/rA0tG4E7ATBXfY/z+zByuRX5sO3KLBwXfTBLMnm1+EymDMQHN7kn0fyktNB+8ddgwpOHlxRbZM9Ig/JGARaRo= ; Message-ID: <20050829162320.9576.qmail@web53211.mail.yahoo.com> Received: from [200.138.208.43] by web53211.mail.yahoo.com via HTTP; Mon, 29 Aug 2005 16:23:20 GMT Date: Mon, 29 Aug 2005 16:23:20 +0000 (GMT) From: Carlos Henrique Reimer Subject: shared buffers To: pgsql-performance@postgresql.org MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="0-1569728133-1125332600=:9424" Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.47 required=5 tests=[DNS_FROM_RFC_ABUSE=0.374, HTML_50_60=0.095, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/493 X-Sequence-Number: 14240 --0-1569728133-1125332600=:9424 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit Hi, I�ve configured postgresql to use 1GB of shared buffers but meminfo and "top" are indicanting 0 shared buffers page. Why? It�s a Linux Redhat 9 box with 4GB RAM and postgresql 7.3. Thanks in advance! Reimer --------------------------------- Yahoo! Acesso Gr�tis: Internet r�pida e gr�tis. Instale o discador agora! --0-1569728133-1125332600=:9424 Content-Type: text/html; charset=iso-8859-1 Content-Transfer-Encoding: 8bit
Hi,
 
I�ve configured postgresql to use 1GB of shared buffers but meminfo and "top" are indicanting 0 shared buffers page. Why?
 
It�s a Linux Redhat 9 box with 4GB RAM and postgresql 7.3.
 
Thanks in advance!
 
Reimer

Yahoo! Acesso Gr�tis: Internet r�pida e gr�tis. Instale o discador agora! --0-1569728133-1125332600=:9424-- From pgsql-performance-owner@postgresql.org Mon Aug 29 14:27:33 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 424F5D7337 for ; Mon, 29 Aug 2005 14:27:32 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 41970-06 for ; Mon, 29 Aug 2005 17:27:29 +0000 (GMT) Received: from rwcrmhc12.comcast.net (rwcrmhc13.comcast.net [216.148.227.118]) by svr1.postgresql.org (Postfix) with ESMTP id C3218D6FEB for ; Mon, 29 Aug 2005 14:27:27 -0300 (ADT) Received: from jefftrout.com ([24.147.120.205]) by comcast.net (rwcrmhc13) with SMTP id <2005082917272601500a0cgue>; Mon, 29 Aug 2005 17:27:27 +0000 Received: (qmail 17864 invoked from network); 29 Aug 2005 17:28:24 -0000 Received: from c-24-147-120-205.hsd1.ma.comcast.net (HELO ?192.168.0.106?) (24.147.120.205) by 192.168.0.109 with SMTP; 29 Aug 2005 17:28:24 -0000 In-Reply-To: <26476.1125259249@sss.pgh.pa.us> References: <26476.1125259249@sss.pgh.pa.us> Mime-Version: 1.0 (Apple Message framework v733) Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: <486EB962-5B80-49B4-AF01-0C2DD77CC9EF@torgo.978.org> Cc: postgres performance Content-Transfer-Encoding: 7bit From: Jeff Trout Subject: Re: OSX & Performance Date: Mon, 29 Aug 2005 13:27:25 -0400 To: Tom Lane X-Mailer: Apple Mail (2.733) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[AWL=-0.000, RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/494 X-Sequence-Number: 14241 On Aug 28, 2005, at 4:00 PM, Tom Lane wrote: > > Hmm, can you provide a test case for other people to poke at? > I'l try to put one together as small as I can make it. The table in question is roughly 22M rows. There are about 8k rows per timestamp (day granularity). > I see -O2 when building PG (CVS tip) on a fully up-to-date 10.4.2 > machine. Maybe something odd in your environment, like a preset > CFLAGS setting? > 8.0.3 doesn't have any optimization flags 8.1beta1 doesn't have any optimization ie: gcc -no-cpp-precomp -Wall -Wmissing-prototypes -Wpointer-arith - Wdeclaration-after-statement -Wold-style-definition -Wendif-labels - fno-strict-aliasing -I../../src/port -I../../src/include -c thread.c -o thread_srv.o I'm on 10.4.2, xcode 2.1 Using built-in specs. Target: powerpc-apple-darwin8 Configured with: /private/var/tmp/gcc/gcc-5026.obj~19/src/configure -- disable-checking --prefix=/usr --mandir=/share/man --enable- languages=c,objc,c++,obj-c++ --program-transform-name=/^[cg][^+.-]*$/ s/$/-4.0/ --with-gxx-include-dir=/include/gcc/darwin/4.0/c++ -- build=powerpc-apple-darwin8 --host=powerpc-apple-darwin8 -- target=powerpc-apple-darwin8 Thread model: posix gcc version 4.0.0 (Apple Computer, Inc. build 5026) The snapshot on ftp.psotgresql.org (dated 8/29) also runs with no optimization. No cflags are set. need to see anything from config.log? -- Jeff Trout http://www.jefftrout.com/ http://www.stuarthamm.net/ From pgsql-performance-owner@postgresql.org Mon Aug 29 14:57:38 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CA1D3D7C6E for ; Mon, 29 Aug 2005 14:57:37 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 50979-05 for ; Mon, 29 Aug 2005 17:57:35 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 4D9CBD7A1F for ; Mon, 29 Aug 2005 14:57:34 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7THvYxj011274; Mon, 29 Aug 2005 13:57:35 -0400 (EDT) To: Jeff Trout Cc: postgres performance Subject: Re: OSX & Performance In-reply-to: <486EB962-5B80-49B4-AF01-0C2DD77CC9EF@torgo.978.org> References: <26476.1125259249@sss.pgh.pa.us> <486EB962-5B80-49B4-AF01-0C2DD77CC9EF@torgo.978.org> Comments: In-reply-to Jeff Trout message dated "Mon, 29 Aug 2005 13:27:25 -0400" Date: Mon, 29 Aug 2005 13:57:34 -0400 Message-ID: <11273.1125338254@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/495 X-Sequence-Number: 14242 Jeff Trout writes: > On Aug 28, 2005, at 4:00 PM, Tom Lane wrote: >> I see -O2 when building PG (CVS tip) on a fully up-to-date 10.4.2 >> machine. Maybe something odd in your environment, like a preset >> CFLAGS setting? > 8.0.3 doesn't have any optimization flags > 8.1beta1 doesn't have any optimization > ie: gcc -no-cpp-precomp -Wall -Wmissing-prototypes -Wpointer-arith - > Wdeclaration-after-statement -Wold-style-definition -Wendif-labels - > fno-strict-aliasing -I../../src/port -I../../src/include -c > thread.c -o thread_srv.o You must have CFLAGS set to empty in your build environment, because configure will certainly default to -O2 if not overridden. It works fine for me on OS X. Maybe you want to trace through the configure script and see why it's doing something else? regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 29 15:08:30 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AC365D7CC1 for ; Mon, 29 Aug 2005 15:07:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 50570-09 for ; Mon, 29 Aug 2005 18:07:20 +0000 (GMT) Received: from web35207.mail.mud.yahoo.com (web35207.mail.mud.yahoo.com [66.163.179.86]) by svr1.postgresql.org (Postfix) with SMTP id 54852D7CB1 for ; Mon, 29 Aug 2005 15:07:17 -0300 (ADT) Received: (qmail 74299 invoked by uid 60001); 29 Aug 2005 18:07:17 -0000 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Message-ID:Received:Date:From:Subject:To:Cc:In-Reply-To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=HLoGm6F3K510fyFbvU5ytEh4st9qXXXpmOc7A00iKpZVX6lr2lXhHIWwYiVNPsTh2vFahLb2bvchJusoKMfOGCAQ10CNuJLopP3/Uq0Wx7gFu7I6QWYqeTgWyjbxfsqC5In4GKuE0JDpu9NqP0SrZEApRx0ToblHoknYkjLGWZs= ; Message-ID: <20050829180717.74297.qmail@web35207.mail.mud.yahoo.com> Received: from [65.218.233.130] by web35207.mail.mud.yahoo.com via HTTP; Mon, 29 Aug 2005 11:07:17 PDT Date: Mon, 29 Aug 2005 11:07:17 -0700 (PDT) From: asif ali Subject: Re: Weird performance drop after VACUUM To: Michael Fuhr Cc: pgsql-performance@postgresql.org In-Reply-To: <20050827014126.GA33637@winnie.fuhr.org> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.28 required=5 tests=[AWL=-0.093, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/496 X-Sequence-Number: 14243 Michael The database is on the same system. What I am doing is only "VACUUM analyze conversion_table" I did the the same thing on a newly created database. And got the same result. So after "VACUUM analyze" performance dropped. Please see this. Runtime changes from "7755.115" to "14859.291" ms explain analyze select keyword_id,sum(daily_impressions) as daily_impressions , sum(daily_clicks) as daily_clicks, COALESCE(sum(daily_cpc::double precision),0) as daily_cpc, sum(daily_revenues)as daily_revenues, sum(daily_actions)as daily_actions ,count(daily_cpc) as count from conversion_table c where c.conversion_date BETWEEN '2005-06-07' and '2005-08-17' group by keyword_Id "HashAggregate (cost=18686.51..18686.54 rows=2 width=52) (actual time=7585.827..7720.370 rows=55717 loops=1)" " -> Index Scan using conversion_table_pk on conversion_table c (cost=0.00..18599.25 rows=4986 width=52) (actual time=0.129..2882.066 rows=885493 loops=1)" " Index Cond: ((conversion_date >= '2005-06-07'::date) AND (conversion_date <= '2005-08-17'::date))" "Total runtime: 7755.115 ms" VACUUM analyze conversion_table explain analyze select keyword_id,sum(daily_impressions) as daily_impressions , sum(daily_clicks) as daily_clicks, COALESCE(sum(daily_cpc::double precision),0) as daily_cpc, sum(daily_revenues)as daily_revenues, sum(daily_actions)as daily_actions ,count(daily_cpc) as count from conversion_table c where c.conversion_date BETWEEN '2005-06-07' and '2005-08-17' group by keyword_Id "GroupAggregate (cost=182521.76..200287.99 rows=20093 width=37) (actual time=8475.580..12618.793 rows=55717 loops=1)" " -> Sort (cost=182521.76..184698.58 rows=870730 width=37) (actual time=8475.246..9418.068 rows=885493 loops=1)" " Sort Key: keyword_id" " -> Seq Scan on conversion_table c (cost=0.00..27336.12 rows=870730 width=37) (actual time=0.007..1520.788 rows=885493 loops=1)" " Filter: ((conversion_date >= '2005-06-07'::date) AND (conversion_date <= '2005-08-17'::date))" "Total runtime: 14859.291 ms" --- Michael Fuhr wrote: > On Fri, Aug 26, 2005 at 05:10:49PM -0700, asif ali > wrote: > > "GroupAggregate (cost=195623.66..206672.52 > rows=20132 > > width=16) (actual time=8205.283..10139.369 > rows=55291 > > loops=1)" > > " -> Sort (cost=195623.66..198360.71 > rows=1094820 > > width=16) (actual time=8205.114..9029.501 > rows=863883 > > loops=1)" > > " Sort Key: keyword_id" > > " -> Seq Scan on keyword_conversion_table > c > > (cost=0.00..29990.83 rows=1094820 width=16) > (actual > > time=0.057..1422.319 rows=863883 loops=1)" > > " Filter: ((conversion_date >= > > '2005-06-07'::date) AND (conversion_date <= > > '2005-08-17'::date))" > > "Total runtime: 14683.617 ms" > > What are your effective_cache_size and work_mem > (8.x) or sort_mem (7.x) > settings? How much RAM does the machine have? If > you have enough > memory then raising those variables should result in > better plans; > you might also want to experiment with > random_page_cost. Be careful > not to set work_mem/sort_mem too high, though. See > "Run-time > Configuration" in the "Server Run-time Environment" > chapter of the > documentation for more information about these > variables. > > -- > Michael Fuhr > > ---------------------------(end of > broadcast)--------------------------- > TIP 4: Have you searched our list archives? > > http://archives.postgresql.org > ____________________________________________________ Start your day with Yahoo! - make it your home page http://www.yahoo.com/r/hs From pgsql-performance-owner@postgresql.org Mon Aug 29 15:42:17 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D888BD735E for ; Mon, 29 Aug 2005 15:41:54 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 64036-04 for ; Mon, 29 Aug 2005 18:41:51 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id 48C52D7275 for ; Mon, 29 Aug 2005 15:41:47 -0300 (ADT) Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: Limit + group + join Date: Mon, 29 Aug 2005 14:41:46 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3417DD207@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Limit + group + join Thread-Index: AcWrcps0SaqA9Qz8SG6XfEbjw6qgEgBVh7Bg From: "Merlin Moncure" To: "Tobias Brox" Cc: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.056 required=5 tests=[AWL=0.006, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/497 X-Sequence-Number: 14244 Tobias wrote: > Splendid :-) Unfortunately we will not be upgrading for some monthes > still, > but anyway I'm happy. This provides yet another good argument for > upgrading > sooner. I'm also happy to see such a perfect match: >=20 > - A problem that can be reduced from beeing complex and > production-specific, to simple and easily reproducible. >=20 > - Enthusiastic people testing it and pinpointing even more precisely what > conditions will cause the condition >=20 > - Programmers actually fixing the issue >=20 > - Testers verifying that it was fixed >=20 > Long live postgresql! :-) In the last three or so years since I've been really active with postgresql, I've found two or three issues/bugs which I was able to reproduce and reduce to a test case. In all instances the fix was in cvs literally within minutes. Merlin From pgsql-performance-owner@postgresql.org Mon Aug 29 16:56:17 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 82398D6E21 for ; Mon, 29 Aug 2005 16:56:16 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 80587-02 for ; Mon, 29 Aug 2005 19:56:14 +0000 (GMT) Received: from hub.org (hub.org [200.46.204.220]) by svr1.postgresql.org (Postfix) with ESMTP id E5623D7CC1 for ; Mon, 29 Aug 2005 16:56:13 -0300 (ADT) Received: from localhost (av.hub.org [200.46.204.144]) by hub.org (Postfix) with ESMTP id 5B29664B903 for ; Mon, 29 Aug 2005 16:56:16 -0300 (ADT) Received: from hub.org ([200.46.204.220]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78445-06 for ; Mon, 29 Aug 2005 19:56:11 +0000 (GMT) Received: from ganymede.hub.org (blk-222-82-85.eastlink.ca [24.222.82.85]) by hub.org (Postfix) with ESMTP id AED2764B8FF for ; Mon, 29 Aug 2005 16:56:13 -0300 (ADT) Received: by ganymede.hub.org (Postfix, from userid 1000) id 5522B3687D; Mon, 29 Aug 2005 16:56:15 -0300 (ADT) Received: from localhost (localhost [127.0.0.1]) by ganymede.hub.org (Postfix) with ESMTP id 5414F35A0A for ; Mon, 29 Aug 2005 16:56:15 -0300 (ADT) Date: Mon, 29 Aug 2005 16:56:15 -0300 (ADT) From: "Marc G. Fournier" X-X-Sender: scrappy@ganymede.hub.org To: pgsql-performance@postgresql.org Subject: getting an index to work with partial indices ... Message-ID: <20050829165335.X1044@ganymede.hub.org> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed X-Virus-Scanned: by amavisd-new at hub.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.032 required=5 tests=[AWL=-0.018, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/498 X-Sequence-Number: 14245 Try as I might, I can't seem to get it to work ... table has >9million rows in it, I've created an index "using btree ( priority ) where priority < 0;", where the table distribution looks like: priority | count ----------+--------- -2 | 138435 -1 | 943250 1 | 3416 9 | 1134171 | 7276960 (5 rows) And it still won't use the index: # explain update table set priority = -3 where priority = -1; QUERY PLAN ------------------------------------------------------------------ Seq Scan on table (cost=0.00..400735.90 rows=993939 width=278) Filter: (priority = -1) (2 rows) But, ti will if I try 'priority = -2' ... what is teh threshhold for using the index? obviously 10% of the records is too high ... thanks ... ---- Marc G. Fournier Hub.Org Networking Services (http://www.hub.org) Email: scrappy@hub.org Yahoo!: yscrappy ICQ: 7615664 From pgsql-performance-owner@postgresql.org Mon Aug 29 17:16:35 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2A348D7F71 for ; Mon, 29 Aug 2005 17:15:31 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83652-04 for ; Mon, 29 Aug 2005 20:15:27 +0000 (GMT) Received: from window.monsterlabs.com (window.monsterlabs.com [216.183.105.176]) by svr1.postgresql.org (Postfix) with SMTP id E6CA9D7E62 for ; Mon, 29 Aug 2005 17:15:26 -0300 (ADT) Received: (qmail 9936 invoked from network); 29 Aug 2005 20:15:27 -0000 Received: from pcp0012204803pcs.blairblvd.tn.nash.comcast.net (HELO ?192.168.15.110?) (69.245.49.69) by 0 with SMTP; 29 Aug 2005 20:15:27 -0000 In-Reply-To: <2773CAC687FD5F4689F526998C7E4E5F43F392@au3010avexu1.global.avaya.com> References: <2773CAC687FD5F4689F526998C7E4E5F43F392@au3010avexu1.global.avaya.com> Mime-Version: 1.0 (Apple Message framework v734) Content-Type: multipart/alternative; boundary=Apple-Mail-33-1010433658 Message-Id: <2408BBD7-2C6D-4FAD-82ED-A8E9DC07D83A@sitening.com> Cc: From: "Thomas F. O'Connell" Subject: Re: Need indexes on empty tables for good performance ? Date: Mon, 29 Aug 2005 15:15:21 -0500 To: "Lenard, Rohan (Rohan)" X-Mailer: Apple Mail (2.734) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.033 required=5 tests=[AWL=0.005, HTML_60_70=0.027, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/499 X-Sequence-Number: 14246 --Apple-Mail-33-1010433658 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=WINDOWS-1252; delsp=yes; format=flowed Rohan, You should note that in Postgres, indexes are not inherited by child =20 tables. Also, it seems difficult to select from a child table whose name you =20 don't know unless you access the parent. And if you are accessing the =20= data via the parent, I'm reasonably certain that you will find that =20 indexes aren't used (even if they exist on the children) as a result =20 of the way the children are accessed. -- Thomas F. O'Connell Co-Founder, Information Architect Sitening, LLC Strategic Open Source: Open Your i=99 http://www.sitening.com/ 110 30th Avenue North, Suite 6 Nashville, TN 37203-6320 615-469-5150 615-469-5151 (fax) On Aug 22, 2005, at 10:41 PM, Lenard, Rohan (Rohan) wrote: > I've read that indexes aren't used for COUNT(*) and I've noticed =20 > (7.3.x) with EXPLAIN that indexes never seem to be used on empty =20 > tables - is there any reason to have indexes on empty tables, or =20 > will postgresql never use them. > > This is not as silly as it sounds - with table inheritance you =20 > might have table children with the data and a parent that is =20 > empty. It'd be nice to make sure postgresql knows to never really =20 > look at the parent - especially is you don't know the names of all =20 > the children .. > > Thoughts ? > > thx, > Rohan --Apple-Mail-33-1010433658 Content-Transfer-Encoding: quoted-printable Content-Type: text/html; charset=WINDOWS-1252 Rohan,

You should note that in = Postgres, indexes are not inherited by child tables.

Also, it seems difficult to = select from a child table whose name you don't know unless you access = the parent. And if you are accessing the data via the parent, I'm = reasonably certain that you will find that indexes aren't used (even if = they exist on the children) as a result of the way the children are = accessed.

http://www.sitening.com/
11= 0 30th Avenue North, Suite 6
Nashville, TN = 37203-6320
615-469-5150
615-469-5151 (fax)
=

On Aug 22, 2005, at 10:41 PM, Lenard, Rohan (Rohan) = wrote:

I've read that indexes aren't used for = COUNT(*) and I've noticed (7.3.x) with EXPLAIN that indexes never seem = to be used on empty tables - is there any reason to have indexes on = empty tables, or will postgresql never use them.
=
=A0
This is not = as silly as it sounds - with table inheritance you might have table = children with the data and a parent that is empty.=A0 It'd be nice to = make sure postgresql knows to never really look at the parent - = especially is you don't know the names of all the children = ..
=A0
Thoughts = ?
=A0
thx,
=A0 = Rohan

= --Apple-Mail-33-1010433658-- From pgsql-performance-owner@postgresql.org Mon Aug 29 17:20:53 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B1ADBD7DBD; Mon, 29 Aug 2005 17:19:06 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83627-08; Mon, 29 Aug 2005 20:19:05 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id C5E0BD7CC1; Mon, 29 Aug 2005 17:19:03 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7TKJ5Nv012791; Mon, 29 Aug 2005 16:19:05 -0400 (EDT) To: "Marc G. Fournier" Cc: pgsql-performance@postgresql.org Subject: Re: getting an index to work with partial indices ... In-reply-to: <20050829165335.X1044@ganymede.hub.org> References: <20050829165335.X1044@ganymede.hub.org> Comments: In-reply-to "Marc G. Fournier" message dated "Mon, 29 Aug 2005 16:56:15 -0300" Date: Mon, 29 Aug 2005 16:19:05 -0400 Message-ID: <12790.1125346745@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/500 X-Sequence-Number: 14247 "Marc G. Fournier" writes: > But, ti will if I try 'priority = -2' ... what is teh threshhold for using > the index? obviously 10% of the records is too high ... Depends on a lot of factors, but usually somewhere between 1% and 10%. (The new bitmap index scan code in 8.1 should be workable for higher percentages.) If this doesn't seem to square with reality for you, you might try reducing random_page_cost. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Aug 29 17:29:31 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 28AFDD7F9D for ; Mon, 29 Aug 2005 17:29:04 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 87635-10 for ; Mon, 29 Aug 2005 20:29:00 +0000 (GMT) Received: from tigger.fuhr.org (tigger.fuhr.org [63.214.45.158]) by svr1.postgresql.org (Postfix) with ESMTP id 6600DD7F31 for ; Mon, 29 Aug 2005 17:28:58 -0300 (ADT) Received: from winnie.fuhr.org (winnie.fuhr.org [10.1.0.1]) by tigger.fuhr.org (8.13.3/8.13.3) with ESMTP id j7TKSvsS023648 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK); Mon, 29 Aug 2005 14:28:59 -0600 (MDT) (envelope-from mfuhr@winnie.fuhr.org) Received: from winnie.fuhr.org (localhost [127.0.0.1]) by winnie.fuhr.org (8.13.3/8.13.3) with ESMTP id j7TKSvvx099568; Mon, 29 Aug 2005 14:28:57 -0600 (MDT) (envelope-from mfuhr@winnie.fuhr.org) Received: (from mfuhr@localhost) by winnie.fuhr.org (8.13.3/8.13.3/Submit) id j7TKSuEs099567; Mon, 29 Aug 2005 14:28:56 -0600 (MDT) (envelope-from mfuhr) Date: Mon, 29 Aug 2005 14:28:56 -0600 From: Michael Fuhr To: asif ali Cc: pgsql-performance@postgresql.org Subject: Re: Weird performance drop after VACUUM Message-ID: <20050829202856.GA99478@winnie.fuhr.org> References: <20050827014126.GA33637@winnie.fuhr.org> <20050829180717.74297.qmail@web35207.mail.mud.yahoo.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20050829180717.74297.qmail@web35207.mail.mud.yahoo.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.005 required=5 tests=[AWL=0.005] X-Spam-Level: X-Archive-Number: 200508/501 X-Sequence-Number: 14248 On Mon, Aug 29, 2005 at 11:07:17AM -0700, asif ali wrote: > The database is on the same system. > What I am doing is only "VACUUM analyze > conversion_table" > > I did the the same thing on a newly created database. > And got the same result. So after "VACUUM analyze" > performance dropped. > Please see this. Runtime changes from "7755.115" to > "14859.291" ms As has been pointed out a couple of times, you're getting a different plan after VACUUM ANALYZE because the row count estimates are more accurate. Unfortunately the more accurate estimates result in a query plan that's slower than the plan for the less accurate estimates. PostgreSQL *thinks* the plan will be faster but your results show that it isn't, so you might need to adjust some of the planner's cost constants. A asked some questions that you didn't answer, so I'll ask them again: What's your effective_cache_size setting? What's your work_mem (8.x) or sort_mem (7.x) setting? What's your random_page_cost setting? How much available RAM does the machine have? What version of PostgreSQL are you running? Various tuning guides give advice on how to set the above and other configuration variables. Here's one such guide: http://www.powerpostgresql.com/PerfList/ -- Michael Fuhr From pgsql-performance-owner@postgresql.org Mon Aug 29 18:26:13 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D7E2DD7D79 for ; Mon, 29 Aug 2005 18:09:24 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 97038-04 for ; Mon, 29 Aug 2005 21:09:21 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [67.100.216.10]) by svr1.postgresql.org (Postfix) with ESMTP id 967DDD7E0E for ; Mon, 29 Aug 2005 18:09:16 -0300 (ADT) Received: by flake.decibel.org (Postfix, from userid 1001) id EF90E152AE; Mon, 29 Aug 2005 16:09:17 -0500 (CDT) Date: Mon, 29 Aug 2005 16:09:17 -0500 From: "Jim C. Nasby" To: tobbe Cc: pgsql-performance@postgresql.org Subject: Re: Performance for relative large DB Message-ID: <20050829210917.GC11282@pervasive.com> References: <1124799264.282119.167760@g47g2000cwa.googlegroups.com> <60wtmcptzg.fsf@dba2.int.libertyrms.com> <1124864702.822522.64410@g49g2000cwa.googlegroups.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1124864702.822522.64410@g49g2000cwa.googlegroups.com> X-Operating-System: FreeBSD 4.11-RELEASE-p10 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.9i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.01 required=5 tests=[AWL=0.010] X-Spam-Level: X-Archive-Number: 200508/502 X-Sequence-Number: 14249 On Tue, Aug 23, 2005 at 11:25:02PM -0700, tobbe wrote: > Hi Chris. > > Thanks for the answer. > Sorry that i was a bit unclear. > > 1) We update around 20.000 posts per night. Doesn't seem like a lot at all. > 2) What i meant was that we suspect that the DBMS called PervasiveSQL > that we are using today is much to small. That's why we're looking for > alternatives. Just so no one gets confused, PervasiveSQL is our Btrieve-based database; it has nothing to do with Pervasive Posgres or PosgreSQL. Also, feel free to contact me off-list if you'd like our help with this. > Today we base our solution much on using querry-specific tables created > at night, so instead of doing querrys direct on the "post" table (with > 4-6M rows) at daytime, we have the data pre-aligned in several much > smaller tables. This is just to make the current DBMS coop with our > amount of data. > > What I am particulary interested in is if we can expect to run all our > select querrys directly from the "post" table with PostgreSQL. Probably, depending on what those queries are, what hardware you have and how the table is laid out. Unless you've got a really high query load I suspect you could handle this on some fairly mundane hardware... > 3) How well does postgres work with load balancing environments. Is it > built-in? As Chris said, there is no built-in solution. PGCluster (http://pgfoundry.org/projects/pgcluster/) is a possible solution should you need clustering/load balancing, but as I mentioned I suspect you should be ok without it. -- Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com Pervasive Software http://pervasive.com 512-569-9461 From pgsql-performance-owner@postgresql.org Tue Aug 30 13:12:54 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id F0932D7F1E for ; Mon, 29 Aug 2005 19:15:42 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 14717-08 for ; Mon, 29 Aug 2005 22:15:39 +0000 (GMT) Received: from tierw.net.avaya.com (tierw.net.avaya.com [198.152.13.100]) by svr1.postgresql.org (Postfix) with ESMTP id 02A57D7EA7 for ; Mon, 29 Aug 2005 19:15:37 -0300 (ADT) Received: from tierw.net.avaya.com (localhost [127.0.0.1]) by tierw.net.avaya.com (Switch-3.1.2/Switch-3.1.0) with ESMTP id j7TM2XE9029771 for ; Mon, 29 Aug 2005 18:02:34 -0400 (EDT) Received: from au3010avexu1.global.avaya.com (h135-27-64-251.avaya.com [135.27.64.251]) by tierw.net.avaya.com (Switch-3.1.2/Switch-3.1.0) with ESMTP id j7TM0YE9027933 for ; Mon, 29 Aug 2005 18:01:13 -0400 (EDT) content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----_=_NextPart_001_01C5ACE6.E73E5078" X-MimeOLE: Produced By Microsoft Exchange V6.0.6603.0 Subject: Re: Need indexes on empty tables for good performance ? Date: Tue, 30 Aug 2005 08:13:37 +1000 Message-ID: <2773CAC687FD5F4689F526998C7E4E5F489B7B@au3010avexu1.global.avaya.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Need indexes on empty tables for good performance ? Thread-Index: AcWs1nJkWo99iBR7RHGnfWZOKlnzvAAEFJGg From: "Lenard, Rohan (Rohan)" To: "Thomas F. O'Connell" Cc: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.017 required=5 tests=[AWL=-0.011, HTML_60_70=0.027, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/523 X-Sequence-Number: 14270 This is a multi-part message in MIME format. ------_=_NextPart_001_01C5ACE6.E73E5078 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Actually the indexes on the child table do seem to get used - I just wanted to make sure there was no penalty not having indexes on the empty parent tables. =20 You are right - the parent is the best way to get at the unknown children ...=20 _____ =20 From: Thomas F. O'Connell [mailto:tfo@sitening.com]=20 Sent: Tuesday, August 30, 2005 6:15 AM To: Lenard, Rohan (Rohan) Cc: pgsql-performance@postgresql.org Subject: Re: [PERFORM] Need indexes on empty tables for good performance ? =09 =09 Rohan,=20 You should note that in Postgres, indexes are not inherited by child tables. Also, it seems difficult to select from a child table whose name you don't know unless you access the parent. And if you are accessing the data via the parent, I'm reasonably certain that you will find that indexes aren't used (even if they exist on the children) as a result of the way the children are accessed. =09 -- Thomas F. O'Connell Co-Founder, Information Architect Sitening, LLC Strategic Open Source: Open Your i(tm) http://www.sitening.com/ 110 30th Avenue North, Suite 6 Nashville, TN 37203-6320 615-469-5150 615-469-5151 (fax) =09 On Aug 22, 2005, at 10:41 PM, Lenard, Rohan (Rohan) wrote: I've read that indexes aren't used for COUNT(*) and I've noticed (7.3.x) with EXPLAIN that indexes never seem to be used on empty tables - is there any reason to have indexes on empty tables, or will postgresql never use them. =20 This is not as silly as it sounds - with table inheritance you might have table children with the data and a parent that is empty. It'd be nice to make sure postgresql knows to never really look at the parent - especially is you don't know the names of all the children .. =20 Thoughts ? =20 thx, Rohan ------_=_NextPart_001_01C5ACE6.E73E5078 Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable
Actually the indexes on the child table do seem = to get used=20 - I just wanted to make sure there was no penalty not = having indexes on the=20 empty parent tables.
 
You are right - the parent is the best way = to get at=20 the unknown children ... 

From: Thomas F. O'Connell=20 [mailto:tfo@sitening.com]
Sent: Tuesday, August 30, 2005 = 6:15=20 AM
To: Lenard, Rohan (Rohan)
Cc:=20 pgsql-performance@postgresql.org
Subject: Re: [PERFORM] Need = indexes=20 on empty tables for good performance ?

Rohan,

You should note that in Postgres, indexes are not inherited by = child=20 tables.

Also, it seems difficult to select from a child table whose name = you=20 don't know unless you access the parent. And if you are accessing the = data via=20 the parent, I'm reasonably certain that you will find that indexes = aren't used=20 (even if they exist on the children) as a result of the way the = children are=20 accessed.

--
Thomas F. O'Connell
Co-Founder, Information Architect
Sitening, LLC

Strategic Open Source: Open Your i™

110 30th Avenue North, Suite 6
Nashville, TN 37203-6320
615-469-5150
615-469-5151 (fax)

On Aug 22, 2005, at 10:41 PM, Lenard, Rohan (Rohan) = wrote:
I've read that=20 indexes aren't used for COUNT(*) and I've noticed (7.3.x) with = EXPLAIN that=20 indexes never seem to be used on empty tables - is there any reason = to have=20 indexes on empty tables, or will postgresql never use=20 them.
 
This is not as=20 silly as it sounds - with table inheritance you might have table = children=20 with the data and a parent that is empty.  It'd be nice to make = sure=20 postgresql knows to never really look at the parent - especially is = you=20 don't know the names of all the children ..
 
