From pgsql-performance-owner@postgresql.org Fri Oct 1 06:46:28 2004 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 0721732A2B4 for ; Fri, 1 Oct 2004 06:46:28 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 61130-09 for ; Fri, 1 Oct 2004 05:46:24 +0000 (GMT) Received: from bayswater1.ymogen.net (host-154-240-27-217.pobox.net.uk [217.27.240.154]) by svr1.postgresql.org (Postfix) with ESMTP id DAF7032A29B for ; Fri, 1 Oct 2004 06:46:23 +0100 (BST) Received: from [82.68.132.233] (82-68-132-233.dsl.in-addr.zen.co.uk [82.68.132.233]) by bayswater1.ymogen.net (Postfix) with ESMTP id C6DBFA36F8; Fri, 1 Oct 2004 06:46:21 +0100 (BST) Message-ID: <415CEE8E.8060805@ymogen.net> Date: Fri, 01 Oct 2004 06:43:42 +0100 From: Matt Clark User-Agent: Mozilla Thunderbird 0.7.3 (Windows/20040803) X-Accept-Language: en-us, en MIME-Version: 1.0 To: "Jim C. Nasby" Cc: Postgresql Performance Subject: Re: Caching of Queries References: <00e801c4a4b6$5526dd30$8300a8c0@solent> <1096306670.40463.173.camel@jester> <41587867.4090505@ymogen.net> <20040930221107.GP1297@decibel.org> In-Reply-To: <20040930221107.GP1297@decibel.org> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/1 X-Sequence-Number: 8477 >If you're not using a connection pool of some kind then you might as >well forget query plan caching, because your connect overhead will swamp >the planning cost. This does not mean you have to use something like >pgpool (which makes some rather questionable claims IMO); any decent web >application language/environment will support connection pooling. > > Hmm, a question of definition - there's a difference between a pool and a persistent connection. Pretty much all web apps have one connection per process, which is persistent (i.e. not dropped and remade for each request), but not shared between processes, therefore not pooled. From pgsql-performance-owner@postgresql.org Fri Oct 1 07:31:08 2004 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 C9E6932A3A0 for ; Fri, 1 Oct 2004 07:31:07 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 73673-01 for ; Fri, 1 Oct 2004 06:31:03 +0000 (GMT) Received: from purple.west.spy.net (mail.west.spy.net [66.149.231.226]) by svr1.postgresql.org (Postfix) with ESMTP id A8AFC32A2D2 for ; Fri, 1 Oct 2004 07:30:58 +0100 (BST) Received: from [192.168.1.50] (dustinti.west.spy.net [192.168.1.50]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (Client did not present a certificate) by purple.west.spy.net (Postfix) with ESMTP id 4B62D5E for ; Thu, 30 Sep 2004 23:38:13 -0700 (PDT) Mime-Version: 1.0 (Apple Message framework v619) Content-Transfer-Encoding: 7bit Message-Id: <6FA9AB99-1373-11D9-A87D-000A957659CC@spy.net> Content-Type: text/plain; charset=US-ASCII; format=flowed To: pgsql-performance@postgresql.org From: Dustin Sallings Subject: inconsistent/weird index usage Date: Thu, 30 Sep 2004 23:30:49 -0700 X-Mailer: Apple Mail (2.619) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/2 X-Sequence-Number: 8478 To save some time, let me start by saying PostgreSQL 7.4.3 on powerpc-apple-darwin7.4.0, compiled by GCC gcc (GCC) 3.3 20030304 (Apple Computer, Inc. build 1640) OK, now on to details... I'm trying to implement oracle style ``partitions'' in postgres. I've run into my first snag on what should be a fairly quick query. Basically, I started with the following schema and split the ``samples'' table into one table for each year (1999-2004). -- BEGIN SCHEMA create table sensor_types ( sensor_type_id serial, sensor_type text not null, units varchar(10) not null, primary key(sensor_type_id) ); create table sensors ( sensor_id serial, sensor_type_id integer not null, serial char(16) not null, name text not null, low smallint not null, high smallint not null, active boolean default true, primary key(sensor_id), foreign key(sensor_type_id) references sensor_types(sensor_type_id) ); create unique index sensors_byserial on sensors(serial); create table samples ( ts datetime not null, sensor_id integer not null, sample float not null, foreign key(sensor_id) references sensors(sensor_id) ); create index samples_bytime on samples(ts); create unique index samples_bytimeid on samples(ts, sensor_id); -- END SCHEMA Each samples_[year] table looks, and is indexed exactly as the above samples table was by using the following commands: create index samples_1999_bytime on samples_1999(ts); create index samples_2000_bytime on samples_2000(ts); create index samples_2001_bytime on samples_2001(ts); create index samples_2002_bytime on samples_2002(ts); create index samples_2003_bytime on samples_2003(ts); create index samples_2004_bytime on samples_2004(ts); create unique index samples_1999_bytimeid on samples_1999(ts, sensor_id); create unique index samples_2000_bytimeid on samples_2000(ts, sensor_id); create unique index samples_2001_bytimeid on samples_2001(ts, sensor_id); create unique index samples_2002_bytimeid on samples_2002(ts, sensor_id); create unique index samples_2003_bytimeid on samples_2003(ts, sensor_id); create unique index samples_2004_bytimeid on samples_2004(ts, sensor_id); The tables contain the following number of rows: samples_1999 311030 samples_2000 2142245 samples_2001 2706571 samples_2002 3111602 samples_2003 3149316 samples_2004 2375972 The following view creates the illusion of the old ``single-table'' model: create view samples as select * from samples_1999 union select * from samples_2000 union select * from samples_2001 union select * from samples_2002 union select * from samples_2003 union select * from samples_2004 ...along with the following rule on the view for the applications performing inserts: create rule sample_rule as on insert to samples do instead insert into samples_2004 (ts, sensor_id, sample) values(new.ts, new.sensor_id, new.sample) OK, now that that's over with, I have this one particular query that I attempt to run for a report from my phone that no longer works because it tries to do a table scan on *some* of the tables. Why it chooses this table scan, I can't imagine. The query is as follows: select s.serial as serial_num, s.name as name, date(ts) as day, min(sample) as min_temp, avg(sample) as avg_temp, stddev(sample) as stddev_temp, max(sample) as max_temp from samples inner join sensors s using (sensor_id) where ts > current_date - 7 group by serial_num, name, day order by serial_num, day desc explain analyze reports the following (sorry for the horrible wrapping): Sort (cost=1185281.45..1185285.95 rows=1800 width=50) (actual time=82832.106..82832.147 rows=56 loops=1) Sort Key: s.serial, date(samples.ts) -> HashAggregate (cost=1185161.62..1185184.12 rows=1800 width=50) (actual time=82830.624..82831.601 rows=56 loops=1) -> Hash Join (cost=1063980.21..1181539.96 rows=206952 width=50) (actual time=80408.123..81688.590 rows=66389 loops=1) Hash Cond: ("outer".sensor_id = "inner".sensor_id) -> Subquery Scan samples (cost=1063979.10..1155957.38 rows=4598914 width=20) (actual time=80392.477..80922.764 rows=66389 loops=1) -> Unique (cost=1063979.10..1109968.24 rows=4598914 width=20) (actual time=80392.451..80646.761 rows=66389 loops=1) -> Sort (cost=1063979.10..1075476.39 rows=4598914 width=20) (actual time=80392.437..80442.787 rows=66389 loops=1) Sort Key: ts, sensor_id, sample -> Append (cost=0.00..312023.46 rows=4598914 width=20) (actual time=79014.428..80148.396 rows=66389 loops=1) -> Subquery Scan "*SELECT* 1" (cost=0.00..9239.37 rows=103677 width=20) (actual time=4010.181..4010.181 rows=0 loops=1) -> Seq Scan on samples_1999 (cost=0.00..8202.60 rows=103677 width=20) (actual time=4010.165..4010.165 rows=0 loops=1) Filter: (ts > ((('now'::text)::date - 7))::timestamp without time zone) -> Subquery Scan "*SELECT* 2" (cost=0.00..28646.17 rows=714082 width=20) (actual time=44.827..44.827 rows=0 loops=1) -> Index Scan using samples_2000_bytime on samples_2000 (cost=0.00..21505.35 rows=714082 width=20) (actual time=44.818..44.818 rows=0 loops=1) Index Cond: (ts > ((('now'::text)::date - 7))::timestamp without time zone) -> Subquery Scan "*SELECT* 3" (cost=0.00..80393.33 rows=902191 width=20) (actual time=34772.377..34772.377 rows=0 loops=1) -> Seq Scan on samples_2001 (cost=0.00..71371.42 rows=902191 width=20) (actual time=34772.366..34772.366 rows=0 loops=1) Filter: (ts > ((('now'::text)::date - 7))::timestamp without time zone) -> Subquery Scan "*SELECT* 4" (cost=0.00..92424.05 rows=1037201 width=20) (actual time=40072.103..40072.103 rows=0 loops=1) -> Seq Scan on samples_2002 (cost=0.00..82052.04 rows=1037201 width=20) (actual time=40072.090..40072.090 rows=0 loops=1) Filter: (ts > ((('now'::text)::date - 7))::timestamp without time zone) -> Subquery Scan "*SELECT* 5" (cost=0.00..42380.58 rows=1049772 width=20) (actual time=49.455..49.455 rows=0 loops=1) -> Index Scan using samples_2003_bytime on samples_2003 (cost=0.00..31882.86 rows=1049772 width=20) (actual time=49.448..49.448 rows=0 loops=1) Index Cond: (ts > ((('now'::text)::date - 7))::timestamp without time zone) -> Subquery Scan "*SELECT* 6" (cost=0.00..58939.96 rows=791991 width=20) (actual time=65.458..1124.363 rows=66389 loops=1) -> Index Scan using samples_2004_bytime on samples_2004 (cost=0.00..51020.05 rows=791991 width=20) (actual time=65.430..750.336 rows=66389 loops=1) Index Cond: (ts > ((('now'::text)::date - 7))::timestamp without time zone) -> Hash (cost=1.09..1.09 rows=9 width=38) (actual time=15.295..15.295 rows=0 loops=1) -> Seq Scan on sensors s (cost=0.00..1.09 rows=9 width=38) (actual time=15.122..15.187 rows=9 loops=1) Total runtime: 82865.119 ms Essentially, what you can see here is that it's doing an index scan on samples_2000, samples_2003, and samples_2004, but a sequential scan on samples_1999, samples_2001, and samples_2002. It's very strange to me that it would make these choices. If I disable sequential scans altogether for this session, the query runs in under 4 seconds. This is a very cool solution for long-term storage, and isn't terribly hard to manage. I actually have other report queries that seem to be making pretty good index selection currently...but I always want more! :) Does anyone have any suggestions as to how to get this to do what I want? Of course, ideally, it would ignore five of the tables altogether. :) -- SPY My girlfriend asked me which one I like better. pub 1024/3CAE01D5 1994/11/03 Dustin Sallings | Key fingerprint = 87 02 57 08 02 D0 DA D6 C8 0F 3E 65 51 98 D8 BE L_______________________ I hope the answer won't upset her. ____________ From pgsql-performance-owner@postgresql.org Fri Oct 1 14:53:39 2004 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 30092329C83 for ; Fri, 1 Oct 2004 14:53:38 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 97785-09 for ; Fri, 1 Oct 2004 13:53:30 +0000 (GMT) Received: from server07.icaen.uiowa.edu (server07.icaen.uiowa.edu [128.255.17.47]) by svr1.postgresql.org (Postfix) with ESMTP id 9053A32A2E8 for ; Fri, 1 Oct 2004 14:53:30 +0100 (BST) Received: from server11.icaen.uiowa.edu (server11.icaen.uiowa.edu [128.255.17.51]) by server07.icaen.uiowa.edu (8.12.9/8.12.9) with ESMTP id i91DrSSZ002347; sent by ; Fri, 1 Oct 2004 08:53:28 -0500 (CDT) Received: from [192.168.1.11] (12-217-241-0.client.mchsi.com [12.217.241.0]) by server11.icaen.uiowa.edu (8.12.9/smtp-service-1.6) with ESMTP id i91DrLcl011436; (envelope-from ) Fri, 1 Oct 2004 08:53:27 -0500 (CDT) Message-ID: <415D614D.7010001@johnmeinel.com> Date: Fri, 01 Oct 2004 08:53:17 -0500 From: John Meinel User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Dustin Sallings Cc: pgsql-performance@postgresql.org Subject: Re: inconsistent/weird index usage References: <6FA9AB99-1373-11D9-A87D-000A957659CC@spy.net> In-Reply-To: <6FA9AB99-1373-11D9-A87D-000A957659CC@spy.net> X-Enigmail-Version: 0.86.0.0 X-Enigmail-Supports: pgp-inline, pgp-mime Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig6BFBC898713A4E75C9A401D3" X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.66n X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.75 on clamav.icaen.uiowa.edu X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/3 X-Sequence-Number: 8479 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig6BFBC898713A4E75C9A401D3 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Dustin Sallings wrote: > [...] > OK, now that that's over with, I have this one particular query that > I attempt to run for a report from my phone that no longer works because > it tries to do a table scan on *some* of the tables. Why it chooses > this table scan, I can't imagine. The query is as follows: > > select > s.serial as serial_num, > s.name as name, > date(ts) as day, > min(sample) as min_temp, > avg(sample) as avg_temp, > stddev(sample) as stddev_temp, > max(sample) as max_temp > from > samples inner join sensors s using (sensor_id) > where > ts > current_date - 7 > group by > serial_num, name, day > order by > serial_num, day desc > > [ next section heavily clipped for clarity ] -> Seq Scan on samples_1999 (cost rows=103677) (actual rows=0 loops=1) -> Index Scan using samples_2000_bytime on samples_2000 (cost rows=714082 (actual rows=0 loops=1) -> Seq Scan on samples_2001 (cost rows=902191) (actual rows=0 loops=1) -> Seq Scan on samples_2002 (cost rows=1037201) (actual rows=0 loops=1) -> Index Scan using samples_2003_bytime on samples_2003 (cost rows=1049772) (actual rows=0 loops=1) -> Index Scan using samples_2004_bytime on samples_2004 (cost rows=791991) (actual rows=66389 loops=1) [...] > > > Essentially, what you can see here is that it's doing an index scan > on samples_2000, samples_2003, and samples_2004, but a sequential scan > on samples_1999, samples_2001, and samples_2002. It's very strange to > me that it would make these choices. If I disable sequential scans > altogether for this session, the query runs in under 4 seconds. > > This is a very cool solution for long-term storage, and isn't > terribly hard to manage. I actually have other report queries that seem > to be making pretty good index selection currently...but I always want > more! :) Does anyone have any suggestions as to how to get this to do > what I want? > > Of course, ideally, it would ignore five of the tables altogether. :) > > -- > SPY My girlfriend asked me which one I like better. > pub 1024/3CAE01D5 1994/11/03 Dustin Sallings > | Key fingerprint = 87 02 57 08 02 D0 DA D6 C8 0F 3E 65 51 98 D8 BE > L_______________________ I hope the answer won't upset her. ____________ > > Just as a heads up. You have run vacuum analyze before running this query, correct? Because you'll notice that the query planner is thinking that it will have 103677 rows from 1999, 700,000 rows from 2000, 900,000 rows from 2001, etc, etc. Obviously the query planner is not planning well considering it there are only 60,000 rows from 2004, and no rows from anything else. It just seems like it hasn't updated it's statistics to be aware of when the time is on most of the tables. (By the way, an indexed scan returning 0 entries is *really* fast, so I wouldn't worry about ignoring the extra tables. :) I suppose the other question is whether this is a prepared or stored query. Because sometimes the query planner cannot do enough optimization in a stored query. (I ran into this problem where I had 1 column with 500,000+ entries referencing 1 number. If I ran manually, the time was much better because I wasn't using *that* number. With a stored query, it had to take into account that I *might* use that number, and didn't want to do 500,000+ indexed lookups) The only other thing I can think of is that there might be some collision between datetime and date. Like it is thinking it is looking at the time of day when it plans the queries (hence why so many rows), but really it is looking at the date. Perhaps a cast is in order to make it work right. I don't really know. Interesting problem, though. John =:-> --------------enig6BFBC898713A4E75C9A401D3 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFBXWFSJdeBCYSNAAMRAqGCAJ9qpnf3Y6wqz5TVhdcytGaNzk0hdQCgkXWD GiG7bUscL8CcI4zP/qjM7zw= =1TiM -----END PGP SIGNATURE----- --------------enig6BFBC898713A4E75C9A401D3-- From pgsql-performance-owner@postgresql.org Fri Oct 1 15:38:52 2004 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 811B732A26F for ; Fri, 1 Oct 2004 15:38:51 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 15030-08 for ; Fri, 1 Oct 2004 14:38:43 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 2314832A234 for ; Fri, 1 Oct 2004 15:38:46 +0100 (BST) 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 i91Eck99009346; Fri, 1 Oct 2004 10:38:46 -0400 (EDT) To: Dustin Sallings Cc: pgsql-performance@postgresql.org Subject: Re: inconsistent/weird index usage In-reply-to: <6FA9AB99-1373-11D9-A87D-000A957659CC@spy.net> References: <6FA9AB99-1373-11D9-A87D-000A957659CC@spy.net> Comments: In-reply-to Dustin Sallings message dated "Thu, 30 Sep 2004 23:30:49 -0700" Date: Fri, 01 Oct 2004 10:38:46 -0400 Message-ID: <9345.1096641526@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/4 X-Sequence-Number: 8480 Dustin Sallings writes: > The following view creates the illusion of the old ``single-table'' > model: > create view samples as > select * from samples_1999 > union select * from samples_2000 > union select * from samples_2001 > union select * from samples_2002 > union select * from samples_2003 > union select * from samples_2004 You really, really, really want to use UNION ALL not UNION here. > OK, now that that's over with, I have this one particular query that I > attempt to run for a report from my phone that no longer works because > it tries to do a table scan on *some* of the tables. Why it chooses > this table scan, I can't imagine. Most of the problem here comes from the fact that "current_date - 7" isn't reducible to a constant and so the planner is making bad guesses about how much of each table will be scanned. If possible, do the date arithmetic on the client side and send over a simple literal constant. If that's not practical you can fake it with a mislabeled IMMUTABLE function --- see the list archives for previous discussions of the same issue. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 1 15:43:20 2004 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 AD696329E5D for ; Fri, 1 Oct 2004 15:43:18 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 20067-04 for ; Fri, 1 Oct 2004 14:43:10 +0000 (GMT) Received: from anchor-post-32.mail.demon.net (anchor-post-32.mail.demon.net [194.217.242.90]) by svr1.postgresql.org (Postfix) with ESMTP id C2D8D329C6B for ; Fri, 1 Oct 2004 15:43:12 +0100 (BST) Received: from mwynhau.demon.co.uk ([193.237.186.96] helo=mainbox.archonet.com) by anchor-post-32.mail.demon.net with esmtp (Exim 4.42) id 1CDOck-0000a5-7W; Fri, 01 Oct 2004 14:43:06 +0000 Received: from [192.168.1.17] (client17.archonet.com [192.168.1.17]) by mainbox.archonet.com (Postfix) with ESMTP id A9E4F17988; Fri, 1 Oct 2004 15:43:05 +0100 (BST) Message-ID: <415D6CF9.60909@archonet.com> Date: Fri, 01 Oct 2004 15:43:05 +0100 From: Richard Huxton User-Agent: Mozilla Thunderbird 0.7 (X11/20040615) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Dustin Sallings Cc: pgsql-performance@postgresql.org Subject: Re: inconsistent/weird index usage References: <6FA9AB99-1373-11D9-A87D-000A957659CC@spy.net> In-Reply-To: <6FA9AB99-1373-11D9-A87D-000A957659CC@spy.net> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/5 X-Sequence-Number: 8481 Dustin Sallings wrote: > The following view creates the illusion of the old ``single-table'' > model: > > create view samples as > select * from samples_1999 > union select * from samples_2000 > union select * from samples_2001 > union select * from samples_2002 > union select * from samples_2003 > union select * from samples_2004 Try this with UNION ALL (you know there won't be any duplicates) and possibly with some limits too: SELECT * FROM samples_1999 WHERE ts BETWEEN '1999-01-01 00:00:00+00' AND '1999-12-31 11:59:59+00' UNION ALL ... > select > s.serial as serial_num, > s.name as name, > date(ts) as day, > min(sample) as min_temp, > avg(sample) as avg_temp, > stddev(sample) as stddev_temp, > max(sample) as max_temp > from > samples inner join sensors s using (sensor_id) > where > ts > current_date - 7 > group by > serial_num, name, day > order by > serial_num, day desc Try restricting the timestamp too WHERE ts BETWEEN (current_date -7) AND current_timestamp Hopefully that will give the planner enough smarts to know it can skip most of the sample_200x tables. -- Richard Huxton Archonet Ltd From pgsql-performance-owner@postgresql.org Fri Oct 1 16:13:29 2004 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 1A89932A259 for ; Fri, 1 Oct 2004 16:13:28 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 56587-05 for ; Fri, 1 Oct 2004 15:13:10 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [66.143.173.58]) by svr1.postgresql.org (Postfix) with ESMTP id 684A7329E79 for ; Fri, 1 Oct 2004 16:13:07 +0100 (BST) Received: by flake.decibel.org (Postfix, from userid 1001) id 580921C8B0; Fri, 1 Oct 2004 15:13:03 +0000 (GMT) Date: Fri, 1 Oct 2004 10:13:03 -0500 From: "Jim C. Nasby" To: Matt Clark Cc: Postgresql Performance Subject: Re: Caching of Queries Message-ID: <20041001151303.GU1297@decibel.org> References: <00e801c4a4b6$5526dd30$8300a8c0@solent> <1096306670.40463.173.camel@jester> <41587867.4090505@ymogen.net> <20040930221107.GP1297@decibel.org> <415CEE8E.8060805@ymogen.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <415CEE8E.8060805@ymogen.net> X-Operating-System: FreeBSD 4.10-RELEASE-p2 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/6 X-Sequence-Number: 8482 On Fri, Oct 01, 2004 at 06:43:42AM +0100, Matt Clark wrote: > > >If you're not using a connection pool of some kind then you might as > >well forget query plan caching, because your connect overhead will swamp > >the planning cost. This does not mean you have to use something like > >pgpool (which makes some rather questionable claims IMO); any decent web > >application language/environment will support connection pooling. > > > > > Hmm, a question of definition - there's a difference between a pool and > a persistent connection. Pretty much all web apps have one connection > per process, which is persistent (i.e. not dropped and remade for each > request), but not shared between processes, therefore not pooled. OK, that'd work too... the point is if you're re-connecting all the time it doesn't really matter what else you do for performance. -- 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 Fri Oct 1 16:47:09 2004 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 288D532A265 for ; Fri, 1 Oct 2004 16:47:08 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49367-05 for ; Fri, 1 Oct 2004 15:47:03 +0000 (GMT) Received: from bayswater1.ymogen.net (host-154-240-27-217.pobox.net.uk [217.27.240.154]) by svr1.postgresql.org (Postfix) with ESMTP id 1846E32A259 for ; Fri, 1 Oct 2004 16:47:01 +0100 (BST) Received: from solent (unknown [213.165.136.10]) by bayswater1.ymogen.net (Postfix) with ESMTP id A49009EB62; Fri, 1 Oct 2004 16:46:59 +0100 (BST) Reply-To: From: "Matt Clark" To: "'Jim C. Nasby'" Cc: "'Postgresql Performance'" Subject: Re: Caching of Queries Date: Fri, 1 Oct 2004 16:46:59 +0100 Organization: Ymogen Ltd Message-ID: <012601c4a7cd$e358d300$8300a8c0@solent> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook, Build 10.0.6626 In-Reply-To: <20041001151303.GU1297@decibel.org> X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 Importance: Normal X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/7 X-Sequence-Number: 8483 > OK, that'd work too... the point is if you're re-connecting=20 > all the time it doesn't really matter what else you do for=20 > performance. Yeah, although there is the chap who was asking questions on the list recently who had some very long-running code on his app servers, so was best off closing the connection because he had far too many postmaster processes just sitting there idle all the time! But you're right, it's a killer usually. M=20 From pgsql-performance-owner@postgresql.org Fri Oct 1 17:30:11 2004 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 4773C32A615 for ; Fri, 1 Oct 2004 17:30:07 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 65788-03 for ; Fri, 1 Oct 2004 16:30:01 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 7555132A60B for ; Fri, 1 Oct 2004 17:30:01 +0100 (BST) 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 6441602; Fri, 01 Oct 2004 09:31:22 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: Tom Lane Subject: Re: inconsistent/weird index usage Date: Fri, 1 Oct 2004 09:29:36 -0700 User-Agent: KMail/1.6.2 Cc: Dustin Sallings , pgsql-performance@postgresql.org References: <6FA9AB99-1373-11D9-A87D-000A957659CC@spy.net> <9345.1096641526@sss.pgh.pa.us> In-Reply-To: <9345.1096641526@sss.pgh.pa.us> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Message-Id: <200410010929.36026.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/8 X-Sequence-Number: 8484 Tom, > Most of the problem here comes from the fact that "current_date - 7" > isn't reducible to a constant and so the planner is making bad guesses > about how much of each table will be scanned. I thought this was fixed in 7.4. No? -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Fri Oct 1 17:34:27 2004 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 199FC32A2BA for ; Fri, 1 Oct 2004 17:34:24 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 68003-02 for ; Fri, 1 Oct 2004 16:34:18 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 43B6432A2D6 for ; Fri, 1 Oct 2004 17:34:17 +0100 (BST) 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 i91GYFqS010776; Fri, 1 Oct 2004 12:34:15 -0400 (EDT) To: Josh Berkus Cc: Dustin Sallings , pgsql-performance@postgresql.org Subject: Re: inconsistent/weird index usage In-reply-to: <200410010929.36026.josh@agliodbs.com> References: <6FA9AB99-1373-11D9-A87D-000A957659CC@spy.net> <9345.1096641526@sss.pgh.pa.us> <200410010929.36026.josh@agliodbs.com> Comments: In-reply-to Josh Berkus message dated "Fri, 01 Oct 2004 09:29:36 -0700" Date: Fri, 01 Oct 2004 12:34:15 -0400 Message-ID: <10775.1096648455@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/9 X-Sequence-Number: 8485 Josh Berkus writes: >> Most of the problem here comes from the fact that "current_date - 7" >> isn't reducible to a constant and so the planner is making bad guesses >> about how much of each table will be scanned. > I thought this was fixed in 7.4. No? No. It's not fixed as of CVS tip either, although there was some talk of doing something in time for 8.0. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 1 18:11:19 2004 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 7A73532A2DF for ; Fri, 1 Oct 2004 18:11:18 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 82106-01 for ; Fri, 1 Oct 2004 17:11:12 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id C7B8532A246 for ; Fri, 1 Oct 2004 18:11:06 +0100 (BST) 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 6441786 for pgsql-performance@postgresql.org; Fri, 01 Oct 2004 10:12:28 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: Postgresql Performance Subject: Re: Caching of Queries Date: Fri, 1 Oct 2004 10:10:40 -0700 User-Agent: KMail/1.6.2 References: <00e801c4a4b6$5526dd30$8300a8c0@solent> <20040930221107.GP1297@decibel.org> <415CEE8E.8060805@ymogen.net> In-Reply-To: <415CEE8E.8060805@ymogen.net> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410011010.40705.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/11 X-Sequence-Number: 8487 People: Transparent "query caching" is the "industry standard" for how these things are handled. However, Postgres' lack of this feature has made me consider other approaches, and I'm starting to wonder if the "standard" query caching -- where a materialized query result, or some reduction thereof, is cached in database memory -- isn't the best way to cache things. I'm going to abbreviate it "SQC" for the rest of this e-mail. Obviously, the draw of SQC is its transparency to developers. With it, the Java/Perl/PHP programmers and the DBA don't have to communicate at all -- you set it up, give it some RAM, and it "just works". As someone who frequently has to consult based on limited knowledge, I can understand the appeal. However, one of the problems with SQC, aside from the ones already mentioned of stale data and/or cache-clearing, is that (at least in applications like MySQL's) it is indiscriminate and caches, at least breifly, unique queries as readily as common ones. Possibly Oracle's implementation is more sophisticated; I've not had an opportunity. The other half of that problem is that an entire query is cached, rather than just the relevant data to uniquely identify the request to the application. This is bad in two respects; one that the entire query needs to be parsed to see if a new query is materially equivalent, and that two materially different queries which could utilize overlapping ranges of the same underlying result set must instead cache their results seperately, eating up yet more memory. To explain what I'm talking about, let me give you a counter-example of another approach. I have a data-warehousing application with a web front-end. The data in the application is quite extensive and complex, and only a summary is presented to the public users -- but that summary is a query involving about 30 lines and 16 joins. This summary information is available in 3 slightly different forms. Further, the client has indicated that an up to 1/2 hour delay in data "freshness" is acceptable. The first step is forcing that "materialized" view of the data into memory. Right now I'm working on a reliable way to do that without using Memcached, which won't install on our Solaris servers. Temporary tables have the annoying property of being per-connection, which doesn't work in a pool of 60 connections. The second step, which I completed first due to the lack of technical obstacles, is to replace all queries against this data with calls to a Set-Returning Function (SRF). This allowed me to re-direct where the data was coming from -- presumably the same thing could be done through RULES, but it would have been considerably harder to implement. The first thing the SRF does is check the criteria passed to it against a set of cached (in a table) criteria with that user's permission level which is < 1/2 hour old. If the same criteria are found, then the SRF is returned a set of row identifiers for the materialized view (MV), and looks up the rows in the MV and returns those to the web client. If no identical set of criteria are found, then the query is run to get a set of identifiers which are then cached, and the SRF returns the queried rows. Once I surmount the problem of storing all the caching information in protected memory, the advantages of this approach over SQC are several: 1) The materialized data is available in 3 different forms; a list, a detail view, and a spreadsheet. Each form as somewhat different columns and different rules about ordering, which would likely confuse an SQC planner. In this implementation, all 3 forms are able to share the same cache. 2) The application is comparing only sets of unambguous criteria rather than long queries which would need to be compared in planner form in order to determine query equivalence. 3) With the identifier sets, we are able to cache other information as well, such as a count of rows, further limiting the number of queries we must run. 4) This approach is ideally suited to the pagination and re-sorting common to a web result set. As only the identifiers are cached, the results can be re-sorted and broken in to pages after the cache read, a fast, all-in-memory operation. In conclusion, what I'm saying is that while forms of transparent query caching (plan, materialized or whatever) may be desirable for other reasons, it's quite possible to acheive a superior level of "query caching" through tight integration with the front-end application. If people are interested in this, I'd love to see some suggestions on ways to force the materialized view into dedicated memory. -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Fri Oct 1 19:26:44 2004 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 D2668329E79 for ; Fri, 1 Oct 2004 19:26:42 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 28516-06 for ; Fri, 1 Oct 2004 18:26:40 +0000 (GMT) Received: from fdd00lnhub.fds.com (external.fds.com [208.15.90.2]) by svr1.postgresql.org (Postfix) with ESMTP id 5CFE8329DAE for ; Fri, 1 Oct 2004 19:26:41 +0100 (BST) MIME-Version: 1.0 X-Mailer: Lotus Notes Release 6.5 September 18, 2003 Subject: Slow update/insert process To: Message-ID: From: Patrick Hatcher Date: Fri, 1 Oct 2004 11:14:08 -0700 X-MIMETrack: Serialize by Router on FDD00LNHUB/FSG/SVR/FDD(Release 6.5.2|June 01, 2004) at 10/01/2004 02:26:09 PM, Serialize complete at 10/01/2004 02:26:09 PM Content-Type: multipart/alternative; boundary="=_alternative 0064C27E88256F20_=" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.8 tagged_above=0.0 required=5.0 tests=HTML_30_40, HTML_MESSAGE, LINES_OF_YELLING X-Spam-Level: X-Archive-Number: 200410/12 X-Sequence-Number: 8488 This is a multipart message in MIME format. --=_alternative 0064C27E88256F20_= Content-Type: text/plain; charset="US-ASCII" Pg: 7.4.5 RH 7.3 8g Ram 200 g drive space RAID0+1 Tables vacuum on a nightly basis The following process below takes 8 hours to run on 90k records and I'm not sure where to being to look for the bottleneck. This isn't the only updating on this database that seems to take a long time to complete. Is there something I should be looking for in my conf settings? TIA Patrick SQL: ---Bring back only selected records to run through the update process. --Without the function the SQL takes < 10secs to return 90,000 records SELECT count(pm.pm_delta_function_amazon(upc.keyp_upc,'amazon')) FROM mdc_upc upc JOIN public.mdc_products prod ON upc.keyf_products = prod.keyp_products JOIN public.mdc_price_post_inc price ON prod.keyp_products = price.keyf_product JOIN public.mdc_attribute_product ap on ap.keyf_products = prod.keyp_products and keyf_attribute=22 WHERE upper(trim(ap.attributevalue)) NOT IN ('ESTEE LAUDER', 'CLINIQUE','ORGINS','PRESCRIPTIVES','LANC?ME','CHANEL','ARAMIS','M.A.C','TAG HEUER') AND keyf_producttype<>222 AND prod.action_publish = 1; Function: CREATE OR REPLACE FUNCTION pm.pm_delta_function_amazon(int4, "varchar") RETURNS bool AS 'DECLARE varkeyf_upc ALIAS FOR $1; varPassword ALIAS FOR $2; varRealMD5 varchar; varDeltaMD5 varchar; varLastTouchDate date; varQuery text; varQuery1 text; varQueryMD5 text; varQueryRecord record; varFuncStatus boolean := false; BEGIN -- Check the password IF varPassword <> \'amazon\' THEN Return false; END IF; -- Get the md5 hash for this product SELECT into varQueryRecord md5(upc.keyp_upc || prod.description || pm.pm_price_post_inc(prod.keyp_products)) AS md5 FROM public.mdc_upc upc JOIN public.mdc_products prod ON upc.keyf_products = prod.keyp_products JOIN public.mdc_price_post_inc price ON price.keyf_product = prod.keyp_products WHERE upc.keyp_upc = varkeyf_upc LIMIT 1 ; IF NOT FOUND THEN RAISE EXCEPTION \'varRealMD5 is NULL. UPC ID is %\', varkeyf_upc; ELSE varRealMD5:=varQueryRecord.md5; END IF; -- Check that the product is in the delta table and return its hash for comparison SELECT into varQueryRecord md5_hash,last_touch_date FROM pm.pm_delta_master_amazon WHERE keyf_upc = varkeyf_upc LIMIT 1; IF NOT FOUND THEN -- ADD and exit INSERT INTO pm.pm_delta_master_amazon (keyf_upc,status,md5_hash,last_touch_date) values (varkeyf_upc,\'add\',varRealMD5,CURRENT_DATE); varFuncStatus:=true; RETURN varFuncStatus; ELSE --Update the record --- If the hash matches then set the record to HOLD IF varRealMD5 = varQueryRecord.md5_hash THEN UPDATE pm.pm_delta_master_amazon SET status= \'hold\', last_touch_date = CURRENT_DATE WHERE keyf_upc = varkeyf_upc AND last_touch_date <> CURRENT_DATE; varFuncStatus:=true; ELSE -- ELSE mark the item as ADD UPDATE pm.pm_delta_master_amazon SET status= \'add\', last_touch_date = CURRENT_DATE WHERE keyf_upc = varkeyf_upc; varFuncStatus:=true; END IF; END IF; RETURN varFuncStatus; END;' LANGUAGE 'plpgsql' IMMUTABLE; TableDef CREATE TABLE pm.pm_delta_master_amazon ( keyf_upc int4 , status varchar(6) , md5_hash varchar(40) , last_touch_date date ) GO CREATE INDEX status_idx ON pm.pm_delta_master_amazon(status) GO CONF -------- # WRITE AHEAD LOG #--------------------------------------------------------------------------- # - Settings - #fsync = true # turns forced synchronization on or off #wal_sync_method = fsync # the default varies across platforms: # fsync, fdatasync, open_sync, or open_datasync wal_buffers = 32 # min 4, 8KB each # - Checkpoints - checkpoint_segments = 50 # in logfile segments, min 1, 16MB each checkpoint_timeout = 600 # 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 Patrick Hatcher Macys.Com --=_alternative 0064C27E88256F20_= Content-Type: text/html; charset="US-ASCII"
Pg: 7.4.5
RH 7.3
8g Ram
200 g drive space
RAID0+1
Tables vacuum on a nightly basis

The following process below takes 8 hours to run on 90k records and I'm not sure where to being to look for the bottleneck.  This isn't the only updating on this database that seems to take a long time to complete.  Is there something I should be looking for in my conf settings?  

TIA
Patrick


SQL:
---Bring back only selected records to run through the update process.
--Without the function the SQL takes < 10secs to return 90,000 records
SELECT count(pm.pm_delta_function_amazon(upc.keyp_upc,'amazon'))
FROM mdc_upc upc
JOIN public.mdc_products prod ON upc.keyf_products = prod.keyp_products
JOIN public.mdc_price_post_inc price ON prod.keyp_products = price.keyf_product
JOIN public.mdc_attribute_product ap on ap.keyf_products = prod.keyp_products and keyf_attribute=22
WHERE
upper(trim(ap.attributevalue)) NOT IN ('ESTEE LAUDER', 'CLINIQUE','ORGINS','PRESCRIPTIVES','LANC?ME','CHANEL','ARAMIS','M.A.C','TAG HEUER')
AND keyf_producttype<>222
AND prod.action_publish = 1;


Function:

CREATE OR REPLACE FUNCTION pm.pm_delta_function_amazon(int4, "varchar")
 RETURNS bool AS
'DECLARE
   varkeyf_upc                ALIAS FOR $1;
   varPassword                ALIAS FOR $2;
   varRealMD5                varchar;
   varDeltaMD5                varchar;
   varLastTouchDate        date;
   varQuery                 text;
   varQuery1                 text;
   varQueryMD5                text;
   varQueryRecord        record;
   varFuncStatus        boolean := false;
   
BEGIN

-- Check the password
 IF varPassword <> \'amazon\' THEN
   Return false;
 END IF;


--  Get the md5 hash for this product
 SELECT into varQueryRecord md5(upc.keyp_upc || prod.description || pm.pm_price_post_inc(prod.keyp_products)) AS md5
   FROM public.mdc_upc upc
   JOIN public.mdc_products prod ON upc.keyf_products = prod.keyp_products
   JOIN public.mdc_price_post_inc price ON price.keyf_product = prod.keyp_products
   WHERE upc.keyp_upc = varkeyf_upc LIMIT 1 ;
 

 IF NOT FOUND THEN
   RAISE EXCEPTION \'varRealMD5 is NULL. UPC ID is %\', varkeyf_upc;
 ELSE
   varRealMD5:=varQueryRecord.md5;
 END IF;

--  Check that the product is in the delta table and return its hash for comparison
 SELECT into varQueryRecord md5_hash,last_touch_date
   FROM pm.pm_delta_master_amazon
   WHERE keyf_upc = varkeyf_upc LIMIT 1;

 IF NOT FOUND THEN
   -- ADD and exit
   INSERT INTO pm.pm_delta_master_amazon (keyf_upc,status,md5_hash,last_touch_date)
   values (varkeyf_upc,\'add\',varRealMD5,CURRENT_DATE);
   varFuncStatus:=true;
   RETURN varFuncStatus;
 ELSE
   --Update the record
     --- If the hash matches then set the record to HOLD
   IF varRealMD5 = varQueryRecord.md5_hash THEN
       UPDATE pm.pm_delta_master_amazon
         SET status= \'hold\',
         last_touch_date =  CURRENT_DATE
       WHERE keyf_upc =  varkeyf_upc AND last_touch_date <> CURRENT_DATE;
       varFuncStatus:=true;
   ELSE
       --  ELSE mark the item as ADD
       UPDATE pm.pm_delta_master_amazon
         SET status= \'add\',
         last_touch_date =  CURRENT_DATE
       WHERE keyf_upc =  varkeyf_upc;
       varFuncStatus:=true;
   END IF;
  END IF;

 RETURN varFuncStatus;
END;'
 LANGUAGE 'plpgsql' IMMUTABLE;


TableDef
CREATE TABLE pm.pm_delta_master_amazon (
    keyf_upc               int4 ,
    status                 varchar(6) ,
    md5_hash               varchar(40) ,
    last_touch_date        date
    )
GO

CREATE INDEX status_idx
    ON pm.pm_delta_master_amazon(status)
GO




CONF
--------
# WRITE AHEAD LOG
#---------------------------------------------------------------------------

# - Settings -

#fsync = true                   # turns forced synchronization on or off
#wal_sync_method = fsync        # the default varies across platforms:
                                # fsync, fdatasync, open_sync, or open_datasync
wal_buffers = 32                # min 4, 8KB each

# - Checkpoints -

checkpoint_segments = 50        # in logfile segments, min 1, 16MB each
checkpoint_timeout = 600        # 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


Patrick Hatcher
Macys.Com
 --=_alternative 0064C27E88256F20_=-- From pgsql-performance-owner@postgresql.org Fri Oct 1 20:49:11 2004 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 A8CC232A20D for ; Fri, 1 Oct 2004 20:49:03 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 26646-08 for ; Fri, 1 Oct 2004 19:49:00 +0000 (GMT) Received: from hotmail.com (bay18-dav5.bay18.hotmail.com [65.54.187.185]) by svr1.postgresql.org (Postfix) with ESMTP id 06F0532A208 for ; Fri, 1 Oct 2004 20:49:01 +0100 (BST) Received: from mail pickup service by hotmail.com with Microsoft SMTPSVC; Fri, 1 Oct 2004 12:49:00 -0700 Received: from 67.81.98.198 by BAY18-DAV5.phx.gbl with DAV; Fri, 01 Oct 2004 19:48:53 +0000 X-Originating-IP: [67.81.98.198] X-Originating-Email: [awerman2@hotmail.com] X-Sender: awerman2@hotmail.com From: "Aaron Werman" To: , "Patrick Hatcher" References: Subject: Re: Slow update/insert process Date: Fri, 1 Oct 2004 15:48:50 -0400 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_004E_01C4A7CE.2553D890" 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 Message-ID: X-OriginalArrivalTime: 01 Oct 2004 19:49:00.0526 (UTC) FILETIME=[B265E0E0:01C4A7EF] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.8 tagged_above=0.0 required=5.0 tests=HTML_30_40, HTML_MESSAGE, LINES_OF_YELLING X-Spam-Level: X-Archive-Number: 200410/13 X-Sequence-Number: 8489 This is a multi-part message in MIME format. ------=_NextPart_000_004E_01C4A7CE.2553D890 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Some quick notes: - Using a side effect of a function to update the database feels bad to me - how long does the SELECT into varQueryRecord md5(upc.keyp.... function take / what does it's explain look like? - There are a lot of non-indexed columns on that delta master table, such a= s keyf_upc.=20 I'm guessing you're doing 90,000 x {a lot of slow scans} - My temptation would be to rewrite the processing to do a pass of updates,= a pass of inserts,=20 and then the SELECT ----- Original Message -----=20 From: Patrick Hatcher=20 To: pgsql-performance@postgresql.org=20 Sent: Friday, October 01, 2004 2:14 PM Subject: [PERFORM] Slow update/insert process Pg: 7.4.5=20 RH 7.3=20 8g Ram=20 200 g drive space=20 RAID0+1=20 Tables vacuum on a nightly basis=20 The following process below takes 8 hours to run on 90k records and I'm n= ot sure where to being to look for the bottleneck. This isn't the only upd= ating on this database that seems to take a long time to complete. Is ther= e something I should be looking for in my conf settings?=20=20=20 TIA=20 Patrick=20 SQL:=20 ---Bring back only selected records to run through the update process.=20 --Without the function the SQL takes < 10secs to return 90,000 records=20 SELECT count(pm.pm_delta_function_amazon(upc.keyp_upc,'amazon'))=20 FROM mdc_upc upc=20 JOIN public.mdc_products prod ON upc.keyf_products =3D prod.keyp_products= =20 JOIN public.mdc_price_post_inc price ON prod.keyp_products =3D price.keyf= _product=20 JOIN public.mdc_attribute_product ap on ap.keyf_products =3D prod.keyp_pr= oducts and keyf_attribute=3D22=20 WHERE=20 upper(trim(ap.attributevalue)) NOT IN ('ESTEE LAUDER', 'CLINIQUE','ORGINS= ','PRESCRIPTIVES','LANC?ME','CHANEL','ARAMIS','M.A.C','TAG HEUER')=20 AND keyf_producttype<>222=20 AND prod.action_publish =3D 1;=20 Function:=20 CREATE OR REPLACE FUNCTION pm.pm_delta_function_amazon(int4, "varchar") RETURNS bool AS 'DECLARE varkeyf_upc ALIAS FOR $1; varPassword ALIAS FOR $2; varRealMD5 varchar; varDeltaMD5 varchar; varLastTouchDate date; varQuery text; varQuery1 text; varQueryMD5 text; varQueryRecord record; varFuncStatus boolean :=3D false; =20=20=20=20=20 BEGIN -- Check the password IF varPassword <> \'amazon\' THEN Return false; END IF; -- Get the md5 hash for this product SELECT into varQueryRecord md5(upc.keyp_upc || prod.description || pm.pm= _price_post_inc(prod.keyp_products)) AS md5 FROM public.mdc_upc upc JOIN public.mdc_products prod ON upc.keyf_products =3D prod.keyp_produ= cts JOIN public.mdc_price_post_inc price ON price.keyf_product =3D prod.ke= yp_products WHERE upc.keyp_upc =3D varkeyf_upc LIMIT 1 ; =20=20=20 IF NOT FOUND THEN RAISE EXCEPTION \'varRealMD5 is NULL. UPC ID is %\', varkeyf_upc; ELSE varRealMD5:=3DvarQueryRecord.md5; END IF; -- Check that the product is in the delta table and return its hash for = comparison=20 SELECT into varQueryRecord md5_hash,last_touch_date=20 FROM pm.pm_delta_master_amazon WHERE keyf_upc =3D varkeyf_upc LIMIT 1; IF NOT FOUND THEN -- ADD and exit INSERT INTO pm.pm_delta_master_amazon (keyf_upc,status,md5_hash,last_t= ouch_date) values (varkeyf_upc,\'add\',varRealMD5,CURRENT_DATE); varFuncStatus:=3Dtrue; RETURN varFuncStatus; ELSE --Update the record=20 --- If the hash matches then set the record to HOLD IF varRealMD5 =3D varQueryRecord.md5_hash THEN UPDATE pm.pm_delta_master_amazon SET status=3D \'hold\', last_touch_date =3D CURRENT_DATE WHERE keyf_upc =3D varkeyf_upc AND last_touch_date <> CURRENT_DAT= E;=20 varFuncStatus:=3Dtrue; ELSE -- ELSE mark the item as ADD UPDATE pm.pm_delta_master_amazon SET status=3D \'add\', last_touch_date =3D CURRENT_DATE WHERE keyf_upc =3D varkeyf_upc; varFuncStatus:=3Dtrue; END IF; END IF; RETURN varFuncStatus; END;' LANGUAGE 'plpgsql' IMMUTABLE; TableDef=20 CREATE TABLE pm.pm_delta_master_amazon (=20 keyf_upc int4 ,=20 status varchar(6) ,=20 md5_hash varchar(40) ,=20 last_touch_date date=20 )=20 GO=20 CREATE INDEX status_idx=20 ON pm.pm_delta_master_amazon(status)=20 GO=20 CONF=20 --------=20 # WRITE AHEAD LOG=20 #------------------------------------------------------------------------= ---=20 # - Settings -=20 #fsync =3D true # turns forced synchronization on or of= f=20 #wal_sync_method =3D fsync # the default varies across platforms:= =20 # fsync, fdatasync, open_sync, or open_da= tasync=20 wal_buffers =3D 32 # min 4, 8KB each=20 # - Checkpoints -=20 checkpoint_segments =3D 50 # in logfile segments, min 1, 16MB each= =20 checkpoint_timeout =3D 600 # range 30-3600, in seconds=20 #checkpoint_warning =3D 30 # 0 is off, in seconds=20 #commit_delay =3D 0 # range 0-100000, in microseconds=20 #commit_siblings =3D 5 # range 1-1000=20 Patrick Hatcher Macys.Com ------=_NextPart_000_004E_01C4A7CE.2553D890 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Some quick notes:
 
- Using a side effect of a function to upd= ate the=20 database feels bad to me
- how long does the  SELECT into varQueryRecord=20 md5(upc.keyp....
  function take / what does it's expl= ain look=20 like?
- There are a lot of non-indexed columns o= n that=20 delta master table, such as keyf_upc.=20
   I'm guessing you'= re doing=20 90,000 x {a lot of slow scans}
- My temptation would be to rewrite the pr= ocessing=20 to do a pass of updates, a pass of inserts,
  and then the SELECT
----- Original Message -----
Fro= m:=20 Patrick= =20 Hatcher
To: pgsql-performance@postgr= esql.org=20
Sent: Friday, October 01, 2004 2:1= 4=20 PM
Subject: [PERFORM] Slow update/ins= ert=20 process


Pg: 7.4.5
RH=20 7.3
8g Ram
200 g drive space
RAID0+1
Tables vacuu= m on a=20 nightly basis

The followi= ng=20 process below takes 8 hours to run on 90k records and I'm not sure where = to=20 being to look for the bottleneck.  This isn't the only updating on t= his=20 database that seems to take a long time to complete.  Is there somet= hing=20 I should be looking for in my conf settings?  

TIA
Patrick


SQL:= =20
---Bring back only selected records = to run=20 through the update process.
-= -Without=20 the function the SQL takes < 10secs to return 90,000 records=20
SELECT=20 count(pm.pm_delta_function_amazon(upc.keyp_upc,'amazon'))
FROM mdc_upc upc
JOIN public.mdc_products prod ON upc.keyf_products =3D=20 prod.keyp_products
JOIN=20 public.mdc_price_post_inc price ON prod.keyp_products =3D=20 price.keyf_product
JOIN=20 public.mdc_attribute_product ap on ap.keyf_products =3D prod.keyp_product= s and=20 keyf_attribute=3D22
WHERE=20
upper(trim(ap.attributevalue)= ) NOT IN=20 ('ESTEE LAUDER',=20 'CLINIQUE','ORGINS','PRESCRIPTIVES','LANC?ME','CHANEL','ARAMIS','M.A.C','= TAG=20 HEUER')
AND=20 keyf_producttype<>222
A= ND=20 prod.action_publish =3D 1;


Function:

C= REATE OR=20 REPLACE FUNCTION pm.pm_delta_function_amazon(int4, "varchar")
 RE= TURNS=20 bool AS
'DECLARE
   varkeyf_upc       &nbs= p;=20        ALIAS FOR $1;
   varPassword &nbs= p;=20              ALIAS FOR $2;
 = =20  varRealMD5              =20  varchar;
   varDeltaMD5         &n= bsp;=20      varchar;
   varLastTouchDate   &nbs= p;=20    date;
   varQuery         &= nbsp;=20       text;
   varQuery1      = =20           text;
   varQueryMD5 &nbs= p;=20              text;
 =20  varQueryRecord        record;
 =20  varFuncStatus        boolean :=3D false;
&nb= sp;=20  
BEGIN

-- Check the password
 IF varPassword <= >=20 \'amazon\' THEN
   Return false;
 END IF;

--=20  Get the md5 hash for this product
 SELECT into varQueryReco= rd=20 md5(upc.keyp_upc || prod.description ||=20 pm.pm_price_post_inc(prod.keyp_products)) AS md5
   FROM=20 public.mdc_upc upc
   JOIN public.mdc_products prod ON=20 upc.keyf_products =3D prod.keyp_products
   JOIN=20 public.mdc_price_post_inc price ON price.keyf_product =3D=20 prod.keyp_products
   WHERE upc.keyp_upc =3D varkeyf_upc LIM= IT 1=20 ;
 

 IF NOT FOUND THEN
   RAISE EXCEPTIO= N=20 \'varRealMD5 is NULL. UPC ID is %\', varkeyf_upc;
 ELSE
 = =20  varRealMD5:=3DvarQueryRecord.md5;
 END IF;

--  = Check=20 that the product is in the delta table and return its hash for comparison= =20
 SELECT into varQueryRecord md5_hash,last_touch_date
 = =20  FROM pm.pm_delta_master_amazon
   WHERE keyf_upc =3D= =20 varkeyf_upc LIMIT 1;

 IF NOT FOUND THEN
   -- AD= D and=20 exit
   INSERT INTO pm.pm_delta_master_amazon=20 (keyf_upc,status,md5_hash,last_touch_date)
   values=20 (varkeyf_upc,\'add\',varRealMD5,CURRENT_DATE);
 =20  varFuncStatus:=3Dtrue;
   RETURN=20 varFuncStatus;
 ELSE
   --Update the record
&nbs= p;=20    --- If the hash matches then set the record to HOLD
 = ;=20  IF varRealMD5 =3D varQueryRecord.md5_hash THEN
    &nb= sp;=20  UPDATE pm.pm_delta_master_amazon
       =20  SET status=3D \'hold\',
       =20  last_touch_date =3D  CURRENT_DATE
      &nbs= p;WHERE=20 keyf_upc =3D  varkeyf_upc AND last_touch_date <> CURRENT_DATE;= =20
       varFuncStatus:=3Dtrue;
 =20  ELSE
       --  ELSE mark the item as= =20 ADD
       UPDATE pm.pm_delta_master_amazon
&nb= sp;=20        SET status=3D \'add\',
     =  =20  last_touch_date =3D  CURRENT_DATE
      &nbs= p;WHERE=20 keyf_upc =3D  varkeyf_upc;
     =20  varFuncStatus:=3Dtrue;
   END IF;
  END=20 IF;

 RETURN varFuncStatus;
END;'
 LANGUAGE 'plpgsq= l'=20 IMMUTABLE;



TableDe= f=20
CREATE TABLE pm.pm_delta_master_amaz= on (=20
    keyf_upc  =  =20           int4 ,
    status             &= nbsp;=20   varchar(6) ,
  &n= bsp;=20 md5_hash               varchar(40) ,=20
    last_touch_date  =  =20    date
  &nbs= p;=20 )
GO

CREATE INDEX status_idx
    ON=20 pm.pm_delta_master_amazon(status)
GO



CONF
--------
# WRITE AHE= AD=20 LOG
#---------------------------------------------------------------= ------------=20

# - Settings -

#fsync =3D true         &n= bsp;  =20       # turns forced synchronization on or off
= #wal_sync_method =3D fsync      = ;  #=20 the default varies across platforms:
                  &= nbsp;=20             # fsync, fdatasync, open_sync, = or=20 open_datasync
wal_buffers =3D= 32  =20              # min 4, 8KB each= =20

# - Checkpoints -
checkpoint_segments =3D 50      = ;  #=20 in logfile segments, min 1, 16MB each
checkpoint_timeout =3D 600        # range 30= -3600, in=20 seconds
#checkpoint_warning = =3D 30=20        # 0 is off, in seconds
#commit_delay =3D 0        = ;  =20     # range 0-100000, in microseconds
#commit_siblings =3D 5       &n= bsp;  =20  # range 1-1000