Thoughts=20 ?
 
thx,
 =20 = Rohan

------_=_NextPart_001_01C5ACE6.E73E5078-- From pgsql-performance-owner@postgresql.org Mon Aug 29 19:59:16 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 75383D7F31 for ; Mon, 29 Aug 2005 19:59:14 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 25531-10 for ; Mon, 29 Aug 2005 22:59:13 +0000 (GMT) Received: from web35204.mail.mud.yahoo.com (web35204.mail.mud.yahoo.com [66.163.179.83]) by svr1.postgresql.org (Postfix) with SMTP id 52972D7F1E for ; Mon, 29 Aug 2005 19:59:09 -0300 (ADT) Received: (qmail 69967 invoked by uid 60001); 29 Aug 2005 22:59:12 -0000 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Message-ID:Received:Date:From:Subject:To:Cc:In-Reply-To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=JTicJ7aGs+Zeadh2ZMP3406znto2Jl38AgbZXtcRQgDr19CWTDZQIk7Vz/aYQsU0H6zvuFO1XJzmrIxnUw+qQJkzNhdLiSwP3kSdVjefA2vq76vPcH88TNafqoVwSN/k6TbAJLNunR2aUz0ZPKtUOQelJa0t1s+MrnJf5VGX3OY= ; Message-ID: <20050829225912.69965.qmail@web35204.mail.mud.yahoo.com> Received: from [65.218.233.130] by web35204.mail.mud.yahoo.com via HTTP; Mon, 29 Aug 2005 15:59:12 PDT Date: Mon, 29 Aug 2005 15:59:12 -0700 (PDT) From: asif ali Subject: Re: Weird performance drop after VACUUM To: Michael Fuhr Cc: pgsql-performance@postgresql.org In-Reply-To: <20050829202856.GA99478@winnie.fuhr.org> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.28 required=5 tests=[AWL=-0.094, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/503 X-Sequence-Number: 14250 Michael, The effective_cache_size, random_page_cost, work_mem were set to default. (commented). I have changed the setting of these and now the performance is better see below. "HashAggregate (cost=42573.89..42925.52 rows=20093 width=37) (actual time=5273.984..5430.733 rows=55717 loops=1)" " -> Seq Scan on keyword_conversion_table c (cost=0.00..27336.12 rows=870730 width=37) (actual time=0.052..1405.576 rows=885493 loops=1)" " Filter: ((conversion_date >= '2005-06-07'::date) AND (conversion_date <= '2005-08-17'::date))" "Total runtime: 5463.764 ms" Thanks a lot --- Michael Fuhr wrote: > On Mon, Aug 29, 2005 at 11:07:17AM -0700, asif ali > wrote: > > The database is on the same system. > > What I am doing is only "VACUUM analyze > > conversion_table" > > > > I did the the same thing on a newly created > database. > > And got the same result. So after "VACUUM analyze" > > performance dropped. > > Please see this. Runtime changes from "7755.115" > to > > "14859.291" ms > > As has been pointed out a couple of times, you're > getting a different > plan after VACUUM ANALYZE because the row count > estimates are more > accurate. Unfortunately the more accurate estimates > result in a > query plan that's slower than the plan for the less > accurate > estimates. PostgreSQL *thinks* the plan will be > faster but your > results show that it isn't, so you might need to > adjust some of the > planner's cost constants. > > A asked some questions that you didn't answer, so > I'll ask them again: > > What's your effective_cache_size setting? > What's your work_mem (8.x) or sort_mem (7.x) > setting? > What's your random_page_cost setting? > How much available RAM does the machine have? > What version of PostgreSQL are you running? > > Various tuning guides give advice on how to set the > above and other > configuration variables. Here's one such guide: > > http://www.powerpostgresql.com/PerfList/ > > -- > Michael Fuhr > > ---------------------------(end of > broadcast)--------------------------- > TIP 3: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faq > ____________________________________________________ Start your day with Yahoo! - make it your home page http://www.yahoo.com/r/hs From pgsql-performance-owner@postgresql.org Mon Aug 29 22:26:38 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D65C1D82F7 for ; Mon, 29 Aug 2005 22:26:32 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 60830-02 for ; Tue, 30 Aug 2005 01:26:30 +0000 (GMT) Received: from houston.familyhealth.com.au (houston.au.fhnetwork.com [203.22.197.21]) by svr1.postgresql.org (Postfix) with ESMTP id AEE72D82D1 for ; Mon, 29 Aug 2005 22:26:27 -0300 (ADT) Received: from houston.familyhealth.com.au (localhost [127.0.0.1]) by houston.familyhealth.com.au (Postfix) with ESMTP id 092452505A; Tue, 30 Aug 2005 09:26:27 +0800 (WST) Received: from [127.0.0.1] (work-40.internal [192.168.0.40]) by houston.familyhealth.com.au (Postfix) with ESMTP id 296B424FF1; Tue, 30 Aug 2005 09:26:26 +0800 (WST) Message-ID: <4313B5F3.9000400@familyhealth.com.au> Date: Tue, 30 Aug 2005 09:27:15 +0800 From: Christopher Kings-Lynne User-Agent: Mozilla Thunderbird 1.0.6 (Windows/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Carlos Henrique Reimer Cc: pgsql-performance@postgresql.org Subject: Re: shared buffers References: <20050829162320.9576.qmail@web53211.mail.yahoo.com> In-Reply-To: <20050829162320.9576.qmail@web53211.mail.yahoo.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 8bit X-familyhealth-MailScanner-Information: Please contact the ISP for more information X-familyhealth-MailScanner: Found to be clean X-familyhealth-MailScanner-From: chriskl@familyhealth.com.au X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.062 required=5 tests=[AWL=0.012, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/504 X-Sequence-Number: 14251 > I�ve configured postgresql to use 1GB of shared buffers but meminfo and > "top" are indicanting 0 shared buffers page. Why? 1GB shared buffers is far too much. Set it back to like 30000 buffers max... Chris From pgsql-performance-owner@postgresql.org Mon Aug 29 22:54:56 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 71E36D7A26 for ; Mon, 29 Aug 2005 22:54:55 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 75632-05 for ; Tue, 30 Aug 2005 01:54:51 +0000 (GMT) Received: from web53204.mail.yahoo.com (web53204.mail.yahoo.com [206.190.49.74]) by svr1.postgresql.org (Postfix) with SMTP id 26703D6E21 for ; Mon, 29 Aug 2005 22:54:50 -0300 (ADT) Received: (qmail 54154 invoked by uid 60001); 30 Aug 2005 01:54:54 -0000 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com.br; h=Message-ID:Received:Date:From:Subject:To:In-Reply-To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=jyf4ihk8PLibD/muw8PHko9j+mgcq0o52j4fUI3GX7xI5iagO85uJ68T5X8tA85R+sgzH+WCTMUc1Qdr9FqPZApYmeDg/tG7lYT0MI09F8H/3UVbDVFp7+JIQvwZcfJ+g4xsJ2F0t0qTXtAETLAlZalMEYM8TUUsOVXd8vagETs= ; Message-ID: <20050830015454.54152.qmail@web53204.mail.yahoo.com> Received: from [201.2.219.48] by web53204.mail.yahoo.com via HTTP; Mon, 29 Aug 2005 22:54:54 ART Date: Mon, 29 Aug 2005 22:54:54 -0300 (ART) From: Carlos Henrique Reimer Subject: Re: shared buffers To: Christopher Kings-Lynne , pgsql-performance@postgresql.org In-Reply-To: <4313B5F3.9000400@familyhealth.com.au> MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="0-841335462-1125366894=:50910" Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.321 required=5 tests=[AWL=-0.140, DNS_FROM_RFC_ABUSE=0.374, HTML_40_50=0.086, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/505 X-Sequence-Number: 14252 --0-841335462-1125366894=:50910 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit I forgot to say that it�s a 12GB database... Ok, I�ll set shared buffers to 30.000 pages but even so "meminfo" and "top" shouldn�t show some shared pages? I heard something about that Redhat 9 can�t handle very well RAM higher than 2GB. Is it right? Thanks in advance! Reimer Christopher Kings-Lynne escreveu: > I�ve configured postgresql to use 1GB of shared buffers but meminfo and > "top" are indicanting 0 shared buffers page. Why? 1GB shared buffers is far too much. Set it back to like 30000 buffers max... Chris --------------------------------- Yahoo! Acesso Gr�tis: Internet r�pida e gr�tis. Instale o discador agora! --0-841335462-1125366894=:50910 Content-Type: text/html; charset=iso-8859-1 Content-Transfer-Encoding: 8bit
I forgot to say that it�s a 12GB database...
 
Ok, I�ll set shared buffers to 30.000 pages but even so "meminfo" and "top" shouldn�t show some shared pages?
 
I heard something about that Redhat 9 can�t handle very well RAM higher than 2GB. Is it right?
Thanks in advance!
 
Reimer

Christopher Kings-Lynne <chriskl@familyhealth.com.au> escreveu:
> I�ve configured postgresql to use 1GB of shared buffers but meminfo and
> "top" are indicanting 0 shared buffers page. Why?

1GB shared buffers is far too much. Set it back to like 30000 buffers
max...

Chris

Yahoo! Acesso Gr�tis: Internet r�pida e gr�tis. Instale o discador agora! --0-841335462-1125366894=:50910-- From pgsql-performance-owner@postgresql.org Mon Aug 29 23:07:33 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 561D9D8208 for ; Mon, 29 Aug 2005 23:07:32 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 88236-07 for ; Tue, 30 Aug 2005 02:07:31 +0000 (GMT) Received: from houston.familyhealth.com.au (houston.au.fhnetwork.com [203.22.197.21]) by svr1.postgresql.org (Postfix) with ESMTP id 06185D818E for ; Mon, 29 Aug 2005 23:07:29 -0300 (ADT) Received: from houston.familyhealth.com.au (localhost [127.0.0.1]) by houston.familyhealth.com.au (Postfix) with ESMTP id EDEEE24FE6; Tue, 30 Aug 2005 10:07:32 +0800 (WST) Received: from [127.0.0.1] (work-40.internal [192.168.0.40]) by houston.familyhealth.com.au (Postfix) with ESMTP id 9EBEA2505D; Tue, 30 Aug 2005 10:07:30 +0800 (WST) Message-ID: <4313BF95.3030105@familyhealth.com.au> Date: Tue, 30 Aug 2005 10:08:21 +0800 From: Christopher Kings-Lynne User-Agent: Mozilla Thunderbird 1.0.6 (Windows/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Carlos Henrique Reimer Cc: pgsql-performance@postgresql.org Subject: Re: shared buffers References: <20050830015454.54152.qmail@web53204.mail.yahoo.com> In-Reply-To: <20050830015454.54152.qmail@web53204.mail.yahoo.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 8bit X-familyhealth-MailScanner-Information: Please contact the ISP for more information X-familyhealth-MailScanner: Found to be clean X-familyhealth-MailScanner-From: chriskl@familyhealth.com.au X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.062 required=5 tests=[AWL=0.012, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/506 X-Sequence-Number: 14253 > I forgot to say that it�s a 12GB database... That's actually not that large. > Ok, I�ll set shared buffers to 30.000 pages but even so "meminfo" and > "top" shouldn�t show some shared pages? Yeah. The reason for not setting buffers so high is because PostgreSQL cannot efficiently manage huge shared buffers, so you're better off giving the RAM to Linux's disk cache. Chris From pgsql-performance-owner@postgresql.org Mon Aug 29 23:16:28 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3D83DD8077 for ; Mon, 29 Aug 2005 23:16:28 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 22707-07 for ; Tue, 30 Aug 2005 02:16:27 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 33381D7A26 for ; Mon, 29 Aug 2005 23:16:25 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7U2GKt7016377; Mon, 29 Aug 2005 22:16:20 -0400 (EDT) To: Carlos Henrique Reimer Cc: Christopher Kings-Lynne , pgsql-performance@postgresql.org Subject: Re: shared buffers In-reply-to: <20050830015454.54152.qmail@web53204.mail.yahoo.com> References: <20050830015454.54152.qmail@web53204.mail.yahoo.com> Comments: In-reply-to Carlos Henrique Reimer message dated "Mon, 29 Aug 2005 22:54:54 -0300" Date: Mon, 29 Aug 2005 22:16:20 -0400 Message-ID: <16376.1125368180@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/507 X-Sequence-Number: 14254 Carlos Henrique Reimer writes: > I heard something about that Redhat 9 can�t handle very well RAM higher than 2GB. Is it right? RHL 9 is certainly pretty long in the tooth. Why aren't you using a more recent distro? regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 30 01:53:35 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 36484D7978 for ; Tue, 30 Aug 2005 01:53:34 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 13915-05 for ; Tue, 30 Aug 2005 04:53:32 +0000 (GMT) Received: from spark.hss.co.in (spark.hss.co.in [203.145.155.21]) by svr1.postgresql.org (Postfix) with ESMTP id 0BE4FD7909 for ; Tue, 30 Aug 2005 01:53:30 -0300 (ADT) Received: from ultra.hss.co.in (ultra [192.168.100.5]) by spark.hss.co.in (8.12.10/8.12.10) with ESMTP id j7U4mF3q020767 for ; Tue, 30 Aug 2005 10:18:21 +0530 (IST) Received: from sandesh.hss.hns.com (localhost [127.0.0.1]) by ultra.hss.co.in (8.10.0/8.10.0) with ESMTP id j7U4sKB09411 for ; Tue, 30 Aug 2005 10:24:20 +0530 (IST) Subject: How to improve Postgres performance To: pgsql-performance@postgresql.org X-Mailer: Lotus Notes Release 6.5 September 26, 2003 Message-ID: From: Hemant Pandey Date: Tue, 30 Aug 2005 10:05:02 +0530 X-MIMETrack: Serialize by Router on Sandesh/HSS(Release 6.0|September 26, 2002) at 30/08/2005 10:23:17 AM MIME-Version: 1.0 Content-type: text/plain; charset=US-ASCII X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.464 required=5 tests=[AWL=-0.000, ROUND_THE_WORLD_LOCAL=0.464] X-Spam-Level: X-Archive-Number: 200508/508 X-Sequence-Number: 14255 Hi All, I am running an application, which connects to the postgres database at initialization time and perform the database operations like Select/Update. Database queries are very simple. On analyzing my application through Quantifier( Performance analyzing tool), I found the most of the time my application is waiting on recv, waiting for Database response. When i run Vacuum on perticular tables, i observed that performance improoves drastically. So please tell me how can i improve database performance through configuration parameters. I had tried to change parameters in postgresql.conf file but of no avail. Now i am trying to Auto Vacuum, but don't know how to run Auto Vacuum. Please help me to solve these problems. Thanks in advance. Hemant *********************** FSS-Unclassified *********************** "DISCLAIMER: This message is proprietary to Flextronics Software Systems Limited (FSS) and is intended solely for the use of the individual to whom it is addressed. It may contain privileged or confidential information and should not be circulated or used for any purpose other than for what it is intended. If you have received this message in error, please notify the originator immediately. If you are not the intended recipient, you are notified that you are strictly prohibited from using, copying, altering, or disclosing the contents of this message. FSS accepts no responsibility for loss or damage arising from the use of the information transmitted by this email including damage from virus." From pgsql-performance-owner@postgresql.org Tue Aug 30 03:04:43 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 90526D7D06 for ; Tue, 30 Aug 2005 03:04:42 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 33414-01 for ; Tue, 30 Aug 2005 06:04:40 +0000 (GMT) Received: from zigo.dhs.org (ua-83-227-204-174.cust.bredbandsbolaget.se [83.227.204.174]) by svr1.postgresql.org (Postfix) with ESMTP id 57352D7CFC for ; Tue, 30 Aug 2005 03:04:37 -0300 (ADT) Received: from zigo.zigo.dhs.org (zigo.zigo.dhs.org [192.168.0.1]) by zigo.dhs.org (Postfix) with ESMTP id C2B528467; Tue, 30 Aug 2005 08:04:35 +0200 (CEST) Date: Tue, 30 Aug 2005 08:04:35 +0200 (CEST) From: Dennis Bjorklund To: Hemant Pandey Cc: pgsql-performance@postgresql.org Subject: Re: How to improve Postgres performance In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=ISO-8859-1 Content-Transfer-Encoding: 8BIT X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.321 required=5 tests=[AWL=-0.103, DNS_FROM_RFC_ABUSE=0.374, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/509 X-Sequence-Number: 14256 On Tue, 30 Aug 2005, Hemant Pandey wrote: > So please tell me how can i improve database performance through > configuration parameters. I had tried to change parameters in > postgresql.conf file but of no avail. > Now i am trying to Auto Vacuum, but don't know how to run Auto Vacuum. The most important part is that you need to run VACUUM ANALYZE regulary. Vacuum can be started each night in a cron job, started from pg_autovacuum when it thinks it's needed, or started in some other way. In any case, it has to be run whenever the data in the database have changed enough. The parameters in the config that is most important in my experience is effective_cache_size and shared_buffers. This is a text I like (it's for pg 7.4 but still useful): http://www.varlena.com/varlena/GeneralBits/Tidbits/perf.html -- /Dennis Bj�rklund From pgsql-performance-owner@postgresql.org Tue Aug 30 04:05:22 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 100C2D77EE for ; Tue, 30 Aug 2005 04:05:22 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 48069-01 for ; Tue, 30 Aug 2005 07:05:19 +0000 (GMT) Received: from floppy.pyrenet.fr (floppy.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id A4F14D7270 for ; Tue, 30 Aug 2005 04:05:15 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id B6C2F30952; Tue, 30 Aug 2005 09:05:16 +0200 (MET DST) From: William Yu X-Newsgroups: pgsql.performance Subject: Re: shared buffers Date: Tue, 30 Aug 2005 00:05:15 -0700 Organization: Hub.Org Networking Services Lines: 13 Message-ID: References: <4313B5F3.9000400@familyhealth.com.au> <20050830015454.54152.qmail@web53204.mail.yahoo.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 8bit X-Complaints-To: usenet@news.hub.org User-Agent: Mozilla Thunderbird 1.0 (X11/20041206) X-Accept-Language: en-us, en In-Reply-To: <20050830015454.54152.qmail@web53204.mail.yahoo.com> To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[AWL=0.024] X-Spam-Level: X-Archive-Number: 200508/510 X-Sequence-Number: 14257 Carlos Henrique Reimer wrote: > I forgot to say that it�s a 12GB database... > > Ok, I�ll set shared buffers to 30.000 pages but even so "meminfo" and > "top" shouldn�t show some shared pages? > > I heard something about that Redhat 9 can�t handle very well RAM higher > than 2GB. Is it right? > Thanks in advance! RH9, like any 32-bit OS, is limited to 2GB address space w/o special tricks. However, it can access > 2GB for the OS disk cache using PAE if you are running the bigmem kernel. From pgsql-performance-owner@postgresql.org Tue Aug 30 09:37:23 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4A8EBD8420 for ; Tue, 30 Aug 2005 09:37:22 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 48925-02 for ; Tue, 30 Aug 2005 12:37:17 +0000 (GMT) Received: from pelego.atua.com.br (unknown [200.248.138.60]) by svr1.postgresql.org (Postfix) with ESMTP id 42EA1D806D for ; Tue, 30 Aug 2005 09:37:16 -0300 (ADT) Received: from [10.0.0.171] (unknown [10.0.0.171]) by pelego.atua.com.br (Postfix) with ESMTP id 7381A13F657 for ; Tue, 30 Aug 2005 09:37:17 -0300 (BRT) Message-ID: <431452FD.3090402@atua.com.br> Date: Tue, 30 Aug 2005 09:37:17 -0300 From: Alvaro Nunes Melo User-Agent: Debian Thunderbird 1.0 (X11/20050116) X-Accept-Language: en-us, en MIME-Version: 1.0 To: PostgreSQL - Performance Subject: RAID Configuration Sugestion Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.027 required=5 tests=[AWL=0.027] X-Spam-Level: X-Archive-Number: 200508/511 X-Sequence-Number: 14258 Hello, We are about to install a new PostgreSQL server, and despite of being a very humble configuration compared to the ones we see in the list, it's the biggest one we've got till now. The server is a Dual Xeon 3.0 with 2 GB RAM and two SCSI disks. Our main doubt is what is the best configuration for the disks. We are thinking about use them in a RAID-0 array. Is this the best option? What do you suggest on partitioning? Separate partitions for the OS, data and pg_xlog? We'll have some time to work on performance tests, and if someone is interested we can provide our results. Thanks in advance, Alvaro From pgsql-performance-owner@postgresql.org Tue Aug 30 09:58:28 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3476BD843C for ; Tue, 30 Aug 2005 09:58:26 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49832-10 for ; Tue, 30 Aug 2005 12:58:21 +0000 (GMT) Received: from wolff.to (wolff.to [66.93.197.194]) by svr1.postgresql.org (Postfix) with SMTP id 9939FD8471 for ; Tue, 30 Aug 2005 09:58:19 -0300 (ADT) Received: (qmail 24048 invoked by uid 500); 30 Aug 2005 12:58:08 -0000 Date: Tue, 30 Aug 2005 07:58:08 -0500 From: Bruno Wolff III To: Alvaro Nunes Melo Cc: PostgreSQL - Performance Subject: Re: RAID Configuration Sugestion Message-ID: <20050830125808.GD22545@wolff.to> Mail-Followup-To: Bruno Wolff III , Alvaro Nunes Melo , PostgreSQL - Performance References: <431452FD.3090402@atua.com.br> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <431452FD.3090402@atua.com.br> User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/512 X-Sequence-Number: 14259 On Tue, Aug 30, 2005 at 09:37:17 -0300, Alvaro Nunes Melo wrote: > > The server is a Dual Xeon 3.0 with 2 GB RAM and two SCSI disks. Our main > doubt is what is the best configuration for the disks. We are thinking > about use them in a RAID-0 array. Is this the best option? What do you > suggest on partitioning? Separate partitions for the OS, data and pg_xlog? You don't have a lot of options with just two disks. What are you trying to accomplish with raid? Raid 0 will possibly give you some speed up, while raid 1 will give you some fault tolerance, some speed of of reads, but cost you half your disk space. From pgsql-performance-owner@postgresql.org Tue Aug 30 10:05:18 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id ADA13D84EB for ; Tue, 30 Aug 2005 10:05:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 51434-10 for ; Tue, 30 Aug 2005 13:05:13 +0000 (GMT) Received: from mail2.bona.us (mail2.bona.us [216.235.153.18]) by svr1.postgresql.org (Postfix) with SMTP id E1037D84DF for ; Tue, 30 Aug 2005 10:05:11 -0300 (ADT) Received: (qmail 7113 invoked by uid 399); 30 Aug 2005 13:10:58 -0000 Received: from unknown (HELO Akshay) (210.212.168.94) by mail2.bona.us with SMTP; 30 Aug 2005 13:10:58 -0000 From: "Akshay Mathur" To: Subject: Observation about db response time Date: Tue, 30 Aug 2005 18:35:30 +0530 Message-ID: <003401c5ad63$8420d940$9701a8c0@pune.wibhu.com> MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0035_01C5AD91.9DD91540" X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook, Build 10.0.2627 Importance: Normal X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.19 required=5 tests=[HTML_90_100=0.189, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/513 X-Sequence-Number: 14260 This is a multi-part message in MIME format. ------=_NextPart_000_0035_01C5AD91.9DD91540 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Hello Friends, We were having a database in pgsql7.4.2 The database was responding very slowly even after full vacuum analyze (select count(*) from some_table_having_18000_records was taking 18 Sec). We took a backup of that db and restored it back. Now the same db on same PC is responding fast (same query is taking 18 ms). But we can't do the same as a solution of slow response. Do anybody has faced similar problem? Is this due to any internal problem of pgsql? Is there any clue to fasten the database? Regards, akshay --------------------------------------- Akshay Mathur SMTS, Product Verification AirTight Networks, Inc. ( www.airtightnetworks.net) O: +91 20 2588 1555 ext 205 F: +91 20 2588 1445 ------=_NextPart_000_0035_01C5AD91.9DD91540 Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable

Hello Friends,

 

We were having a database in pgsql7.4.2 The database = was responding very slowly even after full vacuum analyze (select count(*) from some_table_having_18000_records was = taking 18 Sec).

 

We took a backup of that db and restored it back. Now = the same db on same PC is responding fast (same query is taking 18 = ms).

 

But we can't do the same as a solution of slow = response. Do anybody has faced similar problem? Is this due to any internal problem = of pgsql? Is there any clue to fasten the = database?

 

Regards,

 

akshay<= font size=3D2 face=3DArial>

 

 

-----------------------------------= ----

Akshay Mathur

SMTS, Product = Verification

AirTight = Networks, Inc. (= www.airtightnetworks.net)

O: +91 20 2588 1555 ext = 205

F: +91 20 2588 1445

 

------=_NextPart_000_0035_01C5AD91.9DD91540-- From pgsql-performance-owner@postgresql.org Tue Aug 30 10:15:10 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 46800D8466 for ; Tue, 30 Aug 2005 10:12:43 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 58059-01 for ; Tue, 30 Aug 2005 13:12:37 +0000 (GMT) Received: from frank.wiles.org (frank.wiles.org [24.124.39.75]) by svr1.postgresql.org (Postfix) with ESMTP id 8F21BD7837 for ; Tue, 30 Aug 2005 10:12:35 -0300 (ADT) Received: from kungfu (frank.wiles.org [127.0.0.1]) by frank.wiles.org (8.13.1/8.13.1) with SMTP id j7UDCbOR007840; Tue, 30 Aug 2005 08:12:37 -0500 Date: Tue, 30 Aug 2005 08:13:22 -0500 From: Frank Wiles To: "Akshay Mathur" Cc: pgsql-performance@postgresql.org Subject: Re: Observation about db response time Message-Id: <20050830081322.20a13d4f.frank@wiles.org> In-Reply-To: <003401c5ad63$8420d940$9701a8c0@pune.wibhu.com> References: <003401c5ad63$8420d940$9701a8c0@pune.wibhu.com> X-Mailer: Sylpheed version 1.0.4 (GTK+ 1.2.10; i386-redhat-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.018 required=5 tests=[AWL=0.018] X-Spam-Level: X-Archive-Number: 200508/514 X-Sequence-Number: 14261 On Tue, 30 Aug 2005 18:35:30 +0530 "Akshay Mathur" wrote: > Hello Friends, > > We were having a database in pgsql7.4.2 The database was responding > very slowly even after full vacuum analyze (select count(*) from > some_table_having_18000_records was taking 18 Sec). > > We took a backup of that db and restored it back. Now the same db on > same PC is responding fast (same query is taking 18 ms). > > But we can't do the same as a solution of slow response. Do anybody > has faced similar problem? Is this due to any internal problem of > pgsql? Is there any clue to fasten the database? This could be because you don't have max_fsm_pages and max_fsm_relations setup correctly or are not doing full vacuums often enough. If your database deletes a ton of data as a matter of course then sometimes a full vacuum will not clear up as much space as it could. Try increasing those configuration values and doing vacuums more often. If you should also explore upgrading to the latest 8.0 as you will no doubt see noticeable speed improvements. --------------------------------- Frank Wiles http://www.wiles.org --------------------------------- From pgsql-performance-owner@postgresql.org Tue Aug 30 10:50:35 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 514A1D8420 for ; Tue, 30 Aug 2005 10:50:32 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 65939-02 for ; Tue, 30 Aug 2005 13:50:31 +0000 (GMT) Received: from wolff.to (wolff.to [66.93.197.194]) by svr1.postgresql.org (Postfix) with SMTP id 3FEDED845E for ; Tue, 30 Aug 2005 10:50:29 -0300 (ADT) Received: (qmail 29287 invoked by uid 500); 30 Aug 2005 13:50:18 -0000 Date: Tue, 30 Aug 2005 08:50:18 -0500 From: Bruno Wolff III To: Alvaro Nunes Melo Cc: pgsql-performance@postgresql.org Subject: Re: RAID Configuration Sugestion Message-ID: <20050830135018.GA27789@wolff.to> Mail-Followup-To: Bruno Wolff III , Alvaro Nunes Melo , pgsql-performance@postgresql.org References: <431452FD.3090402@atua.com.br> <20050830125808.GD22545@wolff.to> <43145BE1.9050306@atua.com.br> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <43145BE1.9050306@atua.com.br> User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/515 X-Sequence-Number: 14262 Please keep replies copied to the list so that others may contribute to and learn from the discussion. On Tue, Aug 30, 2005 at 10:15:13 -0300, Alvaro Nunes Melo wrote: > Hello Bruno, > > Bruno Wolff III wrote: > > >On Tue, Aug 30, 2005 at 09:37:17 -0300, > > Alvaro Nunes Melo wrote: > > > > > >>The server is a Dual Xeon 3.0 with 2 GB RAM and two SCSI disks. Our main > >>doubt is what is the best configuration for the disks. We are thinking > >>about use them in a RAID-0 array. Is this the best option? What do you > >>suggest on partitioning? Separate partitions for the OS, data and pg_xlog? > > > Our main goal is performance speedup. Disk space might not be a problem. > I've read a lot here about movig pg_xlog to different partitions, and > we'll surely make tests to see what configuration might be better. This isn't a very good mix of hardware for running postgres. Xeons have some context switching issues for which you will probably see some speed up in 8.1. (So if you aren't going into production for sevral months you might want to be using 8.1beta.) Having only two disk drives is also not a good idea. With what you have you either want to use raid 0 and not worry too much about how the disks are partitioned or use one disk for wal logging and the other for other stuff. There are other people on the list who can probably give you a better idea of which of these options is likely to be better in your case. However, they may need to know more about your raid controller. In particular how much battery backed memory does it have and its model. From pgsql-performance-owner@postgresql.org Tue Aug 30 11:04:50 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 70347D84BD for ; Tue, 30 Aug 2005 11:04:48 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 69939-01 for ; Tue, 30 Aug 2005 14:04:45 +0000 (GMT) Received: from rwcrmhc12.comcast.net (rwcrmhc12.comcast.net [204.127.198.43]) by svr1.postgresql.org (Postfix) with ESMTP id 4C47AD84AE for ; Tue, 30 Aug 2005 11:04:44 -0300 (ADT) Received: from jefftrout.com ([24.147.120.205]) by comcast.net (rwcrmhc12) with SMTP id <2005083014044701400aq57se>; Tue, 30 Aug 2005 14:04:48 +0000 Received: (qmail 31849 invoked from network); 30 Aug 2005 14:05:46 -0000 Received: from c-24-147-120-205.hsd1.ma.comcast.net (HELO ?192.168.0.106?) (24.147.120.205) by 192.168.0.109 with SMTP; 30 Aug 2005 14:05:46 -0000 In-Reply-To: <11273.1125338254@sss.pgh.pa.us> References: <26476.1125259249@sss.pgh.pa.us> <486EB962-5B80-49B4-AF01-0C2DD77CC9EF@torgo.978.org> <11273.1125338254@sss.pgh.pa.us> Mime-Version: 1.0 (Apple Message framework v734) Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: Cc: postgres performance Content-Transfer-Encoding: 7bit From: Jeff Trout Subject: Re: OSX & Performance Date: Tue, 30 Aug 2005 10:04:44 -0400 To: Tom Lane X-Mailer: Apple Mail (2.734) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/516 X-Sequence-Number: 14263 On Aug 29, 2005, at 1:57 PM, Tom Lane wrote: > > You must have CFLAGS set to empty in your build environment, because > configure will certainly default to -O2 if not overridden. It works > fine for me on OS X. Maybe you want to trace through the configure > script and see why it's doing something else? > /me hangs head in shame. Yes. I'd been futzing with various settings and had CFLAGS set to empty instead of cleared out. 8.0.3 and -snapshot (8/29) both seem to now compile with -O2 Anyway, I tried putting together a nice self-data-producing test case but that didn't cause the bug. So I'm trying to get this dump as small as possible (I'll email you a url later). To tide things over, here's the gprof (and shark) output for my query of doom. linux box: 6.36 0.41 0.41 240694 0.00 0.00 _bt_compare 5.97 0.79 0.38 907242 0.00 0.00 AllocSetAlloc 4.55 1.07 0.29 135008 0.00 0.00 hash_any 4.16 1.34 0.27 185684 0.00 0.00 MemoryContextAllocZeroAlig ned 3.30 1.55 0.21 39152 0.01 0.01 localsub 2.98 1.74 0.19 1213172 0.00 0.00 AllocSetFreeIndex 2.83 1.92 0.18 52695 0.00 0.00 nocachegetattr 2.75 2.10 0.17 134775 0.00 0.00 hash_search 2.51 2.25 0.16 47646 0.00 0.01 StrategyBufferLookup 2.28 2.40 0.14 71990 0.00 0.00 fmgr_isbuiltin 2.20 2.54 0.14 33209 0.00 0.00 _bt_moveright 1.88 2.66 0.12 78864 0.00 0.00 comparetup_heap 1.57 2.76 0.10 63485 0.00 0.00 SearchCatCache 1.41 2.85 0.09 39152 0.00 0.00 timesub 1.26 2.93 0.08 325246 0.00 0.00 tas 1.26 3.01 0.08 305883 0.00 0.00 AllocSetFree 1.26 3.09 0.08 162622 0.00 0.00 LWLockAcquire and on osx: (self, total, library, func) 29.0% 29.0% postmaster _bt_checkkeys 15.6% 15.6% postmaster FunctionCall2 10.4% 10.4% libSystem.B.dylib __isnand 9.5% 9.5% postmaster timestamp_cmp_internal 9.3% 9.3% postmaster _bt_step 5.3% 5.3% postmaster timestamp_le 4.9% 4.9% postmaster _bt_next 3.6% 3.6% postmaster dyld_stub___isnand 3.1% 3.1% postmaster timestamp_gt 1.9% 1.9% postmaster int4eq 1.3% 1.3% postmaster BufferGetBlockNumber 0.6% 0.6% postmaster LWLockAcquire 0.5% 0.5% postmaster LWLockRelease 0.4% 0.4% postmaster hash_search On my failed simulated attempt here's what things looked liek (the data should have been relatively similar). linux: 5.39 0.28 0.28 852086 0.00 0.00 AllocSetAlloc 4.90 0.53 0.25 130165 0.00 0.00 hash_any 4.12 0.73 0.21 214061 0.00 0.00 _bt_compare 4.12 0.94 0.21 39152 0.01 0.01 localsub 4.02 1.15 0.20 160487 0.00 0.00 MemoryContextAllocZeroAlig ned 3.24 1.31 0.17 1157316 0.00 0.00 AllocSetFreeIndex 3.14 1.48 0.16 64375 0.00 0.00 fmgr_isbuiltin 2.55 1.60 0.13 56142 0.00 0.00 SearchCatCache 2.35 1.73 0.12 130076 0.00 0.00 hash_search 1.76 1.81 0.09 39152 0.00 0.00 timesub 1.67 1.90 0.09 221469 0.00 0.00 timestamp_cmp_internal 1.67 1.99 0.09 56069 0.00 0.00 MemoryContextCreate 1.57 2.06 0.08 145787 0.00 0.00 LWLockRelease 1.37 2.13 0.07 289119 0.00 0.00 pfree 1.37 2.21 0.07 8002 0.01 0.02 ExecMakeFunctionResult 1.37 2.27 0.07 8000 0.01 0.22 ExecInitIndexScan 1.18 2.33 0.06 291574 0.00 0.00 tas and on osx: (which runs very fast, usually a couple hundred ms faster than the linux box) 5.9% 5.9% postmaster LWLockAcquire 5.2% 5.2% postmaster AllocSetAlloc 4.9% 4.9% postmaster LWLockRelease 3.9% 3.9% postmaster hash_any 3.6% 3.6% postmaster _bt_compare 2.9% 2.9% postmaster hash_search 2.6% 2.6% postmaster MemoryContextAllocZeroAligned 2.6% 2.6% postmaster ExecInitExpr 2.0% 2.0% mach_kernel ml_set_interrupts_enabled 2.0% 2.0% postmaster fmgr_info_cxt_security 2.0% 2.0% postmaster AllocSetFree 1.6% 1.6% postmaster MemoryContextAlloc 1.6% 1.6% postmaster FunctionCall2 1.6% 1.6% postmaster AllocSetDelete 1.6% 1.6% libSystem.B.dylib __isnand which to me anyway, looks like basically the same profile. So there must be something about the exact nature of hte data that is kicking it in the nuts. I tried making a copy of hte table using select into, I get the same performace. Clustered on the index.. same hting. The table is a timestamp (no tz), 2 ints and 4 doubles. The index is on (timestamp, int1) As I said before, I'll send a url along to the dump once it has dumped and I get it somewhere good (unless I get my test data generator to invoke this problem). I could also get you access to this machine, but be warned gprof on tiger is pretty useless from what I've seen. -- Jeff Trout http://www.jefftrout.com/ http://www.stuarthamm.net/ From pgsql-performance-owner@postgresql.org Tue Aug 30 11:46:09 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A3C5BD84E6 for ; Tue, 30 Aug 2005 11:45:20 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78523-02 for ; Tue, 30 Aug 2005 14:45:16 +0000 (GMT) Received: from smtpauth04.mail.atl.earthlink.net (smtpauth04.mail.atl.earthlink.net [209.86.89.64]) by svr1.postgresql.org (Postfix) with ESMTP id B768CD84A0 for ; Tue, 30 Aug 2005 11:45:14 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=TFmJe7fdFb8vD7RUpIffqUaCxGGLKGHfE3NdlDibNdqre8A8MbcoBZeC4S588Rlx; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth04.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1EA7MV-0004BF-17; Tue, 30 Aug 2005 10:45:19 -0400 Message-Id: <6.2.3.4.0.20050830103240.0219b640@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Tue, 30 Aug 2005 10:45:15 -0400 To: Alvaro Nunes Melo , PostgreSQL - Performance From: Ron Subject: Re: RAID Configuration Sugestion In-Reply-To: <431452FD.3090402@atua.com.br> References: <431452FD.3090402@atua.com.br> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bc08336fd781258abe18091a7a0d203f53350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.418 required=5 tests=[AWL=0.044, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/517 X-Sequence-Number: 14264 At 08:37 AM 8/30/2005, Alvaro Nunes Melo wrote: >Hello, > >We are about to install a new PostgreSQL server, and despite of >being a very humble configuration compared to the ones we see in the >list, it's the biggest one we've got till now. > >The server is a Dual Xeon 3.0 with 2 GB RAM and two SCSI disks. Our >main doubt is what is the best configuration for the disks. We are >thinking about use them in a RAID-0 array. Is this the best option? >What do you suggest on partitioning? Separate partitions for the OS, >data and pg_xlog? This is _very_ modest HW. Unless your DB and/or DB load is similarly modest, you are not going to be happy with the performance of your DBMS. At a minimum, for safety reasons you want 4 HDs: 2 for a RAID 1 set for the DB, and 2 for a RAID 1 set for the OS + pg_xlog. 2 extra HDs, even SCSI HDs, is cheap. Especially when compared to the cost of corrupted or lost data. HD's and RAM are cheap enough that you should be able to upgrade in more ways, but do at least that "upgrade"! Beyond that, the best ways to spend you limited $ are highly dependent on your exact DB and its usage pattern. Ron Peacetree From pgsql-performance-owner@postgresql.org Tue Aug 30 11:53:49 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9F0D9D84F7 for ; Tue, 30 Aug 2005 11:53:48 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 77735-08 for ; Tue, 30 Aug 2005 14:53:47 +0000 (GMT) Received: from yertle.kcilink.com (yertle.kcilink.com [65.205.34.180]) by svr1.postgresql.org (Postfix) with ESMTP id D92FDD84F5 for ; Tue, 30 Aug 2005 11:53:45 -0300 (ADT) Received: from [192.168.7.103] (host-103.int.kcilink.com [192.168.7.103]) by yertle.kcilink.com (Postfix) with ESMTP id 21372B80F for ; Tue, 30 Aug 2005 10:53:49 -0400 (EDT) In-Reply-To: <003401c5ad63$8420d940$9701a8c0@pune.wibhu.com> References: <003401c5ad63$8420d940$9701a8c0@pune.wibhu.com> Mime-Version: 1.0 (Apple Message framework v734) X-Priority: 3 (Normal) Content-Type: multipart/alternative; boundary=Apple-Mail-4--1069943359 Message-Id: <4A58CB9B-F92A-40F9-99B8-CD783B1B71C2@khera.org> From: Vivek Khera Subject: Re: Observation about db response time Date: Tue, 30 Aug 2005 10:53:48 -0400 To: Postgresql Performance X-Mailer: Apple Mail (2.734) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.147 required=5 tests=[AWL=-0.113, HTML_60_70=0.027, HTML_FONT_BIG=0.232, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/518 X-Sequence-Number: 14265 --Apple-Mail-4--1069943359 Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed On Aug 30, 2005, at 9:05 AM, Akshay Mathur wrote: > We were having a database in pgsql7.4.2 The database was responding > very slowly even after full vacuum analyze (select count(*) from > some_table_having_18000_records was taking 18 Sec). On a 7.4.2 db, there should probably be no index bloat, but there could be. Does REINDEX on your tables help? If not, then VACUUM FULL followed by REINDEX may help. The latter should result in nearly the same as your dump+restore. And you need to run vacuum often enough to keep your tables from bloating. How often that is depends on your update/delete rate. Also, updating to 8.0 may help. Vivek Khera, Ph.D. +1-301-869-4449 x806 --Apple-Mail-4--1069943359 Content-Transfer-Encoding: quoted-printable Content-Type: text/html; charset=ISO-8859-1
On Aug 30, 2005, = at 9:05 AM, Akshay Mathur wrote:

We were having a database in pgsql7.4.2 The = database was responding very slowly even after full vacuum analyze = (select *) from = some_table_having_18000_records was taking 18 = Sec).

On a 7.4.2 db, there = should probably be no index bloat, but there could be.=A0 Does REINDEX = on your tables help?=A0 If not, then VACUUM FULL followed by REINDEX may = help.=A0 The latter should result in nearly the same as your = dump+restore.=A0 And you need to run vacuum often enough to keep your = tables from bloating.=A0 How often that is depends on your update/delete = rate.

Also,= updating to 8.0 may help.

Vivek Khera, Ph.D.

+1-301-869-4449 x806



= --Apple-Mail-4--1069943359-- From pgsql-performance-owner@postgresql.org Tue Aug 30 13:11:32 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7D6B0D8556 for ; Tue, 30 Aug 2005 13:11:04 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 98564-04 for ; Tue, 30 Aug 2005 16:10:59 +0000 (GMT) Received: from xproxy.gmail.com (xproxy.gmail.com [66.249.82.204]) by svr1.postgresql.org (Postfix) with ESMTP id 2566FD8547 for ; Tue, 30 Aug 2005 13:10:58 -0300 (ADT) Received: by xproxy.gmail.com with SMTP id i31so797060wxd for ; Tue, 30 Aug 2005 09:10:58 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=r++98p1IMH1CtuOt3v7kzpPpjD1MWC8Vd5udydXdXUC0T9QmWGG6TovY6lhBsju8gW1UDgbCpdZRCx93USLLCKYjOdHM+tjO61gb8LxC1/sGnB25j6KrKJVmanQqFALswo5jNwoGAyW2dU4H3mF+iLT9dCm+TxpNFTWKOdL0cdI= Received: by 10.70.90.15 with SMTP id n15mr133462wxb; Tue, 30 Aug 2005 09:10:58 -0700 (PDT) Received: by 10.70.128.19 with HTTP; Tue, 30 Aug 2005 09:10:58 -0700 (PDT) Message-ID: Date: Tue, 30 Aug 2005 11:10:58 -0500 From: Matthew Nuzum Reply-To: newz@bearfruit.org To: pgsql-performance@postgresql.org Subject: Re: RAID Configuration Sugestion In-Reply-To: <20050830135018.GA27789@wolff.to> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <431452FD.3090402@atua.com.br> <20050830125808.GD22545@wolff.to> <43145BE1.9050306@atua.com.br> <20050830135018.GA27789@wolff.to> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/522 X-Sequence-Number: 14269 > > >On Tue, Aug 30, 2005 at 09:37:17 -0300, > > > Alvaro Nunes Melo wrote: > > >>The server is a Dual Xeon 3.0 with 2 GB RAM and two SCSI disks. Our m= ain > > >>doubt is what is the best configuration for the disks. We are thinkin= g > > >>about use them in a RAID-0 array. Is this the best option? What do yo= u > > >>suggest on partitioning? Separate partitions for the OS, data and pg_= xlog? > > > > > Our main goal is performance speedup. Disk space might not be a problem= . > > I've read a lot here about movig pg_xlog to different partitions, and > > we'll surely make tests to see what configuration might be better. >=20 I've set up several servers with a config like this. Its not ideal, but there's no reason you can't enjoy the benefits of a snappy application. The best results I've had involve dedicating one drive to OS, swap, logs, tmp and everything and dedicate one drive to postgres. If you use *nix you can mount the second drive as /var/lib/pgsql (or where ever postgres lives on your server) with noatime as a mount option. In retrospect, you might have saved the money on the second CPU and gotten two more hard drives, but if you're running a dual task server (i.e. LAMP) you may appreciate the second CPU. The beauty of a server like this is that it puts more of the wizardry of creating a fast application into the hands of the app developer, which results in a better db schema, optimized queries and generally *thinking* about the performance of the code. I personally feel that to be a very rewarding aspect of my job. (As a hobby I program microntrollers that run at 4MHz and have only 256 bytes of RAM, so that could just be me.;-) --=20 Matthew Nuzum www.bearfruit.org From pgsql-performance-owner@postgresql.org Tue Aug 30 13:29:46 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3EF32D8547 for ; Tue, 30 Aug 2005 13:15:55 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 96235-10 for ; Tue, 30 Aug 2005 16:15:53 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 5426AD850C for ; Tue, 30 Aug 2005 13:15:52 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7UGFp8g022418; Tue, 30 Aug 2005 12:15:51 -0400 (EDT) To: =?ISO-8859-1?Q?R=E9my_Beaumont?= Cc: pgsql-performance@postgresql.org Subject: Re: High load and iowait but no disk access In-reply-to: References: Comments: In-reply-to =?ISO-8859-1?Q?R=E9my_Beaumont?= message dated "Mon, 29 Aug 2005 09:42:46 -0400" Date: Tue, 30 Aug 2005 12:15:51 -0400 Message-ID: <22417.1125418551@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/526 X-Sequence-Number: 14273 =?ISO-8859-1?Q?R=E9my_Beaumont?= writes: > The stats of the NetApp do confirm that it is sitting idle. Really? > CPU NFS CIFS HTTP Total Net kB/s Disk kB/s Tape kB/s > Cache Cache CP CP Disk DAFS FCP iSCSI FCP kB/s > in out read write read write > age hit time ty util in out > 2% 0 0 0 139 0 0 2788 0 0 0 > 3 96% 0% - 15% 0 139 0 3 2277 > 2% 0 0 0 144 0 0 2504 0 0 0 > 3 96% 0% - 18% 0 144 0 3 2150 > 2% 0 0 0 130 0 0 2212 0 0 0 > 3 96% 0% - 13% 0 130 0 3 1879 > 3% 0 0 0 169 0 0 2937 80 0 0 > 3 96% 0% - 13% 0 169 0 4 2718 > 2% 0 0 0 139 0 0 2448 0 0 0 > 3 96% 0% - 12% 0 139 0 3 2096 I know zip about NetApps, but doesn't the 8th column indicate pretty steady disk reads? regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 30 13:21:26 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 37195D85A5 for ; Tue, 30 Aug 2005 13:19:32 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 99240-04 for ; Tue, 30 Aug 2005 16:19:29 +0000 (GMT) Received: from iad2.emailsrvr.com (iad2.emailsrvr.com [207.97.227.212]) by svr1.postgresql.org (Postfix) with ESMTP id 692E6D85A2 for ; Tue, 30 Aug 2005 13:19:28 -0300 (ADT) Received: from [10.10.20.22] (unknown [216.94.157.147]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (No client certificate requested) (Authenticated sender: remyb@medrium.com) by relay2.r2.iad.emailsrvr.com (SMTP Server) with ESMTP id 6348344C36C; Tue, 30 Aug 2005 12:19:28 -0400 (EDT) In-Reply-To: <22417.1125418551@sss.pgh.pa.us> References: <22417.1125418551@sss.pgh.pa.us> Mime-Version: 1.0 (Apple Message framework v622) Content-Type: text/plain; charset=ISO-8859-1; format=flowed Message-Id: <2fa5224f3e5af53598e33d115a2af980@medrium.com> Content-Transfer-Encoding: quoted-printable Cc: pgsql-performance@postgresql.org From: =?ISO-8859-1?Q?R=E9my_Beaumont?= Subject: Re: High load and iowait but no disk access Date: Tue, 30 Aug 2005 12:19:30 -0400 To: Tom Lane X-Mailer: Apple Mail (2.622) X-Virus-Scanned: OK X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/524 X-Sequence-Number: 14271 On 30-Aug-05, at 12:15, Tom Lane wrote: > =3D?ISO-8859-1?Q?R=3DE9my_Beaumont?=3D writes: >> The stats of the NetApp do confirm that it is sitting idle. > > Really? > >> CPU NFS CIFS HTTP Total Net kB/s Disk kB/s Tape kB/s >> Cache Cache CP CP Disk DAFS FCP iSCSI FCP kB/s >> in out read write read write >> age hit time ty util in out >> 2% 0 0 0 139 0 0 2788 0 0 0 >> 3 96% 0% - 15% 0 139 0 3 2277 >> 2% 0 0 0 144 0 0 2504 0 0 0 >> 3 96% 0% - 18% 0 144 0 3 2150 >> 2% 0 0 0 130 0 0 2212 0 0 0 >> 3 96% 0% - 13% 0 130 0 3 1879 >> 3% 0 0 0 169 0 0 2937 80 0 0 >> 3 96% 0% - 13% 0 169 0 4 2718 >> 2% 0 0 0 139 0 0 2448 0 0 0 >> 3 96% 0% - 12% 0 139 0 3 2096 > > I know zip about NetApps, but doesn't the 8th column indicate pretty > steady disk reads? Yes, but they are very low. At 15% usage, it's pretty much sitting idle if you consider that the OS=20= reports that one of the processor is spending more then 80% of it's=20 time in IOwait. R=E9my > > regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 30 13:26:03 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 76BE3D85B0 for ; Tue, 30 Aug 2005 13:25:34 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 99012-07 for ; Tue, 30 Aug 2005 16:25:31 +0000 (GMT) Received: from vms048pub.verizon.net (vms048pub.verizon.net [206.46.252.48]) by svr1.postgresql.org (Postfix) with ESMTP id 59926D85BA for ; Tue, 30 Aug 2005 13:25:27 -0300 (ADT) Received: from osgiliath.mathom.us ([151.200.14.55]) by vms048.mailsrvcs.net (Sun Java System Messaging Server 6.2-2.05 (built Apr 28 2005)) with ESMTPA id <0IM100E7PMYE2E40@vms048.mailsrvcs.net> for pgsql-performance@postgresql.org; Tue, 30 Aug 2005 11:25:27 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by osgiliath.mathom.us (Postfix) with ESMTP id A973F60040E for ; Tue, 30 Aug 2005 12:25:25 -0400 (EDT) Received: from osgiliath.mathom.us ([127.0.0.1]) by localhost (osgiliath [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 03023-04 for ; Tue, 30 Aug 2005 12:25:25 -0400 (EDT) Received: by osgiliath.mathom.us (Postfix, from userid 1000) id 377DD60041D; Tue, 30 Aug 2005 12:25:25 -0400 (EDT) Date: Tue, 30 Aug 2005 12:25:25 -0400 From: Michael Stone Subject: Re: High load and iowait but no disk access In-reply-to: To: pgsql-performance@postgresql.org Mail-Followup-To: pgsql-performance@postgresql.org Message-id: <20050830162525.GU14921@mathom.us> MIME-version: 1.0 Content-type: text/plain; charset=iso-8859-1; format=flowed Content-transfer-encoding: 8BIT Content-disposition: inline X-Pgp-Fingerprint: 53 FF 38 00 E7 DD 0A 9C 84 52 84 C5 EE DF 7C 88 X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mathom.us References: User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.04 required=5 tests=[AWL=-0.010, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/525 X-Sequence-Number: 14272 On Mon, Aug 29, 2005 at 09:42:46AM -0400, R�my Beaumont wrote: >We have been trying to pinpoint what originally seem to be a I/O >bottleneck but which now seems to be an issue with either Postgresql or >RHES 3. Nope, it's an IO bottleneck. >The behavior we see is that when running queries that do random reads >on disk, IOWAIT goes over 80% and actual disk IO falls to a crawl at a >throughput bellow 3000kB/s That's the sign of an IO bottleneck. >The stats of the NetApp do confirm that it is sitting idle. Doing an >strace on the Postgresql process shows that is it doing seeks and >reads. > >So my question is where is this iowait time spent ? Waiting for the seeks. postgres doesn't do async io, so it requests a block, waits for it to come in, then requests another block, etc. The utilization on the netapp isn't going to be high because it doesn't have a queue of requests and can't do readahead because the IO is random. The only way to improve the situation would be to reduce the latency of the seeks. If I read the numbers right you're only getting about 130 seeks/s, which ain't great. I don't know how much latency the netapp adds in the this configuration; have you tried benchmarking direct-attach disks? Mike Stone From pgsql-performance-owner@postgresql.org Tue Aug 30 13:31:23 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5381CD84DF for ; Tue, 30 Aug 2005 13:29:06 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 01229-06 for ; Tue, 30 Aug 2005 16:29:04 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 76F68D8507 for ; Tue, 30 Aug 2005 13:29:03 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7UGT3X2022551; Tue, 30 Aug 2005 12:29:03 -0400 (EDT) To: =?ISO-8859-1?Q?R=E9my_Beaumont?= Cc: pgsql-performance@postgresql.org Subject: Re: High load and iowait but no disk access In-reply-to: <2fa5224f3e5af53598e33d115a2af980@medrium.com> References: <22417.1125418551@sss.pgh.pa.us> <2fa5224f3e5af53598e33d115a2af980@medrium.com> Comments: In-reply-to =?ISO-8859-1?Q?R=E9my_Beaumont?= message dated "Tue, 30 Aug 2005 12:19:30 -0400" Date: Tue, 30 Aug 2005 12:29:02 -0400 Message-ID: <22550.1125419342@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/527 X-Sequence-Number: 14274 =?ISO-8859-1?Q?R=E9my_Beaumont?= writes: > On 30-Aug-05, at 12:15, Tom Lane wrote: >> I know zip about NetApps, but doesn't the 8th column indicate pretty >> steady disk reads? > Yes, but they are very low. Sure, but that's more or less what you'd expect if the thing is randomly seeking all over the disk :-(. Just because it's a NetApp doesn't mean it's got zero seek time. You did not say what sort of query this is, but I gather that it's doing an indexscan on a table that is not at all in index order. Possible solutions involve reverting to a seqscan (have you forced the planner to choose an indexscan here, either directly or by lowering random_page_cost?) or CLUSTERing the table by the index (which would need to be repeated periodically, so it's not a great answer). regards, tom lane From pgsql-performance-owner@postgresql.org Tue Sep 6 01:12:59 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A9494D85B4 for ; Tue, 30 Aug 2005 13:39:25 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 03743-06 for ; Tue, 30 Aug 2005 16:39:22 +0000 (GMT) Received: from gghcwest.com (adsl-71-128-90-172.dsl.pltn13.pacbell.net [71.128.90.172]) by svr1.postgresql.org (Postfix) with ESMTP id 8F47CD85AE for ; Tue, 30 Aug 2005 13:39:20 -0300 (ADT) Received: from kane (pc111.gghcwest.com [192.168.168.111]) by gghcwest.com (8.12.10/8.12.9) with ESMTP id j7UGdGmd003954; Tue, 30 Aug 2005 09:39:16 -0700 Received: from jwb by kane with local (Exim 3.36 #1 (Debian)) id 1EA98n-0000jF-00; Tue, 30 Aug 2005 09:39:17 -0700 Subject: Re: Observation about db response time From: "Jeffrey W. Baker" To: Frank Wiles Cc: Akshay Mathur , pgsql-performance@postgresql.org In-Reply-To: <20050830081322.20a13d4f.frank@wiles.org> References: <003401c5ad63$8420d940$9701a8c0@pune.wibhu.com> <20050830081322.20a13d4f.frank@wiles.org> Content-Type: text/plain Content-Transfer-Encoding: 7bit Date: Tue, 30 Aug 2005 09:39:17 -0700 Message-Id: <1125419957.2658.0.camel@kane> Mime-Version: 1.0 X-Mailer: Evolution 2.0.4 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 required=5 tests=[FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200509/81 X-Sequence-Number: 14389 On Tue, 2005-08-30 at 08:13 -0500, Frank Wiles wrote: > On Tue, 30 Aug 2005 18:35:30 +0530 > "Akshay Mathur" wrote: > > > Hello Friends, > > > > We were having a database in pgsql7.4.2 The database was responding > > very slowly even after full vacuum analyze (select count(*) from > > some_table_having_18000_records was taking 18 Sec). > > > > We took a backup of that db and restored it back. Now the same db on > > same PC is responding fast (same query is taking 18 ms). > > > > But we can't do the same as a solution of slow response. Do anybody > > has faced similar problem? Is this due to any internal problem of > > pgsql? Is there any clue to fasten the database? > > This could be because you don't have max_fsm_pages and > max_fsm_relations setup correctly or are not doing full vacuums > often enough. > > If your database deletes a ton of data as a matter of course then > sometimes a full vacuum will not clear up as much space as it could. > > Try increasing those configuration values and doing vacuums more > often. > > If you should also explore upgrading to the latest 8.0 as you will > no doubt see noticeable speed improvements. This can also be caused by index bloat. VACUUM does not clear out the index. You must use REINDEX for that. -jwb From pgsql-performance-owner@postgresql.org Tue Aug 30 13:42:19 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B3BD6D8536 for ; Tue, 30 Aug 2005 13:42:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 03120-08 for ; Tue, 30 Aug 2005 16:42:12 +0000 (GMT) Received: from iad2.emailsrvr.com (iad2.emailsrvr.com [207.97.227.212]) by svr1.postgresql.org (Postfix) with ESMTP id 0B9A7D84DF for ; Tue, 30 Aug 2005 13:42:11 -0300 (ADT) Received: from [10.10.20.22] (unknown [216.94.157.147]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (No client certificate requested) (Authenticated sender: remyb@medrium.com) by relay3.r3.iad.emailsrvr.com (SMTP Server) with ESMTP id 834F244C547; Tue, 30 Aug 2005 12:42:11 -0400 (EDT) In-Reply-To: <22550.1125419342@sss.pgh.pa.us> References: <22417.1125418551@sss.pgh.pa.us> <2fa5224f3e5af53598e33d115a2af980@medrium.com> <22550.1125419342@sss.pgh.pa.us> Mime-Version: 1.0 (Apple Message framework v622) Content-Type: text/plain; charset=ISO-8859-1; format=flowed Message-Id: Content-Transfer-Encoding: quoted-printable Cc: pgsql-performance@postgresql.org From: =?ISO-8859-1?Q?R=E9my_Beaumont?= Subject: Re: High load and iowait but no disk access Date: Tue, 30 Aug 2005 12:42:13 -0400 To: Tom Lane X-Mailer: Apple Mail (2.622) X-Virus-Scanned: OK X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/528 X-Sequence-Number: 14275 On 30-Aug-05, at 12:29, Tom Lane wrote: > =3D?ISO-8859-1?Q?R=3DE9my_Beaumont?=3D writes: >> On 30-Aug-05, at 12:15, Tom Lane wrote: >>> I know zip about NetApps, but doesn't the 8th column indicate pretty >>> steady disk reads? > >> Yes, but they are very low. > > Sure, but that's more or less what you'd expect if the thing is=20 > randomly > seeking all over the disk :-(. Just because it's a NetApp doesn't = mean > it's got zero seek time. Per NetApp, the disk utilization percentage they report does include=20 seek time, not just read/write operations. NetApp has been involved in trying to figure out what is going on and=20 their claim is that the NetApp filer is not IO bound. > > You did not say what sort of query this is, but I gather that it's=20 > doing > an indexscan on a table that is not at all in index order. Yes, most of those queries are doing an indexscan. It's a fresh=20 restore of our production database that we have vacuumed/analyzed. > Possible > solutions involve reverting to a seqscan (have you forced the planner=20= > to > choose an indexscan here, either directly or by lowering=20 > random_page_cost?) No. > or CLUSTERing the table by the index (which would need to be repeated > periodically, so it's not a great answer). Will try to cluster the tables and see if it changes anything. Still=20 doesn't explain what is going on with those seeks. Thanks, R=E9my > > regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 30 13:56:38 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6AB94D85AE for ; Tue, 30 Aug 2005 13:56:36 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 08187-04 for ; Tue, 30 Aug 2005 16:56:33 +0000 (GMT) Received: from hosting.commandprompt.com (128.commandprompt.com [207.173.200.128]) by svr1.postgresql.org (Postfix) with ESMTP id E958AD85A8 for ; Tue, 30 Aug 2005 13:56:32 -0300 (ADT) Received: from [192.168.1.100] (clbb-248.saw.net [64.146.135.248]) (authenticated bits=0) by hosting.commandprompt.com (8.13.4/8.13.4) with ESMTP id j7UGrwcg029183; Tue, 30 Aug 2005 09:53:58 -0700 Message-ID: <43148FC4.4000407@commandprompt.com> Date: Tue, 30 Aug 2005 09:56:36 -0700 From: "Joshua D. Drake" User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050727) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Ron Cc: Alvaro Nunes Melo , PostgreSQL - Performance Subject: Re: RAID Configuration Sugestion References: <431452FD.3090402@atua.com.br> <6.2.3.4.0.20050830103240.0219b640@pop.earthlink.net> In-Reply-To: <6.2.3.4.0.20050830103240.0219b640@pop.earthlink.net> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Greylist: Sender succeded SMTP AUTH authentication, not delayed by milter-greylist-1.6 (hosting.commandprompt.com [192.168.1.101]); Tue, 30 Aug 2005 09:53:59 -0700 (PDT) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.022 required=5 tests=[AWL=0.022] X-Spam-Level: X-Archive-Number: 200508/529 X-Sequence-Number: 14276 Ron wrote: > At 08:37 AM 8/30/2005, Alvaro Nunes Melo wrote: > >> Hello, >> >> We are about to install a new PostgreSQL server, and despite of being >> a very humble configuration compared to the ones we see in the list, >> it's the biggest one we've got till now. >> >> The server is a Dual Xeon 3.0 with 2 GB RAM and two SCSI disks. Our >> main doubt is what is the best configuration for the disks. We are >> thinking about use them in a RAID-0 array. Is this the best option? >> What do you suggest on partitioning? Separate partitions for the OS, >> data and pg_xlog? > > > This is _very_ modest HW. Unless your DB and/or DB load is similarly > modest, you are not going to be happy with the performance of your DBMS. Well that is a pretty blanket statement. I have many customers who happily run in less hardware that what is mentioned above. It all depends on the application itself and how the database is utilized. > At a minimum, for safety reasons you want 4 HDs: 2 for a RAID 1 set > for the DB, and 2 for a RAID 1 set for the OS + pg_xlog. > 2 extra HDs, even SCSI HDs, is cheap. Especially when compared to the > cost of corrupted or lost data. Your real test is going to be prototyping the performance you need. A single RAID 1 mirror (don't use RAID 0) may be more than enough. However based on the fact that you speced Xeons my guess is you spent money on CPUs when you should have spent money on hard drives. If you still have the budget, I would suggest considering either what Ron suggested or possibly using a 4 drive RAID 10 instead. If you can't afford to put a couple more SCSI disks it may be worth while to put a software RAID 1 with ATA disks for the OS and swap and then use straight SCSI hardware RAID 1 for the DB. That will allow you to push any swap operations off to the OS disks without sacrificing the performance and reliability of the database itself. Sincerely, Joshua D. Drake > > HD's and RAM are cheap enough that you should be able to upgrade in > more ways, but do at least that "upgrade"! > > Beyond that, the best ways to spend you limited $ are highly dependent > on your exact DB and its usage pattern. > > Ron Peacetree > > > > ---------------------------(end of broadcast)--------------------------- > TIP 1: if posting/reading through Usenet, please send an appropriate > subscribe-nomail command to majordomo@postgresql.org so that your > message can get through to the mailing list cleanly From pgsql-performance-owner@postgresql.org Tue Aug 30 14:43:28 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9451DD864B for ; Tue, 30 Aug 2005 14:43:11 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 20731-01 for ; Tue, 30 Aug 2005 17:43:07 +0000 (GMT) Received: from mail.travelamericas.com (unknown [206.130.134.147]) by svr1.postgresql.org (Postfix) with SMTP id 99DFCD85C4 for ; Tue, 30 Aug 2005 14:43:05 -0300 (ADT) Received: (qmail 7310 invoked from network); 30 Aug 2005 17:43:06 -0000 Received: from unknown (HELO ?10.0.0.253?) (10.0.0.253) by verkiel.travelamericas.com with SMTP; 30 Aug 2005 17:43:06 -0000 Message-ID: <43149AA9.4080902@travelamericas.com> Date: Tue, 30 Aug 2005 10:43:05 -0700 From: Chris Travers User-Agent: Mozilla Thunderbird 1.0 (X11/20041206) X-Accept-Language: en-us, en MIME-Version: 1.0 To: "Lenard, Rohan (Rohan)" Cc: "Thomas F. O'Connell" , pgsql-performance@postgresql.org Subject: Re: Need indexes on empty tables for good performance ? References: <2773CAC687FD5F4689F526998C7E4E5F489B7B@au3010avexu1.global.avaya.com> In-Reply-To: <2773CAC687FD5F4689F526998C7E4E5F489B7B@au3010avexu1.global.avaya.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.052 required=5 tests=[AWL=0.002, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/530 X-Sequence-Number: 14277 Lenard, Rohan (Rohan) wrote: > Actually the indexes on the child table do seem to get used - I just > wanted to make sure there was no penalty not having indexes on the > empty parent tables. > > You are right - the parent is the best way to get at the unknown > children ... Indexes are created in the inheritance process, iirc. However, index entries are not inherited, which means that index-based unique constraints don't properly get inherited. Best Wishes, Chris Travers Metatron Technology Consulting From pgsql-performance-owner@postgresql.org Tue Aug 30 15:20:30 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D7920D8622 for ; Tue, 30 Aug 2005 15:16:31 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 26678-09 for ; Tue, 30 Aug 2005 18:16:28 +0000 (GMT) Received: from smtpauth05.mail.atl.earthlink.net (smtpauth05.mail.atl.earthlink.net [209.86.89.65]) by svr1.postgresql.org (Postfix) with ESMTP id C8003D8613 for ; Tue, 30 Aug 2005 15:16:27 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=VeYw/HueHM3JX1nTD6NFsNMkG3ZCbUEFiL6GR0pAY83N0bCleGYD0e9kBK/XdGS3; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth05.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1EAAen-0000TT-Gm; Tue, 30 Aug 2005 14:16:25 -0400 Message-Id: <6.2.3.4.0.20050830132902.021a5308@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Tue, 30 Aug 2005 14:16:23 -0400 To: "Joshua D. Drake" , From: Ron Subject: Re: RAID Configuration Sugestion In-Reply-To: <43148FC4.4000407@commandprompt.com> References: <431452FD.3090402@atua.com.br> <6.2.3.4.0.20050830103240.0219b640@pop.earthlink.net> <43148FC4.4000407@commandprompt.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bce55e80f265605b9174f79ddc68331a8f350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.412 required=5 tests=[AWL=0.038, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/531 X-Sequence-Number: 14278 At 12:56 PM 8/30/2005, Joshua D. Drake wrote: >Ron wrote: > >>At 08:37 AM 8/30/2005, Alvaro Nunes Melo wrote: >> >>>Hello, >>> >>>We are about to install a new PostgreSQL server, and despite of >>>being a very humble configuration compared to the ones we see in >>>the list, it's the biggest one we've got till now. >>> >>>The server is a Dual Xeon 3.0 with 2 GB RAM and two SCSI disks. >>>Our main doubt is what is the best configuration for the disks. We >>>are thinking about use them in a RAID-0 array. Is this the best >>>option? What do you suggest on partitioning? Separate partitions >>>for the OS, data and pg_xlog? >> >> >>This is _very_ modest HW. Unless your DB and/or DB load is >>similarly modest, you are not going to be happy with the >>performance of your DBMS. > >Well that is a pretty blanket statement. I have many customers who >happily run in less hardware that what is mentioned above. >It all depends on the application itself and how the database is utilized. If your customers "run happily" on 2 HD's, then IME they have very modest DB storage and/or DB performance needs. For safety reasons, the best thing to do if you only have 2 HD's is to run them as a RAID 1 with everything on them. The slightly better performing but considerably less safe alternative is to put the OS + logs on 1 HD and the DB on the other. Any resemblance to a semi-serious OLTP load will reduce either such system to an HD IO bound one with poor IO rates. If, as above, your DBMS is bounded by the performance of one HD, then you are AT BEST getting the raw IO rate of such a device: say ~70-80MB/s in average sustained raw sequential IO. Files system overhead and any seeking behavior will rapidly reduce that number to considerably less. Consider that the CPU <-> memory IO subsystem is easily capable of ~3.2GBps. So you are talking about slowing the DB server to at most ~1/40, maybe even as little as ~1/200, its potential under such circumstances. If your DB can fit completely in RAM and/or does light duty write IO, this may not be a serious issue. OTOH, once you start using those HD's to any reasonable extent, most of the rest of the investment you've made in server HW is wasted. As I keep saying, the highest priority in purchasing a DBMS is to make sure you have enough HD IO bandwidth. RAM comes second, and CPU is a distant third. >>At a minimum, for safety reasons you want 4 HDs: 2 for a RAID 1 set >>for the DB, and 2 for a RAID 1 set for the OS + pg_xlog. >>2 extra HDs, even SCSI HDs, is cheap. Especially when compared to >>the cost of corrupted or lost data. > >Your real test is going to be prototyping the performance you need. >A single RAID 1 mirror (don't use RAID 0) may be more >than enough. However based on the fact that you speced Xeons my >guess is you spent money on CPUs when you should have >spent money on hard drives. I agree with Josh on both points. Don't use RAID 0 for persistent data unless you like losing data. Spend more on HDs and RAM and less on CPU's (fast FSB is far more important than high clock rate. In general buy the highest FSB with the slowest clock rate.). If fact, if you are that strapped for cash, exchange those 2 SCSI HD's for their $ equivalent in SATA HD's. The extra spindles will be well worth it. >If you still have the budget, I would suggest considering either >what Ron suggested or possibly using a 4 drive RAID 10 instead. IME, with only 4 HDs, it's usually better to split them them into two RAID 1's (one for the db, one for everything else including the logs) than it is to put everything on one RAID 10. YMMV. Ron Peacetree From pgsql-performance-owner@postgresql.org Tue Aug 30 15:30:30 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B203FD7F46 for ; Tue, 30 Aug 2005 15:30:29 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 31395-07 for ; Tue, 30 Aug 2005 18:30:15 +0000 (GMT) Received: from iglassmail.istructure.com (istructure.com [24.199.154.122]) by svr1.postgresql.org (Postfix) with ESMTP id 5724DD808F for ; Tue, 30 Aug 2005 15:30:12 -0300 (ADT) Received: from iglassnoc3 ([192.168.177.128]) by iglassmail.istructure.com with Microsoft SMTPSVC(6.0.3790.211); Tue, 30 Aug 2005 14:30:12 -0400 From: "Woody Woodring" To: =?iso-8859-1?Q?'R=E9my_Beaumont'?= , Subject: Re: High load and iowait but no disk access Date: Tue, 30 Aug 2005 14:30:07 -0400 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_022F_01C5AD6F.51C5CDA0" X-Mailer: Microsoft Office Outlook, Build 11.0.6353 thread-index: AcWtfP6puY7NIfQiSwWnwk9yBgadUQAD8ONw X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 In-Reply-To: Message-ID: X-OriginalArrivalTime: 30 Aug 2005 18:30:12.0574 (UTC) FILETIME=[DBE067E0:01C5AD90] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.148 required=5 tests=[AWL=-0.148, HTML_10_20=0.295, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/533 X-Sequence-Number: 14280 This is a multi-part message in MIME format. ------=_NextPart_000_022F_01C5AD6F.51C5CDA0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Have you tried a different kernel? We run with a netapp over NFS = without any issues, but we have seen high IO-wait on other Dell boxes (running = and not running postgres) and RHES 3. We have replaced a Dell PowerEdge 350 running RH 7.3 with a PE750 with more memory running RHES3 and it be = bogged down with IO waits due to syslog messages writing to the disk, the old slower server could handle it fine. I don't know if it is a Dell thing = or a RH kernel, but we try different kernels on our boxes to try to find one = that works better. We have not found one that stands out over another consistently but we have been moving away from Update 2 kernel (2.4.21-15.ELsmp) due to server lockup issues. Unfortunately we get the best disk throughput on our few remaining 7.3 boxes. =20 Woody =20 IGLASS Networks www.iglass.net _____ =20 From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of R=E9my = Beaumont Sent: Monday, August 29, 2005 9:43 AM To: pgsql-performance@postgresql.org Subject: [PERFORM] High load and iowait but no disk access We have been trying to pinpoint what originally seem to be a I/O = bottleneck but which now seems to be an issue with either Postgresql or RHES 3. We have the following test environment on which we can reproduce the problem: 1) Test System A Dell 6650 Quad Xeon Pentium 4 8 Gig of RAM OS: RHES 3 update 2 Storage: NetApp FAS270 connected using an FC card using 10 disks 2) Test System B Dell Dual Xeon Pentium III 2 Gig o RAM OS: RHES 3 update 2 Storage: NetApp FAS920 connected using an FC card using 28 disks Our Database size is around 30G.=20 The behavior we see is that when running queries that do random reads on disk, IOWAIT goes over 80% and actual disk IO falls to a crawl at a throughput bellow 3000kB/s (We usually average 40000 kB/s to 80000 kB/s = on sequential read operations on the netapps) The stats of the NetApp do confirm that it is sitting idle. Doing an = strace on the Postgresql process shows that is it doing seeks and reads. So my question is where is this iowait time spent ? Is there a way to pinpoint the problem in more details ? We are able to reproduce this behavior with Postgresql 7.4.8 and 8.0.3 I have included the output of top,vmstat,strace and systat from the = Netapp from System B while running a single query that generates this behavior. R=E9my top output: 06:27:28 up 5 days, 16:59, 6 users, load average: 1.04, 1.30, 1.01 72 processes: 71 sleeping, 1 running, 0 zombie, 0 stopped CPU states: cpu user nice system irq softirq iowait idle total 2.7% 0.0% 1.0% 0.1% 0.2% 46.0% 49.5% cpu00 0.2% 0.0% 0.2% 0.0% 0.2% 2.2% 97.2% cpu01 5.3% 0.0% 1.9% 0.3% 0.3% 89.8% 1.9% Mem: 2061696k av, 2043936k used, 17760k free, 0k shrd, 3916k buff 1566332k actv, 296648k in_d, 30504k in_c Swap: 16771584k av, 21552k used, 16750032k free 1933772k cached PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND 30960 postgres 15 0 13424 10M 9908 D 2.7 0.5 2:00 1 postmaster 30538 root 15 0 1080 764 524 S 0.7 0.0 0:43 0 sshd 1 root 15 0 496 456 436 S 0.0 0.0 0:08 0 init 2 root RT 0 0 0 0 SW 0.0 0.0 0:00 0 migration/0 3 root RT 0 0 0 0 SW 0.0 0.0 0:00 1 migration/1 4 root 15 0 0 0 0 SW 0.0 0.0 0:01 0 keventd 5 root 34 19 0 0 0 SWN 0.0 0.0 0:00 0 ksoftirqd/0 6 root 34 19 0 0 0 SWN 0.0 0.0 0:00 1 ksoftirqd/1 9 root 15 0 0 0 0 SW 0.0 0.0 0:24 1 bdflush 7 root 15 0 0 0 0 SW 0.0 0.0 6:53 1 kswapd 8 root 15 0 0 0 0 SW 0.0 0.0 8:44 1 kscand 10 root 15 0 0 0 0 SW 0.0 0.0 0:13 0 kupdated 11 root 25 0 0 0 0 SW 0.0 0.0 0:00 0 mdrecoveryd 17 root 15 0 0 0 0 SW 0.0 0.0 0:00 0 ahc_dv_0 vmstat output=20 procs memory swap io system cpu r b swpd free buff cache si so bi bo in cs us sy id wa 0 1 21552 17796 4872 1931928 2 3 3 1 27 6 2 1 7 3 0 1 21552 18044 4880 1931652 0 0 1652 0 397 512 1 2 50 47 0 1 21552 17976 4896 1931664 0 0 2468 0 407 552 2 2 50 47 1 0 21552 17984 4896 1931608 0 0 2124 0 418 538 3 3 48 46 0 1 21552 18028 4900 1931536 0 0 1592 0 385 509 1 3 50 46 0 1 21552 18040 4916 1931488 0 0 1620 820 419 581 2 2 50 46 0 1 21552 17968 4916 1931536 0 4 1708 4 402 554 3 1 50 46 1 1 21552 18052 4916 1931388 0 0 1772 0 409 531 3 1 49 47 0 1 21552 17912 4924 1931492 0 0 1772 0 408 565 3 1 48 48 0 1 21552 17932 4932 1931440 0 4 1356 4 391 545 5 0 49 46 0 1 21552 18320 4944 1931016 0 4 1500 840 414 571 1 1 48 50 0 1 21552 17872 4944 1931440 0 0 2116 0 392 496 1 5 46 48 0 1 21552 18060 4944 1931232 0 0 2232 0 423 597 1 2 48 49 1 1 21552 17684 4944 1931584 0 0 1752 0 395 537 1 1 50 48 0 1 21552 18000 4944 1931240 0 0 1576 0 401 549 0 1 50 49 NetApp stats: CPU NFS CIFS HTTP Total Net kB/s Disk kB/s Tape kB/s Cache Cache CP CP = Disk DAFS FCP iSCSI FCP kB/s in out read write read write age hit time ty util in out 2% 0 0 0 139 0 0 2788 0 0 0 3 96% 0% - 15% 0 139 0 3 2277 2% 0 0 0 144 0 0 2504 0 0 0 3 96% 0% - 18% 0 144 0 3 2150 2% 0 0 0 130 0 0 2212 0 0 0 3 96% 0% - 13% 0 130 0 3 1879 3% 0 0 0 169 0 0 2937 80 0 0 3 96% 0% - 13% 0 169 0 4 2718 2% 0 0 0 139 0 0 2448 0 0 0 3 96% 0% - 12% 0 139 0 3 2096 2% 0 0 0 137 0 0 2116 0 0 0 3 96% 0% - 10% 0 137 0 3 1892 3% 0 0 0 107 0 0 2660 812 0 0 3 96% 24% T 20% 0 107 0 3 1739 2% 0 0 0 118 0 0 1788 0 0 0 3 96% 0% - 13% 0 118 0 3 1608 2% 0 0 0 136 0 0 2228 0 0 0 3 96% 0% - 11% 0 136 0 3 2018 2% 0 0 0 119 0 0 1940 0 0 0 3 96% 0% - 13% 0 119 0 3 1998 2% 0 0 0 136 0 0 2175 0 0 0 3 96% 0% - 14% 0 136 0 3 1929 2% 0 0 0 133 0 0 1924 0 0 0 3 96% 0% - 19% 0 133 0 3 2292 2% 0 0 0 115 0 0 2044 0 0 0 3 96% 0% - 11% 0 115 0 3 1682 2% 0 0 0 134 0 0 2256 0 0 0 3 96% 0% - 12% 0 134 0 3 2096 2% 0 0 0 112 0 0 2184 0 0 0 3 96% 0% - 12% 0 112 0 3 1633 2% 0 0 0 163 0 0 2348 0 0 0 3 96% 0% - 13% 0 163 0 4 2421 2% 0 0 0 120 0 0 2056 184 0 0 3 96% 8% T 14% 0 120 0 3 1703 strace output: read(55, "\4\0\0\0\10fm}\1\0\0\0p\0\264\0\0 \2 \230\236\320\0020"..., = 8192) =3D 8192 _llseek(55, 857997312, [857997312], SEEK_SET) =3D 0 read(55, "\4\0\0\0\\\315\321|\1\0\0\0p\0\354\0\0 \2 \250\236\260"..., = 8192) =3D 8192 _llseek(55, 883220480, [883220480], SEEK_SET) =3D 0 read(55, "\4\0\0\0T\17a~\1\0\0\0p\0\20\1\0 \2 \270\236\220\2D\235"..., = 8192) =3D 8192 _llseek(55, 858005504, [858005504], SEEK_SET) =3D 0 read(55, "\4\0\0\0\300\356\321|\1\0\0\0p\0\330\0\0 \2 \260\236\240"..., 8192) =3D 8192 _llseek(55, 857964544, [857964544], SEEK_SET) =3D 0 read(55, "\4\0\0\0lH\321|\1\0\0\0p\0<\1\0 \2 \300\236\200\2p\235"..., = 8192) =3D 8192 _llseek(55, 857956352, [857956352], SEEK_SET) =3D 0 read(55, "\4\0\0\0l\'\321|\1\0\0\0p\0\320\0\0 \2 \260\236\240\2\\"..., = 8192) =3D 8192 _llseek(55, 910802944, [910802944], SEEK_SET) =3D 0 read(55, "\4\0\0\0\10}\25\200\1\0\0\0l\0\274\1\0 \2 \250\236\260"..., = 8192) =3D 8192 _llseek(55, 857948160, [857948160], SEEK_SET) =3D 0 read(55, "\4\0\0\0\370\5\321|\1\0\0\0p\0\350\0\0 \2 \230\236\320"..., = 8192) =3D 8192 _llseek(56, 80371712, [80371712], SEEK_SET) =3D 0 read(56, "\4\0\0\0Lf \217\1\0\0\0p\0\f\1\0 \2 \250\236\260\2T\235"..., = 8192) =3D 8192 read(102, "\2\0\34\0001\236\0\0\1\0\0\0\t\0\0\00020670\0\0\0B\6\0"..., = 8192) =3D 8192 _llseek(55, 857939968, [857939968], SEEK_SET) =3D 0 read(55, "\4\0\0\0\244\344\320|\1\0\0\0l\0\230\1\0 \2 \244\236\270"..., 8192) =3D 8192 _llseek(55, 857923584, [857923584], SEEK_SET) =3D 0 read(55, "\4\0\0\0\224\242\320|\1\0\0\0p\0|\0\0 \2 \234\236\310\002"..., 8192) =3D 8192 _llseek(55, 57270272, [57270272], SEEK_SET) =3D 0 read(55, "\4\0\0\0\3204FK\1\0\0\0t\0\340\0\0 \2 \310\236j\2\214\235"..., 8192) =3D 8192 _llseek(55, 870727680, [870727680], SEEK_SET) =3D 0 read(55, "\4\0\0\0x>\233}\1\0\0\0p\0@\1\0 \2 \250\236\260\2X\235"..., = 8192) =3D 8192 _llseek(55, 1014734848, [1014734848], SEEK_SET) =3D 0 read(55, "\4\0\0\0\34\354\201\206\1\0\0\0p\0p\0\0 \2 \264\236\230"..., = 8192) =3D 8192 _llseek(55, 857874432, [857874432], SEEK_SET) =3D 0 read(55, "\4\0\0\0\214\331\317|\1\0\0\0l\0\324\1\0 \2 \224\236\330"..., 8192) =3D 8192 _llseek(55, 760872960, [760872960], SEEK_SET) =3D 0 read(55, "\4\0\0\0\30\257\321v\1\0\0\0p\0\230\0\0 \2 \234\236\310"..., = 8192) =3D 8192 _llseek(55, 891715584, [891715584], SEEK_SET) =3D 0 read(55, "\4\0\0\0\370\220\347~\1\0\0\0p\0P\1\0 \2 \230\236\320\2"..., = 8192) =3D 8192 _llseek(55, 857858048, [857858048], SEEK_SET) =3D 0 read(55, "\4\0\0\0\250\227\317|\1\0\0\0p\0\264\0\0 \2 \254\236\250"..., 8192) =3D 8192 _llseek(55, 666910720, [666910720], SEEK_SET) =3D 0 read(55, "\4\0\0\0x\206\3q\1\0\0\0p\0004\1\0 \2 \254\236\242\2P\235"..., 8192) =3D 8192 _llseek(55, 857841664, [857841664], SEEK_SET) =3D 0 read(55, "\4\0\0\0dT\317|\1\0\0\0p\0\224\0\0 \2 \214\236\350\2\30"..., = 8192) =3D 8192 ------=_NextPart_000_022F_01C5AD6F.51C5CDA0 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Have you tried a different kernel?  We run = with a=20 netapp over NFS without any issues, but we have seen high IO-wait on = other Dell=20 boxes (running  and not running postgres) and RHES 3.  We have = replaced a Dell PowerEdge 350 running RH 7.3  with a PE750 with = more memory=20 running RHES3 and it be bogged down with IO waits due to syslog messages = writing=20 to the disk, the old slower server could handle it fine.  I don't = know if=20 it is a Dell thing or a RH kernel, but we try different kernels on our = boxes to=20 try to find one that works better.  We have not found one that = stands out=20 over another consistently but we have been moving away from Update = 2 kernel=20 (2.4.21-15.ELsmp) due to server lockup issues.  Unfortunately we = get the=20 best disk throughput on our few remaining 7.3 boxes.
 