= Patrick=20 Hatcher
Macys.Com
 ------=_NextPart_000_004E_01C4A7CE.2553D890-- From pgsql-performance-owner@postgresql.org Sat Oct 2 23:02:42 2004 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 D95EF32A154 for ; Fri, 1 Oct 2004 23:07:03 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 76137-02 for ; Fri, 1 Oct 2004 22:06:53 +0000 (GMT) Received: from shire.ontko.com (shire.ontko.com [199.164.165.1]) by svr1.postgresql.org (Postfix) with ESMTP id 72D0F329D5F for ; Fri, 1 Oct 2004 23:06:54 +0100 (BST) Received: from cormallen.doxpop.com (ryanv@cormallen.doxpop.com [199.164.165.137]) by shire.ontko.com (8.12.3/8.12.3/Debian-7.1) with ESMTP id i91M6skX009733 for ; Fri, 1 Oct 2004 17:06:54 -0500 From: Ryan VanMiddlesworth To: pgsql-performance@postgresql.org Subject: Query planner problem Date: Fri, 1 Oct 2004 17:06:54 -0500 User-Agent: KMail/1.7 MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200410011706.54231.ryan@vanmiddlesworth.org> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/17 X-Sequence-Number: 8493 Okay, I've got two queries that I think the planner should reduce to be logically equivalent but it's not happening. The example queries below have been simplified as much as possible while still producing the problem. What I'm trying to do is create a single prepared statement that can handle null parameters rather than have to dynamically generate the statement in my app code based on supplied parameters. Basically the date constants below would be substituted with parameters supplied on a web search form (or nulls). Here is the query and EXPLAIN that runs quickly: SELECT case_id FROM case_data WHERE case_filed_date > '2004-09-16' AND case_filed_date < '2004-09-20' QUERY PLAN ------------------------------------------------------------- Index Scan using case_data_case_filed_date on case_data (cost=0.00..13790.52 rows=3614 width=18) Index Cond: ((case_filed_date > '2004-09-16'::date) AND (case_filed_date < '2004-09-20'::date)) And here is the query and EXPLAIN from the version that I believe the planner should reduce to be logically equivalent: SELECT case_id FROM case_data WHERE (('2004-09-16' IS NULL) OR (case_filed_date > '2004-09-16')) AND (('2004-09-20' IS NULL) OR (case_filed_date < '2004-09-20')) QUERY PLAN ------------------------------------------------------------- Seq Scan on case_data (cost=0.00..107422.02 rows=27509 width=18) Filter: ((('2004-09-16' IS NULL) OR (case_filed_date > '2004-09-16'::date)) AND (('2004-09-20' IS NULL) OR (case_filed_date < '2004-09-20'::date))) I was hoping that the null comparisons would get folded out by the planner relatively cheaply. But as you can see, the first query uses indexes and the second one uses sequence scans, thereby taking much longer. I guess my question is - is there a better way to accomplish what I'm doing in SQL or am I going to have to dynamically generate the statement based on supplied parameters? Thanks, Ryan From pgsql-performance-owner@postgresql.org Sat Oct 2 04:15:33 2004 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 710E432A514 for ; Sat, 2 Oct 2004 04:15:24 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 35761-04 for ; Sat, 2 Oct 2004 03:15:17 +0000 (GMT) Received: from hotmail.com (bay18-dav14.bay18.hotmail.com [65.54.187.194]) by svr1.postgresql.org (Postfix) with ESMTP id 6E0CA32A509 for ; Sat, 2 Oct 2004 04:15:16 +0100 (BST) Received: from mail pickup service by hotmail.com with Microsoft SMTPSVC; Fri, 1 Oct 2004 20:14:00 -0700 Received: from 67.81.98.198 by bay18-dav14.bay18.hotmail.com with DAV; Sat, 02 Oct 2004 03:13:28 +0000 X-Originating-IP: [67.81.98.198] X-Originating-Email: [awerman2@hotmail.com] X-Sender: awerman2@hotmail.com From: "Aaron Werman" To: "Josh Berkus" , "Postgresql Performance" References: <00e801c4a4b6$5526dd30$8300a8c0@solent> <20040930221107.GP1297@decibel.org> <415CEE8E.8060805@ymogen.net> <200410011010.40705.josh@agliodbs.com> Subject: Re: Caching of Queries Date: Fri, 1 Oct 2004 23:13:28 -0400 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 8bit 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 Message-ID: X-OriginalArrivalTime: 02 Oct 2004 03:14:00.0519 (UTC) FILETIME=[DCD57D70:01C4A82D] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/14 X-Sequence-Number: 8490 I'm not sure I understand your req fully. If the same request is repeatedly done with same parameters, you could implement a proxy web server with a croned script to purge stale pages. If there is substantially the same data being summarized, doing your own summary tables works; if accessed enough, they're in memory. I interleaved some notes into your posting. ----- Original Message ----- From: "Josh Berkus" To: "Postgresql Performance" Sent: Friday, October 01, 2004 1:10 PM Subject: Re: [PERFORM] Caching of Queries > People: > > Transparent "query caching" is the "industry standard" for how these things > are handled. However, Postgres' lack of this feature has made me consider > other approaches, and I'm starting to wonder if the "standard" query caching > -- where a materialized query result, or some reduction thereof, is cached in > database memory -- isn't the best way to cache things. I'm going to > abbreviate it "SQC" for the rest of this e-mail. > > Obviously, the draw of SQC is its transparency to developers. With it, the > Java/Perl/PHP programmers and the DBA don't have to communicate at all -- you > set it up, give it some RAM, and it "just works". As someone who frequently > has to consult based on limited knowledge, I can understand the appeal. My sense is that pg is currently unique among popular dbmses in having the majority of applications being homegrown (a chicken / egg / advocacy issue - if I install a CMS, I'm not the DBA or the PHP programmer - and I don't want to change the code; we'll see more about this when native WinPg happens). > > However, one of the problems with SQC, aside from the ones already mentioned > of stale data and/or cache-clearing, is that (at least in applications like > MySQL's) it is indiscriminate and caches, at least breifly, unique queries as > readily as common ones. Possibly Oracle's implementation is more > sophisticated; I've not had an opportunity. I'm not sure I agree here. Stale data and caching choice are optimizer/buffer manager choices and implementation can decide whether to allow stale data. These are design choices involving development effort and choices of where to spend server cycles and memory. All buffering choices cache unique objects, I'm not sure why this is bad (but sensing you want control of the choices). FWIW, this is my impression of other dbmses. In MySQL, a global cache can be specified with size and globally, locally, or through statement hints in queries to suggest caching results. I don't believe that these could be used as common subexpressions (with an exception of MERGE table component results). The optimizer knows nothing about the cached results - SQL select statements are hashed, and can be replaced by the the cached statement/results on a match. In DB2 and Oracle result sets are not cached. They have rich sets of materialized view features (that match your requirements). They allow a materialized view to be synchronous with table updates or asynchronous. Synchronous is often an unrealistic option, and asynchronous materialized views are refreshed at a specified schedule. The optimizers allow "query rewrite" (in Oracle it is a session option) so one can connect to the database and specify that the optimizer is allowed to replace subexpressions with data from (possibly stale) materialized views. SQL Server 2K has more restrictive synchronous MVs, but I've never used them. So, in your example use in Oracle, you would need to define appropriate MVs with a � hour refresh frequency, and hope that the planner would use them in your queries. The only change in the app is on connection you would allow use of asynchronous stale data. You're suggesting an alternative involving identifying common, but expensive, subexpressions and generating MVs for them. This is a pretty sophisticated undertaking, and probably requires some theory research to determine if it's viable. > > The other half of that problem is that an entire query is cached, rather than > just the relevant data to uniquely identify the request to the application. > This is bad in two respects; one that the entire query needs to be parsed to > see if a new query is materially equivalent, and that two materially > different queries which could utilize overlapping ranges of the same > underlying result set must instead cache their results separately, eating up > yet more memory. There are two separate issues. The cost of parse/optimization and the cost of results retrieval. Other dbmses hash statement text. This is a good thing, and probably 3 orders of magnitude faster than parse and optimization. (Oracle also has options to replace literals with parameters and match parse trees instead of text, expecting parse costs to be less than planning costs.) MySQL on a match simply returns the result set. Oracle and DB2 attempt to rewrite queries to use the DBA selected extracts. The MySQL approach seems to be almost what you're describing: all it needs is the statement hash, statement, and result set. The rest of your wish list, identifying and caching data to satisfy multiple request is what query rewrite does - as long as you've created the appropriate MV. > > To explain what I'm talking about, let me give you a counter-example of > another approach. > > I have a data-warehousing application with a web front-end. The data in the > application is quite extensive and complex, and only a summary is presented > to the public users -- but that summary is a query involving about 30 lines > and 16 joins. This summary information is available in 3 slightly different > forms. Further, the client has indicated that an up to 1/2 hour delay in > data "freshness" is acceptable. This sounds like a requirement for a summary table - if the data can be summarized appropriately, and a regular refresh process. > > The first step is forcing that "materialized" view of the data into memory. > Right now I'm working on a reliable way to do that without using Memcached, > which won't install on our Solaris servers. Temporary tables have the > annoying property of being per-connection, which doesn't work in a pool of 60 > connections. I'm not clear on your desire to keep the data in memory. If it is because of I/O cost of the summary table, database buffers should be caching it. If you want to store calculated results, again - why not use a summary table? The con of summary tables is the customization / denormalization of the data, and the need to have programs use them instead of source data - you seem to be willing to do each of these things. > > The second step, which I completed first due to the lack of technical > obstacles, is to replace all queries against this data with calls to a > Set-Returning Function (SRF). This allowed me to re-direct where the data > was coming from -- presumably the same thing could be done through RULES, but > it would have been considerably harder to implement. > > The first thing the SRF does is check the criteria passed to it against a set > of cached (in a table) criteria with that user's permission level which is < > 1/2 hour old. If the same criteria are found, then the SRF is returned a > set of row identifiers for the materialized view (MV), and looks up the rows > in the MV and returns those to the web client. > > If no identical set of criteria are found, then the query is run to get a set > of identifiers which are then cached, and the SRF returns the queried rows. > > Once I surmount the problem of storing all the caching information in > protected memory, the advantages of this approach over SQC are several: You are creating summary data on demand. I have had problems with this approach, mostly because it tends to cost more than doing it in batch and adds latency (unfortunately adding to peak load - so I tend to prefer periodic extract/summarize programs). In either approach why don't you want pg to cache the data? The result also feels more like persisted object data than typical rdbms processing. > > 1) The materialized data is available in 3 different forms; a list, a detail > view, and a spreadsheet. Each form as somewhat different columns and > different rules about ordering, which would likely confuse an SQC planner. > In this implementation, all 3 forms are able to share the same cache. I'm not clear what the issue here is. Are you summarizing data differently or using some business rules to identify orthogonal queries? > > 2) The application is comparing only sets of unambiguous criteria rather than > long queries which would need to be compared in planner form in order to > determine query equivalence. > > 3) With the identifier sets, we are able to cache other information as well, > such as a count of rows, further limiting the number of queries we must run. > > 4) This approach is ideally suited to the pagination and re-sorting common to > a web result set. As only the identifiers are cached, the results can be > re-sorted and broken in to pages after the cache read, a fast, all-in-memory > operation. > > In conclusion, what I'm saying is that while forms of transparent query > caching (plan, materialized or whatever) may be desirable for other reasons, > it's quite possible to achieve a superior level of "query caching" through > tight integration with the front-end application. This looks like you're building an object store to support a custom app that periodically or on demand pulls rdbms data mart data. The description of the use seems either static, suggesting summary tables or dynamic, suggesting that you're mimicking some function of a periodically extracted OLAP cube. > > If people are interested in this, I'd love to see some suggestions on ways to > force the materialized view into dedicated memory. Can you identify your objections to summarizing the data and letting pg buffer it? /Aaron From pgsql-performance-owner@postgresql.org Sat Oct 2 21:02:55 2004 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 062E9329EDF for ; Sat, 2 Oct 2004 21:02:46 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 64223-01 for ; Sat, 2 Oct 2004 20:02:41 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id CFAF4329EDA for ; Sat, 2 Oct 2004 21:02:40 +0100 (BST) 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 6448983; Sat, 02 Oct 2004 13:04:03 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: "Aaron Werman" Subject: Re: Caching of Queries Date: Sat, 2 Oct 2004 13:04:21 -0700 User-Agent: KMail/1.6.2 Cc: "Postgresql Performance" References: <00e801c4a4b6$5526dd30$8300a8c0@solent> <200410011010.40705.josh@agliodbs.com> In-Reply-To: MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410021304.21281.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/15 X-Sequence-Number: 8491 Aaron, > I'm not sure I understand your req fully. I'm not surprised. I got wrapped up in an overly involved example and completely left off the points I was illustrating. So here's the points, in brief: 1) Query caching is not a single problem, but rather several different problems requiring several different solutions. 2) Of these several different solutions, any particular query result caching implementation (but particularly MySQL's) is rather limited in its applicability, partly due to the tradeoffs required. Per your explanation, Oracle has improved this by offering a number of configurable options. 3) Certain other caching problems would be solved in part by the ability to construct "in-memory" tables which would be non-durable and protected from cache-flushing. This is what I'm interested in chatting about. BTW, I AM using a summary table. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Sat Oct 2 22:08:07 2004 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 A6CD2329E52 for ; Sat, 2 Oct 2004 22:08:02 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 76250-03 for ; Sat, 2 Oct 2004 21:07:57 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id DFCFC329DCD for ; Sat, 2 Oct 2004 22:07:59 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i92L7rbx077650 for ; Sat, 2 Oct 2004 21:07:53 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i92KjUV9072856 for pgsql-performance@postgresql.org; Sat, 2 Oct 2004 20:45:30 GMT From: William Yu X-Newsgroups: comp.databases.postgresql.performance Subject: Re: Caching of Queries Date: Sat, 02 Oct 2004 13:47:26 -0700 Organization: Hub.Org Networking Services Lines: 23 Message-ID: References: <00e801c4a4b6$5526dd30$8300a8c0@solent> <200410011010.40705.josh@agliodbs.com> <200410021304.21281.josh@agliodbs.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit X-Complaints-To: usenet@news.hub.org User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.3) Gecko/20040910 X-Accept-Language: en-us, en In-Reply-To: <200410021304.21281.josh@agliodbs.com> To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/16 X-Sequence-Number: 8492 Josh Berkus wrote: > 1) Query caching is not a single problem, but rather several different > problems requiring several different solutions. > > 2) Of these several different solutions, any particular query result caching > implementation (but particularly MySQL's) is rather limited in its > applicability, partly due to the tradeoffs required. Per your explanation, > Oracle has improved this by offering a number of configurable options. > > 3) Certain other caching problems would be solved in part by the ability to > construct "in-memory" tables which would be non-durable and protected from > cache-flushing. This is what I'm interested in chatting about. Just my 2 cents on this whole issue. I would lean towards having result caching in pgpool versus the main backend. I want every ounce of memory on a database server devoted to the database. Caching results would double the effect of cache flushing ... ie, now both the results and the pages used to build the results are in memory pushing out other stuff to disk that may be just as important. If it was in pgpool or something similar, I could devote a separate machine just for caching results leaving the db server untouched. From pgsql-performance-owner@postgresql.org Sat Oct 2 23:52:52 2004 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 A52D9329D58 for ; Sat, 2 Oct 2004 23:52:51 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 97466-04 for ; Sat, 2 Oct 2004 22:52:45 +0000 (GMT) Received: from mail.pws.com.au (mail.pws.com.au [210.23.138.139]) by svr1.postgresql.org (Postfix) with SMTP id 6DFC7329D4D for ; Sat, 2 Oct 2004 23:52:46 +0100 (BST) Received: (qmail 4249 invoked by uid 0); 3 Oct 2004 08:52:42 +1000 Received: from cpe-203-45-44-212.vic.bigpond.net.au (HELO wizzard.pws.com.au) (russell@pws.com.au@203.45.44.212) by mail.pws.com.au with SMTP; 3 Oct 2004 08:52:42 +1000 From: Russell Smith To: Ryan VanMiddlesworth Subject: Re: Query planner problem Date: Sun, 3 Oct 2004 08:50:28 +1000 User-Agent: KMail/1.7 Cc: pgsql-performance@postgresql.org References: <200410011706.54231.ryan@vanmiddlesworth.org> In-Reply-To: <200410011706.54231.ryan@vanmiddlesworth.org> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200410030850.28345.mr-russ@pws.com.au> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/19 X-Sequence-Number: 8495 On Sat, 2 Oct 2004 08:06 am, Ryan VanMiddlesworth wrote: [snip] > > > Here is the query and EXPLAIN that runs quickly: > SELECT case_id FROM case_data > WHERE case_filed_date > '2004-09-16' > AND case_filed_date < '2004-09-20' > > QUERY PLAN > ------------------------------------------------------------- > Index Scan using case_data_case_filed_date on case_data > (cost=0.00..13790.52 rows=3614 width=18) > Index Cond: ((case_filed_date > '2004-09-16'::date) > AND (case_filed_date < '2004-09-20'::date)) > > > And here is the query and EXPLAIN from the version that I believe the planner > should reduce to be logically equivalent: > SELECT case_id FROM case_data > WHERE (('2004-09-16' IS NULL) OR (case_filed_date > '2004-09-16')) > AND (('2004-09-20' IS NULL) OR (case_filed_date < '2004-09-20')) > > QUERY PLAN > ------------------------------------------------------------- > Seq Scan on case_data (cost=0.00..107422.02 rows=27509 width=18) > Filter: ((('2004-09-16' IS NULL) OR (case_filed_date > '2004-09-16'::date)) > AND (('2004-09-20' IS NULL) OR (case_filed_date < '2004-09-20'::date))) > > > I was hoping that the null comparisons would get folded out by the planner > relatively cheaply. But as you can see, the first query uses indexes and the > second one uses sequence scans, thereby taking much longer. I guess my > question is - is there a better way to accomplish what I'm doing in SQL or am > I going to have to dynamically generate the statement based on supplied > parameters? > The Index does not store NULL values, so you have to do a tables scan to find NULL values. That means the second query cannot use an Index, even if it wanted to. Regards Russell Smith From pgsql-performance-owner@postgresql.org Sat Oct 2 23:49:23 2004 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 632A2329DB1 for ; Sat, 2 Oct 2004 23:49:21 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 98901-02 for ; Sat, 2 Oct 2004 22:49:15 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 341DF329D9F for ; Sat, 2 Oct 2004 23:49:17 +0100 (BST) 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 6449455; Sat, 02 Oct 2004 15:50:40 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: William Yu Subject: Re: Caching of Queries Date: Sat, 2 Oct 2004 15:50:58 -0700 User-Agent: KMail/1.6.2 Cc: pgsql-performance@postgresql.org References: <00e801c4a4b6$5526dd30$8300a8c0@solent> <200410021304.21281.josh@agliodbs.com> In-Reply-To: MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410021550.58484.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/18 X-Sequence-Number: 8494 William, > Just my 2 cents on this whole issue. I would lean towards having result > caching in pgpool versus the main backend. I want every ounce of memory > on a database server devoted to the database. Caching results would > double the effect of cache flushing ... ie, now both the results and the > pages used to build the results are in memory pushing out other stuff to > disk that may be just as important. > > If it was in pgpool or something similar, I could devote a separate > machine just for caching results leaving the db server untouched. Oddly, Joe Conway just mentioned the same idea to me. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Sun Oct 3 00:21:07 2004 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 C5BE6329E53 for ; Sun, 3 Oct 2004 00:21:05 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 04698-03 for ; Sat, 2 Oct 2004 23:20:58 +0000 (GMT) Received: from fuji.krosing.net (217-159-136-226-dsl.kt.estpak.ee [217.159.136.226]) by svr1.postgresql.org (Postfix) with ESMTP id 69665329E7B for ; Sun, 3 Oct 2004 00:20:59 +0100 (BST) Received: from fuji.krosing.net (localhost [127.0.0.1]) by fuji.krosing.net (8.12.11/8.12.11) with ESMTP id i92NKnJ5005745; Sun, 3 Oct 2004 02:20:49 +0300 Received: (from hannu@localhost) by fuji.krosing.net (8.12.11/8.12.11/Submit) id i92NKgYf005744; Sun, 3 Oct 2004 02:20:42 +0300 X-Authentication-Warning: fuji.krosing.net: hannu set sender to hannu@tm.ee using -f Subject: Re: inconsistent/weird index usage From: Hannu Krosing To: Tom Lane Cc: Josh Berkus , Dustin Sallings , pgsql-performance@postgresql.org In-Reply-To: <10775.1096648455@sss.pgh.pa.us> References: <6FA9AB99-1373-11D9-A87D-000A957659CC@spy.net> <9345.1096641526@sss.pgh.pa.us> <200410010929.36026.josh@agliodbs.com> <10775.1096648455@sss.pgh.pa.us> Content-Type: text/plain Content-Transfer-Encoding: 7bit Message-Id: <1096759241.5508.2.camel@fuji.krosing.net> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Date: Sun, 03 Oct 2004 02:20:41 +0300 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/20 X-Sequence-Number: 8496 On R, 2004-10-01 at 19:34, Tom Lane wrote: > Josh Berkus writes: > >> Most of the problem here comes from the fact that "current_date - 7" > >> isn't reducible to a constant and so the planner is making bad guesses > >> about how much of each table will be scanned. > > > I thought this was fixed in 7.4. No? > > No. It's not fixed as of CVS tip either, although there was some talk > of doing something in time for 8.0. That's weird - my 7.4.2 databases did not consider (now()-'15 min'::interval) to be a constant whereas 7.4.5 does (i.e. it does use index scan on index on datetime column) Is this somehow different for date types ? -------------- Hannu From pgsql-performance-owner@postgresql.org Sun Oct 3 00:21:28 2004 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 3C743329E70 for ; Sun, 3 Oct 2004 00:21:24 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 05712-01 for ; Sat, 2 Oct 2004 23:21:16 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 64ECC329E55 for ; Sun, 3 Oct 2004 00:21:18 +0100 (BST) 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 i92NLHvq026955; Sat, 2 Oct 2004 19:21:17 -0400 (EDT) To: Ryan VanMiddlesworth Cc: pgsql-performance@postgresql.org Subject: Re: Query planner problem In-reply-to: <200410011706.54231.ryan@vanmiddlesworth.org> References: <200410011706.54231.ryan@vanmiddlesworth.org> Comments: In-reply-to Ryan VanMiddlesworth message dated "Fri, 01 Oct 2004 17:06:54 -0500" Date: Sat, 02 Oct 2004 19:21:16 -0400 Message-ID: <26954.1096759276@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/21 X-Sequence-Number: 8497 Ryan VanMiddlesworth writes: > And here is the query and EXPLAIN from the version that I believe the planner > should reduce to be logically equivalent: > SELECT case_id FROM case_data > WHERE (('2004-09-16' IS NULL) OR (case_filed_date > '2004-09-16')) > AND (('2004-09-20' IS NULL) OR (case_filed_date < '2004-09-20')) > I was hoping that the null comparisons would get folded out by the planner > relatively cheaply. You could teach eval_const_expressions about simplifying NullTest nodes if you think it's important enough. regards, tom lane From pgsql-performance-owner@postgresql.org Sun Oct 3 00:44:26 2004 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 E645A329E70 for ; Sun, 3 Oct 2004 00:44:25 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 09422-09 for ; Sat, 2 Oct 2004 23:44:19 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 6EC91329E53 for ; Sun, 3 Oct 2004 00:44:21 +0100 (BST) 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 i92NiLi8027190; Sat, 2 Oct 2004 19:44:22 -0400 (EDT) To: Hannu Krosing Cc: Josh Berkus , Dustin Sallings , pgsql-performance@postgresql.org Subject: Re: inconsistent/weird index usage In-reply-to: <1096759241.5508.2.camel@fuji.krosing.net> References: <6FA9AB99-1373-11D9-A87D-000A957659CC@spy.net> <9345.1096641526@sss.pgh.pa.us> <200410010929.36026.josh@agliodbs.com> <10775.1096648455@sss.pgh.pa.us> <1096759241.5508.2.camel@fuji.krosing.net> Comments: In-reply-to Hannu Krosing message dated "Sun, 03 Oct 2004 02:20:41 +0300" Date: Sat, 02 Oct 2004 19:44:21 -0400 Message-ID: <27189.1096760661@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/22 X-Sequence-Number: 8498 Hannu Krosing writes: >> No. It's not fixed as of CVS tip either, although there was some talk >> of doing something in time for 8.0. > That's weird - my 7.4.2 databases did not consider (now()-'15 > min'::interval) to be a constant whereas 7.4.5 does (i.e. it does use > index scan on index on datetime column) The question isn't whether it can use it as an indexscan bound; the question is whether it can derive an accurate rowcount estimate. The issue is exactly that STABLE functions work for one but not the other. regards, tom lane From pgsql-performance-owner@postgresql.org Sun Oct 3 14:18:16 2004 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 7FAC7329DBF for ; Sun, 3 Oct 2004 14:18:11 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 54552-08 for ; Sun, 3 Oct 2004 13:18:03 +0000 (GMT) Received: from boutiquenumerique.com (gailleton-2-82-67-9-10.fbx.proxad.net [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id EB58E329D83 for ; Sun, 3 Oct 2004 14:18:02 +0100 (BST) Received: (qmail 5130 invoked from network); 3 Oct 2004 15:18:25 +0200 Received: from unknown (HELO musicbox) (192.168.0.2) by gailleton-2-82-67-9-10.fbx.proxad.net with SMTP; 3 Oct 2004 15:18:25 +0200 Date: Sun, 03 Oct 2004 15:20:49 +0200 To: "Postgresql Performance" Subject: Re: Caching of Queries References: <00e801c4a4b6$5526dd30$8300a8c0@solent> <1096306670.40463.173.camel@jester> <41587867.4090505@ymogen.net> <20040930221107.GP1297@decibel.org> From: =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= 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: <20040930221107.GP1297@decibel.org> User-Agent: Opera M2/7.53 (Linux, build 737) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/23 X-Sequence-Number: 8499 > pgpool (which makes some rather questionable claims IMO); any decent web > application language/environment will support connection pooling. That's why it should not be tied to something specific as pgpool. If you want performance, which is the case here, usually you have a webserver serving static files, and an application server serving dynamic pages. This is not necessarily a huge application server, it can be as simple as an Apache instance serving static files, with a special path mod_proxy'ed to another instance of apache acting as an application server. IMHO this is a nice way to do it, because you have a light weight static files server which can spawn many processes without using precious resources like memory and postgres connections, and a specialized server which has a lot less processes, each one having more size, a db connection, etc. The connexions are permanent, of course, so there is no connection overhead. The proxy has an extra advantage buffering the data from the "app server" and sending it back slowly to the client, so the app server can then very quickly process the next request instead of hogging a db connection while the html is slowly trickled back to the client. IMHO the standard PHP way of doing things (just one server) is wrong because every server process, even if it's serving static files, hogs a connection and thus needs an extra layer for pooling. Thus, I see query result caching as a way to pushing further architectures which are already optimized for performance, not as a band-aid for poor design solutions like the one-apache server with pooling. Now, a proposition : Here is where we are now, a typical slow query : PREPARE myquery(text,integer) EXECUTE myquery('john',2) My proposition : PREPARE myquery(text,integer) PLANNED USING ('john',2) CACHED IF $1 IS NOT NULL AND $2 IS NOT NULL DEPENDS ON $1, $2 MAXIMUM CACHE TIME '5 minute'::interval MINIMUM CACHE TIME '1 minute'::interval MAXIMUM CACHE SIZE 2000000 AS SELECT count(*) as number FROM mytable WHERE myname=$2 AND myfield>=$1; EXECUTE myquery('john',2) Explainations : ----------- PLANNED USING ('john',2) Tells the planner to compute the stored query plan using the given parameters. This is independent from caching but could be a nice feature as it would avoid the possibility of storing a bad query plan. ----------- CACHED IF $1 IS NOT NULL AND $2 IS NOT NULL Specifies that the result is to be cached. There is an optional condition (here, IF ...) telling postgres of when and where it should cache, or not cache. It could be useful to avoid wasting cache space. ----------- DEPENDS ON $1, $2 Defines the cache key. I don't know if this is useful, as the query parameters make a pretty obvious cache key so why repeat them. It could be used to add other data as a cache key, like : DEPENDS ON (SELECT somefunction($1)) Also a syntax for specifying which tables should be watched for updates, and which should be ignored, could be interesting. ----------- MAXIMUM CACHE TIME '5 minute'::interval Pretty obvious. ----------- MINIMUM CACHE TIME '1 minute'::interval This query is a count and I want a fast but imprecise count. Thus, I specify a minimum cache time of 1 minute, meaning that the result will stay in the cache even if the tables change. This is dangerous, so I'd suggest the following : MINIMUM CACHE TIME CASE WHEN result.number>10 THEN '1 minute'::interval ELSE '5 second'::interval Thus the cache time is an expression ; it is evaluated after performed the query. There needs to be a way to access the 'count' result, which I called 'result.number' because of the SELECT count() as number. The result could also be used in the CACHE IF. The idea here is that the count will vary over time, but we accept some imprecision to gain speed. SWho cares if there are 225 or 227 messages in a forum thread counter anyway ? However, if there are 2 messages, first caching the query is less necessary because it's fast, and second a variation in the count will be much easier to spot, thus we specify a shorter cache duration for small counts and a longer duration for large counts. For queries returning result sets, this is not usable of course, but a special feature for speeding count() queries would be welcome ! ----------- MAXIMUM CACHE SIZE 2000000 Pretty obvious. Size in bytes. For queries returning several rows, MIN/MAX on result rows could be useful also : MAXIMUM RESULT ROWS nnn Or maybe : CACHE IF (select count(*) from result) > nnn Thinking about it, using prepared queries seems a bad idea ; maybe the cache should act on the result of functions. This would force the application programmers to put the queries they want to optimize in functions, but as function code is shared between connections and prepared statements are not, maybe it would be easier to implement, and would shield against some subtle bugs, like PREPARing the different queries under the same name... In that case the cache manager would also know if the function returns SETOF or not, which would be interesting. What do you think of these optimizations ? Right now, a count() query cache could be implemented as a simple plsql function with a table as the cache, by the way. From pgsql-performance-owner@postgresql.org Sun Oct 3 14:25:22 2004 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 3AE82329D41 for ; Sun, 3 Oct 2004 14:25:08 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 57412-01 for ; Sun, 3 Oct 2004 13:24:59 +0000 (GMT) Received: from boutiquenumerique.com (gailleton-2-82-67-9-10.fbx.proxad.net [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 3E141329D3E for ; Sun, 3 Oct 2004 14:24:57 +0100 (BST) Received: (qmail 5333 invoked from network); 3 Oct 2004 15:25:20 +0200 Received: from unknown (HELO musicbox) (192.168.0.2) by gailleton-2-82-67-9-10.fbx.proxad.net with SMTP; 3 Oct 2004 15:25:20 +0200 To: "Postgresql Performance" Subject: Re: Caching of Queries References: <00e801c4a4b6$5526dd30$8300a8c0@solent> <20040930221107.GP1297@decibel.org> <415CEE8E.8060805@ymogen.net> <200410011010.40705.josh@agliodbs.com> Message-ID: Date: Sun, 03 Oct 2004 15:27:46 +0200 From: =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= 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: <200410011010.40705.josh@agliodbs.com> User-Agent: Opera M2/7.53 (Linux, build 737) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/24 X-Sequence-Number: 8500 > 1) The materialized data is available in 3 different forms; a list, a > detail > view, and a spreadsheet. Each form as somewhat different columns and > different rules about ordering, which would likely confuse an SQC > planner. > In this implementation, all 3 forms are able to share the same cache. See my proposal to cache function results. You can create a cached function and : SELECT your rows FROM cached_function(parameters) WHERE ... ORDER BY... GROUP BY... will only fetch the function result from the cache, and then the only additional costs are the ORDER and GROUP BY... the query parsing is very simple, it's just a select, and a "cached function scan" I think caching can be made much more powerful if it is made usable like this. I mean, not only cache a query and its result, but being able to use cached queries internally like this and manipulaing them, adds value to the cached data and allows storing less data in the cache because duplicates are avoided. Thus we could use cached results in CHECK() conditions, inside plsql functions, anywhere... From pgsql-performance-owner@postgresql.org Sun Oct 3 14:27:33 2004 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 DA4B7329F19 for ; Sun, 3 Oct 2004 14:27:26 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 55660-10 for ; Sun, 3 Oct 2004 13:27:17 +0000 (GMT) Received: from boutiquenumerique.com (gailleton-2-82-67-9-10.fbx.proxad.net [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id C1F9F329F17 for ; Sun, 3 Oct 2004 14:27:17 +0100 (BST) Received: (qmail 5407 invoked from network); 3 Oct 2004 15:27:40 +0200 Received: from unknown (HELO musicbox) (192.168.0.2) by gailleton-2-82-67-9-10.fbx.proxad.net with SMTP; 3 Oct 2004 15:27:40 +0200 To: pgsql-performance@postgresql.org Subject: Re: Caching of Queries References: <00e801c4a4b6$5526dd30$8300a8c0@solent> <200410011010.40705.josh@agliodbs.com> <200410021304.21281.josh@agliodbs.com> Message-ID: Date: Sun, 03 Oct 2004 15:30:05 +0200 From: =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= 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: User-Agent: Opera M2/7.53 (Linux, build 737) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/25 X-Sequence-Number: 8501 > If it was in pgpool or something similar, I could devote a separate > machine just for caching results leaving the db server untouched. BUT you would be limited to caching complete queries. There is a more efficient strategy... From pgsql-performance-owner@postgresql.org Sun Oct 3 16:26:35 2004 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 C01C8329CFB for ; Sun, 3 Oct 2004 16:26:34 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 85717-06 for ; Sun, 3 Oct 2004 15:26:33 +0000 (GMT) Received: from sraigw.sra.co.jp (sraigw.sra.co.jp [202.32.10.2]) by svr1.postgresql.org (Postfix) with ESMTP id 5DA60329C70 for ; Sun, 3 Oct 2004 16:26:32 +0100 (BST) Received: from srascb.sra.co.jp (srascb [133.137.8.65]) by sraigw.sra.co.jp (Postfix) with ESMTP id 91615633AD; Mon, 4 Oct 2004 00:26:30 +0900 (JST) Received: from srascb.sra.co.jp (localhost [127.0.0.1]) by localhost.sra.co.jp (Postfix) with ESMTP id 795E410CD06; Mon, 4 Oct 2004 00:26:30 +0900 (JST) Received: from sranhm.sra.co.jp (sranhm [133.137.44.16]) by srascb.sra.co.jp (Postfix) with ESMTP id 5FCA510CD04; Mon, 4 Oct 2004 00:26:30 +0900 (JST) Received: from localhost (sraihb-hub.sra.co.jp [133.137.8.6]) by sranhm.sra.co.jp (8.9.3+3.2W/3.7W-srambox) with ESMTP id AAA15518; Mon, 4 Oct 2004 00:26:29 +0900 Date: Mon, 04 Oct 2004 00:28:23 +0900 (JST) Message-Id: <20041004.002823.85416338.t-ishii@sra.co.jp> To: matt@ymogen.net Cc: pg@rbt.ca, awerman2@hotmail.com, scottakirkwood@gmail.com, pgsql-performance@postgresql.org Subject: Re: Caching of Queries From: Tatsuo Ishii In-Reply-To: <415887AE.6070909@ymogen.net> References: <41587867.4090505@ymogen.net> <1096317453.40463.226.camel@jester> <415887AE.6070909@ymogen.net> X-Mailer: Mew version 2.3 on Emacs 20.7 / Mule 4.1 =?iso-2022-jp?B?KBskQjAqGyhCKQ==?= 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/26 X-Sequence-Number: 8502 > >>More to the point though, I think this is a feature that really really > >>should be in the DB, because then it's trivial for people to use. > >> > >> > > > >How does putting it into PGPool make it any less trivial for people to > >use? > > > The answers are at http://www2b.biglobe.ne.jp/~caco/pgpool/index-e.html > . Specifically, it's a separate application that needs configuration, > the homepage has no real discussion of the potential pitfalls of pooling > and what this implementation does to get around them, you get the idea. I don't know what you are exactly referring to in above URL when you are talking about "potential pitfalls of pooling". Please explain more. -- Tatsuo Ishii From pgsql-performance-owner@postgresql.org Sun Oct 3 16:27:10 2004 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 EDC9D329D76 for ; Sun, 3 Oct 2004 16:27:08 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 85945-05 for ; Sun, 3 Oct 2004 15:26:51 +0000 (GMT) Received: from boutiquenumerique.com (gailleton-2-82-67-9-10.fbx.proxad.net [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id E4EB5329E2C for ; Sun, 3 Oct 2004 16:26:49 +0100 (BST) Received: (qmail 8856 invoked from network); 3 Oct 2004 17:27:12 +0200 Received: from unknown (HELO musicbox) (192.168.0.2) by gailleton-2-82-67-9-10.fbx.proxad.net with SMTP; 3 Oct 2004 17:27:12 +0200 To: Shiar , pgsql-performance@postgresql.org Subject: Re: index not used when using function References: <20040929194131.GH22917@shiar.org> Message-ID: From: =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= 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: Sun, 03 Oct 2004 17:29:37 +0200 In-Reply-To: <20040929194131.GH22917@shiar.org> User-Agent: Opera M2/7.53 (Linux, build 737) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/27 X-Sequence-Number: 8503 Maybe add an order by artist to force a groupaggregate ? > Hi all, a small question: > > I've got this table "songs" and an index on column artist. Since > there's about > one distinct artist for every 10 rows, it would be nice if it could use > this > index when counting artists. It doesn't however: > > lyrics=> EXPLAIN ANALYZE SELECT count(DISTINCT artist) FROM songs; > Aggregate (cost=31961.26..31961.26 rows=1 width=14) (actual > time=808.863..808.864 rows=1 loops=1) > -> Seq Scan on songs (cost=0.00..31950.41 rows=4341 width=14) > (actual time=26.801..607.172 rows=25207 loops=1) > Total runtime: 809.106 ms > > Even with enable_seqscan to off, it just can't seem to use the index. > The same > query without the count() works just fine: > > lyrics=> EXPLAIN ANALYZE SELECT DISTINCT artist FROM songs; > Unique (cost=0.00..10814.96 rows=828 width=14) (actual > time=0.029..132.903 rows=3280 loops=1) > -> Index Scan using songs_artist_key on songs (cost=0.00..10804.11 > rows=4341 width=14) (actual time=0.027..103.448 rows=25207 loops=1) > Total runtime: 135.697 ms > > Of course I can just take the number of rows from the latter query, but > I'm > still wondering why it can't use indexes with functions. > > Thanks From pgsql-performance-owner@postgresql.org Sun Oct 3 22:57:03 2004 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 8D333329E6F for ; Sun, 3 Oct 2004 22:57:02 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 71331-05 for ; Sun, 3 Oct 2004 21:56:51 +0000 (GMT) Received: from stark.xeocode.com (gsstark.mtl.istop.com [66.11.160.162]) by svr1.postgresql.org (Postfix) with ESMTP id C947E329E3D for ; Sun, 3 Oct 2004 22:56:53 +0100 (BST) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1CEELT-0002xE-00; Sun, 03 Oct 2004 17:56:43 -0400 To: Russell Smith Cc: Ryan VanMiddlesworth , pgsql-performance@postgresql.org Subject: Re: Query planner problem References: <200410011706.54231.ryan@vanmiddlesworth.org> <200410030850.28345.mr-russ@pws.com.au> In-Reply-To: <200410030850.28345.mr-russ@pws.com.au> From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 03 Oct 2004 17:56:42 -0400 Message-ID: <878yan1n51.fsf@stark.xeocode.com> Lines: 12 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/28 X-Sequence-Number: 8504 Russell Smith writes: > The Index does not store NULL values This is false. Though the fact that NULL values are indexed in postgres doesn't help with this poster's actual problem. -- greg From pgsql-performance-owner@postgresql.org Mon Oct 4 18:27:00 2004 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 C3B3E329F5E; Mon, 4 Oct 2004 18:26:57 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 88179-07; Mon, 4 Oct 2004 17:26:52 +0000 (GMT) Received: from fdd00lnhub.fds.com (external.fds.com [208.15.90.2]) by svr1.postgresql.org (Postfix) with ESMTP id 0AA46329EA5; Mon, 4 Oct 2004 18:26:51 +0100 (BST) In-Reply-To: To: "Patrick Hatcher Cc: pgsql-performance@postgresql.org, pgsql-performance-owner@postgresql.org Subject: Re: Slow update/insert process MIME-Version: 1.0 X-Mailer: Lotus Notes Release 6.5 September 18, 2003 Message-ID: From: Patrick Hatcher Date: Mon, 4 Oct 2004 10:14:13 -0700 X-MIMETrack: Serialize by Router on FDD00LNHUB/FSG/SVR/FDD(Release 6.5.2|June 01, 2004) at 10/04/2004 01:26:20 PM, Serialize complete at 10/04/2004 01:26:20 PM Content-Type: multipart/alternative; boundary="=_alternative 005FDF4C88256F23_=" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=3.3 tagged_above=0.0 required=5.0 tests=HTML_30_40, HTML_MESSAGE, LINES_OF_YELLING, LINES_OF_YELLING_2, TO_HAS_SPACES X-Spam-Level: *** X-Archive-Number: 200410/29 X-Sequence-Number: 8505 This is a multipart message in MIME format. --=_alternative 005FDF4C88256F23_= Content-Type: text/plain; charset="US-ASCII" Thanks for the help. I found the culprit. The user had created a function within the function ( pm.pm_price_post_inc(prod.keyp_products)). Once this was fixed the time dropped dramatically. Patrick Hatcher Macys.Com Legacy Integration Developer 415-422-1610 office HatcherPT - AIM Patrick Hatcher Sent by: pgsql-performance-owner@postgresql.org 10/01/04 11:14 AM To cc Subject [PERFORM] Slow update/insert process Pg: 7.4.5 RH 7.3 8g Ram 200 g drive space RAID0+1 Tables vacuum on a nightly basis The following process below takes 8 hours to run on 90k records and I'm not sure where to being to look for the bottleneck. This isn't the only updating on this database that seems to take a long time to complete. Is there something I should be looking for in my conf settings? TIA Patrick SQL: ---Bring back only selected records to run through the update process. --Without the function the SQL takes < 10secs to return 90,000 records SELECT count(pm.pm_delta_function_amazon(upc.keyp_upc,'amazon')) FROM mdc_upc upc JOIN public.mdc_products prod ON upc.keyf_products = prod.keyp_products JOIN public.mdc_price_post_inc price ON prod.keyp_products = price.keyf_product JOIN public.mdc_attribute_product ap on ap.keyf_products = prod.keyp_products and keyf_attribute=22 WHERE upper(trim(ap.attributevalue)) NOT IN ('ESTEE LAUDER', 'CLINIQUE','ORGINS','PRESCRIPTIVES','LANC?ME','CHANEL','ARAMIS','M.A.C','TAG HEUER') AND keyf_producttype<>222 AND prod.action_publish = 1; Function: CREATE OR REPLACE FUNCTION pm.pm_delta_function_amazon(int4, "varchar") RETURNS bool AS 'DECLARE varkeyf_upc ALIAS FOR $1; varPassword ALIAS FOR $2; varRealMD5 varchar; varDeltaMD5 varchar; varLastTouchDate date; varQuery text; varQuery1 text; varQueryMD5 text; varQueryRecord record; varFuncStatus boolean := false; BEGIN -- Check the password IF varPassword <> \'amazon\' THEN Return false; END IF; -- Get the md5 hash for this product SELECT into varQueryRecord md5(upc.keyp_upc || prod.description || pm.pm_price_post_inc(prod.keyp_products)) AS md5 FROM public.mdc_upc upc JOIN public.mdc_products prod ON upc.keyf_products = prod.keyp_products JOIN public.mdc_price_post_inc price ON price.keyf_product = prod.keyp_products WHERE upc.keyp_upc = varkeyf_upc LIMIT 1 ; IF NOT FOUND THEN RAISE EXCEPTION \'varRealMD5 is NULL. UPC ID is %\', varkeyf_upc; ELSE varRealMD5:=varQueryRecord.md5; END IF; -- Check that the product is in the delta table and return its hash for comparison SELECT into varQueryRecord md5_hash,last_touch_date FROM pm.pm_delta_master_amazon WHERE keyf_upc = varkeyf_upc LIMIT 1; IF NOT FOUND THEN -- ADD and exit INSERT INTO pm.pm_delta_master_amazon (keyf_upc,status,md5_hash,last_touch_date) values (varkeyf_upc,\'add\',varRealMD5,CURRENT_DATE); varFuncStatus:=true; RETURN varFuncStatus; ELSE --Update the record --- If the hash matches then set the record to HOLD IF varRealMD5 = varQueryRecord.md5_hash THEN UPDATE pm.pm_delta_master_amazon SET status= \'hold\', last_touch_date = CURRENT_DATE WHERE keyf_upc = varkeyf_upc AND last_touch_date <> CURRENT_DATE; varFuncStatus:=true; ELSE -- ELSE mark the item as ADD UPDATE pm.pm_delta_master_amazon SET status= \'add\', last_touch_date = CURRENT_DATE WHERE keyf_upc = varkeyf_upc; varFuncStatus:=true; END IF; END IF; RETURN varFuncStatus; END;' LANGUAGE 'plpgsql' IMMUTABLE; TableDef CREATE TABLE pm.pm_delta_master_amazon ( keyf_upc int4 , status varchar(6) , md5_hash varchar(40) , last_touch_date date ) GO CREATE INDEX status_idx ON pm.pm_delta_master_amazon(status) GO CONF -------- # WRITE AHEAD LOG #--------------------------------------------------------------------------- # - Settings - #fsync = true # turns forced synchronization on or off #wal_sync_method = fsync # the default varies across platforms: # fsync, fdatasync, open_sync, or open_datasync wal_buffers = 32 # min 4, 8KB each # - Checkpoints - checkpoint_segments = 50 # in logfile segments, min 1, 16MB each checkpoint_timeout = 600 # 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 Patrick Hatcher Macys.Com --=_alternative 005FDF4C88256F23_= Content-Type: text/html; charset="US-ASCII"
Thanks for the help.
I found the culprit.  The user had created a function within the function ( pm.pm_price_post_inc(prod.keyp_products)). Once this was fixed the time dropped dramatically.

Patrick Hatcher
Macys.Com
Legacy Integration Developer
415-422-1610 office
HatcherPT - AIM


Patrick Hatcher <PHatcher@macys.com>
Sent by: pgsql-performance-owner@postgresql.org

10/01/04 11:14 AM

To
<pgsql-performance@postgresql.org>
cc
Subject
[PERFORM] Slow update/insert process






Pg: 7.4.5
RH 7.3
8g Ram
200 g drive space
RAID0+1
Tables vacuum on a nightly basis

The following process below takes 8 hours to run on 90k records and I'm not sure where to being to look for the bottleneck.  This isn't the only updating on this database that seems to take a long time to complete.  Is there something I should be looking for in my conf settings?  

TIA
Patrick


SQL:
---Bring back only selected records to run through the update process.
--Without the function the SQL takes < 10secs to return 90,000 records
SELECT count(pm.pm_delta_function_amazon(upc.keyp_upc,'amazon'))
FROM mdc_upc upc
JOIN public.mdc_products prod ON upc.keyf_products = prod.keyp_products
JOIN public.mdc_price_post_inc price ON prod.keyp_products = price.keyf_product
JOIN public.mdc_attribute_product ap on ap.keyf_products = prod.keyp_products and keyf_attribute=22
WHERE
upper(trim(ap.attributevalue)) NOT IN ('ESTEE LAUDER', 'CLINIQUE','ORGINS','PRESCRIPTIVES','LANC?ME','CHANEL','ARAMIS','M.A.C','TAG HEUER')
AND keyf_producttype<>222
AND prod.action_publish = 1;


Function:

CREATE OR REPLACE FUNCTION pm.pm_delta_function_amazon(int4, "varchar")
RETURNS bool AS
'DECLARE
  varkeyf_upc                ALIAS FOR $1;
  varPassword                ALIAS FOR $2;
  varRealMD5                varchar;
  varDeltaMD5                varchar;
  varLastTouchDate        date;
  varQuery                 text;
  varQuery1                 text;
  varQueryMD5                text;
  varQueryRecord        record;
  varFuncStatus        boolean := false;
 
BEGIN

-- Check the password
IF varPassword <> \'amazon\' THEN
  Return false;
END IF;


--  Get the md5 hash for this product
SELECT into varQueryRecord md5(upc.keyp_upc || prod.description || pm.pm_price_post_inc(prod.keyp_products)) AS md5
  FROM public.mdc_upc upc
  JOIN public.mdc_products prod ON upc.keyf_products = prod.keyp_products
  JOIN public.mdc_price_post_inc price ON price.keyf_product = prod.keyp_products
  WHERE upc.keyp_upc = varkeyf_upc LIMIT 1 ;


IF NOT FOUND THEN
  RAISE EXCEPTION \'varRealMD5 is NULL. UPC ID is %\', varkeyf_upc;
ELSE
  varRealMD5:=varQueryRecord.md5;
END IF;

--  Check that the product is in the delta table and return its hash for comparison
SELECT into varQueryRecord md5_hash,last_touch_date
  FROM pm.pm_delta_master_amazon
  WHERE keyf_upc = varkeyf_upc LIMIT 1;

IF NOT FOUND THEN
  -- ADD and exit
  INSERT INTO pm.pm_delta_master_amazon (keyf_upc,status,md5_hash,last_touch_date)
  values (varkeyf_upc,\'add\',varRealMD5,CURRENT_DATE);
  varFuncStatus:=true;
  RETURN varFuncStatus;
ELSE
  --Update the record
    --- If the hash matches then set the record to HOLD
  IF varRealMD5 = varQueryRecord.md5_hash THEN
      UPDATE pm.pm_delta_master_amazon
        SET status= \'hold\',
        last_touch_date =  CURRENT_DATE
      WHERE keyf_upc =  varkeyf_upc AND last_touch_date <> CURRENT_DATE;
      varFuncStatus:=true;
  ELSE
      --  ELSE mark the item as ADD
      UPDATE pm.pm_delta_master_amazon
        SET status= \'add\',
        last_touch_date =  CURRENT_DATE
      WHERE keyf_upc =  varkeyf_upc;
      varFuncStatus:=true;
  END IF;
 END IF;

RETURN varFuncStatus;
END;'
LANGUAGE 'plpgsql' IMMUTABLE;



TableDef
CREATE TABLE pm.pm_delta_master_amazon (
   keyf_upc               int4 ,
   status                 varchar(6) ,
   md5_hash               varchar(40) ,
   last_touch_date        date
   )
GO

CREATE INDEX status_idx
   ON pm.pm_delta_master_amazon(status)
GO




CONF
--------
# WRITE AHEAD LOG
#---------------------------------------------------------------------------

# - Settings -

#fsync = true                   # turns forced synchronization on or off
#wal_sync_method = fsync        # the default varies across platforms:
                               # fsync, fdatasync, open_sync, or open_datasync
wal_buffers = 32                # min 4, 8KB each

# - Checkpoints -

checkpoint_segments = 50        # in logfile segments, min 1, 16MB each
checkpoint_timeout = 600        # 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