Woody
 
IGLASS Networks

=
From: = pgsql-performance-owner@postgresql.org=20 [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of = R=E9my=20 Beaumont
Sent: Monday, August 29, 2005 9:43 AM
To:=20 pgsql-performance@postgresql.org
Subject: [PERFORM] High load = and=20 iowait but no disk access

We have been trying to pinpoint what originally seem to be a = I/O=20 bottleneck but which now seems to be an issue with either Postgresql or = RHES=20 3.

We have the following test environment on which we can = reproduce the=20 problem:

1) Test System A
Dell 6650 Quad Xeon Pentium 4
8 = Gig of=20 RAM
OS: RHES 3 update 2
Storage: NetApp FAS270 connected using an = FC card=20 using 10 disks

2) Test System B
Dell Dual Xeon Pentium = III
2 Gig o=20 RAM
OS: RHES 3 update 2
Storage: NetApp FAS920 connected using an = FC card=20 using 28 disks

Our Database size is around 30G.

The = behavior we=20 see is that when running queries that do random reads on disk, IOWAIT = goes over=20 80% and actual disk IO falls to a crawl at a throughput bellow 3000kB/s = (We=20 usually average 40000 kB/s to 80000 kB/s on sequential read operations = on the=20 netapps)

The stats of the NetApp do confirm that it is sitting = idle.=20 Doing an strace on the Postgresql process shows that is it doing seeks = and=20 reads.

So my question is where is this iowait time spent ?
Is = there a=20 way to pinpoint the problem in more details ?
We are able to = reproduce this=20 behavior with Postgresql 7.4.8 and 8.0.3

I have included the = output of=20 top,vmstat,strace and systat from the Netapp from System B while running = a=20 single query that generates this behavior.

R=E9my

top=20 output:
06:27:28 up 5 days, 16:59, 6 users, load average: 1.04, 1.30, = 1.01
72 processes: 71 sleeping, 1 running, 0 zombie, 0 stopped
CPU = states:=20 cpu user nice system irq softirq iowait idle
total 2.7% 0.0% 1.0% = 0.1% 0.2%=20 46.0% 49.5%
cpu00 0.2% 0.0% 0.2% 0.0% 0.2% 2.2% 97.2%
cpu01 5.3% = 0.0% 1.9%=20 0.3% 0.3% 89.8% 1.9%
Mem: 2061696k av, 2043936k used, 17760k free, 0k = shrd,=20 3916k buff
1566332k actv, 296648k in_d, 30504k in_c
Swap: = 16771584k av,=20 21552k used, 16750032k free 1933772k cached

PID USER PRI NI SIZE = RSS=20 SHARE STAT %CPU %MEM TIME CPU COMMAND
30960 postgres 15 0 13424 10M = 9908 D=20 2.7 0.5 2:00 1 postmaster
30538 root 15 0 1080 764 524 S 0.7 0.0 0:43 = 0=20 sshd
1 root 15 0 496 456 436 S 0.0 0.0 0:08 0 init
2 root RT 0 0 0 = 0 SW=20 0.0 0.0 0:00 0 migration/0
3 root RT 0 0 0 0 SW 0.0 0.0 0:00 1=20 migration/1
4 root 15 0 0 0 0 SW 0.0 0.0 0:01 0 keventd
5 root 34 = 19 0 0 0=20 SWN 0.0 0.0 0:00 0 ksoftirqd/0
6 root 34 19 0 0 0 SWN 0.0 0.0 0:00 1=20 ksoftirqd/1
9 root 15 0 0 0 0 SW 0.0 0.0 0:24 1 bdflush
7 root 15 = 0 0 0 0=20 SW 0.0 0.0 6:53 1 kswapd
8 root 15 0 0 0 0 SW 0.0 0.0 8:44 1 = kscand
10=20 root 15 0 0 0 0 SW 0.0 0.0 0:13 0 kupdated
11 root 25 0 0 0 0 SW 0.0 = 0.0 0:00=20 0 mdrecoveryd
17 root 15 0 0 0 0 SW 0.0 0.0 0:00 0 = ahc_dv_0


vmstat=20 output
procs memory swap io system cpu
r b swpd free buff cache = si so bi=20 bo in cs us sy id wa
0 1 21552 17796 4872 1931928 2 3 3 1 27 6 2 1 7 = 3
0 1=20 21552 18044 4880 1931652 0 0 1652 0 397 512 1 2 50 47
0 1 21552 17976 = 4896=20 1931664 0 0 2468 0 407 552 2 2 50 47
1 0 21552 17984 4896 1931608 0 0 = 2124 0=20 418 538 3 3 48 46
0 1 21552 18028 4900 1931536 0 0 1592 0 385 509 1 3 = 50=20 46
0 1 21552 18040 4916 1931488 0 0 1620 820 419 581 2 2 50 46
0 1 = 21552=20 17968 4916 1931536 0 4 1708 4 402 554 3 1 50 46
1 1 21552 18052 4916 = 1931388=20 0 0 1772 0 409 531 3 1 49 47
0 1 21552 17912 4924 1931492 0 0 1772 0 = 408 565=20 3 1 48 48
0 1 21552 17932 4932 1931440 0 4 1356 4 391 545 5 0 49 = 46
0 1=20 21552 18320 4944 1931016 0 4 1500 840 414 571 1 1 48 50
0 1 21552 = 17872 4944=20 1931440 0 0 2116 0 392 496 1 5 46 48
0 1 21552 18060 4944 1931232 0 0 = 2232 0=20 423 597 1 2 48 49
1 1 21552 17684 4944 1931584 0 0 1752 0 395 537 1 1 = 50=20 48
0 1 21552 18000 4944 1931240 0 0 1576 0 401 549 0 1 50=20 49


NetApp stats:
CPU NFS CIFS HTTP Total Net kB/s Disk = kB/s Tape=20 kB/s Cache Cache CP CP Disk DAFS FCP iSCSI FCP kB/s
in out read write = read=20 write age hit time ty util in out
2% 0 0 0 139 0 0 2788 0 0 0 3 96% = 0% - 15%=20 0 139 0 3 2277
2% 0 0 0 144 0 0 2504 0 0 0 3 96% 0% - 18% 0 144 0 3=20 2150
2% 0 0 0 130 0 0 2212 0 0 0 3 96% 0% - 13% 0 130 0 3 1879
3% = 0 0 0=20 169 0 0 2937 80 0 0 3 96% 0% - 13% 0 169 0 4 2718
2% 0 0 0 139 0 0 = 2448 0 0 0=20 3 96% 0% - 12% 0 139 0 3 2096
2% 0 0 0 137 0 0 2116 0 0 0 3 96% 0% - = 10% 0=20 137 0 3 1892
3% 0 0 0 107 0 0 2660 812 0 0 3 96% 24% T 20% 0 107 0 3=20 1739
2% 0 0 0 118 0 0 1788 0 0 0 3 96% 0% - 13% 0 118 0 3 1608
2% = 0 0 0=20 136 0 0 2228 0 0 0 3 96% 0% - 11% 0 136 0 3 2018
2% 0 0 0 119 0 0 = 1940 0 0 0=20 3 96% 0% - 13% 0 119 0 3 1998
2% 0 0 0 136 0 0 2175 0 0 0 3 96% 0% - = 14% 0=20 136 0 3 1929
2% 0 0 0 133 0 0 1924 0 0 0 3 96% 0% - 19% 0 133 0 3 = 2292
2%=20 0 0 0 115 0 0 2044 0 0 0 3 96% 0% - 11% 0 115 0 3 1682
2% 0 0 0 134 0 = 0 2256=20 0 0 0 3 96% 0% - 12% 0 134 0 3 2096
2% 0 0 0 112 0 0 2184 0 0 0 3 96% = 0% -=20 12% 0 112 0 3 1633
2% 0 0 0 163 0 0 2348 0 0 0 3 96% 0% - 13% 0 163 0 = 4=20 2421
2% 0 0 0 120 0 0 2056 184 0 0 3 96% 8% T 14% 0 120 0 3=20 1703

strace output:
read(55, = "\4\0\0\0\10fm}\1\0\0\0p\0\264\0\0 \2=20 \230\236\320\0020"..., 8192) =3D 8192
_llseek(55, 857997312, = [857997312],=20 SEEK_SET) =3D 0
read(55, "\4\0\0\0\\\315\321|\1\0\0\0p\0\354\0\0 \2=20 \250\236\260"..., 8192) =3D 8192
_llseek(55, 883220480, [883220480], = SEEK_SET)=20 =3D 0
read(55, "\4\0\0\0T\17a~\1\0\0\0p\0\20\1\0 \2 = \270\236\220\2D\235"...,=20 8192) =3D 8192
_llseek(55, 858005504, [858005504], SEEK_SET) =3D = 0
read(55,=20 "\4\0\0\0\300\356\321|\1\0\0\0p\0\330\0\0 \2 \260\236\240"..., 8192) =3D = 8192
_llseek(55, 857964544, [857964544], SEEK_SET) =3D 0
read(55,=20 "\4\0\0\0lH\321|\1\0\0\0p\0<\1\0 \2 \300\236\200\2p\235"..., 8192) = =3D=20 8192
_llseek(55, 857956352, [857956352], SEEK_SET) =3D 0
read(55,=20 "\4\0\0\0l\'\321|\1\0\0\0p\0\320\0\0 \2 \260\236\240\2\\"..., 8192) =3D=20 8192
_llseek(55, 910802944, [910802944], SEEK_SET) =3D 0
read(55,=20 "\4\0\0\0\10}\25\200\1\0\0\0l\0\274\1\0 \2 \250\236\260"..., 8192) =3D=20 8192
_llseek(55, 857948160, [857948160], SEEK_SET) =3D 0
read(55,=20 "\4\0\0\0\370\5\321|\1\0\0\0p\0\350\0\0 \2 \230\236\320"..., 8192) =3D=20 8192
_llseek(56, 80371712, [80371712], SEEK_SET) =3D 0
read(56, = "\4\0\0\0Lf=20 \217\1\0\0\0p\0\f\1\0 \2 \250\236\260\2T\235"..., 8192) =3D = 8192
read(102,=20 "\2\0\34\0001\236\0\0\1\0\0\0\t\0\0\00020670\0\0\0B\6\0"..., 8192) =3D=20 8192
_llseek(55, 857939968, [857939968], SEEK_SET) =3D 0
read(55,=20 "\4\0\0\0\244\344\320|\1\0\0\0l\0\230\1\0 \2 \244\236\270"..., 8192) =3D = 8192
_llseek(55, 857923584, [857923584], SEEK_SET) =3D 0
read(55,=20 "\4\0\0\0\224\242\320|\1\0\0\0p\0|\0\0 \2 \234\236\310\002"..., 8192) = =3D=20 8192
_llseek(55, 57270272, [57270272], SEEK_SET) =3D 0
read(55,=20 "\4\0\0\0\3204FK\1\0\0\0t\0\340\0\0 \2 \310\236j\2\214\235"..., 8192) = =3D=20 8192
_llseek(55, 870727680, [870727680], SEEK_SET) =3D 0
read(55,=20 "\4\0\0\0x>\233}\1\0\0\0p\0@\1\0 \2 \250\236\260\2X\235"..., 8192) = =3D=20 8192
_llseek(55, 1014734848, [1014734848], SEEK_SET) =3D = 0
read(55,=20 "\4\0\0\0\34\354\201\206\1\0\0\0p\0p\0\0 \2 \264\236\230"..., 8192) =3D=20 8192
_llseek(55, 857874432, [857874432], SEEK_SET) =3D 0
read(55,=20 "\4\0\0\0\214\331\317|\1\0\0\0l\0\324\1\0 \2 \224\236\330"..., 8192) =3D = 8192
_llseek(55, 760872960, [760872960], SEEK_SET) =3D 0
read(55,=20 "\4\0\0\0\30\257\321v\1\0\0\0p\0\230\0\0 \2 \234\236\310"..., 8192) =3D=20 8192
_llseek(55, 891715584, [891715584], SEEK_SET) =3D 0
read(55,=20 "\4\0\0\0\370\220\347~\1\0\0\0p\0P\1\0 \2 \230\236\320\2"..., 8192) =3D=20 8192
_llseek(55, 857858048, [857858048], SEEK_SET) =3D 0
read(55,=20 "\4\0\0\0\250\227\317|\1\0\0\0p\0\264\0\0 \2 \254\236\250"..., 8192) =3D = 8192
_llseek(55, 666910720, [666910720], SEEK_SET) =3D 0
read(55,=20 "\4\0\0\0x\206\3q\1\0\0\0p\0004\1\0 \2 \254\236\242\2P\235"..., 8192) = =3D=20 8192
_llseek(55, 857841664, [857841664], SEEK_SET) =3D 0
read(55,=20 "\4\0\0\0dT\317|\1\0\0\0p\0\224\0\0 \2 \214\236\350\2\30"..., 8192) =3D=20 8192


------=_NextPart_000_022F_01C5AD6F.51C5CDA0-- From pgsql-performance-owner@postgresql.org Tue Aug 30 15:29:29 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id EEB71D8633 for ; Tue, 30 Aug 2005 15:28:48 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 29130-06 for ; Tue, 30 Aug 2005 18:28:44 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id DC68FD85A0 for ; Tue, 30 Aug 2005 15:28:43 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [64.81.245.111] (account josh@agliodbs.com HELO temoku.sf.agliodbs.com) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7995263; Tue, 30 Aug 2005 11:31:01 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: High load and iowait but no disk access Date: Tue, 30 Aug 2005 11:32:06 -0700 User-Agent: KMail/1.8 Cc: =?utf-8?q?R=C3=A9my_Beaumont?= References: In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200508301132.06904.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.048 required=5 tests=[AWL=-0.002, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/532 X-Sequence-Number: 14279 Remy, > The behavior we see is that when running queries that do random reads > on disk, IOWAIT goes over 80% and actual disk IO falls to a crawl at a > throughput bellow 3000kB/s (We usually average 40000 kB/s to 80000 kB/s > on sequential read operations on the netapps) This seems pretty low for a NetApp -- you should be able to manage up to 180mb/s, if not higher. Are you sure it's configured correctly? -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Tue Aug 30 15:49:02 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E688BD7BC4 for ; Tue, 30 Aug 2005 15:37:35 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 32771-07 for ; Tue, 30 Aug 2005 18:37:32 +0000 (GMT) Received: from ll.mit.edu (LLMAIL.LL.MIT.EDU [129.55.12.40]) by svr1.postgresql.org (Postfix) with ESMTP id 944A5D7B71 for ; Tue, 30 Aug 2005 15:37:30 -0300 (ADT) Received: (from smtp@localhost) by ll.mit.edu (8.12.10/8.8.8) id j7UIbTtY003609 for ; Tue, 30 Aug 2005 14:37:29 -0400 (EDT) Received: from UNKNOWN( ), claiming to be "sty.llan" via SMTP by llmail, id smtpdAAAddaahb; Tue Aug 30 14:36:19 2005 Date: Tue, 30 Aug 2005 14:36:18 -0400 From: george young To: pgsql-performance@postgresql.org Subject: Re: Observation about db response time Message-Id: <20050830143618.2c10144d.gry@ll.mit.edu> In-Reply-To: <2527.192.168.1.175.1125158337.squirrel@mailserv> References: <2527.192.168.1.175.1125158337.squirrel@mailserv> Reply-To: gry@ll.mit.edu Organization: MIT Lincoln Laboratory X-Mailer: Sylpheed version 2.0.1 (GTK+ 2.8.0; i686-pc-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.019 required=5 tests=[AWL=0.019] X-Spam-Level: X-Archive-Number: 200508/535 X-Sequence-Number: 14282 On Sat, 27 Aug 2005 21:28:57 +0530 (IST) threw this fish to the penguins: > Hello Friends, > We were having a database in pgsql7.4. The database was responding very > slowly even after full vacuum (select > count(*) from some_table_having_18000_records was taking 18 Sec). One comment here: "select count(*)" may seem like a good benchmark, but it's not generally. If your application really depends on this number, fine. Otherwise, you should measure performance with a real query from your application. The "select count(*)" can be very slow because it does not use indexes. > We took a backup of that db and restored it back. Now the same db on > same PC is responding fast (same query is taking 18 ms). This sounds like some index is getting gooped up. If you do a lot of deleting from tables, your indexes can collect dead space that vacuum can not reclaim. Try in sql "reindex table my_slow_table" for a suspect table. In the contrib directory of the postgresql distribution there is a script called "reindexdb". You can run this to reindex your whole database. I also wonder about file system slowdowns. What hardware/OS/filesystem are you using? -- George -- "Are the gods not just?" "Oh no, child. What would become of us if they were?" (CSL) From pgsql-performance-owner@postgresql.org Tue Aug 30 15:44:03 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5E50BD7F83 for ; Tue, 30 Aug 2005 15:42:39 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 34352-06 for ; Tue, 30 Aug 2005 18:42:37 +0000 (GMT) Received: from iad2.emailsrvr.com (iad2.emailsrvr.com [207.97.227.212]) by svr1.postgresql.org (Postfix) with ESMTP id DE32ED7F46 for ; Tue, 30 Aug 2005 15:42:35 -0300 (ADT) Received: from [10.10.20.22] (unknown [216.94.157.147]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (No client certificate requested) (Authenticated sender: remyb@medrium.com) by relay3.r3.iad.emailsrvr.com (SMTP Server) with ESMTP id 11F7544C5E8; Tue, 30 Aug 2005 14:42:36 -0400 (EDT) In-Reply-To: <200508301132.06904.josh@agliodbs.com> References: <200508301132.06904.josh@agliodbs.com> Mime-Version: 1.0 (Apple Message framework v622) Content-Type: text/plain; charset=ISO-8859-1; format=flowed Message-Id: <22924e579ad7e7e0deb5133eb905618b@medrium.com> Content-Transfer-Encoding: quoted-printable Cc: pgsql-performance@postgresql.org From: =?ISO-8859-1?Q?R=E9my_Beaumont?= Subject: Re: High load and iowait but no disk access Date: Tue, 30 Aug 2005 14:42:38 -0400 To: josh@agliodbs.com X-Mailer: Apple Mail (2.622) X-Virus-Scanned: OK X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/534 X-Sequence-Number: 14281 On 30-Aug-05, at 14:32, Josh Berkus wrote: > Remy, > >> The behavior we see is that when running queries that do random reads >> on disk, IOWAIT goes over 80% and actual disk IO falls to a crawl at = a >> throughput bellow 3000kB/s (We usually average 40000 kB/s to 80000=20 >> kB/s >> on sequential read operations on the netapps) > > This seems pretty low for a NetApp -- you should be able to manage up=20= > to > 180mb/s, if not higher. Are you sure it's configured correctly? Hi Josh, The config has been reviewed by NetApp. We do get rates higher then=20 80mb/s, but on average, that's what we get. Do you have NetApp filers deployed ? How many spindles do you have in your volume ? On which OS are you running Postgres ? Thanks, R=E9my > > --=20 > --Josh > > Josh Berkus > Aglio Database Solutions > San Francisco > > ---------------------------(end of=20 > broadcast)--------------------------- > TIP 9: In versions below 8.0, the planner will ignore your desire to > choose an index scan if your joining column's datatypes do not > match From pgsql-performance-owner@postgresql.org Tue Aug 30 15:58:53 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AE028D76E9 for ; Tue, 30 Aug 2005 15:46:17 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 35668-02 for ; Tue, 30 Aug 2005 18:46:14 +0000 (GMT) Received: from vt-pe2550-001.VANTAGE.vantage.com (unknown [64.80.203.244]) by svr1.postgresql.org (Postfix) with ESMTP id E1970D6F02 for ; Tue, 30 Aug 2005 15:46:11 -0300 (ADT) X-MimeOLE: Produced By Microsoft Exchange V6.0.6249.0 content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----_=_NextPart_001_01C5AD93.16A8668F" Subject: Re: High load and iowait but no disk access Date: Tue, 30 Aug 2005 14:46:10 -0400 Message-ID: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098C1C@vt-pe2550-001.vantage.vantage.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] High load and iowait but no disk access Thread-Index: AcWtfP6puY7NIfQiSwWnwk9yBgadUQAD8ONwAAFHUkA= From: "Anjan Dave" To: "Woody Woodring" , =?iso-8859-1?Q?R=E9my_Beaumont?= , X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.059 required=5 tests=[AWL=0.002, HTML_30_40=0.056, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/538 X-Sequence-Number: 14285 This is a multi-part message in MIME format. ------_=_NextPart_001_01C5AD93.16A8668F Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable I have seen references of changing the kernel io scheduler at boot = time...not sure if it applies to RHEL3.0, or will help, but try setting = 'elevator=3Ddeadline' during boot time or via grub.conf. Have you tried = running a simple 'dd' on the LUN? The drives are in RAID10 = configuration, right? =20 Thanks, Anjan _____ =20 From: Woody Woodring [mailto:george.woodring@iglass.net]=20 Sent: Tuesday, August 30, 2005 2:30 PM To: 'R=E9my Beaumont'; pgsql-performance@postgresql.org Subject: Re: [PERFORM] High load and iowait but no disk access =20 Have you tried a different kernel? We run with a netapp over NFS = without any issues, but we have seen high IO-wait on other Dell boxes = (running and not running postgres) and RHES 3. We have replaced a Dell = PowerEdge 350 running RH 7.3 with a PE750 with more memory running = RHES3 and it be bogged down with IO waits due to syslog messages writing = to the disk, the old slower server could handle it fine. I don't know = if it is a Dell thing or a RH kernel, but we try different kernels on = our boxes to try to find one that works better. We have not found one = that stands out over another consistently but we have been moving away = from Update 2 kernel (2.4.21-15.ELsmp) due to server lockup issues. = Unfortunately we get the best disk throughput on our few remaining 7.3 = boxes. =20 Woody =20 IGLASS Networks www.iglass.net =20 _____ =20 From: pgsql-performance-owner@postgresql.org = [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of R=E9my = Beaumont Sent: Monday, August 29, 2005 9:43 AM To: pgsql-performance@postgresql.org Subject: [PERFORM] High load and iowait but no disk access We have been trying to pinpoint what originally seem to be a I/O = bottleneck but which now seems to be an issue with either Postgresql or = RHES 3. We have the following test environment on which we can reproduce the = problem: 1) Test System A Dell 6650 Quad Xeon Pentium 4 8 Gig of RAM OS: RHES 3 update 2 Storage: NetApp FAS270 connected using an FC card using 10 disks 2) Test System B Dell Dual Xeon Pentium III 2 Gig o RAM OS: RHES 3 update 2 Storage: NetApp FAS920 connected using an FC card using 28 disks Our Database size is around 30G.=20 The behavior we see is that when running queries that do random reads on = disk, IOWAIT goes over 80% and actual disk IO falls to a crawl at a = throughput bellow 3000kB/s (We usually average 40000 kB/s to 80000 kB/s = on sequential read operations on the netapps) The stats of the NetApp do confirm that it is sitting idle. Doing an = strace on the Postgresql process shows that is it doing seeks and reads. So my question is where is this iowait time spent ? Is there a way to pinpoint the problem in more details ? We are able to reproduce this behavior with Postgresql 7.4.8 and 8.0.3 I have included the output of top,vmstat,strace and systat from the = Netapp from System B while running a single query that generates this = behavior. R=E9my top output: 06:27:28 up 5 days, 16:59, 6 users, load average: 1.04, 1.30, 1.01 72 processes: 71 sleeping, 1 running, 0 zombie, 0 stopped CPU states: cpu user nice system irq softirq iowait idle total 2.7% 0.0% 1.0% 0.1% 0.2% 46.0% 49.5% cpu00 0.2% 0.0% 0.2% 0.0% 0.2% 2.2% 97.2% cpu01 5.3% 0.0% 1.9% 0.3% 0.3% 89.8% 1.9% Mem: 2061696k av, 2043936k used, 17760k free, 0k shrd, 3916k buff 1566332k actv, 296648k in_d, 30504k in_c Swap: 16771584k av, 21552k used, 16750032k free 1933772k cached PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND 30960 postgres 15 0 13424 10M 9908 D 2.7 0.5 2:00 1 postmaster 30538 root 15 0 1080 764 524 S 0.7 0.0 0:43 0 sshd 1 root 15 0 496 456 436 S 0.0 0.0 0:08 0 init 2 root RT 0 0 0 0 SW 0.0 0.0 0:00 0 migration/0 3 root RT 0 0 0 0 SW 0.0 0.0 0:00 1 migration/1 4 root 15 0 0 0 0 SW 0.0 0.0 0:01 0 keventd 5 root 34 19 0 0 0 SWN 0.0 0.0 0:00 0 ksoftirqd/0 6 root 34 19 0 0 0 SWN 0.0 0.0 0:00 1 ksoftirqd/1 9 root 15 0 0 0 0 SW 0.0 0.0 0:24 1 bdflush 7 root 15 0 0 0 0 SW 0.0 0.0 6:53 1 kswapd 8 root 15 0 0 0 0 SW 0.0 0.0 8:44 1 kscand 10 root 15 0 0 0 0 SW 0.0 0.0 0:13 0 kupdated 11 root 25 0 0 0 0 SW 0.0 0.0 0:00 0 mdrecoveryd 17 root 15 0 0 0 0 SW 0.0 0.0 0:00 0 ahc_dv_0 vmstat output=20 procs memory swap io system cpu r b swpd free buff cache si so bi bo in cs us sy id wa 0 1 21552 17796 4872 1931928 2 3 3 1 27 6 2 1 7 3 0 1 21552 18044 4880 1931652 0 0 1652 0 397 512 1 2 50 47 0 1 21552 17976 4896 1931664 0 0 2468 0 407 552 2 2 50 47 1 0 21552 17984 4896 1931608 0 0 2124 0 418 538 3 3 48 46 0 1 21552 18028 4900 1931536 0 0 1592 0 385 509 1 3 50 46 0 1 21552 18040 4916 1931488 0 0 1620 820 419 581 2 2 50 46 0 1 21552 17968 4916 1931536 0 4 1708 4 402 554 3 1 50 46 1 1 21552 18052 4916 1931388 0 0 1772 0 409 531 3 1 49 47 0 1 21552 17912 4924 1931492 0 0 1772 0 408 565 3 1 48 48 0 1 21552 17932 4932 1931440 0 4 1356 4 391 545 5 0 49 46 0 1 21552 18320 4944 1931016 0 4 1500 840 414 571 1 1 48 50 0 1 21552 17872 4944 1931440 0 0 2116 0 392 496 1 5 46 48 0 1 21552 18060 4944 1931232 0 0 2232 0 423 597 1 2 48 49 1 1 21552 17684 4944 1931584 0 0 1752 0 395 537 1 1 50 48 0 1 21552 18000 4944 1931240 0 0 1576 0 401 549 0 1 50 49 NetApp stats: CPU NFS CIFS HTTP Total Net kB/s Disk kB/s Tape kB/s Cache Cache CP CP = Disk DAFS FCP iSCSI FCP kB/s in out read write read write age hit time ty util in out 2% 0 0 0 139 0 0 2788 0 0 0 3 96% 0% - 15% 0 139 0 3 2277 2% 0 0 0 144 0 0 2504 0 0 0 3 96% 0% - 18% 0 144 0 3 2150 2% 0 0 0 130 0 0 2212 0 0 0 3 96% 0% - 13% 0 130 0 3 1879 3% 0 0 0 169 0 0 2937 80 0 0 3 96% 0% - 13% 0 169 0 4 2718 2% 0 0 0 139 0 0 2448 0 0 0 3 96% 0% - 12% 0 139 0 3 2096 2% 0 0 0 137 0 0 2116 0 0 0 3 96% 0% - 10% 0 137 0 3 1892 3% 0 0 0 107 0 0 2660 812 0 0 3 96% 24% T 20% 0 107 0 3 1739 2% 0 0 0 118 0 0 1788 0 0 0 3 96% 0% - 13% 0 118 0 3 1608 2% 0 0 0 136 0 0 2228 0 0 0 3 96% 0% - 11% 0 136 0 3 2018 2% 0 0 0 119 0 0 1940 0 0 0 3 96% 0% - 13% 0 119 0 3 1998 2% 0 0 0 136 0 0 2175 0 0 0 3 96% 0% - 14% 0 136 0 3 1929 2% 0 0 0 133 0 0 1924 0 0 0 3 96% 0% - 19% 0 133 0 3 2292 2% 0 0 0 115 0 0 2044 0 0 0 3 96% 0% - 11% 0 115 0 3 1682 2% 0 0 0 134 0 0 2256 0 0 0 3 96% 0% - 12% 0 134 0 3 2096 2% 0 0 0 112 0 0 2184 0 0 0 3 96% 0% - 12% 0 112 0 3 1633 2% 0 0 0 163 0 0 2348 0 0 0 3 96% 0% - 13% 0 163 0 4 2421 2% 0 0 0 120 0 0 2056 184 0 0 3 96% 8% T 14% 0 120 0 3 1703 strace output: read(55, "\4\0\0\0\10fm}\1\0\0\0p\0\264\0\0 \2 \230\236\320\0020"..., = 8192) =3D 8192 _llseek(55, 857997312, [857997312], SEEK_SET) =3D 0 read(55, "\4\0\0\0\\\315\321|\1\0\0\0p\0\354\0\0 \2 \250\236\260"..., = 8192) =3D 8192 _llseek(55, 883220480, [883220480], SEEK_SET) =3D 0 read(55, "\4\0\0\0T\17a~\1\0\0\0p\0\20\1\0 \2 \270\236\220\2D\235"..., = 8192) =3D 8192 _llseek(55, 858005504, [858005504], SEEK_SET) =3D 0 read(55, "\4\0\0\0\300\356\321|\1\0\0\0p\0\330\0\0 \2 \260\236\240"..., = 8192) =3D 8192 _llseek(55, 857964544, [857964544], SEEK_SET) =3D 0 read(55, "\4\0\0\0lH\321|\1\0\0\0p\0<\1\0 \2 \300\236\200\2p\235"..., = 8192) =3D 8192 _llseek(55, 857956352, [857956352], SEEK_SET) =3D 0 read(55, "\4\0\0\0l\'\321|\1\0\0\0p\0\320\0\0 \2 \260\236\240\2\\"..., = 8192) =3D 8192 _llseek(55, 910802944, [910802944], SEEK_SET) =3D 0 read(55, "\4\0\0\0\10}\25\200\1\0\0\0l\0\274\1\0 \2 \250\236\260"..., = 8192) =3D 8192 _llseek(55, 857948160, [857948160], SEEK_SET) =3D 0 read(55, "\4\0\0\0\370\5\321|\1\0\0\0p\0\350\0\0 \2 \230\236\320"..., = 8192) =3D 8192 _llseek(56, 80371712, [80371712], SEEK_SET) =3D 0 read(56, "\4\0\0\0Lf \217\1\0\0\0p\0\f\1\0 \2 \250\236\260\2T\235"..., = 8192) =3D 8192 read(102, "\2\0\34\0001\236\0\0\1\0\0\0\t\0\0\00020670\0\0\0B\6\0"..., = 8192) =3D 8192 _llseek(55, 857939968, [857939968], SEEK_SET) =3D 0 read(55, "\4\0\0\0\244\344\320|\1\0\0\0l\0\230\1\0 \2 \244\236\270"..., = 8192) =3D 8192 _llseek(55, 857923584, [857923584], SEEK_SET) =3D 0 read(55, "\4\0\0\0\224\242\320|\1\0\0\0p\0|\0\0 \2 \234\236\310\002"..., = 8192) =3D 8192 _llseek(55, 57270272, [57270272], SEEK_SET) =3D 0 read(55, "\4\0\0\0\3204FK\1\0\0\0t\0\340\0\0 \2 \310\236j\2\214\235"..., = 8192) =3D 8192 _llseek(55, 870727680, [870727680], SEEK_SET) =3D 0 read(55, "\4\0\0\0x>\233}\1\0\0\0p\0@\1\0 \2 \250\236\260\2X\235"..., = 8192) =3D 8192 _llseek(55, 1014734848, [1014734848], SEEK_SET) =3D 0 read(55, "\4\0\0\0\34\354\201\206\1\0\0\0p\0p\0\0 \2 \264\236\230"..., = 8192) =3D 8192 _llseek(55, 857874432, [857874432], SEEK_SET) =3D 0 read(55, "\4\0\0\0\214\331\317|\1\0\0\0l\0\324\1\0 \2 \224\236\330"..., = 8192) =3D 8192 _llseek(55, 760872960, [760872960], SEEK_SET) =3D 0 read(55, "\4\0\0\0\30\257\321v\1\0\0\0p\0\230\0\0 \2 \234\236\310"..., = 8192) =3D 8192 _llseek(55, 891715584, [891715584], SEEK_SET) =3D 0 read(55, "\4\0\0\0\370\220\347~\1\0\0\0p\0P\1\0 \2 \230\236\320\2"..., = 8192) =3D 8192 _llseek(55, 857858048, [857858048], SEEK_SET) =3D 0 read(55, "\4\0\0\0\250\227\317|\1\0\0\0p\0\264\0\0 \2 \254\236\250"..., = 8192) =3D 8192 _llseek(55, 666910720, [666910720], SEEK_SET) =3D 0 read(55, "\4\0\0\0x\206\3q\1\0\0\0p\0004\1\0 \2 \254\236\242\2P\235"..., = 8192) =3D 8192 _llseek(55, 857841664, [857841664], SEEK_SET) =3D 0 read(55, "\4\0\0\0dT\317|\1\0\0\0p\0\224\0\0 \2 \214\236\350\2\30"..., = 8192) =3D 8192 ------_=_NextPart_001_01C5AD93.16A8668F Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable

I have seen references of changing = the kernel io scheduler at boot time…not sure if it applies to = RHEL3.0, or will help, but try setting ‘elevator=3Ddeadline’ during boot = time or via grub.conf. Have you tried running a simple ‘dd’ on the = LUN? The drives are in RAID10 configuration, right?

 

Thanks,

=

Anjan

From: Woody = Woodring [mailto:george.woodring@iglass.net]
Sent: Tuesday, August 30, = 2005 2:30 PM
To: 'R=E9my Beaumont'; pgsql-performance@postgresql.org
Subject: Re: [PERFORM] = High load and iowait but no disk access

 

Have you tried a different = kernel?  We run with a netapp over NFS without any issues, but we have seen high = IO-wait on other Dell boxes (running  and not running postgres) and RHES = 3.  We have replaced a Dell PowerEdge 350 running RH 7.3  with a PE750 = with more memory running RHES3 and it be bogged down with IO waits due to = syslog messages writing to the disk, the old slower server could handle it = fine.  I don't know if it is a Dell thing or a RH kernel, but we try different = kernels on our boxes to try to find one that works better.  We have not = found one that stands out over another consistently but we have been moving = away from Update 2 kernel (2.4.21-15.ELsmp) due to server lockup = issues.  Unfortunately we get the best disk throughput on our few remaining 7.3 = boxes.

 

Woody

 

IGLASS = Networks

www.iglass.net

 

From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of R=E9my Beaumont
Sent: Monday, August 29, = 2005 9:43 AM
To: pgsql-performance@postgresql.org
Subject: [PERFORM] High = load and iowait but no disk access

We have been trying to pinpoint what originally seem to be a I/O bottleneck but which now seems to be an issue with either Postgresql or = RHES 3.

We have the following test environment on which we can reproduce the = problem:

1) Test System A
Dell 6650 Quad Xeon Pentium 4
8 Gig of RAM
OS: RHES 3 update 2
Storage: NetApp FAS270 connected using an FC card using 10 disks

2) Test System B
Dell Dual Xeon Pentium III
2 Gig o RAM
OS: RHES 3 update 2
Storage: NetApp FAS920 connected using an FC card using 28 disks

Our Database size is around 30G.

The behavior we see is that when running queries that do random reads on = disk, IOWAIT goes over 80% and actual disk IO falls to a crawl at a throughput = bellow 3000kB/s (We usually average 40000 kB/s to 80000 kB/s on sequential read operations on the netapps)

The stats of the NetApp do confirm that it is sitting idle. Doing an = strace on the Postgresql process shows that is it doing seeks and reads.

So my question is where is this iowait time spent ?
Is there a way to pinpoint the problem in more details ?
We are able to reproduce this behavior with Postgresql 7.4.8 and = 8.0.3

I have included the output of top,vmstat,strace and systat from the = Netapp from System B while running a single query that generates this behavior.

R=E9my

top output:
06:27:28 up 5 days, 16:59, 6 users, load average: 1.04, 1.30, 1.01
72 processes: 71 sleeping, 1 running, 0 zombie, 0 stopped
CPU states: cpu user nice system irq softirq iowait idle
total 2.7% 0.0% 1.0% 0.1% 0.2% 46.0% 49.5%
cpu00 0.2% 0.0% 0.2% 0.0% 0.2% 2.2% 97.2%
cpu01 5.3% 0.0% 1.9% 0.3% 0.3% 89.8% 1.9%
Mem: 2061696k av, 2043936k used, 17760k free, 0k shrd, 3916k buff
1566332k actv, 296648k in_d, 30504k in_c
Swap: 16771584k av, 21552k used, 16750032k free 1933772k cached

PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND
30960 postgres 15 0 13424 10M 9908 D 2.7 0.5 2:00 1 postmaster
30538 root 15 0 1080 764 524 S 0.7 0.0 0:43 0 sshd
1 root 15 0 496 456 436 S 0.0 0.0 0:08 0 init
2 root RT 0 0 0 0 SW 0.0 0.0 0:00 0 migration/0
3 root RT 0 0 0 0 SW 0.0 0.0 0:00 1 migration/1
4 root 15 0 0 0 0 SW 0.0 0.0 0:01 0 keventd
5 root 34 19 0 0 0 SWN 0.0 0.0 0:00 0 ksoftirqd/0
6 root 34 19 0 0 0 SWN 0.0 0.0 0:00 1 ksoftirqd/1
9 root 15 0 0 0 0 SW 0.0 0.0 0:24 1 bdflush
7 root 15 0 0 0 0 SW 0.0 0.0 6:53 1 kswapd
8 root 15 0 0 0 0 SW 0.0 0.0 8:44 1 kscand
10 root 15 0 0 0 0 SW 0.0 0.0 0:13 0 kupdated
11 root 25 0 0 0 0 SW 0.0 0.0 0:00 0 mdrecoveryd
17 root 15 0 0 0 0 SW 0.0 0.0 0:00 0 ahc_dv_0


vmstat output
procs memory swap io system cpu
r b swpd free buff cache si so bi bo in cs us sy id wa
0 1 21552 17796 4872 1931928 2 3 3 1 27 6 2 1 7 3
0 1 21552 18044 4880 1931652 0 0 1652 0 397 512 1 2 50 47
0 1 21552 17976 4896 1931664 0 0 2468 0 407 552 2 2 50 47
1 0 21552 17984 4896 1931608 0 0 2124 0 418 538 3 3 48 46
0 1 21552 18028 4900 1931536 0 0 1592 0 385 509 1 3 50 46
0 1 21552 18040 4916 1931488 0 0 1620 820 419 581 2 2 50 46
0 1 21552 17968 4916 1931536 0 4 1708 4 402 554 3 1 50 46
1 1 21552 18052 4916 1931388 0 0 1772 0 409 531 3 1 49 47
0 1 21552 17912 4924 1931492 0 0 1772 0 408 565 3 1 48 48
0 1 21552 17932 4932 1931440 0 4 1356 4 391 545 5 0 49 46
0 1 21552 18320 4944 1931016 0 4 1500 840 414 571 1 1 48 50
0 1 21552 17872 4944 1931440 0 0 2116 0 392 496 1 5 46 48
0 1 21552 18060 4944 1931232 0 0 2232 0 423 597 1 2 48 49
1 1 21552 17684 4944 1931584 0 0 1752 0 395 537 1 1 50 48
0 1 21552 18000 4944 1931240 0 0 1576 0 401 549 0 1 50 49


NetApp stats:
CPU NFS CIFS HTTP Total Net kB/s Disk kB/s Tape kB/s Cache Cache CP CP = Disk DAFS FCP iSCSI FCP kB/s
in out read write read write age hit time ty util in out
2% 0 0 0 139 0 0 2788 0 0 0 3 96% 0% - 15% 0 139 0 3 2277
2% 0 0 0 144 0 0 2504 0 0 0 3 96% 0% - 18% 0 144 0 3 2150
2% 0 0 0 130 0 0 2212 0 0 0 3 96% 0% - 13% 0 130 0 3 1879
3% 0 0 0 169 0 0 2937 80 0 0 3 96% 0% - 13% 0 169 0 4 2718
2% 0 0 0 139 0 0 2448 0 0 0 3 96% 0% - 12% 0 139 0 3 2096
2% 0 0 0 137 0 0 2116 0 0 0 3 96% 0% - 10% 0 137 0 3 1892
3% 0 0 0 107 0 0 2660 812 0 0 3 96% 24% T 20% 0 107 0 3 1739
2% 0 0 0 118 0 0 1788 0 0 0 3 96% 0% - 13% 0 118 0 3 1608
2% 0 0 0 136 0 0 2228 0 0 0 3 96% 0% - 11% 0 136 0 3 2018
2% 0 0 0 119 0 0 1940 0 0 0 3 96% 0% - 13% 0 119 0 3 1998
2% 0 0 0 136 0 0 2175 0 0 0 3 96% 0% - 14% 0 136 0 3 1929
2% 0 0 0 133 0 0 1924 0 0 0 3 96% 0% - 19% 0 133 0 3 2292
2% 0 0 0 115 0 0 2044 0 0 0 3 96% 0% - 11% 0 115 0 3 1682
2% 0 0 0 134 0 0 2256 0 0 0 3 96% 0% - 12% 0 134 0 3 2096
2% 0 0 0 112 0 0 2184 0 0 0 3 96% 0% - 12% 0 112 0 3 1633
2% 0 0 0 163 0 0 2348 0 0 0 3 96% 0% - 13% 0 163 0 4 2421
2% 0 0 0 120 0 0 2056 184 0 0 3 96% 8% T 14% 0 120 0 3 1703

strace output:
read(55, "\4\0\0\0\10fm}\1\0\0\0p\0\264\0\0 \2 = \230\236\320\0020"..., 8192) =3D 8192
_llseek(55, 857997312, [857997312], SEEK_SET) =3D 0
read(55, "\4\0\0\0\\\315\321|\1\0\0\0p\0\354\0\0 \2 = \250\236\260"..., 8192) =3D 8192
_llseek(55, 883220480, [883220480], SEEK_SET) =3D 0
read(55, "\4\0\0\0T\17a~\1\0\0\0p\0\20\1\0 \2 = \270\236\220\2D\235"..., 8192) =3D 8192
_llseek(55, 858005504, [858005504], SEEK_SET) =3D 0
read(55, "\4\0\0\0\300\356\321|\1\0\0\0p\0\330\0\0 \2 \260\236\240"..., 8192) =3D 8192
_llseek(55, 857964544, [857964544], SEEK_SET) =3D 0
read(55, "\4\0\0\0lH\321|\1\0\0\0p\0<\1\0 \2 = \300\236\200\2p\235"..., 8192) =3D 8192
_llseek(55, 857956352, [857956352], SEEK_SET) =3D 0
read(55, "\4\0\0\0l\'\321|\1\0\0\0p\0\320\0\0 \2 \260\236\240\2\\"..., 8192) =3D 8192
_llseek(55, 910802944, [910802944], SEEK_SET) =3D 0
read(55, "\4\0\0\0\10}\25\200\1\0\0\0l\0\274\1\0 \2 = \250\236\260"..., 8192) =3D 8192
_llseek(55, 857948160, [857948160], SEEK_SET) =3D 0
read(55, "\4\0\0\0\370\5\321|\1\0\0\0p\0\350\0\0 \2 = \230\236\320"..., 8192) =3D 8192
_llseek(56, 80371712, [80371712], SEEK_SET) =3D 0
read(56, "\4\0\0\0Lf \217\1\0\0\0p\0\f\1\0 \2 \250\236\260\2T\235"..., 8192) =3D 8192
read(102, "\2\0\34\0001\236\0\0\1\0\0\0\t\0\0\00020670\0\0\0B\6\0"..., = 8192) =3D 8192
_llseek(55, 857939968, [857939968], SEEK_SET) =3D 0
read(55, "\4\0\0\0\244\344\320|\1\0\0\0l\0\230\1\0 \2 = \244\236\270"..., 8192) =3D 8192
_llseek(55, 857923584, [857923584], SEEK_SET) =3D 0
read(55, "\4\0\0\0\224\242\320|\1\0\0\0p\0|\0\0 \2 \234\236\310\002"..., 8192) =3D 8192
_llseek(55, 57270272, [57270272], SEEK_SET) =3D 0
read(55, "\4\0\0\0\3204FK\1\0\0\0t\0\340\0\0 \2 = \310\236j\2\214\235"..., 8192) =3D 8192
_llseek(55, 870727680, [870727680], SEEK_SET) =3D 0
read(55, "\4\0\0\0x>\233}\1\0\0\0p\0@\1\0 \2 \250\236\260\2X\235"..., 8192) =3D 8192
_llseek(55, 1014734848, [1014734848], SEEK_SET) =3D 0
read(55, "\4\0\0\0\34\354\201\206\1\0\0\0p\0p\0\0 \2 \264\236\230"..., 8192) =3D 8192
_llseek(55, 857874432, [857874432], SEEK_SET) =3D 0
read(55, "\4\0\0\0\214\331\317|\1\0\0\0l\0\324\1\0 \2 \224\236\330"..., 8192) =3D 8192
_llseek(55, 760872960, [760872960], SEEK_SET) =3D 0
read(55, "\4\0\0\0\30\257\321v\1\0\0\0p\0\230\0\0 \2 \234\236\310"..., 8192) =3D 8192
_llseek(55, 891715584, [891715584], SEEK_SET) =3D 0
read(55, "\4\0\0\0\370\220\347~\1\0\0\0p\0P\1\0 \2 \230\236\320\2"..., 8192) =3D 8192
_llseek(55, 857858048, [857858048], SEEK_SET) =3D 0
read(55, "\4\0\0\0\250\227\317|\1\0\0\0p\0\264\0\0 \2 \254\236\250"..., 8192) =3D 8192
_llseek(55, 666910720, [666910720], SEEK_SET) =3D 0
read(55, "\4\0\0\0x\206\3q\1\0\0\0p\0004\1\0 \2 \254\236\242\2P\235"..., 8192) =3D 8192
_llseek(55, 857841664, [857841664], SEEK_SET) =3D 0
read(55, "\4\0\0\0dT\317|\1\0\0\0p\0\224\0\0 \2 \214\236\350\2\30"..., 8192) =3D 8192