Patrick Hatcher
Macys.Com
--=_alternative 005FDF4C88256F23_=-- From pgsql-performance-owner@postgresql.org Mon Oct 4 18:39:04 2004 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 21BFE329F8A for ; Mon, 4 Oct 2004 18:39:03 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 92935-04 for ; Mon, 4 Oct 2004 17:38:59 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id F1789329F6A for ; Mon, 4 Oct 2004 18:38:59 +0100 (BST) 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 6456240; Mon, 04 Oct 2004 10:40:21 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: Steve Atkins Subject: Re: Performance suggestions for an update-mostly database? Date: Mon, 4 Oct 2004 10:38:14 -0700 User-Agent: KMail/1.6.2 Cc: pgsql-performance@postgresql.org References: <20041004174019.GA25671@gp.word-to-the-wise.com> In-Reply-To: <20041004174019.GA25671@gp.word-to-the-wise.com> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410041038.14571.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/31 X-Sequence-Number: 8507 Steve, > I'm used to performance tuning on a select-heavy database, but this > will have a very different impact on the system. Does anyone have any > experience with an update heavy system, and have any performance hints > or hardware suggestions? Minimal/no indexes on the table(s). Raise checkpoint_segments and consider using commit_siblings/commit_delay if it's a multi-stream application. Figure out ways to do inserts instead of updates where possible, and COPY instead of insert, where possible. Put your WAL on its own disk resource. I'm a little curious as to what kind of app would be 95% writes. A log? -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Mon Oct 4 18:33:36 2004 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 B74B432A05F for ; Mon, 4 Oct 2004 18:33:34 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91468-02 for ; Mon, 4 Oct 2004 17:33:29 +0000 (GMT) Received: from gp.word-to-the-wise.com (gp.word-to-the-wise.com [64.71.176.18]) by svr1.postgresql.org (Postfix) with ESMTP id B577E32A050 for ; Mon, 4 Oct 2004 18:33:28 +0100 (BST) Received: by gp.word-to-the-wise.com (Postfix, from userid 500) id 451CF90000E; Mon, 4 Oct 2004 10:40:19 -0700 (PDT) Date: Mon, 4 Oct 2004 10:40:19 -0700 From: Steve Atkins To: pgsql-performance@postgresql.org Subject: Performance suggestions for an update-mostly database? Message-ID: <20041004174019.GA25671@gp.word-to-the-wise.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.4.1i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/30 X-Sequence-Number: 8506 I'm putting together a system where the operation mix is likely to be >95% update, <5% select on primary key. I'm used to performance tuning on a select-heavy database, but this will have a very different impact on the system. Does anyone have any experience with an update heavy system, and have any performance hints or hardware suggestions? Cheers, Steve From pgsql-performance-owner@postgresql.org Mon Oct 4 18:53:49 2004 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 C4B3432A42A for ; Mon, 4 Oct 2004 18:53:48 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 96984-03 for ; Mon, 4 Oct 2004 17:53:43 +0000 (GMT) Received: from gp.word-to-the-wise.com (gp.word-to-the-wise.com [64.71.176.18]) by svr1.postgresql.org (Postfix) with ESMTP id 92DDC32A417 for ; Mon, 4 Oct 2004 18:53:44 +0100 (BST) Received: by gp.word-to-the-wise.com (Postfix, from userid 500) id 64D7190000E; Mon, 4 Oct 2004 11:00:36 -0700 (PDT) Date: Mon, 4 Oct 2004 11:00:36 -0700 From: Steve Atkins To: pgsql-performance@postgresql.org Subject: Re: Performance suggestions for an update-mostly database? Message-ID: <20041004180036.GA26390@gp.word-to-the-wise.com> References: <20041004174019.GA25671@gp.word-to-the-wise.com> <200410041038.14571.josh@agliodbs.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <200410041038.14571.josh@agliodbs.com> User-Agent: Mutt/1.4.1i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/32 X-Sequence-Number: 8508 On Mon, Oct 04, 2004 at 10:38:14AM -0700, Josh Berkus wrote: > Steve, > > > I'm used to performance tuning on a select-heavy database, but this > > will have a very different impact on the system. Does anyone have any > > experience with an update heavy system, and have any performance hints > > or hardware suggestions? > > Minimal/no indexes on the table(s). Raise checkpoint_segments and consider > using commit_siblings/commit_delay if it's a multi-stream application. > Figure out ways to do inserts instead of updates where possible, and COPY > instead of insert, where possible. Put your WAL on its own disk resource. Thanks. > I'm a little curious as to what kind of app would be 95% writes. A log? It's the backend to a web application. The applications mix of queries is pretty normal, but it uses a large, in-core, write-through cache between the business logic and the database. It has more than usual locality on queries over short time periods, so the vast majority of reads should be answered out of the cache and not touch the database. In some ways something like Berkeley DB might be a better match to the frontend, but I'm comfortable with PostgreSQL and prefer to have the power of SQL commandline for when I need it. Cheers, Steve From pgsql-performance-owner@postgresql.org Mon Oct 4 20:00:36 2004 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 83A49329F69 for ; Mon, 4 Oct 2004 20:00:34 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 18946-09 for ; Mon, 4 Oct 2004 19:00:26 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 53336329E91 for ; Mon, 4 Oct 2004 20:00:26 +0100 (BST) 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 6456684; Mon, 04 Oct 2004 12:01:48 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: Steve Atkins Subject: Re: Performance suggestions for an update-mostly database? Date: Mon, 4 Oct 2004 12:02:03 -0700 User-Agent: KMail/1.6.2 Cc: pgsql-performance@postgresql.org References: <20041004174019.GA25671@gp.word-to-the-wise.com> <200410041038.14571.josh@agliodbs.com> <20041004180036.GA26390@gp.word-to-the-wise.com> In-Reply-To: <20041004180036.GA26390@gp.word-to-the-wise.com> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410041202.03548.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/33 X-Sequence-Number: 8509 Steve, > In some ways something like Berkeley DB might be a better match to the > frontend, but I'm comfortable with PostgreSQL and prefer to have the > power of SQL commandline for when I need it. Well, if data corruption is not a concern, you can always turn off checkpointing. This will save you a fair amount of overhead. You could also look at telegraphCQ. It's not prodcucton yet, but their idea of "streams" as data sources really seems to fit with what you're talking about. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Mon Oct 4 20:18:59 2004 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 8560D329E32 for ; Mon, 4 Oct 2004 20:18:58 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 27014-07 for ; Mon, 4 Oct 2004 19:18:52 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [66.143.173.58]) by svr1.postgresql.org (Postfix) with ESMTP id 69A8C329D9F for ; Mon, 4 Oct 2004 20:18:52 +0100 (BST) Received: by flake.decibel.org (Postfix, from userid 1001) id 2D86C1C8F8; Mon, 4 Oct 2004 19:18:50 +0000 (GMT) Date: Mon, 4 Oct 2004 14:18:50 -0500 From: "Jim C. Nasby" To: Josh Berkus Cc: Postgresql Performance Subject: Re: Caching of Queries Message-ID: <20041004191850.GS1297@decibel.org> References: <00e801c4a4b6$5526dd30$8300a8c0@solent> <20040930221107.GP1297@decibel.org> <415CEE8E.8060805@ymogen.net> <200410011010.40705.josh@agliodbs.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <200410011010.40705.josh@agliodbs.com> X-Operating-System: FreeBSD 4.10-RELEASE-p2 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/34 X-Sequence-Number: 8510 On Fri, Oct 01, 2004 at 10:10:40AM -0700, Josh Berkus wrote: > Transparent "query caching" is the "industry standard" for how these things > are handled. However, Postgres' lack of this feature has made me consider > other approaches, and I'm starting to wonder if the "standard" query caching > -- where a materialized query result, or some reduction thereof, is cached in > database memory -- isn't the best way to cache things. I'm going to > abbreviate it "SQC" for the rest of this e-mail. Not to quibble, but are you sure that's the standard? Oracle and DB2 don't do this, and I didn't think MSSQL did either. What they do do is cache query *plans*. This is a *huge* deal in Oracle; search http://asktom.oracle.com for 'soft parse'. In any case, I think a means of marking some specific queries as being cachable is an excellent idea; perfect for 'static data' scenarios. What I don't know is how much will be saved. -- 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 Oct 4 20:23:01 2004 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 6DFE132A47E for ; Mon, 4 Oct 2004 20:22:59 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 28057-04 for ; Mon, 4 Oct 2004 19:22:53 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [66.143.173.58]) by svr1.postgresql.org (Postfix) with ESMTP id D63BA32A318 for ; Mon, 4 Oct 2004 20:22:54 +0100 (BST) Received: by flake.decibel.org (Postfix, from userid 1001) id 365271C8F8; Mon, 4 Oct 2004 19:22:54 +0000 (GMT) Date: Mon, 4 Oct 2004 14:22:54 -0500 From: "Jim C. Nasby" To: Josh Berkus Cc: Steve Atkins , pgsql-performance@postgresql.org Subject: Re: Performance suggestions for an update-mostly database? Message-ID: <20041004192254.GT1297@decibel.org> References: <20041004174019.GA25671@gp.word-to-the-wise.com> <200410041038.14571.josh@agliodbs.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <200410041038.14571.josh@agliodbs.com> X-Operating-System: FreeBSD 4.10-RELEASE-p2 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/35 X-Sequence-Number: 8511 And obviously make sure you're vacuuming frequently. On Mon, Oct 04, 2004 at 10:38:14AM -0700, Josh Berkus wrote: > Steve, > > > I'm used to performance tuning on a select-heavy database, but this > > will have a very different impact on the system. Does anyone have any > > experience with an update heavy system, and have any performance hints > > or hardware suggestions? > > Minimal/no indexes on the table(s). Raise checkpoint_segments and consider > using commit_siblings/commit_delay if it's a multi-stream application. > Figure out ways to do inserts instead of updates where possible, and COPY > instead of insert, where possible. Put your WAL on its own disk resource. > > I'm a little curious as to what kind of app would be 95% writes. A log? > > -- > Josh Berkus > Aglio Database Solutions > San Francisco > > ---------------------------(end of broadcast)--------------------------- > TIP 2: you can get off all lists at once with the unregister command > (send "unregister YourEmailAddressHere" to majordomo@postgresql.org) > -- 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 Oct 5 00:28:05 2004 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 A7E7B329E01 for ; Tue, 5 Oct 2004 00:28:01 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 98525-04 for ; Mon, 4 Oct 2004 23:27:55 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.193]) by svr1.postgresql.org (Postfix) with ESMTP id 7ADC1329D9F for ; Tue, 5 Oct 2004 00:27:56 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 73so3431920rnk for ; Mon, 04 Oct 2004 16:27:52 -0700 (PDT) Received: by 10.38.208.76 with SMTP id f76mr2869849rng; Mon, 04 Oct 2004 16:27:51 -0700 (PDT) Received: by 10.38.151.40 with HTTP; Mon, 4 Oct 2004 16:27:51 -0700 (PDT) Message-ID: <59b2d39b0410041627d405f39@mail.gmail.com> Date: Mon, 4 Oct 2004 16:27:51 -0700 From: Miles Keaton Reply-To: Miles Keaton To: pgsql-performance@postgresql.org Subject: would number of fields in a table affect search-query time? 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/36 X-Sequence-Number: 8512 would the number of fields in a table significantly affect the search-query time? (meaning: less fields = much quicker response?) I have this database table of items with LOTS of properties per-item, that takes a LONG time to search. So as I was benchmarking it against SQLite, MySQL and some others, I exported just a few fields for testing, into all three databases. What surprised me the most is that the subset, even in the original database, gave search results MUCH faster than the full table! I know I'm being vague, but does anyone know if this is just common knowledge ("duh! of course!") or if I should be looking at is as a problem to fix? From pgsql-performance-owner@postgresql.org Tue Oct 5 00:32:42 2004 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 4BAD132A2B4 for ; Tue, 5 Oct 2004 00:32:16 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 98136-10 for ; Mon, 4 Oct 2004 23:32:09 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id E5F0B329E01 for ; Tue, 5 Oct 2004 00:32:10 +0100 (BST) 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 i94NW93t007325; Mon, 4 Oct 2004 19:32:10 -0400 (EDT) To: Miles Keaton Cc: pgsql-performance@postgresql.org Subject: Re: would number of fields in a table affect search-query time? In-reply-to: <59b2d39b0410041627d405f39@mail.gmail.com> References: <59b2d39b0410041627d405f39@mail.gmail.com> Comments: In-reply-to Miles Keaton message dated "Mon, 04 Oct 2004 16:27:51 -0700" Date: Mon, 04 Oct 2004 19:32:09 -0400 Message-ID: <7324.1096932729@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/37 X-Sequence-Number: 8513 Miles Keaton writes: > What surprised me the most is that the subset, even in the original > database, gave search results MUCH faster than the full table! The subset table's going to be physically much smaller, so it could just be that this reflects smaller I/O load. Hard to tell without a lot more detail about what case you were testing. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Oct 5 00:34:49 2004 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 317B4329E01 for ; Tue, 5 Oct 2004 00:34:44 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 01525-02 for ; Mon, 4 Oct 2004 23:34:38 +0000 (GMT) Received: from trofast.sesse.net (trofast.sesse.net [129.241.93.32]) by svr1.postgresql.org (Postfix) with ESMTP id 95EC2329DF7 for ; Tue, 5 Oct 2004 00:34:37 +0100 (BST) Received: from sesse by trofast.sesse.net with local (Exim 3.36 #1 (Debian)) id 1CEcLk-0003UO-00 for ; Tue, 05 Oct 2004 01:34:36 +0200 Date: Tue, 5 Oct 2004 01:34:36 +0200 From: "Steinar H. Gunderson" To: pgsql-performance@postgresql.org Subject: Re: would number of fields in a table affect search-query time? Message-ID: <20041004233436.GB13394@uio.no> Mail-Followup-To: pgsql-performance@postgresql.org References: <59b2d39b0410041627d405f39@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <59b2d39b0410041627d405f39@mail.gmail.com> X-Operating-System: Linux 2.6.6 on a i686 User-Agent: Mutt/1.5.6+20040818i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/38 X-Sequence-Number: 8514 On Mon, Oct 04, 2004 at 04:27:51PM -0700, Miles Keaton wrote: > would the number of fields in a table significantly affect the > search-query time? More fields = larger records = fewer records per page = if you read in everything, you'll need more I/O. > I have this database table of items with LOTS of properties per-item, > that takes a LONG time to search. It's a bit hard to say anything without seeing your actual tables and queries; I'd guess you either have a lot of matches or you're doing a sequential scan. You might want to normalize your tables, but again, it's hard to say anything without seeing your actual data. /* Steinar */ -- Homepage: http://www.sesse.net/ From pgsql-performance-owner@postgresql.org Tue Oct 5 03:07:28 2004 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 8C9B5329F40 for ; Tue, 5 Oct 2004 03:07:27 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 33280-01 for ; Tue, 5 Oct 2004 02:07:18 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 7CF0D329D20 for ; Tue, 5 Oct 2004 03:07:20 +0100 (BST) 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 6458521; Mon, 04 Oct 2004 19:08:41 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: Miles Keaton Subject: Re: would number of fields in a table affect search-query time? Date: Mon, 4 Oct 2004 19:06:33 -0700 User-Agent: KMail/1.6.2 Cc: pgsql-performance@postgresql.org References: <59b2d39b0410041627d405f39@mail.gmail.com> In-Reply-To: <59b2d39b0410041627d405f39@mail.gmail.com> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410041906.33483.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/39 X-Sequence-Number: 8515 Miles, > would the number of fields in a table significantly affect the > search-query time? Yes. In addition to the issues mentioned previously, there is the issue of criteria; an OR query on 8 fields is going to take longer to filter than an OR query on 2 fields. Anyway, I think maybe you should tell us more about your database design. Often the fastest solution involves a more sophisticated approach toward querying your tables. -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Tue Oct 5 07:39:39 2004 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 AF642329CB7 for ; Tue, 5 Oct 2004 07:39:32 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 92176-02 for ; Tue, 5 Oct 2004 06:39:26 +0000 (GMT) Received: from pd3mo1so.prod.shaw.ca (shawidc-mo1.cg.shawcable.net [24.71.223.10]) by svr1.postgresql.org (Postfix) with ESMTP id 1CF2732A369 for ; Tue, 5 Oct 2004 07:39:25 +0100 (BST) Received: from pd3mr3so.prod.shaw.ca (pd3mr3so-qfe3.prod.shaw.ca [10.0.141.179]) by l-daemon (Sun ONE Messaging Server 6.0 HotFix 1.01 (built Mar 15 2004)) with ESMTP id <0I5300F7PMHPI420@l-daemon> for pgsql-performance@postgresql.org; Tue, 05 Oct 2004 00:39:25 -0600 (MDT) Received: from pn2ml10so.prod.shaw.ca ([10.0.121.80]) by pd3mr3so.prod.shaw.ca (Sun ONE Messaging Server 6.0 HotFix 1.01 (built Mar 15 2004)) with ESMTP id <0I5300JQ4MHPSQ10@pd3mr3so.prod.shaw.ca> for pgsql-performance@postgresql.org; Tue, 05 Oct 2004 00:39:25 -0600 (MDT) Received: from [192.168.1.10] (S01060050bac04c93.ed.shawcable.net [68.148.193.184]) by l-daemon (iPlanet Messaging Server 5.2 HotFix 1.18 (built Jul 28 2003)) with ESMTP id <0I530015CMHPQJ@l-daemon> for pgsql-performance@postgresql.org; Tue, 05 Oct 2004 00:39:25 -0600 (MDT) Date: Tue, 05 Oct 2004 00:39:23 -0600 From: Patrick Clery Subject: Re: Comparing user attributes with bitwise operators In-reply-to: <87sm9eluc3.fsf@stark.xeocode.com> To: Greg Stark Cc: pgsql-performance@postgresql.org Message-id: <200410050039.24230.patrick@phpforhire.com> MIME-version: 1.0 Content-type: text/plain; charset=iso-8859-1 Content-transfer-encoding: 7bit Content-disposition: inline References: <200409160141.37604.patrick@phpforhire.com> <200409182126.13686.patrick@phpforhire.com> <87sm9eluc3.fsf@stark.xeocode.com> User-Agent: KMail/1.7 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.9 tagged_above=0.0 required=5.0 tests=FAKE_HELO_SHAW_CA X-Spam-Level: X-Archive-Number: 200410/40 X-Sequence-Number: 8516 Sorry I have taken this long to reply, Greg, but here are the results of the personals site done with contrib/intarray: The first thing I did was add a serial column to the attributes table. So instead of having a unique constraint on (attribute_id,value_id), every row has a unique value: datingsite=> \d attribute_names Table "public.attribute_names" Column | Type | Modifiers ----------------+-----------------------+--------------------------------------------------------------------------- attribute_id | integer | not null default nextval('public.attribute_names_attribute_id_seq'::text) attribute_name | character varying(50) | not null Indexes: "attribute_names_pkey" PRIMARY KEY, btree (attribute_id) "attribute_names_attribute_id_key" UNIQUE, btree (attribute_id, attribute_name an example insert: insert into attribute_names (attribute_name) values ('languages'); datingsite=> \d attribute_values Table "public.attribute_values" Column | Type | Modifiers --------------+------------------------+------------------------------------------------------------------------ attribute_id | integer | not null order_id | integer | not null default (nextval('order_id_seq'::text) - 1) label | character varying(255) | not null value_id | integer | not null default nextval('public.attribute_values_value_id_seq'::text) Indexes: "attribute_values_pkey" PRIMARY KEY, btree (value_id) Foreign-key constraints: "attribute_values_attribute_id_fkey" FOREIGN KEY (attribute_id) REFERENCES attribute_names(attribute_id) an example insert (22 is the attribute_id of "languages"): insert into attribute_values (attribute_id, label) values (22, 'English'); The "value_id" column is where the integers inside the int[] arrays will reference. Even age (between 18-99) and height (between 48-84) have rows for every possible choice, as well as "Ask me!" where a user could choose to leave that blank. Here is "the int[] table": create table people_attributes ( person_id int references people (person_id) on delete cascade initially deferred, askmecount int not null default 0, age int not null references attribute_values(value_id) on delete restrict, gender int not null references attribute_values(value_id) on delete restrict, bodytype int not null references attribute_values(value_id) on delete restrict, children int not null references attribute_values(value_id) on delete restrict, drinking int not null references attribute_values(value_id) on delete restrict, education int not null references attribute_values(value_id) on delete restrict, ethnicity int not null references attribute_values(value_id) on delete restrict, eyecolor int not null references attribute_values(value_id) on delete restrict, haircolor int not null references attribute_values(value_id) on delete restrict, hairstyle int not null references attribute_values(value_id) on delete restrict, height int not null references attribute_values(value_id) on delete restrict, income int not null references attribute_values(value_id) on delete restrict, languages int[] not null, occupation int not null references attribute_values(value_id) on delete restrict, orientation int not null references attribute_values(value_id) on delete restrict, relation int not null references attribute_values(value_id) on delete restrict, religion int not null references attribute_values(value_id) on delete restrict, smoking int not null references attribute_values(value_id) on delete restrict, want_children int not null references attribute_values(value_id) on delete restrict, weight int not null references attribute_values(value_id) on delete restrict, seeking int[] not null, primary key (person_id) ) without oids; If you'll notice that "seeking" and "languages" are both int[] types. I did this because those will be multiple choice. The index was created like so: create index people_attributes_search on people_attributes using gist ( (array[ age, gender, orientation, children, drinking, education, ethnicity, eyecolor, haircolor, hairstyle, height, income, occupation, relation, religion, smoking, want_children, weight ] + seeking + languages) gist__int_ops ); seeking and languages are appended with the intarray + op. I'm not going to go too in depth on how this query was generated since that was mostly done with the PHP side of things, but from the structure it should be obvious. I did, however, have to generate a few SQL functions using Smarty templates since it would be way to tedious to map all these values by hand. There are 96,000 rows (people) in the people_attributes table. Here is what is going on in the following query: "Show me all single (48) females (88) who are heterosexual (92) age between 18 and 31 (95|96|97|98|99|100|101|102|103| 104|105|106|107|108)" EXPLAIN ANALYZE SELECT * FROM people_attributes pa WHERE person_id <> 1 AND (ARRAY[age, gender, orientation, children, drinking, education, ethnicity, eyecolor, haircolor, hairstyle, height, income, occupation, relation, religion, smoking, want_children, weight] + seeking + languages) @@ '48 & 89 & 92 & ( 95 | 96 | 97 | 98 | 99 | 100 | 101 | 102 | 103 | 104 | 105 | 106 | 107 | 108 )' Index Scan using people_attributes_search on people_attributes pa (cost=0.00..386.45 rows=96 width=140) (actual time=0.057..19.266 rows=516 loops=1) Index Cond: (((ARRAY[age, gender, orientation, children, drinking, education, ethnicity, eyecolor, haircolor, hairstyle, height, income, occupation, relation, religion, smoking, want_children, weight] + seeking) + languages) @@ '48 & 89 & 92 & ( ( ( ( ( ( ( ( ( ( ( ( ( 95 | 96 ) | 97 ) | 98 ) | 99 ) | 100 ) | 101 ) | 102 ) | 103 ) | 104 ) | 105 ) | 106 ) | 107 ) | 108 )'::query_int) Filter: (person_id <> 1) Total runtime: 21.646 ms The speed only seems to vary significant on very broad searches, e.g: "All females." But once the threshold of about 2 or more attributes is met, the times are very acceptable. If we get a little more specific by adding "non-smokers and non-drinkers between 18 and 22", slight improvements: EXPLAIN ANALYZE SELECT * FROM people_attributes pa WHERE person_id <> 1 AND (ARRAY[age, gender, orientation, children, drinking, education, ethnicity, eyecolor, haircolor, hairstyle, height, income, occupation, relation, religion, smoking, want_children, weight] + seeking + languages) @@ '48 & 89 & 92 & ( 95 | 96 | 97 | 98) & 67 & 2' Index Scan using people_attributes_search on people_attributes pa (cost=0.00..386.45 rows=96 width=140) (actual time=0.077..13.090 rows=32 loops=1) Index Cond: (((ARRAY[age, gender, orientation, children, drinking, education, ethnicity, eyecolor, haircolor, hairstyle, height, income, occupation, relation, religion, smoking, want_children, weight] + seeking) + languages) @@ '48 & 89 & 92 & ( ( ( 95 | 96 ) | 97 ) | 98 ) & 67 & 2'::query_int) Filter: (person_id <> 1) Total runtime: 13.393 ms All in all, my final thoughts on this are that it is "hella faster" than the previous methods. Vertical tables for your user attributes will not work for a personals site -- there are just too many conditions to be able to efficiently use an index. Out of all the methods I have tried, verticle table was not even remotely scalable on large amounts of data. Horizontal table is the way to go, but it wouldn't perform like this if not for the intarray module. The array method works quite nicely, especially for the columns like "languages" and "seeking" that are multiple choice. However, even though this method is fast, I still might opt for caching the results because the "real world" search query involves a lot more and will be executed non-stop. But to have it run this fast the first time certainly helps. The only drawback I can think of is that the attributes no longer have values like 1,2,3 -- instead they could be any integer value. This puts a spin on the programming side of things, which required me to write "code that writes code" on a few occassions during the attribute "mapping" process. For example, keeping an associative array of all the attributes without fetching that data from the database each time. My advice: if you're not a masochist, use a template engine (or simply parse out a print_r() ) to create these PHP arrays or SQL functions. Greg, thanks a lot for the advice. I owe you a beer ;) On Saturday 18 September 2004 23:07, you wrote: > Patrick Clery writes: > > PLAN > > ------------------------------------------------------------------------- > >-------------------------------------------------------------------------- > >-------------- Limit (cost=6.03..6.03 rows=1 width=68) (actual > > time=69.391..69.504 rows=10 loops=1) -> Sort (cost=6.03..6.03 rows=1 > > width=68) (actual time=69.381..69.418 rows=10 loops=1) Sort Key: age > > -> Index Scan using people_attributes_search on > > people_attributes (cost=0.00..6.02 rows=1 width=68) (actual > > time=0.068..61.648 rows=937 loops=1) Index Cond: > > (('{30,31,32,33,34,35,36,37,38,39,40}'::integer[] && age) AND > > ('{2}'::integer[] && gender) AND ('{1,2,4}'::integer[] && orientation)) > > Total runtime: 69.899 ms > > (6 rows) > > ... > > > - Is there a way of speeding up the sort? > > The sort seems to have only taken 8ms out of 69ms or just over 10%. As long > as the index scan doesn't match too many records the sort should never be > any slower so it shouldn't be the performance bottleneck. You might > consider putting a subquery inside the order by with a limit to ensure that > the sort never gets more than some safe maximum. Something like: > > select * from (select * from people_attributes where ... limit 1000) order > by age limit 10 > > This means if the query matches more than 1000 it won't be sorted properly > by age; you'll get the top 10 out of some random subset. But you're > protected against ever having to sort more than 1000 records. > > > - Will using queries like " WHERE orientation IN (1,2,4) " be any > > better/worse? > > Well they won't use the GiST index, so no. If there was a single column > with a btree index then this would be the cleanest way to go. > > > - The queries with the GiST index are faster, but is it of any benefit > > when the int[] arrays all contain a single value? > > Well you've gone from 5 minutes to 60ms. You'll have to do more than one > test to be sure but it sure seems like it's of some benefit. > > If they're always a single value you could make it an expression index > instead and not have to change your data model. > > Just have the fields be integers individually and make an index as: > > create index idx on people_attributes using gist ( > (array[age]) gist__int_ops, > (array[gender]) gist__int_ops, > ... > ) > > > However I would go one step further. I would make the index simply: > > create index idx on people_attributes using gist ( > (array[age,gender,orientation,...]) gist__int_ops > ) > > And ensure that all of these attributes have distinct domains. Ie, that > they don't share any values. There are 4 billion integer values available > so that shouldn't be an issue. > > Then you could use query_int to compare them the way you want. You > misunderstood how query_int is supposed to work. You index an array column > and then you can check it against a query_int just as you're currently > checking for overlap. Think of @@ as a more powerful version of the overlap > operator that can do complex logical expressions. > > The equivalent of > > where '{30,31,32,33,34,35,36,37,38,39,40}'::int[] && age > and '{2}'::int[] && gender > and '{1,2,4}'::int[] && orientation > > would then become: > > WHERE array[age,gender,orientation] @@ > '(30|31|32|33|34|35|36|37|38|39|40)&(2)&(1|2|4)' > > except you would have to change orientation and gender to not both have a > value of 2. > > You might consider doing the expression index a bit of overkill actually. > You might consider just storing a column "attributes" with an integer array > directly in the table. > > You would also want a table that lists the valid attributes to be sure not > to have any overlaps: > > 1 age 1 > 2 age 2 > ... > 101 gender male > 102 gender female > 103 orientation straight > 104 orientation gay > 105 orientation bi > 106 bodytype scrawny > ... > > > - Is there any hope for this structure? > > You'll have to test this carefully. I tried using GiST indexes for my > project and found that I couldn't load the data and build the GiST indexes > fast enough. You have to test the costs of building and maintaining this > index, especially since it has so many columns in it. > > But it looks like your queries are in trouble without it so hopefully it'll > be ok on the insert/update side for you. From pgsql-performance-owner@postgresql.org Mon Oct 11 13:03:28 2004 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 CBCFA32A080 for ; Tue, 5 Oct 2004 07:49:36 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 93944-02 for ; Tue, 5 Oct 2004 06:49:30 +0000 (GMT) Received: from vertigo.530collins.mel.nella.net.au (vertigo.530collins.mel.nella.net.au [202.76.191.244]) by svr1.postgresql.org (Postfix) with ESMTP id A014632A482 for ; Tue, 5 Oct 2004 07:49:27 +0100 (BST) Received: from capricorn ([202.61.206.7]) by vertigo.530collins.mel.nella.net.au (8.12.8/8.12.8) with SMTP id i956rN3r017717 for ; Tue, 5 Oct 2004 16:53:26 +1000 From: "Chris Hutchinson" To: Subject: EXPLAIN ANALYZE much slower than running query normally Date: Tue, 5 Oct 2004 16:49:26 +1000 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 (9.0.2910.0) X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 Importance: Normal X--MailScanner-Information: Please contact the ISP for more information X--MailScanner: Found to be clean X--MailScanner-SpamScore: sss X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/130 X-Sequence-Number: 8606 Running a trivial query in v7.4.2 (installed with fedora core2) using EXPLAIN ANALYZE is taking considerably longer than just running the query (2mins vs 6 secs). I was using this query to quickly compare a couple of systems after installing a faster disk. Is this sort of slowdown to be expected? Here's the query: ---------------------------------------- [chris@fedora tmp]$ time psql dbt << ENDSQL > select count(*) from etab; > ENDSQL count --------- 9646782 (1 row) real 0m6.532s user 0m0.005s sys 0m0.002s [chris@fedora tmp]$ time psql dbt << ENDSQL > explain analyze select count(*) from etab; > ENDSQL QUERY PLAN ---------------------------------------------------------------------------- --------------- ----------------------------- Aggregate (cost=182029.78..182029.78 rows=1 width=0) (actual time=112701.488..112701.493 rows=1 loops=1) -> Seq Scan on etab (cost=0.00..157912.82 rows=9646782 width=0) (actual time=0.053..578 59.120 rows=9646782 loops=1) Total runtime: 112701.862 ms (3 rows) real 1m52.716s user 0m0.003s sys 0m0.005s --------------------------------------- Thanks in advance for any clues. Chris Hutchinson From pgsql-performance-owner@postgresql.org Tue Oct 5 16:35:48 2004 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 6FFDD329F70 for ; Tue, 5 Oct 2004 16:35:46 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 50932-04 for ; Tue, 5 Oct 2004 15:35:36 +0000 (GMT) Received: from bayswater1.ymogen.net (host-154-240-27-217.pobox.net.uk [217.27.240.154]) by svr1.postgresql.org (Postfix) with ESMTP id 33790329CDD for ; Tue, 5 Oct 2004 16:35:35 +0100 (BST) Received: from solent (unknown [213.165.136.10]) by bayswater1.ymogen.net (Postfix) with ESMTP id A24ABA3405; Tue, 5 Oct 2004 16:35:28 +0100 (BST) Reply-To: From: "Matt Clark" To: "'Tatsuo Ishii'" Cc: , , , Subject: Re: Caching of Queries Date: Tue, 5 Oct 2004 16:35:28 +0100 Organization: Ymogen Ltd Message-ID: <012501c4aaf0$f12d0930$8300a8c0@solent> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook, Build 10.0.6626 Importance: Normal X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 In-Reply-To: <20041004.002823.85416338.t-ishii@sra.co.jp> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/41 X-Sequence-Number: 8517 > I don't know what you are exactly referring to in above URL=20 > when you are talking about "potential pitfalls of pooling".=20 > Please explain more. Sorry, I wasn't implying that pgpool doesn't deal with the issues, just that some people aren't necessarily aware of them up front. For instance, pgpool does an 'abort transaction' and a 'reset all' in lieu of a full reconnect (of course, since a full reconnect is exactly what we are trying to avoid). Is this is enough to guarantee that a given pooled connection behaves exactly as a non-pooled connection would from a client perspective? For instance, temporary tables are usually dropped at the end of a session, so a client (badly coded perhaps) that does not already use persistent connections might be confused when the sequence 'connect, create temp table foo ..., disconnect, connect, create temp table foo ...' results in the error 'Relation 'foo' already exists'. From pgsql-performance-owner@postgresql.org Tue Oct 5 17:22:00 2004 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 0675C329E12 for ; Tue, 5 Oct 2004 17:21:58 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 66659-03 for ; Tue, 5 Oct 2004 16:21:48 +0000 (GMT) Received: from snowwhite.lulu.com (mail.lulu.com [66.193.5.50]) by svr1.postgresql.org (Postfix) with ESMTP id 2956832A332 for ; Tue, 5 Oct 2004 17:21:47 +0100 (BST) Received: from [10.0.0.32] (meatwad.rdu.lulu.com [10.0.0.32]) (authenticated bits=0) by snowwhite.lulu.com (8.12.8/8.12.8) with ESMTP id i95GMTbO017772 (version=TLSv1/SSLv3 cipher=RC4-MD5 bits=128 verify=NO) for ; Tue, 5 Oct 2004 12:22:32 -0400 Message-ID: <4162CA14.6080206@lulu.com> Date: Tue, 05 Oct 2004 12:21:40 -0400 From: Bill Montgomery User-Agent: Mozilla Thunderbird 0.7.3 (X11/20040803) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Excessive context switching on SMP Xeons 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/42 X-Sequence-Number: 8518 All, I realize the excessive-context-switching-on-xeon issue has been discussed at length in the past, but I wanted to follow up and verify my conclusion from those discussions: On a 2-way or 4-way Xeon box, there is no way to avoid excessive (30,000-60,000 per second) context switches when using PostgreSQL 7.4.5 to query a data set small enough to fit into main memory under a significant load. I am experiencing said symptom on two different dual-Xeon boxes, both Dells with ServerWorks chipsets, running the latest RH9 and RHEL3 kernels, respectively. The databases are 90% read, 10% write, and are small enough to fit entirely into main memory, between pg shared buffers and kernel buffers. We recently invested in an solid-state storage device (http://www.superssd.com/products/ramsan-320/) to help write performance. Our entire pg data directory is stored on it. Regrettably (and in retrospect, unsurprisingly) we found that opening up the I/O bottleneck does little for write performance when the server is under load, due to the bottleneck created by excessive context switching. Is the only solution then to move to a different SMP architecture such as Itanium 2 or Opteron? If so, should we expect to see an additional benefit from running PostgreSQL on a 64-bit architecture, versus 32-bit, context switching aside? Alternatively, are there good 32-bit SMP architectures to consider other than Xeon, given the high cost of Itanium 2 and Opteron systems? More generally, how have others scaled "up" their PostgreSQL environments? We will eventually have to invent some "outward" scalability within the logic of our application (e.g. do read-only transactions against a pool of Slony-I subscribers), but in the short term we still have an urgent need to scale upward. Thoughts? General wisdom? Best Regards, Bill Montgomery From pgsql-performance-owner@postgresql.org Tue Oct 5 17:33:43 2004 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 66508329F35 for ; Tue, 5 Oct 2004 17:33:41 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 70476-05 for ; Tue, 5 Oct 2004 16:33:31 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id A1F52329F26 for ; Tue, 5 Oct 2004 17:33:30 +0100 (BST) 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 6461122 for pgsql-performance@postgresql.org; Tue, 05 Oct 2004 09:34:52 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Comparing user attributes with bitwise operators Date: Tue, 5 Oct 2004 09:32:38 -0700 User-Agent: KMail/1.6.2 References: <200409160141.37604.patrick@phpforhire.com> <87sm9eluc3.fsf@stark.xeocode.com> <200410050039.24230.patrick@phpforhire.com> In-Reply-To: <200410050039.24230.patrick@phpforhire.com> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410050932.38595.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/43 X-Sequence-Number: 8519 Patrick, First off, thanks for posting this solution! I love to see a new demo of The Power of Postgres(tm) and have been wondering about this particular problem since it came up on IRC. > The array method works quite nicely, especially for the > columns like "languages" and "seeking" that are multiple choice. However, > even though this method is fast, I still might opt for caching the results > because the "real world" search query involves a lot more and will be > executed non-stop. But to have it run this fast the first time certainly > helps. Now, for the bad news: you need to test having a large load of users updating their data. The drawback to GiST indexes is that they are low-concurrency, because the updating process needs to lock the whole index (this has been on our TODO list for about a decade, but it's a hard problem). -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Tue Oct 5 17:48:37 2004 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 EDEA132A332 for ; Tue, 5 Oct 2004 17:48:32 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 75505-04 for ; Tue, 5 Oct 2004 16:48:30 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id C19AC329FEA for ; Tue, 5 Oct 2004 17:48:29 +0100 (BST) 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 6461203 for pgsql-performance@postgresql.org; Tue, 05 Oct 2004 09:49:52 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Excessive context switching on SMP Xeons Date: Tue, 5 Oct 2004 09:47:36 -0700 User-Agent: KMail/1.6.2 References: <4162CA14.6080206@lulu.com> In-Reply-To: <4162CA14.6080206@lulu.com> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410050947.36174.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/44 X-Sequence-Number: 8520 Bill, > I realize the excessive-context-switching-on-xeon issue has been > discussed at length in the past, but I wanted to follow up and verify my > conclusion from those discussions: First off, the good news: Gavin Sherry and OSDL may have made some progress on this. We'll be testing as soon as OSDL gets the Scalable Test Platform running again. If you have the CS problem (which I don't think you do, see below) and a test box, I'd be thrilled to have you test it. > On a 2-way or 4-way Xeon box, there is no way to avoid excessive > (30,000-60,000 per second) context switches when using PostgreSQL 7.4.5 > to query a data set small enough to fit into main memory under a > significant load. Hmmm ... some clarification: 1) I don't really consider a CS of 30,000 to 60,000 on Xeon to be excessive. People demonstrating the problem on dual or quad Xeon reported CS levels of 150,000 or more. So you probably don't have this issue at all -- depending on the load, your level could be considered "normal". 2) The problem is not limited to Xeon, Linux, or x86 architecture. It has been demonstrated, for example, on 8-way Solaris machines. It's just worse (and thus more noticable) on Xeon. > I am experiencing said symptom on two different dual-Xeon boxes, both > Dells with ServerWorks chipsets, running the latest RH9 and RHEL3 > kernels, respectively. The databases are 90% read, 10% write, and are > small enough to fit entirely into main memory, between pg shared buffers > and kernel buffers. Ah. Well, you do have the worst possible architecture for PostgreSQL-SMP performance. The ServerWorks chipset is badly flawed (the company is now, I believe, bankrupt from recalled products) and Xeons have several performance issues on databases based on online tests. > We recently invested in an solid-state storage device > (http://www.superssd.com/products/ramsan-320/) to help write > performance. Our entire pg data directory is stored on it. Regrettably > (and in retrospect, unsurprisingly) we found that opening up the I/O > bottleneck does little for write performance when the server is under > load, due to the bottleneck created by excessive context switching. Well, if you're CPU-bound, improved I/O won't help you, no. > Is > the only solution then to move to a different SMP architecture such as > Itanium 2 or Opteron? If so, should we expect to see an additional > benefit from running PostgreSQL on a 64-bit architecture, versus 32-bit, > context switching aside? Your performance will almost certainly be better for a variety of reasons on Opteron/Itanium. However, I'm still not convinced that you have the CS bug. > Alternatively, are there good 32-bit SMP > architectures to consider other than Xeon, given the high cost of > Itanium 2 and Opteron systems? AthalonMP appears to be less suseptible to the CS bug than Xeon, and the effect of the bug is not as severe. However, a quad-Opteron box can be built for less than $6000; what's your standard for "expensive"? If you don't have that much money, then you may be stuck for options. > More generally, how have others scaled "up" their PostgreSQL > environments? We will eventually have to invent some "outward" > scalability within the logic of our application (e.g. do read-only > transactions against a pool of Slony-I subscribers), but in the short > term we still have an urgent need to scale upward. Thoughts? General > wisdom? As long as you're on x86, scaling outward is the way to go. If you want to continue to scale upwards, ask Andrew Sullivan about his experiences running PostgreSQL on big IBM boxes. But if you consider an quad-Opteron server expensive, I don't think that's an option for you. Overall, though, I'm not convinced that you have the CS bug and I think it's more likely that you have a few "bad queries" which are dragging down the whole system. Troubleshoot those and your CPU-bound problems may go away. -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Tue Oct 5 22:08:49 2004 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 784A4329E64 for ; Tue, 5 Oct 2004 22:08:47 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 55104-05 for ; Tue, 5 Oct 2004 21:08:35 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id D854732A0B4 for ; Tue, 5 Oct 2004 22:08:37 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i95L8Xc1056190 for ; Tue, 5 Oct 2004 21:08:34 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i95L8UIW056176 for pgsql-performance@postgresql.org; Tue, 5 Oct 2004 21:08:30 GMT From: Gaetano Mendola X-Newsgroups: comp.databases.postgresql.performance Subject: Re: Excessive context switching on SMP Xeons Date: Tue, 05 Oct 2004 23:08:23 +0200 Organization: PYRENET Midi-pyrenees Provider Lines: 117 Message-ID: <41630D47.2020704@bigfoot.com> References: <4162CA14.6080206@lulu.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Complaints-To: abuse@pyrenet.fr To: Bill Montgomery User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en In-Reply-To: <4162CA14.6080206@lulu.com> X-Enigmail-Version: 0.86.1.0 X-Enigmail-Supports: pgp-inline, pgp-mime To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/46 X-Sequence-Number: 8522 Bill Montgomery wrote: > All, > > I realize the excessive-context-switching-on-xeon issue has been > discussed at length in the past, but I wanted to follow up and verify my > conclusion from those discussions: > > On a 2-way or 4-way Xeon box, there is no way to avoid excessive > (30,000-60,000 per second) context switches when using PostgreSQL 7.4.5 > to query a data set small enough to fit into main memory under a > significant load. > > I am experiencing said symptom on two different dual-Xeon boxes, both > Dells with ServerWorks chipsets, running the latest RH9 and RHEL3 > kernels, respectively. The databases are 90% read, 10% write, and are > small enough to fit entirely into main memory, between pg shared buffers > and kernel buffers. > I don't know if my box is not loaded enough but I have a dual-Xeon box, by DELL with the HT enabled and I'm not experiencing this kind of CS problem, normaly hour CS is around 100000 per second. # cat /proc/version Linux version 2.4.9-e.24smp (bhcompile@porky.devel.redhat.com) (gcc version 2.96 20000731 (Red Hat Linux 7.2 2.96-118.7.2)) #1 SMP Tue May 27 16:07:39 EDT 2003 # cat /proc/cpuinfo processor : 0 vendor_id : GenuineIntel cpu family : 15 model : 2 model name : Intel(R) Xeon(TM) CPU 2.80GHz stepping : 7 cpu MHz : 2787.139 cache size : 512 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm bogomips : 5557.45 processor : 1 vendor_id : GenuineIntel cpu family : 15 model : 2 model name : Intel(R) Xeon(TM) CPU 2.80GHz stepping : 7 cpu MHz : 2787.139 cache size : 512 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm bogomips : 5570.56 processor : 2 vendor_id : GenuineIntel cpu family : 15 model : 2 model name : Intel(R) Xeon(TM) CPU 2.80GHz stepping : 7 cpu MHz : 2787.139 cache size : 512 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm bogomips : 5570.56 processor : 3 vendor_id : GenuineIntel cpu family : 15 model : 2 model name : Intel(R) Xeon(TM) CPU 2.80GHz stepping : 7 cpu MHz : 2787.139 cache size : 512 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm bogomips : 5570.56 Regards Gaetano Mendola From pgsql-performance-owner@postgresql.org Tue Oct 5 22:08:49 2004 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 44A6C329D9F for ; Tue, 5 Oct 2004 22:08:47 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 54878-06 for ; Tue, 5 Oct 2004 21:08:36 +0000 (GMT) Received: from snowwhite.lulu.com (mail.lulu.com [66.193.5.50]) by svr1.postgresql.org (Postfix) with ESMTP id C6D7032A082 for ; Tue, 5 Oct 2004 22:08:36 +0100 (BST) Received: from [10.0.0.32] (meatwad.rdu.lulu.com [10.0.0.32]) (authenticated bits=0) by snowwhite.lulu.com (8.12.8/8.12.8) with ESMTP id i95L9LbO022592 (version=TLSv1/SSLv3 cipher=RC4-MD5 bits=128 verify=NO) for ; Tue, 5 Oct 2004 17:09:21 -0400 Message-ID: <41630D50.3020308@lulu.com> Date: Tue, 05 Oct 2004 17:08:32 -0400 From: Bill Montgomery User-Agent: Mozilla Thunderbird 0.7.3 (X11/20040803) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Re: Excessive context switching on SMP Xeons References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> In-Reply-To: <200410050947.36174.josh@agliodbs.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/45 X-Sequence-Number: 8521 Thanks for the helpful response. Josh Berkus wrote: > First off, the good news: Gavin Sherry and OSDL may have made some > progress > >on this. We'll be testing as soon as OSDL gets the Scalable Test Platform >running again. If you have the CS problem (which I don't think you do, see >below) and a test box, I'd be thrilled to have you test it. > I'd be thrilled to test it too, if for no other reason that to determine whether what I'm experiencing really is the "CS problem". >1) I don't really consider a CS of 30,000 to 60,000 on Xeon to be excessive. >People demonstrating the problem on dual or quad Xeon reported CS levels of >150,000 or more. So you probably don't have this issue at all -- depending >on the load, your level could be considered "normal". > Fair enough. I never see nearly this much context switching on my dual Xeon boxes running dozens (sometimes hundreds) of concurrent apache processes, but I'll concede this could just be due to the more parallel nature of a bunch of independent apache workers. >>I am experiencing said symptom on two different dual-Xeon boxes, both >>Dells with ServerWorks chipsets, running the latest RH9 and RHEL3 >>kernels, respectively. The databases are 90% read, 10% write, and are >>small enough to fit entirely into main memory, between pg shared buffers >>and kernel buffers. >> > >Ah. Well, you do have the worst possible architecture for PostgreSQL-SMP >performance. The ServerWorks chipset is badly flawed (the company is now, I >believe, bankrupt from recalled products) and Xeons have several performance >issues on databases based on online tests. > Hence my desire for recommendations on alternate architectures ;-) >AthalonMP appears to be less suseptible to the CS bug than Xeon, and the >effect of the bug is not as severe. However, a quad-Opteron box can be >built for less than $6000; what's your standard for "expensive"? If you >don't have that much money, then you may be stuck for options. > Being a 24x7x365 shop, and these servers being mission critical, I require vendors that can offer 24x7 4-hour part replacement, like Dell or IBM. I haven't seen 4-way 64-bit boxes meeting that requirement for less than $20,000, and that's for a very minimally configured box. A suitably configured pair will likely end up costing $50,000 or more. I would like to avoid an unexpected expense of that size, unless there's no other good alternative. That said, I'm all ears for a cheaper alternative that meets my support and performance requirements. >Overall, though, I'm not convinced that you have the CS bug and I think it's >more likely that you have a few "bad queries" which are dragging down the >whole system. Troubleshoot those and your CPU-bound problems may go away. > You may be right, but to compare apples to apples, here's some vmstat output from a pgbench run: [billm@xxx billm]$ pgbench -i -s 20 pgbench [billm@xxx billm]$ pgbench -s 20 -t 500 -c 100 pgbench starting vacuum...end. transaction type: TPC-B (sort of) scaling factor: 20 number of clients: 100 number of transactions per client: 500 number of transactions actually processed: 50000/50000 tps = 369.717832 (including connections establishing) tps = 370.852058 (excluding connections establishing) and some of the vmstat output... [billm@poe billm]$ vmstat 1 procs memory swap io system cpu r b swpd free buff cache si so bi bo in cs us sy wa id 0 1 0 863108 220620 1571924 0 0 4 64 34 50 1 0 0 98 0 1 0 863092 220620 1571932 0 0 0 3144 171 2037 3 3 47 47 0 1 0 863084 220620 1571956 0 0 0 5840 202 3702 6 3 46 45 1 1 0 862656 220620 1572420 0 0 0 12948 631 42093 69 22 5 5 11 0 0 862188 220620 1572828 0 0 0 12644 531 41330 70 23 2 5 9 0 0 862020 220620 1573076 0 0 0 8396 457 28445 43 17 17 22 9 0 0 861620 220620 1573556 0 0 0 13564 726 44330 72 22 2 5 8 1 0 861248 220620 1573980 0 0 0 12564 660 43667 65 26 2 7 3 1 0 860704 220624 1574236 0 0 0 14588 646 41176 62 25 5 8 0 1 0 860440 220624 1574476 0 0 0 42184 865 31704 44 23 15 18 8 0 0 860320 220624 1574628 0 0 0 10796 403 19971 31 10 29 29 0 1 0 860040 220624 1574884 0 0 0 23588 654 36442 49 20 13 17 0 1 0 859984 220624 1574932 0 0 0 4940 229 3884 5 3 45 46 0 1 0 859940 220624 1575004 0 0 0 12140 355 13454 20 10 35 35 0 1 0 859904 220624 1575044 0 0 0 5044 218 6922 11 5 41 43 1 1 0 859868 220624 1575052 0 0 0 4808 199 2029 3 3 47 48 0 1 0 859720 220624 1575180 0 0 0 21596 485 18075 28 13 29 30 11 1 0 859372 220624 1575532 0 0 0 24520 609 41409 62 33 2 3 While pgbench does not generate quite as high a number of CS as our app, it is an apples-to-apples comparison, and rules out the possibility of poorly written queries in our app. Still, 40k CS/sec seems high to me. While pgbench is just a synthetic benchmark, and not necessarily the best benchmark, yada yada, 370 tps seems like pretty poor performance. I've benchmarked the IO subsystem at 70MB/s of random 8k writes, yet pgbench typically doesn't use more than 10MB/s of that bandwidth (a little more at checkpoints). So I guess the question is this: now that I've opened up the IO bottleneck that exists on most database servers, am I really truly CPU bound now, and not just suffering from poorly handled spinlocks on my Xeon/ServerWorks platform? If so, is the expense of a 64-bit system worth it, or is the price/performance for PostgreSQL still better on an alternative 32-bit platform, like AthlonMP? Best Regards, Bill Montgomery From pgsql-performance-owner@postgresql.org Tue Oct 5 23:37:25 2004 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 99C54329F2F for ; Tue, 5 Oct 2004 23:37:16 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 77413-05 for ; Tue, 5 Oct 2004 22:37:08 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id E658E329EF2 for ; Tue, 5 Oct 2004 23:37:09 +0100 (BST) 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 6462887 for pgsql-performance@postgresql.org; Tue, 05 Oct 2004 15:38:32 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Excessive context switching on SMP Xeons Date: Tue, 5 Oct 2004 15:38:51 -0700 User-Agent: KMail/1.6.2 References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <41630D50.3020308@lulu.com> In-Reply-To: <41630D50.3020308@lulu.com> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410051538.51664.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/47 X-Sequence-Number: 8523 Bill, > I'd be thrilled to test it too, if for no other reason that to determine > whether what I'm experiencing really is the "CS problem". Hmmm ... Gavin's patch is built against 8.0, and any version of the patch would require linux 2.6, probably 2.6.7 minimum. Can you test on that linux version? Do you have the resources to back-port Gavin's patch? > Fair enough. I never see nearly this much context switching on my dual > Xeon boxes running dozens (sometimes hundreds) of concurrent apache > processes, but I'll concede this could just be due to the more parallel > nature of a bunch of independent apache workers. Certainly could be. Heavy CSes only happen when you have a number of long-running processes with contention for RAM in my experience. If Apache is dispatching thing quickly enough, they'd never arise. > Hence my desire for recommendations on alternate architectures ;-) Well, you could certainly stay on Xeon if there's better support availability. Just get off Dell *650's. > Being a 24x7x365 shop, and these servers being mission critical, I > require vendors that can offer 24x7 4-hour part replacement, like Dell > or IBM. I haven't seen 4-way 64-bit boxes meeting that requirement for > less than $20,000, and that's for a very minimally configured box. A > suitably configured pair will likely end up costing $50,000 or more. I > would like to avoid an unexpected expense of that size, unless there's > no other good alternative. That said, I'm all ears for a cheaper > alternative that meets my support and performance requirements. No, you're going to pay through the nose for that support level. It's how things work. > tps = 369.717832 (including connections establishing) > tps = 370.852058 (excluding connections establishing) Doesn't seem too bad to me. Have anything to compare it to? What's in your postgresql.conf? --Josh From pgsql-performance-owner@postgresql.org Tue Oct 5 23:56:19 2004 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 C023632A626 for ; Tue, 5 Oct 2004 23:55:36 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 82753-05 for ; Tue, 5 Oct 2004 22:55:32 +0000 (GMT) Received: from pw6tiger.de (pw6tiger.de [217.160.140.116]) by svr1.postgresql.org (Postfix) with SMTP id EFA5A32A130 for ; Tue, 5 Oct 2004 23:55:32 +0100 (BST) Received: (qmail 29414 invoked by uid 505); 5 Oct 2004 22:55:29 -0000 Received: from vygen@gmx.de by planwerk6 by uid 89 with qmail-scanner-1.14 (virus scan: Clear:. Processed in 0.390171 secs); 05 Oct 2004 22:55:29 -0000 Received: from p508cd17a.dip0.t-ipconnect.de (HELO ?192.168.1.102?) (vygen@planwerk6.de@80.140.209.122) by pw6tiger.de with SMTP; 5 Oct 2004 22:55:28 -0000 From: Janning Vygen To: pgsql-performance@postgresql.org Subject: slow rule on update Date: Wed, 6 Oct 2004 00:55:04 +0200 User-Agent: KMail/1.7 MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200410060055.06211.vygen@gmx.de> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/48 X-Sequence-Number: 8524 Hi, (pg_version 7.4.2, i do run vacuum analyze on the whole database frequently and just before executing statements below) i dont know if anyone can help me because i dont know really where the problem is, but i try. If any further information is needed i'll be glad to send. my real rule much longer (more calculation instead of "+ 1") but this shortcut has the same disadvantages in performance: CREATE RULE ru_sp_update AS ON UPDATE TO Spiele DO UPDATE punktecache SET pc_punkte = pc_punkte + 1 FROM Spieletipps AS stip NATURAL JOIN tippspieltage2spiele AS tspt2sp WHERE punktecache.tr_kurzname = stip.tr_kurzname AND punktecache.mg_name = stip.mg_name AND punktecache.tspt_name = tspt2sp.tspt_name AND stip.sp_id = OLD.sp_id ; punktecache is a materialized view which should be updated by this rule # \d punktecache Table "public.punktecache" Column | Type | Modifiers -------------+----------+----------- tr_kurzname | text | not null mg_name | text | not null tspt_name | text | not null pc_punkte | smallint | not null Indexes: "pk_punktecache" primary key, btree (tr_kurzname, mg_name, tspt_name) Foreign-key constraints: "fk_mitglieder" FOREIGN KEY (tr_kurzname, mg_name) REFERENCES mitglieder(tr_kurzname, mg_name) ON UPDATE CASCADE ON DELETE CASCADE "fk_tippspieltage" FOREIGN KEY (tr_kurzname, tspt_name) REFERENCES tippspieltage(tr_kurzname, tspt_name) ON UPDATE CASCADE ON DELETE CASCADE my update statement: explain analyze UPDATE spiele SET sp_heimtore = spup.spup_heimtore, sp_gasttore = spup.spup_gasttore, sp_abpfiff = spup.spup_abpfiff FROM spieleupdates AS spup WHERE spiele.sp_id = spup.sp_id; and output from explain [did i post explain's output right? i just copied it, but i wonder if there is a more pretty print like method to post explain's output?] Nested Loop (cost=201.85..126524.78 rows=1 width=45) (actual time=349.694..290491.442 rows=100990 loops=1) -> Nested Loop (cost=201.85..126518.97 rows=1 width=57) (actual time=349.623..288222.145 rows=100990 loops=1) -> Hash Join (cost=201.85..103166.61 rows=4095 width=64) (actual time=131.376..8890.220 rows=102472 loops=1) Hash Cond: (("outer".tspt_name = "inner".tspt_name) AND ("outer".tr_kurzname = "inner".tr_kurzname)) -> Seq Scan on punktecache (cost=0.00..40970.20 rows=2065120 width=45) (actual time=0.054..4356.321 rows=2065120 loops=1) -> Hash (cost=178.16..178.16 rows=4738 width=35) (actual time=102.259..102.259 rows=0 loops=1) -> Nested Loop (cost=0.00..178.16 rows=4738 width=35) (actual time=17.262..88.076 rows=10519 loops=1) -> Seq Scan on spieleupdates spup (cost=0.00..0.00 rows=1 width=4) (actual time=0.015..0.024 rows=1 loops=1) -> Index Scan using ix_tspt2sp_fk_spiele on tippspieltage2spiele tspt2sp (cost=0.00..118.95 rows=4737 width=31) (actual time=17.223..69.486 rows=10519 loops=1) Index Cond: ("outer".sp_id = tspt2sp.sp_id) -> Index Scan using pk_spieletipps on spieletipps stip (cost=0.00..5.69 rows=1 width=25) (actual time=2.715..2.717 rows=1 loops=102472) Index Cond: (("outer".tr_kurzname = stip.tr_kurzname) AND ("outer".mg_name = stip.mg_name) AND ("outer".sp_id = stip.sp_id)) -> Index Scan using pk_spiele on spiele (cost=0.00..5.78 rows=1 width=4) (actual time=0.012..0.014 rows=1 loops=100990) Index Cond: (spiele.sp_id = "outer".sp_id) Total runtime: 537319.321 ms Can this be made any faster? Can you give me a hint where to start research? My guess is that the update statement inside the rule doesnt really uses the index on punktecache, but i dont know why and i dont know how to change it. Any hint or help is is very appreciated. kind regards janning From pgsql-performance-owner@postgresql.org Wed Oct 6 01:42:24 2004 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 0BA7A329E93 for ; Wed, 6 Oct 2004 01:42:20 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 06983-03 for ; Wed, 6 Oct 2004 00:42:07 +0000 (GMT) Received: from khan.acc.umu.se (khan.acc.umu.se [130.239.18.139]) by svr1.postgresql.org (Postfix) with ESMTP id 5BB42329D41 for ; Wed, 6 Oct 2004 01:42:09 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by amavisd-new (Postfix) with ESMTP id 22D0DD223 for ; Wed, 6 Oct 2004 02:42:08 +0200 (MEST) Received: from shaka.acc.umu.se (shaka.acc.umu.se [130.239.18.148]) by khan.acc.umu.se (Postfix) with ESMTP id 79363D2E2 for ; Wed, 6 Oct 2004 02:42:05 +0200 (MEST) Received: by shaka.acc.umu.se (Postfix, from userid 23132) id 385C2F; Wed, 6 Oct 2004 02:42:05 +0200 (MEST) Date: Wed, 6 Oct 2004 02:42:04 +0200 From: Nichlas =?iso-8859-1?Q?L=F6fdahl?= To: pgsql-performance@postgresql.org Subject: Planner picks the wrong plan? Message-ID: <20041006004204.GA13074@shaka.acc.umu.se> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.4.1i X-Virus-Scanned: by amavisd-new at acc.umu.se X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/49 X-Sequence-Number: 8525 Hello! I'm using Postgres 7.4.5, sort_mem is 8192. Tables analyzed / vacuumed. Here's a function I'm using to get an age from the user's birthday: agey(date) -> SELECT date_part('year', age($1::timestamp)) The problem is, why do the plans differ so much between Q1 & Q3 below? Something with age() being a non-IMMUTABLE function? Q1: explain analyze SELECT al.pid, al.owner, al.title, al.front, al.created_at, al.n_images, u.username as owner_str, u.image as owner_image, u.puid as owner_puid FROM albums al , users u WHERE u.uid = al.owner AND al.security='a' AND al.n_images > 0 AND date_part('year', age(u.born)) > 17 AND date_part('year', age(u.born)) < 20 AND city = 1 ORDER BY al.id DESC LIMIT 9; QUERY PLAN ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Limit (cost=5700.61..5700.63 rows=9 width=183) (actual time=564.291..564.299 rows=9 loops=1) -> Sort (cost=5700.61..5700.82 rows=83 width=183) (actual time=564.289..564.291 rows=9 loops=1) Sort Key: al.id -> Nested Loop (cost=0.00..5697.97 rows=83 width=183) (actual time=30.029..526.211 rows=4510 loops=1) -> Seq Scan on users u (cost=0.00..5311.05 rows=86 width=86) (actual time=5.416..421.264 rows=3021 loops=1) Filter: ((date_part('year'::text, age((('now'::text)::date)::timestamp with time zone, (born)::timestamp with time zone)) > 17::double precision) AND (date_part('year'::text, age((('now'::text)::date)::timestamp with time zone, (born)::timestamp with time zone)) < 20::double precision) AND (city = 1)) -> Index Scan using albums_owner_key on albums al (cost=0.00..4.47 rows=2 width=101) (actual time=0.014..0.025 rows=1 loops=3021) Index Cond: ("outer".uid = al."owner") Filter: (("security" = 'a'::bpchar) AND (n_images > 0)) Total runtime: 565.120 ms (10 rows) Result when removing the second age-check (AND date_part('year', age(u.born)) < 20): Q2: explain analyze SELECT al.pid, al.owner, al.title, al.front, al.created_at, al.n_images, u.username as owner_str, u.image as owner_image, u.puid as owner_puid FROM albums al, users u WHERE u.uid = al.owner AND al.security='a' AND al.n_images > 0 AND date_part('year', age(u.born)) > 17 AND city = 1 ORDER BY al.id DESC LIMIT 9; --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Limit (cost=0.00..140.95 rows=9 width=183) (actual time=0.217..2.474 rows=9 loops=1) -> Nested Loop (cost=0.00..86200.99 rows=5504 width=183) (actual time=0.216..2.464 rows=9 loops=1) -> Index Scan Backward using albums_id_key on albums al (cost=0.00..2173.32 rows=27610 width=101) (actual time=0.086..1.080 rows=40 loops=1) Filter: (("security" = 'a'::bpchar) AND (n_images > 0)) -> Index Scan using users_pkey on users u (cost=0.00..3.03 rows=1 width=86) (actual time=0.031..0.031 rows=0 loops=40) Index Cond: (u.uid = "outer"."owner") Filter: ((date_part('year'::text, age((('now'::text)::date)::timestamp with time zone, (born)::timestamp with time zone)) > 17::double precision) AND (city = 1)) Total runtime: 2.611 ms (8 rows) Trying another approach: adding a separate "stale" age-column to the users-table: alter table users add column age smallint; update users set age=date_part('year'::text, age((('now'::text)::date)::timestamp with time zone, (born)::timestamp with time zone)); analyze users; Result with separate column: Q3: explain analyze SELECT al.pid, al.owner, al.title, al.front, al.created_at, al.n_images, u.username as owner_str, u.image as owner_image, u.puid as owner_puid FROM albums al , users u WHERE u.uid = al.owner AND al.security='a' AND al.n_images > 0 AND age > 17 AND age < 20 AND city = 1 ORDER BY al.id DESC LIMIT 9; -------------------------------------------------------------------------------------------------------------------------------------------------------- Limit (cost=0.00..263.40 rows=9 width=183) (actual time=0.165..2.832 rows=9 loops=1) -> Nested Loop (cost=0.00..85925.69 rows=2936 width=183) (actual time=0.163..2.825 rows=9 loops=1) -> Index Scan Backward using albums_id_key on albums al (cost=0.00..2173.32 rows=27610 width=101) (actual time=0.043..1.528 rows=56 loops=1) Filter: (("security" = 'a'::bpchar) AND (n_images > 0)) -> Index Scan using users_pkey on users u (cost=0.00..3.02 rows=1 width=86) (actual time=0.020..0.020 rows=0 loops=56) Index Cond: (u.uid = "outer"."owner") Filter: ((age > 17) AND (age < 20) AND (city = 1)) Total runtime: 2.973 ms (8 rows) My question is, why doesn't the planner pick the same plan for Q1 & Q3? /Nichlas From pgsql-performance-owner@postgresql.org Wed Oct 6 04:24:52 2004 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 93992329FDA for ; Wed, 6 Oct 2004 04:24:51 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 40314-09 for ; Wed, 6 Oct 2004 03:24:49 +0000 (GMT) Received: from roue.portalpotty.net (roue.portalpotty.net [69.44.62.206]) by svr1.postgresql.org (Postfix) with ESMTP id 03764329FD7 for ; Wed, 6 Oct 2004 04:24:48 +0100 (BST) Received: from max by roue.portalpotty.net with local (Exim 4.34) id 1CF2Q2-0001sB-A4 for pgsql-performance@postgresql.org; Tue, 05 Oct 2004 20:24:46 -0700 Date: Tue, 5 Oct 2004 23:24:46 -0400 From: Max Baker To: pgsql-performance@postgresql.org Subject: test post Message-ID: <20041006032446.GB7136@warped.org> Mail-Followup-To: pgsql-performance@postgresql.org Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.4.1i X-AntiAbuse: This header was added to track abuse, please include it with any abuse report X-AntiAbuse: Primary Hostname - roue.portalpotty.net X-AntiAbuse: Original Domain - postgresql.org X-AntiAbuse: Originator/Caller UID/GID - [514 32003] / [47 12] X-AntiAbuse: Sender Address Domain - roue.portalpotty.net X-Source: /usr/bin/mutt X-Source-Args: mutt X-Source-Dir: /home/max X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/50 X-Sequence-Number: 8526 please ignore if this goes through. They've been bouncing and I'm trying to find out why. -m From pgsql-performance-owner@postgresql.org Wed Oct 6 04:30:17 2004 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 BDB5632A00F for ; Wed, 6 Oct 2004 04:30:15 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 43204-08 for ; Wed, 6 Oct 2004 03:30:06 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 7E8CC32A003 for ; Wed, 6 Oct 2004 04:30:04 +0100 (BST) 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 i963Turd020688; Tue, 5 Oct 2004 23:29:57 -0400 (EDT) To: Nichlas =?iso-8859-1?Q?L=F6fdahl?= Cc: pgsql-performance@postgresql.org Subject: Re: Planner picks the wrong plan? In-reply-to: <20041006004204.GA13074@shaka.acc.umu.se> References: <20041006004204.GA13074@shaka.acc.umu.se> Comments: In-reply-to Nichlas =?iso-8859-1?Q?L=F6fdahl?= message dated "Wed, 06 Oct 2004 02:42:04 +0200" Date: Tue, 05 Oct 2004 23:29:56 -0400 Message-ID: <20687.1097033396@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/51 X-Sequence-Number: 8527 Nichlas =?iso-8859-1?Q?L=F6fdahl?= writes: > My question is, why doesn't the planner pick the same plan for Q1 & Q3? I think it's mostly that after you've added and ANALYZEd the "age" column, the planner has a pretty good idea of how many rows will pass the "age > 17 AND age < 20" condition. It can't do very much with the equivalent condition in the original form, though, and in fact ends up drastically underestimating the number of matching rows (86 vs reality of 3021). That leads directly to a bad plan choice :-( regards, tom lane From pgsql-performance-owner@postgresql.org Wed Oct 6 05:50:18 2004 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 3C601329FA5 for ; Wed, 6 Oct 2004 05:50:17 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 64517-08 for ; Wed, 6 Oct 2004 04:50:07 +0000 (GMT) Received: from wren.rentec.com (wren.rentec.com [65.213.84.9]) by svr1.postgresql.org (Postfix) with ESMTP id 53AFB329FA0 for ; Wed, 6 Oct 2004 05:50:05 +0100 (BST) Received: from [172.16.160.107] (stange-dhcp1.rentec.com [172.16.160.107]) by wren.rentec.com (8.12.9/8.12.1) with ESMTP id i964nueP010922; Wed, 6 Oct 2004 00:49:57 -0400 (EDT) Message-ID: <41636D8E.1090403@rentec.com> Date: Tue, 05 Oct 2004 23:59:10 -0400 From: Alan Stange User-Agent: Mozilla Thunderbird 0.7 (X11/20040615) X-Accept-Language: en-us, en MIME-Version: 1.0 To: josh@agliodbs.com Cc: pgsql-performance@postgresql.org Subject: Re: Excessive context switching on SMP Xeons References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <41630D50.3020308@lulu.com> <200410051538.51664.josh@agliodbs.com> In-Reply-To: <200410051538.51664.josh@agliodbs.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/52 X-Sequence-Number: 8528 A few quick random observations on the Xeon v. Opteron comparison: - running a dual Xeon with hyperthreading turned on really isn't the same as having a quad cpu system. I haven't seen postgresql specific benchmarks, but the general case has been that HT is a benefit in a few particular work loads but with no benefit in general. - We're running postgresql 8 (in production!) on a dual Opteron 250, Linux 2.6, 8GB memory, 1.7TB of attached fiber channel disk, etc. This machine is fast. A dual 2.8 Ghz Xeon with 512K caches (with or without HT enabled) simlpy won't be in the same performance league as this dual Opteron system (assuming identical disk systems, etc). We run a Linux 2.6 kernel because it scales under load so much better than the 2.4 kernels. The units we're using (and we have a lot of them) are SunFire v20z. You can get a dualie Opteron 250 for $7K with 4GB memory from Sun. My personal experience with this setup in a mission critical config is to not depend on 4 hour spare parts, but to spend the money and install the spare in the rack. Naturally, one can go cheaper with slower cpus, different vendors, etc. I don't care to go into the whole debate of Xeon v. Opteron here. We also have a lot of dual Xeon systems. In every comparison I've done with our codes, the dual Opteron clearly outperforms the dual Xeon, when running on one and both cpus. -- Alan Josh Berkus wrote: >Bill, > > > >>I'd be thrilled to test it too, if for no other reason that to determine >>whether what I'm experiencing really is the "CS problem". >> >> > >Hmmm ... Gavin's patch is built against 8.0, and any version of the patch >would require linux 2.6, probably 2.6.7 minimum. Can you test on that linux >version? Do you have the resources to back-port Gavin's patch? > > > >>Fair enough. I never see nearly this much context switching on my dual >>Xeon boxes running dozens (sometimes hundreds) of concurrent apache >>processes, but I'll concede this could just be due to the more parallel >>nature of a bunch of independent apache workers. >> >> > >Certainly could be. Heavy CSes only happen when you have a number of >long-running processes with contention for RAM in my experience. If Apache >is dispatching thing quickly enough, they'd never arise. > > > >>Hence my desire for recommendations on alternate architectures ;-) >> >> > >Well, you could certainly stay on Xeon if there's better support availability. >Just get off Dell *650's. > > > >>Being a 24x7x365 shop, and these servers being mission critical, I >>require vendors that can offer 24x7 4-hour part replacement, like Dell >>or IBM. I haven't seen 4-way 64-bit boxes meeting that requirement for >>less than $20,000, and that's for a very minimally configured box. A >>suitably configured pair will likely end up costing $50,000 or more. I >>would like to avoid an unexpected expense of that size, unless there's >>no other good alternative. That said, I'm all ears for a cheaper >>alternative that meets my support and performance requirements. >> >> > >No, you're going to pay through the nose for that support level. It's how >things work. > > > >>tps = 369.717832 (including connections establishing) >>tps = 370.852058 (excluding connections establishing) >> >> > >Doesn't seem too bad to me. Have anything to compare it to? > >What's in your postgresql.conf? > >--Josh > >---------------------------(end of broadcast)--------------------------- >TIP 5: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faqs/FAQ.html > > From pgsql-performance-owner@postgresql.org Mon Oct 11 13:03:40 2004 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 AA8BD32A05D for ; Wed, 6 Oct 2004 08:31:08 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 98677-02 for ; Wed, 6 Oct 2004 07:30:57 +0000 (GMT) Received: from www.hanseo.hs.kr (unknown [210.90.30.1]) by svr1.postgresql.org (Postfix) with ESMTP id 4FAB232A256 for ; Wed, 6 Oct 2004 08:30:56 +0100 (BST) Received: from [192.168.123.8] ([211.238.40.5]) (authenticated bits=0) by www.hanseo.hs.kr (8.12.9/8.12.9) with ESMTP id i967Ud0X003214 for ; Wed, 6 Oct 2004 16:30:44 +0900 Message-ID: <41639F3C.4050103@siche.net> Date: Wed, 06 Oct 2004 16:31:08 +0900 From: HyunSung Jang User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: ko-kr, ko, en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: why my query is not using index?? 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/131 X-Sequence-Number: 8607 postgres=# explain ANALYZE select * from test where today < '2004-01-01'; QUERY PLAN ---------------------------------------------------------------------------------------------------- Seq Scan on test (cost=0.00..19.51 rows=334 width=44) (actual time=0.545..2.429 rows=721 loops=1) Filter: (today < '2004-01-01 00:00:00'::timestamp without time zone) Total runtime: 3.072 ms (3 rows) postgres=# explain ANALYZE select * from test where today > '2003-01-01' and today < '2004-01-01'; QUERY PLAN ----------------------------------------------------------------------------------------------------------------------------------------------- Index Scan using idx_today on test (cost=0.00..18.89 rows=6 width=44) (actual time=0.055..1.098 rows=365 loops=1) Index Cond: ((today > '2003-01-01 00:00:00'::timestamp without time zone) AND (today < '2004-01-01 00:00:00'::timestamp without time zone)) Total runtime: 1.471 ms (3 rows) hello I was expected 1st query should using index, but it doesn't 2nd query doing perfect as you see. can you explain to me why it's not doing that i expected?? now I'm currently using postgresql 8.0pre3 on linux /hyunsung jang. From pgsql-performance-owner@postgresql.org Wed Oct 6 08:44:26 2004 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 AA6A132A07F for ; Wed, 6 Oct 2004 08:44:23 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 01996-02 for ; Wed, 6 Oct 2004 07:44:12 +0000 (GMT) Received: from stark.xeocode.com (gsstark.mtl.istop.com [66.11.160.162]) by svr1.postgresql.org (Postfix) with ESMTP id E1D6B32A3F6 for ; Wed, 6 Oct 2004 08:44:13 +0100 (BST) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1CF6Sw-0004sI-00; Wed, 06 Oct 2004 03:44:02 -0400 To: Alan Stange Cc: josh@agliodbs.com, pgsql-performance@postgresql.org Subject: Re: Excessive context switching on SMP Xeons References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <41630D50.3020308@lulu.com> <200410051538.51664.josh@agliodbs.com> <41636D8E.1090403@rentec.com> In-Reply-To: <41636D8E.1090403@rentec.com> From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 06 Oct 2004 03:44:02 -0400 Message-ID: <87acv0z3y5.fsf@stark.xeocode.com> Lines: 29 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/53 X-Sequence-Number: 8529 Alan Stange writes: > A few quick random observations on the Xeon v. Opteron comparison: > > - running a dual Xeon with hyperthreading turned on really isn't the same as > having a quad cpu system. I haven't seen postgresql specific benchmarks, but > the general case has been that HT is a benefit in a few particular work > loads but with no benefit in general. Part of the FUD with hyperthreading did have a kernel of truth that lied in older kernels' schedulers. For example with Linux until recently the kernel can easily end up scheduling two processes on the two virtual processors of one single physical processor, leaving the other physical processor totally idle. With modern kernels' schedulers I would expect hyperthreading to live up to its billing of adding 10% to 20% performance. Ie., a dual Xeon machine with hyperthreading won't be as fast as four processors, but it should be 10-20% faster than a dual Xeon without hyperthreading. As with all things that will only help if you're bound by the right limited resource to begin with. If you're I/O bound it isn't going to help. I would expect Postgres with its heavy demand on memory bandwidth and shared memory could potentially benefit more than usual from being able to context switch during pipeline stalls. -- greg From pgsql-performance-owner@postgresql.org Wed Oct 6 10:29:28 2004 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 4B882329F04 for ; Wed, 6 Oct 2004 10:29:27 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 30492-01 for ; Wed, 6 Oct 2004 09:29:15 +0000 (GMT) Received: from mail.hamburg.cityline.net (ns.mcs-hh.de [194.77.146.129]) by svr1.postgresql.org (Postfix) with ESMTP id 8863532A188 for ; Wed, 6 Oct 2004 10:29:15 +0100 (BST) Received: from [212.1.56.10] (helo=aconitin.toxine.lan) by mail.hamburg.cityline.net with asmtp (TLSv1:RC4-MD5:128) (Exim 4.32) id 1CF86j-0004rG-Si for pgsql-performance@postgresql.org; Wed, 06 Oct 2004 11:29:13 +0200 From: Ole Langbehn Organization: freiheit.com To: pgsql-performance@postgresql.org Subject: sequential scan on select distinct Date: Wed, 6 Oct 2004 11:30:58 +0200 User-Agent: KMail/1.7 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Message-Id: <200410061130.58625.ole@freiheit.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/54 X-Sequence-Number: 8530 Hi, I'm using Postgres 7.4.5. Tables are analyzed & vacuumed. I am wondering why postgresql never uses an index on queries of the type 'select distinct ...' while e.g. mysql uses the index on the same query. See the following explains: postgresql: explain analyze select distinct "land" from "customer_dim"; ---------------------------------------------------------------------------= ------------------------------------------------------------+ QUERY PLAN = | ---------------------------------------------------------------------------= ------------------------------------------------------------+ Unique (cost=3D417261.85..430263.66 rows=3D18 width=3D15) (actual time=3D= 45875.235..67204.694 rows=3D103 loops=3D1) | -> Sort (cost=3D417261.85..423762.75 rows=3D2600362 width=3D15) (actua= l time=3D45875.226..54114.473 rows=3D2600362 loops=3D1) | Sort Key: land = | -> Seq Scan on customer_dim (cost=3D0.00..84699.62 rows=3D260036= 2 width=3D15) (actual time=3D0.048..10733.227 rows=3D2600362 loops=3D1) | Total runtime: 67246.465 ms = | ---------------------------------------------------------------------------= ------------------------------------------------------------+ mysql: explain select DISTINCT `customer_dim`.`land` from `customer_dim`; --------------+-------+---------------+---------------+---------+--------+-= --------+-------------+ table | type | possible_keys | key | key_len | ref | = rows | Extra | --------------+-------+---------------+---------------+---------+--------+-= --------+-------------+ customer_dim | index | [NULL] | IDX_cstd_land | 81 | [NULL] | = 2600362 | Using index | --------------+-------+---------------+---------------+---------+--------+-= --------+-------------+ 1 row in result (first row: 8 msec; total: 9 msec) The result set contains 103 rows (but i get this behavior with every query = of this kind). My tables consist of at least a million rows. The indexes on the column 'land' are standard indexes, so in case of postgresql, it's a btree-index. I've tried to change the index type, but to= no avail. So, why doesn't postgresql use the index, and (how) could i persuade postgr= esql to use an index for this type of query? TiA --=20 Ole Langbehn freiheit.com technologies gmbh Theodorstr. 42-90 / 22761 Hamburg, Germany fon =A0 =A0 =A0 +49 (0)40 / 890584-0 fax =A0 =A0 =A0 +49 (0)40 / 890584-20 Freie Software durch B=FCcherkauf f=F6rdern | http://bookzilla.de/ From pgsql-benchmarks-owner@postgresql.org Wed Oct 6 11:15:49 2004 X-Original-To: pgsql-benchmarks-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 34DD6329D7F; Wed, 6 Oct 2004 11:15:46 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 40870-09; Wed, 6 Oct 2004 10:15:33 +0000 (GMT) Received: from mailer.unicite.fr.netcentrex.net (unknown [62.161.167.249]) by svr1.postgresql.org (Postfix) with ESMTP id 5B0F4329CDD; Wed, 6 Oct 2004 11:15:33 +0100 (BST) Received: from [192.168.101.228] ([192.168.101.228]) by mailer.unicite.fr.netcentrex.net with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2657.72) id 42CX026B; Wed, 6 Oct 2004 12:15:32 +0200 Message-ID: <4163C5C3.3030304@fr.netcentrex.net> Date: Wed, 06 Oct 2004 12:15:31 +0200 From: =?ISO-8859-1?Q?=22Alban_M=E9dici_=28NetCentrex=29=22?= Organization: NetCentrex User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-benchmarks@postgresql.org, pgsql-performance@postgresql.org Subject: stats on cursor and query execution troubleshooting Content-Type: multipart/alternative; boundary="------------040806000209030206000904" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.5 tagged_above=0.0 required=5.0 tests=HTML_40_50, HTML_MESSAGE X-Spam-Level: X-Archive-Number: 200410/1 X-Sequence-Number: 18 This is a multi-part message in MIME format. --------------040806000209030206000904 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: quoted-printable Hi, I'm looking for the statistic of memory, CPU, filesystem access while=20 executing some regular SQL query, and I want to compare them to same kind of results while executing a cursor function. The stat collector give me good results (sequencial scans , acceded=20 tuple .....) for regular query but nor for cursor (as explain in the=20 documentation) For more results, i have activated some log level in the postgresql.conf : show_query_stats =3D true But I have some trouble in the results interpretation such as : ! system usage stats: ! 2.776053 elapsed 1.880000 user 0.520000 system sec ! [1.910000 user 0.540000 sys total] ! 0/0 [0/0] filesystem blocks in/out ! 5/1 [319/148] page faults/reclaims, 0 [0] swaps ! 0 [0] signals rcvd, 0/0 [0/0] messages rcvd/sent ! 0/0 [0/0] voluntary/involuntary context switches ! postgres usage stats: ! Shared blocks: 3877 read, 0 written, buffer hit rate= =3D 0.00% ! Local blocks: 0 read, 0 written, buffer hit rate= =3D 0.00% ! Direct blocks: 0 read, 0 written Here is result done after fetching ALL a row with 178282 records in the=20 table. looking at the i/o stat of linux I saw a filesystem access while=20 executing this request but not in the previous log !!! ! 0/0 [0/0] filesystem blocks in/out I'm running postgresql 7.2.4 under redhat 7.2 Does am i wrong in my interpretation ? Does any newest postgresql version could told me execution paln for a=20 fetch AND better stats ? thx Ps: please excuse my poor english --=20 Alban M=E9dici R&D software engineer ------------------------------ you can contact me @ : http://www.netcentrex.net ------------------------------ --------------040806000209030206000904 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Hi,

I'm looking for the statistic of memory,  CPU,  filesystem access while executing some regular SQL query,  and I want to compare them to
same kind of results while executing a cursor function.

The stat collector give me good results (sequencial scans ,  acceded tuple .....)  for regular query but nor for cursor (as explain in the documentation)

For more results, i have activated some log level in the postgresql.conf :
    show_query_stats = true

But I have some trouble in the results interpretation  such as :

! system usage stats:
!       2.776053 elapsed 1.880000 user 0.520000 system sec
!       [1.910000 user 0.540000 sys total]
!       0/0 [0/0] filesystem blocks in/out
!       5/1 [319/148] page faults/reclaims, 0 [0] swaps
!       0 [0] signals rcvd, 0/0 [0/0] messages rcvd/sent
!       0/0 [0/0] voluntary/involuntary context switches
! postgres usage stats:
!       Shared blocks:       3877 read,          0 written, buffer hit rate = 0.00%
!       Local  blocks:          0 read,          0 written, buffer hit rate = 0.00%
!       Direct blocks:          0 read,          0 written

Here is  result done after fetching ALL a row with 178282 records in the table.
looking at the i/o stat of linux I saw a filesystem access while executing this request but not in the previous log !!!
!       0/0 [0/0] filesystem blocks in/out
I'm running postgresql 7.2.4  under redhat 7.2

Does am i wrong in my interpretation ?