------_=_NextPart_001_01C5AD93.16A8668F-- From pgsql-performance-owner@postgresql.org Tue Aug 30 15:54:22 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 0E20AD7F46 for ; Tue, 30 Aug 2005 15:50:24 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 34634-08 for ; Tue, 30 Aug 2005 18:50:16 +0000 (GMT) Received: from xproxy.gmail.com (xproxy.gmail.com [66.249.82.205]) by svr1.postgresql.org (Postfix) with ESMTP id 528ECD7DD9 for ; Tue, 30 Aug 2005 15:50:15 -0300 (ADT) Received: by xproxy.gmail.com with SMTP id i31so822926wxd for ; Tue, 30 Aug 2005 11:50:16 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=NQ9rOruHU2UuaX5blFauvBJI7D9xJLsWEREhkNDFUPZZ7scS6lQsrTO0bwjlxlvofn5AVYmEXyHrp1bIBAaDJ0HJONs/7T91NYnnQlMFfMHqe/ri+jrzVA8dNCGyNTtyD93/P3R/4IxG2TVvDnAnqVGcSPFqRYDx7vHlThVeJVk= Received: by 10.70.78.19 with SMTP id a19mr138440wxb; Tue, 30 Aug 2005 11:50:16 -0700 (PDT) Received: by 10.70.128.19 with HTTP; Tue, 30 Aug 2005 11:50:16 -0700 (PDT) Message-ID: Date: Tue, 30 Aug 2005 13:50:16 -0500 From: Matthew Nuzum Reply-To: newz@bearfruit.org To: pgsql-performance@postgresql.org Subject: Re: RAID Configuration Sugestion In-Reply-To: <6.2.3.4.0.20050830132902.021a5308@pop.earthlink.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline References: <431452FD.3090402@atua.com.br> <6.2.3.4.0.20050830103240.0219b640@pop.earthlink.net> <43148FC4.4000407@commandprompt.com> <6.2.3.4.0.20050830132902.021a5308@pop.earthlink.net> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.024 required=5 tests=[RCVD_BY_IP=0.024] X-Spam-Level: X-Archive-Number: 200508/537 X-Sequence-Number: 14284 On 8/30/05, Ron wrote: > >If you still have the budget, I would suggest considering either > >what Ron suggested or possibly using a 4 drive RAID 10 instead. >=20 > IME, with only 4 HDs, it's usually better to split them them into two > RAID 1's (one for the db, one for everything else including the logs) > than it is to put everything on one RAID 10. YMMV. This coresponds to what I have observed as well. Of course, we all know that work loads varry. Just a note for the OP who has only two drives, there are tools for a variety of OSs that monitor the S.M.A.R.T. features of the drive and give an early warning in case it senses impending failure. I've caught two drives before failure with these types of tools. Also note that when reading discussions of this nature you must take into consideration the value of your data. For some people, restoring from a nightly backup is inconvienent, but not life-or-death. Some people even do twice-daily backups so that in case of a failure they can recover with little loss of data. This might be a good way to mitigate the cost of expensive server hardware. If you cannot afford to lose any data then you need to consider it imperitive to use some type of RAID setup (not RAID 0) and to achieve great performance you'll want more than 2 drives. --=20 Matthew Nuzum www.bearfruit.org From pgsql-performance-owner@postgresql.org Tue Aug 30 15:54:00 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id ACA01D8581 for ; Tue, 30 Aug 2005 15:50:35 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 35972-03 for ; Tue, 30 Aug 2005 18:50:33 +0000 (GMT) Received: from iad2.emailsrvr.com (iad2.emailsrvr.com [207.97.227.212]) by svr1.postgresql.org (Postfix) with ESMTP id 66535D76E9 for ; Tue, 30 Aug 2005 15:50:32 -0300 (ADT) Received: from [10.10.20.22] (unknown [216.94.157.147]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (No client certificate requested) (Authenticated sender: remyb@medrium.com) by relay3.r3.iad.emailsrvr.com (SMTP Server) with ESMTP id D11BF44C5EE; Tue, 30 Aug 2005 14:50:32 -0400 (EDT) In-Reply-To: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098C1C@vt-pe2550-001.vantage.vantage.com> References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098C1C@vt-pe2550-001.vantage.vantage.com> Mime-Version: 1.0 (Apple Message framework v622) Content-Type: text/plain; charset=WINDOWS-1252; format=flowed Message-Id: <6bb032ec58fb35eb6e5a3cefd3dd9360@medrium.com> Content-Transfer-Encoding: quoted-printable Cc: From: =?ISO-8859-1?Q?R=E9my_Beaumont?= Subject: Re: High load and iowait but no disk access Date: Tue, 30 Aug 2005 14:50:35 -0400 To: "Anjan Dave" X-Mailer: Apple Mail (2.622) X-Virus-Scanned: OK X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/536 X-Sequence-Number: 14283 On 30-Aug-05, at 14:46, Anjan Dave wrote: > I have seen references of changing the kernel io scheduler at boot=20 > time=85not sure if it applies to RHEL3.0, or will help, but try = setting=20 > =91elevator=3Ddeadline=92 during boot time or via grub.conf. That's only for RHEL 4.0. > Have you tried running a simple =91dd=92 on the LUN? We get amazing performance using dd. > The drives are in RAID10 configuration, right? NetApp has their own type of raid format (RAID4 aka WAFL) R=E9my > =A0 > Thanks, > Anjan > > From: Woody Woodring [mailto:george.woodring@iglass.net] > Sent: Tuesday, August 30, 2005 2:30 PM > To: 'R=E9my Beaumont'; pgsql-performance@postgresql.org > Subject: Re: [PERFORM] High load and iowait but no disk access > =A0 > Have you tried a different kernel?=A0 We run with a netapp over NFS=20 > without any issues, but we have seen high IO-wait on other Dell boxes=20= > (running=A0 and not running postgres) and RHES 3.=A0 We have replaced = a=20 > Dell PowerEdge 350 running RH 7.3 =A0with a PE750 with more memory=20 > running RHES3 and it be bogged down with IO waits due to syslog=20 > messages writing to the disk, the old slower server could handle it=20 > fine.=A0 I don't know if it is a Dell thing or a RH kernel, but we try=20= > different kernels on our boxes to try to find one that works better.=A0=20= > We have not found one that stands out over another=A0consistently but = we=20 > have been moving away from Update 2 kernel (2.4.21-15.ELsmp) due to=20 > server lockup issues.=A0 Unfortunately we get the best disk throughput=20= > on our few remaining 7.3 boxes. > =A0 > Woody > =A0 > IGLASS Networks > www.iglass.net > =A0 > > > From: pgsql-performance-owner@postgresql.org=20 > [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of R=E9my=20 > Beaumont > Sent: Monday, August 29, 2005 9:43 AM > To: pgsql-performance@postgresql.org > Subject: [PERFORM] High load and iowait but no disk access > We have been trying to pinpoint what originally seem to be a I/O=20 > bottleneck but which now seems to be an issue with either Postgresql=20= > or RHES 3. > > We have the following test environment on which we can reproduce the=20= > problem: > > 1) Test System A > Dell 6650 Quad Xeon Pentium 4 > 8 Gig of RAM > OS: RHES 3 update 2 > Storage: NetApp FAS270 connected using an FC card using 10 disks > > 2) Test System B > Dell Dual Xeon Pentium III > 2 Gig o RAM > OS: RHES 3 update 2 > Storage: NetApp FAS920 connected using an FC card using 28 disks > > Our Database size is around 30G. > > The behavior we see is that when running queries that do random reads=20= > on disk, IOWAIT goes over 80% and actual disk IO falls to a crawl at a=20= > throughput bellow 3000kB/s (We usually average 40000 kB/s to 80000=20 > kB/s on sequential read operations on the netapps) > > The stats of the NetApp do confirm that it is sitting idle. Doing an=20= > strace on the Postgresql process shows that is it doing seeks and=20 > reads. > > So my question is where is this iowait time spent ? > Is there a way to pinpoint the problem in more details ? > We are able to reproduce this behavior with Postgresql 7.4.8 and = 8.0.3 > > I have included the output of top,vmstat,strace and systat from the=20= > Netapp from System B while running a single query that generates this=20= > behavior. > > R=E9my > > top output: > 06:27:28 up 5 days, 16:59, 6 users, load average: 1.04, 1.30, 1.01 > 72 processes: 71 sleeping, 1 running, 0 zombie, 0 stopped > CPU states: cpu user nice system irq softirq iowait idle > total 2.7% 0.0% 1.0% 0.1% 0.2% 46.0% 49.5% > cpu00 0.2% 0.0% 0.2% 0.0% 0.2% 2.2% 97.2% > cpu01 5.3% 0.0% 1.9% 0.3% 0.3% 89.8% 1.9% > Mem: 2061696k av, 2043936k used, 17760k free, 0k shrd, 3916k buff > 1566332k actv, 296648k in_d, 30504k in_c > Swap: 16771584k av, 21552k used, 16750032k free 1933772k cached > > PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND > 30960 postgres 15 0 13424 10M 9908 D 2.7 0.5 2:00 1 postmaster > 30538 root 15 0 1080 764 524 S 0.7 0.0 0:43 0 sshd > 1 root 15 0 496 456 436 S 0.0 0.0 0:08 0 init > 2 root RT 0 0 0 0 SW 0.0 0.0 0:00 0 migration/0 > 3 root RT 0 0 0 0 SW 0.0 0.0 0:00 1 migration/1 > 4 root 15 0 0 0 0 SW 0.0 0.0 0:01 0 keventd > 5 root 34 19 0 0 0 SWN 0.0 0.0 0:00 0 ksoftirqd/0 > 6 root 34 19 0 0 0 SWN 0.0 0.0 0:00 1 ksoftirqd/1 > 9 root 15 0 0 0 0 SW 0.0 0.0 0:24 1 bdflush > 7 root 15 0 0 0 0 SW 0.0 0.0 6:53 1 kswapd > 8 root 15 0 0 0 0 SW 0.0 0.0 8:44 1 kscand > 10 root 15 0 0 0 0 SW 0.0 0.0 0:13 0 kupdated > 11 root 25 0 0 0 0 SW 0.0 0.0 0:00 0 mdrecoveryd > 17 root 15 0 0 0 0 SW 0.0 0.0 0:00 0 ahc_dv_0 > > > vmstat output > procs memory swap io system cpu > r b swpd free buff cache si so bi bo in cs us sy id wa > 0 1 21552 17796 4872 1931928 2 3 3 1 27 6 2 1 7 3 > 0 1 21552 18044 4880 1931652 0 0 1652 0 397 512 1 2 50 47 > 0 1 21552 17976 4896 1931664 0 0 2468 0 407 552 2 2 50 47 > 1 0 21552 17984 4896 1931608 0 0 2124 0 418 538 3 3 48 46 > 0 1 21552 18028 4900 1931536 0 0 1592 0 385 509 1 3 50 46 > 0 1 21552 18040 4916 1931488 0 0 1620 820 419 581 2 2 50 46 > 0 1 21552 17968 4916 1931536 0 4 1708 4 402 554 3 1 50 46 > 1 1 21552 18052 4916 1931388 0 0 1772 0 409 531 3 1 49 47 > 0 1 21552 17912 4924 1931492 0 0 1772 0 408 565 3 1 48 48 > 0 1 21552 17932 4932 1931440 0 4 1356 4 391 545 5 0 49 46 > 0 1 21552 18320 4944 1931016 0 4 1500 840 414 571 1 1 48 50 > 0 1 21552 17872 4944 1931440 0 0 2116 0 392 496 1 5 46 48 > 0 1 21552 18060 4944 1931232 0 0 2232 0 423 597 1 2 48 49 > 1 1 21552 17684 4944 1931584 0 0 1752 0 395 537 1 1 50 48 > 0 1 21552 18000 4944 1931240 0 0 1576 0 401 549 0 1 50 49 > > > NetApp stats: > CPU NFS CIFS HTTP Total Net kB/s Disk kB/s Tape kB/s Cache Cache CP=20= > CP Disk DAFS FCP iSCSI FCP kB/s > in out read write read write age hit time ty util in out > 2% 0 0 0 139 0 0 2788 0 0 0 3 96% 0% - 15% 0 139 0 3 2277 > 2% 0 0 0 144 0 0 2504 0 0 0 3 96% 0% - 18% 0 144 0 3 2150 > 2% 0 0 0 130 0 0 2212 0 0 0 3 96% 0% - 13% 0 130 0 3 1879 > 3% 0 0 0 169 0 0 2937 80 0 0 3 96% 0% - 13% 0 169 0 4 2718 > 2% 0 0 0 139 0 0 2448 0 0 0 3 96% 0% - 12% 0 139 0 3 2096 > 2% 0 0 0 137 0 0 2116 0 0 0 3 96% 0% - 10% 0 137 0 3 1892 > 3% 0 0 0 107 0 0 2660 812 0 0 3 96% 24% T 20% 0 107 0 3 1739 > 2% 0 0 0 118 0 0 1788 0 0 0 3 96% 0% - 13% 0 118 0 3 1608 > 2% 0 0 0 136 0 0 2228 0 0 0 3 96% 0% - 11% 0 136 0 3 2018 > 2% 0 0 0 119 0 0 1940 0 0 0 3 96% 0% - 13% 0 119 0 3 1998 > 2% 0 0 0 136 0 0 2175 0 0 0 3 96% 0% - 14% 0 136 0 3 1929 > 2% 0 0 0 133 0 0 1924 0 0 0 3 96% 0% - 19% 0 133 0 3 2292 > 2% 0 0 0 115 0 0 2044 0 0 0 3 96% 0% - 11% 0 115 0 3 1682 > 2% 0 0 0 134 0 0 2256 0 0 0 3 96% 0% - 12% 0 134 0 3 2096 > 2% 0 0 0 112 0 0 2184 0 0 0 3 96% 0% - 12% 0 112 0 3 1633 > 2% 0 0 0 163 0 0 2348 0 0 0 3 96% 0% - 13% 0 163 0 4 2421 > 2% 0 0 0 120 0 0 2056 184 0 0 3 96% 8% T 14% 0 120 0 3 1703 > > strace output: > read(55, "\4\0\0\0\10fm}\1\0\0\0p\0\264\0\0 \2 \230\236\320\0020"...,=20= > 8192) =3D 8192 > _llseek(55, 857997312, [857997312], SEEK_SET) =3D 0 > read(55, "\4\0\0\0\\\315\321|\1\0\0\0p\0\354\0\0 \2 \250\236\260"...,=20= > 8192) =3D 8192 > _llseek(55, 883220480, [883220480], SEEK_SET) =3D 0 > read(55, "\4\0\0\0T\17a~\1\0\0\0p\0\20\1\0 \2=20 > \270\236\220\2D\235"..., 8192) =3D 8192 > _llseek(55, 858005504, [858005504], SEEK_SET) =3D 0 > read(55, "\4\0\0\0\300\356\321|\1\0\0\0p\0\330\0\0 \2=20 > \260\236\240"..., 8192) =3D 8192 > _llseek(55, 857964544, [857964544], SEEK_SET) =3D 0 > read(55, "\4\0\0\0lH\321|\1\0\0\0p\0<\1\0 \2 \300\236\200\2p\235"...,=20= > 8192) =3D 8192 > _llseek(55, 857956352, [857956352], SEEK_SET) =3D 0 > read(55, "\4\0\0\0l\'\321|\1\0\0\0p\0\320\0\0 \2=20 > \260\236\240\2\\"..., 8192) =3D 8192 > _llseek(55, 910802944, [910802944], SEEK_SET) =3D 0 > read(55, "\4\0\0\0\10}\25\200\1\0\0\0l\0\274\1\0 \2 \250\236\260"...,=20= > 8192) =3D 8192 > _llseek(55, 857948160, [857948160], SEEK_SET) =3D 0 > read(55, "\4\0\0\0\370\5\321|\1\0\0\0p\0\350\0\0 \2 \230\236\320"...,=20= > 8192) =3D 8192 > _llseek(56, 80371712, [80371712], SEEK_SET) =3D 0 > read(56, "\4\0\0\0Lf \217\1\0\0\0p\0\f\1\0 \2=20 > \250\236\260\2T\235"..., 8192) =3D 8192 > read(102,=20 > "\2\0\34\0001\236\0\0\1\0\0\0\t\0\0\00020670\0\0\0B\6\0"..., 8192) =3D=20= > 8192 > _llseek(55, 857939968, [857939968], SEEK_SET) =3D 0 > read(55, "\4\0\0\0\244\344\320|\1\0\0\0l\0\230\1\0 \2=20 > \244\236\270"..., 8192) =3D 8192 > _llseek(55, 857923584, [857923584], SEEK_SET) =3D 0 > read(55, "\4\0\0\0\224\242\320|\1\0\0\0p\0|\0\0 \2=20 > \234\236\310\002"..., 8192) =3D 8192 > _llseek(55, 57270272, [57270272], SEEK_SET) =3D 0 > read(55, "\4\0\0\0\3204FK\1\0\0\0t\0\340\0\0 \2=20 > \310\236j\2\214\235"..., 8192) =3D 8192 > _llseek(55, 870727680, [870727680], SEEK_SET) =3D 0 > read(55, "\4\0\0\0x>\233}\1\0\0\0p\0@\1\0 \2 \250\236\260\2X\235"...,=20= > 8192) =3D 8192 > _llseek(55, 1014734848, [1014734848], SEEK_SET) =3D 0 > read(55, "\4\0\0\0\34\354\201\206\1\0\0\0p\0p\0\0 \2=20 > \264\236\230"..., 8192) =3D 8192 > _llseek(55, 857874432, [857874432], SEEK_SET) =3D 0 > read(55, "\4\0\0\0\214\331\317|\1\0\0\0l\0\324\1\0 \2=20 > \224\236\330"..., 8192) =3D 8192 > _llseek(55, 760872960, [760872960], SEEK_SET) =3D 0 > read(55, "\4\0\0\0\30\257\321v\1\0\0\0p\0\230\0\0 \2=20 > \234\236\310"..., 8192) =3D 8192 > _llseek(55, 891715584, [891715584], SEEK_SET) =3D 0 > read(55, "\4\0\0\0\370\220\347~\1\0\0\0p\0P\1\0 \2=20 > \230\236\320\2"..., 8192) =3D 8192 > _llseek(55, 857858048, [857858048], SEEK_SET) =3D 0 > read(55, "\4\0\0\0\250\227\317|\1\0\0\0p\0\264\0\0 \2=20 > \254\236\250"..., 8192) =3D 8192 > _llseek(55, 666910720, [666910720], SEEK_SET) =3D 0 > read(55, "\4\0\0\0x\206\3q\1\0\0\0p\0004\1\0 \2=20 > \254\236\242\2P\235"..., 8192) =3D 8192 > _llseek(55, 857841664, [857841664], SEEK_SET) =3D 0 > read(55, "\4\0\0\0dT\317|\1\0\0\0p\0\224\0\0 \2=20 > \214\236\350\2\30"..., 8192) =3D 8192 > > > From pgsql-performance-owner@postgresql.org Tue Aug 30 16:27:35 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id DD180D7D44 for ; Tue, 30 Aug 2005 16:27:33 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 44752-02 for ; Tue, 30 Aug 2005 19:27:31 +0000 (GMT) Received: from hosting.commandprompt.com (128.commandprompt.com [207.173.200.128]) by svr1.postgresql.org (Postfix) with ESMTP id 888EFD7BC4 for ; Tue, 30 Aug 2005 16:27:28 -0300 (ADT) Received: from [192.168.1.100] (clbb-248.saw.net [64.146.135.248]) (authenticated bits=0) by hosting.commandprompt.com (8.13.4/8.13.4) with ESMTP id j7UJOvjZ024904; Tue, 30 Aug 2005 12:24:57 -0700 Message-ID: <4314B329.20009@commandprompt.com> Date: Tue, 30 Aug 2005 12:27:37 -0700 From: "Joshua D. Drake" User-Agent: Mozilla Thunderbird 1.0.6 (X11/20050727) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Ron Cc: pgsql-performance@postgresql.org Subject: Re: RAID Configuration Sugestion References: <431452FD.3090402@atua.com.br> <6.2.3.4.0.20050830103240.0219b640@pop.earthlink.net> <43148FC4.4000407@commandprompt.com> <6.2.3.4.0.20050830132902.021a5308@pop.earthlink.net> In-Reply-To: <6.2.3.4.0.20050830132902.021a5308@pop.earthlink.net> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Greylist: Sender succeded SMTP AUTH authentication, not delayed by milter-greylist-1.6 (hosting.commandprompt.com [192.168.1.101]); Tue, 30 Aug 2005 12:24:58 -0700 (PDT) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.022 required=5 tests=[AWL=0.022] X-Spam-Level: X-Archive-Number: 200508/539 X-Sequence-Number: 14286 > >> If you still have the budget, I would suggest considering either what >> Ron suggested or possibly using a 4 drive RAID 10 instead. > > > IME, with only 4 HDs, it's usually better to split them them into two > RAID 1's (one for the db, one for everything else including the logs) > than it is to put everything on one RAID 10. YMMV. Really? That's interesting. My experience is different, I assume SCSI? Software/Hardware Raid? Sincerely, Joshua D. Drake > > > Ron Peacetree > From pgsql-performance-owner@postgresql.org Tue Aug 30 16:42:01 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 413F9D8583 for ; Tue, 30 Aug 2005 16:34:43 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 43716-08 for ; Tue, 30 Aug 2005 19:34:40 +0000 (GMT) Received: from mail0.rawbw.com (mail0.rawbw.com [198.144.192.41]) by svr1.postgresql.org (Postfix) with ESMTP id 7C5C5D84F9 for ; Tue, 30 Aug 2005 16:34:38 -0300 (ADT) Received: (from www@localhost) by mail0.rawbw.com (8.11.6p2/8.11.6) id j7UJYe130017 for pgsql-performance@postgresql.org; Tue, 30 Aug 2005 12:34:40 -0700 (PDT) Received: from cybs-gw.ic3.com (cybs-gw.ic3.com [66.185.177.10]) by webmail.rawbw.com (IMP) with HTTP for ; Tue, 30 Aug 2005 12:34:40 -0700 Message-ID: <1125430480.4314b4d00a886@webmail.rawbw.com> Date: Tue, 30 Aug 2005 12:34:40 -0700 From: mudfoot@rawbw.com To: pgsql-performance@postgresql.org Subject: Re: High load and iowait but no disk access References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098C1C@vt-pe2550-001.vantage.vantage.com> In-Reply-To: <4BAFBB6B9CC46F41B2AD7D9F4BBAF785098C1C@vt-pe2550-001.vantage.vantage.com> MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit User-Agent: Internet Messaging Program (IMP) 3.2.1 X-Originating-IP: 66.185.177.10 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.178 required=5 tests=[NO_REAL_NAME=0.178] X-Spam-Level: X-Archive-Number: 200508/540 X-Sequence-Number: 14287 This might be optimal behavior from the hardware. Random reads are hard to optimize for--except if you have enough physical memory to hold the entire dataset. Cached reads (either in array controller or OS buffer cache) should return nearly immediately. But random reads probably aren't cached. And any read-ahead alogorithms or other types of performance enhancements in the hardware or OS go out the window--because the behavior isn't predictable. Each time a drive spindle needs to move to a new track, it requires at least a couple of miliseconds. Sequential reads only require this movement infrequently. But random reads may be forcing this movement for every IO operation. Since the bottleneck in random reads is the physical hard drives themselves, everything else stands around waiting. Fancy hardware can optimize everything else -- writes with write cache, sequential reads with read-ahead and read cache. But there's no real solution to a purely random read workload except perhaps creating different disk groups to help avoid spindle contention. I like this tool: http://www.soliddata.com/products/iotest.html It allows you to select pure workloads (read/write/sequential/random), and it runs against raw devices, so you bypass the OS buffer cache. When I've run it I've always seen sequential activity get much much higher throughput than random. Quoting Anjan Dave : > I have seen references of changing the kernel io scheduler at boot time...not > sure if it applies to RHEL3.0, or will help, but try setting > 'elevator=deadline' during boot time or via grub.conf. Have you tried running > a simple 'dd' on the LUN? The drives are in RAID10 configuration, right? > > > > Thanks, > > Anjan > > _____ > > From: Woody Woodring [mailto:george.woodring@iglass.net] > Sent: Tuesday, August 30, 2005 2:30 PM > To: 'R�my Beaumont'; pgsql-performance@postgresql.org > Subject: Re: [PERFORM] High load and iowait but no disk access > > > > Have you tried a different kernel? We run with a netapp over NFS without any > issues, but we have seen high IO-wait on other Dell boxes (running and not > running postgres) and RHES 3. We have replaced a Dell PowerEdge 350 running > RH 7.3 with a PE750 with more memory running RHES3 and it be bogged down > with IO waits due to syslog messages writing to the disk, the old slower > server could handle it fine. I don't know if it is a Dell thing or a RH > kernel, but we try different kernels on our boxes to try to find one that > works better. We have not found one that stands out over another > consistently but we have been moving away from Update 2 kernel > (2.4.21-15.ELsmp) due to server lockup issues. Unfortunately we get the best > disk throughput on our few remaining 7.3 boxes. > > > > Woody > > > > IGLASS Networks > > www.iglass.net > > > > _____ > > From: pgsql-performance-owner@postgresql.org > [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of R�my Beaumont > Sent: Monday, August 29, 2005 9:43 AM > To: pgsql-performance@postgresql.org > Subject: [PERFORM] High load and iowait but no disk access > > We have been trying to pinpoint what originally seem to be a I/O bottleneck > but which now seems to be an issue with either Postgresql or RHES 3. > > We have the following test environment on which we can reproduce the > problem: > > 1) Test System A > Dell 6650 Quad Xeon Pentium 4 > 8 Gig of RAM > OS: RHES 3 update 2 > Storage: NetApp FAS270 connected using an FC card using 10 disks > > 2) Test System B > Dell Dual Xeon Pentium III > 2 Gig o RAM > OS: RHES 3 update 2 > Storage: NetApp FAS920 connected using an FC card using 28 disks > > Our Database size is around 30G. > > The behavior we see is that when running queries that do random reads on > disk, IOWAIT goes over 80% and actual disk IO falls to a crawl at a > throughput bellow 3000kB/s (We usually average 40000 kB/s to 80000 kB/s on > sequential read operations on the netapps) > > The stats of the NetApp do confirm that it is sitting idle. Doing an strace > on the Postgresql process shows that is it doing seeks and reads. > > So my question is where is this iowait time spent ? > Is there a way to pinpoint the problem in more details ? > We are able to reproduce this behavior with Postgresql 7.4.8 and 8.0.3 > > I have included the output of top,vmstat,strace and systat from the Netapp > from System B while running a single query that generates this behavior. > > R�my > > top output: > 06:27:28 up 5 days, 16:59, 6 users, load average: 1.04, 1.30, 1.01 > 72 processes: 71 sleeping, 1 running, 0 zombie, 0 stopped > CPU states: cpu user nice system irq softirq iowait idle > total 2.7% 0.0% 1.0% 0.1% 0.2% 46.0% 49.5% > cpu00 0.2% 0.0% 0.2% 0.0% 0.2% 2.2% 97.2% > cpu01 5.3% 0.0% 1.9% 0.3% 0.3% 89.8% 1.9% > Mem: 2061696k av, 2043936k used, 17760k free, 0k shrd, 3916k buff > 1566332k actv, 296648k in_d, 30504k in_c > Swap: 16771584k av, 21552k used, 16750032k free 1933772k cached > > PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND > 30960 postgres 15 0 13424 10M 9908 D 2.7 0.5 2:00 1 postmaster > 30538 root 15 0 1080 764 524 S 0.7 0.0 0:43 0 sshd > 1 root 15 0 496 456 436 S 0.0 0.0 0:08 0 init > 2 root RT 0 0 0 0 SW 0.0 0.0 0:00 0 migration/0 > 3 root RT 0 0 0 0 SW 0.0 0.0 0:00 1 migration/1 > 4 root 15 0 0 0 0 SW 0.0 0.0 0:01 0 keventd > 5 root 34 19 0 0 0 SWN 0.0 0.0 0:00 0 ksoftirqd/0 > 6 root 34 19 0 0 0 SWN 0.0 0.0 0:00 1 ksoftirqd/1 > 9 root 15 0 0 0 0 SW 0.0 0.0 0:24 1 bdflush > 7 root 15 0 0 0 0 SW 0.0 0.0 6:53 1 kswapd > 8 root 15 0 0 0 0 SW 0.0 0.0 8:44 1 kscand > 10 root 15 0 0 0 0 SW 0.0 0.0 0:13 0 kupdated > 11 root 25 0 0 0 0 SW 0.0 0.0 0:00 0 mdrecoveryd > 17 root 15 0 0 0 0 SW 0.0 0.0 0:00 0 ahc_dv_0 > > > vmstat output > procs memory swap io system cpu > r b swpd free buff cache si so bi bo in cs us sy id wa > 0 1 21552 17796 4872 1931928 2 3 3 1 27 6 2 1 7 3 > 0 1 21552 18044 4880 1931652 0 0 1652 0 397 512 1 2 50 47 > 0 1 21552 17976 4896 1931664 0 0 2468 0 407 552 2 2 50 47 > 1 0 21552 17984 4896 1931608 0 0 2124 0 418 538 3 3 48 46 > 0 1 21552 18028 4900 1931536 0 0 1592 0 385 509 1 3 50 46 > 0 1 21552 18040 4916 1931488 0 0 1620 820 419 581 2 2 50 46 > 0 1 21552 17968 4916 1931536 0 4 1708 4 402 554 3 1 50 46 > 1 1 21552 18052 4916 1931388 0 0 1772 0 409 531 3 1 49 47 > 0 1 21552 17912 4924 1931492 0 0 1772 0 408 565 3 1 48 48 > 0 1 21552 17932 4932 1931440 0 4 1356 4 391 545 5 0 49 46 > 0 1 21552 18320 4944 1931016 0 4 1500 840 414 571 1 1 48 50 > 0 1 21552 17872 4944 1931440 0 0 2116 0 392 496 1 5 46 48 > 0 1 21552 18060 4944 1931232 0 0 2232 0 423 597 1 2 48 49 > 1 1 21552 17684 4944 1931584 0 0 1752 0 395 537 1 1 50 48 > 0 1 21552 18000 4944 1931240 0 0 1576 0 401 549 0 1 50 49 > > > NetApp stats: > CPU NFS CIFS HTTP Total Net kB/s Disk kB/s Tape kB/s Cache Cache CP CP Disk > DAFS FCP iSCSI FCP kB/s > in out read write read write age hit time ty util in out > 2% 0 0 0 139 0 0 2788 0 0 0 3 96% 0% - 15% 0 139 0 3 2277 > 2% 0 0 0 144 0 0 2504 0 0 0 3 96% 0% - 18% 0 144 0 3 2150 > 2% 0 0 0 130 0 0 2212 0 0 0 3 96% 0% - 13% 0 130 0 3 1879 > 3% 0 0 0 169 0 0 2937 80 0 0 3 96% 0% - 13% 0 169 0 4 2718 > 2% 0 0 0 139 0 0 2448 0 0 0 3 96% 0% - 12% 0 139 0 3 2096 > 2% 0 0 0 137 0 0 2116 0 0 0 3 96% 0% - 10% 0 137 0 3 1892 > 3% 0 0 0 107 0 0 2660 812 0 0 3 96% 24% T 20% 0 107 0 3 1739 > 2% 0 0 0 118 0 0 1788 0 0 0 3 96% 0% - 13% 0 118 0 3 1608 > 2% 0 0 0 136 0 0 2228 0 0 0 3 96% 0% - 11% 0 136 0 3 2018 > 2% 0 0 0 119 0 0 1940 0 0 0 3 96% 0% - 13% 0 119 0 3 1998 > 2% 0 0 0 136 0 0 2175 0 0 0 3 96% 0% - 14% 0 136 0 3 1929 > 2% 0 0 0 133 0 0 1924 0 0 0 3 96% 0% - 19% 0 133 0 3 2292 > 2% 0 0 0 115 0 0 2044 0 0 0 3 96% 0% - 11% 0 115 0 3 1682 > 2% 0 0 0 134 0 0 2256 0 0 0 3 96% 0% - 12% 0 134 0 3 2096 > 2% 0 0 0 112 0 0 2184 0 0 0 3 96% 0% - 12% 0 112 0 3 1633 > 2% 0 0 0 163 0 0 2348 0 0 0 3 96% 0% - 13% 0 163 0 4 2421 > 2% 0 0 0 120 0 0 2056 184 0 0 3 96% 8% T 14% 0 120 0 3 1703 > > strace output: > read(55, "\4\0\0\0\10fm}\1\0\0\0p\0\264\0\0 \2 \230\236\320\0020"..., 8192) = > 8192 > _llseek(55, 857997312, [857997312], SEEK_SET) = 0 > read(55, "\4\0\0\0\\\315\321|\1\0\0\0p\0\354\0\0 \2 \250\236\260"..., 8192) = > 8192 > _llseek(55, 883220480, [883220480], SEEK_SET) = 0 > read(55, "\4\0\0\0T\17a~\1\0\0\0p\0\20\1\0 \2 \270\236\220\2D\235"..., 8192) > = 8192 > _llseek(55, 858005504, [858005504], SEEK_SET) = 0 > read(55, "\4\0\0\0\300\356\321|\1\0\0\0p\0\330\0\0 \2 \260\236\240"..., 8192) > = 8192 > _llseek(55, 857964544, [857964544], SEEK_SET) = 0 > read(55, "\4\0\0\0lH\321|\1\0\0\0p\0<\1\0 \2 \300\236\200\2p\235"..., 8192) = > 8192 > _llseek(55, 857956352, [857956352], SEEK_SET) = 0 > read(55, "\4\0\0\0l\'\321|\1\0\0\0p\0\320\0\0 \2 \260\236\240\2\\"..., 8192) > = 8192 > _llseek(55, 910802944, [910802944], SEEK_SET) = 0 > read(55, "\4\0\0\0\10}\25\200\1\0\0\0l\0\274\1\0 \2 \250\236\260"..., 8192) = > 8192 > _llseek(55, 857948160, [857948160], SEEK_SET) = 0 > read(55, "\4\0\0\0\370\5\321|\1\0\0\0p\0\350\0\0 \2 \230\236\320"..., 8192) = > 8192 > _llseek(56, 80371712, [80371712], SEEK_SET) = 0 > read(56, "\4\0\0\0Lf \217\1\0\0\0p\0\f\1\0 \2 \250\236\260\2T\235"..., 8192) > = 8192 > read(102, "\2\0\34\0001\236\0\0\1\0\0\0\t\0\0\00020670\0\0\0B\6\0"..., 8192) > = 8192 > _llseek(55, 857939968, [857939968], SEEK_SET) = 0 > read(55, "\4\0\0\0\244\344\320|\1\0\0\0l\0\230\1\0 \2 \244\236\270"..., 8192) > = 8192 > _llseek(55, 857923584, [857923584], SEEK_SET) = 0 > read(55, "\4\0\0\0\224\242\320|\1\0\0\0p\0|\0\0 \2 \234\236\310\002"..., > 8192) = 8192 > _llseek(55, 57270272, [57270272], SEEK_SET) = 0 > read(55, "\4\0\0\0\3204FK\1\0\0\0t\0\340\0\0 \2 \310\236j\2\214\235"..., > 8192) = 8192 > _llseek(55, 870727680, [870727680], SEEK_SET) = 0 > read(55, "\4\0\0\0x>\233}\1\0\0\0p\0@\1\0 \2 \250\236\260\2X\235"..., 8192) = > 8192 > _llseek(55, 1014734848, [1014734848], SEEK_SET) = 0 > read(55, "\4\0\0\0\34\354\201\206\1\0\0\0p\0p\0\0 \2 \264\236\230"..., 8192) > = 8192 > _llseek(55, 857874432, [857874432], SEEK_SET) = 0 > read(55, "\4\0\0\0\214\331\317|\1\0\0\0l\0\324\1\0 \2 \224\236\330"..., 8192) > = 8192 > _llseek(55, 760872960, [760872960], SEEK_SET) = 0 > read(55, "\4\0\0\0\30\257\321v\1\0\0\0p\0\230\0\0 \2 \234\236\310"..., 8192) > = 8192 > _llseek(55, 891715584, [891715584], SEEK_SET) = 0 > read(55, "\4\0\0\0\370\220\347~\1\0\0\0p\0P\1\0 \2 \230\236\320\2"..., 8192) > = 8192 > _llseek(55, 857858048, [857858048], SEEK_SET) = 0 > read(55, "\4\0\0\0\250\227\317|\1\0\0\0p\0\264\0\0 \2 \254\236\250"..., 8192) > = 8192 > _llseek(55, 666910720, [666910720], SEEK_SET) = 0 > read(55, "\4\0\0\0x\206\3q\1\0\0\0p\0004\1\0 \2 \254\236\242\2P\235"..., > 8192) = 8192 > _llseek(55, 857841664, [857841664], SEEK_SET) = 0 > read(55, "\4\0\0\0dT\317|\1\0\0\0p\0\224\0\0 \2 \214\236\350\2\30"..., 8192) > = 8192 > > > > > > From pgsql-performance-owner@postgresql.org Tue Aug 30 18:14:00 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AF5C0D7EE6 for ; Tue, 30 Aug 2005 18:13:59 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 66555-09 for ; Tue, 30 Aug 2005 21:13:56 +0000 (GMT) Received: from web31606.mail.mud.yahoo.com (web31606.mail.mud.yahoo.com [68.142.198.152]) by svr1.postgresql.org (Postfix) with SMTP id 4E111D7DD2 for ; Tue, 30 Aug 2005 18:13:54 -0300 (ADT) Received: (qmail 48140 invoked by uid 60001); 30 Aug 2005 21:13:56 -0000 Message-ID: <20050830211356.48138.qmail@web31606.mail.mud.yahoo.com> Received: from [209.217.70.195] by web31606.mail.mud.yahoo.com via HTTP; Tue, 30 Aug 2005 14:13:55 PDT Date: Tue, 30 Aug 2005 14:13:55 -0700 (PDT) From: Markus Benne Reply-To: markus@m-bass.com Subject: When to do a vacuum for highly active table To: pgsql-performance@postgresql.org MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/541 X-Sequence-Number: 14288 We have a highly active table that has virtually all entries updated every 5 minutes. Typical size of the table is 50,000 entries, and entries have grown fat. We are currently vaccuming hourly, and towards the end of the hour we are seeing degradation, when compared to the top of the hour. Vaccum is slowly killing our system, as it is starting to take up to 10 minutes, and load at the time of vacuum is 6+ on a Linux box. During the vacuum, overall system is goin unresponsive, then comes back once vacuum completes. If we run vacuum less frequently, degradation continues to the point that we can't keep up with the throughput, plus vacuum takes longer anyway. Becoming quite a pickle:-) We are thinking of splitting the table in two: the part the updates often, and the part the updates infrequently as we suspect that record size impacts vacuum. Any ideas? Thanks, Mark ----------------- From pgsql-performance-owner@postgresql.org Tue Sep 6 01:12:44 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 077F9D707D for ; Tue, 30 Aug 2005 18:25:54 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 72055-03 for ; Tue, 30 Aug 2005 21:25:51 +0000 (GMT) Received: from linux.finestmedia.tv (linux.finestmedia.tv [213.180.31.12]) by svr1.postgresql.org (Postfix) with ESMTP id AFAF0D706F for ; Tue, 30 Aug 2005 18:25:49 -0300 (ADT) Received: from localhost (localhost [127.0.0.1]) by linux.finestmedia.tv (Postfix) with ESMTP id 4FC6718E71B for ; Wed, 31 Aug 2005 00:25:49 +0300 (EEST) Received: from linux.finestmedia.tv ([127.0.0.1]) by localhost (linux.finestmedia.tv [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 29451-04 for ; Wed, 31 Aug 2005 00:25:46 +0300 (EEST) Received: from err (err.fm.sise [192.168.2.19]) by linux.finestmedia.tv (Postfix) with ESMTP id 5472718E718 for ; Wed, 31 Aug 2005 00:25:45 +0300 (EEST) From: "Rigmor Ukuhe" To: Subject: Re: When to do a vacuum for highly active table Date: Wed, 31 Aug 2005 00:25:44 +0300 Organization: Finestmedia MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit X-Mailer: Microsoft Office Outlook, Build 11.0.6353 Thread-Index: AcWtqCLO1Qji3rAwRQSxeM5zsbuD4QAAJ1CA X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2527 In-Reply-To: <20050830211356.48138.qmail@web31606.mail.mud.yahoo.com> Message-Id: <20050830212545.5472718E718@linux.finestmedia.tv> X-Virus-Scanned: by amavisd-new at finestmedia.ee X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.026 required=5 tests=[AWL=0.026] X-Spam-Level: X-Archive-Number: 200509/80 X-Sequence-Number: 14388 > -----Original Message----- > From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance- > owner@postgresql.org] On Behalf Of Markus Benne > Sent: Wednesday, August 31, 2005 12:14 AM > To: pgsql-performance@postgresql.org > Subject: [PERFORM] When to do a vacuum for highly active table > > We have a highly active table that has virtually all > entries updated every 5 minutes. Typical size of the > table is 50,000 entries, and entries have grown fat. > > We are currently vaccuming hourly, and towards the end > of the hour we are seeing degradation, when compared > to the top of the hour. > > Vaccum is slowly killing our system, as it is starting > to take up to 10 minutes, and load at the time of > vacuum is 6+ on a Linux box. During the vacuum, > overall system is goin unresponsive, then comes back > once vacuum completes. Play with vacuum_cost_delay option. In our case it made BIG difference (going from very heavy hitting to almost unnoticed vacuuming.) Hope it helps. Rigmor Ukuhe > > If we run vacuum less frequently, degradation > continues to the point that we can't keep up with the > throughput, plus vacuum takes longer anyway. > > Becoming quite a pickle:-) > > We are thinking of splitting the table in two: the > part the updates often, and the part the updates > infrequently as we suspect that record size impacts > vacuum. > > Any ideas? > > > Thanks, > Mark > > ----------------- > > ---------------------------(end of broadcast)--------------------------- > TIP 2: Don't 'kill -9' the postmaster From pgsql-performance-owner@postgresql.org Tue Aug 30 18:30:49 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3817ED86EE for ; Tue, 30 Aug 2005 18:30:43 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 70857-10 for ; Tue, 30 Aug 2005 21:30:24 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id BEC6CD86AF for ; Tue, 30 Aug 2005 18:29:15 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7ULTH1Z024876; Tue, 30 Aug 2005 17:29:17 -0400 (EDT) To: markus@m-bass.com Cc: pgsql-performance@postgresql.org Subject: Re: When to do a vacuum for highly active table In-reply-to: <20050830211356.48138.qmail@web31606.mail.mud.yahoo.com> References: <20050830211356.48138.qmail@web31606.mail.mud.yahoo.com> Comments: In-reply-to Markus Benne message dated "Tue, 30 Aug 2005 14:13:55 -0700" Date: Tue, 30 Aug 2005 17:29:17 -0400 Message-ID: <24875.1125437357@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/542 X-Sequence-Number: 14289 Markus Benne writes: > We have a highly active table that has virtually all > entries updated every 5 minutes. Typical size of the > table is 50,000 entries, and entries have grown fat. > We are currently vaccuming hourly, and towards the end > of the hour we are seeing degradation, when compared > to the top of the hour. On something like this, you really need to be vacuuming more often not less so; I'd think about how to do it every five or ten minutes rather than backing off. With only 50K rows it should really not take more than a couple of seconds to do the vacuum. When you wait till there are 600K dead rows, it's going to take awhile, plus you are suffering across-the-board performance degradation from all the dead rows. If you are using PG 8.0, there are some "vacuum cost" knobs you can fiddle with to slow down vacuum so it doesn't impose as much I/O load. Ideally you could get it to where you could run vacuum as often as you need to without noticing much impact on foreground processing. If you're not using 8.0 ... maybe it's time to update. Another thing you might want to do is look at "vacuum verbose" output, which will give you some idea of the time spent in each step. It might be there are specific aspects that could be improved. > We are thinking of splitting the table in two: the > part the updates often, and the part the updates > infrequently as we suspect that record size impacts > vacuum. You just said that virtually all rows update constantly --- where's the "infrequent" part? regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 30 19:15:40 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BAD23D86C6 for ; Tue, 30 Aug 2005 19:15:37 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 84803-01 for ; Tue, 30 Aug 2005 22:15:25 +0000 (GMT) Received: from mail.envyfinancial.com (unknown [206.248.142.186]) by svr1.postgresql.org (Postfix) with ESMTP id D5082D86C4 for ; Tue, 30 Aug 2005 19:15:20 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by mail.envyfinancial.com (Postfix) with ESMTP id 94AE81C73D; Tue, 30 Aug 2005 18:05:03 -0400 (EDT) Received: from mail.envyfinancial.com ([127.0.0.1]) by localhost (mark.mielke.cc [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 27349-07; Tue, 30 Aug 2005 18:05:03 -0400 (EDT) Received: by mail.envyfinancial.com (Postfix, from userid 500) id 6A5931C742; Tue, 30 Aug 2005 18:05:03 -0400 (EDT) Date: Tue, 30 Aug 2005 18:05:03 -0400 From: mark@mark.mielke.cc To: Tom Lane Cc: markus@m-bass.com, pgsql-performance@postgresql.org Subject: Re: When to do a vacuum for highly active table Message-ID: <20050830220503.GA27740@mark.mielke.cc> References: <20050830211356.48138.qmail@web31606.mail.mud.yahoo.com> <24875.1125437357@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <24875.1125437357@sss.pgh.pa.us> User-Agent: Mutt/1.4.2.1i X-Virus-Scanned: amavisd-new at mail.envyfinancial.com X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.208 required=5 tests=[AWL=0.030, NO_REAL_NAME=0.178] X-Spam-Level: X-Archive-Number: 200508/543 X-Sequence-Number: 14290 On Tue, Aug 30, 2005 at 05:29:17PM -0400, Tom Lane wrote: > Markus Benne writes: > > We have a highly active table that has virtually all > > entries updated every 5 minutes. Typical size of the > > table is 50,000 entries, and entries have grown fat. > ... > > We are thinking of splitting the table in two: the > > part the updates often, and the part the updates > > infrequently as we suspect that record size impacts > > vacuum. > You just said that virtually all rows update constantly --- where's > the "infrequent" part? I think he means splitting it vertically, instead of horizontally, and it sounds like an excellent idea, if a large enough portion of each record is in fact mostly fixed. Otherwise, PostgreSQL is copying data multiple times, only to have the data expire as part of a dead row. I've already started to notice such issues with postgresql - but more because I'm using low-end hardware, and I'm projecting the effect for when our database becomes much larger with much higher demand on the database. This is the sort of scenario where a database without transactional integrity would significantly out-perform one designed around it. If records are fixed sized, and updated in place, these problems would occur far less often. Is it heresy to suggest MySQL in here? :-) I switched from MySQL to PostgreSQL several months ago, and haven't looked back - but they do work differently, and for certain uses, one can destroy the other. Using a MyISAM table would be the way I would go with this sort of problem. Cheers, mark -- mark@mielke.cc / markm@ncf.ca / markm@nortel.com __________________________ . . _ ._ . . .__ . . ._. .__ . . . .__ | Neighbourhood Coder |\/| |_| |_| |/ |_ |\/| | |_ | |/ |_ | | | | | | \ | \ |__ . | | .|. |__ |__ | \ |__ | Ottawa, Ontario, Canada One ring to rule them all, one ring to find them, one ring to bring them all and in the darkness bind them... http://mark.mielke.cc/ From pgsql-performance-owner@postgresql.org Tue Aug 30 20:04:09 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 92617D86A7 for ; Tue, 30 Aug 2005 20:04:08 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 94807-07 for ; Tue, 30 Aug 2005 23:04:05 +0000 (GMT) Received: from floppy.pyrenet.fr (news.pyrenet.fr [194.116.145.2]) by svr1.postgresql.org (Postfix) with ESMTP id 07736D8691 for ; Tue, 30 Aug 2005 20:04:04 -0300 (ADT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id CC26430956; Wed, 31 Aug 2005 01:04:03 +0200 (MET DST) From: Chris Browne X-Newsgroups: pgsql.performance Subject: Re: When to do a vacuum for highly active table Date: Tue, 30 Aug 2005 18:05:38 -0400 Organization: cbbrowne Computing Inc Lines: 33 Message-ID: <6064tn6pxp.fsf@dba2.int.libertyrms.com> References: <20050830211356.48138.qmail@web31606.mail.mud.yahoo.com> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 X-Complaints-To: usenet@news.hub.org User-Agent: Gnus/5.1007 (Gnus v5.10.7) XEmacs/21.4.17 (Jumbo Shrimp, linux) Cancel-Lock: sha1:Mb6tP0jYqjMBqYOk/ujn4/x8Npg= To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.101 required=5 tests=[AWL=0.101] X-Spam-Level: X-Archive-Number: 200508/548 X-Sequence-Number: 14295 thing@m-bass.com (Markus Benne) writes: > We have a highly active table that has virtually all > entries updated every 5 minutes. Typical size of the > table is 50,000 entries, and entries have grown fat. > > We are currently vaccuming hourly, and towards the end > of the hour we are seeing degradation, when compared > to the top of the hour. You're not vacuuming the table nearly often enough. You should vacuum this table every five minutes, and possibly more often than that. [We have some tables like that, albeit smaller than 50K entries, which we vacuum once per minute in production...] > We are thinking of splitting the table in two: the part the updates > often, and the part the updates infrequently as we suspect that > record size impacts vacuum. There's *some* merit to that. You might discover that there's a "hot spot" that needs to be vacuumed once per minute. But it may be simpler to just hit the table with a vacuum once every few minutes even though some tuples are seldom updated. -- output = reverse("gro.gultn" "@" "enworbbc") http://cbbrowne.com/info/spreadsheets.html Signs of a Klingon Programmer #3: "By filing this TPR you have challenged the honor of my family. Prepare to die!" From pgsql-performance-owner@postgresql.org Tue Aug 30 19:19:21 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6A3ADD85E9 for ; Tue, 30 Aug 2005 19:19:20 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 82130-10 for ; Tue, 30 Aug 2005 22:19:18 +0000 (GMT) Received: from smtp102.biz.mail.mud.yahoo.com (smtp102.biz.mail.mud.yahoo.com [68.142.200.237]) by svr1.postgresql.org (Postfix) with SMTP id 937EFD84FF for ; Tue, 30 Aug 2005 19:19:16 -0300 (ADT) Received: (qmail 68845 invoked from network); 30 Aug 2005 22:19:19 -0000 Received: from unknown (HELO ?192.168.3.1?) (ralph.mason@telogis.com@203.98.10.169 with plain) by smtp102.biz.mail.mud.yahoo.com with SMTP; 30 Aug 2005 22:19:18 -0000 Message-ID: <4314DBE0.9070708@telogis.com> Date: Wed, 31 Aug 2005 10:21:20 +1200 From: Ralph Mason User-Agent: Mozilla Thunderbird 1.0.6 (Windows/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: 'Real' auto vacuum? Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/544 X-Sequence-Number: 14291 This is a wild and crazy thought which I am sure is invalid for some good reason. But why can't postgres just vacuum itself as it goes along? When a row is orphaned it's added to a list of possibly available rows. When a new row is needed the list of possible rows is examined and the first one with a transaction id less then the lowest running transaction id is chosen to be the new row? These rows can be in a heap so it's really fast to find one. Like magic - no more vacuuming. No more holes for people to fall into. Is this an oversimplification of the problem? Ralph From pgsql-performance-owner@postgresql.org Tue Aug 30 19:45:46 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6A80AD8686 for ; Tue, 30 Aug 2005 19:45:45 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91135-02 for ; Tue, 30 Aug 2005 22:45:38 +0000 (GMT) Received: from mail.envyfinancial.com (unknown [206.248.142.186]) by svr1.postgresql.org (Postfix) with ESMTP id 4AEE6D870F for ; Tue, 30 Aug 2005 19:45:36 -0300 (ADT) Received: from localhost (localhost.localdomain [127.0.0.1]) by mail.envyfinancial.com (Postfix) with ESMTP id 5895F1C7A7; Tue, 30 Aug 2005 18:35:19 -0400 (EDT) Received: from mail.envyfinancial.com ([127.0.0.1]) by localhost (mark.mielke.cc [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 28101-03; Tue, 30 Aug 2005 18:35:19 -0400 (EDT) Received: by mail.envyfinancial.com (Postfix, from userid 500) id 15D581C7CA; Tue, 30 Aug 2005 18:35:19 -0400 (EDT) Date: Tue, 30 Aug 2005 18:35:19 -0400 From: mark@mark.mielke.cc To: Ralph Mason Cc: pgsql-performance@postgresql.org Subject: Re: 'Real' auto vacuum? Message-ID: <20050830223518.GA28168@mark.mielke.cc> References: <4314DBE0.9070708@telogis.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <4314DBE0.9070708@telogis.com> User-Agent: Mutt/1.4.2.1i X-Virus-Scanned: amavisd-new at mail.envyfinancial.com X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.205 required=5 tests=[AWL=0.027, NO_REAL_NAME=0.178] X-Spam-Level: X-Archive-Number: 200508/545 X-Sequence-Number: 14292 On Wed, Aug 31, 2005 at 10:21:20AM +1200, Ralph Mason wrote: > This is a wild and crazy thought which I am sure is invalid for some > good reason. > > But why can't postgres just vacuum itself as it goes along? > > When a row is orphaned it's added to a list of possibly available rows. > When a new row is needed the list of possible rows is examined and the > first one with a transaction id less then the lowest running transaction > id is chosen to be the new row? These rows can be in a heap so it's > really fast to find one. > > Like magic - no more vacuuming. No more holes for people to fall into. Yes please. :-) > Is this an oversimplification of the problem? But, yeah. It's probably not that easy, especially with really big databases. Where is this free list stored? How efficient is it to keep track of the lowest running transaction at all times? How does one synchronize access to this free list, to ensure that processes don't block up waiting for access to the free list? Is the fre list journalled to prevent corruption, and the accidental re-use of a still in use row? And, there would be a cost to scanning this list on every insert or update. As an outsider (like you?) I see the current model as a design flaw as well. A neat and tidy model on paper. Not so nice in real life. The need to vacuum in batch mode, to keep the database from dying, seems intuitively bad. I think there must be answers to this problem. Even simple optimizations, such as defining a table such that any delete or update within a table, upon commit, will attempt to vacuum just the rows that should not be considered free for any new transactions. If it's in use by an active transaction, oh well. It can be picked up by a batch run of vacuum. If it's free though - let's do it now. I think any optimizations we come up with, will be more happily accepted with a working patch that causes no breakage... :-) Cheers, mark -- mark@mielke.cc / markm@ncf.ca / markm@nortel.com __________________________ . . _ ._ . . .__ . . ._. .__ . . . .__ | Neighbourhood Coder |\/| |_| |_| |/ |_ |\/| | |_ | |/ |_ | | | | | | \ | \ |__ . | | .|. |__ |__ | \ |__ | Ottawa, Ontario, Canada One ring to rule them all, one ring to find them, one ring to bring them all and in the darkness bind them... http://mark.mielke.cc/ From pgsql-performance-owner@postgresql.org Tue Aug 30 19:55:14 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7B422D86EA for ; Tue, 30 Aug 2005 19:55:12 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 90789-06 for ; Tue, 30 Aug 2005 22:55:10 +0000 (GMT) Received: from davinci.ethosmedia.com (server227.ethosmedia.com [209.128.84.227]) by svr1.postgresql.org (Postfix) with ESMTP id CF932D86D4 for ; Tue, 30 Aug 2005 19:55:09 -0300 (ADT) X-EthosMedia-Virus-Scanned: no infections found Received: from [64.81.245.111] (account josh@agliodbs.com HELO temoku.sf.agliodbs.com) by davinci.ethosmedia.com (CommuniGate Pro SMTP 4.1.8) with ESMTP id 7998323; Tue, 30 Aug 2005 15:57:29 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: 'Real' auto vacuum? Date: Tue, 30 Aug 2005 15:58:34 -0700 User-Agent: KMail/1.8 Cc: Ralph Mason References: <4314DBE0.9070708@telogis.com> In-Reply-To: <4314DBE0.9070708@telogis.com> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200508301558.34697.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.048 required=5 tests=[AWL=-0.002, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/546 X-Sequence-Number: 14293 Ralph, > When a row is orphaned it's added to a list of possibly available rows. > When a new row is needed the list of possible rows is examined and the > first one with a transaction id less then the lowest running transaction > id is chosen to be the new row? These rows can be in a heap so it's > really fast to find one. This is the long-term plan. However, it's actually a lot harder than it sounds. Patches welcome. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Tue Aug 30 20:02:46 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id BD931D6E40 for ; Tue, 30 Aug 2005 20:02:45 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 94332-03 for ; Tue, 30 Aug 2005 23:02:30 +0000 (GMT) Received: from smtpauth09.mail.atl.earthlink.net (smtpauth09.mail.atl.earthlink.net [209.86.89.69]) by svr1.postgresql.org (Postfix) with ESMTP id 56989D8668 for ; Tue, 30 Aug 2005 20:02:29 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=FuHj2/mPj6faXiIrgDUtUtwA/RQ4TAL4KmNXJJRYugP0fkNNRFdk+W9JR0pdkcAI; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth09.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1EAF7f-0004sd-Kq; Tue, 30 Aug 2005 19:02:31 -0400 Message-Id: <6.2.3.4.0.20050830154823.05ac9308@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Tue, 30 Aug 2005 19:02:28 -0400 To: "Joshua D. Drake" , pgsql-performance@postgresql.org From: Ron Subject: Re: RAID Configuration Sugestion In-Reply-To: <4314B329.20009@commandprompt.com> References: <431452FD.3090402@atua.com.br> <6.2.3.4.0.20050830103240.0219b640@pop.earthlink.net> <43148FC4.4000407@commandprompt.com> <6.2.3.4.0.20050830132902.021a5308@pop.earthlink.net> <4314B329.20009@commandprompt.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bc5c1988b3457bd03d90ce542a24f176e2350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.406 required=5 tests=[AWL=0.032, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/547 X-Sequence-Number: 14294 At 03:27 PM 8/30/2005, Joshua D. Drake wrote: >>>If you still have the budget, I would suggest considering either >>>what Ron suggested or possibly using a 4 drive RAID 10 instead. >> >> >>IME, with only 4 HDs, it's usually better to split them them into >>two RAID 1's (one for the db, one for everything else including the >>logs) than it is to put everything on one RAID 10. YMMV. > >Really? That's interesting. My experience is different, I assume >SCSI? Software/Hardware Raid? The issue exists regardless of technologies used, although the technology used does affect when things become an irritation or serious problem. The issue with "everything on the same HD set" seems to be that under light loads anything works reasonably well, but as load increases contention between DB table access, OS access, and xlog writes can cause performance problems. In particular, _everything_ else hangs while logs are being written with "everything on the same HD set". Thus leaving you with the nasty choices of small log writes that cause more seeking behavior, and the resultant poor overall HD IO performance, or large log writes that basically freeze the server until they are done. Having the logs on a different HD, and if possible different IO bus, reduces this effect to a minimum and seems to be a better choice than the "shared everything" approach. Although this effect seems largest when there are fewest HDs, the general pattern is that one should use as many spindles as one can make use of and that they should be as dedicated as possible in their purpose(s). That's why the TPC bench marked systems tend to have literally 100's of HD's and they tend to be split into very focused purposes. Ron Peacetree From pgsql-performance-owner@postgresql.org Tue Aug 30 20:05:12 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 295A6D86CE for ; Tue, 30 Aug 2005 20:05:11 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 95447-03 for ; Tue, 30 Aug 2005 23:05:04 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 826A4D86A6 for ; Tue, 30 Aug 2005 20:05:03 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7UN54nF025547; Tue, 30 Aug 2005 19:05:04 -0400 (EDT) To: mark@mark.mielke.cc Cc: markus@m-bass.com, pgsql-performance@postgresql.org Subject: Re: When to do a vacuum for highly active table In-reply-to: <20050830220503.GA27740@mark.mielke.cc> References: <20050830211356.48138.qmail@web31606.mail.mud.yahoo.com> <24875.1125437357@sss.pgh.pa.us> <20050830220503.GA27740@mark.mielke.cc> Comments: In-reply-to mark@mark.mielke.cc message dated "Tue, 30 Aug 2005 18:05:03 -0400" Date: Tue, 30 Aug 2005 19:05:04 -0400 Message-ID: <25546.1125443104@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/549 X-Sequence-Number: 14296 mark@mark.mielke.cc writes: > I think he means splitting it vertically, instead of horizontally, and > it sounds like an excellent idea, if a large enough portion of each > record is in fact mostly fixed. Otherwise, PostgreSQL is copying data > multiple times, only to have the data expire as part of a dead row. Only up to a point. Fields that are wide enough to get toasted out-of-line (multiple Kb) do not get physically copied if there's a row update that doesn't affect them. We don't really have enough information about his table to guess whether there's any point in manually partitioning the columns, but my leaning would be "probably not" --- the overhead in joining the resulting two tables would be high enough that you'd need a heck of a big improvement to justify it. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Aug 30 20:37:20 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AEECBD867A for ; Tue, 30 Aug 2005 20:37:19 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 01375-09 for ; Tue, 30 Aug 2005 23:37:14 +0000 (GMT) Received: from smtp102.biz.mail.mud.yahoo.com (smtp102.biz.mail.mud.yahoo.com [68.142.200.237]) by svr1.postgresql.org (Postfix) with SMTP id 4CF14D8653 for ; Tue, 30 Aug 2005 20:37:13 -0300 (ADT) Received: (qmail 38594 invoked from network); 30 Aug 2005 23:37:17 -0000 Received: from unknown (HELO ?192.168.3.1?) (ralph.mason@telogis.com@203.98.10.169 with plain) by smtp102.biz.mail.mud.yahoo.com with SMTP; 30 Aug 2005 23:37:16 -0000 Message-ID: <4314EE25.30605@telogis.com> Date: Wed, 31 Aug 2005 11:39:17 +1200 From: Ralph Mason User-Agent: Mozilla Thunderbird 1.0.6 (Windows/20050716) X-Accept-Language: en-us, en MIME-Version: 1.0 To: mark@mark.mielke.cc Cc: pgsql-performance@postgresql.org Subject: Re: 'Real' auto vacuum? References: <4314DBE0.9070708@telogis.com> <20050830223518.GA28168@mark.mielke.cc> In-Reply-To: <20050830223518.GA28168@mark.mielke.cc> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/550 X-Sequence-Number: 14297 mark@mark.mielke.cc wrote: >But, yeah. It's probably not that easy, especially with really big >databases. Where is this free list stored? How efficient is it to keep >track of the lowest running transaction at all times? How does one >synchronize access to this free list, to ensure that processes don't >block up waiting for access to the free list? Is the fre list >journalled to prevent corruption, and the accidental re-use of a still >in use row? And, there would be a cost to scanning this list on every >insert or update. > > I suspect the freelist could be stored as an index, and just handily postgres supports those out of the box. There would be a cost yes, but then what is the cost of adding pages to the file all the time? I guess as with all things there is no one size fits all, so perhaps you could turn it off - although I expect for 99.9% of the cases 'on' would be the better choice. If it gets broken there is already the reindex code that can fix it. A coherency / fixing / recover of a table command would probably be a useful tool anyway. >As an outsider (like you?) I see the current model as a design flaw as >well. A neat and tidy model on paper. Not so nice in real life. The >need to vacuum in batch mode, to keep the database from dying, seems >intuitively bad. > > We have a script that vacuums the database every 5 minutes, excessive - yes, but turns out that any less is no good really. I think that this is sub optimal, the DB work keeps running, but the vacuum can slow down other tasks. It also probably flushes data that we would need out of the page cache so it can look at data that isn't used often as the vacuum runs. Not the most optimal data access pattern I could imagine. >I think there must be answers to this problem. Even simple >optimizations, such as defining a table such that any delete or update >within a table, upon commit, will attempt to vacuum just the rows that >should not be considered free for any new transactions. If it's in >use by an active transaction, oh well. It can be picked up by a batch >run of vacuum. If it's free though - let's do it now. > > Anything would be good - I think it's the achilles heel of postgres. Perhaps there is something simple like that could fix 95% of the problem. >I think any optimizations we come up with, will be more happily accepted >with a working patch that causes no breakage... :-) > > > I am sure they would. Cheers Ralph From pgsql-performance-owner@postgresql.org Tue Aug 30 21:04:45 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9188CD86AE for ; Tue, 30 Aug 2005 21:04:44 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 08824-05 for ; Wed, 31 Aug 2005 00:04:42 +0000 (GMT) Received: from vms040pub.verizon.net (vms040pub.verizon.net [206.46.252.40]) by svr1.postgresql.org (Postfix) with ESMTP id 4CA3FD86A9 for ; Tue, 30 Aug 2005 21:04:41 -0300 (ADT) Received: from osgiliath.mathom.us ([151.200.14.55]) by vms040.mailsrvcs.net (Sun Java System Messaging Server 6.2-2.05 (built Apr 28 2005)) with ESMTPA id <0IM200H6487V0US1@vms040.mailsrvcs.net> for pgsql-performance@postgresql.org; Tue, 30 Aug 2005 19:04:44 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by osgiliath.mathom.us (Postfix) with ESMTP id 7601260043E for ; Tue, 30 Aug 2005 20:04:43 -0400 (EDT) Received: from osgiliath.mathom.us ([127.0.0.1]) by localhost (osgiliath [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 11456-04-2 for ; Tue, 30 Aug 2005 20:04:43 -0400 (EDT) Received: by osgiliath.mathom.us (Postfix, from userid 1000) id 3B11D60041D; Tue, 30 Aug 2005 20:04:43 -0400 (EDT) Date: Tue, 30 Aug 2005 20:04:43 -0400 From: Michael Stone Subject: Re: RAID Configuration Sugestion In-reply-to: <6.2.3.4.0.20050830154823.05ac9308@pop.earthlink.net> To: pgsql-performance@postgresql.org Mail-Followup-To: pgsql-performance@postgresql.org Message-id: <20050831000443.GZ14921@mathom.us> MIME-version: 1.0 Content-type: text/plain; charset=us-ascii; format=flowed Content-disposition: inline X-Pgp-Fingerprint: 53 FF 38 00 E7 DD 0A 9C 84 52 84 C5 EE DF 7C 88 X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mathom.us References: <431452FD.3090402@atua.com.br> <6.2.3.4.0.20050830103240.0219b640@pop.earthlink.net> <43148FC4.4000407@commandprompt.com> <6.2.3.4.0.20050830132902.021a5308@pop.earthlink.net> <4314B329.20009@commandprompt.com> <6.2.3.4.0.20050830154823.05ac9308@pop.earthlink.net> User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.04 required=5 tests=[AWL=-0.010, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/551 X-Sequence-Number: 14298 On Tue, Aug 30, 2005 at 07:02:28PM -0400, Ron wrote: >purpose(s). That's why the TPC bench marked systems tend to have >literally 100's of HD's and they tend to be split into very focused >purposes. Of course, TPC benchmark systems are constructed such that cost and storage capacity are irrelevant--in the real world things tend to be more complicated. Mike Stone From pgsql-performance-owner@postgresql.org Tue Aug 30 21:41:45 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 19519D845F for ; Tue, 30 Aug 2005 21:41:44 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 20568-02 for ; Wed, 31 Aug 2005 00:41:42 +0000 (GMT) Received: from smtpauth03.mail.atl.earthlink.net (smtpauth03.mail.atl.earthlink.net [209.86.89.63]) by svr1.postgresql.org (Postfix) with ESMTP id AC623D7BC4 for ; Tue, 30 Aug 2005 21:41:40 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=X8m3ogjwA7itRKelI5VyL92f5h5qmM8mIOlCL4KglD13uFb/EVDxP52TxTxpyYRQ; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth03.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1EAGfg-0005Kj-1H; Tue, 30 Aug 2005 20:41:44 -0400 Message-Id: <6.2.3.4.0.20050830203126.0221feb8@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Tue, 30 Aug 2005 20:41:40 -0400 To: Michael Stone , pgsql-performance@postgresql.org From: Ron Subject: Re: RAID Configuration Sugestion In-Reply-To: <20050831000443.GZ14921@mathom.us> References: <431452FD.3090402@atua.com.br> <6.2.3.4.0.20050830103240.0219b640@pop.earthlink.net> <43148FC4.4000407@commandprompt.com> <6.2.3.4.0.20050830132902.021a5308@pop.earthlink.net> <4314B329.20009@commandprompt.com> <6.2.3.4.0.20050830154823.05ac9308@pop.earthlink.net> <20050831000443.GZ14921@mathom.us> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bc44d7f6b7cee5b1b2aad82a9aecd5c5ff350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.4 required=5 tests=[AWL=0.026, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/552 X-Sequence-Number: 14299 At 08:04 PM 8/30/2005, Michael Stone wrote: >On Tue, Aug 30, 2005 at 07:02:28PM -0400, Ron wrote: >>purpose(s). That's why the TPC bench marked systems tend to have >>literally 100's of HD's and they tend to be split into very focused purposes. > >Of course, TPC benchmark systems are constructed such that cost and >storage capacity are irrelevant--in the real world things tend to be >more complicated. The scary thing is that I've worked on RW production systems that bore a striking resemblance to a TPC benchmark system. As you can imagine, they uniformly belonged to BIG organizations (read: lot's 'o $$$) who were using the systems for mission critical stuff where either it was company existence threatening for the system to be done, or they would lose much $$$ per min of down time, or both. Financial institutions, insurance companies, central data mines for Fortune 2000 companies, etc _all_ build systems that push the state of the art in how much storage can be managed and how many HDs, CPUs, RAM DIMMs, etc are usable. Historically, this has been the sole province of Oracle and DB2 on the SW side and equally outrageously priced custom HW. Clearly, I'd like to see PostgreSQL change that ;-) Ron Peacetree From pgsql-performance-owner@postgresql.org Tue Aug 30 21:43:38 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D6954D6E40 for ; Tue, 30 Aug 2005 21:43:36 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 19136-06 for ; Wed, 31 Aug 2005 00:43:35 +0000 (GMT) Received: from vms042pub.verizon.net (vms042pub.verizon.net [206.46.252.42]) by svr1.postgresql.org (Postfix) with ESMTP id D9C37D86BE for ; Tue, 30 Aug 2005 21:43:33 -0300 (ADT) Received: from osgiliath.mathom.us ([151.200.14.55]) by vms042.mailsrvcs.net (Sun Java System Messaging Server 6.2-2.05 (built Apr 28 2005)) with ESMTPA id <0IM200HZ7A0LHJM1@vms042.mailsrvcs.net> for pgsql-performance@postgresql.org; Tue, 30 Aug 2005 19:43:37 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by osgiliath.mathom.us (Postfix) with ESMTP id 802FF600652; Tue, 30 Aug 2005 20:43:32 -0400 (EDT) Received: from osgiliath.mathom.us ([127.0.0.1]) by localhost (osgiliath [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 12171-04-5; Tue, 30 Aug 2005 20:43:32 -0400 (EDT) Received: by osgiliath.mathom.us (Postfix, from userid 1000) id 5185D60041D; Tue, 30 Aug 2005 20:43:32 -0400 (EDT) Date: Tue, 30 Aug 2005 20:43:32 -0400 From: Michael Stone Subject: Re: RAID Configuration Sugestion In-reply-to: <6.2.3.4.0.20050830203126.0221feb8@pop.earthlink.net> To: Ron Cc: Michael Stone , pgsql-performance@postgresql.org Mail-Followup-To: Ron , Michael Stone , pgsql-performance@postgresql.org Message-id: <20050831004332.GA14921@mathom.us> MIME-version: 1.0 Content-type: text/plain; charset=us-ascii; format=flowed Content-disposition: inline X-Pgp-Fingerprint: 53 FF 38 00 E7 DD 0A 9C 84 52 84 C5 EE DF 7C 88 X-Virus-Scanned: by amavisd-new-20030616-p10 (Debian) at mathom.us References: <431452FD.3090402@atua.com.br> <6.2.3.4.0.20050830103240.0219b640@pop.earthlink.net> <43148FC4.4000407@commandprompt.com> <6.2.3.4.0.20050830132902.021a5308@pop.earthlink.net> <4314B329.20009@commandprompt.com> <6.2.3.4.0.20050830154823.05ac9308@pop.earthlink.net> <20050831000443.GZ14921@mathom.us> <6.2.3.4.0.20050830203126.0221feb8@pop.earthlink.net> User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.039 required=5 tests=[AWL=-0.011, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/553 X-Sequence-Number: 14300 On Tue, Aug 30, 2005 at 08:41:40PM -0400, Ron wrote: >The scary thing is that I've worked on RW production systems that >bore a striking resemblance to a TPC benchmark system. As you can >imagine, they uniformly belonged to BIG organizations (read: lot's 'o >$$$) who were using the systems for mission critical stuff where >either it was company existence threatening for the system to be >done, or they would lose much $$$ per min of down time, or both. Yeah, and that market is relevant to someone with one dell server and 2 hard disks how? Mike Stone From pgsql-performance-owner@postgresql.org Tue Aug 30 23:51:04 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id AD4E4D6EBA for ; Tue, 30 Aug 2005 23:51:03 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 50664-10 for ; Wed, 31 Aug 2005 02:50:57 +0000 (GMT) Received: from smtpauth06.mail.atl.earthlink.net (smtpauth06.mail.atl.earthlink.net [209.86.89.66]) by svr1.postgresql.org (Postfix) with ESMTP id 80148D6E65 for ; Tue, 30 Aug 2005 23:50:55 -0300 (ADT) DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=dk20050327; d=earthlink.net; b=LIZrgEQ1vh0PGzpk3LTad4UNC+uPiBZXUg+sGJB8YD0ZW+bTncjHcy5dBAFQrVfz; h=Received:Message-Id:X-Mailer:Date:To:From:Subject:In-Reply-To:References:Mime-Version:Content-Type:X-ELNK-Trace:X-Originating-IP; Received: from [24.41.3.216] (helo=RJP-01.earthlink.net) by smtpauth06.mail.atl.earthlink.net with asmtp (Exim 4.34) id 1EAIgm-0008PZ-CI; Tue, 30 Aug 2005 22:51:00 -0400 Message-Id: <6.2.3.4.0.20050830224550.05ac9828@pop.earthlink.net> X-Mailer: QUALCOMM Windows Eudora Version 6.2.3.4 Date: Tue, 30 Aug 2005 22:50:57 -0400 To: Michael Stone , pgsql-performance@postgresql.org From: Ron Subject: Re: RAID Configuration Sugestion In-Reply-To: <20050831004332.GA14921@mathom.us> References: <431452FD.3090402@atua.com.br> <6.2.3.4.0.20050830103240.0219b640@pop.earthlink.net> <43148FC4.4000407@commandprompt.com> <6.2.3.4.0.20050830132902.021a5308@pop.earthlink.net> <4314B329.20009@commandprompt.com> <6.2.3.4.0.20050830154823.05ac9308@pop.earthlink.net> <20050831000443.GZ14921@mathom.us> <6.2.3.4.0.20050830203126.0221feb8@pop.earthlink.net> <20050831004332.GA14921@mathom.us> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed X-ELNK-Trace: acd68a6551193be5d780f4a490ca69563f9fea00a6dd62bc05468f44d6420521445079c5f3598995350badd9bab72f9c350badd9bab72f9c350badd9bab72f9c X-Originating-IP: 24.41.3.216 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.395 required=5 tests=[AWL=0.021, DNS_FROM_RFC_ABUSE=0.374] X-Spam-Level: X-Archive-Number: 200508/554 X-Sequence-Number: 14301 At 08:43 PM 8/30/2005, Michael Stone wrote: >On Tue, Aug 30, 2005 at 08:41:40PM -0400, Ron wrote: >>The scary thing is that I've worked on RW production systems that >>bore a striking resemblance to a TPC benchmark system. As you can >>imagine, they uniformly belonged to BIG organizations (read: lot's >>'o $$$) who were using the systems for mission critical stuff where >>either it was company existence threatening for the system to be >>done, or they would lose much $$$ per min of down time, or both. > >Yeah, and that market is relevant to someone with one dell server >and 2 hard disks how? Because successful small companies that _start_ with one small server and 2 HDs grow to _become_ companies that need far more HW; ...and in the perfect world their SW scales to their increased needs... _Without_ exponentially increasing their costs or overhead (as Oracle and DB2 currently do) THIS is the real long term promise of OS DBMS. Ron Peacetree From pgsql-performance-owner@postgresql.org Wed Aug 31 02:18:02 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id EF206D880E for ; Wed, 31 Aug 2005 02:18:01 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 99800-07 for ; Wed, 31 Aug 2005 05:17:58 +0000 (GMT) Received: from calypso.bi.lt (calypso.bi.lt [213.226.153.10]) by svr1.postgresql.org (Postfix) with ESMTP id 087BCD8831 for ; Wed, 31 Aug 2005 02:17:54 -0300 (ADT) Received: by calypso.bi.lt (Postfix, from userid 101) id E7E9C48054F; Wed, 31 Aug 2005 08:17:53 +0300 (EEST) X-Original-To: pgsql-performance@postgresql.org Received: from B027543 (inet.bee.lt [213.226.131.30]) by calypso.bi.lt (Postfix) with SMTP id D60F94804CC for ; Wed, 31 Aug 2005 08:17:53 +0300 (EEST) Message-ID: <00f601c5adeb$56fcb910$f20214ac@bite.lt> From: "Mindaugas Riauba" To: References: <4314DBE0.9070708@telogis.com> <200508301558.34697.josh@agliodbs.com> Subject: Re: 'Real' auto vacuum? Date: Wed, 31 Aug 2005 08:17:53 +0300 MIME-Version: 1.0 Content-Type: text/plain; charset="ISO-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 6.00.2800.1506 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1506 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.09 required=5 tests=[AWL=0.090] X-Spam-Level: X-Archive-Number: 200508/555 X-Sequence-Number: 14302 > > When a row is orphaned it's added to a list of possibly available rows. > > When a new row is needed the list of possible rows is examined and the > > first one with a transaction id less then the lowest running transaction > > id is chosen to be the new row? These rows can be in a heap so it's > > really fast to find one. > > This is the long-term plan. However, it's actually a lot harder than it > sounds. Patches welcome. Some ETA? Since that would be the most welcome addition for us. We have few very heavily updated databases where table bloat and constant vacuuming is killing performance. Mindaugas From pgsql-performance-owner@postgresql.org Tue Sep 6 00:46:09 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 10F6AD834F for ; Wed, 31 Aug 2005 07:22:50 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 85447-05 for ; Wed, 31 Aug 2005 10:22:47 +0000 (GMT) Received: from vey.eyjar.is (vey.eyjar.is [194.144.184.1]) by svr1.postgresql.org (Postfix) with ESMTP id A9EBFD8347 for ; Wed, 31 Aug 2005 07:22:45 -0300 (ADT) Received: from www.eyjar.is (localhost [127.0.0.1]) by vey.eyjar.is (8.12.9-20030918-gdh-p1/8.12.9) with SMTP id j7VAMhUO020800; Wed, 31 Aug 2005 10:22:44 GMT Received: from 213.220.100.208 (SquirrelMail authenticated user gdh) by www.eyjar.is with HTTP; Wed, 31 Aug 2005 10:22:44 -0000 (GMT) Message-ID: <1913.213.220.100.208.1125483764.squirrel@www.eyjar.is> Date: Wed, 31 Aug 2005 10:22:44 -0000 (GMT) Subject: Query slow after VACUUM ANALYZE From: gdh@eyjar.is To: pgsql-performance@postgresql.org Cc: gdh@eyjar.is User-Agent: SquirrelMail/1.4.2 MIME-Version: 1.0 Content-Type: text/plain;charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Priority: 3 Importance: Normal X-Spam-Search-By: milter1.eyjar.is X-Spam-Search-Via: spamass-milter 0.5pre3 X-Spam-DCC: SIHOPE-DCC-3: milter1.eyjar.is 1085; Body=1 Fuz1=1 Fuz2=1 X-Virus-Scanned: by AMaViS - amavis-milter (http://www.amavis.org/) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.178 required=5 tests=[NO_REAL_NAME=0.178] X-Spam-Level: X-Archive-Number: 200509/73 X-Sequence-Number: 14381 Hi all I'm having a strange problem with a query which looks like this: SELECT id FROM orders WHERE id NOT IN (SELECT order_id FROM orders_items); The id fields are varchars (32), both indexed. The number of rows in the tables are about 60000. Now, the really strange part is if I delete all data from orders_items, run VACUUM ANALYZE, then import all the data, the query finshes in about 3 seconds. Then I run VACUUM ANALYZE, and *after* the vacuum, the query takes about 30 minutes to run. The data is the same and this is the only query running, and the machine load is effectively none. EXPLAIN'ng the query shows, before VACUUM ANALYZE, shows this: QUERY PLAN ------------------------------------------------------------------------- Seq Scan on orders (cost=0.00..12184.14 rows=29526 width=33) Filter: (NOT (hashed subplan)) SubPlan -> Seq Scan on orders_items (cost=0.00..0.00 rows=1 width=33) After the vacuum, the plan is like this: QUERY PLAN -------------------------------------------------------------------------------- Seq Scan on fsi_orders (cost=0.00..40141767.46 rows=29526 width=33) Filter: (NOT (subplan)) SubPlan -> Seq Scan on fsi_orders_items (cost=0.00..1208.12 rows=60412 width=33) Any ideas what I can do to make the query running in < 10 seconds? Thanks, Gu�mundur. From pgsql-performance-owner@postgresql.org Tue Sep 6 01:02:08 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 762C4D8373 for ; Wed, 31 Aug 2005 07:26:28 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 84495-07 for ; Wed, 31 Aug 2005 10:26:24 +0000 (GMT) Received: from vey.eyjar.is (vey.eyjar.is [194.144.184.1]) by svr1.postgresql.org (Postfix) with ESMTP id 8CBB5D8368 for ; Wed, 31 Aug 2005 07:26:23 -0300 (ADT) Received: from www.eyjar.is (localhost [127.0.0.1]) by vey.eyjar.is (8.12.9-20030918-gdh-p1/8.12.9) with SMTP id j7VAQMUO021308; Wed, 31 Aug 2005 10:26:22 GMT Received: from 213.220.100.208 (SquirrelMail authenticated user gdh) by www.eyjar.is with HTTP; Wed, 31 Aug 2005 10:26:22 -0000 (GMT) Message-ID: <1923.213.220.100.208.1125483982.squirrel@www.eyjar.is> In-Reply-To: <1913.213.220.100.208.1125483764.squirrel@www.eyjar.is> References: <1913.213.220.100.208.1125483764.squirrel@www.eyjar.is> Date: Wed, 31 Aug 2005 10:26:22 -0000 (GMT) Subject: Re: Query slow after VACUUM ANALYZE From: gdh@eyjar.is To: gdh@eyjar.is Cc: pgsql-performance@postgresql.org, gdh@eyjar.is User-Agent: SquirrelMail/1.4.2 MIME-Version: 1.0 Content-Type: text/plain;charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Priority: 3 Importance: Normal X-Spam-Search-By: milter1.eyjar.is X-Spam-Search-Via: spamass-milter 0.5pre3 X-Spam-DCC: NIET: milter1.eyjar.is 1080; Body=1 Fuz1=1 Fuz2=1 X-Virus-Scanned: by AMaViS - amavis-milter (http://www.amavis.org/) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.178 required=5 tests=[NO_REAL_NAME=0.178] X-Spam-Level: X-Archive-Number: 200509/78 X-Sequence-Number: 14386 Hi again [..] > > QUERY PLAN > ------------------------------------------------------------------------- > Seq Scan on orders (cost=0.00..12184.14 rows=29526 width=33) > Filter: (NOT (hashed subplan)) > SubPlan > -> Seq Scan on orders_items (cost=0.00..0.00 rows=1 width=33) > > After the vacuum, the plan is like this: > > QUERY PLAN > -------------------------------------------------------------------------------- > Seq Scan on fsi_orders (cost=0.00..40141767.46 rows=29526 width=33) > Filter: (NOT (subplan)) > SubPlan > -> Seq Scan on fsi_orders_items (cost=0.00..1208.12 rows=60412 > width=33) > This, of course, should be "orders", not "fsi_orders", and "orders_items", not "fsi_orders_items". Sorry for the confusion. Additional info: I'm running PostgreSQL 7.4.8. Thanks, Gu�mundur. From pgsql-performance-owner@postgresql.org Wed Aug 31 10:51:43 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2EF52D8527 for ; Wed, 31 Aug 2005 10:51:42 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 46809-09 for ; Wed, 31 Aug 2005 13:51:40 +0000 (GMT) Received: from mail.libertyrms.com (unknown [207.219.45.62]) by svr1.postgresql.org (Postfix) with ESMTP id 872D6D8509 for ; Wed, 31 Aug 2005 10:51:38 -0300 (ADT) Received: from dba5.int.libertyrms.com ([10.1.3.44]) by mail.libertyrms.com with esmtp (Exim 4.22) id 1EAT0B-0005xV-Ej for pgsql-performance@postgresql.org; Wed, 31 Aug 2005 09:51:43 -0400 Message-ID: <4315B5EF.5020509@ca.afilias.info> Date: Wed, 31 Aug 2005 09:51:43 -0400 From: Brad Nicholson User-Agent: Mozilla Thunderbird 0.8 (X11/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Re: 'Real' auto vacuum? References: <4314DBE0.9070708@telogis.com> <200508301558.34697.josh@agliodbs.com> <00f601c5adeb$56fcb910$f20214ac@bite.lt> In-Reply-To: <00f601c5adeb$56fcb910$f20214ac@bite.lt> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-SA-Exim-Mail-From: bnichols@ca.afilias.info X-SA-Exim-Scanned: No; SAEximRunCond expanded to false X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.311 required=5 tests=[AWL=-0.220, FORGED_RCVD_HELO=0.05, INFO_TLD=0.481] X-Spam-Level: X-Archive-Number: 200508/556 X-Sequence-Number: 14303 Mindaugas Riauba wrote: >>>When a row is orphaned it's added to a list of possibly available rows. >>>When a new row is needed the list of possible rows is examined and the >>>first one with a transaction id less then the lowest running transaction >>>id is chosen to be the new row? These rows can be in a heap so it's >>>really fast to find one. >>> >>> >>This is the long-term plan. However, it's actually a lot harder than it >>sounds. Patches welcome. >> >> > > Some ETA? Since that would be the most welcome addition for us. We >have few very heavily updated databases where table bloat and constant >vacuuming is killing performance. > > > How often are you vacuuming (the definition of 'constantly' tends to vary)? Are you vacuuming the whole database each time? If so, identify which tables are being updated frequently, and vacuum those often. Vacuum other tables less frequently. Also, are you you using VACUUM FULL (if so, you certainly don't want to be). -- Brad Nicholson 416-673-4106 bnichols@ca.afilias.info Database Administrator, Afilias Canada Corp. From pgsql-performance-owner@postgresql.org Wed Aug 31 15:25:21 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B389FD852F for ; Wed, 31 Aug 2005 15:25:19 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 10680-06 for ; Wed, 31 Aug 2005 18:25:16 +0000 (GMT) Received: from web53209.mail.yahoo.com (web53209.mail.yahoo.com [206.190.49.79]) by svr1.postgresql.org (Postfix) with SMTP id 2B811D842B for ; Wed, 31 Aug 2005 15:25:15 -0300 (ADT) Received: (qmail 89933 invoked by uid 60001); 31 Aug 2005 18:25:15 -0000 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com.br; h=Message-ID:Received:Date:From:Subject:To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=F1lwtU6uvUNLkADAkLvTSFq+O50g8f5T6ynl8XBGWFyPh8fLJ1hx7R1Q5GK+0Jxw4XmLgBy1swoGRyBw9y8U6hhgsDiuoD4APiVjTqo9md6zsfTcqwTEAakyhMFmtXByKSY0Lx9nwZw04Gp8H6yh3acsa4gjdqki584Ue3ENec0= ; Message-ID: <20050831182515.89931.qmail@web53209.mail.yahoo.com> Received: from [201.24.119.165] by web53209.mail.yahoo.com via HTTP; Wed, 31 Aug 2005 15:25:15 ART Date: Wed, 31 Aug 2005 15:25:15 -0300 (ART) From: Carlos Henrique Reimer Subject: Swapping To: pgsql-performance@postgresql.org MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="0-353803500-1125512715=:87763" Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.596 required=5 tests=[AWL=0.165, DNS_FROM_RFC_ABUSE=0.374, HTML_30_40=0.056, HTML_MESSAGE=0.001] X-Spam-Level: X-Archive-Number: 200508/557 X-Sequence-Number: 14304 --0-353803500-1125512715=:87763 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit Hi, I�m trying to tune a linux box with a 12 GB database and 4 GB RAM. First of all I would like to stop the swapping, so the shared_buffers and sort_mem were decreased but even so it started swapping two hours after DBMS started up. I would like to know some suggestions how to discover why is it swapping? I�ve collected the following data from the environment and saved at http://paginas.terra.com.br/educacao/rei/dados.htm 1. select version() 2. uname -a 3. cat /proc/cpuinfo 4. cat /proc/meminfo 5. vmstat 5 6. pg_stat_activity 7. postgresql.conf Thanks in advance! Reimer __________________________________________________ Converse com seus amigos em tempo real com o Yahoo! Messenger http://br.download.yahoo.com/messenger/ --0-353803500-1125512715=:87763 Content-Type: text/html; charset=iso-8859-1 Content-Transfer-Encoding: 8bit
Hi,
 