Does any newest postgresql version could told me execution paln for a fetch AND better stats ?

thx

Ps: please excuse my poor english

-- 
Alban Médici
R&D software engineer
------------------------------
you can contact me @ :
http://www.netcentrex.net
------------------------------
--------------040806000209030206000904-- From pgsql-performance-owner@postgresql.org Wed Oct 6 11:16:50 2004 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 CC07432A3A5 for ; Wed, 6 Oct 2004 11:16:39 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 41331-09 for ; Wed, 6 Oct 2004 10:16:31 +0000 (GMT) Received: from boutiquenumerique.com (gailleton-2-82-67-9-10.fbx.proxad.net [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 61CA332A2ED for ; Wed, 6 Oct 2004 11:16:21 +0100 (BST) Received: (qmail 3818 invoked from network); 6 Oct 2004 12:16:46 +0200 Received: from unknown (HELO musicbox) (192.168.0.2) by gailleton-2-82-67-9-10.fbx.proxad.net with SMTP; 6 Oct 2004 12:16:46 +0200 To: "Ole Langbehn" , pgsql-performance@postgresql.org Subject: Re: sequential scan on select distinct References: <200410061130.58625.ole@freiheit.com> Message-ID: Date: Wed, 06 Oct 2004 12:19:24 +0200 From: =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= 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: <200410061130.58625.ole@freiheit.com> User-Agent: Opera M2/7.53 (Linux, build 737) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/56 X-Sequence-Number: 8532 You could try : explain analyze select "land" from "customer_dim" group by "land"; It will be a lot faster but I can't make it use the index on my machine... Example : create table dummy as (select id, id%255 as number from a large table with 1M rows); so we have a table with 256 (0-255) disctinct "number" values. -------------------------------------------------------------------------------- => explain analyze select distinct number from dummy; Unique (cost=69.83..74.83 rows=200 width=4) (actual time=13160.490..14414.004 rows=255 loops=1) -> Sort (cost=69.83..72.33 rows=1000 width=4) (actual time=13160.483..13955.792 rows=1000000 loops=1) Sort Key: number -> Seq Scan on dummy (cost=0.00..20.00 rows=1000 width=4) (actual time=0.052..1759.145 rows=1000000 loops=1) Total runtime: 14442.872 ms => Horribly slow because it has to sort 1M rows for the Unique. -------------------------------------------------------------------------------- => explain analyze select number from dummy group by number; HashAggregate (cost=22.50..22.50 rows=200 width=4) (actual time=1875.214..1875.459 rows=255 loops=1) -> Seq Scan on dummy (cost=0.00..20.00 rows=1000 width=4) (actual time=0.107..1021.014 rows=1000000 loops=1) Total runtime: 1875.646 ms => A lot faster because it HashAggregates instead of sorting (but still seq scan) -------------------------------------------------------------------------------- Now : create index dummy_idx on dummy(number); Let's try again. -------------------------------------------------------------------------------- explain analyze select distinct number from dummy; Unique (cost=0.00..35301.00 rows=200 width=4) (actual time=0.165..21781.732 rows=255 loops=1) -> Index Scan using dummy_idx on dummy (cost=0.00..32801.00 rows=1000000 width=4) (actual time=0.162..21154.752 rows=1000000 loops=1) Total runtime: 21782.270 ms => Index scan the whole table. argh. I should have ANALYZized. -------------------------------------------------------------------------------- explain analyze select number from dummy group by number; HashAggregate (cost=17402.00..17402.00 rows=200 width=4) (actual time=1788.425..1788.668 rows=255 loops=1) -> Seq Scan on dummy (cost=0.00..14902.00 rows=1000000 width=4) (actual time=0.048..960.063 rows=1000000 loops=1) Total runtime: 1788.855 ms => Still the same... -------------------------------------------------------------------------------- Let's make a function : The function starts at the lowest number and advances to the next number in the index until they are all exhausted. CREATE OR REPLACE FUNCTION sel_distinct() RETURNS SETOF INTEGER LANGUAGE plpgsql AS ' DECLARE pos INTEGER; BEGIN SELECT INTO pos number FROM dummy ORDER BY number ASC LIMIT 1; IF NOT FOUND THEN RAISE NOTICE ''no records.''; RETURN; END IF; LOOP RETURN NEXT pos; SELECT INTO pos number FROM dummy WHERE number>pos ORDER BY number ASC LIMIT 1; IF NOT FOUND THEN RETURN; END IF; END LOOP; END; '; explain analyze select * from sel_distinct(); Function Scan on sel_distinct (cost=0.00..12.50 rows=1000 width=4) (actual time=215.472..215.696 rows=255 loops=1) Total runtime: 215.839 ms That's better ! -------------------------------------------------------------------------------- Why not use DESC instead of ASC ? CREATE OR REPLACE FUNCTION sel_distinct() RETURNS SETOF INTEGER LANGUAGE plpgsql AS ' DECLARE pos INTEGER; BEGIN SELECT INTO pos number FROM dummy ORDER BY number DESC LIMIT 1; IF NOT FOUND THEN RAISE NOTICE ''no records.''; RETURN; END IF; LOOP RETURN NEXT pos; SELECT INTO pos number FROM dummy WHERE number; Wed, 6 Oct 2004 14:11:58 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91108-04 for ; Wed, 6 Oct 2004 13:11:52 +0000 (GMT) Received: from wren.rentec.com (wren.rentec.com [65.213.84.9]) by svr1.postgresql.org (Postfix) with ESMTP id DC9B532A160 for ; Wed, 6 Oct 2004 14:11:53 +0100 (BST) Received: from [172.26.132.145] (hoopoe.rentec.com [172.26.132.145]) by wren.rentec.com (8.12.9/8.12.1) with ESMTP id i96DBpeP003622; Wed, 6 Oct 2004 09:11:51 -0400 (EDT) Message-ID: <4163EF16.3030804@rentec.com> Date: Wed, 06 Oct 2004 09:11:50 -0400 From: Alan Stange Reply-To: stange@rentec.com Organization: Renaissance Technologies Corp. User-Agent: Mozilla Thunderbird 0.8 (X11/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Greg Stark Cc: pgsql-performance@postgresql.org Subject: Re: Excessive context switching on SMP Xeons References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <41630D50.3020308@lulu.com> <200410051538.51664.josh@agliodbs.com> <41636D8E.1090403@rentec.com> <87acv0z3y5.fsf@stark.xeocode.com> In-Reply-To: <87acv0z3y5.fsf@stark.xeocode.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=0.3 tagged_above=0.0 required=5.0 tests=BETTERMEMORY X-Spam-Level: X-Archive-Number: 200410/57 X-Sequence-Number: 8533 Greg Stark wrote: >Alan Stange writes: > > >>A few quick random observations on the Xeon v. Opteron comparison: >> >>- running a dual Xeon with hyperthreading turned on really isn't the same as >>having a quad cpu system. I haven't seen postgresql specific benchmarks, but >>the general case has been that HT is a benefit in a few particular work >>loads but with no benefit in general. >> >> >Part of the FUD with hyperthreading did have a kernel of truth that lied in >older kernels' schedulers. For example with Linux until recently the kernel >can easily end up scheduling two processes on the two virtual processors of >one single physical processor, leaving the other physical processor totally >idle. > >With modern kernels' schedulers I would expect hyperthreading to live up to >its billing of adding 10% to 20% performance. Ie., a dual Xeon machine with >hyperthreading won't be as fast as four processors, but it should be 10-20% >faster than a dual Xeon without hyperthreading. > >As with all things that will only help if you're bound by the right limited >resource to begin with. If you're I/O bound it isn't going to help. I would >expect Postgres with its heavy demand on memory bandwidth and shared memory >could potentially benefit more than usual from being able to context switch >during pipeline stalls. > > All true. I'd be surprised if HT on an older 2.8 Ghz Xeon with only a 512K cache will see any real benefit. The dual Xeon is already memory starved, now further increase the memory pressure on the caches (because the 512K is now "shared" by two virtual processors) and you probably won't see a gain. It's memory stalls all around. To be clear, the context switch in this case isn't a kernel context switch but a "virtual cpu" context switch. The probable reason we see dual Opteron boxes way outperforming dual Xeons boxes is exactly because of Postgresql's heavy demand on memory. The Opteron's have a much better memory system. A quick search on google or digging around in the comp.arch archives will provide lots of details. HP's web site has (had?) some benchmarks comparing these systems. HP sells both Xeon and Opteron systems, so the comparison were quite "fair". Their numbers showed the Opteron handily outperfoming the Xeons. -- Alan From pgsql-benchmarks-owner@postgresql.org Wed Oct 6 15:16:51 2004 X-Original-To: pgsql-benchmarks-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2793F329E8E; Wed, 6 Oct 2004 15:16:50 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 12764-06; Wed, 6 Oct 2004 14:16:44 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id D2716329E64; Wed, 6 Oct 2004 15:16:43 +0100 (BST) 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 i96EGeKS025577; Wed, 6 Oct 2004 10:16:40 -0400 (EDT) To: =?ISO-8859-1?Q?=22Alban_M=E9dici_=28NetCentrex=29=22?= Cc: pgsql-benchmarks@postgresql.org, pgsql-performance@postgresql.org Subject: Re: [PERFORM] stats on cursor and query execution troubleshooting In-reply-to: <4163C5C3.3030304@fr.netcentrex.net> References: <4163C5C3.3030304@fr.netcentrex.net> Comments: In-reply-to =?ISO-8859-1?Q?=22Alban_M=E9dici_=28NetCentrex=29=22?= message dated "Wed, 06 Oct 2004 12:15:31 +0200" Date: Wed, 06 Oct 2004 10:16:39 -0400 Message-ID: <25576.1097072199@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/2 X-Sequence-Number: 19 =?ISO-8859-1?Q?=22Alban_M=E9dici_=28NetCentrex=29=22?= writes: > I'm looking for the statistic of memory, CPU, filesystem access while=20 > executing some regular SQL query, and I want to compare them to > same kind of results while executing a cursor function. I think your second query is finding all the disk pages it needs in kernel disk cache, because they were all read in by the first query. This has little to do with cursor versus non cursor, and everything to do with hitting recently-read data again. regards, tom lane From pgsql-performance-owner@postgresql.org Wed Oct 6 16:45:40 2004 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 4854032A039 for ; Wed, 6 Oct 2004 16:45:39 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49427-03 for ; Wed, 6 Oct 2004 15:45:35 +0000 (GMT) Received: from snowwhite.lulu.com (mail.lulu.com [66.193.5.50]) by svr1.postgresql.org (Postfix) with ESMTP id F136F329DB1 for ; Wed, 6 Oct 2004 16:45:33 +0100 (BST) Received: from [10.0.0.32] (meatwad.rdu.lulu.com [10.0.0.32]) (authenticated bits=0) by snowwhite.lulu.com (8.12.8/8.12.8) with ESMTP id i96FkJbO002919 (version=TLSv1/SSLv3 cipher=RC4-MD5 bits=128 verify=NO) for ; Wed, 6 Oct 2004 11:46:19 -0400 Message-ID: <4164131A.6080601@lulu.com> Date: Wed, 06 Oct 2004 11:45:30 -0400 From: Bill Montgomery User-Agent: Mozilla Thunderbird 0.7.3 (X11/20040803) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Re: Excessive context switching on SMP Xeons References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <41630D50.3020308@lulu.com> <200410051538.51664.josh@agliodbs.com> In-Reply-To: <200410051538.51664.josh@agliodbs.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/59 X-Sequence-Number: 8535 Josh Berkus wrote: >>I'd be thrilled to test it too, if for no other reason that to determine >>whether what I'm experiencing really is the "CS problem". >> >> > >Hmmm ... Gavin's patch is built against 8.0, and any version of the patch >would require linux 2.6, probably 2.6.7 minimum. Can you test on that linux >version? Do you have the resources to back-port Gavin's patch? > > I don't currently have any SMP Xeon systems running a 2.6 kernel, but it could be arranged. As for back-porting the patch to 7.4.5, probably so, but I'd have to see it first. >>tps = 369.717832 (including connections establishing) >>tps = 370.852058 (excluding connections establishing) >> >> > >Doesn't seem too bad to me. Have anything to compare it to? > > Yes, about 280 tps on the same machine with the data directory on a 3-disk RAID 5 w/ a 128MB cache, rather than the SSD. I was expecting a much larger increase, given that the RAID does about 3MB/s of random 8k writes, and the SSD device does about 70MB/s of random 8k writes. Said differently, I thought my CPU bottleneck would be much higher, as to allow for more than a 30% increase in pgbench TPS when I took the IO bottleneck out of the equation. (That said, I'm not tuning for pgbench, but it is a useful comparison that everyone on the list is familiar with, and takes out the possibility that my app just has a bunch of poorly written queries). >What's in your postgresql.conf? > > Some relevant parameters: shared_buffers = 16384 sort_mem = 2048 vacuum_mem = 16384 max_fsm_pages = 200000 max_fsm_relations = 10000 fsync = true wal_sync_method = fsync wal_buffers = 32 checkpoint_segments = 6 effective_cache_size = 262144 random_page_cost = 0.25 Everything else is left at the default (or not relevant to this post). Anything blatantly stupid in there for my setup? Thanks, Bill Montgomery From pgsql-performance-owner@postgresql.org Wed Oct 6 17:08:49 2004 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 7E943329E8B for ; Wed, 6 Oct 2004 17:08:43 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 55100-07 for ; Wed, 6 Oct 2004 16:08:02 +0000 (GMT) Received: from mail.hamburg.cityline.net (ns.mcs-hh.de [194.77.146.129]) by svr1.postgresql.org (Postfix) with ESMTP id 4D706329E4B for ; Wed, 6 Oct 2004 17:08:01 +0100 (BST) Received: from [212.1.56.10] (helo=aconitin.toxine.lan) by mail.hamburg.cityline.net with asmtp (TLSv1:RC4-MD5:128) (Exim 4.32) id 1CFEKb-00073J-4j for pgsql-performance@postgresql.org; Wed, 06 Oct 2004 18:07:57 +0200 From: Ole Langbehn Organization: freiheit.com To: pgsql-performance@postgresql.org Subject: Re: sequential scan on select distinct Date: Wed, 6 Oct 2004 18:09:43 +0200 User-Agent: KMail/1.7 References: <200410061130.58625.ole@freiheit.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: <200410061809.43549.ole@freiheit.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/60 X-Sequence-Number: 8536 Am Mittwoch, 6. Oktober 2004 12:19 schrieb Pierre-Fr=E9d=E9ric Caillaud: > You could try : > > explain analyze select "land" from "customer_dim" group by "land"; > It will be a lot faster but I can't make it use the index on my machine.= .. this already speeds up my queries to about 1/4th of the time, which is abou= t=20 the range of mysql and oracle. > > Example : > > [..] > > Hum hum ! Again, a lot better ! > Index scan backwards seems a lot faster than index scan forwards. Why, I > don't know, but here you go from 15 seconds to 14 milliseconds... thanks for this very extensive answer, it helped me a lot. > > I don't know WHY (oh why) postgres does not use this kind of strategy > when distinct'ing an indexed field... Anybody got an idea ? That's the big question I still would like to see answered too. Can anyone= =20 tell us? TiA --=20 Ole Langbehn From pgsql-performance-owner@postgresql.org Wed Oct 6 17:41:36 2004 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 366C3329ED4 for ; Wed, 6 Oct 2004 17:41:34 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 68294-08 for ; Wed, 6 Oct 2004 16:41:30 +0000 (GMT) Received: from stark.xeocode.com (gsstark.mtl.istop.com [66.11.160.162]) by svr1.postgresql.org (Postfix) with ESMTP id E7E96329CA4 for ; Wed, 6 Oct 2004 17:41:28 +0100 (BST) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1CFEqm-00085M-00; Wed, 06 Oct 2004 12:41:12 -0400 To: =?iso-8859-1?q?Pierre-Fr=E9d=E9ric_Caillaud?= Cc: "Ole Langbehn" , pgsql-performance@postgresql.org Subject: Re: sequential scan on select distinct References: <200410061130.58625.ole@freiheit.com> In-Reply-To: From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 06 Oct 2004 12:41:12 -0400 Message-ID: <874ql7ztnb.fsf@stark.xeocode.com> Lines: 39 User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.3 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/61 X-Sequence-Number: 8537 Pierre-Fr�d�ric Caillaud writes: > I don't know WHY (oh why) postgres does not use this kind of strategy > when distinct'ing an indexed field... Anybody got an idea ? Well there are two questions here. Why given the current plans available does postgres choose a sequential scan instead of an index scan. And why isn't there this kind of "skip index scan" available. Postgres chooses a sequential scan with a sort (or hash aggregate) over an index scan because it expects it to be faster. sequential scans are much faster than random access scans of indexes, plus index scans need to read many more blocks. If you're finding the index scan to be just as fast as sequential scans you might consider lowering random_page_cost closer to 1.0. But note that you may be getting fooled by a testing methodology where more things are cached than would be in production. why isn't a "skip index scan" plan available? Well, nobody's written the code yet. It would part of the same code needed to get an index scan used for: select y,min(x) from bar group by y And possibly also related to the TODO item: Use index to restrict rows returned by multi-key index when used with non-consecutive keys to reduce heap accesses For an index on col1,col2,col3, and a WHERE clause of col1 = 5 and col3 = 9, spin though the index checking for col1 and col3 matches, rather than just col1 Note that the optimizer would have to make a judgement call based on the expected number of distinct values. If you had much more than 256 distinct values then the your plpgsql function wouldn't have performed well at all. -- greg From pgsql-performance-owner@postgresql.org Wed Oct 6 18:28:11 2004 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 379F232A400 for ; Wed, 6 Oct 2004 18:28:07 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83506-10 for ; Wed, 6 Oct 2004 17:28:01 +0000 (GMT) Received: from fe08.axelero.hu (fe08.axelero.hu [195.228.240.96]) by svr1.postgresql.org (Postfix) with ESMTP id 1798632A3FB for ; Wed, 6 Oct 2004 18:27:59 +0100 (BST) Received: from fe08 (localhost-02 [127.0.2.1]) by fe08.axelero.hu (8.12.11/8.12.11) with SMTP id i96HQxcf092268 for ; Wed, 6 Oct 2004 19:26:59 +0200 (CEST) Received: from fe08.axelero.hu [127.0.2.1] via SMTP gateway by fe08 [195.228.240.96]; id A068691C51D at Wed, 06 Oct 2004 19:26:59 +0200 Received: from fejleszt4 (121-248-182-81.adsl-fixip.axelero.hu [81.182.248.121]) by fe08.axelero.hu (8.12.11/8.12.11) with SMTP id i96HQw2i092255 for ; Wed, 6 Oct 2004 19:26:58 +0200 (CEST) Message-ID: <019301c4abc9$ef251ea0$0403a8c0@fejleszt4> From: =?iso-8859-1?Q?SZUCS_G=E1bor?= To: References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <41630D50.3020308@lulu.com> <200410051538.51664.josh@agliodbs.com> <4164131A.6080601@lulu.com> Subject: Re: Excessive context switching on SMP Xeons Date: Wed, 6 Oct 2004 19:28:45 +0200 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 6.00.2800.1437 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 X-RAVMilter-Version: 8.4.3(snapshot 20030217) (fe08.axelero.hu) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/62 X-Sequence-Number: 8538 Hmmm... I may be mistaken (I think last time I read about optimization params was in 7.3 docs), but doesn't RPC < 1 mean that random read is faster than sequential read? In your case, do you really think reading randomly is 4x faster than reading sequentially? Doesn't seem to make sense, even with a zillion-disk array. Theoretically. Also not sure, but sort_mem and vacuum_mem seem to be too small to me. G. %----------------------- cut here -----------------------% \end ----- Original Message ----- From: "Bill Montgomery" Sent: Wednesday, October 06, 2004 5:45 PM > Some relevant parameters: > shared_buffers = 16384 > sort_mem = 2048 > vacuum_mem = 16384 > max_fsm_pages = 200000 > max_fsm_relations = 10000 > fsync = true > wal_sync_method = fsync > wal_buffers = 32 > checkpoint_segments = 6 > effective_cache_size = 262144 > random_page_cost = 0.25 From pgsql-performance-owner@postgresql.org Wed Oct 6 18:31:28 2004 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 3AED732A260 for ; Wed, 6 Oct 2004 18:31:26 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 84278-07 for ; Wed, 6 Oct 2004 17:31:21 +0000 (GMT) Received: from boutiquenumerique.com (gailleton-2-82-67-9-10.fbx.proxad.net [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 25427329EE4 for ; Wed, 6 Oct 2004 18:31:19 +0100 (BST) Received: (qmail 17839 invoked from network); 6 Oct 2004 19:31:43 +0200 Received: from unknown (HELO musicbox) (192.168.0.2) by gailleton-2-82-67-9-10.fbx.proxad.net with SMTP; 6 Oct 2004 19:31:43 +0200 Date: Wed, 06 Oct 2004 19:34:22 +0200 To: pgsql-performance@postgresql.org Subject: Re: sequential scan on select distinct References: <200410061130.58625.ole@freiheit.com> <874ql7ztnb.fsf@stark.xeocode.com> From: =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= 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: <874ql7ztnb.fsf@stark.xeocode.com> User-Agent: Opera M2/7.53 (Linux, build 737) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/63 X-Sequence-Number: 8539 There are even three questions here : - given that 'SELECT DISTINCT field FROM table' is exactly the same as 'SELECT field FROM table GROUP BY field", postgres could transform the first into the second and avoid itself a (potentially killer) sort. On my example the table was not too large but on a very large table, sorting all the values and then discinct'ing them does not look too appealing. Currently Postgres does Sort+Unique, but there could be a DistinctSort instead of a Sort, that is a thing that sorts and removes the duplicates at the same time. Not that much complicated to code than a sort, and much faster in this case. Or there could be a DistinctHash, which would be similar or rather identical to a HashAggregate and would again skip the sort. It would (as a bonus) speed up queries like UNION (not ALL), that kind of things. For example : explain (select number from dummy) union (select number from dummy); Unique (cost=287087.62..297087.62 rows=2000000 width=4) -> Sort (cost=287087.62..292087.62 rows=2000000 width=4) Sort Key: number -> Append (cost=0.00..49804.00 rows=2000000 width=4) -> Subquery Scan "*SELECT* 1" (cost=0.00..24902.00 rows=1000000 width=4) -> Seq Scan on dummy (cost=0.00..14902.00 rows=1000000 width=4) -> Subquery Scan "*SELECT* 2" (cost=0.00..24902.00 rows=1000000 width=4) -> Seq Scan on dummy (cost=0.00..14902.00 rows=1000000 width=4) This is scary ! I can rewrite it as such (and the planner could, too) : explain select * from ((select number from dummy) union all (select number from dummy)) as foo group by number; HashAggregate (cost=74804.00..74804.00 rows=200 width=4) -> Subquery Scan foo (cost=0.00..69804.00 rows=2000000 width=4) -> Append (cost=0.00..49804.00 rows=2000000 width=4) -> Subquery Scan "*SELECT* 1" (cost=0.00..24902.00 rows=1000000 width=4) -> Seq Scan on dummy (cost=0.00..14902.00 rows=1000000 width=4) -> Subquery Scan "*SELECT* 2" (cost=0.00..24902.00 rows=1000000 width=4) -> Seq Scan on dummy (cost=0.00..14902.00 rows=1000000 width=4) which avoids a large sort... However there must be cases in which performing a sort is faster, like when there are a lot of distinct values and the HashAggregate becomes huge too. > Well there are two questions here. Why given the current plans available > does > postgres choose a sequential scan instead of an index scan. And why isn't Well because it needs to get all the rows in the table in order. in this case seq scan+sort is about twice as fast as index scan. Interestingly, once I ANALYZED the table, postgres will chooses to index-scan, which is slower. > there this kind of "skip index scan" available. It would be really nice to have a skip index scan available. I have an other idea, lets call it the indexed sequential scan : When pg knows there are a lot of rows to access, it will ignore the index and seqscan. This is because index access is very random, thus slow. However postgres could implement an "indexed sequential scan" where : - the page numbers for the matching rows are looked up in the index (this is fast as an index has good locality) - the page numbers are grouped so we have a list of pages with one and only one instance of each page number - the list is then sorted so we have page numbers in-order - the pages are loaded in sorted order (doing a kind of partial sequential scan) which would be faster than reading them randomly. Other ideas later > Postgres chooses a sequential scan with a sort (or hash aggregate) over > an > index scan because it expects it to be faster. sequential scans are much > faster than random access scans of indexes, plus index scans need to > read many > more blocks. If you're finding the index scan to be just as fast as > sequential > scans you might consider lowering random_page_cost closer to 1.0. But > note > that you may be getting fooled by a testing methodology where more > things are > cached than would be in production. > > why isn't a "skip index scan" plan available? Well, nobody's written the > code > yet. It would part of the same code needed to get an index scan used for: > > select y,min(x) from bar group by y > > And possibly also related to the TODO item: > > Use index to restrict rows returned by multi-key index when used with > non-consecutive keys to reduce heap accesses > > For an index on col1,col2,col3, and a WHERE clause of col1 = 5 and > col3 = > 9, spin though the index checking for col1 and col3 matches, rather > than > just col1 > > > Note that the optimizer would have to make a judgement call based on the > expected number of distinct values. If you had much more than 256 > distinct > values then the your plpgsql function wouldn't have performed well at > all. > From pgsql-performance-owner@postgresql.org Wed Oct 6 19:55:23 2004 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 AE2AC329D20 for ; Wed, 6 Oct 2004 19:55:22 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 12724-02 for ; Wed, 6 Oct 2004 18:55:17 +0000 (GMT) Received: from pd4mo1so.prod.shaw.ca (shawidc-mo1.cg.shawcable.net [24.71.223.10]) by svr1.postgresql.org (Postfix) with ESMTP id C95A432A0C4 for ; Wed, 6 Oct 2004 19:55:14 +0100 (BST) Received: from pd3mr3so.prod.shaw.ca (pd3mr3so-qfe3.prod.shaw.ca [10.0.141.179]) by l-daemon (Sun ONE Messaging Server 6.0 HotFix 1.01 (built Mar 15 2004)) with ESMTP id <0I5600MYUF7RJL60@l-daemon> for pgsql-performance@postgresql.org; Wed, 06 Oct 2004 12:55:03 -0600 (MDT) Received: from pn2ml9so.prod.shaw.ca ([10.0.121.7]) by pd3mr3so.prod.shaw.ca (Sun ONE Messaging Server 6.0 HotFix 1.01 (built Mar 15 2004)) with ESMTP id <0I5600GDAF7RGJD0@pd3mr3so.prod.shaw.ca> for pgsql-performance@postgresql.org; Wed, 06 Oct 2004 12:55:03 -0600 (MDT) Received: from [192.168.1.10] (S01060050bac04c93.ed.shawcable.net [68.148.193.184]) by l-daemon (iPlanet Messaging Server 5.2 HotFix 1.18 (built Jul 28 2003)) with ESMTP id <0I5600J7SF7Q3C@l-daemon> for pgsql-performance@postgresql.org; Wed, 06 Oct 2004 12:55:03 -0600 (MDT) Date: Wed, 06 Oct 2004 12:55:02 -0600 From: Patrick Clery Subject: Re: Comparing user attributes with bitwise operators In-reply-to: <200410050932.38595.josh@agliodbs.com> To: pgsql-performance@postgresql.org Message-id: <200410061255.03052.patrick@phpforhire.com> MIME-version: 1.0 Content-type: text/plain; charset=iso-8859-1 Content-transfer-encoding: 7bit Content-disposition: inline References: <200409160141.37604.patrick@phpforhire.com> <200410050039.24230.patrick@phpforhire.com> <200410050932.38595.josh@agliodbs.com> User-Agent: KMail/1.7 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.9 tagged_above=0.0 required=5.0 tests=FAKE_HELO_SHAW_CA X-Spam-Level: X-Archive-Number: 200410/64 X-Sequence-Number: 8540 Another problem I should note is that when I first insert all the data into the people_attributes table ("the int[] table"), the GiST index is not used: THE INDEX: "people_attributes_search" gist ((ARRAY[age, gender, orientation, children, drinking, education, ethnicity, eyecolor, haircolor, hairstyle, height, income, occupation, relation, religion, smoking, w ant_children, weight] + seeking + languages)) PART OF THE QUERY PLAN: Seq Scan on people_attributes pa (cost=0.00..0.00 rows=1 width=20) Filter: (((ARRAY[age, gender, orientation, children, drinking, education, ethnicity, eyecolor, haircolor, hairstyle, height, income, occupation, relation, religion, smoking, want_children, weight] + seeking) + languages) @@ '( ( 4 | 5 ) | 6 ) & 88 & 48 & ( 69 | 70 ) & 92 & ( ( ( ( ( ( ( ( ( ( ( ( ( 95 | 96 ) | 97 ) | 98 ) | 99 ) | 100 ) | 101 ) | 102 ) | 103 ) | 104 ) | 105 ) | 106 ) | 107 ) | 108 ) & ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( 190 | 191 ) | 192 ) | 193 ) | 194 ) | 195 ) | 196 ) | 197 ) | 198 ) | 199 ) | 200 ) | 201 ) | 202 ) | 203 ) | 204 ) | 205 ) | 206 ) | 207 ) | 208 ) | 209 ) | 210 ) | 211 ) | 212 ) | 213 ) | 214 ) | 215 ) | 216 ) | 217 ) | 218 ) | 219 ) | 220 ) | 221 ) | 222 ) | 223 ) | 224 ) | 225 ) | 226 ) | 227 ) | 228 ) | 229 ) | 230 ) | 231 ) | 232 ) | 233 ) | 234 ) | 235 ) | 236 ) | 237 ) | 238 ) | 239 ) | 240 ) | 241 ) | 242 ) | 243 )'::query_int) So I run "VACUUM ANALYZE people_attributes", then run again: PART OF THE QUERY PLAN: Index Scan using people_attributes_pkey on people_attributes pa (cost=0.00..5.32 rows=1 width=20) Index Cond: (pa.person_id = "outer".person_id) Filter: (((ARRAY[age, gender, orientation, children, drinking, education, ethnicity, eyecolor, haircolor, hairstyle, height, income, occupation, relation, religion, smoking, want_children, weight] + seeking) + languages) @@ '( ( 4 | 5 ) | 6 ) & 88 & 48 & ( 69 | 70 ) & 92 & ( ( ( ( ( ( ( ( ( ( ( ( ( 95 | 96 ) | 97 ) | 98 ) | 99 ) | 100 ) | 101 ) | 102 ) | 103 ) | 104 ) | 105 ) | 106 ) | 107 ) | 108 ) & ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( 190 | 191 ) | 192 ) | 193 ) | 194 ) | 195 ) | 196 ) | 197 ) | 198 ) | 199 ) | 200 ) | 201 ) | 202 ) | 203 ) | 204 ) | 205 ) | 206 ) | 207 ) | 208 ) | 209 ) | 210 ) | 211 ) | 212 ) | 213 ) | 214 ) | 215 ) | 216 ) | 217 ) | 218 ) | 219 ) | 220 ) | 221 ) | 222 ) | 223 ) | 224 ) | 225 ) | 226 ) | 227 ) | 228 ) | 229 ) | 230 ) | 231 ) | 232 ) | 233 ) | 234 ) | 235 ) | 236 ) | 237 ) | 238 ) | 239 ) | 240 ) | 241 ) | 242 ) | 243 )'::query_int) Still not using the index. I'm trying to DROP INDEX and recreate it, but the query just stalls. I remember last time this situation happened that I just dropped and recreated the index, and voila it was using the index again. Now I can't seem to get this index to drop. Here's the table structure: Column | Type | Modifiers ---------------+-----------+-------------------- person_id | integer | not null askmecount | integer | not null default 0 age | integer | not null gender | integer | not null bodytype | integer | not null children | integer | not null drinking | integer | not null education | integer | not null ethnicity | integer | not null eyecolor | integer | not null haircolor | integer | not null hairstyle | integer | not null height | integer | not null income | integer | not null languages | integer[] | not null occupation | integer | not null orientation | integer | not null relation | integer | not null religion | integer | not null smoking | integer | not null want_children | integer | not null weight | integer | not null seeking | integer[] | not null Indexes: "people_attributes_pkey" PRIMARY KEY, btree (person_id) "people_attributes_search" gist ((ARRAY[age, gender, orientation, children, drinking, education, ethnicity, eyecolor, haircolor, hairstyle, height, income, occupation, relation, religion, smoking, w ant_children, weight] + seeking + languages)) Foreign-key constraints: "people_attributes_weight_fkey" FOREIGN KEY (weight) REFERENCES attribute_values(value_id) ON DEL ETE RESTRICT "people_attributes_person_id_fkey" FOREIGN KEY (person_id) REFERENCES people(person_id) ON DELETE CASCADE DEFERRABLE INITIALLY DEFERRED "people_attributes_age_fkey" FOREIGN KEY (age) REFERENCES attribute_values(value_id) ON DELETE RE STRICT "people_attributes_gender_fkey" FOREIGN KEY (gender) REFERENCES attribute_values(value_id) ON DEL ETE RESTRICT "people_attributes_bodytype_fkey" FOREIGN KEY (bodytype) REFERENCES attribute_values(value_id) ON DELETE RESTRICT "people_attributes_children_fkey" FOREIGN KEY (children) REFERENCES attribute_values(value_id) ON DELETE RESTRICT "people_attributes_drinking_fkey" FOREIGN KEY (drinking) REFERENCES attribute_values(value_id) ON DELETE RESTRICT "people_attributes_education_fkey" FOREIGN KEY (education) REFERENCES attribute_values(value_id) ON DELETE RESTRICT "people_attributes_ethnicity_fkey" FOREIGN KEY (ethnicity) REFERENCES attribute_values(value_id) ON DELETE RESTRICT "people_attributes_eyecolor_fkey" FOREIGN KEY (eyecolor) REFERENCES attribute_values(value_id) ON DELETE RESTRICT "people_attributes_haircolor_fkey" FOREIGN KEY (haircolor) REFERENCES attribute_values(value_id) ON DELETE RESTRICT "people_attributes_hairstyle_fkey" FOREIGN KEY (hairstyle) REFERENCES attribute_values(value_id) ON DELETE RESTRICT "people_attributes_height_fkey" FOREIGN KEY (height) REFERENCES attribute_values(value_id) ON DELETE RESTRICT "people_attributes_income_fkey" FOREIGN KEY (income) REFERENCES attribute_values(value_id) ON DELETE RESTRICT "people_attributes_occupation_fkey" FOREIGN KEY (occupation) REFERENCES attribute_values(value_id ) ON DELETE RESTRICT "people_attributes_orientation_fkey" FOREIGN KEY (orientation) REFERENCES attribute_values(value_ id) ON DELETE RESTRICT "people_attributes_relation_fkey" FOREIGN KEY (relation) REFERENCES attribute_values(value_id) ON DELETE RESTRICT "people_attributes_religion_fkey" FOREIGN KEY (religion) REFERENCES attribute_values(value_id) ON DELETE RESTRICT "people_attributes_smoking_fkey" FOREIGN KEY (smoking) REFERENCES attribute_values(value_id) ON D ELETE RESTRICT "people_attributes_want_children_fkey" FOREIGN KEY (want_children) REFERENCES attribute_values(va lue_id) ON DELETE RESTRICT Is it all the foreign keys that are stalling the drop? I have done VACUUM ANALYZE on the entire db. Could anyone offer some insight as to why this index is not being used or why the index is not dropping easily? On Tuesday 05 October 2004 10:32, you wrote: > Patrick, > > First off, thanks for posting this solution! I love to see a new demo of > The Power of Postgres(tm) and have been wondering about this particular > problem since it came up on IRC. > > > The array method works quite nicely, especially for the > > columns like "languages" and "seeking" that are multiple choice. However, > > even though this method is fast, I still might opt for caching the > > results because the "real world" search query involves a lot more and > > will be executed non-stop. But to have it run this fast the first time > > certainly helps. > > Now, for the bad news: you need to test having a large load of users > updating their data. The drawback to GiST indexes is that they are > low-concurrency, because the updating process needs to lock the whole index > (this has been on our TODO list for about a decade, but it's a hard > problem). From pgsql-performance-owner@postgresql.org Wed Oct 6 20:28:06 2004 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 4CD8D32A039 for ; Wed, 6 Oct 2004 20:28:00 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 21457-02 for ; Wed, 6 Oct 2004 19:27:54 +0000 (GMT) Received: from pd2mo3so.prod.shaw.ca (shawidc-mo1.cg.shawcable.net [24.71.223.10]) by svr1.postgresql.org (Postfix) with ESMTP id A8D3C32A00E for ; Wed, 6 Oct 2004 20:27:55 +0100 (BST) Received: from pd2mr3so.prod.shaw.ca (pd2mr3so-qfe3.prod.shaw.ca [10.0.141.108]) by l-daemon (Sun ONE Messaging Server 6.0 HotFix 1.01 (built Mar 15 2004)) with ESMTP id <0I56005G2GQIS6C0@l-daemon> for pgsql-performance@postgresql.org; Wed, 06 Oct 2004 13:27:54 -0600 (MDT) Received: from pn2ml9so.prod.shaw.ca ([10.0.121.7]) by pd2mr3so.prod.shaw.ca (Sun ONE Messaging Server 6.0 HotFix 1.01 (built Mar 15 2004)) with ESMTP id <0I56003UFGQIAI40@pd2mr3so.prod.shaw.ca> for pgsql-performance@postgresql.org; Wed, 06 Oct 2004 13:27:54 -0600 (MDT) Received: from [192.168.1.10] (S01060050bac04c93.ed.shawcable.net [68.148.193.184]) by l-daemon (iPlanet Messaging Server 5.2 HotFix 1.18 (built Jul 28 2003)) with ESMTP id <0I5600F0XGQIM4@l-daemon> for pgsql-performance@postgresql.org; Wed, 06 Oct 2004 13:27:54 -0600 (MDT) Date: Wed, 06 Oct 2004 13:27:55 -0600 From: Patrick Clery Subject: Re: Comparing user attributes with bitwise operators In-reply-to: <200410061325.38807.patrick@phpforhire.com> To: pgsql-performance@postgresql.org Message-id: <200410061327.55989.patrick@phpforhire.com> MIME-version: 1.0 Content-type: text/plain; charset=iso-8859-1 Content-transfer-encoding: 7bit Content-disposition: inline References: <200409160141.37604.patrick@phpforhire.com> <200410061255.03052.patrick@phpforhire.com> <200410061325.38807.patrick@phpforhire.com> User-Agent: KMail/1.7 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.9 tagged_above=0.0 required=5.0 tests=FAKE_HELO_SHAW_CA X-Spam-Level: X-Archive-Number: 200410/65 X-Sequence-Number: 8541 Err... I REINDEX'ed it and it is now using the index. :) I'd still appreciate if anyone could tell me why this needs to be reindexed. Is the index not updated when the records are inserted? > On Wednesday 06 October 2004 12:55, I wrote: > > Another problem I should note is that when I first insert all the data > > into the people_attributes table ("the int[] table"), the GiST index is > > not used: > > > > THE INDEX: > > "people_attributes_search" gist ((ARRAY[age, gender, orientation, > > children, drinking, education, > > ethnicity, eyecolor, haircolor, hairstyle, height, income, occupation, > > relation, religion, smoking, w > > ant_children, weight] + seeking + languages)) > > > > PART OF THE QUERY PLAN: > > Seq Scan on people_attributes pa (cost=0.00..0.00 rows=1 width=20) > > Filter: (((ARRAY[age, gender, orientation, children, > > drinking, education, ethnicity, eyecolor, haircolor, hairstyle, height, > > income, occupation, relation, religion, smoking, want_children, weight] + > > seeking) + languages) @@ '( ( 4 | 5 ) | 6 ) & 88 & 48 & ( 69 | 70 ) & 92 > > & ( ( ( ( ( ( ( ( ( ( ( ( ( 95 | 96 ) | 97 ) | 98 ) | 99 ) | 100 ) | 101 > > ) | 102 ) | 103 ) | 104 ) | 105 ) | 106 ) | 107 ) | 108 ) & > > ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( > > ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( 190 > > > > | 191 ) | 192 ) | 193 ) | 194 ) | 195 ) | 196 ) | 197 ) | 198 ) | 199 ) | > > > > 200 ) | 201 ) | 202 ) | 203 ) | 204 ) | 205 ) | 206 ) | 207 ) | 208 ) | > > 209 ) > > > > | 210 ) | 211 ) | 212 ) | 213 ) | 214 ) | 215 ) | 216 ) | 217 ) | 218 ) | > > > > 219 ) | 220 ) | 221 ) | 222 ) | 223 ) | 224 ) | 225 ) | 226 ) | 227 ) | > > 228 ) > > > > | 229 ) | 230 ) | 231 ) | 232 ) | 233 ) | 234 ) | 235 ) | 236 ) | 237 ) | > > > > 238 ) | 239 ) | 240 ) | 241 ) | 242 ) | 243 )'::query_int) > > > > > > So I run "VACUUM ANALYZE people_attributes", then run again: > > > > PART OF THE QUERY PLAN: > > Index Scan using people_attributes_pkey on people_attributes pa > > (cost=0.00..5.32 rows=1 width=20) > > Index Cond: (pa.person_id = "outer".person_id) > > Filter: (((ARRAY[age, gender, orientation, children, drinking, > > education, ethnicity, eyecolor, haircolor, hairstyle, height, income, > > occupation, relation, religion, smoking, want_children, weight] + > > seeking) + languages) @@ '( ( 4 | 5 ) | 6 ) & 88 & 48 & ( 69 | 70 ) & 92 > > & ( ( ( ( ( ( ( ( ( ( ( ( ( 95 | 96 ) | 97 ) | 98 ) | 99 ) | 100 ) | 101 > > ) | 102 ) | 103 ) | 104 ) | 105 ) | 106 ) | 107 ) | 108 ) & > > ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( > > ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( 190 > > > > | 191 ) | 192 ) | 193 ) | 194 ) | 195 ) | 196 ) | 197 ) | 198 ) | 199 ) | > > > > 200 ) | 201 ) | 202 ) | 203 ) | 204 ) | 205 ) | 206 ) | 207 ) | 208 ) | > > 209 ) > > > > | 210 ) | 211 ) | 212 ) | 213 ) | 214 ) | 215 ) | 216 ) | 217 ) | 218 ) | > > > > 219 ) | 220 ) | 221 ) | 222 ) | 223 ) | 224 ) | 225 ) | 226 ) | 227 ) | > > 228 ) > > > > | 229 ) | 230 ) | 231 ) | 232 ) | 233 ) | 234 ) | 235 ) | 236 ) | 237 ) | > > > > 238 ) | 239 ) | 240 ) | 241 ) | 242 ) | 243 )'::query_int) > > > > Still not using the index. I'm trying to DROP INDEX and recreate it, but > > the query just stalls. I remember last time this situation happened that > > I just dropped and recreated the index, and voila it was using the index > > again. Now I can't seem to get this index to drop. Here's the table > > structure: > > > > > > Column | Type | Modifiers > > ---------------+-----------+-------------------- > > person_id | integer | not null > > askmecount | integer | not null default 0 > > age | integer | not null > > gender | integer | not null > > bodytype | integer | not null > > children | integer | not null > > drinking | integer | not null > > education | integer | not null > > ethnicity | integer | not null > > eyecolor | integer | not null > > haircolor | integer | not null > > hairstyle | integer | not null > > height | integer | not null > > income | integer | not null > > languages | integer[] | not null > > occupation | integer | not null > > orientation | integer | not null > > relation | integer | not null > > religion | integer | not null > > smoking | integer | not null > > want_children | integer | not null > > weight | integer | not null > > seeking | integer[] | not null > > Indexes: > > "people_attributes_pkey" PRIMARY KEY, btree (person_id) > > "people_attributes_search" gist ((ARRAY[age, gender, orientation, > > children, drinking, education, > > ethnicity, eyecolor, haircolor, hairstyle, height, income, occupation, > > relation, religion, smoking, w > > ant_children, weight] + seeking + languages)) > > Foreign-key constraints: > > "people_attributes_weight_fkey" FOREIGN KEY (weight) REFERENCES > > attribute_values(value_id) ON DEL > > ETE RESTRICT > > "people_attributes_person_id_fkey" FOREIGN KEY (person_id) REFERENCES > > people(person_id) ON DELETE > > CASCADE DEFERRABLE INITIALLY DEFERRED > > "people_attributes_age_fkey" FOREIGN KEY (age) REFERENCES > > attribute_values(value_id) ON DELETE RE > > STRICT > > "people_attributes_gender_fkey" FOREIGN KEY (gender) REFERENCES > > attribute_values(value_id) ON DEL > > ETE RESTRICT > > "people_attributes_bodytype_fkey" FOREIGN KEY (bodytype) REFERENCES > > attribute_values(value_id) ON > > DELETE RESTRICT > > "people_attributes_children_fkey" FOREIGN KEY (children) REFERENCES > > attribute_values(value_id) ON > > DELETE RESTRICT > > "people_attributes_drinking_fkey" FOREIGN KEY (drinking) REFERENCES > > attribute_values(value_id) ON > > DELETE RESTRICT > > "people_attributes_education_fkey" FOREIGN KEY (education) REFERENCES > > attribute_values(value_id) > > ON DELETE RESTRICT > > "people_attributes_ethnicity_fkey" FOREIGN KEY (ethnicity) REFERENCES > > attribute_values(value_id) > > ON DELETE RESTRICT > > "people_attributes_eyecolor_fkey" FOREIGN KEY (eyecolor) REFERENCES > > attribute_values(value_id) ON > > DELETE RESTRICT > > "people_attributes_haircolor_fkey" FOREIGN KEY (haircolor) REFERENCES > > attribute_values(value_id) > > ON DELETE RESTRICT > > "people_attributes_hairstyle_fkey" FOREIGN KEY (hairstyle) REFERENCES > > attribute_values(value_id) > > ON DELETE RESTRICT > > "people_attributes_height_fkey" FOREIGN KEY (height) REFERENCES > > attribute_values(value_id) ON DELETE RESTRICT > > "people_attributes_income_fkey" FOREIGN KEY (income) REFERENCES > > attribute_values(value_id) ON DELETE RESTRICT > > "people_attributes_occupation_fkey" FOREIGN KEY (occupation) > > REFERENCES attribute_values(value_id > > ) ON DELETE RESTRICT > > "people_attributes_orientation_fkey" FOREIGN KEY (orientation) > > REFERENCES attribute_values(value_ > > id) ON DELETE RESTRICT > > "people_attributes_relation_fkey" FOREIGN KEY (relation) REFERENCES > > attribute_values(value_id) ON > > DELETE RESTRICT > > "people_attributes_religion_fkey" FOREIGN KEY (religion) REFERENCES > > attribute_values(value_id) ON > > DELETE RESTRICT > > "people_attributes_smoking_fkey" FOREIGN KEY (smoking) REFERENCES > > attribute_values(value_id) ON D > > ELETE RESTRICT > > "people_attributes_want_children_fkey" FOREIGN KEY (want_children) > > REFERENCES attribute_values(va > > lue_id) ON DELETE RESTRICT > > > > > > Is it all the foreign keys that are stalling the drop? I have done VACUUM > > ANALYZE on the entire db. Could anyone offer some insight as to why this > > index is not being used or why the index is not dropping easily? > > > > On Tuesday 05 October 2004 10:32, you wrote: > > > Patrick, > > > > > > First off, thanks for posting this solution! I love to see a new demo > > > of The Power of Postgres(tm) and have been wondering about this > > > particular problem since it came up on IRC. > > > > > > > The array method works quite nicely, especially for the > > > > columns like "languages" and "seeking" that are multiple choice. > > > > However, even though this method is fast, I still might opt for > > > > caching the results because the "real world" search query involves a > > > > lot more and will be executed non-stop. But to have it run this fast > > > > the first time certainly helps. > > > > > > Now, for the bad news: you need to test having a large load of users > > > updating their data. The drawback to GiST indexes is that they are > > > low-concurrency, because the updating process needs to lock the whole > > > index (this has been on our TODO list for about a decade, but it's a > > > hard problem). > > > > ---------------------------(end of broadcast)--------------------------- > > TIP 6: Have you searched our list archives? > > > > http://archives.postgresql.org From pgsql-performance-owner@postgresql.org Wed Oct 6 20:40:52 2004 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 BC605329EE3 for ; Wed, 6 Oct 2004 20:40:50 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 26505-05 for ; Wed, 6 Oct 2004 19:40:45 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 265B132A4E0 for ; Wed, 6 Oct 2004 20:40:45 +0100 (BST) 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 i96JcQme029195; Wed, 6 Oct 2004 15:38:26 -0400 (EDT) To: Greg Stark Cc: =?iso-8859-1?q?Pierre-Fr=E9d=E9ric_Caillaud?= , "Ole Langbehn" , pgsql-performance@postgresql.org Subject: Re: sequential scan on select distinct In-reply-to: <874ql7ztnb.fsf@stark.xeocode.com> References: <200410061130.58625.ole@freiheit.com> <874ql7ztnb.fsf@stark.xeocode.com> Comments: In-reply-to Greg Stark message dated "06 Oct 2004 12:41:12 -0400" Date: Wed, 06 Oct 2004 15:38:26 -0400 Message-ID: <29194.1097091506@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/67 X-Sequence-Number: 8543 Greg Stark writes: > why isn't a "skip index scan" plan available? Well, nobody's written the code > yet. I don't really think it would be a useful plan anyway. What *would* be useful is to support HashAggregate as an implementation alternative for DISTINCT --- currently I believe we only consider that for GROUP BY. The DISTINCT planning code is fairly old and crufty and hasn't been redesigned lately. regards, tom lane From pgsql-performance-owner@postgresql.org Wed Oct 6 20:40:04 2004 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 D3B1E329FEF for ; Wed, 6 Oct 2004 20:39:12 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 25337-04 for ; Wed, 6 Oct 2004 19:39:06 +0000 (GMT) Received: from stark.xeocode.com (gsstark.mtl.istop.com [66.11.160.162]) by svr1.postgresql.org (Postfix) with ESMTP id 7925D32A493 for ; Wed, 6 Oct 2004 20:38:58 +0100 (BST) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1CFHcm-0000xZ-00; Wed, 06 Oct 2004 15:38:56 -0400 To: Patrick Clery Cc: pgsql-performance@postgresql.org Subject: Re: Comparing user attributes with bitwise operators References: <200409160141.37604.patrick@phpforhire.com> <200410050039.24230.patrick@phpforhire.com> <200410050932.38595.josh@agliodbs.com> <200410061255.03052.patrick@phpforhire.com> In-Reply-To: <200410061255.03052.patrick@phpforhire.com> From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 06 Oct 2004 15:38:56 -0400 Message-ID: <87y8ijy6un.fsf@stark.xeocode.com> Lines: 27 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/66 X-Sequence-Number: 8542 Patrick Clery writes: > PART OF THE QUERY PLAN: > Index Scan using people_attributes_pkey on people_attributes pa (cost=0.00..5.32 rows=1 width=20) > Index Cond: (pa.person_id = "outer".person_id) > Filter: (((ARRAY[age, gender, orientation, children, drinking, You'll probably have to show the rest of the plan for anyone to have much idea what's going on. It seems to be part of a join of some sort and the planner is choosing to drive the join from the wrong table. This may make it awkward to force the right plan using enable_seqscan or anything like that. But GiST indexes don't have very good selectivity estimates so I'm not sure you can hope for the optimizer to guess right on its own. > Is it all the foreign keys that are stalling the drop? I have done VACUUM > ANALYZE on the entire db. Could anyone offer some insight as to why this > index is not being used or why the index is not dropping easily? I don't think foreign keys cause problems dropping indexes. Foreign key constraints are just checked whenever there's an insert/update/delete. Perhaps you're just underestimating the size of this index and the amount of time it'll take to delete it? Or are there queries actively executing using the index while you're trying to delete it? Or a vacuum running? -- greg From pgsql-performance-owner@postgresql.org Wed Oct 6 21:02:40 2004 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 AA65F32A404 for ; Wed, 6 Oct 2004 21:02:34 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 31452-06 for ; Wed, 6 Oct 2004 20:02:29 +0000 (GMT) Received: from stark.xeocode.com (gsstark.mtl.istop.com [66.11.160.162]) by svr1.postgresql.org (Postfix) with ESMTP id D40BB32A3FA for ; Wed, 6 Oct 2004 21:02:29 +0100 (BST) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1CFHzS-00017E-00; Wed, 06 Oct 2004 16:02:22 -0400 To: Tom Lane Cc: Greg Stark , =?iso-8859-1?q??= =?iso-8859-1?q? Pierre-Fr�d�ric Caillaud?= , "Ole Langbehn" , pgsql-performance@postgresql.org Subject: Re: sequential scan on select distinct References: <200410061130.58625.ole@freiheit.com> <874ql7ztnb.fsf@stark.xeocode.com> <29194.1097091506@sss.pgh.pa.us> In-Reply-To: <29194.1097091506@sss.pgh.pa.us> From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 06 Oct 2004 16:02:22 -0400 Message-ID: <87sm8ry5rl.fsf@stark.xeocode.com> Lines: 31 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-Amavis-Alert: BAD HEADER Non-encoded 8-bit data (char E9 hex) in message header 'Cc' Cc: ...= =?iso-8859-1?q? Pierre-Fr\351d\351ric Caillaud?... ^ X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/68 X-Sequence-Number: 8544 Tom Lane writes: > Greg Stark writes: > > why isn't a "skip index scan" plan available? Well, nobody's written the code > > yet. > > I don't really think it would be a useful plan anyway. Well it would clearly be useful in this test case, where has a small number of distinct values in a large table, and an index on the column. His plpgsql function that emulates such a plan is an order of magnitude faster than the hash aggregate plan even though it has to do entirely separate index scans for each key value. I'm not sure where the break-even point would be, but it would probably be pretty low. Probably somewhere around the order of 1% distinct values in the table. That might be uncommon, but certainly not impossible. But regardless of how uncommon it is, it could be considered important in another sense: when you need it there really isn't any alternative. It's an algorithmic improvement with no bound on the performance difference. Nothing short of using a manually maintained materialized view would bring the performance into the same ballpark. So even if it's only useful occasionally, not having the plan available can leave postgres with no effective plan for what should be an easy query. -- greg From pgsql-performance-owner@postgresql.org Wed Oct 6 21:05:03 2004 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 EAAFC32A02D for ; Wed, 6 Oct 2004 21:05:01 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 31323-08 for ; Wed, 6 Oct 2004 20:04:54 +0000 (GMT) Received: from neomail10.traderonline.com (email.traderonline.com [65.213.231.10]) by svr1.postgresql.org (Postfix) with ESMTP id 1D0AE329FF4 for ; Wed, 6 Oct 2004 21:04:54 +0100 (BST) Received: from dyounger-IBM (64-139-89-109-ubr02b-epensb01-pa.hfc.comcastbusiness.net [64.139.89.109]) by neomail10.traderonline.com (8.12.10/8.12.8) with ESMTP id i96K4of6002273 for ; Wed, 6 Oct 2004 16:04:50 -0400 Message-Id: <6.1.2.0.2.20041006152747.033bcd50@pop.traderonline.com> X-Sender: dylists@mail.ptd.net (Unverified) X-Mailer: QUALCOMM Windows Eudora Version 6.1.2.0 Date: Wed, 06 Oct 2004 16:04:52 -0400 To: pgsql-performance@postgresql.org From: Doug Y Subject: The never ending quest for clarity on shared_buffers 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/69 X-Sequence-Number: 8545 Hello, We recently upgraded os from rh 7.2 (2.4 kernel) to Suse 9.1 (2.6 kernel), and psql from 7.3.4 to 7.4.2 One of the quirks I've noticed is how the queries don't always have the same explain plans on the new psql... but that's a different email I think. My main question is I'm trying to convince the powers that be to let me use persistent DB connections (from apache 2 / php), and my research has yielded conflicting documentation about the shared_buffers setting... real shocker there :) For idle persistent connections, do each of them allocate the memory specified by this setting (shared_buffers * 8k), or is it one pool used by all the connection (which seems the logical conclusion based on the name SHARED_buffers)? Personally I'm more inclined to think the latter choice, but I've seen references that alluded to both cases, but never a definitive answer. For what its worth, shared_buffers is currently set to 50000 (on a 4G system). Also, effective_cache_size is 125000. max_connections is 256, so I don't want to end up with a possible 100G (50k * 8k * 256) of memory tied up... not that it would be possible, but you never know. I typically never see more than a dozen or so concurrent connections to the db (serving 3 web servers), so I'm thinking of actually using something like pgpool to keep about 10 per web server, rather than use traditional persistent connections of 1 per Apache child, which would probably average about 50 per web server. Thanks. From pgsql-performance-owner@postgresql.org Wed Oct 6 21:20:47 2004 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 5B701329F66 for ; Wed, 6 Oct 2004 21:20:42 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 37473-01 for ; Wed, 6 Oct 2004 20:20:37 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 4DC6F329E89 for ; Wed, 6 Oct 2004 21:20:37 +0100 (BST) 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 i96KK3Do029715; Wed, 6 Oct 2004 16:20:04 -0400 (EDT) To: Greg Stark Cc: =?iso-8859-1?q??= =?iso-8859-1?q? Pierre-Fr�d�ric Caillaud?= , "Ole Langbehn" , pgsql-performance@postgresql.org Subject: Re: sequential scan on select distinct In-reply-to: <87sm8ry5rl.fsf@stark.xeocode.com> References: <200410061130.58625.ole@freiheit.com> <874ql7ztnb.fsf@stark.xeocode.com> <29194.1097091506@sss.pgh.pa.us> <87sm8ry5rl.fsf@stark.xeocode.com> Comments: In-reply-to Greg Stark message dated "06 Oct 2004 16:02:22 -0400" Date: Wed, 06 Oct 2004 16:20:03 -0400 Message-ID: <29712.1097094003@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Amavis-Alert: BAD HEADER Non-encoded 8-bit data (char E9 hex) in message header 'Cc' Cc: ...= =?iso-8859-1?q? Pierre-Fr\351d\351ric Caillaud?... ^ X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/70 X-Sequence-Number: 8546 Greg Stark writes: > But regardless of how uncommon it is, it could be considered important in > another sense: when you need it there really isn't any alternative. It's an > algorithmic improvement with no bound on the performance difference. [ shrug... ] There are an infinite number of special cases for which that claim could be made. The more we load down the planner with seldom-useful special cases, the *lower* the overall performance will be, because we'll waste cycles checking for the special cases in every case ... In this particular case, it's not merely a matter of the planner, either. You'd need some new type of plan node in the executor, so there's a pretty fair amount of added code bulk that will have to be written and then maintained. I'm open to being persuaded that this is worth doing, but the bar is going to be high; I think there are a lot of other more-profitable ways to invest our coding effort and planning cycles. regards, tom lane From pgsql-performance-owner@postgresql.org Wed Oct 6 22:36:19 2004 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 63DCD329EEF for ; Wed, 6 Oct 2004 22:36:17 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 57880-02 for ; Wed, 6 Oct 2004 21:36:09 +0000 (GMT) Received: from vsmtp1.tin.it (vsmtp1.tin.it [212.216.176.141]) by svr1.postgresql.org (Postfix) with ESMTP id BB8EA329EBD for ; Wed, 6 Oct 2004 22:36:07 +0100 (BST) Received: from angus.tin.it (82.53.57.193) by vsmtp1.tin.it (7.0.027) id 4162DC020009E4F5 for pgsql-performance@postgresql.org; Wed, 6 Oct 2004 23:36:07 +0200 Message-Id: <6.1.2.0.2.20041006230239.0201bd40@box.tin.it> X-Sender: angusgb@box.tin.it (Unverified) X-Mailer: QUALCOMM Windows Eudora Version 6.1.2.0 Date: Wed, 06 Oct 2004 23:36:05 +0200 To: pgsql-performance@postgresql.org From: Gabriele Bartolini Subject: Data warehousing requirements Mime-Version: 1.0 Content-Type: multipart/mixed; x-avg-checked=avg-ok-13F04943; boundary="=======395821=======" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.1 tagged_above=0.0 required=5.0 tests=RCVD_IN_RFCI X-Spam-Level: X-Archive-Number: 200410/71 X-Sequence-Number: 8547 --=======395821======= Content-Type: text/plain; x-avg-checked=avg-ok-13F04943; charset=us-ascii; format=flowed Content-Transfer-Encoding: 8bit Hi guys, I just discussed about my problem on IRC. I am building a Web usage mining system based on Linux, PostgreSQL and C++ made up of an OLTP database which feeds several and multi-purpose data warehouses about users' behaviour on HTTP servers. I modelled every warehouse using the star schema, with a fact table and then 'n' dimension tables linked using a surrogate ID. Discussing with the guys of the chat, I came up with these conclusions, regarding the warehouse's performance: 1) don't use referential integrity in the facts table 2) use INTEGER and avoid SMALLINT and NUMERIC types for dimensions' IDs 3) use an index for every dimension's ID in the fact table As far as administration is concerned: run VACUUM ANALYSE daily and VACUUM FULL periodically. Is there anything else I should keep in mind? Also, I was looking for advice regarding hardware requirements for a data warehouse system that needs to satisfy online queries. I have indeed no idea at the moment. I can only predict 4 million about records a month in the fact table, does it make sense or not? is it too much? Data needs to be easily backed up and eventually replicated. Having this in mind, what hardware architecture should I look for? How many hard disks do I need, what kind and what RAID solution do you suggest me to adopt (5 or 10 - I think)? Thank you so much, -Gabriele -- Gabriele Bartolini: Web Programmer, ht://Dig & IWA/HWG Member, ht://Check maintainer Current Location: Prato, Toscana, Italia angusgb@tin.it | http://www.prato.linux.it/~gbartolini | ICQ#129221447 > "Leave every hope, ye who enter!", Dante Alighieri, Divine Comedy, The Inferno --=======395821======= Content-Type: text/plain; charset=us-ascii; x-avg=cert; x-avg-checked=avg-ok-13F04943 Content-Disposition: inline --- Outgoing mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com). Version: 6.0.773 / Virus Database: 520 - Release Date: 05/10/2004 --=======395821=======-- From pgsql-performance-owner@postgresql.org Wed Oct 6 23:28:19 2004 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 970B932A0F8 for ; Wed, 6 Oct 2004 23:28:16 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 73209-05 for ; Wed, 6 Oct 2004 22:28:09 +0000 (GMT) Received: from mail.refractions.net (mail.refractions.net [24.68.236.214]) by svr1.postgresql.org (Postfix) with ESMTP id DED0032A0D0 for ; Wed, 6 Oct 2004 23:28:10 +0100 (BST) Received: from lion.animals (lion [192.168.50.200]) by mail.refractions.net (Postfix) with ESMTP id E59902BEC6; Wed, 6 Oct 2004 15:28:09 -0700 (PDT) Received: by lion.animals (Postfix, from userid 88) id 7400AE3BA; Wed, 6 Oct 2004 15:28:09 -0700 (PDT) Received: from [192.168.50.11] (unknown [192.168.50.11]) by lion.animals (Postfix) with ESMTP id 13D26E3B7; Wed, 6 Oct 2004 15:28:08 -0700 (PDT) Message-ID: <41647127.3060407@refractions.net> Date: Wed, 06 Oct 2004 15:26:47 -0700 From: Paul Ramsey User-Agent: Mozilla Thunderbird 0.7.3 (Windows/20040803) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Doug Y Cc: pgsql-performance@postgresql.org Subject: Re: The never ending quest for clarity on shared_buffers References: <6.1.2.0.2.20041006152747.033bcd50@pop.traderonline.com> In-Reply-To: <6.1.2.0.2.20041006152747.033bcd50@pop.traderonline.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/72 X-Sequence-Number: 8548 Doug Y wrote: > For idle persistent connections, do each of them allocate the memory > specified by this setting (shared_buffers * 8k), or is it one pool used > by all the connection (which seems the logical conclusion based on the > name SHARED_buffers)? Personally I'm more inclined to think the latter > choice, but I've seen references that alluded to both cases, but never a > definitive answer. The shared_buffers are shared (go figure) :). It is all one pool shared by all connections. The sort_mem and vacuum_mem are *per*connection* however, so when allocating that size you have to take into account your expected number of concurrent connections. Paul From pgsql-performance-owner@postgresql.org Thu Oct 7 00:56:02 2004 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 B3A37329ED1 for ; Thu, 7 Oct 2004 00:56:00 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 94940-01 for ; Wed, 6 Oct 2004 23:55:53 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id 7AAD0329D62 for ; Thu, 7 Oct 2004 00:55:55 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i96NtpJm094954 for ; Wed, 6 Oct 2004 23:55:51 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i96NmUQG092858 for pgsql-performance@postgresql.org; Wed, 6 Oct 2004 23:48:30 GMT From: Gaetano Mendola X-Newsgroups: comp.databases.postgresql.performance Subject: Re: Excessive context switching on SMP Xeons Date: Thu, 07 Oct 2004 01:48:22 +0200 Organization: PYRENET Midi-pyrenees Provider Lines: 19 Message-ID: <41648446.5020102@bigfoot.com> References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <41630D50.3020308@lulu.com> <200410051538.51664.josh@agliodbs.com> <41636D8E.1090403@rentec.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Complaints-To: abuse@pyrenet.fr To: Alan Stange User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en In-Reply-To: <41636D8E.1090403@rentec.com> X-Enigmail-Version: 0.86.1.0 X-Enigmail-Supports: pgp-inline, pgp-mime To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/73 X-Sequence-Number: 8549 Alan Stange wrote: > A few quick random observations on the Xeon v. Opteron comparison: [SNIP] > I don't care to go into the whole debate of Xeon v. Opteron here. We > also have a lot of dual Xeon systems. In every comparison I've done with > our codes, the dual Opteron clearly outperforms the dual Xeon, when > running on one and both cpus. Here http://www6.tomshardware.com/cpu/20030422/ both were tested and there is a database performance section, unfortunatelly they used MySQL. Regards Gaetano Mendola From pgsql-performance-owner@postgresql.org Thu Oct 7 02:07:01 2004 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 20996329EE9 for ; Thu, 7 Oct 2004 02:07:00 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 08437-05 for ; Thu, 7 Oct 2004 01:06:52 +0000 (GMT) Received: from sraigw.sra.co.jp (sraigw.sra.co.jp [202.32.10.2]) by svr1.postgresql.org (Postfix) with ESMTP id 372FD329CD9 for ; Thu, 7 Oct 2004 02:06:54 +0100 (BST) Received: from srascb.sra.co.jp (srascb [133.137.8.65]) by sraigw.sra.co.jp (Postfix) with ESMTP id AD88062253; Thu, 7 Oct 2004 10:06:53 +0900 (JST) Received: from srascb.sra.co.jp (localhost [127.0.0.1]) by localhost.sra.co.jp (Postfix) with ESMTP id 9470B10CD06; Thu, 7 Oct 2004 10:06:53 +0900 (JST) Received: from sranhm.sra.co.jp (sranhm [133.137.44.16]) by srascb.sra.co.jp (Postfix) with ESMTP id 7513E10CD04; Thu, 7 Oct 2004 10:06:53 +0900 (JST) Received: from localhost (IDENT:t-ishii@hub-es6.sra.co.jp [133.137.44.7]) by sranhm.sra.co.jp (8.9.3+3.2W/3.7W-srambox) with ESMTP id KAA32531; Thu, 7 Oct 2004 10:06:53 +0900 Date: Thu, 07 Oct 2004 10:08:47 +0900 (JST) Message-Id: <20041007.100847.26272104.t-ishii@sra.co.jp> To: matt@ymogen.net Cc: pg@rbt.ca, awerman2@hotmail.com, scottakirkwood@gmail.com, pgsql-performance@postgresql.org Subject: Re: Caching of Queries From: Tatsuo Ishii In-Reply-To: <012501c4aaf0$f12d0930$8300a8c0@solent> References: <20041004.002823.85416338.t-ishii@sra.co.jp> <012501c4aaf0$f12d0930$8300a8c0@solent> X-Mailer: Mew version 2.3 on Emacs 20.7 / Mule 4.1 =?iso-2022-jp?B?KBskQjAqGyhCKQ==?= 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/74 X-Sequence-Number: 8550 > > I don't know what you are exactly referring to in above URL > > when you are talking about "potential pitfalls of pooling". > > Please explain more. > > Sorry, I wasn't implying that pgpool doesn't deal with the issues, just that > some people aren't necessarily aware of them up front. For instance, pgpool > does an 'abort transaction' and a 'reset all' in lieu of a full reconnect > (of course, since a full reconnect is exactly what we are trying to avoid). > Is this is enough to guarantee that a given pooled connection behaves > exactly as a non-pooled connection would from a client perspective? For > instance, temporary tables are usually dropped at the end of a session, so a > client (badly coded perhaps) that does not already use persistent > connections might be confused when the sequence 'connect, create temp table > foo ..., disconnect, connect, create temp table foo ...' results in the > error 'Relation 'foo' already exists'. First, it's not a particular problem with pgpool. As far as I know any connection pool solution has exactly the same problem. Second, it's easy to fix if PostgreSQL provides a functionarity such as:"drop all temporary tables if any". I think we should implement it if we agree that connection pooling should be implemented outside the PostgreSQL engine itself. I think cores agree with this. -- Tatsuo Ishii From pgsql-performance-owner@postgresql.org Thu Oct 7 02:30:07 2004 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 C47A132A5DA for ; Thu, 7 Oct 2004 02:30:06 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 15330-01 for ; Thu, 7 Oct 2004 01:29:53 +0000 (GMT) Received: from sue.samurai.com (sue.samurai.com [205.207.28.74]) by svr1.postgresql.org (Postfix) with ESMTP id 0F0A132A5ED for ; Thu, 7 Oct 2004 02:29:56 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by sue.samurai.com (Postfix) with ESMTP id 18382197CE; Wed, 6 Oct 2004 21:29:57 -0400 (EDT) Received: from sue.samurai.com ([127.0.0.1]) by localhost (sue.samurai.com [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 54699-01-3; Wed, 6 Oct 2004 21:29:55 -0400 (EDT) Received: from localhost (unknown [61.88.101.19]) by sue.samurai.com (Postfix) with ESMTP id 6B1BE197A8; Wed, 6 Oct 2004 21:29:53 -0400 (EDT) Subject: Re: The never ending quest for clarity on shared_buffers From: Neil Conway To: Paul Ramsey Cc: Doug Y , pgsql-performance@postgresql.org In-Reply-To: <41647127.3060407@refractions.net> References: <6.1.2.0.2.20041006152747.033bcd50@pop.traderonline.com> <41647127.3060407@refractions.net> Content-Type: text/plain Message-Id: <1097112547.13119.108.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Thu, 07 Oct 2004 11:29:07 +1000 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/75 X-Sequence-Number: 8551 On Thu, 2004-10-07 at 08:26, Paul Ramsey wrote: > The shared_buffers are shared (go figure) :). It is all one pool shared > by all connections. Yeah, I thought this was pretty clear. Doug, can you elaborate on where you saw the misleading docs? > The sort_mem and vacuum_mem are *per*connection* however, so when > allocating that size you have to take into account your > expected number of concurrent connections. Allocations of size `sort_mem' can actually can actually happen several times within a *single* connection (if the query plan happens to involve a number of sort steps or hash tables) -- the limit is on the amount of memory that will be used for a single sort/hash table. So choosing the right figure is actually a little more complex than that. -Neil From pgsql-performance-owner@postgresql.org Thu Oct 7 06:47:58 2004 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 05A6132AB16 for ; Thu, 7 Oct 2004 06:40:04 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 58139-02 for ; Thu, 7 Oct 2004 05:38:46 +0000 (GMT) Received: from svr4.postgresql.org (svr4.postgresql.org [66.98.251.159]) by svr1.postgresql.org (Postfix) with ESMTP id E1DD332A4B9 for ; Thu, 7 Oct 2004 06:38:46 +0100 (BST) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr4.postgresql.org (Postfix) with ESMTP id 519165AFE00 for ; Thu, 7 Oct 2004 04:18:33 +0000 (GMT) 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 i973CEMt023363; Wed, 6 Oct 2004 23:12:14 -0400 (EDT) To: Tatsuo Ishii Cc: matt@ymogen.net, pg@rbt.ca, awerman2@hotmail.com, scottakirkwood@gmail.com, pgsql-performance@postgresql.org Subject: Re: Caching of Queries In-reply-to: <20041007.100847.26272104.t-ishii@sra.co.jp> References: <20041004.002823.85416338.t-ishii@sra.co.jp> <012501c4aaf0$f12d0930$8300a8c0@solent> <20041007.100847.26272104.t-ishii@sra.co.jp> Comments: In-reply-to Tatsuo Ishii message dated "Thu, 07 Oct 2004 10:08:47 +0900" Date: Wed, 06 Oct 2004 23:12:14 -0400 Message-ID: <23362.1097118734@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/77 X-Sequence-Number: 8553 Tatsuo Ishii writes: > First, it's not a particular problem with pgpool. As far as I know any > connection pool solution has exactly the same problem. Second, it's > easy to fix if PostgreSQL provides a functionarity such as:"drop all > temporary tables if any". I don't like that definition exactly --- it would mean that every time we add more backend-local state, we expect client drivers to know to issue the right incantation to reset that kind of state. I'm thinking we need to invent a command like "RESET CONNECTION" that resets GUC variables, drops temp tables, forgets active NOTIFYs, and generally does whatever else needs to be done to make the session state appear virgin. When we add more such state, we can fix it inside the backend without bothering clients. I now realize that our "RESET ALL" command for GUC variables was not fully thought out. We could possibly redefine it as doing the above, but that might break some applications ... regards, tom lane From pgsql-performance-owner@postgresql.org Thu Oct 7 05:14:01 2004 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 97FCA32AE37 for ; Thu, 7 Oct 2004 05:13:56 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 42791-09 for ; Thu, 7 Oct 2004 04:13:17 +0000 (GMT) Received: from wren.rentec.com (unknown [65.213.84.9]) by svr1.postgresql.org (Postfix) with ESMTP id 4F1B932A5C7 for ; Thu, 7 Oct 2004 05:06:52 +0100 (BST) Received: from [172.16.160.107] (stange-dhcp1.rentec.com [172.16.160.107]) by wren.rentec.com (8.12.9/8.12.1) with ESMTP id i9745LBD002387; Thu, 7 Oct 2004 00:05:21 -0400 (EDT) Message-ID: <4164B48C.9060202@rentec.com> Date: Wed, 06 Oct 2004 23:14:20 -0400 From: Alan Stange User-Agent: Mozilla Thunderbird 0.7 (X11/20040615) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Gaetano Mendola Cc: pgsql-performance@postgresql.org Subject: Re: Excessive context switching on SMP Xeons References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <41630D50.3020308@lulu.com> <200410051538.51664.josh@agliodbs.com> <41636D8E.1090403@rentec.com> <41648446.5020102@bigfoot.com> In-Reply-To: <41648446.5020102@bigfoot.