I�m trying to tune a linux box with a 12 GB database and 4 GB RAM. First of all I would like to stop the swapping, so the shared_buffers and sort_mem were decreased but even so it started swapping two hours after DBMS started up.
 
I would like to know some suggestions how to discover why is it swapping?
 
I�ve collected the following data from the environment and saved at http://paginas.terra.com.br/educacao/rei/dados.htm
 
1. select version()
2. uname -a
3. cat /proc/cpuinfo
4. cat /proc/meminfo
5. vmstat 5
6. pg_stat_activity
7. postgresql.conf
 
Thanks in advance!
 
Reimer

__________________________________________________
Converse com seus amigos em tempo real com o Yahoo! Messenger
http://br.download.yahoo.com/messenger/ --0-353803500-1125512715=:87763-- From pgsql-performance-owner@postgresql.org Wed Aug 31 15:49:22 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id A578FD79BC for ; Wed, 31 Aug 2005 15:49:21 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 16548-06 for ; Wed, 31 Aug 2005 18:49:18 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id B7424D7708 for ; Wed, 31 Aug 2005 15:49:17 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.13.1/8.13.1) with ESMTP id j7VInIZJ003610; Wed, 31 Aug 2005 14:49:18 -0400 (EDT) To: Carlos Henrique Reimer Cc: pgsql-performance@postgresql.org Subject: Re: Swapping In-reply-to: <20050831182515.89931.qmail@web53209.mail.yahoo.com> References: <20050831182515.89931.qmail@web53209.mail.yahoo.com> Comments: In-reply-to Carlos Henrique Reimer message dated "Wed, 31 Aug 2005 15:25:15 -0300" Date: Wed, 31 Aug 2005 14:49:18 -0400 Message-ID: <3609.1125514158@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.006 required=5 tests=[AWL=0.006] X-Spam-Level: X-Archive-Number: 200508/558 X-Sequence-Number: 14305 Carlos Henrique Reimer writes: > I would like to know some suggestions how to discover why is it swapping? Zero swap-in rate and swap-out rates in the single digits do not constitute a swapping problem. It's reasonably likely that that traffic isn't even coming from Postgres, but something else. I'd say ignore it. regards, tom lane From pgsql-performance-owner@postgresql.org Wed Aug 31 16:17:44 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 18F4ED8760 for ; Wed, 31 Aug 2005 16:17:42 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 20036-09 for ; Wed, 31 Aug 2005 19:17:39 +0000 (GMT) Received: from neutron.verseon.com (adsl-63-204-199-153.dsl.snfc21.pacbell.net [63.204.199.153]) by svr1.postgresql.org (Postfix) with ESMTP id E4232D8749 for ; Wed, 31 Aug 2005 16:17:37 -0300 (ADT) content-class: urn:content-classes:message Subject: Big question on insert performance/using COPY FROM Date: Wed, 31 Aug 2005 12:17:38 -0700 Message-ID: <08B420FF5BF7BC42A064212C2EB768801C10BE@neutron.verseon.com> X-MS-Has-Attach: MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable X-MS-TNEF-Correlator: Thread-Topic: Big question on insert performance/using COPY FROM Thread-Index: AcWuYKag40d30iTBS4Gf8Alpk4yOrg== X-MimeOLE: Produced By Microsoft Exchange V6.0.6556.0 From: "Morgan Kita" To: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.05 required=5 tests=[AWL=-0.000, FORGED_RCVD_HELO=0.05] X-Spam-Level: X-Archive-Number: 200508/559 X-Sequence-Number: 14306 Hi, I am currently trying to speed up the insertion of bulk loads to my = database. I have fiddled with all of the parameters that I have seen = suggested(aka checkpoint_segments, checkpoint_timeout, = maintinence_work_mem, and shared buffers) with no success. I even turned = off fysnc with no effect so I am pretty sure the biggest problem is that = the DB is CPU limited at the moment because of the rather weak machine = that postmaster is running on(Athlon 2400+ xp with 512 RAM), but that = will change in the future so I am trying to get performance increases = that don't involve changing the machine at the moment. I am currently inserting into the database through lipqxx's C++ = interface. I am using prepared statements that perform regular inserts. = I would like to use COPY FROM since I have read so much about its = increased performance with respect to INSERT, but I am not sure how to = use it in my case. So let me give you an idea on how the tables are laid = out.=20 The real DB has more tables, but lets say for the sake of argument I = have 3 tables; TB1, TB2, TB3. Lets say that TB1 has a primary key PK1 = and a unique identifier column(type text) UK1 that has an index on it. = TB2 then has a PK2, a UK2(type text) of its own with an index, and a = foreign key FK2 that points to TB1's PK1. TB3 has a PK3 and a FK3 that = points to FK2.=20 TB1 TB2 TB3 -------------- ------------------------------- = ---------------------- PK1, UK1 PK2, UK2, FK2(PK1) PK3, FK3(PK2) Now in lipqxx I am parsing an input list of objects that are then = written to these tables. Each object may produce one row in TB1, one row = in TB2, and one row in TB3. The UK1 and UK2 indentifiers are used to = prevent duplicate entries for TB1 and TB2 respectively. I know COPY FROM = obeys these unique checks; however, my problem is the FKs. So lets say I = try to insert a row into TB1. If it is unique on UK1 then it inserts a = new row with some new primary key int4 identifier and if it is a = duplicate then no insert is done but the already existing row's primary = key identifier is returned. This identifier(duplicate or not) is used = when populating TB2's row as the FK2 identifier. The row that is to be = inserted into TB2 needs the primary key indentifier from the result of = the attempted insert into TB1. Similarily the insert into TB3 needs the = result of the pk indentifier of the attempted insert into TB2. Once that = is done then I move on to parsing the next object for insertion into the = 3 tables. So lets say I want to insert a list of objects using COPY FROM... whats = the way to do it? How can I at the very least get a list of the primary = keys of TB1(newly inserted rows or from already existings row) returned = from the COPY FROM insert into TB1 so I can use them for the COPY FROM = insert into TB2 and so on? Is there a better way to do this? P.S. I am going to setup autovacuum for these bulk loads. My question = though is why for bulkloads is VACUUM useful? I understand that it frees = up dead rows as a result of UPDATE and such, but where are the dead rows = created from plain INSERTS? Thanks, Morgan From pgsql-performance-owner@postgresql.org Wed Aug 31 16:28:07 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4FCB3D8760 for ; Wed, 31 Aug 2005 16:22:33 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 21977-10 for ; Wed, 31 Aug 2005 19:22:31 +0000 (GMT) Received: from mail.nordicbet.com (mail.nordicbet.com [193.69.167.130]) by svr1.postgresql.org (Postfix) with ESMTP id 064C0D8754 for ; Wed, 31 Aug 2005 16:22:29 -0300 (ADT) Received: from [80.203.125.83] (helo=tobias.nordicbet.invalid) (Authenticated Sender=tobias@nordicbet.com) by mail.nordicbet.com with esmtpa (Exim 4.50 #1 (Debian)) id 1EAY7X-0004R0-Bp; Wed, 31 Aug 2005 21:19:49 +0200 Received: by tobias.nordicbet.invalid (Postfix, from userid 500) id 2466AAAEA; Wed, 31 Aug 2005 21:22:17 +0200 (CEST) Date: Wed, 31 Aug 2005 21:22:17 +0200 From: Tobias Brox To: Carlos Henrique Reimer Cc: pgsql-performance@postgresql.org Subject: Re: Swapping Message-ID: <20050831192217.GC5736@tobias.lan> References: <20050831182515.89931.qmail@web53209.mail.yahoo.com> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Disposition: inline In-Reply-To: <20050831182515.89931.qmail@web53209.mail.yahoo.com> Organization: Group Nordicbet User-Agent: Mutt/1.5.10i Content-Transfer-Encoding: quoted-printable X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/560 X-Sequence-Number: 14307 [Carlos Henrique Reimer - Wed at 03:25:15PM -0300] > I=B4m trying to tune a linux box with a 12 GB database and 4 GB RAM. Fi= rst > of all I would like to stop the swapping, so the shared_buffers and sor= t_mem > were decreased but even so it started swapping two hours after DBMS sta= rted > up. > =20 > I would like to know some suggestions how to discover why is it swappin= g? I agree with Tom Lane, nothing to worry about. Swapping is not a problem per se, aggressive swapping is a problem. If you are absolutely sure you want to ban all swapping, use "swapoff"? I'd trust linux to handle swap/cache sensibly. Eventually, become involv= ed with kernel hacking ;-) --=20 Notice of Confidentiality: This email is sent unencrypted over the networ= k, and may be stored on several email servers; it can be read by third parti= es as easy as a postcard. Do not rely on email for confidential information= . From pgsql-performance-owner@postgresql.org Wed Aug 31 17:37:59 2005 X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4BEE4D8695 for ; Wed, 31 Aug 2005 16:33:40 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 26357-02 for ; Wed, 31 Aug 2005 19:33:35 +0000 (GMT) Received: from mail.nordicbet.com (mail.nordicbet.com [193.69.167.130]) by svr1.postgresql.org (Postfix) with ESMTP id 2C949D84B4 for ; Wed, 31 Aug 2005 16:33:32 -0300 (ADT) Received: from [80.203.125.83] (helo=tobias.nordicbet.invalid) (Authenticated Sender=tobias@nordicbet.com) by mail.nordicbet.com with esmtpa (Exim 4.50 #1 (Debian)) id 1EAYIJ-0004y4-0k; Wed, 31 Aug 2005 21:30:56 +0200 Received: by tobias.nordicbet.invalid (Postfix, from userid 500) id 76AC5AAEA; Wed, 31 Aug 2005 21:33:17 +0200 (CEST) Date: Wed, 31 Aug 2005 21:33:17 +0200 From: Tobias Brox To: Tobias Brox Cc: Carlos Henrique Reimer , pgsql-performance@postgresql.org Subject: Re: Swapping Message-ID: <20050831193317.GD5736@tobias.lan> References: <20050831182515.89931.qmail@web53209.mail.yahoo.com> <20050831192217.GC5736@tobias.lan> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20050831192217.GC5736@tobias.lan> Organization: Group Nordicbet User-Agent: Mutt/1.5.10i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0 required=5 tests=[none] X-Spam-Level: X-Archive-Number: 200508/561 X-Sequence-Number: 14308 [Tobias Brox - Wed at 09:22:17PM +0200] > I'd trust linux to handle swap/cache sensibly. Eventually, become involved > with kernel hacking ;-) Of course, there are also some files in /proc/sys/vm that you may want to peek into, for tuning the swapping. Particularly, at later 2.6-kernels (I'm running 2.6.12) you have the file /proc/sys/vm/swappiness, where the number should be some percentage. I'm not completely sure how it works, but I suppose that the higher you set it, the more likely it is to swap out memory not beeing used. I think the default setting is probably sane, but you may want to google a bit about it. -- Notice of Confidentiality: This email is sent unencrypted over the network, and may be stored on several email servers; it can be read by third parties as easy as a postcard. Do not rely on email for confidential information.