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/76 X-Sequence-Number: 8552 Here's a few numbers from the Opteron 250. If I get some time I'll post a more comprehensive comparison including some other systems. The system is a Sun v20z. Dual Opteron 250, 2.4Ghz, Linux 2.6, 8 GB memory. I did a compile and install of pg 8.0 beta 3. I created a data base on a tmpfs file system and ran pgbench. Everything was "out of the box", meaning I did not tweak any config files. I used this for pgbench: $ pgbench -i -s 32 and this for pgbench invocations: $ pgbench -s 32 -c 1 -t 10000 -v clients tps 1 1290 2 1780 4 1760 8 1680 16 1376 32 904 How are these results useful? In some sense, this is a speed of light number for the Opteron 250. You'll never go faster on this system with a real storage subsystem involved instead of a tmpfs file system. It's also a set of numbers that anyone else can reproduce as we don't have to deal with any differences in file systems, disk subsystems, networking, etc. Finally, it's a set of results that anyone else can compute on Xeon's or other systems and make a simple (and naive) comparisons. Just to stay on topic: vmstat reported about 30K cs / second while this was running the 1 and 2 client cases. -- Alan From pgsql-performance-owner@postgresql.org Thu Oct 7 08:46:30 2004 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 4EE0032ABB9 for ; Thu, 7 Oct 2004 08:46:28 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78834-10 for ; Thu, 7 Oct 2004 07:45:55 +0000 (GMT) Received: from sraigw.sra.co.jp (sraigw.sra.co.jp [202.32.10.2]) by svr1.postgresql.org (Postfix) with ESMTP id 7F0CB32A822 for ; Thu, 7 Oct 2004 08:44:58 +0100 (BST) Received: from srascb.sra.co.jp (srascb [133.137.8.65]) by sraigw.sra.co.jp (Postfix) with ESMTP id EAA2062072; Thu, 7 Oct 2004 16:16:48 +0900 (JST) Received: from srascb.sra.co.jp (localhost [127.0.0.1]) by localhost.sra.co.jp (Postfix) with ESMTP id CA96010CD06; Thu, 7 Oct 2004 16:16:48 +0900 (JST) Received: from sranhm.sra.co.jp (sranhm [133.137.44.16]) by srascb.sra.co.jp (Postfix) with ESMTP id A55F110CD04; Thu, 7 Oct 2004 16:16:48 +0900 (JST) Received: from localhost (IDENT:t-ishii@srapc2345.sra.co.jp [133.137.44.184]) by sranhm.sra.co.jp (8.9.3+3.2W/3.7W-srambox) with ESMTP id QAA15134; Thu, 7 Oct 2004 16:16:48 +0900 Date: Thu, 07 Oct 2004 16:18:42 +0900 (JST) Message-Id: <20041007.161842.77062117.t-ishii@sra.co.jp> To: tgl@sss.pgh.pa.us Cc: matt@ymogen.net, pg@rbt.ca, awerman2@hotmail.com, scottakirkwood@gmail.com, pgsql-performance@postgresql.org Subject: Re: Caching of Queries From: Tatsuo Ishii In-Reply-To: <23362.1097118734@sss.pgh.pa.us> References: <012501c4aaf0$f12d0930$8300a8c0@solent> <20041007.100847.26272104.t-ishii@sra.co.jp> <23362.1097118734@sss.pgh.pa.us> X-Mailer: Mew version 2.3 on Emacs 20.7 / Mule 4.1 =?iso-2022-jp?B?KBskQjAqGyhCKQ==?= 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/78 X-Sequence-Number: 8554 > Tatsuo Ishii writes: > > First, it's not a particular problem with pgpool. As far as I know any > > connection pool solution has exactly the same problem. Second, it's > > easy to fix if PostgreSQL provides a functionarity such as:"drop all > > temporary tables if any". > > I don't like that definition exactly --- it would mean that every time > we add more backend-local state, we expect client drivers to know to > issue the right incantation to reset that kind of state. > > I'm thinking we need to invent a command like "RESET CONNECTION" that > resets GUC variables, drops temp tables, forgets active NOTIFYs, and > generally does whatever else needs to be done to make the session state > appear virgin. When we add more such state, we can fix it inside the > backend without bothering clients. Great. It's much better than I propose. > I now realize that our "RESET ALL" command for GUC variables was not > fully thought out. We could possibly redefine it as doing the above, > but that might break some applications ... > > regards, tom lane > From pgsql-performance-owner@postgresql.org Thu Oct 7 12:43:44 2004 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 09FAF32A1A9 for ; Thu, 7 Oct 2004 12:43:41 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 26108-06 for ; Thu, 7 Oct 2004 11:42:47 +0000 (GMT) Received: from hotmail.com (bay18-dav8.bay18.hotmail.com [65.54.187.188]) by svr1.postgresql.org (Postfix) with ESMTP id C9F4732AAE7 for ; Thu, 7 Oct 2004 12:30:37 +0100 (BST) Received: from mail pickup service by hotmail.com with Microsoft SMTPSVC; Thu, 7 Oct 2004 04:30:01 -0700 Received: from 67.81.98.198 by BAY18-DAV8.phx.gbl with DAV; Thu, 07 Oct 2004 11:29:53 +0000 X-Originating-IP: [67.81.98.198] X-Originating-Email: [awerman2@hotmail.com] X-Sender: awerman2@hotmail.com From: "Aaron Werman" To: , "Gabriele Bartolini" References: <6.1.2.0.2.20041006230239.0201bd40@box.tin.it> Subject: Re: Data warehousing requirements Date: Thu, 7 Oct 2004 07:30:07 -0400 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 6.00.2800.1437 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 Message-ID: X-OriginalArrivalTime: 07 Oct 2004 11:30:01.0026 (UTC) FILETIME=[FB8B3A20:01C4AC60] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/79 X-Sequence-Number: 8555 Consider how the fact table is going to be used, and review hacking it up based on usage. Fact tables should be fairly narrow, so if there are extra columns beyond keys and dimension keys consider breaking it into parallel tables (vertical partitioning). Horizontal partitioning is your friend; especially if it is large - consider slicing the data into chunks. If the fact table is date driven it might be worthwhile to break it into separate tables based on date key. This wins in reducing the working set of queries and in buffering. If there is a real hotspot, such as current month's activity, you might want to keep a separate table with just the (most) active data.Static tables of unchanged data can simplify backups, etc., as well. Consider summary tables if you know what type of queries you'll hit. Especially here, MVCC is not your friend because it has extra work to do for aggregate functions. Cluster helps if you bulk load. In most warehouses, the data is downstream data from existing operational systems. Because of that you're not able to use database features to preserve integrity. In most cases, the data goes through an extract/transform/load process - and the output is considered acceptable. So, no RI is correct for star or snowflake design. Pretty much no anything else that adds intelligence - no triggers, no objects, no constraints of any sort. Many designers try hard to avoid nulls. On the hardware side - RAID5 might work here because of the low volume if you can pay the write performance penalty. To size hardware you need to estimate load in terms of transaction type (I usually make bucket categories of small, medium, and large effort needs) and transaction rate. Then try to estimate how much CPU and I/O they'll use. /Aaron "Let us not speak of them; but look, and pass on." ----- Original Message ----- From: "Gabriele Bartolini" To: Sent: Wednesday, October 06, 2004 5:36 PM Subject: [PERFORM] Data warehousing requirements > Hi guys, > > I just discussed about my problem on IRC. I am building a Web usage > mining system based on Linux, PostgreSQL and C++ made up of an OLTP > database which feeds several and multi-purpose data warehouses about users' > behaviour on HTTP servers. > > I modelled every warehouse using the star schema, with a fact table and > then 'n' dimension tables linked using a surrogate ID. > > Discussing with the guys of the chat, I came up with these conclusions, > regarding the warehouse's performance: > > 1) don't use referential integrity in the facts table > 2) use INTEGER and avoid SMALLINT and NUMERIC types for dimensions' IDs > 3) use an index for every dimension's ID in the fact table > > As far as administration is concerned: run VACUUM ANALYSE daily and > VACUUM FULL periodically. > > Is there anything else I should keep in mind? > > Also, I was looking for advice regarding hardware requirements for a > data warehouse system that needs to satisfy online queries. I have indeed > no idea at the moment. I can only predict 4 million about records a month > in the fact table, does it make sense or not? is it too much? > > Data needs to be easily backed up and eventually replicated. > > Having this in mind, what hardware architecture should I look for? How > many hard disks do I need, what kind and what RAID solution do you suggest > me to adopt (5 or 10 - I think)? > > Thank you so much, > -Gabriele > -- > Gabriele Bartolini: Web Programmer, ht://Dig & IWA/HWG Member, ht://Check > maintainer > Current Location: Prato, Toscana, Italia > angusgb@tin.it | http://www.prato.linux.it/~gbartolini | ICQ#129221447 > > "Leave every hope, ye who enter!", Dante Alighieri, Divine Comedy, The > Inferno > ---------------------------------------------------------------------------- ---- > > --- > Outgoing mail is certified Virus Free. > Checked by AVG anti-virus system (http://www.grisoft.com). > Version: 6.0.773 / Virus Database: 520 - Release Date: 05/10/2004 > ---------------------------------------------------------------------------- ---- > > ---------------------------(end of broadcast)--------------------------- > TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org > From pgsql-performance-owner@postgresql.org Thu Oct 7 13:00:25 2004 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 7490832A2F7 for ; Thu, 7 Oct 2004 13:00:22 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 28992-05 for ; Thu, 7 Oct 2004 11:59:37 +0000 (GMT) Received: from boutiquenumerique.com (gailleton-2-82-67-9-10.fbx.proxad.net [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id AEDED32A009 for ; Thu, 7 Oct 2004 12:58:08 +0100 (BST) Received: (qmail 27673 invoked from network); 7 Oct 2004 13:58:30 +0200 Received: from unknown (HELO musicbox) (192.168.0.2) by gailleton-2-82-67-9-10.fbx.proxad.net with SMTP; 7 Oct 2004 13:58:30 +0200 In-Reply-To: <29194.1097091506@sss.pgh.pa.us> Organization: =?iso-8859-15?Q?La_Boutique_Num=E9rique?= References: <200410061130.58625.ole@freiheit.com> <874ql7ztnb.fsf@stark.xeocode.com> <29194.1097091506@sss.pgh.pa.us> Subject: Re: sequential scan on select distinct To: pgsql-performance@postgresql.org Message-ID: From: =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= Content-Type: text/plain; format=flowed; delsp=yes; charset=iso-8859-15 MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Date: Thu, 07 Oct 2004 14:01:13 +0200 User-Agent: Opera M2/7.53 (Linux, build 737) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/80 X-Sequence-Number: 8556 > I don't really think it would be a useful plan anyway. What *would* be > useful is to support HashAggregate as an implementation alternative for > DISTINCT --- currently I believe we only consider that for GROUP BY. > The DISTINCT planning code is fairly old and crufty and hasn't been > redesigned lately. > > regards, tom lane I see this as a minor annoyance only because I can write GROUP BY instead of DISTINCT and get the speed boost. It probably annoys people trying to port applications to postgres though, forcing them to rewrite their queries. * SELECT DISTINCT : 21442.296 ms (by default, uses an index scan) disabling index_scan => Sort + Unique : 14512.105 ms * GROUP BY : 1793.651 ms using HashAggregate * skip index scan by function : 13.833 ms The HashAggregate speed boost is good, but rather pathetic compared to a "skip index scan" ; but it's still worth having if updating the DISTINCT code is easy. Note that it would also benefit UNION queries which apparently use DISTINCT internally and currently produce this : ------------------------------------------------------------------------------ explain analyze select number from ((select number from dummy) union (select number from dummy)) as foo; Subquery Scan foo (cost=287087.62..317087.62 rows=2000000 width=4) (actual time=33068.776..35575.330 rows=255 loops=1) -> Unique (cost=287087.62..297087.62 rows=2000000 width=4) (actual time=33068.763..35574.126 rows=255 loops=1) -> Sort (cost=287087.62..292087.62 rows=2000000 width=4) (actual time=33068.757..34639.180 rows=2000000 loops=1) Sort Key: number -> Append (cost=0.00..49804.00 rows=2000000 width=4) (actual time=0.055..7412.551 rows=2000000 loops=1) -> Subquery Scan "*SELECT* 1" (cost=0.00..24902.00 rows=1000000 width=4) (actual time=0.054..3104.165 rows=1000000 loops=1) -> Seq Scan on dummy (cost=0.00..14902.00 rows=1000000 width=4) (actual time=0.051..1792.348 rows=1000000 loops=1) -> Subquery Scan "*SELECT* 2" (cost=0.00..24902.00 rows=1000000 width=4) (actual time=0.048..3034.462 rows=1000000 loops=1) -> Seq Scan on dummy (cost=0.00..14902.00 rows=1000000 width=4) (actual time=0.044..1718.682 rows=1000000 loops=1) Total runtime: 36265.662 ms ------------------------------------------------------------------------------ But could instead do this : explain analyze select number from ((select number from dummy) union all (select number from dummy)) as foo group by number; HashAggregate (cost=74804.00..74804.00 rows=200 width=4) (actual time=10753.648..10753.890 rows=255 loops=1) -> Subquery Scan foo (cost=0.00..69804.00 rows=2000000 width=4) (actual time=0.059..8992.084 rows=2000000 loops=1) -> Append (cost=0.00..49804.00 rows=2000000 width=4) (actual time=0.055..6688.639 rows=2000000 loops=1) -> Subquery Scan "*SELECT* 1" (cost=0.00..24902.00 rows=1000000 width=4) (actual time=0.054..2749.708 rows=1000000 loops=1) -> Seq Scan on dummy (cost=0.00..14902.00 rows=1000000 width=4) (actual time=0.052..1640.427 rows=1000000 loops=1) -> Subquery Scan "*SELECT* 2" (cost=0.00..24902.00 rows=1000000 width=4) (actual time=0.038..2751.916 rows=1000000 loops=1) -> Seq Scan on dummy (cost=0.00..14902.00 rows=1000000 width=4) (actual time=0.034..1637.818 rows=1000000 loops=1) Total runtime: 10754.120 ms ------------------------------------------------------------------------------ A 3x speedup, but still a good thing to have. When I LIMIT the two subqueries to 100k rows instead of a million, the times are about equal. When I LIMIT one of the subqueries to 100k and leave the other to 1M, UNION ALL 17949.609 ms UNION + GROUP BY 6130.417 ms Still some performance to be gained... ------------------------------------------------------------------------------ Of course it can't use a skip index scan on a subquery, but I could instead : I know it's pretty stupid to use the same table twice but it's just an example. However, if you think about table partitions and views, a "select distinct number" from a view having multiple partitions would yield this type of query, and that table partitioning seems like a hot subject lately. let's create a dummy example view : create view dummy_view as (select * from dummy) union all (select * from dummy); explain analyze select number from dummy_view group by number; HashAggregate (cost=74804.00..74804.00 rows=200 width=4) (actual time=10206.456..10206.713 rows=255 loops=1) -> Subquery Scan dummy_view (cost=0.00..69804.00 rows=2000000 width=4) (actual time=0.060..8431.776 rows=2000000 loops=1) -> Append (cost=0.00..49804.00 rows=2000000 width=8) (actual time=0.055..6122.125 rows=2000000 loops=1) -> Subquery Scan "*SELECT* 1" (cost=0.00..24902.00 rows=1000000 width=8) (actual time=0.054..2456.566 rows=1000000 loops=1) -> Seq Scan on dummy (cost=0.00..14902.00 rows=1000000 width=8) (actual time=0.048..1107.151 rows=1000000 loops=1) -> Subquery Scan "*SELECT* 2" (cost=0.00..24902.00 rows=1000000 width=8) (actual time=0.036..2471.748 rows=1000000 loops=1) -> Seq Scan on dummy (cost=0.00..14902.00 rows=1000000 width=8) (actual time=0.031..1104.482 rows=1000000 loops=1) Total runtime: 10206.945 ms A smarter planner could rewrite it into this : select number from ((select distinct number from dummy) union (select distinct number from dummy)) as foo; and notice it would index-skip-scan the two partitions (here, example with my function) explain analyze select number from ((select sel_distinct as number from sel_distinct()) union all (select sel_distinct as number from sel_distinct())) as foo group by number; HashAggregate (cost=70.00..70.00 rows=200 width=4) (actual time=29.078..29.332 rows=255 loops=1) -> Subquery Scan foo (cost=0.00..65.00 rows=2000 width=4) (actual time=13.378..28.587 rows=510 loops=1) -> Append (cost=0.00..45.00 rows=2000 width=4) (actual time=13.373..28.003 rows=510 loops=1) -> Subquery Scan "*SELECT* 1" (cost=0.00..22.50 rows=1000 width=4) (actual time=13.373..13.902 rows=255 loops=1) -> Function Scan on sel_distinct (cost=0.00..12.50 rows=1000 width=4) (actual time=13.367..13.619 rows=255 loops=1) -> Subquery Scan "*SELECT* 2" (cost=0.00..22.50 rows=1000 width=4) (actual time=13.269..13.800 rows=255 loops=1) -> Function Scan on sel_distinct (cost=0.00..12.50 rows=1000 width=4) (actual time=13.263..13.512 rows=255 loops=1) Total runtime: 29.569 ms So, if a query with UNION or UNION ALL+DISTINCT tries to put DISTINCT inside the subqueries and yields an index skip scan, here is a massive speedup. You will tell me "but if the UNION ALL has 10 subqueries, planning is going to take forever !" Well not necessarily. The above query with 10 subqueries UNIONALLed then GROUPed takes : UNION : 320509.522 ms (the Sort + Unique truly becomes humongous). UNION ALL + GROUP : 54586.759 ms (you see there is already interest in rewiring DISTINCT/UNION) skip scan + UNION : 147.941 ms skip scan + UNION ALL + group : 147.313 ms > Well it would clearly be useful in this test case, where has a small > number of distinct values in a large table, and an index on the column. > His plpgsql function that emulates such a plan is an order of magnitude > faster than the hash aggregate plan even though it has to do entirely > separate index scans for each key value. Actually, it is more like two orders of magnitude (100x faster) : in fact the time for a seq scan is O(N rows) whereas the time for the skip index scan should be, if I'm not mistaken, something like O((N distinct values) * (log N rows)) ; in my case there are 256 distinct values for 1M rows and a speedup of 100x, so if there were 10M rows the speedup would be like 300x (depending on the base of the log which I assume is 2). And if the skip index scan is implemented in postgres instead of in a function, it could be much, much faster... > [ shrug... ] There are an infinite number of special cases for which > that claim could be made. The more we load down the planner with > seldom-useful special cases, the *lower* the overall performance will > be, because we'll waste cycles checking for the special cases in every > case ... In a general way, you are absolutely right... special-casing a case for a speedup of 2x for instance would be worthless... but we are considering a HUGE speedup here. And, if this mode is only used for DISTINCT and GROUP BY queries, no planning cycles will be wasted at all on queries which do not use DISTINCT nor GROUP BY. Present state is that DISTINCT and UNION are slow with or without using the GROUP BY trick. Including the index skip scan in the planning options would only happen when appropriate cases are detected. This detection would be very fast. The index skip scan would then speed up the query so much that the additional planning cost would not matter. If there are many distinct values, so that seq scan is faster than skip scan, the query will be slow enough anyway so that the additional planning cost does not matter. The only problem cases are queries with small tables where startup time is important, but in that case the planner has stats about the number of rows in the table, and again excluding skip scan from the start would be fast. Lateral thought : Create a new index type which only indexes one row for each value. This index would use very little space and would be very fast to update (on my table it would index only 256 values). Keep the Index Scan code and all, but use this index type when you can. This solution is less general and also has a few drawbacks. Another thought : \d dummy Table �public.dummy� Colonne | Type | Modificateurs ---------+---------+--------------- id | integer | number | integer | Index : �dummy_idx� btree (number) �dummy_idx_2� btree (number, id) explain analyze select * from dummy where id=1; Seq Scan on dummy (cost=0.00..17402.00 rows=1 width=8) (actual time=274.480..1076.092 rows=1 loops=1) Filter: (id = 1) Total runtime: 1076.168 ms explain analyze select * from dummy where number between 0 and 256 and id=1; Index Scan using dummy_idx_2 on dummy (cost=0.00..6.02 rows=1 width=8) (actual time=1.449..332.020 rows=1 loops=1) Index Cond: ((number >= 0) AND (number <= 256) AND (id = 1)) Total runtime: 332.112 ms In this case we have no index on id, but using a skip index scan, emulated by the "between" to force use of the (number,id) index, even though it must look in all the 256 possible values for number, still speeds it up by 3x. Interestingly, with only 16 distinct values, the time is quite the same. Thus, the "skip index scan" could be used in cases where there is a multicolumn index, but the WHERE misses a column. This would not waste planning cycles because : - If the index we need exists and there is no "distinct" or "group by" without aggregate, the planner does not even consider using the skip index scan. - If the index we need does not exist, the planner only loses the cycles needed to check if there is a multicolumn index which may be used. In this case, either there is no such index, and a seq scan is chosen, which will be slow, so the time wasted for the check is negligible ; or an index is found and can be used, and the time gained by the skip index scan is well amortized. Currently one has to carefully consider which queries will be used frequently and need indexes, and which ones are infrequent and don't justify an index (but these queries will be very slow). With the skip index scan, these less frequent queries don't always mean a seq scan. Thus people will need to create less infrequently used indexes, and will have a higher INSERT/UPDATE speed/ ------------------------------------------------------------------------------------------------ The skip scan would also be a winner on this type of query which is a killer, a variant of the famous 'TOP 10' query : EXPLAIN SELECT max(id), number FROM dummy GROUP BY number; -> 2229.141 ms Postgres uses a Seq scan + HashAggregate. Come on, we have an index btree (number, id), use it ! A simple customization on my skip scan emulation function takes 13.683 ms... I know that Postgres does not translate max() on on indexed column to ORDER BY column DESC LIMIT 1, because it would be extremely hard to implement due to the general nature of aggregates which is a very good thing. It does not bother me because I can still write ORDER BY column DESC LIMIT 1. Where it does bother me is if I want the highest ID from each number, which can only be expressed by SELECT max(id), number FROM dummy GROUP BY number; and not with LIMITs. Suppose I want the first 10 higher id's for each number, which is another variant on the "killer top 10 query". I'm stuck, I cannot even use max(), I have to write a custom aggregate which would keep the 10 highest values, which would be very slow, so I have to use my function and put a LIMIT 10 instead of a LIMIT 1 in each query, along with a FOR and some other conditions to check if there are less than 10 id's for a number, etc, which more or less amounts to "select the next number, then select the associated id's". It'll still be fast a lot faster than seq scan, but it gets more and more complicated. However I'd like to write : select number,id from dummy ORDER BY number DESC, id DESC MULTILIMIT 50,10; The MULTILIMIT means "I want 50 numbers and 10 id's for each number." MULTILIMIT NULL,10 would mean "I want all numbers and 10 id's for each number." NULL is not mandatory, it could also be -1, a keyword or something. MULTILIMIT could simply be LIMIT too, because LIMIT takes one parameter. The OFFSET clause could also evolve accordingly. And this would naturally use a skip index scan, and benefit a whole class of queries which have traditionnaly been difficult to get right... Conclusion : smarting up the DISTINCT planner has the following benefits : - speedup on DISTINCT - speedup on UNION which seems to use DISTINCT internally index skip scan has the following benefits : - massive speedup (x100) on queries involving DISTINCT or its GROUP BY variant - same thing (x300) on UNION queries if the parser tries to rewrite the query and put the DISTINCT inside the subqueries - paves the way for a MULTILIMIT which gives an elegant, and very efficient way of expressing traditionnaly difficult queries like the "Top 10 by category" which are used quite often and give headaches to dba's. - Possibility to use a multicolumn index with a WHERE not including all left columns index skip scan has the following drawbacks : - more complexity - additional planning time This last drawback is in fact, limited because : - It is easy and fast to know when the index skip scan will never be used, so in most queries which won't need it, the possibility can be eliminated without wasting cycles in planning - When it is used, the performance gains are so massive that it is justified - People who use many queries where planning time is significant comparing to execution time are probably using SQL functions or prepared queries. Enough arguments, maybe not to convince you, but to have a second thought on it ? --------------------------------------------------------------- Side Note : What do you think about the idea of an "UniqueSort" which would do sort+unique in one pass ? This could also be simple to code, and would also offer advantages to all queries using UNION. The sort would be faster and consume less storage space because the data size would diminish as duplicates are eliminated along the way. From pgsql-performance-owner@postgresql.org Thu Oct 7 16:35:48 2004 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 6B2A232A0D0 for ; Thu, 7 Oct 2004 16:35:46 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 06019-02 for ; Thu, 7 Oct 2004 15:35:37 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id A9C3A32A3D2 for ; Thu, 7 Oct 2004 16:35:37 +0100 (BST) 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 i97FZZS9029075; Thu, 7 Oct 2004 11:35:35 -0400 (EDT) To: =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= Cc: pgsql-performance@postgresql.org Subject: Re: sequential scan on select distinct In-reply-to: References: <200410061130.58625.ole@freiheit.com> <874ql7ztnb.fsf@stark.xeocode.com> <29194.1097091506@sss.pgh.pa.us> Comments: In-reply-to =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= message dated "Thu, 07 Oct 2004 14:01:13 +0200" Date: Thu, 07 Oct 2004 11:35:34 -0400 Message-ID: <29074.1097163334@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/81 X-Sequence-Number: 8557 =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= writes: > Present state is that DISTINCT and UNION are slow with or without using > the GROUP BY trick. Including the index skip scan in the planning options > would only happen when appropriate cases are detected. This detection > would be very fast. You have no basis whatever for making that last assertion; and since it's the critical point, I don't intend to let you slide by without backing it up. I think that looking for relevant indexes would be nontrivial; the more so in cases like you've been armwaving about just above, where you have to find a relevant index for each of several subqueries. The fact that the optimization wins a lot when it wins is agreed, but the time spent trying to apply it when it doesn't work is a cost that has to be set against that. I don't accept your premise that every query for which skip-index isn't relevant is so slow that planning time does not matter. regards, tom lane From pgsql-performance-owner@postgresql.org Thu Oct 7 16:48:57 2004 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 6C58D32A0F8 for ; Thu, 7 Oct 2004 16:48:55 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 10588-02 for ; Thu, 7 Oct 2004 15:48:52 +0000 (GMT) Received: from snowwhite.lulu.com (mail.lulu.com [66.193.5.50]) by svr1.postgresql.org (Postfix) with ESMTP id 9B8B132A04B for ; Thu, 7 Oct 2004 16:48:49 +0100 (BST) Received: from [10.0.0.32] (meatwad.rdu.lulu.com [10.0.0.32]) (authenticated bits=0) by snowwhite.lulu.com (8.12.8/8.12.8) with ESMTP id i97FnVbO022152 (version=TLSv1/SSLv3 cipher=RC4-MD5 bits=128 verify=NO); Thu, 7 Oct 2004 11:49:32 -0400 Message-ID: <41656559.70100@lulu.com> Date: Thu, 07 Oct 2004 11:48:41 -0400 From: Bill Montgomery User-Agent: Mozilla Thunderbird 0.7.3 (X11/20040803) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Alan Stange Cc: Gaetano Mendola , pgsql-performance@postgresql.org Subject: Re: Excessive context switching on SMP Xeons References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <41630D50.3020308@lulu.com> <200410051538.51664.josh@agliodbs.com> <41636D8E.1090403@rentec.com> <41648446.5020102@bigfoot.com> <4164B48C.9060202@rentec.com> In-Reply-To: <4164B48C.9060202@rentec.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/82 X-Sequence-Number: 8558 Alan Stange wrote: > Here's a few numbers from the Opteron 250. If I get some time I'll > post a more comprehensive comparison including some other systems. > > The system is a Sun v20z. Dual Opteron 250, 2.4Ghz, Linux 2.6, 8 GB > memory. I did a compile and install of pg 8.0 beta 3. I created a > data base on a tmpfs file system and ran pgbench. Everything was "out > of the box", meaning I did not tweak any config files. > > I used this for pgbench: > $ pgbench -i -s 32 > > and this for pgbench invocations: > $ pgbench -s 32 -c 1 -t 10000 -v > > > clients tps 1 1290 2 > 1780 4 1760 8 1680 > 16 1376 32 904 The same test on a Dell PowerEdge 1750, Dual Xeon 3.2 GHz, 512k cache, HT on, Linux 2.4.21-20.ELsmp (RHEL 3), 4GB memory, pg 7.4.5: $ pgbench -i -s 32 pgbench $ pgbench -s 32 -c 1 -t 10000 -v clients tps avg CS/sec ------- ----- ---------- 1 601 48,000 2 889 77,000 4 1006 80,000 8 985 59,000 16 966 47,000 32 913 46,000 Far less performance that the Dual Opterons with a low number of clients, but the gap narrows as the number of clients goes up. Anyone smarter than me care to explain? Anyone have a 4-way Opteron to run the same benchmark on? -Bill > How are these results useful? In some sense, this is a speed of light > number for the Opteron 250. You'll never go faster on this system > with a real storage subsystem involved instead of a tmpfs file > system. It's also a set of numbers that anyone else can reproduce as > we don't have to deal with any differences in file systems, disk > subsystems, networking, etc. Finally, it's a set of results that > anyone else can compute on Xeon's or other systems and make a simple > (and naive) comparisons. > > > Just to stay on topic: vmstat reported about 30K cs / second while > this was running the 1 and 2 client cases. > > -- Alan From pgsql-performance-owner@postgresql.org Thu Oct 7 18:11:26 2004 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 950C6329D41 for ; Thu, 7 Oct 2004 18:11:23 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 37432-04 for ; Thu, 7 Oct 2004 17:11:13 +0000 (GMT) Received: from vsmtp3.tin.it (vsmtp3alice.tin.it [212.216.176.143]) by svr1.postgresql.org (Postfix) with ESMTP id 72F8B329E5F for ; Thu, 7 Oct 2004 18:11:07 +0100 (BST) Received: from angus.tin.it (82.51.64.122) by vsmtp3.tin.it (7.0.027) id 414B175C008A43F4; Thu, 7 Oct 2004 19:07:07 +0200 Message-Id: <6.1.2.0.2.20041007183909.0201f310@box.tin.it> X-Sender: angusgb@box.tin.it (Unverified) X-Mailer: QUALCOMM Windows Eudora Version 6.1.2.0 Date: Thu, 07 Oct 2004 19:07:04 +0200 To: "Aaron Werman" , From: Gabriele Bartolini Subject: Re: Data warehousing requirements In-Reply-To: References: <6.1.2.0.2.20041006230239.0201bd40@box.tin.it> Mime-Version: 1.0 Content-Type: multipart/mixed; x-avg-checked=avg-ok-16C74A51; boundary="=======7EC77ACB=======" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.1 tagged_above=0.0 required=5.0 tests=RCVD_IN_RFCI X-Spam-Level: X-Archive-Number: 200410/84 X-Sequence-Number: 8560 --=======7EC77ACB======= Content-Type: text/plain; x-avg-checked=avg-ok-16C74A51; charset=us-ascii; format=flowed Content-Transfer-Encoding: 8bit At 13.30 07/10/2004, Aaron Werman wrote: >Consider how the fact table is going to be used, and review hacking it up >based on usage. Fact tables should be fairly narrow, so if there are extra >columns beyond keys and dimension keys consider breaking it into parallel >tables (vertical partitioning). Hmm ... I have only an extra column. Sorry if I ask you to confirm this, but practically vertical partitioning allows me to divide a table into 2 tables (like if I cut them vertically, right?) having the same key. If I had 2 extra columns, that could be the case, couldn't it? >Horizontal partitioning is your friend; especially if it is large - consider >slicing the data into chunks. If the fact table is date driven it might be >worthwhile to break it into separate tables based on date key. This wins in >reducing the working set of queries and in buffering. If there is a real >hotspot, such as current month's activity, you might want to keep a separate >table with just the (most) active data.Static tables of unchanged data can >simplify backups, etc., as well. In this case, you mean I can chunk data into: "facts_04_08" for the august 2004 facts. Is this the case? Otherwise, is it right my point of view that I can get good results by using a different approach, based on mixing vertical partitioning and the CLUSTER facility of PostgreSQL? Can I vertically partition also dimension keys from the fact table or not? However, this subject is awesome and interesting. Far out ... data warehousing seems to be really continous modeling, doesn't it! :-) >Consider summary tables if you know what type of queries you'll hit. At this stage, I can't predict it yet. But of course I need some sort of summary. I will keep it in mind. >Especially here, MVCC is not your friend because it has extra work to do for >aggregate functions. Why does it have extra work? Do you mind being more precise, Aaron? It is really interesting. (thanks) >Cluster helps if you bulk load. Is it maybe because I can update or build them once the load operation has finished? >In most warehouses, the data is downstream data from existing operational >systems. That's my case too. >Because of that you're not able to use database features to >preserve integrity. In most cases, the data goes through an >extract/transform/load process - and the output is considered acceptable. >So, no RI is correct for star or snowflake design. Pretty much no anything >else that adds intelligence - no triggers, no objects, no constraints of any >sort. Many designers try hard to avoid nulls. That's another interesting argument. Again, I had in mind the space efficiency principle and I decided to use null IDs for dimension tables if I don't have the information. I noticed though that in those cases I can't use any index and performances result very poor. I have a dimension table 'categories' referenced through the 'id_category' field in the facts table. I decided to set it to NULL in case I don't have any category to associate to it. I believe it is better to set a '0' value if I don't have any category, allowing me not to use a "SELECT * from facts where id_category IS NULL" which does not use the INDEX I had previously created on that field. >On the hardware side - RAID5 might work here because of the low volume if >you can pay the write performance penalty. To size hardware you need to >estimate load in terms of transaction type (I usually make bucket categories >of small, medium, and large effort needs) and transaction rate. Then try to >estimate how much CPU and I/O they'll use. Thank you so much again Aaron. Your contribution has been really important to me. Ciao, -Gabriele >"Let us not speak of them; but look, and pass on." P.S.: Dante rules ... :-) -- Gabriele Bartolini: Web Programmer, ht://Dig & IWA/HWG Member, ht://Check maintainer Current Location: Prato, Toscana, Italia angusgb@tin.it | http://www.prato.linux.it/~gbartolini | ICQ#129221447 > "Leave every hope, ye who enter!", Dante Alighieri, Divine Comedy, The Inferno --=======7EC77ACB======= Content-Type: text/plain; charset=us-ascii; x-avg=cert; x-avg-checked=avg-ok-16C74A51 Content-Disposition: inline --- Outgoing mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com). Version: 6.0.773 / Virus Database: 520 - Release Date: 05/10/2004 --=======7EC77ACB=======-- From pgsql-performance-owner@postgresql.org Thu Oct 7 18:08:21 2004 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 2AA4732A088 for ; Thu, 7 Oct 2004 18:08:20 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 35483-08 for ; Thu, 7 Oct 2004 17:08:17 +0000 (GMT) Received: from stark.xeocode.com (gsstark.mtl.istop.com [66.11.160.162]) by svr1.postgresql.org (Postfix) with ESMTP id D7EE332A063 for ; Thu, 7 Oct 2004 18:08:15 +0100 (BST) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1CFbkS-0001VQ-00; Thu, 07 Oct 2004 13:08:12 -0400 To: =?iso-8859-1?q?Pierre-Fr=E9d=E9ric_Caillaud?= Cc: "Tom Lane" , "Greg Stark" , pgsql-performance@postgresql.org Subject: Re: sequential scan on select distinct References: <200410061130.58625.ole@freiheit.com> <874ql7ztnb.fsf@stark.xeocode.com> <29194.1097091506@sss.pgh.pa.us> In-Reply-To: From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 07 Oct 2004 13:08:11 -0400 Message-ID: <87pt3u78xw.fsf@stark.xeocode.com> Lines: 21 User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.3 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/83 X-Sequence-Number: 8559 Pierre-Fr�d�ric Caillaud writes: > I see this as a minor annoyance only because I can write GROUP BY > instead of DISTINCT and get the speed boost. It probably annoys people > trying to port applications to postgres though, forcing them to rewrite > their queries. Yeah, really DISTINCT and DISTINCT ON are just special cases of GROUP BY. It seems it makes more sense to put the effort into GROUP BY and just have DISTINCT and DISTINCT ON go through the same code path. Effectively rewriting it internally as a GROUP BY. The really tricky part is that a DISTINCT ON needs to know about a first() aggregate. And to make optimal use of indexes, a last() aggregate as well. And ideally the planner/executor needs to know something is magic about first()/last() (and potentially min()/max() at some point) and that they don't need the complete set of tuples to calculate their results. -- greg From pgsql-performance-owner@postgresql.org Thu Oct 7 18:19:32 2004 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 836D4329E63 for ; Thu, 7 Oct 2004 18:19:30 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 38820-09 for ; Thu, 7 Oct 2004 17:19:20 +0000 (GMT) Received: from mail.hamburg.cityline.net (jewel.mcs-hh.de [194.77.146.129]) by svr1.postgresql.org (Postfix) with ESMTP id 469B1329D41 for ; Thu, 7 Oct 2004 18:19:20 +0100 (BST) Received: from [212.1.56.10] (helo=aconitin.toxine.lan) by mail.hamburg.cityline.net with asmtp (TLSv1:RC4-MD5:128) (Exim 4.32) id 1CFbvC-0006e4-0w; Thu, 07 Oct 2004 19:19:18 +0200 From: Ole Langbehn Organization: freiheit.com To: =?iso-8859-15?q?Pierre-Fr=E9d=E9ric_Caillaud?= Subject: Re: sequential scan on select distinct Date: Thu, 7 Oct 2004 19:21:08 +0200 User-Agent: KMail/1.7 Cc: pgsql-performance@postgresql.org References: <200410061130.58625.ole@freiheit.com> <29194.1097091506@sss.pgh.pa.us> 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: <200410071921.08685.ole@freiheit.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/85 X-Sequence-Number: 8561 Am Donnerstag, 7. Oktober 2004 14:01 schrieb Pierre-Fr=E9d=E9ric Caillaud: > Side Note : > > What do you think about the idea of an "UniqueSort" which would do > sort+unique in one pass ?=20 This is what oracle does and it is quite fast with it... --=20 Ole Langbehn freiheit.com technologies gmbh Theodorstr. 42-90 / 22761 Hamburg, Germany fon =A0 =A0 =A0 +49 (0)40 / 890584-0 fax =A0 =A0 =A0 +49 (0)40 / 890584-20 Freie Software durch B=FCcherkauf f=F6rdern | http://bookzilla.de/ From pgsql-performance-owner@postgresql.org Thu Oct 7 18:26:49 2004 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 2C3BA329EB4 for ; Thu, 7 Oct 2004 18:26:47 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 41438-08 for ; Thu, 7 Oct 2004 17:26:44 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 1DEA0329E94 for ; Thu, 7 Oct 2004 18:26:44 +0100 (BST) 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 i97HQfej000200; Thu, 7 Oct 2004 13:26:41 -0400 (EDT) To: Ole Langbehn Cc: =?iso-8859-15?q?Pierre-Fr=E9d=E9ric_Caillaud?= , pgsql-performance@postgresql.org Subject: Re: sequential scan on select distinct In-reply-to: <200410071921.08685.ole@freiheit.com> References: <200410061130.58625.ole@freiheit.com> <29194.1097091506@sss.pgh.pa.us> <200410071921.08685.ole@freiheit.com> Comments: In-reply-to Ole Langbehn message dated "Thu, 07 Oct 2004 19:21:08 +0200" Date: Thu, 07 Oct 2004 13:26:41 -0400 Message-ID: <199.1097170001@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/86 X-Sequence-Number: 8562 Ole Langbehn writes: >> What do you think about the idea of an "UniqueSort" which would do >> sort+unique in one pass ? > This is what oracle does and it is quite fast with it... Hashing is at least as fast, if not faster. regards, tom lane From pgsql-performance-owner@postgresql.org Thu Oct 7 19:15:51 2004 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 37146329FD7 for ; Thu, 7 Oct 2004 19:15:50 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 57108-06 for ; Thu, 7 Oct 2004 18:15:39 +0000 (GMT) Received: from mail3.panix.com (mail3.panix.com [166.84.1.74]) by svr1.postgresql.org (Postfix) with ESMTP id 1A5B8329F32 for ; Thu, 7 Oct 2004 19:15:39 +0100 (BST) Received: from panix2.panix.com (panix2.panix.com [166.84.1.2]) by mail3.panix.com (Postfix) with ESMTP id 3638F981E9; Thu, 7 Oct 2004 14:15:38 -0400 (EDT) Received: (from adler@localhost) by panix2.panix.com (8.11.6p2-a/8.8.8/PanixN1.1) id i97IFcB14523; Thu, 7 Oct 2004 14:15:38 -0400 (EDT) Date: Thu, 7 Oct 2004 14:15:38 -0400 From: Michael Adler To: Bill Montgomery Cc: Alan Stange , Gaetano Mendola , pgsql-performance@postgresql.org Subject: Re: Excessive context switching on SMP Xeons Message-ID: <20041007181537.GA1693@pobox.com> References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <41630D50.3020308@lulu.com> <200410051538.51664.josh@agliodbs.com> <41636D8E.1090403@rentec.com> <41648446.5020102@bigfoot.com> <4164B48C.9060202@rentec.com> <41656559.70100@lulu.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <41656559.70100@lulu.com> User-Agent: Mutt/1.4.2.1i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/87 X-Sequence-Number: 8563 On Thu, Oct 07, 2004 at 11:48:41AM -0400, Bill Montgomery wrote: > Alan Stange wrote: > > The same test on a Dell PowerEdge 1750, Dual Xeon 3.2 GHz, 512k cache, > HT on, Linux 2.4.21-20.ELsmp (RHEL 3), 4GB memory, pg 7.4.5: > > Far less performance that the Dual Opterons with a low number of > clients, but the gap narrows as the number of clients goes up. Anyone > smarter than me care to explain? You'll have to wait for someone smarter than you, but I will posit this: Did you use a tmpfs filesystem like Alan? You didn't mention either way. Alan did that as an attempt remove IO as a variable. -Mike From pgsql-performance-owner@postgresql.org Thu Oct 7 19:29:19 2004 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 ACCA1329F49 for ; Thu, 7 Oct 2004 19:29:15 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 59971-07 for ; Thu, 7 Oct 2004 18:29:04 +0000 (GMT) Received: from ram.rentec.com (ram.rentec.com [192.5.35.66]) by svr1.postgresql.org (Postfix) with ESMTP id A53EE329D69 for ; Thu, 7 Oct 2004 19:29:04 +0100 (BST) Received: from [172.26.132.145] (hoopoe.rentec.com [172.26.132.145]) by ram.rentec.com (8.12.9/8.12.1) with ESMTP id i97ISsi2002393; Thu, 7 Oct 2004 14:28:54 -0400 (EDT) Message-ID: <41658AE6.3070700@rentec.com> Date: Thu, 07 Oct 2004 14:28:54 -0400 From: Alan Stange Reply-To: stange@rentec.com Organization: Renaissance Technologies Corp. User-Agent: Mozilla Thunderbird 0.8 (X11/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Bill Montgomery Cc: Gaetano Mendola , pgsql-performance@postgresql.org Subject: Re: Excessive context switching on SMP Xeons References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <41630D50.3020308@lulu.com> <200410051538.51664.josh@agliodbs.com> <41636D8E.1090403@rentec.com> <41648446.5020102@bigfoot.com> <4164B48C.9060202@rentec.com> <41656559.70100@lulu.com> In-Reply-To: <41656559.70100@lulu.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/88 X-Sequence-Number: 8564 Bill Montgomery wrote: > Alan Stange wrote: > >> Here's a few numbers from the Opteron 250. If I get some time I'll >> post a more comprehensive comparison including some other systems. >> >> The system is a Sun v20z. Dual Opteron 250, 2.4Ghz, Linux 2.6, 8 GB >> memory. I did a compile and install of pg 8.0 beta 3. I created a >> data base on a tmpfs file system and ran pgbench. Everything was >> "out of the box", meaning I did not tweak any config files. >> >> I used this for pgbench: >> $ pgbench -i -s 32 >> >> and this for pgbench invocations: >> $ pgbench -s 32 -c 1 -t 10000 -v >> >> >> clients tps 1 1290 2 >> 1780 4 1760 8 1680 >> 16 1376 32 904 > > > > The same test on a Dell PowerEdge 1750, Dual Xeon 3.2 GHz, 512k cache, > HT on, Linux 2.4.21-20.ELsmp (RHEL 3), 4GB memory, pg 7.4.5: > > $ pgbench -i -s 32 pgbench > $ pgbench -s 32 -c 1 -t 10000 -v > > clients tps avg CS/sec > ------- ----- ---------- > 1 601 48,000 > 2 889 77,000 > 4 1006 80,000 > 8 985 59,000 > 16 966 47,000 > 32 913 46,000 > > Far less performance that the Dual Opterons with a low number of > clients, but the gap narrows as the number of clients goes up. Anyone > smarter than me care to explain? boy, did Thunderbird ever botch the format of the table I entered... I thought the falloff at 32 clients was a bit steep as well. One thought that crossed my mind is that "pgbench -s 32 -c 32 ..." might not be valid. From the pgbench README: -s scaling_factor this should be used with -i (initialize) option. number of tuples generated will be multiple of the scaling factor. For example, -s 100 will imply 10M (10,000,000) tuples in the accounts table. default is 1. NOTE: scaling factor should be at least as large as the largest number of clients you intend to test; else you'll mostly be measuring update contention. Another possible cause is the that pgbench process is cpu starved and isn't able to keep driving the postgresql processes. So I ran pgbench from another system with all else the same. The numbers were a bit smaller but otherwise similar. I then reran everything using -s 64: clients tps 1 1254 2 1645 4 1713 8 1548 16 1396 32 1060 Still starting to head down a bit. In the 32 client case, the system was ~60% user time, ~25% sytem and ~15% idle. Anyway, the machine is clearly hitting some contention somewhere. It could be in the tmpfs code, VM system, etc. -- Alan From pgsql-performance-owner@postgresql.org Thu Oct 7 19:48:56 2004 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 4916D32A51A for ; Thu, 7 Oct 2004 19:48:54 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 68051-03 for ; Thu, 7 Oct 2004 18:48:51 +0000 (GMT) Received: from snowwhite.lulu.com (mail.lulu.com [66.193.5.50]) by svr1.postgresql.org (Postfix) with ESMTP id 2CA6632A50B for ; Thu, 7 Oct 2004 19:48:51 +0100 (BST) Received: from [10.0.0.32] (meatwad.rdu.lulu.com [10.0.0.32]) (authenticated bits=0) by snowwhite.lulu.com (8.12.8/8.12.8) with ESMTP id i97InabO025282 (version=TLSv1/SSLv3 cipher=RC4-MD5 bits=128 verify=NO); Thu, 7 Oct 2004 14:49:37 -0400 Message-ID: <41658F8D.9090204@lulu.com> Date: Thu, 07 Oct 2004 14:48:45 -0400 From: Bill Montgomery User-Agent: Mozilla Thunderbird 0.7.3 (X11/20040803) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Michael Adler Cc: pgsql-performance@postgresql.org Subject: Re: Excessive context switching on SMP Xeons References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <41630D50.3020308@lulu.com> <200410051538.51664.josh@agliodbs.com> <41636D8E.1090403@rentec.com> <41648446.5020102@bigfoot.com> <4164B48C.9060202@rentec.com> <41656559.70100@lulu.com> <20041007181537.GA1693@pobox.com> In-Reply-To: <20041007181537.GA1693@pobox.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/89 X-Sequence-Number: 8565 Michael Adler wrote: >On Thu, Oct 07, 2004 at 11:48:41AM -0400, Bill Montgomery wrote: > > >>Alan Stange wrote: >> >>The same test on a Dell PowerEdge 1750, Dual Xeon 3.2 GHz, 512k cache, >>HT on, Linux 2.4.21-20.ELsmp (RHEL 3), 4GB memory, pg 7.4.5: >> >>Far less performance that the Dual Opterons with a low number of >>clients, but the gap narrows as the number of clients goes up. Anyone >>smarter than me care to explain? >> >> > >You'll have to wait for someone smarter than you, but I will posit >this: Did you use a tmpfs filesystem like Alan? You didn't mention >either way. Alan did that as an attempt remove IO as a variable. > >-Mike > > Yes, I should have been more explicit. My goal was to replicate his experiment as closely as possible in my environment, so I did run my postgres data directory on a tmpfs. -Bill Montgomery From pgsql-performance-owner@postgresql.org Mon Oct 11 13:03:48 2004 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 ABBB032A08B for ; Thu, 7 Oct 2004 21:26:16 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 01157-04 for ; Thu, 7 Oct 2004 20:26:08 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id BA927329E8B for ; Thu, 7 Oct 2004 21:26:07 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i97KQ4Jm002823 for ; Thu, 7 Oct 2004 20:26:04 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i97K02mH094943 for pgsql-performance@postgresql.org; Thu, 7 Oct 2004 20:00:02 GMT From: Mischa Sandberg Reply-To: ischamay.andbergsay@activestateway.com User-Agent: Mozilla Thunderbird 0.7.3 (X11/20040803) X-Accept-Language: en-us, en MIME-Version: 1.0 X-Newsgroups: comp.databases.postgresql.performance Subject: Re: sequential scan on select distinct References: <200410061130.58625.ole@freiheit.com> <29194.1097091506@sss.pgh.pa.us> <200410071921.08685.ole@freiheit.com> <199.1097170001@sss.pgh.pa.us> In-Reply-To: <199.1097170001@sss.pgh.pa.us> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Lines: 22 Message-ID: <8fh9d.25089$MV5.20638@clgrps13> Date: Thu, 07 Oct 2004 20:00:04 GMT To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.1 tagged_above=0.0 required=5.0 tests=NO_DNS_FOR_FROM X-Spam-Level: * X-Archive-Number: 200410/132 X-Sequence-Number: 8608 Tom Lane wrote: > Ole Langbehn writes: > >>>What do you think about the idea of an "UniqueSort" which would do >>>sort+unique in one pass ? > >>This is what oracle does and it is quite fast with it... > Hashing is at least as fast, if not faster. > > regards, tom lane I got good mileage in a different SQL engine, by combining the hash-aggregate and sort nodes into a single operator. The hash table was just an index into the equivalent of the heap used for generating runs. That gave me partially aggregated data, or eliminated duplicate keys, without extra memory overhead of the hash-aggregation node below the sort. Memory was scarce then ... :-) BTW I'm really puzzled that Oracle is pushing 'index skip scan' as a new feature. Wasn't this in the original Oracle Rdb --- one of Gennady Antoshenkov's tweaks? From pgsql-performance-owner@postgresql.org Thu Oct 7 23:48:48 2004 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 8214C32A0A4 for ; Thu, 7 Oct 2004 23:48:41 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 40301-10 for ; Thu, 7 Oct 2004 22:48:37 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 7628F32A04B for ; Thu, 7 Oct 2004 23:48:38 +0100 (BST) 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 6473573; Thu, 07 Oct 2004 15:50:01 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Data warehousing requirements Date: Thu, 7 Oct 2004 15:50:20 -0700 User-Agent: KMail/1.6.2 Cc: Gabriele Bartolini , "Aaron Werman" References: <6.1.2.0.2.20041006230239.0201bd40@box.tin.it> <6.1.2.0.2.20041007183909.0201f310@box.tin.it> In-Reply-To: <6.1.2.0.2.20041007183909.0201f310@box.tin.it> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Message-Id: <200410071550.20665.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/90 X-Sequence-Number: 8566 Gabriele, > That's another interesting argument. Again, I had in mind the space > efficiency principle and I decided to use null IDs for dimension tables if > I don't have the information. I noticed though that in those cases I can't > use any index and performances result very poor. For one thing, this is false optimization; a NULL isn't saving you any table size on an INT or BIGINT column. NULLs are only smaller on variable-width columns. If you're going to start counting bytes, make sure it's an informed count. More importantly, you should never, ever allow null FKs on a star-topology database. LEFT OUTER JOINs are vastly less efficient than INNER JOINs in a query, and the difference between having 20 outer joins for your data view, vs 20 regular joins, can easily be a difference of 100x in execution time. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Fri Oct 8 03:03:25 2004 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 7520432A601 for ; Fri, 8 Oct 2004 03:03:23 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 22512-10 for ; Fri, 8 Oct 2004 02:03:19 +0000 (GMT) Received: by svr1.postgresql.org (Postfix, from userid 1001) id 2829032B4C1; Fri, 8 Oct 2004 03:50:02 +0100 (BST) Received: from hotmail.com (bay18-dav1.bay18.hotmail.com [65.54.187.181]) by svr1.postgresql.org (Postfix) with ESMTP id A0A2632AC6D for ; Fri, 8 Oct 2004 02:20:01 +0100 (BST) Received: from mail pickup service by hotmail.com with Microsoft SMTPSVC; Thu, 7 Oct 2004 18:20:01 -0700 Received: from 67.81.98.198 by BAY18-DAV1.phx.gbl with DAV; Fri, 08 Oct 2004 01:19:29 +0000 X-Originating-IP: [67.81.98.198] X-Originating-Email: [awerman2@hotmail.com] X-Sender: awerman2@hotmail.com From: "Aaron Werman" To: , "Gabriele Bartolini" References: <6.1.2.0.2.20041006230239.0201bd40@box.tin.it> <6.1.2.0.2.20041007183909.0201f310@box.tin.it> Subject: Re: Data warehousing requirements Date: Thu, 7 Oct 2004 21:19:44 -0400 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 6.00.2800.1437 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 Message-ID: X-OriginalArrivalTime: 08 Oct 2004 01:20:01.0210 (UTC) FILETIME=[EEC441A0:01C4ACD4] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/91 X-Sequence-Number: 8567 ----- Original Message ----- From: "Gabriele Bartolini" To: "Aaron Werman" ; Sent: Thursday, October 07, 2004 1:07 PM Subject: Re: [PERFORM] Data warehousing requirements > At 13.30 07/10/2004, Aaron Werman wrote: > >Consider how the fact table is going to be used, and review hacking it up > >based on usage. Fact tables should be fairly narrow, so if there are extra > >columns beyond keys and dimension keys consider breaking it into parallel > >tables (vertical partitioning). > > Hmm ... I have only an extra column. Sorry if I ask you to confirm this, > but practically vertical partitioning allows me to divide a table into 2 > tables (like if I cut them vertically, right?) having the same key. If I > had 2 extra columns, that could be the case, couldn't it? Yes - it's splitting a table's columns and copying the PK. If you have only one column and it's narrow - partitioning becomes harder to justify. > > >Horizontal partitioning is your friend; especially if it is large - consider > >slicing the data into chunks. If the fact table is date driven it might be > >worthwhile to break it into separate tables based on date key. This wins in > >reducing the working set of queries and in buffering. If there is a real > >hotspot, such as current month's activity, you might want to keep a separate > >table with just the (most) active data.Static tables of unchanged data can > >simplify backups, etc., as well. > > In this case, you mean I can chunk data into: "facts_04_08" for the august > 2004 facts. Is this the case? Exactly. The problem is when you need to query across the chunks. There was a discussion here of creating views ala create view facts as select * from facts_04_07 where datekey between '01/07/2004' and '31/07/2004' union all select * from facts_04_08 where datekey between '01/08/2004' and '31/08/2004' union all select * from facts_04_09 where datekey between '01/09/2004' and '30/09/2004' ... hoping the restrictions would help the planner prune chunks out. Has anyone tried this? > > Otherwise, is it right my point of view that I can get good results by > using a different approach, based on mixing vertical partitioning and the > CLUSTER facility of PostgreSQL? Can I vertically partition also dimension > keys from the fact table or not? If you can do that, you probably should beyond a star schema. The standard definition of a star schema is a single very large fact table with very small dimension tables. The point of a star is that it can be used to efficiantly restrict results out by merging the dimensional restrictions and only extracting matches from the fact table. E.g., select count(*) from people_fact, /* 270M */ states_dim, /* only 50 something */ gender_dim, /* 2 */ age_dim /* say 115 */ where age_dim.age > 65 and gender_dim.gender = 'F' and states_dim.state_code in ('PR', 'ME') and age_dim.age_key = people_fact.age_key and gender_dim.gender_key = people_fact.gender_key and states_dim.state_key = people_fact.state_key (I had to write out this trivial query because most DBAs don't realize going in how ugly star queries are.) If you split the fact table so ages were in a vertical partition you would optimize queries which didn't use the age data, but if you needed the age data, you would have to join two large tables - which is not a star query. What you're thinking about on the cluster front is fun. You can split groups of dimension keys off to seperate vertical partitions, but you can only cluster each on a single key. So you need to split each one off, which results in your inventing the index! (-: > > However, this subject is awesome and interesting. Far out ... data > warehousing seems to be really continous modeling, doesn't it! :-) > > >Consider summary tables if you know what type of queries you'll hit. > > At this stage, I can't predict it yet. But of course I need some sort of > summary. I will keep it in mind. > > >Especially here, MVCC is not your friend because it has extra work to do for > >aggregate functions. > > Why does it have extra work? Do you mind being more precise, Aaron? It is > really interesting. (thanks) The standard reasons - that a lot of queries that seem intuitively to be resolvable statically or through indices have to walk the data to find current versions. Keeping aggregates (especially if you can allow them to be slightly stale) can reduce lots of reads. A big goal of horizontal partitioning is to give the planner some way of reducing the query scope. > > >Cluster helps if you bulk load. > > Is it maybe because I can update or build them once the load operation has > finished? If you have streaming loads, clustering can be a pain to implement well. > > >In most warehouses, the data is downstream data from existing operational > >systems. > > That's my case too. > > >Because of that you're not able to use database features to > >preserve integrity. In most cases, the data goes through an > >extract/transform/load process - and the output is considered acceptable. > >So, no RI is correct for star or snowflake design. Pretty much no anything > >else that adds intelligence - no triggers, no objects, no constraints of any > >sort. Many designers try hard to avoid nulls. > > That's another interesting argument. Again, I had in mind the space > efficiency principle and I decided to use null IDs for dimension tables if > I don't have the information. I noticed though that in those cases I can't > use any index and performances result very poor. > > I have a dimension table 'categories' referenced through the 'id_category' > field in the facts table. I decided to set it to NULL in case I don't have > any category to associate to it. I believe it is better to set a '0' value > if I don't have any category, allowing me not to use a "SELECT * from facts > where id_category IS NULL" which does not use the INDEX I had previously > created on that field. (Sorry for being a pain in the neck, but BTW - that is not a star query; it should be SELECT facts.* from facts, id_dim where facts.id_key = id_dim.id_key and id_dim.id_category IS NULL [and it really gets to the whole problem of indexing low cardinality fields]) > > >On the hardware side - RAID5 might work here because of the low volume if > >you can pay the write performance penalty. To size hardware you need to > >estimate load in terms of transaction type (I usually make bucket categories > >of small, medium, and large effort needs) and transaction rate. Then try to > >estimate how much CPU and I/O they'll use. > > Thank you so much again Aaron. Your contribution has been really important > to me. > > Ciao, > -Gabriele > > >"Let us not speak of them; but look, and pass on." > > P.S.: Dante rules ... :-) :-) that quote was not a reference to anyone in this group! Good luck, /Aaron > > -- > Gabriele Bartolini: Web Programmer, ht://Dig & IWA/HWG Member, ht://Check > maintainer > Current Location: Prato, Toscana, Italia > angusgb@tin.it | http://www.prato.linux.it/~gbartolini | ICQ#129221447 > > "Leave every hope, ye who enter!", Dante Alighieri, Divine Comedy, The > Inferno > From pgsql-performance-owner@postgresql.org Fri Oct 8 03:43:40 2004 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 0E978329E46 for ; Fri, 8 Oct 2004 03:43:39 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 40814-09 for ; Fri, 8 Oct 2004 02:43:37 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 4B850329E42 for ; Fri, 8 Oct 2004 03:43:35 +0100 (BST) 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 i982hXd3027540; Thu, 7 Oct 2004 22:43:33 -0400 (EDT) To: josh@agliodbs.com Cc: pgsql-performance@postgresql.org, Gabriele Bartolini , "Aaron Werman" Subject: Re: Data warehousing requirements In-reply-to: <200410071550.20665.josh@agliodbs.com> References: <6.1.2.0.2.20041006230239.0201bd40@box.tin.it> <6.1.2.0.2.20041007183909.0201f310@box.tin.it> <200410071550.20665.josh@agliodbs.com> Comments: In-reply-to Josh Berkus message dated "Thu, 07 Oct 2004 15:50:20 -0700" Date: Thu, 07 Oct 2004 22:43:33 -0400 Message-ID: <27539.1097203413@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/92 X-Sequence-Number: 8568 Josh Berkus writes: > For one thing, this is false optimization; a NULL isn't saving you any table > size on an INT or BIGINT column. NULLs are only smaller on variable-width > columns. Uh ... not true. The column will not be stored, either way. Now if you had a row that otherwise had no nulls, the first null in the column will cause a null-columns-bitmap to be added, which might more than eat up the savings from storing a single int or bigint. But after the first null, each additional null in a row is a win, free-and-clear, whether it's fixed-width or not. (There are also some alignment considerations that might cause the savings to vanish.) > More importantly, you should never, ever allow null FKs on a star-topology > database. LEFT OUTER JOINs are vastly less efficient than INNER JOINs in a > query, and the difference between having 20 outer joins for your data view, > vs 20 regular joins, can easily be a difference of 100x in execution time. It's not so much that they are necessarily inefficient as that they constrain the planner's freedom of action. You need to think a lot more carefully about the order of joining than when you use inner joins. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 8 06:54:41 2004 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 AF8B432A04D for ; Fri, 8 Oct 2004 06:54:36 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 90991-01 for ; Fri, 8 Oct 2004 05:54:33 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id B733C32A04A for ; Fri, 8 Oct 2004 06:54:32 +0100 (BST) 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 6474953; Thu, 07 Oct 2004 22:55:55 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Data warehousing requirements Date: Thu, 7 Oct 2004 22:53:26 -0700 User-Agent: KMail/1.6.2 Cc: Tom Lane , Gabriele Bartolini , "Aaron Werman" References: <6.1.2.0.2.20041006230239.0201bd40@box.tin.it> <200410071550.20665.josh@agliodbs.com> <27539.1097203413@sss.pgh.pa.us> In-Reply-To: <27539.1097203413@sss.pgh.pa.us> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Message-Id: <200410072253.26169.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/93 X-Sequence-Number: 8569 Tom, Well, I sit corrected. Obviously I misread that. > It's not so much that they are necessarily inefficient as that they > constrain the planner's freedom of action. You need to think a lot more > carefully about the order of joining than when you use inner joins. I've also found that OUTER JOINS constrain the types of joins that can/will be used as well as the order. Maybe you didn't intend it that way, but (for example) OUTER JOINs seem much more likely to use expensive merge joins. -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-benchmarks-owner@postgresql.org Fri Oct 8 09:47:56 2004 X-Original-To: pgsql-benchmarks-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6CF7A329E47; Fri, 8 Oct 2004 09:47:51 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 32175-09; Fri, 8 Oct 2004 08:47:44 +0000 (GMT) Received: from mailer.unicite.fr.netcentrex.net (unknown [62.161.167.249]) by svr1.postgresql.org (Postfix) with ESMTP id 777A7329E51; Fri, 8 Oct 2004 09:47:43 +0100 (BST) Received: from [192.168.101.228] ([192.168.101.228]) by mailer.unicite.fr.netcentrex.net with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2657.72) id 42CX03QZ; Fri, 8 Oct 2004 10:47:42 +0200 Message-ID: <4166542E.2020809@fr.netcentrex.net> Date: Fri, 08 Oct 2004 10:47:42 +0200 From: =?ISO-8859-1?Q?=22Alban_M=E9dici_=28NetCentrex=29=22?= User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: pgsql-benchmarks@postgresql.org, pgsql-performance@postgresql.org Subject: Re: [PERFORM] stats on cursor and query execution troubleshooting References: <4163C5C3.3030304@fr.netcentrex.net> <25576.1097072199@sss.pgh.pa.us> In-Reply-To: <25576.1097072199@sss.pgh.pa.us> Content-Type: multipart/alternative; boundary="------------030409030803050401070308" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.5 tagged_above=0.0 required=5.0 tests=HTML_20_30, HTML_MESSAGE X-Spam-Level: X-Archive-Number: 200410/3 X-Sequence-Number: 20 This is a multi-part message in MIME format. --------------030409030803050401070308 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: quoted-printable Thanks for your repply, but I still don"t understand why the statistic=20 logs : ! 0/0 [0/0] filesystem blocks in/out it told me there is no hard disk access, I'm sure there is, I heard my=20 HDD, and see activity using gkrellm (even using my first query ; big=20 select *) ? 2004-10-08 10:40:05 DEBUG: query: select * from "LINE_Line"; 2004-10-08 10:40:53 DEBUG: QUERY STATISTICS ! system usage stats: ! 48.480196 elapsed 42.010000 user 0.700000 system sec ! [42.030000 user 0.720000 sys total] ! 0/0 [0/0] filesystem blocks in/out ! 6/23 [294/145] page faults/reclaims, 0 [0] swaps ! 0 [0] signals rcvd, 0/0 [0/0] messages rcvd/sent ! 0/0 [0/0] voluntary/involuntary context switches ! postgres usage stats: ! Shared blocks: 3902 read, 0 written, buffer hit=20 rate =3D 11.78% ! Local blocks: 0 read, 0 written, buffer hit=20 rate =3D 0.00% ! Direct blocks: 0 read, 0 written looking at the web some logs, I saw those fields filled (i/o filesystem) Does my postgresql.conf missing an option or is therer a known bug of my=20 postgresql server 7.2.4 ? thx regards Alban M=E9dici on 06/10/2004 16:16 Tom Lane said the following: >=3D?ISO-8859-1?Q?=3D22Alban_M=3DE9dici_=3D28NetCentrex=3D29=3D22?=3D writes: >=20=20 > >>I'm looking for the statistic of memory, CPU, filesystem access while= =3D20 >>executing some regular SQL query, and I want to compare them to >>same kind of results while executing a cursor function. >>=20=20=20=20 >> > >I think your second query is finding all the disk pages it needs in >kernel disk cache, because they were all read in by the first query. >This has little to do with cursor versus non cursor, and everything >to do with hitting recently-read data again. > > regards, tom lane > >=20=20 > --=20 Alban M=E9dici R&D software engineer ------------------------------ you can contact me @ : http://www.netcentrex.net ------------------------------ --------------030409030803050401070308 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Thanks for your repply,  but I still don"t understand why the statistic logs   :
 
!       0/0 [0/0] filesystem blocks in/out
it told me there is no hard disk access, I'm sure there is,  I heard my HDD,  and see activity using gkrellm (even using my first query ; big select *) ?

2004-10-08 10:40:05 DEBUG:  query: select * from "LINE_Line";
2004-10-08 10:40:53 DEBUG:  QUERY STATISTICS
! system usage stats:
!       48.480196 elapsed 42.010000 user 0.700000 system sec
!       [42.030000 user 0.720000 sys total]
!       0/0 [0/0] filesystem blocks in/out
!       6/23 [294/145] page faults/reclaims, 0 [0] swaps
!       0 [0] signals rcvd, 0/0 [0/0] messages rcvd/sent
!       0/0 [0/0] voluntary/involuntary context switches
! postgres usage stats:
!       Shared blocks:       3902 read,          0 written, buffer hit rate = 11.78%
!       Local  blocks:          0 read,          0 written, buffer hit rate = 0.00%
!       Direct blocks:          0 read,          0 written


looking at the web some logs,  I saw those fields filled (i/o filesystem)
Does my postgresql.conf missing an option or is therer a known bug of my postgresql server  7.2.4 ?



thx
regards

Alban Médici


on 06/10/2004 16:16 Tom Lane said the following: 
=?ISO-8859-1?Q?=22Alban_M=E9dici_=28NetCentrex=29=22?= <amedici@fr.netcentrex.net> writes:
  
I'm looking for the statistic of memory,  CPU,  filesystem access while=20
executing some regular SQL query,  and I want to compare them to
same kind of results while executing a cursor function.
    

I think your second query is finding all the disk pages it needs in
kernel disk cache, because they were all read in by the first query.
This has little to do with cursor versus non cursor, and everything
to do with hitting recently-read data again.

			regards, tom lane

  

-- 
Alban Médici
R&D software engineer
------------------------------
you can contact me @ :
http://www.netcentrex.net
------------------------------
--------------030409030803050401070308-- From pgsql-performance-owner@postgresql.org Fri Oct 8 09:51:56 2004 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 A044832AE5E for ; Fri, 8 Oct 2004 09:51:53 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 35873-05 for ; Fri, 8 Oct 2004 08:51:48 +0000 (GMT) Received: from boutiquenumerique.com (gailleton-2-82-67-9-10.fbx.proxad.net [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 3250932AC69 for ; Fri, 8 Oct 2004 09:51:46 +0100 (BST) Received: (qmail 10834 invoked from network); 8 Oct 2004 10:52:12 +0200 Received: from unknown (HELO musicbox) (192.168.0.2) by gailleton-2-82-67-9-10.fbx.proxad.net with SMTP; 8 Oct 2004 10:52:12 +0200 Date: Fri, 08 Oct 2004 10:54:59 +0200 To: pgsql-performance@postgresql.org Subject: Re: sequential scan on select distinct References: <200410061130.58625.ole@freiheit.com> <874ql7ztnb.fsf@stark.xeocode.com> <29194.1097091506@sss.pgh.pa.us> <87pt3u78xw.fsf@stark.xeocode.com> From: =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= 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: <87pt3u78xw.fsf@stark.xeocode.com> User-Agent: Opera M2/7.53 (Linux, build 737) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/95 X-Sequence-Number: 8571 > The really tricky part is that a DISTINCT ON needs to know about a > first() > aggregate. And to make optimal use of indexes, a last() aggregate as > well. And > ideally the planner/executor needs to know something is magic about > first()/last() (and potentially min()/max() at some point) and that they > don't > need the complete set of tuples to calculate their results. I'm going to be accused of hand-waving again, but please pardon me, I'm enthusiastic, and I like to propose new idead, you can kick me if you don't like them or if I put out too much uninformed bull ! Idea : The aggregate accumulation function could have a way to say : "stop ! I've had enough of these values ! Get on with the next item in the GROUP BY clause !" I don't know how, or if, the planner could use this (guess: no) or the index scan use this (guess: no) but it would at least save the function calls. I'd guess this idea is quite useless. Aggregates could have an additional attribute saying how much values it will need ('max_rows' maybe). This would prevent the creation of "magic" aggregates for max() (which is a kind of special-casing), keep it generic (so users can create magic aggregates like this). Aggregates already consist of a bunch of functions (start, accumulate, return retuls) so this could be just another element in this set. This information would be known ahead of time and could influence the query plans too. I'm going to wave my hand and say "not too much planning cost" because I guess the aggregate details are fetched during planning so fetching one more attribute would not be that long... For instance first() would have max_rows=1, and users could code a "first N accumulator-in-array" which would have max_rows=N... This does not solve the problem of min() and max() which need max_rows=1 only if the result is sorted... hum... maybe another attribute like max_rows_sorted = 1 for max() and -1 for min() meaning 'first 1' or 'last 1' (or first N or last N)... according to the "order by" clause it would be known that the 'first N' of an 'order by ... asc' is the same as the 'last N' from an 'order by ... desc' ??? From pgsql-performance-owner@postgresql.org Fri Oct 8 10:11:26 2004 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 5A20A329F24 for ; Fri, 8 Oct 2004 10:11:24 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 41833-04 for ; Fri, 8 Oct 2004 09:11:18 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.196]) by svr1.postgresql.org (Postfix) with ESMTP id 65AC932AF62 for ; Fri, 8 Oct 2004 10:11:17 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 77so6816rnk for ; Fri, 08 Oct 2004 02:11:16 -0700 (PDT) Received: by 10.38.65.39 with SMTP id n39mr48739rna; Fri, 08 Oct 2004 02:11:16 -0700 (PDT) Received: by 10.38.163.22 with HTTP; Fri, 8 Oct 2004 02:11:16 -0700 (PDT) Message-ID: <758d5e7f041008021163fb40eb@mail.gmail.com> Date: Fri, 8 Oct 2004 11:11:16 +0200 From: Dawid Kuroczko Reply-To: Dawid Kuroczko To: pgsql-performance@postgresql.org Subject: integer[] indexing. 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=0.9 tagged_above=0.0 required=5.0 tests=FROM_ENDS_IN_NUMS X-Spam-Level: X-Archive-Number: 200410/96 X-Sequence-Number: 8572 I have a large table with a column: ids integer[] not null most of these entries (over 95%) contain only one array element, some can contain up to 10 array elements. seqscan is naturally slow. GIST on int_array works nice, but GIST isn't exactly a speed daemon when it comes to updating. So I thought, why not create partial indexes? CREATE INDEX one_element_array_index ON table ((ids[1])) WHERE icount(ids) <= 1; CREATE INDEX many_element_array_index ON table USING GIST (ids) WHERE icount(ids) > 1; Now, if I select WHERE icount(ids) <= 1 AND ids[1] = 33 I get lightning fast results. If I select WHERE icount(ids) > 1 AND ids && '{33}' -- I get them even faster. But when I phrase the query: SELECT * FROM table WHERE (icount(ids) <= 1 AND ids[1] = 33) OR (icount(ids) > 1 AND ids && '{33}'); Planner insists on using seqscan. Even with enable_seqscan = off; Any hints, comments? :) [ I think thsese partial indexes take best of two worlds, only if planner wanted to take advantage of it... :) ] Regards, Dawid From pgsql-performance-owner@postgresql.org Fri Oct 8 10:26:29 2004 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 1FA5F32A343 for ; Fri, 8 Oct 2004 10:26:28 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 46826-01 for ; Fri, 8 Oct 2004 09:26:21 +0000 (GMT) Received: from boutiquenumerique.com (gailleton-2-82-67-9-10.fbx.proxad.net [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 57C8832A0EB for ; Fri, 8 Oct 2004 10:26:21 +0100 (BST) Received: (qmail 12098 invoked from network); 8 Oct 2004 11:26:48 +0200 Received: from unknown (HELO musicbox) (192.168.0.2) by gailleton-2-82-67-9-10.fbx.proxad.net with SMTP; 8 Oct 2004 11:26:48 +0200 To: "Dawid Kuroczko" , pgsql-performance@postgresql.org Subject: Re: integer[] indexing. References: <758d5e7f041008021163fb40eb@mail.gmail.com> Message-ID: From: =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= 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: Fri, 08 Oct 2004 11:29:35 +0200 In-Reply-To: <758d5e7f041008021163fb40eb@mail.gmail.com> User-Agent: Opera M2/7.53 (Linux, build 737) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.3 tagged_above=0.0 required=5.0 tests=UPPERCASE_25_50 X-Spam-Level: X-Archive-Number: 200410/97 X-Sequence-Number: 8573 disclaimer : brainless proposition (SELECT * FROM table WHERE (icount(ids) <= 1 AND ids[1] = 33) UNION ALL (SELECT * FROM table WHERE (icount(ids) > 1 AND ids && '{33}')); From pgsql-performance-owner@postgresql.org Fri Oct 8 10:54:10 2004 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 E28D332A090 for ; Fri, 8 Oct 2004 10:54:04 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 52331-08 for ; Fri, 8 Oct 2004 09:53:58 +0000 (GMT) Received: from boutiquenumerique.com (gailleton-2-82-67-9-10.fbx.proxad.net [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 56423329E83 for ; Fri, 8 Oct 2004 10:53:59 +0100 (BST) Received: (qmail 13015 invoked from network); 8 Oct 2004 11:54:26 +0200 Received: from unknown (HELO musicbox) (192.168.0.2) by gailleton-2-82-67-9-10.fbx.proxad.net with SMTP; 8 Oct 2004 11:54:26 +0200 To: pgsql-performance@postgresql.org Subject: Re: sequential scan on select distinct References: <200410061130.58625.ole@freiheit.com> <29194.1097091506@sss.pgh.pa.us> <200410071921.08685.ole@freiheit.com> <199.1097170001@sss.pgh.pa.us> Message-ID: Date: Fri, 08 Oct 2004 11:57:13 +0200 From: =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= 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: <199.1097170001@sss.pgh.pa.us> User-Agent: Opera M2/7.53 (Linux, build 737) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/98 X-Sequence-Number: 8574 > Hashing is at least as fast, if not faster. > regards, tom lane Probably quite faster if the dataset is not huge... UniqueSort would be useful for GROUP BY x ORDER BY x though From pgsql-performance-owner@postgresql.org Mon Oct 11 13:02:48 2004 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 1C83432A31D for ; Fri, 8 Oct 2004 13:10:37 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91914-08 for ; Fri, 8 Oct 2004 12:10:29 +0000 (GMT) Received: from scaup.mail.pas.earthlink.net (scaup.mail.pas.earthlink.net [207.217.120.49]) by svr1.postgresql.org (Postfix) with ESMTP id B3AE032A311 for ; Fri, 8 Oct 2004 13:10:32 +0100 (BST) Received: from lsanca1-ar6-4-62-201-153.lsanca1.elnk.dsl.genuity.net ([4.62.201.153] helo=bsd.mvh) by scaup.mail.pas.earthlink.net with esmtp (Exim 3.33 #1) id 1CFtZv-0001FB-00 for pgsql-performance@postgresql.org; Fri, 08 Oct 2004 05:10:31 -0700 Received: from localhost (localhost [127.0.0.1]) by bsd.mvh (Postfix) with ESMTP id 6489B5485C for ; Fri, 8 Oct 2004 05:10:31 -0700 (PDT) Received: from bsd.mvh ([127.0.0.1]) by localhost (bsd.mvh [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 73471-07 for ; Fri, 8 Oct 2004 05:10:30 -0700 (PDT) Received: by bsd.mvh (Postfix, from userid 1000) id EEB485486C; Fri, 8 Oct 2004 05:10:29 -0700 (PDT) From: Mike Harding To: pgsql-performance@postgresql.org Subject: COPY slows down? Message-Id: <20041008121029.EEB485486C@bsd.mvh> Date: Fri, 8 Oct 2004 05:10:29 -0700 (PDT) X-Virus-Scanned: by amavisd-new at bsd.mvh X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/129 X-Sequence-Number: 8605 I just ran a COPY of a million records several times, and each time I ran it it ran apparently exponentially slower. If I do an insert of 10 million records, even with 2 indexes (same table) it doesn't appear to slow down at all. Any ideas? - Mike H. (I apologize for the ^Hs) Script started on Wed Oct 6 08:37:32 2004 bash-3.00$ psql Welcome to psql 7.4.5, the PostgreSQL interactive terminal. Type: \copyright for distribution terms \h for help with SQL commands \? for help on internal slash commands \g or terminate with semicolon to execute query \q to quit mvh=# \timing Timing is on. mvh=# \timingreindex table bgtest;mvh=# delete from bgtest;mvh=# copy bgtest from '/home/mvh/databasestuff/dbdmp/bgdump'; COPY Time: 69796.130 ms mvh=# vacuum analyze; VACUUM Time: 19148.621 ms mvh=# vacuum analyze;mvh=# copy bgtest from '/home/mvh/databasestuff/dbdmp/bgdump'; COPY Time: 89189.939 ms mvh=# copy bgtest from '/home/mvh/databasestuff/dbdmp/bgdump';mvh=# vacuum analyze; VACUUM Time: 26814.670 ms mvh=# vacuum analyze;mvh=# copy bgtest from '/home/mvh/databasestuff/dbdmp/bgdump'; COPY Time: 131131.982 ms mvh=# copy bgtest from '/home/mvh/databasestuff/dbdmp/bgdump';mvh=# vacuum analyze; VACUUM Time: 64997.264 ms mvh=# vacuum analyze;mvh=# copy bgtest from '/home/mvh/databasestuff/dbdmp/bgdump'; COPY Time: 299977.697 ms mvh=# copy bgtest from '/home/mvh/databasestuff/dbdmp/bgdump';mvh=# vacuum analyze; VACUUM Time: 103541.716 ms mvh=# vacuum analyze;mvh=# copy bgtest from '/home/mvh/databasestuff/dbdmp/bgdump'; COPY Time: 455292.600 ms mvh=# copy bgtest from '/home/mvh/databasestuff/dbdmp/bgdump';mvh=# vacuum analyze; VACUUM Time: 138910.015 ms mvh=# vacuum analyze;mvh=# copy bgtest from '/home/mvh/databasestuff/dbdmp/bgdump'; COPY Time: 612119.661 ms mvh=# copy bgtest from '/home/mvh/databasestuff/dbdmp/bgdump';mvh=# vacuum analyze; VACUUM Time: 151331.243 ms mvh=# \q bash-3.00$ exit Script done on Wed Oct 6 10:43:04 2004 From pgsql-performance-owner@postgresql.org Fri Oct 8 13:29:06 2004 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 144B032A446 for ; Fri, 8 Oct 2004 13:29:05 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 00267-03 for ; Fri, 8 Oct 2004 12:28:56 +0000 (GMT) Received: from main.gmane.org (main.gmane.org [80.91.229.2]) by svr1.postgresql.org (Postfix) with ESMTP id AB5CF32A46B for ; Fri, 8 Oct 2004 13:28:58 +0100 (BST) Received: from list by main.gmane.org with local (Exim 3.35 #1 (Debian)) id 1CFtrk-0002Yw-00 for ; Fri, 08 Oct 2004 14:28:56 +0200 Received: from srv.protecting.net ([212.126.218.242]) by main.gmane.org with esmtp (Gmexim 0.1 (Debian)) id 1AlnuQ-0007hv-00 for ; Fri, 08 Oct 2004 14:28:56 +0200 Received: from hf0722x by srv.protecting.net with local (Gmexim 0.1 (Debian)) id 1AlnuQ-0007hv-00 for ; Fri, 08 Oct 2004 14:28:56 +0200 X-Injected-Via-Gmane: http://gmane.org/ To: pgsql-performance@postgresql.org From: Harald Fuchs Subject: Re: integer[] indexing. Date: 08 Oct 2004 14:28:52 +0200 Organization: Linux Private Site Lines: 24 Message-ID: References: <758d5e7f041008021163fb40eb@mail.gmail.com> Reply-To: hf0722x@protecting.net Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII X-Complaints-To: usenet@sea.gmane.org X-Gmane-NNTP-Posting-Host: srv.protecting.net User-Agent: Gnus/5.0808 (Gnus v5.8.8) Emacs/20.7 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/99 X-Sequence-Number: 8575 In article , =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= writes: > disclaimer : brainless proposition > (SELECT * FROM table WHERE (icount(ids) <= 1 AND ids[1] = 33) > UNION ALL > (SELECT * FROM table WHERE (icount(ids) > 1 AND ids && '{33}')); I guess my proposition is even more brainless :-) If 95% of all records have only one value, how about putting the first (and most often only) value into a separate column with a btree index on it? Something like that: CREATE TABLE tbl ( -- other columns id1 INT NOT NULL, idN INT[] NULL ); CREATE INDEX tbl_id1_ix ON tbl (id1); If id1 is selective enough, you probably don't need another index on idn. From pgsql-performance-owner@postgresql.org Fri Oct 8 13:38:20 2004 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 C493A32A701 for ; Fri, 8 Oct 2004 13:38:18 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 01328-07 for ; Fri, 8 Oct 2004 12:38:10 +0000 (GMT) Received: from moutng.kundenserver.de (moutng.kundenserver.de [212.227.126.191]) by svr1.postgresql.org (Postfix) with ESMTP id 87C4E32A6BE for ; Fri, 8 Oct 2004 13:38:12 +0100 (BST) Received: from [212.227.126.203] (helo=mrvnet.kundenserver.de) by moutng.kundenserver.de with esmtp (Exim 3.35 #1) id 1CFu0d-0005w4-00; Fri, 08 Oct 2004 14:38:07 +0200 Received: from [172.23.4.132] (helo=config5.kundenserver.de) by mrvnet.kundenserver.de with esmtp (Exim 3.35 #1) id 1CFu0c-0006fr-00; Fri, 08 Oct 2004 14:38:06 +0200 Received: from www-data by config5.kundenserver.de with local (Exim 3.35 #1 (Debian)) id 1CFu0c-0002BW-00; Fri, 08 Oct 2004 14:38:06 +0200 To: =?iso-8859-1?Q?Josh_Berkus?= Subject: =?iso-8859-1?Q?Re:_Re:__Data_warehousing_requirements?= From: Cc: , =?iso-8859-1?Q?Tom_Lane?= , =?iso-8859-1?Q?Gabriele_Bartolini?= , =?iso-8859-1?Q?Aaron_Werman?= Message-Id: <28292295$10972388454166893d146f88.44056315@config5.schlund.de> X-Binford: 6100 (more power) X-Originating-From: 28292295 X-Mailer: Webmail X-Routing: UK X-Received: from config5 by 193.23.116.11 with HTTP id 28292295 for josh@agliodbs.com; Fri, 8 Oct 2004 14:38:01 +0200 Content-Type: text/plain; charset="iso-8859-1" Mime-Version: 1.0 Content-Transfer-Encoding: 8bit X-Priority: 3 Date: Fri, 8 Oct 2004 14:38:01 +0200 X-Provags-ID: kundenserver.de abuse@kundenserver.de ident:@172.23.4.132 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.3 tagged_above=0.0 required=5.0 tests=NO_REAL_NAME X-Spam-Level: X-Archive-Number: 200410/100 X-Sequence-Number: 8576 Josh Berkus wrote on 08.10.2004, 07:53:26: > > > It's not so much that they are necessarily inefficient as that they > > constrain the planner's freedom of action. You need to think a lot more > > carefully about the order of joining than when you use inner joins. > > I've also found that OUTER JOINS constrain the types of joins that can/will be > used as well as the order. Maybe you didn't intend it that way, but (for > example) OUTER JOINs seem much more likely to use expensive merge joins. > Unfortunately, yes thats true - thats is for correctness, not an optimization decision. Outer joins constrain you on both join order AND on join type. Nested loops and hash joins avoid touching all rows in the right hand table, which is exactly what you don't want when you have a right outer join to perform, since you wish to include rows in that table when there is no match. Thus, we MUST choose a merge join even when (if it wasn't an outer join) we would have chosen a nested loops or hash. Best Regards, Simon Riggs From pgsql-benchmarks-owner@postgresql.org Fri Oct 8 14:56:09 2004 X-Original-To: pgsql-benchmarks-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 27CDB32A090; Fri, 8 Oct 2004 14:56:04 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 27480-02; Fri, 8 Oct 2004 13:55:54 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id AFDC4329E3E; Fri, 8 Oct 2004 14:55:57 +0100 (BST) 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 i98DtuVB011895; Fri, 8 Oct 2004 09:55:57 -0400 (EDT) To: =?ISO-8859-1?Q?=22Alban_M=E9dici_=28NetCentrex=29=22?= Cc: pgsql-benchmarks@postgresql.org, pgsql-performance@postgresql.org Subject: Re: [PERFORM] stats on cursor and query execution troubleshooting In-reply-to: <4166542E.2020809@fr.netcentrex.net> References: <4163C5C3.3030304@fr.netcentrex.net> <25576.1097072199@sss.pgh.pa.us> <4166542E.2020809@fr.netcentrex.net> Comments: In-reply-to =?ISO-8859-1?Q?=22Alban_M=E9dici_=28NetCentrex=29=22?= message dated "Fri, 08 Oct 2004 10:47:42 +0200" Date: Fri, 08 Oct 2004 09:55:56 -0400 Message-ID: <11894.1097243756@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/4 X-Sequence-Number: 21 =?ISO-8859-1?Q?=22Alban_M=E9dici_=28NetCentrex=29=22?= writes: > Thanks for your repply, but I still don"t understand why the statistic > logs : > ! 0/0 [0/0] filesystem blocks in/out > it told me there is no hard disk access, I'm sure there is, Complain to your friendly local kernel hacker. We just report what getrusage() tells us; so if the number is wrong then it's a kernel bug. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 8 15:03:45 2004 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 CAE2732A821 for ; Fri, 8 Oct 2004 15:03:42 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 30015-06 for ; Fri, 8 Oct 2004 14:03:33 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 7E6C832A641 for ; Fri, 8 Oct 2004 15:03:35 +0100 (BST) 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 i98E3RQn011973; Fri, 8 Oct 2004 10:03:27 -0400 (EDT) To: Dawid Kuroczko Cc: pgsql-performance@postgresql.org Subject: Re: integer[] indexing. In-reply-to: <758d5e7f041008021163fb40eb@mail.gmail.com> References: <758d5e7f041008021163fb40eb@mail.gmail.com> Comments: In-reply-to Dawid Kuroczko message dated "Fri, 08 Oct 2004 11:11:16 +0200" Date: Fri, 08 Oct 2004 10:03:26 -0400 Message-ID: <11972.1097244206@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.1 tagged_above=0.0 required=5.0 tests=MAILTO_TO_SPAM_ADDR X-Spam-Level: * X-Archive-Number: 200410/102 X-Sequence-Number: 8578 Dawid Kuroczko writes: > But when I phrase the query: > SELECT * FROM table WHERE (icount(ids) <= 1 AND ids[1] = 33) OR > (icount(ids) > 1 AND ids && '{33}'); > Planner insists on using seqscan. Even with enable_seqscan = off; The OR-index-scan mechanism isn't currently smart enough to use partial indexes that are only valid for some of the OR'd clauses rather than all of them. Feel free to fix it ;-). (This might not even be very hard; I haven't looked.) regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 8 15:23:02 2004 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 97C2132B1CA for ; Fri, 8 Oct 2004 15:22:58 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 36347-07 for ; Fri, 8 Oct 2004 14:22:47 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 616D832B1A7 for ; Fri, 8 Oct 2004 15:22:51 +0100 (BST) 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 i98EMiRm012170; Fri, 8 Oct 2004 10:22:44 -0400 (EDT) To: simon@2ndquadrant.com Cc: =?iso-8859-1?Q?Josh_Berkus?= , pgsql-performance@postgresql.org, =?iso-8859-1?Q?Gabriele_Bartolini?= , =?iso-8859-1?Q?Aaron_Werman?= Subject: Re: =?iso-8859-1?Q?Re:_Re:__Data_warehousing_requirements?= In-reply-to: <28292295$10972388454166893d146f88.44056315@config5.schlund.de> References: <28292295$10972388454166893d146f88.44056315@config5.schlund.de> Comments: In-reply-to message dated "Fri, 08 Oct 2004 14:38:01 +0200" Date: Fri, 08 Oct 2004 10:22:44 -0400 Message-ID: <12169.1097245364@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/103 X-Sequence-Number: 8579 writes: > Unfortunately, yes thats true - thats is for correctness, not an > optimization decision. Outer joins constrain you on both join order AND > on join type. Nested loops and hash joins avoid touching all rows in > the right hand table, which is exactly what you don't want when you > have a right outer join to perform, since you wish to include rows in > that table when there is no match. Thus, we MUST choose a merge join > even when (if it wasn't an outer join) we would have chosen a nested > loops or hash. The alternative of course is to flip it around to be a left outer join so that we can use those plan types. But depending on the relative sizes of the two tables this may be a loser. If you are using a FULL join then it is indeed true that mergejoin is the only supported plan type. I don't think that was at issue here though. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 8 15:47:57 2004 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 5B9E2329EE6 for ; Fri, 8 Oct 2004 15:47:55 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 45195-05 for ; Fri, 8 Oct 2004 14:47:43 +0000 (GMT) Received: from mail.deg.cc (unknown [64.139.134.201]) by svr1.postgresql.org (Postfix) with ESMTP id CF3D5329F72 for ; Fri, 8 Oct 2004 15:47:43 +0100 (BST) Received: from deg.cc (unknown [198.70.16.205]) by mail.deg.cc (Postfix) with ESMTP id A0E145608A for ; Fri, 8 Oct 2004 10:47:40 -0400 (EDT) Message-ID: <4166A8E8.4060102@deg.cc> Date: Fri, 08 Oct 2004 10:49:12 -0400 From: Pallav Kalva User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.4.2) Gecko/20040308 X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Query Tuning 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/104 X-Sequence-Number: 8580 Hi, I have a problem with the below query, when i do explain on the below query on my live database it doesnt use any index specified on the tables and does seq scan on the table which is 400k records. But if i copy the same table onto a different database on a different machine it uses all the indexes specified and query runs much quicker. I ran analyze, vacuum analyze and rebuilt indexes on the live database but still there is no difference in the performance. Can anyone tell why this odd behavior ? Thanks! Query -------- SELECT a.total as fsbos, b.total as foreclosures, c.total as auctions, d.latestDate as lastUpdated FROM ((SELECT count(1) as total FROM Properties p INNER JOIN Datasources ds ON p.datasource = ds.sourceId WHERE p.countyState = 'GA' AND ds.sourceType = 'fsbo' AND p.status in (1,2) )) a, ((SELECT count(1) as total FROM Properties p INNER JOIN Datasources ds ON p.datasource = ds.sourceId WHERE p.countyState = 'GA' AND ds.sourceType = 'foreclosure' AND (p.status in (1,2) OR (p.status = 0 AND p.LastReviewed2 >= current_timestamp - INTERVAL '14 days') ) )) b, ((SELECT count(1) as total FROM Properties p WHERE p.datasource = 1087 AND p.countyState = 'GA' AND p.status in (1,2) )) c, ((SELECT to_char(max(entryDate2), 'MM/DD/YYYY HH24:MI') as latestDate FROM Properties p WHERE p.countyState = 'GA' )) d Explain from the Live database ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Nested Loop (cost=1334730.95..1334731.02 rows=1 width=56) -> Nested Loop (cost=1026932.25..1026932.30 rows=1 width=24) -> Nested Loop (cost=704352.11..704352.14 rows=1 width=16) -> Subquery Scan b (cost=375019.89..375019.90 rows=1 width=8) -> Aggregate (cost=375019.89..375019.89 rows=1 width=0) -> Hash Join (cost=308.72..374844.49 rows=70158 width=0) Hash Cond: ("outer".datasource = "inner".sourceid) -> Seq Scan on properties p (cost=0.00..373289.10 rows=72678 width=4) Filter: ((countystate = 'GA'::bpchar) AND ((status = 0) OR (status = 1) OR (status = 2)) AND ((lastreviewed2 >= (('now'::text)::timestamp(6) with time zone - '14 days'::interval)) OR (status = 1) OR (status = 2))) -> Hash (cost=288.05..288.05 rows=8267 width=4) -> Seq Scan on datasources ds (cost=0.00..288.05 rows=8267 width=4) Filter: ((sourcetype)::text = 'foreclosure'::text) -> Subquery Scan c (cost=329332.22..329332.23 rows=1 width=8) -> Aggregate (cost=329332.22..329332.22 rows=1 width=0) -> Seq Scan on properties p (cost=0.00..329321.06 rows=4464 width=0) Filter: ((datasource = 1087) AND (countystate = 'GA'::bpchar) AND ((status = 1) OR (status = 2))) -> Subquery Scan a (cost=322580.14..322580.15 rows=1 width=8) -> Aggregate (cost=322580.14..322580.14 rows=1 width=0) -> Hash Join (cost=288.24..322579.28 rows=344 width=0) Hash Cond: ("outer".datasource = "inner".sourceid) -> Seq Scan on properties p (cost=0.00..321993.05 rows=39273 width=4) Filter: ((countystate = 'GA'::bpchar) AND ((status = 1) OR (status = 2))) -> Hash (cost=288.05..288.05 rows=75 width=4) -> Seq Scan on datasources ds (cost=0.00..288.05 rows=75 width=4) Filter: ((sourcetype)::text = 'fsbo'::text) -> Subquery Scan d (cost=307798.70..307798.72 rows=1 width=32) -> Aggregate (cost=307798.70..307798.71 rows=1 width=8) -> Seq Scan on properties p (cost=0.00..307337.04 rows=184666 width=8) Filter: (countystate = 'GA'::bpchar) Explain on the Copy of the Live database for the same query ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Nested Loop (cost=5380.81..5380.88 rows=1 width=56) -> Nested Loop (cost=3714.30..3714.35 rows=1 width=48) -> Nested Loop (cost=2687.15..2687.18 rows=1 width=40) -> Subquery Scan a (cost=1022.76..1022.77 rows=1 width=8) -> Aggregate (cost=1022.76..1022.76 rows=1 width=0) -> Nested Loop (cost=0.00..1022.75 rows=2 width=0) -> Seq Scan on datasources ds (cost=0.00..4.44 rows=2 width=4) Filter: ((sourcetype)::text = 'fsbo'::text) -> Index Scan using idx_properties_datasourcestateauctiondate on properties p (cost=0.00..509.14 rows=2 width=4) Index Cond: (p.datasource = "outer".sourceid) Filter: ((countystate = 'GA'::bpchar) AND ((status = 1) OR (status = 2))) -> Subquery Scan d (cost=1664.39..1664.40 rows=1 width=32) -> Aggregate (cost=1664.39..1664.39 rows=1 width=8) -> Index Scan using properties_idx_search on properties p (cost=0.00..1663.35 rows=416 width=8) Index Cond: (countystate = 'GA'::bpchar) -> Subquery Scan b (cost=1027.15..1027.16 rows=1 width=8) -> Aggregate (cost=1027.15..1027.15 rows=1 width=0) -> Nested Loop (cost=0.00..1027.14 rows=3 width=0) -> Seq Scan on datasources ds (cost=0.00..4.44 rows=2 width=4) Filter: ((sourcetype)::text = 'foreclosure'::text) -> Index Scan using idx_properties_datasourcestateauctiondate on properties p (cost=0.00..511.32 rows=3 width=4) Index Cond: (p.datasource = "outer".sourceid) Filter: ((countystate = 'GA'::bpchar) AND ((status = 0) OR (status = 1) OR (status = 2)) AND ((lastreviewed2 >= (('now'::text)::timestamp(6) with time zone - '14 days'::interval)) OR (status = 1) OR (status = 2))) -> Subquery Scan c (cost=1666.51..1666.52 rows=1 width=8) -> Aggregate (cost=1666.51..1666.51 rows=1 width=0) -> Index Scan using properties_idx_search on properties p (cost=0.00..1666.46 rows=18 width=0) Index Cond: (countystate = 'GA'::bpchar) Filter: ((datasource = 1087) AND ((status = 1) OR (status = 2))) From pgsql-performance-owner@postgresql.org Fri Oct 8 17:43:14 2004 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 5908832A704 for ; Fri, 8 Oct 2004 17:43:11 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 82826-04 for ; Fri, 8 Oct 2004 16:42:59 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 327F132A6A5 for ; Fri, 8 Oct 2004 17:42:59 +0100 (BST) 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 i98GguGn013658; Fri, 8 Oct 2004 12:42:57 -0400 (EDT) To: Pallav Kalva Cc: pgsql-performance@postgresql.org Subject: Re: Query Tuning In-reply-to: <4166A8E8.4060102@deg.cc> References: <4166A8E8.4060102@deg.cc> Comments: In-reply-to Pallav Kalva message dated "Fri, 08 Oct 2004 10:49:12 -0400" Date: Fri, 08 Oct 2004 12:42:56 -0400 Message-ID: <13657.1097253776@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.1 tagged_above=0.0 required=5.0 tests=NO_DNS_FOR_FROM X-Spam-Level: * X-Archive-Number: 200410/105 X-Sequence-Number: 8581 Pallav Kalva writes: > I have a problem with the below query, when i do explain on the > below query on my live database it doesnt use any index specified on the > tables and does seq scan on the table which is 400k records. But if i > copy the same table onto a different database on a different machine it > uses all the indexes specified and query runs much quicker. It looks to me like you've never vacuumed/analyzed the copy, and so you get a different plan there. The fact that that plan is better than the one made with statistics is unhappy making :-( ... but when you only show us EXPLAIN output rather than EXPLAIN ANALYZE, it's impossible to speculate about why. Also, what PG version is this? regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 8 18:28:37 2004 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 2CFBA32AABA for ; Fri, 8 Oct 2004 18:28:36 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 94752-10 for ; Fri, 8 Oct 2004 17:27:54 +0000 (GMT) Received: from mail.deg.cc (unknown [64.139.134.201]) by svr1.postgresql.org (Postfix) with ESMTP id A2EAA329E64 for ; Fri, 8 Oct 2004 18:27:54 +0100 (BST) Received: from deg.cc (unknown [198.70.16.205]) by mail.deg.cc (Postfix) with ESMTP id 711FB560C1; Fri, 8 Oct 2004 13:27:49 -0400 (EDT) Message-ID: <4166CE72.2050900@deg.cc> Date: Fri, 08 Oct 2004 13:29:22 -0400 From: Pallav Kalva User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.4.2) Gecko/20040308 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: pgsql-performance@postgresql.org Subject: Re: Query Tuning References: <4166A8E8.4060102@deg.cc> <13657.1097253776@sss.pgh.pa.us> In-Reply-To: <13657.1097253776@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=1.1 tagged_above=0.0 required=5.0 tests=NO_DNS_FOR_FROM X-Spam-Level: * X-Archive-Number: 200410/106 X-Sequence-Number: 8582 Tom Lane wrote: >Pallav Kalva writes: > > >> I have a problem with the below query, when i do explain on the >>below query on my live database it doesnt use any index specified on the >>tables and does seq scan on the table which is 400k records. But if i >>copy the same table onto a different database on a different machine it >>uses all the indexes specified and query runs much quicker. >> >> > >It looks to me like you've never vacuumed/analyzed the copy, and so you >get a different plan there. The fact that that plan is better than the >one made with statistics is unhappy making :-( ... but when you only >show us EXPLAIN output rather than EXPLAIN ANALYZE, it's impossible to >speculate about why. Also, what PG version is this? > > regards, tom lane > > > Thanks! for the quick reply. I cant run the EXPLAIN ANALYZE on the live database because, it takes lot of time and hols up lot of other queries on the table. The postgres version I am using is 7.4 . when you say " i never vacuum/analyxed the copy" you mean the Live database ? or the copy of the live database ? . I run vacuum database daily on my live database as a part of daily maintanence. From pgsql-performance-owner@postgresql.org Fri Oct 8 20:32:17 2004 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 DB4EA329E3B for ; Fri, 8 Oct 2004 20:32:15 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 34411-07 for ; Fri, 8 Oct 2004 19:31:47 +0000 (GMT) Received: from server.gpdnet.co.uk (gpdnet.plus.com [212.56.100.243]) by svr1.postgresql.org (Postfix) with ESMTP id 8B47E329F5B for ; Fri, 8 Oct 2004 20:31:46 +0100 (BST) Received: from gary (unknown [192.168.1.2]) by server.gpdnet.co.uk (Postfix) with ESMTP id 49E15FBBB7 for ; Fri, 8 Oct 2004 20:31:53 +0100 (BST) From: "Gary Doades" To: pgsql-performance@postgresql.org Date: Fri, 08 Oct 2004 20:32:22 +0100 MIME-Version: 1.0 Subject: Odd planner choice? Message-ID: <4166F956.21282.29EC515A@localhost> X-mailer: Pegasus Mail for Windows (v4.12a) Content-type: Multipart/Alternative; boundary="Alt-Boundary-24044.703353178" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=2.1 tagged_above=0.0 required=5.0 tests=HTML_40_50, HTML_MESSAGE, HTML_TITLE_EMPTY, NO_DNS_FOR_FROM X-Spam-Level: ** X-Archive-Number: 200410/107 X-Sequence-Number: 8583 --Alt-Boundary-24044.703353178 Content-type: text/plain; charset=US-ASCII Content-transfer-encoding: 7BIT Content-description: Mail message body I'm looking at one of my standard queries and have encountered some strange performance problems. The query below is to search for vacant staff member date/time slots given a series of target date/times. The data contained in the booking_plan/staff_booking tables contain the existing bookings, so I'm looking for "clashing" bookings to eliminate them from a candidate list. The query is: select distinct b.staff_id from staff_booking b, booking_plan bp, t_search_reqt_dates rd where b.booking_id = bp.booking_id and rd.datetime_from <= bp.datetime_to and rd.datetime_to >= bp.datetime_from AND bp.booking_date between rd.reqt_date-1 and rd.reqt_date+1 and rd.search_id = 13 and rd.reqt_date between '2004-09-30' AND '2005-12-31' There are 197877 rows in staff_booking, 573416 rows in booking_plan and 26 rows in t_search_reqt_dates. The t_search reqt_dates is a temp table created and populated with the target date/times. The temp table is *not* analyzed, all the other are. The "good" query plan comes with the criteria on search_id and reqt_date given in the last two lines in the query. Note all the rows in the temp table are search_id = 13 and all the rows are between the two dates, so the whole 26 rows is always pulled out. In this case it is doing exactly what I expect. It is pulling all rows from the t_search_reqt_dates table, then pulling the relevant records from the booking_plan and then hashing with staff_booking. Excellent performance. The problem is I don't need the clauses for search_id and reqt_dates as the whole table is always read anyway. The good plan is because the planner thinks just one row will be read from t_search_reqt_dates. If I remove the redundant clauses, the planner now estimates 1000 rows returned from the table, not unreasonable since it has no statistics. But *why* in that case, with *more* estimated rows does it choose to materialize that table (26 rows) 573416 times!!! whenever it estimates more than one row it chooses the bad plan. I really want to remove the redundant clauses, but I can't. If I analyse the table, then it knows there are 26 rows and chooses the "bad" plan whatever I do. Any ideas??? Cheers, Gary. -------------------- Plans for above query ------------------------ Good QUERY PLAN Unique (cost=15440.83..15447.91 rows=462 width=4) (actual time=1342.000..1342.000 rows=110 loops=1) -> Sort (cost=15440.83..15444.37 rows=7081 width=4) (actual time=1342.000..1342.000 rows=2173 loops=1) Sort Key: b.staff_id -> Hash Join (cost=10784.66..15350.26 rows=7081 width=4) (actual time=601.000..1331.000 rows=2173 loops=1) Hash Cond: ("outer".booking_id = "inner".booking_id) -> Seq Scan on staff_booking b (cost=0.00..4233.39 rows=197877 width=8) (actual time=0.000..400.000 rows=197877 loops=1) -> Hash (cost=10781.12..10781.12 rows=7080 width=4) (actual time=591.000..591.000 rows=0 loops=1) -> Nested Loop (cost=0.00..10781.12 rows=7080 width=4) (actual time=10.000..581.000 rows=2173 loops=1) Join Filter: (("outer".datetime_from <= "inner".datetime_to) AND ("outer".datetime_to >= "inner".datetime_from)) -> Seq Scan on t_search_reqt_dates rd (cost=0.00..16.50 rows=1 width=20) (actual time=0.000..0.000 rows=26 loops=1) Filter: ((search_id = 13) AND (reqt_date >= '2004-09-30'::date) AND (reqt_date <= '2005-12-31'::date)) -> Index Scan using booking_plan_idx2 on booking_plan bp (cost=0.00..10254.91 rows=63713 width=24) (actual time=0.000..11.538 rows=5871 loops=26) Index Cond: ((bp.booking_date >= ("outer".reqt_date - 1)) AND (bp.booking_date <= ("outer".reqt_date + 1))) Total runtime: 1342.000 ms Bad QUERY PLAN Unique (cost=7878387.29..7885466.50 rows=462 width=4) (actual time=41980.000..41980.000 rows=110 loops=1) -> Sort (cost=7878387.29..7881926.90 rows=7079211 width=4) (actual time=41980.000..41980.000 rows=2173 loops=1) Sort Key: b.staff_id -> Nested Loop (cost=5314.32..7480762.73 rows=7079211 width=4) (actual time=6579.000..41980.000 rows=2173 loops=1) Join Filter: (("inner".datetime_from <= "outer".datetime_to) AND ("inner".datetime_to >= "outer".datetime_from) AND ("outer".booking_date >= ("inner".reqt_date - 1)) AND ("outer".booking_date <= ("inner".reqt_date + 1))) -> Hash Join (cost=5299.32..26339.73 rows=573416 width=24) (actual time=2413.000..7832.000 rows=573416 loops=1) Hash Cond: ("outer".booking_id = "inner".booking_id) -> Seq Scan on booking_plan bp (cost=0.00..7646.08 rows=573416 width=24) (actual time=0.000..1201.000 rows=573416 loops=1) -> Hash (cost=4233.39..4233.39 rows=197877 width=8) (actual time=811.000..811.000 rows=0 loops=1) -> Seq Scan on staff_booking b (cost=0.00..4233.39 rows=197877 width=8) (actual time=0.000..430.000 rows=197877 loops=1) -> Materialize (cost=15.00..20.00 rows=1000 width=20) (actual time=0.001..0.019 rows=26 loops=573416) -> Seq Scan on t_search_reqt_dates rd (cost=0.00..15.00 rows=1000 width=20) (actual time=0.000..0.000 rows=26 loops=1) Total runtime: 41980.000 ms --Alt-Boundary-24044.703353178 Content-type: text/html; charset=US-ASCII Content-transfer-encoding: 7BIT Content-description: Mail message body
I'm looking at one of my standard queries and have encountered some strange performance problems.

The query below is to search for vacant staff member date/time slots given a series of target date/times. The data contained in the booking_plan/staff_booking tables contain the existing bookings, so I'm looking for "clashing" bookings to eliminate them from a candidate list.

The query is:

select distinct b.staff_id  from staff_booking b, booking_plan bp, t_search_reqt_dates rd
where b.booking_id = bp.booking_id
and rd.datetime_from <= bp.datetime_to and rd.datetime_to >= bp.datetime_from
AND bp.booking_date between rd.reqt_date-1 and rd.reqt_date+1
and rd.search_id = 13
and rd.reqt_date between '2004-09-30' AND '2005-12-31'

There are 197877 rows in staff_booking, 573416 rows in booking_plan and 26 rows in t_search_reqt_dates.

The t_search reqt_dates is a temp table created and populated with the target date/times. The temp table is *not* analyzed, all the other are.

The "good" query plan comes with the criteria on search_id and reqt_date given in the last two lines in the query. Note all the rows in the temp table are search_id = 13 and all the rows are between the two dates, so the whole 26 rows is always pulled out.

In this case it is doing exactly what I expect. It is pulling all rows from the t_search_reqt_dates table, then pulling the relevant records from the booking_plan and then hashing with staff_booking. Excellent performance.

The problem is I don't need the clauses for search_id and reqt_dates as the whole table is always read anyway. The good plan is because the planner thinks just one row will be read from t_search_reqt_dates.

If I remove the redundant clauses, the planner now estimates 1000 rows returned from the table, not unreasonable since it has no statistics. But *why* in that case, with *more* estimated rows does it choose to materialize that table (26 rows) 573416 times!!!

whenever it estimates more than one row it chooses the bad plan.

I really want to remove the redundant clauses, but I can't. If I analyse the table, then it knows there are 26 rows and chooses the "bad" plan whatever I do.

Any ideas???

Cheers,
Gary.

-------------------- Plans for above query ------------------------

Good QUERY PLAN
Unique  (cost=15440.83..15447.91 rows=462 width=4) (actual time=1342.000..1342.000 rows=110 loops=1)
  ->  Sort  (cost=15440.83..15444.37 rows=7081 width=4) (actual time=1342.000..1342.000 rows=2173 loops=1)
        Sort Key: b.staff_id
        ->  Hash Join  (cost=10784.66..15350.26 rows=7081 width=4) (actual time=601.000..1331.000 rows=2173 loops=1)
              Hash Cond: ("outer".booking_id = "inner".booking_id)
              ->  Seq Scan on staff_booking b  (cost=0.00..4233.39 rows=197877 width=8) (actual time=0.000..400.000 rows=197877 loops=1)
              ->  Hash  (cost=10781.12..10781.12 rows=7080 width=4) (actual time=591.000..591.000 rows=0 loops=1)
                    ->  Nested Loop  (cost=0.00..10781.12 rows=7080 width=4) (actual time=10.000..581.000 rows=2173 loops=1)
                          Join Filter: (("outer".datetime_from <= "inner".datetime_to) AND ("outer".datetime_to >= "inner".datetime_from))
                          ->  Seq Scan on t_search_reqt_dates rd  (cost=0.00..16.50 rows=1 width=20) (actual time=0.000..0.000 rows=26 loops=1)
                                Filter: ((search_id = 13) AND (reqt_date >= '2004-09-30'::date) AND (reqt_date <= '2005-12-31'::date))
                          ->  Index Scan using booking_plan_idx2 on booking_plan bp (cost=0.00..10254.91 rows=63713 width=24) (actual time=0.000..11.538 rows=5871 loops=26)
                                Index Cond: ((bp.booking_date >= ("outer".reqt_date - 1)) AND (bp.booking_date <= ("outer".reqt_date + 1)))
Total runtime: 1342.000 ms


Bad QUERY PLAN
Unique  (cost=7878387.29..7885466.50 rows=462 width=4) (actual time=41980.000..41980.000 rows=110 loops=1)
  ->  Sort  (cost=7878387.29..7881926.90 rows=7079211 width=4) (actual time=41980.000..41980.000 rows=2173 loops=1)
        Sort Key: b.staff_id
        ->  Nested Loop  (cost=5314.32..7480762.73 rows=7079211 width=4) (actual time=6579.000..41980.000 rows=2173 loops=1)
              Join Filter: (("inner".datetime_from <= "outer".datetime_to) AND ("inner".datetime_to >= "outer".datetime_from) AND ("outer".booking_date >= ("inner".reqt_date - 1)) AND ("outer".booking_date <= ("inner".reqt_date + 1)))
              ->  Hash Join  (cost=5299.32..26339.73 rows=573416 width=24) (actual time=2413.000..7832.000 rows=573416 loops=1)
                    Hash Cond: ("outer".booking_id = "inner".booking_id)
                    ->  Seq Scan on booking_plan bp  (cost=0.00..7646.08 rows=573416 width=24) (actual time=0.000..1201.000 rows=573416 loops=1)
                    ->  Hash  (cost=4233.39..4233.39 rows=197877 width=8) (actual time=811.000..811.000 rows=0 loops=1)
                          ->  Seq Scan on staff_booking b  (cost=0.00..4233.39 rows=197877 width=8) (actual time=0.000..430.000 rows=197877 loops=1)
              ->  Materialize  (cost=15.00..20.00 rows=1000 width=20) (actual time=0.001..0.019 rows=26 loops=573416)
                    ->  Seq Scan on t_search_reqt_dates rd  (cost=0.00..15.00 rows=1000 width=20) (actual time=0.000..0.000 rows=26 loops=1)
Total runtime: 41980.000 ms
--Alt-Boundary-24044.703353178-- From pgsql-performance-owner@postgresql.org Fri Oct 8 20:46:13 2004 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 C32C432A187 for ; Fri, 8 Oct 2004 20:46:10 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 38300-06 for ; Fri, 8 Oct 2004 19:45:25 +0000 (GMT) Received: from server.gpdnet.co.uk (gpdnet.plus.com [212.56.100.243]) by svr1.postgresql.org (Postfix) with ESMTP id 5739E32A42A for ; Fri, 8 Oct 2004 20:45:26 +0100 (BST) Received: from gary (unknown [192.168.1.2]) by server.gpdnet.co.uk (Postfix) with ESMTP id 6D762FBBB7 for ; Fri, 8 Oct 2004 20:45:34 +0100 (BST) From: "Gary Doades" To: pgsql-performance@postgresql.org Date: Fri, 08 Oct 2004 20:46:03 +0100 MIME-Version: 1.0 Subject: Re: Odd planner choice? Message-ID: <4166FC8B.19594.29F8D953@localhost> In-reply-to: <4166F956.21282.29EC515A@localhost> X-mailer: Pegasus Mail for Windows (v4.12a) Content-type: text/plain; charset=US-ASCII Content-transfer-encoding: 7BIT Content-description: Mail message body X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.1 tagged_above=0.0 required=5.0 tests=NO_DNS_FOR_FROM X-Spam-Level: * X-Archive-Number: 200410/108 X-Sequence-Number: 8584 Oops, forgot to mention: PostgreSQL 8.0 beta 2 Windows. Thanks, Gary. On 8 Oct 2004 at 20:32, Gary Doades wrote: > > I'm looking at one of my standard queries and have encountered some strange performance > problems. > > The query below is to search for vacant staff member date/time slots given a series of target > date/times. The data contained in the booking_plan/staff_booking tables contain the existing > bookings, so I'm looking for "clashing" bookings to eliminate them from a candidate list. > > The query is: > > select distinct b.staff_id from staff_booking b, booking_plan bp, t_search_reqt_dates rd > where b.booking_id = bp.booking_id > and rd.datetime_from <= bp.datetime_to and rd.datetime_to >= bp.datetime_from > AND bp.booking_date between rd.reqt_date-1 and rd.reqt_date+1 > and rd.search_id = 13 > and rd.reqt_date between '2004-09-30' AND '2005-12-31' > > There are 197877 rows in staff_booking, 573416 rows in booking_plan and 26 rows in > t_search_reqt_dates. > > The t_search reqt_dates is a temp table created and populated with the target date/times. The > temp table is *not* analyzed, all the other are. > > The "good" query plan comes with the criteria on search_id and reqt_date given in the last two > lines in the query. Note all the rows in the temp table are search_id = 13 and all the rows are > between the two dates, so the whole 26 rows is always pulled out. > > In this case it is doing exactly what I expect. It is pulling all rows from the t_search_reqt_dates > table, then pulling the relevant records from the booking_plan and then hashing with > staff_booking. Excellent performance. > > The problem is I don't need the clauses for search_id and reqt_dates as the whole table is > always read anyway. The good plan is because the planner thinks just one row will be read from > t_search_reqt_dates. > > If I remove the redundant clauses, the planner now estimates 1000 rows returned from the table, > not unreasonable since it has no statistics. But *why* in that case, with *more* estimated rows > does it choose to materialize that table (26 rows) 573416 times!!! > > whenever it estimates more than one row it chooses the bad plan. > > I really want to remove the redundant clauses, but I can't. If I analyse the table, then it knows > there are 26 rows and chooses the "bad" plan whatever I do. > > Any ideas??? > > Cheers, > Gary. > > -------------------- Plans for above query ------------------------ > > Good QUERY PLAN > Unique (cost=15440.83..15447.91 rows=462 width=4) (actual time=1342.000..1342.000 > rows=110 loops=1) > -> Sort (cost=15440.83..15444.37 rows=7081 width=4) (actual time=1342.000..1342.000 > rows=2173 loops=1) > Sort Key: b.staff_id > -> Hash Join (cost=10784.66..15350.26 rows=7081 width=4) (actual > time=601.000..1331.000 rows=2173 loops=1) > Hash Cond: ("outer".booking_id = "inner".booking_id) > -> Seq Scan on staff_booking b (cost=0.00..4233.39 rows=197877 width=8) (actual > time=0.000..400.000 rows=197877 loops=1) > -> Hash (cost=10781.12..10781.12 rows=7080 width=4) (actual > time=591.000..591.000 rows=0 loops=1) > -> Nested Loop (cost=0.00..10781.12 rows=7080 width=4) (actual > time=10.000..581.000 rows=2173 loops=1) > Join Filter: (("outer".datetime_from <= "inner".datetime_to) AND > ("outer".datetime_to >= "inner".datetime_from)) > -> Seq Scan on t_search_reqt_dates rd (cost=0.00..16.50 rows=1 width=20) > (actual time=0.000..0.000 rows=26 loops=1) > Filter: ((search_id = 13) AND (reqt_date >= '2004-09-30'::date) AND > (reqt_date <= '2005-12-31'::date)) > -> Index Scan using booking_plan_idx2 on booking_plan bp > (cost=0.00..10254.91 rows=63713 width=24) (actual time=0.000..11.538 rows=5871 loops=26) > Index Cond: ((bp.booking_date >= ("outer".reqt_date - 1)) AND > (bp.booking_date <= ("outer".reqt_date + 1))) > Total runtime: 1342.000 ms > > > Bad QUERY PLAN > Unique (cost=7878387.29..7885466.50 rows=462 width=4) (actual time=41980.000..41980.000 > rows=110 loops=1) > -> Sort (cost=7878387.29..7881926.90 rows=7079211 width=4) (actual > time=41980.000..41980.000 rows=2173 loops=1) > Sort Key: b.staff_id > -> Nested Loop (cost=5314.32..7480762.73 rows=7079211 width=4) (actual > time=6579.000..41980.000 rows=2173 loops=1) > Join Filter: (("inner".datetime_from <= "outer".datetime_to) AND ("inner".datetime_to >= > "outer".datetime_from) AND ("outer".booking_date >= ("inner".reqt_date - 1)) AND > ("outer".booking_date <= ("inner".reqt_date + 1))) > -> Hash Join (cost=5299.32..26339.73 rows=573416 width=24) (actual > time=2413.000..7832.000 rows=573416 loops=1) > Hash Cond: ("outer".booking_id = "inner".booking_id) > -> Seq Scan on booking_plan bp (cost=0.00..7646.08 rows=573416 width=24) > (actual time=0.000..1201.000 rows=573416 loops=1) > -> Hash (cost=4233.39..4233.39 rows=197877 width=8) (actual > time=811.000..811.000 rows=0 loops=1) > -> Seq Scan on staff_booking b (cost=0.00..4233.39 rows=197877 width=8) > (actual time=0.000..430.000 rows=197877 loops=1) > -> Materialize (cost=15.00..20.00 rows=1000 width=20) (actual time=0.001..0.019 > rows=26 loops=573416) > -> Seq Scan on t_search_reqt_dates rd (cost=0.00..15.00 rows=1000 width=20) > (actual time=0.000..0.000 rows=26 loops=1) > Total runtime: 41980.000 ms > From pgsql-performance-owner@postgresql.org Fri Oct 8 21:05:20 2004 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 68A18329ED7 for ; Fri, 8 Oct 2004 21:05:17 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 44059-06 for ; Fri, 8 Oct 2004 20:04:49 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id B5DEF32A518 for ; Fri, 8 Oct 2004 21:04:48 +0100 (BST) 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 i98K4g31015494; Fri, 8 Oct 2004 16:04:43 -0400 (EDT) To: "Gary Doades" Cc: pgsql-performance@postgresql.org Subject: Re: Odd planner choice? In-reply-to: <4166F956.21282.29EC515A@localhost> References: <4166F956.21282.29EC515A@localhost> Comments: In-reply-to "Gary Doades" message dated "Fri, 08 Oct 2004 20:32:22 +0100" Date: Fri, 08 Oct 2004 16:04:42 -0400 Message-ID: <15493.1097265882@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.1 tagged_above=0.0 required=5.0 tests=NO_DNS_FOR_FROM X-Spam-Level: * X-Archive-Number: 200410/109 X-Sequence-Number: 8585 "Gary Doades" writes: > If I remove the redundant clauses, the planner now estimates 1000 rows returned from > the table, not unreasonable since it has no statistics. But *why* in that case, with *more* > estimated rows does it choose to materialize that table (26 rows) 573416 times!!! It isn't. It's materializing that once and scanning it 573416 times, once for each row in the outer relation. And this is not a bad plan given the estimates. If it had stuck to what you call the good plan, and there *had* been 1000 rows in the temp table, that plan would have run 1000 times longer than it did. As a general rule, if your complaint is that you get a bad plan for an unanalyzed table, the response is going to be "so analyze the table". regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 8 21:29:58 2004 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 B98FB32A088 for ; Fri, 8 Oct 2004 21:29:55 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 53183-01 for ; Fri, 8 Oct 2004 20:29:21 +0000 (GMT) Received: from server.gpdnet.co.uk (gpdnet.plus.com [212.56.100.243]) by svr1.postgresql.org (Postfix) with ESMTP id 58C71329E3B for ; Fri, 8 Oct 2004 21:29:19 +0100 (BST) Received: from gary (unknown [192.168.1.2]) by server.gpdnet.co.uk (Postfix) with ESMTP id DE04CFBBB7 for ; Fri, 8 Oct 2004 21:29:27 +0100 (BST) From: "Gary Doades" To: pgsql-performance@postgresql.org Date: Fri, 08 Oct 2004 21:29:57 +0100 MIME-Version: 1.0 Subject: Re: Odd planner choice? Message-ID: <416706D5.13084.2A2108B2@localhost> In-reply-to: <15493.1097265882@sss.pgh.pa.us> References: <4166F956.21282.29EC515A@localhost> X-mailer: Pegasus Mail for Windows (v4.12a) Content-type: text/plain; charset=US-ASCII Content-transfer-encoding: 7BIT Content-description: Mail message body X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.1 tagged_above=0.0 required=5.0 tests=NO_DNS_FOR_FROM X-Spam-Level: * X-Archive-Number: 200410/110 X-Sequence-Number: 8586 On 8 Oct 2004 at 16:04, Tom Lane wrote: > "Gary Doades" writes: > > If I remove the redundant clauses, the planner now estimates 1000 rows returned from > > the table, not unreasonable since it has no statistics. But *why* in that case, with *more* > > estimated rows does it choose to materialize that table (26 rows) 573416 times!!! > > It isn't. It's materializing that once and scanning it 573416 times, > once for each row in the outer relation. And this is not a bad plan > given the estimates. If it had stuck to what you call the good plan, > and there *had* been 1000 rows in the temp table, that plan would have > run 1000 times longer than it did. > > As a general rule, if your complaint is that you get a bad plan for an > unanalyzed table, the response is going to be "so analyze the table". > The problem is in this case is that if I *do* analyse the table I *always* get the bad plan. Bad in this case meaning the query takes a lot longer. I'm still not sure why it can't choose the better plan by just reading the 26 rows once and index scan the booking_plan table 26 times (as in the "good" plan). OK, with 1000 row estimate I can see that index scanning 1000 times into the booking_plan table would take some time, but the even if planner estimates 5 rows it still produces the same slow query. If I analyze the table it then knows there are 26 rows and therefore always goes slow. This is why I am not analyzing this table, to fool the planner into thinking there is only one row and produce a much faster access plan. Not ideal I know. Just using one redundant clause I now get: select distinct b.staff_id from staff_booking b, booking_plan bp, t_search_reqt_dates rd where b.booking_id = bp.booking_id and rd.datetime_from <= bp.datetime_to and rd.datetime_to >= bp.datetime_from AND bp.booking_date between rd.reqt_date-1 and rd.reqt_date+1 and rd.search_id = 13 QUERY PLAN Unique (cost=50885.97..50921.37 rows=462 width=4) (actual time=35231.000..35241.000 rows=110 loops=1) -> Sort (cost=50885.97..50903.67 rows=35397 width=4) (actual time=35231.000..35241.000 rows=2173 loops=1) Sort Key: b.staff_id -> Hash Join (cost=44951.32..50351.07 rows=35397 width=4) (actual time=34530.000..35231.000 rows=2173 loops=1) Hash Cond: ("outer".booking_id = "inner".booking_id) -> Seq Scan on staff_booking b (cost=0.00..4233.39 rows=197877 width=8) (actual time=0.000..351.000 rows=197877 loops=1) -> Hash (cost=44933.62..44933.62 rows=35397 width=4) (actual time=34530.000..34530.000 rows=0 loops=1) -> Nested Loop (cost=15.50..44933.62 rows=35397 width=4) (actual time=8342.000..34520.000 rows=2173 loops=1) Join Filter: (("inner".datetime_from <= "outer".datetime_to) AND ("inner".datetime_to >= "outer".datetime_from) AND ("outer".booking_date >= ("inner".reqt_date - 1)) AND ("outer".booking_date <= ("inner".reqt_date + 1))) -> Seq Scan on booking_plan bp (cost=0.00..7646.08 rows=573416 width=24) (actual time=0.000..1053.000 rows=573416 loops=1) -> Materialize (cost=15.50..15.53 rows=5 width=20) (actual time=0.001..0.019 rows=26 loops=573416) -> Seq Scan on t_search_reqt_dates rd (cost=0.00..15.50 rows=5 width=20) (actual time=0.000..0.000 rows=26 loops=1) Filter: (search_id = 13) Total runtime: 35241.000 ms If this is the only answer for now, then fair enough I will just have to do more testing. Regards, Gary. From pgsql-performance-owner@postgresql.org Fri Oct 8 22:42:15 2004 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 6D5EF329DBF for ; Fri, 8 Oct 2004 22:42:13 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 71225-10 for ; Fri, 8 Oct 2004 21:41:31 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 95FDE32A004 for ; Fri, 8 Oct 2004 22:41:32 +0100 (BST) 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 6480235; Fri, 08 Oct 2004 14:42:55 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: First set of OSDL Shared Mem scalability results, some wierdness ... Date: Fri, 8 Oct 2004 14:43:16 -0700 User-Agent: KMail/1.6.2 Cc: testperf-general@pgfoundry.org MIME-Version: 1.0 Content-Disposition: inline Message-Id: <200410081443.16109.josh@agliodbs.com> 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=1.1 tagged_above=0.0 required=5.0 tests=NO_DNS_FOR_FROM X-Spam-Level: * X-Archive-Number: 200410/111 X-Sequence-Number: 8587 Folks, I'm hoping that some of you can shed some light on this. I've been trying to peg the "sweet spot" for shared memory using OSDL's equipment. With Jan's new ARC patch, I was expecting that the desired amount of shared_buffers to be greatly increased. This has not turned out to be the case. The first test series was using OSDL's DBT2 (OLTP) test, with 150 "warehouses". All tests were run on a 4-way Pentium III 700mhz 3.8GB RAM system hooked up to a rather high-end storage device (14 spindles). Tests were on PostgreSQL 8.0b3, Linux 2.6.7. Here's a top-level summary: shared_buffers % RAM NOTPM20* 1000 0.2% 1287 23000 5% 1507 46000 10% 1481 69000 15% 1382 92000 20% 1375 115000 25% 1380 138000 30% 1344 * = New Order Transactions Per Minute, last 20 Minutes Higher is better. The maximum possible is 1800. As you can see, the "sweet spot" appears to be between 5% and 10% of RAM, which is if anything *lower* than recommendations for 7.4! This result is so surprising that I want people to take a look at it and tell me if there's something wrong with the tests or some bottlenecking factor that I've not seen. in order above: http://khack.osdl.org/stp/297959/ http://khack.osdl.org/stp/297960/ http://khack.osdl.org/stp/297961/ http://khack.osdl.org/stp/297962/ http://khack.osdl.org/stp/297963/ http://khack.osdl.org/stp/297964/ http://khack.osdl.org/stp/297965/ Please note that many of the Graphs in these reports are broken. For one thing, some aren't recorded (flat lines) and the CPU usage graph has mislabeled lines. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Fri Oct 8 23:13:16 2004 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 6368432A10B for ; Fri, 8 Oct 2004 23:13:14 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78395-10 for ; Fri, 8 Oct 2004 22:13:02 +0000 (GMT) Received: from mx1.neopolitan.us (mx1.neopolitan.us [65.87.16.224]) by svr1.postgresql.org (Postfix) with ESMTP id 85DE232A12D for ; Fri, 8 Oct 2004 23:13:03 +0100 (BST) Received: from [65.87.16.98] (HELO vulture.corp.neopolitan.com) by mx1.neopolitan.us (CommuniGate Pro SMTP 4.1.8) with ESMTP-TLS id 6056326; Fri, 08 Oct 2004 15:13:00 -0700 Subject: Re: First set of OSDL Shared Mem scalability results, From: "J. Andrew Rogers" To: josh@agliodbs.com, pgsql-performance@postgresql.org In-Reply-To: <200410081443.16109.josh@agliodbs.com> References: <200410081443.16109.josh@agliodbs.com> Content-Type: text/plain Organization: Message-Id: <1097273580.25588.18.camel@vulture.corp.neopolitan.com> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.2.2 (1.2.2-5) Date: 08 Oct 2004 15:13:00 -0700 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/112 X-Sequence-Number: 8588 I have an idea that makes some assumptions about internals that I think are correct. When you have a huge number of buffers in a list that has to be traversed to look for things in cache, e.g. 100k, you will generate an almost equivalent number of cache line misses on the processor to jump through all those buffers. As I understand it (and I haven't looked so I could be wrong), the buffer cache is searched by traversing it sequentially. OTOH, it seems reasonable to me that the OS disk cache may actually be using a tree structure that would generate vastly fewer cache misses by comparison to find a buffer. This could mean a substantial linear search cost as a function of the number of buffers, big enough to rise above the noise floor when you have hundreds of thousands of buffers. Cache misses start to really add up when a code path generates many, many thousands of them, and differences in the access path between the buffer cache and disk cache would be reflected when you have that many buffers. I've seen these types of unexpected performance anomalies before that got traced back to code patterns and cache efficiency and gotten integer factors improvements by making some seemingly irrelevant code changes. So I guess my question would be 1) are my assumptions about the internals correct, and 2) if they are, is there a way to optimize searching the buffer cache so that a search doesn't iterate over a really long buffer list that is bottlenecked on cache line replacement. My random thought of the day, j. andrew rogers From pgsql-performance-owner@postgresql.org Fri Oct 8 23:21:44 2004 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 64BE6329DBF for ; Fri, 8 Oct 2004 23:21:43 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 79799-10 for ; Fri, 8 Oct 2004 22:21:31 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 57C0132A12D for ; Fri, 8 Oct 2004 23:21:33 +0100 (BST) 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 i98MLSRV021366; Fri, 8 Oct 2004 18:21:28 -0400 (EDT) To: josh@agliodbs.com Cc: pgsql-performance@postgresql.org, testperf-general@pgfoundry.org, Jan Wieck Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... In-reply-to: <200410081443.16109.josh@agliodbs.com> References: <200410081443.16109.josh@agliodbs.com> Comments: In-reply-to Josh Berkus message dated "Fri, 08 Oct 2004 14:43:16 -0700" Date: Fri, 08 Oct 2004 18:21:28 -0400 Message-ID: <21365.1097274088@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/113 X-Sequence-Number: 8589 Josh Berkus writes: > Here's a top-level summary: > shared_buffers % RAM NOTPM20* > 1000 0.2% 1287 > 23000 5% 1507 > 46000 10% 1481 > 69000 15% 1382 > 92000 20% 1375 > 115000 25% 1380 > 138000 30% 1344 > As you can see, the "sweet spot" appears to be between 5% and 10% of RAM, > which is if anything *lower* than recommendations for 7.4! This doesn't actually surprise me a lot. There are a number of aspects of Postgres that will get slower the more buffers there are. One thing that I hadn't focused on till just now, which is a new overhead in 8.0, is that StrategyDirtyBufferList() scans the *entire* buffer list *every time it's called*, which is to say once per bgwriter loop. And to add insult to injury, it's doing that with the BufMgrLock held (not that it's got any choice). We could alleviate this by changing the API between this function and BufferSync, such that StrategyDirtyBufferList can stop as soon as it's found all the buffers that are going to be written in this bgwriter cycle ... but AFAICS that means abandoning the "bgwriter_percent" knob since you'd never really know how many dirty pages there were altogether. BTW, what is the actual size of the test database (disk footprint wise) and how much of that do you think is heavily accessed during the run? It's possible that the test conditions are such that adjusting shared_buffers isn't going to mean anything anyway. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 8 23:32:51 2004 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 204EE32A10B for ; Fri, 8 Oct 2004 23:32:45 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 85620-03 for ; Fri, 8 Oct 2004 22:32:33 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id C639732A102 for ; Fri, 8 Oct 2004 23:32:35 +0100 (BST) 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 i98MWWwB021490; Fri, 8 Oct 2004 18:32:33 -0400 (EDT) To: "J. Andrew Rogers" Cc: josh@agliodbs.com, pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, In-reply-to: <1097273580.25588.18.camel@vulture.corp.neopolitan.com> References: <200410081443.16109.josh@agliodbs.com> <1097273580.25588.18.camel@vulture.corp.neopolitan.com> Comments: In-reply-to "J. Andrew Rogers" message dated "08 Oct 2004 15:13:00 -0700" Date: Fri, 08 Oct 2004 18:32:32 -0400 Message-ID: <21489.1097274752@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/114 X-Sequence-Number: 8590 "J. Andrew Rogers" writes: > As I understand it (and I haven't looked so I could be wrong), the > buffer cache is searched by traversing it sequentially. You really should look first. The main-line code paths use hashed lookups. There are some cases that do linear searches through the buffer headers or the CDB lists; in theory those are supposed to be non-performance-critical cases, though I am suspicious that some are not (see other response). In any case, those structures are considerably more compact than the buffers proper, and I doubt that cache misses per se are the killer factor. This does raise a question for Josh though, which is "where's the oprofile results?" If we do have major problems at the level of cache misses then oprofile would be able to prove it. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 8 23:40:05 2004 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 21E0632A10B for ; Fri, 8 Oct 2004 23:40:02 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 85227-06 for ; Fri, 8 Oct 2004 22:39:49 +0000 (GMT) Received: from trofast.sesse.net (trofast.sesse.net [129.241.93.32]) by svr1.postgresql.org (Postfix) with ESMTP id 04769329E5C for ; Fri, 8 Oct 2004 23:39:51 +0100 (BST) Received: from sesse by trofast.sesse.net with local (Exim 3.36 #1 (Debian)) id 1CG3Or-0003vH-00 for ; Sat, 09 Oct 2004 00:39:45 +0200 Date: Sat, 9 Oct 2004 00:39:45 +0200 From: "Steinar H. Gunderson" To: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, Message-ID: <20041008223945.GA15067@uio.no> Mail-Followup-To: pgsql-performance@postgresql.org References: <200410081443.16109.josh@agliodbs.com> <1097273580.25588.18.camel@vulture.corp.neopolitan.com> <21489.1097274752@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <21489.1097274752@sss.pgh.pa.us> X-Operating-System: Linux 2.6.6 on a i686 User-Agent: Mutt/1.5.6+20040818i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/115 X-Sequence-Number: 8591 On Fri, Oct 08, 2004 at 06:32:32PM -0400, Tom Lane wrote: > This does raise a question for Josh though, which is "where's the > oprofile results?" If we do have major problems at the level of cache > misses then oprofile would be able to prove it. Or cachegrind. I've found it to be really effective at pinpointing cache misses in the past (one CPU-intensive routine was sped up by 30% just by avoiding a memory clear). :-) /* Steinar */ -- Homepage: http://www.sesse.net/ From pgsql-performance-owner@postgresql.org Sat Oct 9 00:07:43 2004 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 4F38D32A5DB for ; Sat, 9 Oct 2004 00:06:45 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 94346-01 for ; Fri, 8 Oct 2004 23:06:34 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id C2EB932A42A for ; Sat, 9 Oct 2004 00:06:34 +0100 (BST) 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 6480716; Fri, 08 Oct 2004 16:07:55 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, Date: Fri, 8 Oct 2004 16:08:14 -0700 User-Agent: KMail/1.6.2 Cc: Tom Lane , "J. Andrew Rogers" References: <200410081443.16109.josh@agliodbs.com> <1097273580.25588.18.camel@vulture.corp.neopolitan.com> <21489.1097274752@sss.pgh.pa.us> In-Reply-To: <21489.1097274752@sss.pgh.pa.us> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable Message-Id: <200410081608.14043.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/116 X-Sequence-Number: 8592 Tom, > This does raise a question for Josh though, which is "where's the > oprofile results?" =C2=A0If we do have major problems at the level of cac= he > misses then oprofile would be able to prove it. Missing, I'm afraid. OSDL has been having technical issues with STP all we= ek. Hopefully the next test run will have them. --=20 --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Sat Oct 9 00:30:36 2004 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 189E8329FF2 for ; Sat, 9 Oct 2004 00:30:12 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 95808-08 for ; Fri, 8 Oct 2004 23:29:59 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 48472329F17 for ; Sat, 9 Oct 2004 00:30:01 +0100 (BST) 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 6480816; Fri, 08 Oct 2004 16:31:19 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... Date: Fri, 8 Oct 2004 16:31:41 -0700 User-Agent: KMail/1.6.2 Cc: Tom Lane , testperf-general@pgfoundry.org, Jan Wieck References: <200410081443.16109.josh@agliodbs.com> <21365.1097274088@sss.pgh.pa.us> In-Reply-To: <21365.1097274088@sss.pgh.pa.us> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Message-Id: <200410081631.41545.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/117 X-Sequence-Number: 8593 Tom, > BTW, what is the actual size of the test database (disk footprint wise) > and how much of that do you think is heavily accessed during the run? > It's possible that the test conditions are such that adjusting > shared_buffers isn't going to mean anything anyway. The raw data is 32GB, but a lot of the activity is incremental, that is inserts and updates to recent inserts. Still, according to Mark, most of the data does get queried in the course of filling orders. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Sat Oct 9 04:20:05 2004 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 B841432B43C for ; Sat, 9 Oct 2004 04:20:03 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 45254-05 for ; Sat, 9 Oct 2004 03:19:58 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id 1634632B429 for ; Sat, 9 Oct 2004 04:19:57 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i993Ju2B046079 for ; Sat, 9 Oct 2004 03:19:56 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i992nwbS038767 for pgsql-performance@postgresql.org; Sat, 9 Oct 2004 02:49:58 GMT From: Christopher Browne X-Newsgroups: comp.databases.postgresql.performance Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... Date: Fri, 08 Oct 2004 22:10:19 -0400 Organization: cbbrowne Computing Inc Lines: 48 Message-ID: References: <200410081443.16109.josh@agliodbs.com> 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 (Security Through Obscurity, linux) Cancel-Lock: sha1:MSmAo22qNssFcOru6grzK7aLro0= To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/118 X-Sequence-Number: 8594 josh@agliodbs.com (Josh Berkus) wrote: > I've been trying to peg the "sweet spot" for shared memory using > OSDL's equipment. With Jan's new ARC patch, I was expecting that > the desired amount of shared_buffers to be greatly increased. This > has not turned out to be the case. That doesn't surprise me. My primary expectation would be that ARC would be able to make small buffers much more effective alongside vacuums and seq scans than they used to be. That does not establish anything about the value of increasing the size buffer caches... > This result is so surprising that I want people to take a look at it > and tell me if there's something wrong with the tests or some > bottlenecking factor that I've not seen. I'm aware of two conspicuous scenarios where ARC would be expected to _substantially_ improve performance: 1. When it allows a VACUUM not to throw useful data out of the shared cache in that VACUUM now only 'chews' on one page of the cache; 2. When it allows a Seq Scan to not push useful data out of the shared cache, for much the same reason. I don't imagine either scenario are prominent in the OSDL tests. Increasing the number of cache buffers _is_ likely to lead to some slowdowns: - Data that passes through the cache also passes through kernel cache, so it's recorded twice, and read twice... - The more cache pages there are, the more work is needed for PostgreSQL to manage them. That will notably happen anywhere that there is a need to scan the cache. - If there are any inefficiencies in how the OS kernel manages shared memory, as their size scales, well, that will obviously cause a slowdown. -- If this was helpful, rate me http://www.ntlug.org/~cbbrowne/internet.html "One World. One Web. One Program." -- MICROS~1 hype "Ein Volk, ein Reich, ein Fuehrer" -- Nazi hype (One people, one country, one leader) From pgsql-performance-owner@postgresql.org Sat Oct 9 12:21:03 2004 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 5763232A1F1 for ; Sat, 9 Oct 2004 12:20:58 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 34288-05 for ; Sat, 9 Oct 2004 11:20:50 +0000 (GMT) Received: from sccrmhc13.comcast.net (sccrmhc13.comcast.net [204.127.202.64]) by svr1.postgresql.org (Postfix) with ESMTP id 875FC32A1B6 for ; Sat, 9 Oct 2004 12:20:50 +0100 (BST) Received: from sysexperts.com (c-24-6-183-218.client.comcast.net[24.6.183.218]) by comcast.net (sccrmhc13) with ESMTP id <2004100911204901600c0k3ne>; Sat, 9 Oct 2004 11:20:50 +0000 Received: from localhost (localhost [127.0.0.1]) (uid 1000) by filer with local; Sat, 09 Oct 2004 04:20:48 -0700 id 0000545D.4167C990.00003826 Date: Sat, 9 Oct 2004 04:20:48 -0700 From: Kevin Brown To: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... Message-ID: <20041009112048.GA665@filer> Mail-Followup-To: Kevin Brown , pgsql-performance@postgresql.org References: <200410081443.16109.josh@agliodbs.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline In-Reply-To: Organization: Frobozzco International User-Agent: Mutt/1.5.6+20040818i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/119 X-Sequence-Number: 8595 Christopher Browne wrote: > Increasing the number of cache buffers _is_ likely to lead to some > slowdowns: > > - Data that passes through the cache also passes through kernel > cache, so it's recorded twice, and read twice... Even worse, memory that's used for the PG cache is memory that's not available to the kernel's page cache. Even if the overall memory usage in the system isn't enough to cause some paging to disk, most modern kernels will adjust the page/disk cache size dynamically to fit the memory demands of the system, which in this case means it'll be smaller if running programs need more memory for their own use. This is why I sometimes wonder whether or not it would be a win to use mmap() to access the data and index files -- doing so under a truly modern OS would surely at the very least save a buffer copy (from the page/disk cache to program memory) because the OS could instead direcly map the buffer cache pages directly to the program's memory space. Since PG often has to have multiple files open at the same time, and in a production database many of those files will be rather large, PG would have to limit the size of the mmap()ed region on 32-bit platforms, which means that things like the order of mmap() operations to access various parts of the file can become just as important in the mmap()ed case as it is in the read()/write() case (if not more so!). I would imagine that the use of mmap() on a 64-bit platform would be a much, much larger win because PG would most likely be able to mmap() entire files and let the OS work out how to order disk reads and writes. The biggest problem as I see it is that (I think) mmap() would have to be made to cooperate with malloc() for virtual address space. I suspect issues like this have already been worked out by others, however... -- Kevin Brown kevin@sysexperts.com From pgsql-performance-owner@postgresql.org Sat Oct 9 16:05:30 2004 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 C9946329DBE for ; Sat, 9 Oct 2004 16:05:28 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 79030-09 for ; Sat, 9 Oct 2004 15:05:18 +0000 (GMT) Received: from outbound.mailhop.org (outbound.mailhop.org [63.208.196.171]) by svr1.postgresql.org (Postfix) with ESMTP id 0433832A928 for ; Sat, 9 Oct 2004 16:05:16 +0100 (BST) Received: from ool-43529386.dyn.optonline.net ([67.82.147.134] helo=[192.168.1.9]) by outbound.mailhop.org with esmtpsa (TLSv1:AES256-SHA:256) (Exim 4.42) id 1CGImZ-0000Rs-DG; Sat, 09 Oct 2004 11:05:15 -0400 Message-ID: <4167FD71.9050501@zeut.net> Date: Sat, 09 Oct 2004 11:02:09 -0400 From: Matthew User-Agent: Mozilla Thunderbird 0.7 (Windows/20040616) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Christopher Browne Cc: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some References: <200410081443.16109.josh@agliodbs.com> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Mail-Handler: MailHop Outbound by DynDNS.org X-Originating-IP: 67.82.147.134 X-Report-Abuse-To: abuse@dyndns.org (see http://www.mailhop.org/outbound/abuse.html for abuse reporting information) X-MHO-User: Zeut X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/120 X-Sequence-Number: 8596 Christopher Browne wrote: >josh@agliodbs.com (Josh Berkus) wrote: > > >>This result is so surprising that I want people to take a look at it >>and tell me if there's something wrong with the tests or some >>bottlenecking factor that I've not seen. >> >> >I'm aware of two conspicuous scenarios where ARC would be expected to >_substantially_ improve performance: > > 1. When it allows a VACUUM not to throw useful data out of > the shared cache in that VACUUM now only 'chews' on one > page of the cache; > Right, Josh, I assume you didn't run these test with pg_autovacuum running, which might be interesting. Also how do these numbers compare to 7.4? They may not be what you expected, but they might still be an improvment. Matthew From pgsql-performance-owner@postgresql.org Sat Oct 9 16:07:33 2004 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 12BE732A0F0 for ; Sat, 9 Oct 2004 16:07:32 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83161-02 for ; Sat, 9 Oct 2004 15:07:30 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 0E60B329DBE for ; Sat, 9 Oct 2004 16:07:29 +0100 (BST) 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 i99F7TRX027697; Sat, 9 Oct 2004 11:07:29 -0400 (EDT) To: Kevin Brown Cc: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... In-reply-to: <20041009112048.GA665@filer> References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> Comments: In-reply-to Kevin Brown message dated "Sat, 09 Oct 2004 04:20:48 -0700" Date: Sat, 09 Oct 2004 11:07:28 -0400 Message-ID: <27696.1097334448@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/121 X-Sequence-Number: 8597 Kevin Brown writes: > This is why I sometimes wonder whether or not it would be a win to use > mmap() to access the data and index files -- mmap() is Right Out because it does not afford us sufficient control over when changes to the in-memory data will propagate to disk. The address-space-management problems you describe are also a nasty headache, but that one is the showstopper. regards, tom lane From pgsql-performance-owner@postgresql.org Sat Oct 9 21:37:34 2004 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 B2F3B32A010 for ; Sat, 9 Oct 2004 21:37:28 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49483-06 for ; Sat, 9 Oct 2004 20:37:16 +0000 (GMT) Received: from rwcrmhc13.comcast.net (rwcrmhc13.comcast.net [204.127.198.39]) by svr1.postgresql.org (Postfix) with ESMTP id B975932A00B for ; Sat, 9 Oct 2004 21:37:16 +0100 (BST) Received: from sysexperts.com (c-24-6-183-218.client.comcast.net[24.6.183.218]) by comcast.net (rwcrmhc13) with ESMTP id <20041009203713015003c198e>; Sat, 9 Oct 2004 20:37:13 +0000 Received: from localhost (localhost [127.0.0.1]) (uid 1000) by filer with local; Sat, 09 Oct 2004 13:37:12 -0700 id 000064D1.41684BF8.00005448 Date: Sat, 9 Oct 2004 13:37:12 -0700 From: Kevin Brown To: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... Message-ID: <20041009203710.GB665@filer> Mail-Followup-To: Kevin Brown , pgsql-performance@postgresql.org References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline In-Reply-To: <27696.1097334448@sss.pgh.pa.us> Organization: Frobozzco International User-Agent: Mutt/1.5.6+20040818i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/122 X-Sequence-Number: 8598 Tom Lane wrote: > Kevin Brown writes: > > This is why I sometimes wonder whether or not it would be a win to use > > mmap() to access the data and index files -- > > mmap() is Right Out because it does not afford us sufficient control > over when changes to the in-memory data will propagate to disk. The > address-space-management problems you describe are also a nasty > headache, but that one is the showstopper. Huh? Surely fsync() or fdatasync() of the file descriptor associated with the mmap()ed region at the appropriate times would accomplish much of this? I'm particularly confused since PG's entire approach to disk I/O is predicated on the notion that the OS, and not PG, is the best arbiter of when data hits the disk. Otherwise it would be using raw partitions for the highest-speed data store, yes? Also, there isn't any particular requirement to use mmap() for everything -- you can use traditional open/write/close calls for the WAL and mmap() for the data/index files (but it wouldn't surprise me if this would require some extensive code changes). That said, if it's typical for many changes to made to a page internally before PG needs to commit that page to disk, then your argument makes sense, and that's especially true if we simply cannot have the page written to disk in a partially-modified state (something I can easily see being an issue for the WAL -- would the same hold true of the index/data files?). -- Kevin Brown kevin@sysexperts.com From pgsql-performance-owner@postgresql.org Sat Oct 9 22:01:15 2004 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 AF56832A1BF for ; Sat, 9 Oct 2004 22:01:13 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 55961-09 for ; Sat, 9 Oct 2004 21:01:04 +0000 (GMT) Received: from rwcrmhc11.comcast.net (rwcrmhc11.comcast.net [204.127.198.35]) by svr1.postgresql.org (Postfix) with ESMTP id 4330032A143 for ; Sat, 9 Oct 2004 22:01:03 +0100 (BST) Received: from sysexperts.com (c-24-6-183-218.client.comcast.net[24.6.183.218]) by comcast.net (rwcrmhc11) with ESMTP id <2004100921010201300c3se8e>; Sat, 9 Oct 2004 21:01:02 +0000 Received: from localhost (localhost [127.0.0.1]) (uid 1000) by filer with local; Sat, 09 Oct 2004 14:01:02 -0700 id 000064D1.4168518E.0000554D Date: Sat, 9 Oct 2004 14:01:02 -0700 From: Kevin Brown To: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... Message-ID: <20041009210102.GC665@filer> Mail-Followup-To: Kevin Brown , pgsql-performance@postgresql.org References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline In-Reply-To: <20041009203710.GB665@filer> Organization: Frobozzco International User-Agent: Mutt/1.5.6+20040818i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/123 X-Sequence-Number: 8599 I wrote: > That said, if it's typical for many changes to made to a page > internally before PG needs to commit that page to disk, then your > argument makes sense, and that's especially true if we simply cannot > have the page written to disk in a partially-modified state (something > I can easily see being an issue for the WAL -- would the same hold > true of the index/data files?). Also, even if multiple changes would be made to the page, with the page being valid for a disk write only after all such changes are made, the use of mmap() (in conjunction with an internal buffer that would then be copied to the mmap()ed memory space at the appropriate time) would potentially save a system call over the use of write() (even if write() were used to write out multiple pages). However, there is so much lower-hanging fruit than this that an mmap() implementation almost certainly isn't worth pursuing for this alone. So: it seems to me that mmap() is worth pursuing only if most internal buffers tend to be written to only once or if it's acceptable for a partially modified data/index page to be written to disk (which I suppose could be true for data/index pages in the face of a rock-solid WAL). -- Kevin Brown kevin@sysexperts.com From pgsql-performance-owner@postgresql.org Sun Oct 10 00:06:56 2004 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 B3DDF32A235 for ; Sun, 10 Oct 2004 00:05:48 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78696-07 for ; Sat, 9 Oct 2004 23:05:36 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 84E24329D95 for ; Sun, 10 Oct 2004 00:05:37 +0100 (BST) 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 i99N5bBm004860; Sat, 9 Oct 2004 19:05:37 -0400 (EDT) To: Kevin Brown Cc: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... In-reply-to: <20041009203710.GB665@filer> References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> Comments: In-reply-to Kevin Brown message dated "Sat, 09 Oct 2004 13:37:12 -0700" Date: Sat, 09 Oct 2004 19:05:37 -0400 Message-ID: <4859.1097363137@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/124 X-Sequence-Number: 8600 Kevin Brown writes: > Tom Lane wrote: >> mmap() is Right Out because it does not afford us sufficient control >> over when changes to the in-memory data will propagate to disk. > ... that's especially true if we simply cannot > have the page written to disk in a partially-modified state (something > I can easily see being an issue for the WAL -- would the same hold > true of the index/data files?). You're almost there. Remember the fundamental WAL rule: log entries must hit disk before the data changes they describe. That means that we need not only a way of forcing changes to disk (fsync) but a way of being sure that changes have *not* gone to disk yet. In the existing implementation we get that by just not issuing write() for a given page until we know that the relevant WAL log entries are fsync'd down to disk. (BTW, this is what the LSN field on every page is for: it tells the buffer manager the latest WAL offset that has to be flushed before it can safely write the page.) mmap provides msync which is comparable to fsync, but AFAICS it provides no way to prevent an in-memory change from reaching disk too soon. This would mean that WAL entries would have to be written *and flushed* before we could make the data change at all, which would convert multiple updates of a single page into a series of write-and- wait-for-WAL-fsync steps. Not good. fsync'ing WAL once per transaction is bad enough, once per atomic action is intolerable. There is another reason for doing things this way. Consider a backend that goes haywire and scribbles all over shared memory before crashing. When the postmaster sees the abnormal child termination, it forcibly kills the other active backends and discards shared memory altogether. This gives us fairly good odds that the crash did not affect any data on disk. It's not perfect of course, since another backend might have been in process of issuing a write() when the disaster happens, but it's pretty good; and I think that that isolation has a lot to do with PG's good reputation for not corrupting data in crashes. If we had a large fraction of the address space mmap'd then this sort of crash would be just about guaranteed to propagate corruption into the on-disk files. regards, tom lane From pgsql-performance-owner@postgresql.org Sun Oct 10 10:20:25 2004 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 A767332A9CD for ; Sun, 10 Oct 2004 10:20:17 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 87000-03 for ; Sun, 10 Oct 2004 09:20:12 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id 0981E32A038 for ; Sun, 10 Oct 2004 10:20:13 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i9A9KA2B087637 for ; Sun, 10 Oct 2004 09:20:10 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i9A9JxJA087537 for pgsql-performance@postgresql.org; Sun, 10 Oct 2004 09:19:59 GMT From: Gaetano Mendola X-Newsgroups: comp.databases.postgresql.performance Subject: kernel 2.6 synchronous directory Date: Sun, 10 Oct 2004 11:19:59 +0200 Organization: PYRENET Midi-pyrenees Provider Lines: 16 Message-ID: Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Complaints-To: abuse@pyrenet.fr User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en X-Enigmail-Version: 0.86.1.0 X-Enigmail-Supports: pgp-inline, pgp-mime To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/125 X-Sequence-Number: 8601 Hi all, I'm wondering if setting the $PG_DATA directory as synchronous directory in order to make a crash event more safe will penalyze the performances. If you run a kernel 2.6 the command is: chattr +S $PG_DATA Regards Gaetano Mendola From pgsql-performance-owner@postgresql.org Sun Oct 10 10:50:24 2004 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 5EE0E32B429 for ; Sun, 10 Oct 2004 10:50:21 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 90836-05 for ; Sun, 10 Oct 2004 09:50:16 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id 67DA632B41C for ; Sun, 10 Oct 2004 10:50:15 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i9A9oA2B093342 for ; Sun, 10 Oct 2004 09:50:10 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i9A9POaU088856 for pgsql-performance@postgresql.org; Sun, 10 Oct 2004 09:25:24 GMT From: Gaetano Mendola X-Newsgroups: comp.databases.postgresql.performance Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness Date: Sun, 10 Oct 2004 11:25:23 +0200 Organization: PYRENET Midi-pyrenees Provider Lines: 26 Message-ID: <41690003.2010905@bigfoot.com> References: <200410081443.16109.josh@agliodbs.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Complaints-To: abuse@pyrenet.fr To: Josh Berkus User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en In-Reply-To: <200410081443.16109.josh@agliodbs.com> X-Enigmail-Version: 0.86.1.0 X-Enigmail-Supports: pgp-inline, pgp-mime To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/126 X-Sequence-Number: 8602 Josh Berkus wrote: > Folks, > > I'm hoping that some of you can shed some light on this. > > I've been trying to peg the "sweet spot" for shared memory using OSDL's > equipment. With Jan's new ARC patch, I was expecting that the desired > amount of shared_buffers to be greatly increased. This has not turned out to > be the case. > > The first test series was using OSDL's DBT2 (OLTP) test, with 150 > "warehouses". All tests were run on a 4-way Pentium III 700mhz 3.8GB RAM > system hooked up to a rather high-end storage device (14 spindles). Tests > were on PostgreSQL 8.0b3, Linux 2.6.7. I'd like to see these tests running using the cpu affinity capability in order to oblige a backend to not change CPU during his life, this could drastically increase the cache hit. Regards Gaetano Mendola From pgsql-performance-owner@postgresql.org Sun Oct 10 22:49:02 2004 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 662F032ACAB for ; Sun, 10 Oct 2004 22:48:59 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 79976-10 for ; Sun, 10 Oct 2004 21:48:52 +0000 (GMT) Received: from zigo.dhs.org (as2-4-3.an.g.bonet.se [194.236.34.191]) by svr1.postgresql.org (Postfix) with ESMTP id 10AF132AC9D for ; Sun, 10 Oct 2004 22:48:49 +0100 (BST) Received: from zigo.zigo.dhs.org (zigo.zigo.dhs.org [192.168.0.1]) by zigo.dhs.org (Postfix) with ESMTP id 414BF8467; Sun, 10 Oct 2004 23:48:48 +0200 (CEST) Date: Sun, 10 Oct 2004 23:48:48 +0200 (CEST) From: Dennis Bjorklund To: Josh Berkus Cc: pgsql-performance@postgresql.org, Subject: Re: First set of OSDL Shared Mem scalability results, some In-Reply-To: <200410081443.16109.josh@agliodbs.com> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/127 X-Sequence-Number: 8603 On Fri, 8 Oct 2004, Josh Berkus wrote: > As you can see, the "sweet spot" appears to be between 5% and 10% of RAM, > which is if anything *lower* than recommendations for 7.4! What recommendation is that? To have shared buffers being about 10% of the ram sounds familiar to me. What was recommended for 7.4? In the past we used to say that the worst value is 50% since then the same things might be cached both by pg and the os disk cache. Why do we excpect the shared buffer size sweet spot to change because of the new arc stuff? And why would it make it better to have bigger shared mem? Wouldn't it be the opposit, that now we don't invalidate as much of the cache for vacuums and seq. scan so now we can do as good caching as before but with less shared buffers. That said, testing and getting some numbers of good sizes for shared mem is good. -- /Dennis Bj�rklund From pgsql-performance-owner@postgresql.org Mon Oct 11 10:54:54 2004 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 BA7DF32AA72 for ; Mon, 11 Oct 2004 10:54:52 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 45725-05 for ; Mon, 11 Oct 2004 09:54:45 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.199]) by svr1.postgresql.org (Postfix) with ESMTP id 2FEC732A59E for ; Mon, 11 Oct 2004 10:54:46 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 77so211241rnk for ; Mon, 11 Oct 2004 02:54:41 -0700 (PDT) Received: by 10.38.73.61 with SMTP id v61mr1530119rna; Mon, 11 Oct 2004 02:54:41 -0700 (PDT) Received: by 10.38.163.22 with HTTP; Mon, 11 Oct 2004 02:54:41 -0700 (PDT) Message-ID: <758d5e7f04101102543eab8dca@mail.gmail.com> Date: Mon, 11 Oct 2004 11:54:41 +0200 From: Dawid Kuroczko Reply-To: Dawid Kuroczko To: pgsql-performance@postgresql.org Subject: Views, joins and LIMIT 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=0.9 tagged_above=0.0 required=5.0 tests=FROM_ENDS_IN_NUMS X-Spam-Level: X-Archive-Number: 200410/128 X-Sequence-Number: 8604 I've been wondering... Suppose we have two tables CREATE TABLE messages ( message_id serial PRIMARY KEY, message text NOT NULL ); CREATE TABLE entries ( entry_id serial PRIMARY KEY, message_id integer NOT NULL REFERENCES messages ); And we have a join: SELECT entry_id,message FROM entries NATURAL JOIN messages ORDER BY entry_id DESC LIMIT 10; The typical planners order of doing things is -- join the tables, perform sort, perform limit. But in the above case (which I guess is quite common) few things can be assumed. 1) to perform ORDER BY we don't need any join (entry_id is in our entries table). 2) entries.entry_id references PRIMARY KEY, which is unique, so we will have not less, not more but exactly one row per join (one row from messages per one row from entries) 3) Knowing above, instead of performing join on each of thousands of entries rows, we could perform ORDER BY and LIMIT before JOINing. 4) And then, after LIMITing we could JOIN those 5 rows. This I guess would be quite benefitial for VIEWs. :) Other thing that would be, I guess, benefitial for views would be special handling of lines like this: SELECT entry_id,message_id FROM entries NATURAL JOIN messages; Here there is no reason to perform JOIN at all -- the data will not be used. As above, since entries.message_id IS NOT NULL REFERENCES messages and messages is UNIQUE (PRIMARY KEY) we are sure there will be one-to-one(*) mapping between two tables. And since these keys are not used, no need to waste time and perform JOIN. I wonder what you all think about it. :) Regards, Dawid (*) not exactly one-to-one, because same messages.message_id can be references many times from entries.message_id, but the join will return exactly the same number of lines as would select * from entries; From pgsql-performance-owner@postgresql.org Mon Oct 11 13:16:36 2004 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 7624932A4C8 for ; Mon, 11 Oct 2004 13:16:30 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 89712-05 for ; Mon, 11 Oct 2004 12:16:20 +0000 (GMT) Received: from pw6tiger.de (pw6tiger.de [217.160.140.116]) by svr1.postgresql.org (Postfix) with SMTP id 5550632A3B8 for ; Mon, 11 Oct 2004 13:16:22 +0100 (BST) Received: (qmail 30577 invoked by uid 505); 11 Oct 2004 12:16:22 -0000 Received: from vygen@gmx.de by planwerk6 by uid 89 with qmail-scanner-1.14 (virus scan: Clear:. Processed in 0.139657 secs); 11 Oct 2004 12:16:22 -0000 Received: from isi-dial-146-203.isionline-dialin.de (HELO ?192.168.1.10?) (vygen@planwerk6.de@195.158.146.203) by pw6tiger.de with SMTP; 11 Oct 2004 12:16:22 -0000 From: Janning Vygen To: "Chris Hutchinson" Subject: Re: EXPLAIN ANALYZE much slower than running query normally Date: Mon, 11 Oct 2004 14:15:56 +0200 User-Agent: KMail/1.7 Cc: pgsql-performance@postgresql.org References: In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200410111415.56547.vygen@gmx.de> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/133 X-Sequence-Number: 8609 Am Dienstag, 5. Oktober 2004 08:49 schrieb Chris Hutchinson: > Running a trivial query in v7.4.2 (installed with fedora core2) using > EXPLAIN ANALYZE is taking considerably longer than just running the query > (2mins vs 6 secs). I was using this query to quickly compare a couple of > systems after installing a faster disk. > > Is this sort of slowdown to be expected? no. did you run VACCUM ANALYZE before? you should do it after pg_restore your db to a new filesystem in which order did you ran the queries. If you start your server and run two equal queries, the second one will be much faster because of some or even all data needed to answer the query is still in the shared buffers. janning From pgsql-performance-owner@postgresql.org Mon Oct 11 13:25:32 2004 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 92A8832A9C1 for ; Mon, 11 Oct 2004 13:25:30 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 90995-07 for ; Mon, 11 Oct 2004 12:25:22 +0000 (GMT) Received: from pw6tiger.de (pw6tiger.de [217.160.140.116]) by svr1.postgresql.org (Postfix) with SMTP id 2533232A937 for ; Mon, 11 Oct 2004 13:25:24 +0100 (BST) Received: (qmail 31354 invoked by uid 505); 11 Oct 2004 12:25:25 -0000 Received: from vygen@gmx.de by planwerk6 by uid 89 with qmail-scanner-1.14 (virus scan: Clear:. Processed in 0.170588 secs); 11 Oct 2004 12:25:25 -0000 Received: from isi-dial-146-203.isionline-dialin.de (HELO ?192.168.1.10?) (vygen@planwerk6.de@195.158.146.203) by pw6tiger.de with SMTP; 11 Oct 2004 12:25:24 -0000 From: Janning Vygen To: HyunSung Jang Subject: Re: why my query is not using index?? Date: Mon, 11 Oct 2004 14:25:02 +0200 User-Agent: KMail/1.7 References: <41639F3C.4050103@siche.net> In-Reply-To: <41639F3C.4050103@siche.net> Cc: pgsql-performance@postgresql.org MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200410111425.02671.vygen@gmx.de> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/134 X-Sequence-Number: 8610 Am Mittwoch, 6. Oktober 2004 09:31 schrieben Sie: > postgres=# explain ANALYZE select * from test where today < '2004-01-01'; > QUERY PLAN >------------------------- Seq Scan on test (cost=0.00..19.51 rows=334 > width=44) (actual > time=0.545..2.429 rows=721 loops=1) > Filter: (today < '2004-01-01 00:00:00'::timestamp without time zone) > Total runtime: 3.072 ms > (3 rows) > > postgres=# explain ANALYZE select * from test where today > '2003-01-01' > and today < '2004-01-01'; > QUERY > PLAN > --------------------------------------------------------------- Index > Scan using idx_today on test (cost=0.00..18.89 rows=6 width=44) (actual > time=0.055..1.098 rows=365 loops=1) > Index Cond: ((today > '2003-01-01 00:00:00'::timestamp without time > zone) AND (today < '2004-01-01 00:00:00'::timestamp without time zone)) > Total runtime: 1.471 ms > (3 rows) > > hello > > I was expected 1st query should using index, but it doesn't > 2nd query doing perfect as you see. postgres uses a seq scan if its faster. In your case postgres seems to know that most of your rows have a date < 2004-01-01 and so doesn't need to consult the index if it has to read every page anyway. seq scan can be faster on small tables. try (in psql) "SET enable_seqscan TO off;" before running your query and see how postgres plans it without using seq scan. janning From pgsql-performance-owner@postgresql.org Mon Oct 11 15:14:37 2004 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 72E5932A518 for ; Mon, 11 Oct 2004 15:14:31 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 31047-03 for ; Mon, 11 Oct 2004 14:14:22 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 7B81C32AEB4 for ; Mon, 11 Oct 2004 15:14:25 +0100 (BST) 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 i9BEEIxB004866; Mon, 11 Oct 2004 10:14:19 -0400 (EDT) To: Dawid Kuroczko Cc: pgsql-performance@postgresql.org Subject: Re: Views, joins and LIMIT In-reply-to: <758d5e7f04101102543eab8dca@mail.gmail.com> References: <758d5e7f04101102543eab8dca@mail.gmail.com> Comments: In-reply-to Dawid Kuroczko message dated "Mon, 11 Oct 2004 11:54:41 +0200" Date: Mon, 11 Oct 2004 10:14:18 -0400 Message-ID: <4865.1097504058@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.1 tagged_above=0.0 required=5.0 tests=MAILTO_TO_SPAM_ADDR X-Spam-Level: * X-Archive-Number: 200410/135 X-Sequence-Number: 8611 Dawid Kuroczko writes: > This I guess would be quite benefitial for VIEWs. :) Have you tried it? regression-# SELECT entry_id,message FROM entries NATURAL JOIN messages ORDER BY entry_id DESC LIMIT 10; QUERY PLAN ----------------------------------------------------------------------------------------------------- Limit (cost=0.00..48.88 rows=10 width=36) -> Nested Loop (cost=0.00..4887.52 rows=1000 width=36) -> Index Scan Backward using entries_pkey on entries (cost=0.00..52.00 rows=1000 width=8) -> Index Scan using messages_pkey on messages (cost=0.00..4.82 rows=1 width=36) Index Cond: ("outer".message_id = messages.message_id) (5 rows) > Other thing that would be, I guess, benefitial for views would be > special handling of lines like this: > SELECT entry_id,message_id FROM entries NATURAL JOIN messages; > Here there is no reason to perform JOIN at all -- the data will not be used. > As above, since entries.message_id IS NOT NULL REFERENCES messages > and messages is UNIQUE (PRIMARY KEY) we are sure there will be one-to-one(*) > mapping between two tables. And since these keys are not used, no need to > waste time and perform JOIN. The bang-for-the-buck ratio on that seems much too low. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Oct 11 15:22:12 2004 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 0E4A232C12A for ; Mon, 11 Oct 2004 15:22:10 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 32319-09 for ; Mon, 11 Oct 2004 14:22:01 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id CC9B232C0E7 for ; Mon, 11 Oct 2004 15:22:04 +0100 (BST) 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 i9BEM3YO004973; Mon, 11 Oct 2004 10:22:04 -0400 (EDT) To: Mike Harding Cc: pgsql-performance@postgresql.org Subject: Re: COPY slows down? In-reply-to: <20041008121029.EEB485486C@bsd.mvh> References: <20041008121029.EEB485486C@bsd.mvh> Comments: In-reply-to Mike Harding message dated "Fri, 08 Oct 2004 05:10:29 -0700" Date: Mon, 11 Oct 2004 10:22:03 -0400 Message-ID: <4972.1097504523@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/136 X-Sequence-Number: 8612 Mike Harding writes: > I just ran a COPY of a million records several times, and each time I > ran it it ran apparently exponentially slower. Tell us about indexes, foreign keys involving this table, triggers, rules? Some mention of your PG version would be appropriate, too. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Oct 11 15:26:16 2004 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 741A532A7C4 for ; Mon, 11 Oct 2004 15:26:14 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 36332-04 for ; Mon, 11 Oct 2004 14:26:05 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id BC95232A44B for ; Mon, 11 Oct 2004 15:26:08 +0100 (BST) 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 i9BEQ8f7005033; Mon, 11 Oct 2004 10:26:09 -0400 (EDT) To: HyunSung Jang Cc: pgsql-performance@postgresql.org Subject: Re: why my query is not using index?? In-reply-to: <41639F3C.4050103@siche.net> References: <41639F3C.4050103@siche.net> Comments: In-reply-to HyunSung Jang message dated "Wed, 06 Oct 2004 16:31:08 +0900" Date: Mon, 11 Oct 2004 10:26:08 -0400 Message-ID: <5032.1097504768@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/137 X-Sequence-Number: 8613 HyunSung Jang writes: > can you explain to me why it's not doing that i expected?? Have you ANALYZEd this table recently? The estimated row counts seem way off. regards, tom lane From pgsql-benchmarks-owner@postgresql.org Mon Oct 11 16:39:14 2004 X-Original-To: pgsql-benchmarks-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5A2BB32B0E9; Mon, 11 Oct 2004 16:39:12 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 65553-01; Mon, 11 Oct 2004 15:39:06 +0000 (GMT) Received: from mailer.unicite.fr.netcentrex.net (unknown [62.161.167.249]) by svr1.postgresql.org (Postfix) with ESMTP id 324A132AA59; Mon, 11 Oct 2004 16:39:04 +0100 (BST) Received: from [192.168.101.228] ([192.168.101.228]) by mailer.unicite.fr.netcentrex.net with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2657.72) id 42CX0WPL; Mon, 11 Oct 2004 17:39:02 +0200 Message-ID: <416AA915.1000008@fr.netcentrex.net> Date: Mon, 11 Oct 2004 17:39:01 +0200 From: =?ISO-8859-1?Q?=22Alban_M=E9dici_=28NetCentrex=29=22?= User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-benchmarks@postgresql.org Cc: pgsql-performance@postgresql.org Subject: Re: [PERFORM] stats on cursor and query execution References: <4163C5C3.3030304@fr.netcentrex.net> <25576.1097072199@sss.pgh.pa.us> <4166542E.2020809@fr.netcentrex.net> <11894.1097243756@sss.pgh.pa.us> In-Reply-To: <11894.1097243756@sss.pgh.pa.us> Content-Type: multipart/alternative; boundary="------------060000070001090402020807" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.5 tagged_above=0.0 required=5.0 tests=HTML_20_30, HTML_MESSAGE X-Spam-Level: X-Archive-Number: 200410/5 X-Sequence-Number: 22 This is a multi-part message in MIME format. --------------060000070001090402020807 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: quoted-printable Ok thanks tom, what shall we do without U ? by the way I have look at my kernel and getrusage() is well configure=20 and return good results. i/o stats too. I test an other version of postgresql and now, it works fine. It' seems to be an install bug. thx regards, Alban M=E9dici on 08/10/2004 15:55 Tom Lane said the following: >=3D?ISO-8859-1?Q?=3D22Alban_M=3DE9dici_=3D28NetCentrex=3D29=3D22?=3D writes: >=20=20 > >>Thanks for your repply, but I still don"t understand why the statistic >>logs : >>! 0/0 [0/0] filesystem blocks in/out >>it told me there is no hard disk access, I'm sure there is, >>=20=20=20=20 >> > >Complain to your friendly local kernel hacker. We just report what >getrusage() tells us; so if the number is wrong then it's a kernel bug. > > regards, tom lane > >=20=20 > --=20 Alban M=E9dici R&D software engineer ------------------------------ you can contact me @ : IPPhone : +33 (0)2 31 46 37 68 alban.medici@fr.netcentrex.net http://www.netcentrex.net ------------------------------ --------------060000070001090402020807 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Ok thanks tom, what shall we do without U ?

by the way I have look at my kernel and getrusage() is well configure and return good results.
i/o stats too.

I test an other version of postgresql and now, it works fine.
It' seems to be an install bug.

thx
regards, Alban Médici


on 08/10/2004 15:55 Tom Lane said the following: 
=?ISO-8859-1?Q?=22Alban_M=E9dici_=28NetCentrex=29=22?= <amedici@fr.netcentrex.net> writes:
  
Thanks for your repply,  but I still don"t understand why the statistic
logs   :
!       0/0 [0/0] filesystem blocks in/out
it told me there is no hard disk access, I'm sure there is,
    

Complain to your friendly local kernel hacker.  We just report what
getrusage() tells us; so if the number is wrong then it's a kernel bug.

			regards, tom lane

  

-- 
Alban Médici
R&D software engineer
------------------------------
you can contact me @ :
IPPhone : +33 (0)2 31 46 37 68
alban.medici@fr.netcentrex.net
http://www.netcentrex.net
------------------------------
--------------060000070001090402020807-- From pgsql-performance-owner@postgresql.org Mon Oct 11 18:38:31 2004 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 57B1C32A2F0 for ; Mon, 11 Oct 2004 18:38:27 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 05897-07 for ; Mon, 11 Oct 2004 17:38:20 +0000 (GMT) Received: from smtp100.rog.mail.re2.yahoo.com (smtp100.rog.mail.re2.yahoo.com [206.190.36.78]) by svr1.postgresql.org (Postfix) with SMTP id 5F49D32A19B for ; Mon, 11 Oct 2004 18:38:20 +0100 (BST) Received: from unknown (HELO phlogiston.dydns.org) (a.sullivan@rogers.com@65.49.125.184 with login) by smtp100.rog.mail.re2.yahoo.com with SMTP; 11 Oct 2004 17:38:19 -0000 Received: by phlogiston.dydns.org (Postfix, from userid 1000) id 5FF244054; Mon, 11 Oct 2004 13:38:19 -0400 (EDT) Date: Mon, 11 Oct 2004 13:38:19 -0400 From: Andrew Sullivan To: pgsql-performance@postgresql.org Subject: IBM P-series machines (was: Excessive context switching on SMP Xeons) Message-ID: <20041011173819.GB22076@phlogiston.dyndns.org> Mail-Followup-To: pgsql-performance@postgresql.org References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <200410050947.36174.josh@agliodbs.com> User-Agent: Mutt/1.5.6+20040722i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/139 X-Sequence-Number: 8615 On Tue, Oct 05, 2004 at 09:47:36AM -0700, Josh Berkus wrote: > As long as you're on x86, scaling outward is the way to go. If you want to > continue to scale upwards, ask Andrew Sullivan about his experiences running > PostgreSQL on big IBM boxes. But if you consider an quad-Opteron server > expensive, I don't think that's an option for you. Well, they're not that big, and both Chris Browne and Andrew Hammond are at least as qualified to talk about this as I. But since Josh mentioned it, I'll put some anecdotal rablings here just in case anyone is interested. We used to run our systems on Solaris 7, then 8, on Sun E4500s. We found the performance on those boxes surprisingly bad under certain pathological loads. I ultimately traced this to some remarkably poor shared memory handling by Solaris: during relatively heavy load (in particular, thousands of selects per second on the same set of tuples) we'd see an incredible number of semaphore operations, and concluded that the buffer handling was killing us. I think we could have tuned this away, but for independent reasons we decided to dump Sun gear (the hardware had become unreliable, and we were not satisfied with the service we were getting). We ended up choosing IBM P650s as a replacement. The 650s are not cheap, but boy are they fast. I don't have any numbers I can share, but I can tell you that we recently had a few days in which our write load was as large as the entire write load for last year, and you couldn't tell. It is too early for us to say whether the P series lives up to its billing in terms of relibility: the real reason we use these machines is reliability, so if approaching 100% uptime isn't important to you, the speed may not be worth it. We're also, for the record, doing experiments with Opterons. So far, we're impressed, and you can buy a lot of Opteron for the cost of one P650. A -- Andrew Sullivan | ajs@crankycanuck.ca I remember when computers were frustrating because they *did* exactly what you told them to. That actually seems sort of quaint now. --J.D. Baldwin From pgsql-performance-owner@postgresql.org Tue Oct 12 01:28:33 2004 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 88AA532A6B1 for ; Tue, 12 Oct 2004 01:28:28 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 19369-10 for ; Tue, 12 Oct 2004 00:28:16 +0000 (GMT) Received: from tht.net (vista.tht.net [216.126.88.2]) by svr1.postgresql.org (Postfix) with ESMTP id 7C2B932A623 for ; Tue, 12 Oct 2004 01:28:18 +0100 (BST) Received: from [134.22.69.18] (dyn-69-18.tor.dsl.tht.net [134.22.69.18]) by tht.net (Postfix) with ESMTP id 3098677887; Mon, 11 Oct 2004 14:21:11 -0400 (EDT) Subject: Re: IBM P-series machines (was: Excessive context From: Rod Taylor To: Andrew Sullivan Cc: Postgresql Performance In-Reply-To: <20041011173819.GB22076@phlogiston.dyndns.org> References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <20041011173819.GB22076@phlogiston.dyndns.org> Content-Type: text/plain Message-Id: <1097518852.54644.20.camel@home> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Mon, 11 Oct 2004 14:20:52 -0400 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/148 X-Sequence-Number: 8624 On Mon, 2004-10-11 at 13:38, Andrew Sullivan wrote: > On Tue, Oct 05, 2004 at 09:47:36AM -0700, Josh Berkus wrote: > > As long as you're on x86, scaling outward is the way to go. If you want to > > continue to scale upwards, ask Andrew Sullivan about his experiences running > > PostgreSQL on big IBM boxes. But if you consider an quad-Opteron server > > expensive, I don't think that's an option for you. > The 650s are not cheap, but boy are they fast. I don't have any > numbers I can share, but I can tell you that we recently had a few > days in which our write load was as large as the entire write load > for last year, and you couldn't tell. It is too early for us to say > whether the P series lives up to its billing in terms of relibility: > the real reason we use these machines is reliability, so if > approaching 100% uptime isn't important to you, the speed may not be > worth it. Agreed completely, and the 570 knocks the 650 out of the water -- nearly double the performance for math heavy queries. Beware vendor support for Linux on these things though -- we ran into many of the same issues with vendor support on the IBM machines as we did with the Opterons. From pgsql-performance-owner@postgresql.org Mon Oct 11 21:49:27 2004 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 4F22632A14D for ; Mon, 11 Oct 2004 21:49:21 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 63552-07 for ; Mon, 11 Oct 2004 20:49:18 +0000 (GMT) Received: from mail1.acecape.com (mail1.acecape.com [66.114.74.12]) by svr1.postgresql.org (Postfix) with ESMTP id 9B7D232A10C for ; Mon, 11 Oct 2004 21:49:18 +0100 (BST) Received: from p65-147.acedsl.com (p65-147.acedsl.com [66.114.65.147]) by mail1.acecape.com (8.12.11/8.12.11) with ESMTP id i9BKn56d018997; Mon, 11 Oct 2004 16:49:05 -0400 Date: Mon, 11 Oct 2004 16:49:44 -0400 (EDT) From: Francisco Reyes X-X-Sender: fran@zoraida.natserv.net To: Janning Vygen Cc: HyunSung Jang , pgsql-performance@postgresql.org Subject: Re: why my query is not using index?? In-Reply-To: <200410111425.02671.vygen@gmx.de> Message-ID: <20041011163756.M97379@zoraida.natserv.net> References: <41639F3C.4050103@siche.net> <200410111425.02671.vygen@gmx.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/140 X-Sequence-Number: 8616 On Mon, 11 Oct 2004, Janning Vygen wrote: > postgres uses a seq scan if its faster. In your case postgres seems to know > that most of your rows have a date < 2004-01-01 and so doesn't need to > consult the index if it has to read every page anyway. seq scan can be faster > on small tables. try (in psql) "SET enable_seqscan TO off;" before running > your query and see how postgres plans it without using seq scan. I was about to post and saw this message. I have a query that was using sequential scans. Upon turning seqscan to off it changed to using the index. What does that mean? The tables are under 5k records so I wonder if that is why the optimizer is option, on it's default state, to do sequential scans. I was also wondering if there is a relation between the sequential scans and the fact that my entire query is a series of left joins: (1)FROM Accounts (2)LEFT JOIN Equity_Positions ON Accounts.Account_ID = (3)Equity_Positions.Account_ID (4)LEFT JOIN Equities USING( Equity_ID ) (5)LEFT JOIN Benchmarks USING( Benchmark_ID ) (6)LEFT JOIN Equity_Prices ON Equities.equity_id = Equity_Prices.equity_id (7) AND Equity_Positions.Equity_Date = Equity_Prices.Date (8)LEFT JOIN Benchmark_Positions ON Equities.Benchmark_ID = (9)Benchmark_Positions.Benchmark_ID (10) AND Equity_Positions.Equity_Date = (11)Benchmark_Positions.Benchmark_Date (12)WHERE Client_ID =32 When I saw the default explain I was surprised to see that indexes were not been used. For example the join on lines 4,5 are exactly the primary key of the tables yet a sequential scan was used. The default explain was: Sort (cost=382.01..382.15 rows=56 width=196) Sort Key: accounts.account_group, accounts.account_name, equities.equity_description, equity_positions.equity_date -> Hash Left Join (cost=357.36..380.39 rows=56 width=196) Hash Cond: (("outer".benchmark_id = "inner".benchmark_id) AND ("outer".equity_date = "inner".benchmark_date)) -> Hash Left Join (cost=353.41..375.46 rows=56 width=174) Hash Cond: (("outer".equity_id = "inner".equity_id) AND ("outer".equity_date = "inner".date)) -> Hash Left Join (cost=292.22..296.90 rows=56 width=159) Hash Cond: ("outer".benchmark_id = "inner".benchmark_id) -> Merge Right Join (cost=290.40..294.51 rows=56 width=137) Merge Cond: ("outer".equity_id = "inner".equity_id) -> Sort (cost=47.19..48.83 rows=655 width=70) Sort Key: equities.equity_id -> Seq Scan on equities (cost=0.00..16.55 rows=655 width=70) -> Sort (cost=243.21..243.35 rows=56 width=67) Sort Key: equity_positions.equity_id -> Nested Loop Left Join (cost=0.00..241.58 rows=56 width=67) -> Seq Scan on accounts (cost=0.00..5.80 rows=3 width=44) Filter: (client_id = 32) -> Index Scan using positions_acct_equity_date on equity_positions (cost=0.00..78.30 rows=23 width=27) Index Cond: ("outer".account_id = equity_positions.account_id) -> Hash (cost=1.66..1.66 rows=66 width=22) -> Seq Scan on benchmarks (cost=0.00..1.66 rows=66 width=22) -> Hash (cost=50.79..50.79 rows=2079 width=23) -> Seq Scan on equity_prices (cost=0.00..50.79 rows=2079 width=23) -> Hash (cost=3.30..3.30 rows=130 width=30) -> Seq Scan on benchmark_positions (cost=0.00..3.30 rows=130 width=30) After set enable_seqscan to off; It becomes Sort (cost=490.82..490.96 rows=56 width=196) Sort Key: accounts.account_group, accounts.account_name, equities.equity_description, equity_positions.equity_date -> Merge Left Join (cost=309.75..489.20 rows=56 width=196) Merge Cond: ("outer".benchmark_id = "inner".benchmark_id) Join Filter: ("outer".equity_date = "inner".benchmark_date) -> Nested Loop Left Join (cost=309.75..644.88 rows=56 width=174) -> Merge Left Join (cost=309.75..315.90 rows=56 width=159) Merge Cond: ("outer".benchmark_id = "inner".benchmark_id) -> Sort (cost=309.75..309.89 rows=56 width=137) Sort Key: equities.benchmark_id -> Merge Right Join (cost=254.43..308.12 rows=56 width=137) Merge Cond: ("outer".equity_id = "inner".equity_id) -> Index Scan using equities_pkey on equities (cost=0.00..51.21 rows=655 width=70) -> Sort (cost=254.43..254.57 rows=56 width=67) Sort Key: equity_positions.equity_id -> Nested Loop Left Join (cost=0.00..252.81 rows=56 width=67) -> Index Scan using accounts_pkey on accounts (cost=0.00..17.02 rows=3 width=44) Filter: (client_id = 32) -> Index Scan using positions_acct_equity_date on equity_positions (cost=0.00..78.30 rows=23 width=27) Index Cond: ("outer".account_id = equity_positions.account_id) -> Index Scan using benchmarks_pkey on benchmarks (cost=0.00..5.57 rows=66 width=22) -> Index Scan using equity_prices_equity_date on equity_prices (cost=0.00..5.86 rows=1 width=23) Index Cond: (("outer".equity_id = equity_prices.equity_id) AND ("outer".equity_date = equity_prices.date)) -> Index Scan using benchpositions_acct_equity_date on benchmark_positions (cost=0.00..10.82 rows=130 width=30) From pgsql-performance-owner@postgresql.org Mon Oct 11 22:03:47 2004 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 058DD329E7B for ; Mon, 11 Oct 2004 22:03:46 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 68539-07 for ; Mon, 11 Oct 2004 21:03:35 +0000 (GMT) Received: from mail1.acecape.com (mail1.acecape.com [66.114.74.12]) by svr1.postgresql.org (Postfix) with ESMTP id 1BDC5329D5F for ; Mon, 11 Oct 2004 22:03:36 +0100 (BST) Received: from p65-147.acedsl.com (p65-147.acedsl.com [66.114.65.147]) by mail1.acecape.com (8.12.11/8.12.11) with ESMTP id i9BL3bLP025166 for ; Mon, 11 Oct 2004 17:03:37 -0400 Date: Mon, 11 Oct 2004 17:04:16 -0400 (EDT) From: Francisco Reyes X-X-Sender: fran@zoraida.natserv.net To: PostgreSQL performance Subject: Understanding explains Message-ID: <20041011164950.J97379@zoraida.natserv.net> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/141 X-Sequence-Number: 8617 Is there a tutorial or reference to the different terms that appear on the explain output? Items such as "Nested Loop", "Hash".. Also is there a way to easily tell which of two explains is "worse". Example I am running a query with "set enable_seqscan to off;" and i see the explain now shows index scans, but not sure if is any faster now. I tried "explain analyze" and the "total runtime" for the one with seq_scan off was faster, but after repeathing them they both dropped in time, likely due to data getting cached. Even after the time drops for both the one with seqscan off was always faster. Is there any disadvantage of having the enable_seqscan off? From pgsql-performance-owner@postgresql.org Mon Oct 11 22:06:10 2004 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 0977932A35C for ; Mon, 11 Oct 2004 22:06:05 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 71639-01 for ; Mon, 11 Oct 2004 21:05:57 +0000 (GMT) Received: from vsmtp3.tin.it (vsmtp3alice.tin.it [212.216.176.143]) by svr1.postgresql.org (Postfix) with ESMTP id 16FD332A1C4 for ; Mon, 11 Oct 2004 22:05:57 +0100 (BST) Received: from angus.tin.it (82.51.56.159) by vsmtp3.tin.it (7.0.027) id 414B175C00A42ED0 for pgsql-performance@postgresql.org; Mon, 11 Oct 2004 23:05:57 +0200 Message-Id: <6.1.2.0.2.20041011224414.01fadec0@box.tin.it> X-Sender: angusgb@box.tin.it (Unverified) X-Mailer: QUALCOMM Windows Eudora Version 6.1.2.0 Date: Mon, 11 Oct 2004 23:05:59 +0200 To: pgsql-performance@postgresql.org From: Gabriele Bartolini Subject: Normal case or bad query plan? Mime-Version: 1.0 Content-Type: multipart/mixed; x-avg-checked=avg-ok-55D43834; boundary="=======6D9D418A=======" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.1 tagged_above=0.0 required=5.0 tests=RCVD_IN_RFCI X-Spam-Level: X-Archive-Number: 200410/142 X-Sequence-Number: 8618 --=======6D9D418A======= Content-Type: text/plain; x-avg-checked=avg-ok-55D43834; charset=us-ascii; format=flowed Content-Transfer-Encoding: 8bit Hi guys, please consider this scenario. I have this table: CREATE TABLE ip2location ( ip_address_from BIGINT NOT NULL, ip_address_to BIGINT NOT NULL, id_location BIGINT NOT NULL, PRIMARY KEY (ip_address_from, ip_address_to) ); I created a cluster on its primary key, by running: CLUSTER ip2location_ip_address_from_key ON ip2location; This allowed me to organise data in a more efficient way: the data that is contained are ranges of IP addresses with empty intersections; for every IP class there is a related location's ID. The total number of entries is 1392443. For every IP address I have, an application retrieves the corresponding location's id from the above table, by running a query like: SELECT id_location FROM ip2location WHERE '11020000111' >= ip_address_from AND '11020000111' <= ip_address_to; For instance, by running the 'EXPLAIN ANALYSE' command, I get this "funny" result: QUERY PLAN --------------------------------------------------------------------------------------------------------------------- Seq Scan on ip2location (cost=0.00..30490.65 rows=124781 width=8) (actual time=5338.120..40237.283 rows=1 loops=1) Filter: ((1040878301::bigint >= ip_address_from) AND (1040878301::bigint <= ip_address_to)) Total runtime: 40237.424 ms With other data, that returns an empty set, I get: explain SELECT id_location FROM ip2location WHERE '11020000111' >= ip_address_from AND '11020000111' <= ip_address_to; QUERY PLAN ------------------------------------------------------------------------------------------------------- Index Scan using ip2location_ip_address_from_key on ip2location (cost=0.00..419.16 rows=140 width=8) Index Cond: ((11020000111::bigint >= ip_address_from) AND (11020000111::bigint <= ip_address_to)) I guess the planner chooses the best of the available options for the first case, the sequential scan. This is not confirmed though by the fact that, after I ran "SET enable_scan TO off", I got this: QUERY PLAN ------------------------------------------------------------------------------------------------------------------------------------------------------------ Index Scan using ip2location_ip_address_from_key on ip2location (cost=0.00..31505.73 rows=124781 width=8) (actual time=2780.172..2780.185 rows=1 loops=1) Index Cond: ((1040878301::bigint >= ip_address_from) AND (1040878301::bigint <= ip_address_to)) Total runtime: 2780.359 ms Is this a normal case or should I worry? What am I missing? Do you have any suggestion or comment to do (that would be extremely appreciated)? Is the CLUSTER I created worthwhile or not? Thank you, -Gabriele -- Gabriele Bartolini: Web Programmer, ht://Dig & IWA/HWG Member, ht://Check maintainer Current Location: Prato, Toscana, Italia angusgb@tin.it | http://www.prato.linux.it/~gbartolini | ICQ#129221447 > "Leave every hope, ye who enter!", Dante Alighieri, Divine Comedy, The Inferno --=======6D9D418A======= Content-Type: text/plain; charset=us-ascii; x-avg=cert; x-avg-checked=avg-ok-55D43834 Content-Disposition: inline --- Outgoing mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com). Version: 6.0.773 / Virus Database: 520 - Release Date: 05/10/2004 --=======6D9D418A=======-- From pgsql-performance-owner@postgresql.org Tue Oct 12 05:09:09 2004 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 905E232A572 for ; Tue, 12 Oct 2004 05:09:04 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 72106-07 for ; Tue, 12 Oct 2004 04:08:55 +0000 (GMT) Received: from mail63.csoft.net (leary3.csoft.net [63.111.22.74]) by svr1.postgresql.org (Postfix) with SMTP id 747B2329ECA for ; Tue, 12 Oct 2004 05:08:53 +0100 (BST) Received: (qmail 19966 invoked by uid 1112); 11 Oct 2004 21:17:24 -0000 Date: Mon, 11 Oct 2004 16:17:24 -0500 (EST) From: Kris Jurka X-X-Sender: books@leary.csoft.net To: Gabriele Bartolini Cc: pgsql-performance@postgresql.org Subject: Re: Normal case or bad query plan? In-Reply-To: <6.1.2.0.2.20041011224414.01fadec0@box.tin.it> Message-ID: References: <6.1.2.0.2.20041011224414.01fadec0@box.tin.it> 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=0.6 tagged_above=0.0 required=5.0 tests=DATE_IN_PAST_06_12 X-Spam-Level: X-Archive-Number: 200410/152 X-Sequence-Number: 8628 On Mon, 11 Oct 2004, Gabriele Bartolini wrote: > --------------------------------------------------------------------------------------------------------------------- > Seq Scan on ip2location (cost=0.00..30490.65 rows=124781 width=8) > (actual time=5338.120..40237.283 rows=1 loops=1) > Filter: ((1040878301::bigint >= ip_address_from) AND > (1040878301::bigint <= ip_address_to)) > Total runtime: 40237.424 ms > I believe the problem is that pg's lack of cross-column statistics is producing the poor number of rows estimate. The number of rows mataching just the first 1040878301::bigint >= ip_address_from condition is 122774 which is roughtly 10% of the table. I imagine the query planner believes that the other condition alone will match the other 90% of the table. The problem is that it doesn't know that these two ranges' intersection is actually tiny. The planner assumes a complete or nearly complete overlap so it thinks it will need to fetch 10% of the rows from both the index and the heap and chooses a seqscan. Kris Jurka From pgsql-performance-owner@postgresql.org Mon Oct 11 22:26:18 2004 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 3289632A1CB for ; Mon, 11 Oct 2004 22:26:17 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 74558-06 for ; Mon, 11 Oct 2004 21:26:06 +0000 (GMT) Received: from pw6tiger.de (pw6tiger.de [217.160.140.116]) by svr1.postgresql.org (Postfix) with SMTP id BC05F32A1A4 for ; Mon, 11 Oct 2004 22:26:06 +0100 (BST) Received: (qmail 9227 invoked by uid 505); 11 Oct 2004 21:26:05 -0000 Received: from vygen@gmx.de by planwerk6 by uid 89 with qmail-scanner-1.14 (virus scan: Clear:. Processed in 0.263111 secs); 11 Oct 2004 21:26:05 -0000 Received: from p508cc1e2.dip0.t-ipconnect.de (HELO ?192.168.1.102?) (vygen@planwerk6.de@80.140.193.226) by pw6tiger.de with SMTP; 11 Oct 2004 21:26:05 -0000 From: Janning Vygen To: Francisco Reyes Subject: Re: why my query is not using index?? Date: Mon, 11 Oct 2004 23:26:02 +0200 User-Agent: KMail/1.7 Cc: HyunSung Jang , pgsql-performance@postgresql.org References: <41639F3C.4050103@siche.net> <200410111425.02671.vygen@gmx.de> <20041011163756.M97379@zoraida.natserv.net> In-Reply-To: <20041011163756.M97379@zoraida.natserv.net> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200410112326.02486.vygen@gmx.de> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/143 X-Sequence-Number: 8619 Am Montag, 11. Oktober 2004 22:49 schrieb Francisco Reyes: > On Mon, 11 Oct 2004, Janning Vygen wrote: > > postgres uses a seq scan if its faster. In your case postgres seems to > > know that most of your rows have a date < 2004-01-01 and so doesn't need > > to consult the index if it has to read every page anyway. seq scan can be > > faster on small tables. try (in psql) "SET enable_seqscan TO off;" > > before running your query and see how postgres plans it without using seq > > scan. > > I was about to post and saw this message. > I have a query that was using sequential scans. Upon turning seqscan to > off it changed to using the index. What does that mean? enable_seqscan off means that postgres is not allowed to use seqscan. default is on and postgres decides for each table lookup which method is faster: seq scan or index scan. thats what the planner does: deciding which access method might be the fastest. > The tables are under 5k records so I wonder if that is why the optimizer > is option, on it's default state, to do sequential scans. if you have small tables, postgres is using seqscan to reduce disk lookups. postgresql reads disk pages in 8k blocks. if your whole table is under 8k there is no reason for postgres to load an index from another disk page because it has to load the whole disk anyway. not sure, but i think postgres also analyzes the table to see which values are in there. if you have a huge table with a column of integers and postgres knows that 99% are of value 1 and you are looking for a row with a value of 1, why should it use an index just to see that it has to load the whole table to find a matching row. And that's why you can't make performance tests with small tables. you need test data which is as close as possible to real data. > I was also wondering if there is a relation between the sequential scans > and the fact that my entire query is a series of left joins: no. janning From pgsql-performance-owner@postgresql.org Mon Oct 11 22:34:10 2004 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 95118329F15 for ; Mon, 11 Oct 2004 22:34:07 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78526-01 for ; Mon, 11 Oct 2004 21:33:56 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 7CE31329F2B for ; Mon, 11 Oct 2004 22:33:57 +0100 (BST) 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 i9BLXv1q023516; Mon, 11 Oct 2004 17:33:58 -0400 (EDT) To: Gabriele Bartolini Cc: pgsql-performance@postgresql.org Subject: Re: Normal case or bad query plan? In-reply-to: <6.1.2.0.2.20041011224414.01fadec0@box.tin.it> References: <6.1.2.0.2.20041011224414.01fadec0@box.tin.it> Comments: In-reply-to Gabriele Bartolini message dated "Mon, 11 Oct 2004 23:05:59 +0200" Date: Mon, 11 Oct 2004 17:33:57 -0400 Message-ID: <23515.1097530437@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/144 X-Sequence-Number: 8620 Gabriele Bartolini writes: > QUERY PLAN > --------------------------------------------------------------------------------------------------------------------- > Seq Scan on ip2location (cost=0.00..30490.65 rows=124781 width=8) > (actual time=5338.120..40237.283 rows=1 loops=1) > Filter: ((1040878301::bigint >= ip_address_from) AND > (1040878301::bigint <= ip_address_to)) > Total runtime: 40237.424 ms > Is this a normal case or should I worry? What am I missing? The striking thing about that is the huge difference between estimated rowcount (124781) and actual (1). The planner would certainly have picked an indexscan if it thought the query would select only one row. I suspect that you haven't ANALYZEd this table in a long time, if ever. You really need reasonably up-to-date ANALYZE stats if you want the planner to do an adequate job of planning range queries. It may well be that you need to increase the analyze statistics target for this table, also --- in BIGINT terms the distribution is probably pretty irregular, which will mean you need finer-grain statistics to get good estimates. (BTW, have you looked at the inet datatype to see if that would fit your needs?) regards, tom lane From pgsql-performance-owner@postgresql.org Mon Oct 11 23:03:18 2004 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 4816C329C89 for ; Mon, 11 Oct 2004 23:03:17 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83145-04 for ; Mon, 11 Oct 2004 22:03:07 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.201]) by svr1.postgresql.org (Postfix) with ESMTP id B089732A091 for ; Mon, 11 Oct 2004 23:03:08 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 74so350081rnk for ; Mon, 11 Oct 2004 15:03:07 -0700 (PDT) Received: by 10.38.151.14 with SMTP id y14mr1517843rnd; Mon, 11 Oct 2004 15:03:07 -0700 (PDT) Received: by 10.38.76.43 with HTTP; Mon, 11 Oct 2004 15:03:07 -0700 (PDT) Message-ID: <37d451f70410111503755fb751@mail.gmail.com> Date: Mon, 11 Oct 2004 17:03:07 -0500 From: Rosser Schwarz Reply-To: Rosser Schwarz To: Francisco Reyes Subject: Re: Understanding explains Cc: PostgreSQL performance In-Reply-To: <20041011164950.J97379@zoraida.natserv.net> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: <20041011164950.J97379@zoraida.natserv.net> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/145 X-Sequence-Number: 8621 while you weren't looking, Francisco Reyes wrote: > Is there any disadvantage of having the enable_seqscan off? Plenty. The planner will choose whichever plan looks "cheapest", based on the information it has available (table size, statistics, &c). If a sequential scan looks cheaper, and in your case above it clearly is, the planner will choose that query plan. Setting enable_seqscan = false doesn't actually disable sequential scans; it merely makes them seem radically more expensive to the planner, in hopes of biasing its choice towards another query plan. In your case, that margin made an index scan look less expensive than sequential scan, but your query runtimes clearly suggest otherwise. In general, it's best to let the planner make the appropriate choice without any artificial constraints. I've seen pathalogical cases where the planner makes the wrong choice(s), but upon analysis, they're almost always attributable to poor statistics, long un-vacuumed tables, &c. /rls -- :wq From pgsql-performance-owner@postgresql.org Tue Oct 12 00:50:46 2004 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 1B0A3329EAB for ; Tue, 12 Oct 2004 00:50:44 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 11711-02 for ; Mon, 11 Oct 2004 23:50:32 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 275A2329E65 for ; Tue, 12 Oct 2004 00:50:34 +0100 (BST) 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 6491475 for pgsql-performance@postgresql.org; Mon, 11 Oct 2004 16:51:50 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: PostgreSQL performance Subject: TestPerf Project started Date: Mon, 11 Oct 2004 16:52:14 -0700 User-Agent: KMail/1.6.2 MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Message-Id: <200410111652.14890.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/146 X-Sequence-Number: 8622 Folks, In order to have a place for scripts, graphs, results, etc., I've started the TestPerf project on pgFoundry: http://pgfoundry.org/projects/testperf/ If you are interested in doing performance testing for PostgreSQL, please join the mailing list for the project. I could certainly use some help scripting. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Tue Oct 12 01:21:24 2004 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 9426D32A684 for ; Tue, 12 Oct 2004 01:21:21 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 19534-01 for ; Tue, 12 Oct 2004 00:21:16 +0000 (GMT) Received: from server07.icaen.uiowa.edu (server07.icaen.uiowa.edu [128.255.17.47]) by svr1.postgresql.org (Postfix) with ESMTP id 20FD232A2C3 for ; Tue, 12 Oct 2004 01:21:17 +0100 (BST) Received: from server11.icaen.uiowa.edu (server11.icaen.uiowa.edu [128.255.17.51]) by server07.icaen.uiowa.edu (8.12.9/8.12.9) with ESMTP id i9C0LBSp009554; sent by ; Mon, 11 Oct 2004 19:21:15 -0500 (CDT) Received: from [192.168.1.11] (12-217-241-0.client.mchsi.com [12.217.241.0]) by server11.icaen.uiowa.edu (8.12.9/smtp-service-1.6) with ESMTP id i9C0Cj4p002504; (envelope-from ) Mon, 11 Oct 2004 19:12:45 -0500 (CDT) Message-ID: <416B217D.9090000@johnmeinel.com> Date: Mon, 11 Oct 2004 19:12:45 -0500 From: John Meinel User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Francisco Reyes Cc: pgsql-performance@postgresql.org Subject: Re: why my query is not using index?? References: <41639F3C.4050103@siche.net> <200410111425.02671.vygen@gmx.de> <20041011163756.M97379@zoraida.natserv.net> In-Reply-To: <20041011163756.M97379@zoraida.natserv.net> X-Enigmail-Version: 0.86.0.0 X-Enigmail-Supports: pgp-inline, pgp-mime Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig020F595836BEAD80916D9EEE" X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.66n X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.75 on clamav.icaen.uiowa.edu X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/147 X-Sequence-Number: 8623 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig020F595836BEAD80916D9EEE Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Francisco Reyes wrote: > On Mon, 11 Oct 2004, Janning Vygen wrote: > [...] > When I saw the default explain I was surprised to see that indexes were > not been used. For example the join on lines 4,5 are exactly the primary > key of the tables yet a sequential scan was used. > Note this: > The default explain was: > > Sort (cost=382.01..382.15 rows=56 width=196) > Sort Key: accounts.account_group, accounts.account_name, [...] Versus this: > > After set enable_seqscan to off; > It becomes > > Sort (cost=490.82..490.96 rows=56 width=196) > Sort Key: accounts.account_group, accounts.account_name, [...] Postgres believes that it will cost 382 to do a sequential scan, versus 490 for an indexed scan. Hence why it prefers to do the sequential scan. Try running explain analyze to see if how accurate it is. As Janning mentioned, sometimes sequential scans *are* faster. If the number of entries that will be found is large compared to the number of total entries (I don't know the percentages, but probably >30-40%), then it is faster to just load the data and scan through it, rather than doing a whole bunch of indexed lookups. John =:-> --------------enig020F595836BEAD80916D9EEE Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFBayF9JdeBCYSNAAMRAhXHAJ9Oe5UXhr7l4rMCNpQU88A1ePzZmACdH7td cedOaVSPnnV6RVTC4dpe8fo= =J1/L -----END PGP SIGNATURE----- --------------enig020F595836BEAD80916D9EEE-- From pgsql-performance-owner@postgresql.org Tue Oct 12 01:28:40 2004 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 24F26329E7B for ; Tue, 12 Oct 2004 01:28:34 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 20439-03 for ; Tue, 12 Oct 2004 00:28:29 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id CB908329E65 for ; Tue, 12 Oct 2004 01:28:31 +0100 (BST) 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 i9C0SSGY025941; Mon, 11 Oct 2004 20:28:28 -0400 (EDT) To: "Chris Hutchinson" Cc: pgsql-performance@postgresql.org Subject: Re: EXPLAIN ANALYZE much slower than running query normally In-reply-to: References: Comments: In-reply-to "Chris Hutchinson" message dated "Tue, 05 Oct 2004 16:49:26 +1000" Date: Mon, 11 Oct 2004 20:28:28 -0400 Message-ID: <25940.1097540908@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/149 X-Sequence-Number: 8625 "Chris Hutchinson" writes: > Running a trivial query in v7.4.2 (installed with fedora core2) using > EXPLAIN ANALYZE is taking considerably longer than just running the query > (2mins vs 6 secs). I was using this query to quickly compare a couple of > systems after installing a faster disk. Turning on EXPLAIN ANALYZE will incur two gettimeofday() kernel calls per row (in this particular plan), which is definitely nontrivial overhead if there's not much I/O going on. I couldn't duplicate your results exactly, but I did see a test case with 2.5 million one-column rows go from <4 seconds to 21 seconds, which makes the cost of a gettimeofday about 3.4 microseconds on my machine (Fedora Core 3, P4 running at something over 1Ghz). When I widened the rows to a couple hundred bytes, the raw runtime went up to 30 seconds and the analyzed time to 50, so the overhead per row is pretty constant, as you'd expect. Some tests with a simple loop around a gettimeofday call yielded a value of 2.16 microsec/gettimeofday, so there's some overhead attributable to the EXPLAIN mechanism as well, but the kernel call is clearly the bulk of it. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Oct 12 02:13:46 2004 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 6F0AB32A3C1 for ; Tue, 12 Oct 2004 02:13:42 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 33170-09 for ; Tue, 12 Oct 2004 01:13:37 +0000 (GMT) Received: from stark.xeocode.com (gsstark.mtl.istop.com [66.11.160.162]) by svr1.postgresql.org (Postfix) with ESMTP id E34AD32A2AE for ; Tue, 12 Oct 2004 02:13:38 +0100 (BST) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1CHBEF-0004T2-00; Mon, 11 Oct 2004 21:13:27 -0400 To: John Meinel Cc: Francisco Reyes , pgsql-performance@postgresql.org Subject: Re: why my query is not using index?? References: <41639F3C.4050103@siche.net> <200410111425.02671.vygen@gmx.de> <20041011163756.M97379@zoraida.natserv.net> <416B217D.9090000@johnmeinel.com> In-Reply-To: <416B217D.9090000@johnmeinel.com> From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 11 Oct 2004 21:13:27 -0400 Message-ID: <87fz4k3fig.fsf@stark.xeocode.com> Lines: 15 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/150 X-Sequence-Number: 8626 John Meinel writes: > As Janning mentioned, sometimes sequential scans *are* faster. If the number of > entries that will be found is large compared to the number of total entries (I > don't know the percentages, but probably >30-40%), Actually 30%-40% is unrealistic. The traditional rule of thumb for the break-even point was 10%. In POstgres the actual percentage varies based on how wide the records are and how correlated the location of the records is with the index. Usually it's between 5%-10% but it can be even lower than that sometimes. -- greg From pgsql-performance-owner@postgresql.org Tue Oct 12 02:48:52 2004 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 AFC29329EAB for ; Tue, 12 Oct 2004 02:48:49 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 39405-05 for ; Tue, 12 Oct 2004 01:48:37 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id D3905329E7E for ; Tue, 12 Oct 2004 02:48:40 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i9C1maJ7042300 for ; Tue, 12 Oct 2004 01:48:36 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i9C1ZD83039309 for pgsql-performance@postgresql.org; Tue, 12 Oct 2004 01:35:13 GMT From: Christopher Browne X-Newsgroups: comp.databases.postgresql.performance Subject: Re: IBM P-series machines (was: Excessive context Date: Mon, 11 Oct 2004 21:34:44 -0400 Organization: cbbrowne Computing Inc Lines: 55 Message-ID: References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <20041011173819.GB22076@phlogiston.dyndns.org> <1097518852.54644.20.camel@home> 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 (Security Through Obscurity, linux) Cancel-Lock: sha1:RQnEtCKDkofbB7SGiselPsJNRl8= To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/151 X-Sequence-Number: 8627 pg@rbt.ca (Rod Taylor) wrote: > On Mon, 2004-10-11 at 13:38, Andrew Sullivan wrote: >> On Tue, Oct 05, 2004 at 09:47:36AM -0700, Josh Berkus wrote: >> > As long as you're on x86, scaling outward is the way to go. If >> > you want to continue to scale upwards, ask Andrew Sullivan about >> > his experiences running PostgreSQL on big IBM boxes. But if you >> > consider an quad-Opteron server expensive, I don't think that's >> > an option for you. > >> The 650s are not cheap, but boy are they fast. I don't have any >> numbers I can share, but I can tell you that we recently had a few >> days in which our write load was as large as the entire write load >> for last year, and you couldn't tell. It is too early for us to >> say whether the P series lives up to its billing in terms of >> relibility: the real reason we use these machines is reliability, >> so if approaching 100% uptime isn't important to you, the speed may >> not be worth it. > > Agreed completely, and the 570 knocks the 650 out of the water -- > nearly double the performance for math heavy queries. Beware vendor > support for Linux on these things though -- we ran into many of the > same issues with vendor support on the IBM machines as we did with > the Opterons. The 650s are running AIX, not Linux. Based on the "Signal 11" issue, I'm not certain what would be the 'best' answer. It appears that the problem relates to proprietary bits of AIX libc. In theory, that would have been more easily resolvable with a source-available GLIBC. On the other hand, I'm not sure what happens to support for any of the interesting hardware extensions. I'm not sure, for instance, that we could run HACMP on Linux on this hardware. As for "vendor support" for Opteron, that sure looks like a trainwreck... If you're going through IBM, then they won't want to respond to any issues if you're not running a "bog-standard" RHAS/RHES release from Red Hat. And that, on Opteron, is preposterous, because there's plenty of the bits of Opteron support that only ever got put in Linux 2.6, whilst RHAT is still back in the 2.4 days. In a way, that's just as well, at this point. There's plenty of stuff surrounding this that is still pretty experimental; the longer RHAT waits to support 2.6, the greater the likelihood that Linux support for Opteron will have settled down to the point that the result will actually be supportable by RHAT, and by proxy, by IBM and others. There is some risk that if RHAT waits _too_ long for 2.6, people will have already jumped ship to SuSE. No benefits without risks... -- (reverse (concatenate 'string "gro.mca" "@" "enworbbc")) http://www.ntlug.org/~cbbrowne/rdbms.html If at first you don't succeed, then you didn't do it right! If at first you don't succeed, then skydiving definitely isn't for you. From pgsql-performance-owner@postgresql.org Tue Oct 12 05:55:49 2004 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 000D432A13F for ; Tue, 12 Oct 2004 05:55:46 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 84330-06 for ; Tue, 12 Oct 2004 04:55:37 +0000 (GMT) Received: from mail1.acecape.com (mail1.acecape.com [66.114.74.12]) by svr1.postgresql.org (Postfix) with ESMTP id 510B332C241 for ; Tue, 12 Oct 2004 05:55:36 +0100 (BST) Received: from p65-147.acedsl.com (p65-147.acedsl.com [66.114.65.147]) by mail1.acecape.com (8.12.11/8.12.11) with ESMTP id i9C4tX7c016065; Tue, 12 Oct 2004 00:55:33 -0400 Date: Tue, 12 Oct 2004 00:56:15 -0400 (EDT) From: Francisco Reyes X-X-Sender: fran@zoraida.natserv.net To: John Meinel Cc: pgsql-performance@postgresql.org Subject: Re: why my query is not using index?? In-Reply-To: <416B217D.9090000@johnmeinel.com> Message-ID: <20041012005231.U98867@zoraida.natserv.net> References: <41639F3C.4050103@siche.net> <200410111425.02671.vygen@gmx.de> <20041011163756.M97379@zoraida.natserv.net> <416B217D.9090000@johnmeinel.com> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/153 X-Sequence-Number: 8629 On Mon, 11 Oct 2004, John Meinel wrote: > Postgres believes that it will cost 382 to do a sequential scan, versus 490 > for an indexed scan. Hence why it prefers to do the sequential scan. Try > running explain analyze to see if how accurate it is. With explain analyze I have with sequential scan on Sort (cost=382.01..382.15 rows=56 width=196) (actual time=64.346..64.469 rows=24 loops=1) And with seqscan off Sort (cost=490.82..490.96 rows=56 width=196) (actual time=56.668..56.789 rows=24 loops=1) So I guess that for this particular query I am better off setting the seqscan off. From pgsql-performance-owner@postgresql.org Tue Oct 12 05:59:13 2004 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 09B8632C2A7 for ; Tue, 12 Oct 2004 05:59:12 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 87047-07 for ; Tue, 12 Oct 2004 04:59:02 +0000 (GMT) Received: from mail1.acecape.com (mail1.acecape.com [66.114.74.12]) by svr1.postgresql.org (Postfix) with ESMTP id D914532C2A6 for ; Tue, 12 Oct 2004 05:59:02 +0100 (BST) Received: from p65-147.acedsl.com (p65-147.acedsl.com [66.114.65.147]) by mail1.acecape.com (8.12.11/8.12.11) with ESMTP id i9C4wtJZ017053; Tue, 12 Oct 2004 00:58:55 -0400 Date: Tue, 12 Oct 2004 00:59:37 -0400 (EDT) From: Francisco Reyes X-X-Sender: fran@zoraida.natserv.net To: Rosser Schwarz Cc: PostgreSQL performance Subject: Re: Understanding explains In-Reply-To: <37d451f70410111503755fb751@mail.gmail.com> Message-ID: <20041012005749.G98867@zoraida.natserv.net> References: <20041011164950.J97379@zoraida.natserv.net> <37d451f70410111503755fb751@mail.gmail.com> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/154 X-Sequence-Number: 8630 On Mon, 11 Oct 2004, Rosser Schwarz wrote: > In general, it's best to let the planner make the appropriate choice > without any artificial constraints. As someone suggested ran with Explain analyze. With seqscan_off was better. Ran a vacuum analyze this afternoon so the stats were up to date. Although I will leave the setting as it's default for most of everything I do, it seems that for some reason in this case it mases sense to turn it off. From pgsql-performance-owner@postgresql.org Tue Oct 12 06:26:31 2004 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 BE35F329F0D for ; Tue, 12 Oct 2004 06:26:29 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 92567-06 for ; Tue, 12 Oct 2004 05:26:19 +0000 (GMT) Received: from vsmtp4.tin.it (unknown [212.216.176.150]) by svr1.postgresql.org (Postfix) with ESMTP id 7CF2D329F33 for ; Tue, 12 Oct 2004 06:26:17 +0100 (BST) Received: from angus.tin.it (82.53.64.64) by vsmtp4.tin.it (7.0.027) id 41568C9D006D2FD9; Tue, 12 Oct 2004 07:26:09 +0200 Message-Id: <6.1.2.0.2.20041012072112.01ed8810@box.tin.it> X-Sender: angusgb@box.tin.it (Unverified) X-Mailer: QUALCOMM Windows Eudora Version 6.1.2.0 Date: Tue, 12 Oct 2004 07:26:12 +0200 To: Tom Lane From: Gabriele Bartolini Subject: Re: Normal case or bad query plan? Cc: pgsql-performance@postgresql.org In-Reply-To: <23515.1097530437@sss.pgh.pa.us> References: <6.1.2.0.2.20041011224414.01fadec0@box.tin.it> <23515.1097530437@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: multipart/mixed; x-avg-checked=avg-ok-2C527FEE; boundary="=======636B5308=======" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.1 tagged_above=0.0 required=5.0 tests=RCVD_IN_RFCI X-Spam-Level: X-Archive-Number: 200410/155 X-Sequence-Number: 8631 --=======636B5308======= Content-Type: text/plain; x-avg-checked=avg-ok-2C527FEE; charset=us-ascii; format=flowed Content-Transfer-Encoding: 8bit Hi Tom, thanks for your interest. At 23.33 11/10/2004, Tom Lane wrote: >Gabriele Bartolini writes: > > QUERY PLAN > > > --------------------------------------------------------------------------------------------------------------------- > > Seq Scan on ip2location (cost=0.00..30490.65 rows=124781 width=8) > > (actual time=5338.120..40237.283 rows=1 loops=1) > > Filter: ((1040878301::bigint >= ip_address_from) AND > > (1040878301::bigint <= ip_address_to)) > > Total runtime: 40237.424 ms > > > Is this a normal case or should I worry? What am I missing? > >The striking thing about that is the huge difference between estimated >rowcount (124781) and actual (1). The planner would certainly have >picked an indexscan if it thought the query would select only one row. > >I suspect that you haven't ANALYZEd this table in a long time, if ever. >You really need reasonably up-to-date ANALYZE stats if you want the >planner to do an adequate job of planning range queries. That's the thing ... I had just peformed a VACUUM ANALYSE :-( > It may well be that you need to increase the analyze statistics target > for this table, >also --- in BIGINT terms the distribution is probably pretty irregular, >which will mean you need finer-grain statistics to get good estimates. You mean ... SET STATISTICS for the two columns, don't you? >(BTW, have you looked at the inet datatype to see if that would fit your >needs?) Yes, I know. In other cases I use it. But this is a type of data coming from an external source (www.ip2location.com) and I can't change it. Thank you so much. I will try to play with the grain of the statistics, otherwise - if worse comes to worst - I will simply disable the seq scan after connecting. -Gabriele -- Gabriele Bartolini: Web Programmer, ht://Dig & IWA/HWG Member, ht://Check maintainer Current Location: Prato, Toscana, Italia angusgb@tin.it | http://www.prato.linux.it/~gbartolini | ICQ#129221447 > "Leave every hope, ye who enter!", Dante Alighieri, Divine Comedy, The Inferno --=======636B5308======= Content-Type: text/plain; charset=us-ascii; x-avg=cert; x-avg-checked=avg-ok-2C527FEE Content-Disposition: inline --- Outgoing mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com). Version: 6.0.773 / Virus Database: 520 - Release Date: 05/10/2004 --=======636B5308=======-- From pgsql-performance-owner@postgresql.org Tue Oct 12 06:27:09 2004 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 9E03332AA03 for ; Tue, 12 Oct 2004 06:27:05 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 93247-03 for ; Tue, 12 Oct 2004 05:26:56 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 98625329E25 for ; Tue, 12 Oct 2004 06:26:56 +0100 (BST) 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 i9C5QrYI028284; Tue, 12 Oct 2004 01:26:53 -0400 (EDT) To: Francisco Reyes Cc: John Meinel , pgsql-performance@postgresql.org Subject: Re: why my query is not using index?? In-reply-to: <20041012005231.U98867@zoraida.natserv.net> References: <41639F3C.4050103@siche.net> <200410111425.02671.vygen@gmx.de> <20041011163756.M97379@zoraida.natserv.net> <416B217D.9090000@johnmeinel.com> <20041012005231.U98867@zoraida.natserv.net> Comments: In-reply-to Francisco Reyes message dated "Tue, 12 Oct 2004 00:56:15 -0400" Date: Tue, 12 Oct 2004 01:26:53 -0400 Message-ID: <28283.1097558813@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/156 X-Sequence-Number: 8632 Francisco Reyes writes: > With explain analyze I have with sequential scan on > Sort (cost=382.01..382.15 rows=56 width=196) > (actual time=64.346..64.469 rows=24 loops=1) > And with seqscan off > Sort (cost=490.82..490.96 rows=56 width=196) > (actual time=56.668..56.789 rows=24 loops=1) > So I guess that for this particular query I am better off setting the > seqscan off. For that kind of margin, you'd be a fool to do any such thing. You might want to look at making some adjustment to random_page_cost to bring the estimated costs in line with reality (though I'd counsel taking more than one example into account while you tweak it). But setting seqscan off as a production setting is just a recipe for shooting yourself in the foot. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Oct 12 06:46:40 2004 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 C5F3032BBEA for ; Tue, 12 Oct 2004 06:45:22 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 98035-06 for ; Tue, 12 Oct 2004 05:45:18 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id E090932BB20 for ; Tue, 12 Oct 2004 06:45:16 +0100 (BST) 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 i9C5jFsq028412; Tue, 12 Oct 2004 01:45:16 -0400 (EDT) To: Gabriele Bartolini Cc: pgsql-performance@postgresql.org Subject: Re: Normal case or bad query plan? In-reply-to: <6.1.2.0.2.20041012072112.01ed8810@box.tin.it> References: <6.1.2.0.2.20041011224414.01fadec0@box.tin.it> <23515.1097530437@sss.pgh.pa.us> <6.1.2.0.2.20041012072112.01ed8810@box.tin.it> Comments: In-reply-to Gabriele Bartolini message dated "Tue, 12 Oct 2004 07:26:12 +0200" Date: Tue, 12 Oct 2004 01:45:15 -0400 Message-ID: <28411.1097559915@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/157 X-Sequence-Number: 8633 Gabriele Bartolini writes: > Seq Scan on ip2location (cost=0.00..30490.65 rows=124781 width=8) > (actual time=5338.120..40237.283 rows=1 loops=1) > Filter: ((1040878301::bigint >= ip_address_from) AND > (1040878301::bigint <= ip_address_to)) > Total runtime: 40237.424 ms >> >> I suspect that you haven't ANALYZEd this table in a long time, if ever. >> You really need reasonably up-to-date ANALYZE stats if you want the >> planner to do an adequate job of planning range queries. > That's the thing ... I had just peformed a VACUUM ANALYSE :-( In that case I think Kris Jurka had it right: the problem is the planner doesn't know enough about the relationship of the ip_address_from and ip_address_to columns to realize that this is a very selective query. But actually, even *had* it realized that, it would have had little choice but to use a seqscan, because neither of the independent conditions is really very useful as an index condition by itself. Assuming that this problem is representative of your query load, you really need to recast the data representation to make it more readily searchable. I think you might be able to get somewhere by combining ip_address_from and ip_address_to into a single CIDR column and then using the network-overlap operator to probe for matches to your query address. (This assumes that the from/to pairs are actually meant to represent CIDR subnets; if not you need some other idea.) Another possibility is to convert to a geometric type and use an rtree index with an "overlaps" operator. I'm too tired to work out the details, but try searching for "decorrelation" in the list archives to see some related problems. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Oct 12 07:53:20 2004 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 3710532AC81 for ; Tue, 12 Oct 2004 07:53:15 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 41694-03 for ; Tue, 12 Oct 2004 06:53:04 +0000 (GMT) Received: from bayswater1.ymogen.net (host-154-240-27-217.pobox.net.uk [217.27.240.154]) by svr1.postgresql.org (Postfix) with ESMTP id C30D932A590 for ; Tue, 12 Oct 2004 07:53:02 +0100 (BST) Received: from [82.68.132.233] (82-68-132-233.dsl.in-addr.zen.co.uk [82.68.132.233]) by bayswater1.ymogen.net (Postfix) with ESMTP id 4E504A384A; Tue, 12 Oct 2004 07:53:01 +0100 (BST) Message-ID: <416B7F4C.5070600@ymogen.net> Date: Tue, 12 Oct 2004 07:53:00 +0100 From: Matt Clark User-Agent: Mozilla Thunderbird 0.7.3 (Windows/20040803) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Christopher Browne Cc: pgsql-performance@postgresql.org Subject: Re: IBM P-series machines References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <20041011173819.GB22076@phlogiston.dyndns.org> <1097518852.54644.20.camel@home> In-Reply-To: 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/158 X-Sequence-Number: 8634 >As for "vendor support" for Opteron, that sure looks like a >trainwreck... If you're going through IBM, then they won't want to >respond to any issues if you're not running a "bog-standard" RHAS/RHES >release from Red Hat. And that, on Opteron, is preposterous, because >there's plenty of the bits of Opteron support that only ever got put >in Linux 2.6, whilst RHAT is still back in the 2.4 days. > > > To be fair, they have backported a boatload of 2.6 features to their kernel: http://www.redhat.com/software/rhel/kernel26/ And that page certainly isn't an exhaustive list... M From pgsql-performance-owner@postgresql.org Tue Oct 12 12:43:58 2004 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 04BE8329E84 for ; Tue, 12 Oct 2004 12:43:54 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78033-05 for ; Tue, 12 Oct 2004 11:43:42 +0000 (GMT) Received: from web54110.mail.yahoo.com (web54110.mail.yahoo.com [206.190.37.245]) by svr1.postgresql.org (Postfix) with SMTP id A7457329DB0 for ; Tue, 12 Oct 2004 12:43:43 +0100 (BST) Message-ID: <20041012114343.69531.qmail@web54110.mail.yahoo.com> Received: from [129.94.6.30] by web54110.mail.yahoo.com via HTTP; Tue, 12 Oct 2004 04:43:43 PDT Date: Tue, 12 Oct 2004 04:43:43 -0700 (PDT) From: my ho Subject: Re: execute cursor fetch To: pgsql-performance@postgresql.org In-Reply-To: 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/159 X-Sequence-Number: 8635 Hi, If anyone can help pls, I have a question abt the execution of cursor create/fetch/move , in particular about disk cost. When a cursor is created, is the whole table (with the required columns) got put into memory? otherwise how does it work? (in term of disk read and transfer?) after user issues command move/fetch, how does postgre speed up the query in compare to normal selection? Thanks a lot, regards, MT Ho __________________________________ Do you Yahoo!? Yahoo! Mail Address AutoComplete - You start. We finish. http://promotions.yahoo.com/new_mail From pgsql-performance-owner@postgresql.org Tue Oct 12 13:21:15 2004 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 3453632A207 for ; Tue, 12 Oct 2004 13:21:13 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 88801-06 for ; Tue, 12 Oct 2004 12:21:01 +0000 (GMT) Received: from hotmail.com (bay18-dav8.bay18.hotmail.com [65.54.187.188]) by svr1.postgresql.org (Postfix) with ESMTP id 91A2F32A1EB for ; Tue, 12 Oct 2004 13:21:00 +0100 (BST) Received: from mail pickup service by hotmail.com with Microsoft SMTPSVC; Tue, 12 Oct 2004 05:21:00 -0700 Received: from 67.81.98.198 by BAY18-DAV8.phx.gbl with DAV; Tue, 12 Oct 2004 12:20:46 +0000 X-Originating-IP: [67.81.98.198] X-Originating-Email: [awerman2@hotmail.com] X-Sender: awerman2@hotmail.com From: "Aaron Werman" To: "Kris Jurka" , "Gabriele Bartolini" Cc: References: <6.1.2.0.2.20041011224414.01fadec0@box.tin.it> Subject: Re: Normal case or bad query plan? Date: Tue, 12 Oct 2004 08:20:52 -0400 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 6.00.2800.1437 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 Message-ID: X-OriginalArrivalTime: 12 Oct 2004 12:21:00.0672 (UTC) FILETIME=[EF4CE800:01C4B055] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/160 X-Sequence-Number: 8636 Makes sense. See DB2 8.2 info on their new implementation of cross column statistics. If this is common and you're willing to change code, you can fake that by adding a operation index on some hash function of both columns, and search for both columns and the hash. ----- Original Message ----- From: "Kris Jurka" To: "Gabriele Bartolini" Cc: Sent: Monday, October 11, 2004 5:17 PM Subject: Re: [PERFORM] Normal case or bad query plan? > > > On Mon, 11 Oct 2004, Gabriele Bartolini wrote: > > > -------------------------------------------------------------------------- ------------------------------------------- > > Seq Scan on ip2location (cost=0.00..30490.65 rows=124781 width=8) > > (actual time=5338.120..40237.283 rows=1 loops=1) > > Filter: ((1040878301::bigint >= ip_address_from) AND > > (1040878301::bigint <= ip_address_to)) > > Total runtime: 40237.424 ms > > > > I believe the problem is that pg's lack of cross-column statistics is > producing the poor number of rows estimate. The number of rows mataching > just the first 1040878301::bigint >= ip_address_from condition is 122774 > which is roughtly 10% of the table. I imagine the query planner > believes that the other condition alone will match the other 90% of the > table. The problem is that it doesn't know that these two ranges' > intersection is actually tiny. The planner assumes a complete or nearly > complete overlap so it thinks it will need to fetch 10% of the rows from > both the index and the heap and chooses a seqscan. > > Kris Jurka > > ---------------------------(end of broadcast)--------------------------- > TIP 5: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faqs/FAQ.html > From pgsql-performance-owner@postgresql.org Tue Oct 12 13:36:25 2004 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 8828532A7BC for ; Tue, 12 Oct 2004 13:36:19 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91533-08 for ; Tue, 12 Oct 2004 12:36:10 +0000 (GMT) Received: from boutiquenumerique.com (gailleton-2-82-67-9-10.fbx.proxad.net [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 22A8732A486 for ; Tue, 12 Oct 2004 13:36:10 +0100 (BST) Received: (qmail 12858 invoked from network); 12 Oct 2004 14:36:10 +0200 Received: from unknown (HELO musicbox) (boutiquenumerique-lists@192.168.0.2) by gailleton-2-82-67-9-10.fbx.proxad.net with SMTP; 12 Oct 2004 14:36:10 +0200 To: "my ho" , pgsql-performance@postgresql.org Subject: Re: execute cursor fetch References: <20041012114343.69531.qmail@web54110.mail.yahoo.com> Message-ID: From: =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= 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: Tue, 12 Oct 2004 14:36:10 +0200 In-Reply-To: <20041012114343.69531.qmail@web54110.mail.yahoo.com> User-Agent: Opera M2/7.54 (Linux, build 751) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/161 X-Sequence-Number: 8637 I just discovered this : http://www.postgresql.org/docs/7.4/static/jdbc-query.html#AEN24298 On Tue, 12 Oct 2004 04:43:43 -0700 (PDT), my ho wrote: > Hi, > If anyone can help pls, I have a question abt the > execution of cursor create/fetch/move , in particular > about disk cost. When a cursor is created, is the > whole table (with the required columns) got put into > memory? otherwise how does it work? (in term of disk > read and transfer?) after user issues command > move/fetch, how does postgre speed up the query in > compare to normal selection? > Thanks a lot, > regards, > MT Ho > > > > > __________________________________ > Do you Yahoo!? > Yahoo! Mail Address AutoComplete - You start. We finish. > http://promotions.yahoo.com/new_mail > > ---------------------------(end of broadcast)--------------------------- > TIP 4: Don't 'kill -9' the postmaster > From pgsql-performance-owner@postgresql.org Tue Oct 12 14:05:52 2004 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 8D64F329CFB for ; Tue, 12 Oct 2004 14:05:49 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 00356-09 for ; Tue, 12 Oct 2004 13:05:37 +0000 (GMT) Received: from ucsns.ucs.co.za (ucs.co.za [196.23.43.2]) by svr1.postgresql.org (Postfix) with ESMTP id CED4C32A92B for ; Tue, 12 Oct 2004 14:05:36 +0100 (BST) Received: from localhost (localhost.localdomain [127.0.0.1]) by ucsns.ucs.co.za (Postfix) with ESMTP id 4F52D2BD73 for ; Tue, 12 Oct 2004 15:05:22 +0200 (SAST) Received: from ucsns.ucs.co.za ([127.0.0.1]) by localhost (ucsns.ucs.co.za [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 01807-04 for ; Tue, 12 Oct 2004 15:05:12 +0200 (SAST) Received: from ucspost.ucs.co.za (mailgw1.ucs.co.za [196.23.43.253]) by ucsns.ucs.co.za (Postfix) with ESMTP id 1E4C62BD72 for ; Tue, 12 Oct 2004 15:05:12 +0200 (SAST) Received: from jhb.ucs.co.za (jhb.ucs.co.za [172.31.1.3]) by ucspost.ucs.co.za (Postfix) with ESMTP id 6BD65DA1A6 for ; Tue, 12 Oct 2004 15:05:14 +0200 (SAST) Received: from svb.ucs.co.za (svb.ucs.co.za [172.31.1.148]) by jhb.ucs.co.za (Postfix) with SMTP id 6302497A48 for ; Tue, 12 Oct 2004 15:05:15 +0200 (SAST) Date: Tue, 12 Oct 2004 15:05:15 +0200 From: Stef To: pgsql-performance@postgresql.org Subject: Re: execute cursor fetch Message-ID: <20041012150515.167c5e5e@svb.ucs.co.za> In-Reply-To: References: <20041012114343.69531.qmail@web54110.mail.yahoo.com> X-Mailer: Sylpheed-Claws 0.9.12b (GTK+ 1.2.10; i386-pc-linux-gnu) User-Agent: sillypheed-claws (anti-aliased) X-Face: GFoC95e6r)_TTG>n~=uFLojP=O~4W@Ms]>:.DMm/')(z3\Mwj^XP@? Q:3";lD.OM1"^mDu}2NJ@US:)dO:U*iY5EM50&Tx. X-Operating-System: sid X-X-X: _-^-_ Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable X-Virus-Scanned: by amavisd-new at ucs.co.za X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/162 X-Sequence-Number: 8638 Pierre-Fr=E9d=E9ric Caillaud mentioned : =3D> http://www.postgresql.org/docs/7.4/static/jdbc-query.html#AEN24298 My question is : Is this only true for postgres versions >=3D 7.4 ? I see the same section about "Setting fetch size to turn cursors on and off" is not in the postgres 7.3.7 docs. Does this mean 7.3 the JDBC driver for postgres < 7.4 doesn't support this ? Kind Regards Stefan From pgsql-performance-owner@postgresql.org Tue Oct 12 15:08:07 2004 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 C529A32B1AB for ; Tue, 12 Oct 2004 15:08:05 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 26915-02 for ; Tue, 12 Oct 2004 14:07:53 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 964C132B19A for ; Tue, 12 Oct 2004 15:07:56 +0100 (BST) 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 i9CE7vuv001857; Tue, 12 Oct 2004 10:07:57 -0400 (EDT) To: my ho Cc: pgsql-performance@postgresql.org Subject: Re: execute cursor fetch In-reply-to: <20041012114343.69531.qmail@web54110.mail.yahoo.com> References: <20041012114343.69531.qmail@web54110.mail.yahoo.com> Comments: In-reply-to my ho message dated "Tue, 12 Oct 2004 04:43:43 -0700" Date: Tue, 12 Oct 2004 10:07:56 -0400 Message-ID: <1856.1097590076@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/163 X-Sequence-Number: 8639 my ho writes: > If anyone can help pls, I have a question abt the > execution of cursor create/fetch/move , in particular > about disk cost. When a cursor is created, is the > whole table (with the required columns) got put into > memory? No. The plan is set up and then incrementally executed each time you say FETCH. > how does postgre speed up the query in > compare to normal selection? The only difference from a SELECT is that the planner will prefer "fast-start" plans, on the theory that you may not be intending to retrieve the whole result. For instance it might prefer an indexscan to a seqscan + sort, when it otherwise wouldn't. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Oct 12 15:29:47 2004 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 A52A532A1BB for ; Tue, 12 Oct 2004 15:29:45 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 35685-01 for ; Tue, 12 Oct 2004 14:29:39 +0000 (GMT) Received: from vsmtp4.tin.it (unknown [212.216.176.150]) by svr1.postgresql.org (Postfix) with ESMTP id E6A1732A0AC for ; Tue, 12 Oct 2004 15:29:42 +0100 (BST) Received: from ims3d.cp.tin.it (192.168.70.103) by vsmtp4.tin.it (7.0.027) id 41568C9D00709332; Tue, 12 Oct 2004 16:29:42 +0200 Received: from [192.168.70.227] by ims3d.cp.tin.it with HTTP; Tue, 12 Oct 2004 16:29:36 +0200 Date: Tue, 12 Oct 2004 16:29:36 +0200 Message-ID: <41536BE0000113F4@ims3d.cp.tin.it> In-Reply-To: From: "Gabriele Bartolini" Subject: Re: Normal case or bad query plan? To: "Kris Jurka" Cc: pgsql-performance@postgresql.org MIME-Version: 1.0 Content-Type: text/plain; charset="ISO-8859-15" Content-Transfer-Encoding: quoted-printable X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/164 X-Sequence-Number: 8640 Hi Kris, >I believe the problem is that pg's lack of cross-column statistics is >producing the poor number of rows estimate. The number of rows mataching I got your point now. I had not understood it last night but it makes really sense. >which is roughtly 10% of the table. I imagine the query planner >believes that the other condition alone will match the other 90% of the >table. The problem is that it doesn't know that these two ranges' >intersection is actually tiny. The planner assumes a complete or nearly >complete overlap so it thinks it will need to fetch 10% of the rows from Yep, because it performs those checks separately and it gets 10% for one check and 90% for the other. As Tom says, I should somehow make PostgreSQL see my data as a single entity in order to perform a real range check. I will study some way to obtain it. However, I got better results by specifying the grane of the statistics through "ALTER TABLE ... SET STATISTICS". FYI I set it to 1000 (the maximum) and I reduced the query's estimated time by the 90% (from 40000ms to 4000ms) although much slower than the index scan (200ms). I will play a bit with data types as Tom suggested. For now, thanks anyone who tried and helped me. Ciao, -Gabriele From pgsql-performance-owner@postgresql.org Tue Oct 12 15:41:12 2004 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 367AD32BD1F for ; Tue, 12 Oct 2004 15:41:09 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 40863-01 for ; Tue, 12 Oct 2004 14:40:55 +0000 (GMT) Received: from trofast.sesse.net (trofast.sesse.net [129.241.93.32]) by svr1.postgresql.org (Postfix) with ESMTP id A849832B1DF for ; Tue, 12 Oct 2004 15:40:56 +0100 (BST) Received: from sesse by trofast.sesse.net with local (Exim 3.36 #1 (Debian)) id 1CHNpe-00027y-00 for ; Tue, 12 Oct 2004 16:40:54 +0200 Date: Tue, 12 Oct 2004 16:40:54 +0200 From: "Steinar H. Gunderson" To: pgsql-performance@postgresql.org Subject: Re: Normal case or bad query plan? Message-ID: <20041012144054.GA8166@uio.no> Mail-Followup-To: pgsql-performance@postgresql.org References: <41536BE0000113F4@ims3d.cp.tin.it> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <41536BE0000113F4@ims3d.cp.tin.it> X-Operating-System: Linux 2.6.6 on a i686 User-Agent: Mutt/1.5.6+20040818i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/165 X-Sequence-Number: 8641 On Tue, Oct 12, 2004 at 04:29:36PM +0200, Gabriele Bartolini wrote: > FYI I set it to 1000 (the maximum) and I reduced the query's estimated time > by the 90% (from 40000ms to 4000ms) although much slower than the index > scan (200ms). Note that the estimated times are _not_ in ms. They are in multiples of a disk fetch (?). Thus, you can't compare estimated and real times like that. /* Steinar */ -- Homepage: http://www.sesse.net/ From pgsql-performance-owner@postgresql.org Tue Oct 12 16:26:06 2004 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 1B3C332A026 for ; Tue, 12 Oct 2004 16:26:04 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 58295-02 for ; Tue, 12 Oct 2004 15:25:59 +0000 (GMT) Received: from heimdall.hig.se (heimdall.hig.se [130.243.8.180]) by svr1.postgresql.org (Postfix) with SMTP id CB282329F96 for ; Tue, 12 Oct 2004 16:25:57 +0100 (BST) Received: from webmail.student.hig.se ([130.243.8.161]) by heimdall.hig.se (SAVSMTP 3.1.0.29) with SMTP id M2004101217220519020 for ; Tue, 12 Oct 2004 17:22:05 +0200 Received: from 130.243.14.147 (SquirrelMail authenticated user nd02tsk); by webmail.student.hig.se with HTTP; Tue, 12 Oct 2004 17:45:01 +0200 (CEST) Message-ID: <3388.130.243.14.147.1097595901.squirrel@130.243.14.147> Date: Tue, 12 Oct 2004 17:45:01 +0200 (CEST) Subject: Which plattform do you recommend I run PostgreSQL for best performance? From: nd02tsk@student.hig.se To: pgsql-performance@postgresql.org User-Agent: SquirrelMail/1.4.3 X-Mailer: SquirrelMail/1.4.3 MIME-Version: 1.0 Content-Type: text/plain;charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Priority: 3 (Normal) Importance: Normal X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.3 tagged_above=0.0 required=5.0 tests=NO_REAL_NAME X-Spam-Level: X-Archive-Number: 200410/166 X-Sequence-Number: 8642 Hello I am doing a comparison between MySQL and PostgreSQL. In the MySQL manual it says that MySQL performs best with Linux 2.4 with ReiserFS on x86. Can anyone official, or in the know, give similar information regarding PostgreSQL? Also, any links to benchmarking tests available on the internet between MySQL and PostgreSQL would be appreciated. Thank you! Tim From pgsql-performance-owner@postgresql.org Tue Oct 12 16:47:51 2004 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 2AF9132BFBC for ; Tue, 12 Oct 2004 16:47:46 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 66459-02 for ; Tue, 12 Oct 2004 15:47:36 +0000 (GMT) Received: from bayswater1.ymogen.net (host-154-240-27-217.pobox.net.uk [217.27.240.154]) by svr1.postgresql.org (Postfix) with ESMTP id E208632BF28 for ; Tue, 12 Oct 2004 16:47:36 +0100 (BST) Received: from [82.68.132.233] (82-68-132-233.dsl.in-addr.zen.co.uk [82.68.132.233]) by bayswater1.ymogen.net (Postfix) with ESMTP id 7BA21A2CAE; Tue, 12 Oct 2004 16:47:34 +0100 (BST) Message-ID: <416BFC96.9070506@ymogen.net> Date: Tue, 12 Oct 2004 16:47:34 +0100 From: Matt Clark User-Agent: Mozilla Thunderbird 0.7.3 (Windows/20040803) X-Accept-Language: en-us, en MIME-Version: 1.0 To: nd02tsk@student.hig.se Cc: pgsql-performance@postgresql.org Subject: Re: Which plattform do you recommend I run PostgreSQL for References: <3388.130.243.14.147.1097595901.squirrel@130.243.14.147> In-Reply-To: <3388.130.243.14.147.1097595901.squirrel@130.243.14.147> 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=1.0 tagged_above=0.0 required=5.0 tests=SUBJ_HAS_SPACES X-Spam-Level: X-Archive-Number: 200410/167 X-Sequence-Number: 8643 >In the MySQL manual it says that MySQL performs best with Linux 2.4 with >ReiserFS on x86. Can anyone official, or in the know, give similar >information regarding PostgreSQL? > > I'm neither official, nor in the know, but I do have a spare moment! I can tell you that any *NIX variant on any modern hardware platform will give you good performance, except for Cygwin/x86. Any differences between OSes on the same hardware are completely swamped by far more direct concerns like IO systems, database design, OS tuning etc. Pick the OS you're most familiar with is usually a good recommendation (and not just for Postgres). From pgsql-performance-owner@postgresql.org Tue Oct 12 17:26:36 2004 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 62F7832A35C for ; Tue, 12 Oct 2004 17:26:33 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 80025-09 for ; Tue, 12 Oct 2004 16:26:31 +0000 (GMT) Received: from mail1.acecape.com (mail1.acecape.com [66.114.74.12]) by svr1.postgresql.org (Postfix) with ESMTP id 8C87E32A2FA for ; Tue, 12 Oct 2004 17:26:30 +0100 (BST) Received: from p65-147.acedsl.com (p65-147.acedsl.com [66.114.65.147]) by mail1.acecape.com (8.12.11/8.12.11) with ESMTP id i9CGQSeV026344; Tue, 12 Oct 2004 12:26:29 -0400 Date: Tue, 12 Oct 2004 12:27:14 -0400 (EDT) From: Francisco Reyes X-X-Sender: fran@zoraida.natserv.net To: nd02tsk@student.hig.se Cc: pgsql-performance@postgresql.org Subject: Re: Which plattform do you recommend I run PostgreSQL for In-Reply-To: <3388.130.243.14.147.1097595901.squirrel@130.243.14.147> Message-ID: <20041012121734.Q4531@zoraida.natserv.net> References: <3388.130.243.14.147.1097595901.squirrel@130.243.14.147> 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=1.0 tagged_above=0.0 required=5.0 tests=SUBJ_HAS_SPACES X-Spam-Level: X-Archive-Number: 200410/168 X-Sequence-Number: 8644 On Tue, 12 Oct 2004 nd02tsk@student.hig.se wrote: > In the MySQL manual it says that MySQL performs best with Linux 2.4 with > ReiserFS on x86. Can anyone official, or in the know, give similar > information regarding PostgreSQL? Don't know which OS/filesystem PostgreSQL runs best on, but you should test on whatever OS and filesystem you are most experienced. Whatever speed gain you may get from "best setups" will mean little if the machine crashes and you don't know how to fix it and get it back up quickly. From pgsql-performance-owner@postgresql.org Tue Oct 19 22:03:21 2004 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 E418332C33E for ; Tue, 12 Oct 2004 18:21:50 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 19589-03 for ; Tue, 12 Oct 2004 17:21:42 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id 5D24B32CA51 for ; Tue, 12 Oct 2004 18:18:49 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i9CHImJ7017315 for ; Tue, 12 Oct 2004 17:18:48 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i9CGurEp096229 for pgsql-performance@postgresql.org; Tue, 12 Oct 2004 16:56:53 GMT From: Mischa Sandberg Reply-To: ischamay.andbergsay@activestateway.com User-Agent: Mozilla Thunderbird 0.7.3 (X11/20040803) X-Accept-Language: en-us, en MIME-Version: 1.0 X-Newsgroups: comp.databases.postgresql.performance Subject: Re: Normal case or bad query plan? References: <6.1.2.0.2.20041011224414.01fadec0@box.tin.it> In-Reply-To: <6.1.2.0.2.20041011224414.01fadec0@box.tin.it> Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Lines: 129 Message-ID: Date: Tue, 12 Oct 2004 16:56:53 GMT To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.1 tagged_above=0.0 required=5.0 tests=NO_DNS_FOR_FROM X-Spam-Level: * X-Archive-Number: 200410/286 X-Sequence-Number: 8762 This may sound more elaborate than it's worth, but I don't know of a better way to avoid a table scan. You want to index on a computed value that is a common prefix of your FROM and TO fields. The next step is to search on a fixed SET of prefixes of different lengths. For example, some of your ranges might be common in the first 3 bytes of ipaddr, some in two, some in only one. You create and index on one common prefix of either 1,2 or 3 bytes, for each row. Your query then looks something like (pardon my ignorance in PGSQL) select * from ip2location where ip2prefix in ( network(:myaddr || '/8'), network(:myaddr || '/16'), network(:myaddr || '/24'), :myaddr --- assuming single-address ranges are possible ) and :myaddr between ip_address_from and ip_address_to Although this looks a little gross, it hits very few records. It also adapts cleanly to a join between ip2location and a table of ip addrs. Gabriele Bartolini wrote: > Hi guys, > > please consider this scenario. I have this table: > > CREATE TABLE ip2location ( > ip_address_from BIGINT NOT NULL, > ip_address_to BIGINT NOT NULL, > id_location BIGINT NOT NULL, > PRIMARY KEY (ip_address_from, ip_address_to) > ); > > I created a cluster on its primary key, by running: > CLUSTER ip2location_ip_address_from_key ON ip2location; > > This allowed me to organise data in a more efficient way: the data that > is contained are ranges of IP addresses with empty intersections; for > every IP class there is a related location's ID. The total number of > entries is 1392443. > > For every IP address I have, an application retrieves the corresponding > location's id from the above table, by running a query like: > > SELECT id_location FROM ip2location WHERE '11020000111' >= > ip_address_from AND '11020000111' <= ip_address_to; > > For instance, by running the 'EXPLAIN ANALYSE' command, I get this > "funny" result: > > > QUERY PLAN > --------------------------------------------------------------------------------------------------------------------- > > Seq Scan on ip2location (cost=0.00..30490.65 rows=124781 width=8) > (actual time=5338.120..40237.283 rows=1 loops=1) > Filter: ((1040878301::bigint >= ip_address_from) AND > (1040878301::bigint <= ip_address_to)) > Total runtime: 40237.424 ms > > > With other data, that returns an empty set, I get: > > explain SELECT id_location FROM ip2location WHERE '11020000111' >= > ip_address_from AND '11020000111' <= ip_address_to; > > QUERY PLAN > ------------------------------------------------------------------------------------------------------- > > Index Scan using ip2location_ip_address_from_key on ip2location > (cost=0.00..419.16 rows=140 width=8) > Index Cond: ((11020000111::bigint >= ip_address_from) AND > (11020000111::bigint <= ip_address_to)) > > > I guess the planner chooses the best of the available options for the > first case, the sequential scan. This is not confirmed though by the > fact that, after I ran "SET enable_scan TO off", I got this: > > QUERY PLAN > ------------------------------------------------------------------------------------------------------------------------------------------------------------ > > Index Scan using ip2location_ip_address_from_key on ip2location > (cost=0.00..31505.73 rows=124781 width=8) (actual > time=2780.172..2780.185 rows=1 loops=1) > Index Cond: ((1040878301::bigint >= ip_address_from) AND > (1040878301::bigint <= ip_address_to)) > Total runtime: 2780.359 ms > > > Is this a normal case or should I worry? What am I missing? Do you have > any suggestion or comment to do (that would be extremely appreciated)? > Is the CLUSTER I created worthwhile or not? > > Thank you, > -Gabriele > > -- > Gabriele Bartolini: Web Programmer, ht://Dig & IWA/HWG Member, > ht://Check maintainer > Current Location: Prato, Toscana, Italia > angusgb@tin.it | http://www.prato.linux.it/~gbartolini | ICQ#129221447 > > "Leave every hope, ye who enter!", Dante Alighieri, Divine Comedy, > The Inferno > > > ------------------------------------------------------------------------ > > > --- > Outgoing mail is certified Virus Free. > Checked by AVG anti-virus system (http://www.grisoft.com). > Version: 6.0.773 / Virus Database: 520 - Release Date: 05/10/2004 > > > ------------------------------------------------------------------------ > > > ---------------------------(end of broadcast)--------------------------- > TIP 2: you can get off all lists at once with the unregister command > (send "unregister YourEmailAddressHere" to majordomo@postgresql.org) From pgsql-performance-owner@postgresql.org Tue Oct 12 22:18:57 2004 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 6311332ACB7 for ; Tue, 12 Oct 2004 22:18:54 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 28987-05 for ; Tue, 12 Oct 2004 21:18:50 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id EEE64329EEB for ; Tue, 12 Oct 2004 22:18:51 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i9CLImJ7029941 for ; Tue, 12 Oct 2004 21:18:48 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i9CKo2Bk019231 for pgsql-performance@postgresql.org; Tue, 12 Oct 2004 20:50:02 GMT From: Christopher Browne X-Newsgroups: comp.databases.postgresql.performance Subject: Re: Which plattform do you recommend I run PostgreSQL for best performance? Date: Tue, 12 Oct 2004 16:02:53 -0400 Organization: cbbrowne Computing Inc Lines: 44 Message-ID: References: <3388.130.243.14.147.1097595901.squirrel@130.243.14.147> 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 (Security Through Obscurity, linux) Cancel-Lock: sha1:nlmAziT7r6hsj/8FJNbIoHTLWJw= To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/170 X-Sequence-Number: 8646 In an attempt to throw the authorities off his trail, nd02tsk@student.hig.se transmitted: > I am doing a comparison between MySQL and PostgreSQL. > > In the MySQL manual it says that MySQL performs best with Linux 2.4 with > ReiserFS on x86. Can anyone official, or in the know, give similar > information regarding PostgreSQL? The fastest I have ever seen PostgreSQL run is on an IBM pSeries 650 system using AIX 5.1 and JFS2. There aren't many Linux systems that are anywhere _near_ as fast as that. There's some indication that FreeBSD 4.9, running the Berkeley FFS filesystem might be the quickest way of utilizing pedestrian IA-32 hardware, although it is _much_ more important to have a system for which you have a competent sysadmin than it is to have some "tweaked-out" OS configuration. In practice, competent people generally prefer to have systems that hum along nicely as opposed to systems that have ben "tweaked out" such that any little change will cause them to cave in. Benchmarks are useful in determining: a) Whether or not it is realistic to attempt a project, and b) Whether or not you have made conspicuous errors in configuring your systems. They are notoriously BAD as predictive tools, as the benchmarks sponsored by vendors get tweaked to make the vendors' products look good, as opposed to being written to be useful for prediction. See if you see anything useful from MySQL in this regard: > Also, any links to benchmarking tests available on the internet > between MySQL and PostgreSQL would be appreciated. Most database vendors have licenses that specifically forbid publishing benchmarks. -- (reverse (concatenate 'string "gro.mca" "@" "enworbbc")) http://www3.sympatico.ca/cbbrowne/oses.html Do you know where your towel is? From pgsql-performance-owner@postgresql.org Tue Oct 12 21:30:31 2004 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 31BDF32A611 for ; Tue, 12 Oct 2004 21:30:30 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 12497-01 for ; Tue, 12 Oct 2004 20:30:27 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id AC2C632A526 for ; Tue, 12 Oct 2004 21:30:28 +0100 (BST) 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 6495340; Tue, 12 Oct 2004 13:31:52 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Which plattform do you recommend I run PostgreSQL for best performance? Date: Tue, 12 Oct 2004 13:32:13 -0700 User-Agent: KMail/1.6.2 Cc: nd02tsk@student.hig.se References: <3388.130.243.14.147.1097595901.squirrel@130.243.14.147> In-Reply-To: <3388.130.243.14.147.1097595901.squirrel@130.243.14.147> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410121332.13849.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/169 X-Sequence-Number: 8645 Tim, > In the MySQL manual it says that MySQL performs best with Linux 2.4 with > ReiserFS on x86. Can anyone official, or in the know, give similar > information regarding PostgreSQL? PostgreSQL runs on a lot more platforms than MySQL; it's not even reasonable to compare some of them, like rtLinux, AIX or Cygwin. The only reasonable comparative testing that's been done seems to indicate that: Linux 2.6 is more efficient than FreeBSD which is more efficient than Linux 2.4 all of which are significantly more efficient than Solaris, and ReiserFS, XFS and JFS *seem* to outperform other Linux journalling FSes. However, as others have said, configuration and hardware will probably make more difference than your choice of OS except in extreme cases. And all of the above is being further tested, particularly the filesystems. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Wed Oct 13 02:03:28 2004 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 62FF532B684 for ; Wed, 13 Oct 2004 02:03:26 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 84099-05 for ; Wed, 13 Oct 2004 01:03:18 +0000 (GMT) Received: from candle.pha.pa.us (candle.pha.pa.us [207.106.42.251]) by svr1.postgresql.org (Postfix) with ESMTP id E678632B4B6 for ; Wed, 13 Oct 2004 02:03:20 +0100 (BST) Received: (from pgman@localhost) by candle.pha.pa.us (8.11.6/8.11.6) id i9D12vN08023; Tue, 12 Oct 2004 21:02:57 -0400 (EDT) From: Bruce Momjian Message-Id: <200410130102.i9D12vN08023@candle.pha.pa.us> Subject: Re: Caching of Queries In-Reply-To: <20041007.161842.77062117.t-ishii@sra.co.jp> To: Tatsuo Ishii Date: Tue, 12 Oct 2004 21:02:57 -0400 (EDT) Cc: tgl@sss.pgh.pa.us, matt@ymogen.net, pg@rbt.ca, awerman2@hotmail.com, scottakirkwood@gmail.com, pgsql-performance@postgresql.org X-Mailer: ELM [version 2.4ME+ PL108 (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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/171 X-Sequence-Number: 8647 Added to TODO: * Add RESET CONNECTION command to reset all session state This would include resetting of all variables (RESET ALL), dropping of all temporary tables, removal of any NOTIFYs, etc. This could be used for connection pooling. We could also change RESET ALL to have this functionality. --------------------------------------------------------------------------- Tatsuo Ishii wrote: > > Tatsuo Ishii writes: > > > First, it's not a particular problem with pgpool. As far as I know any > > > connection pool solution has exactly the same problem. Second, it's > > > easy to fix if PostgreSQL provides a functionarity such as:"drop all > > > temporary tables if any". > > > > I don't like that definition exactly --- it would mean that every time > > we add more backend-local state, we expect client drivers to know to > > issue the right incantation to reset that kind of state. > > > > I'm thinking we need to invent a command like "RESET CONNECTION" that > > resets GUC variables, drops temp tables, forgets active NOTIFYs, and > > generally does whatever else needs to be done to make the session state > > appear virgin. When we add more such state, we can fix it inside the > > backend without bothering clients. > > Great. It's much better than I propose. > > > I now realize that our "RESET ALL" command for GUC variables was not > > fully thought out. We could possibly redefine it as doing the above, > > but that might break some applications ... > > > > regards, tom lane > > > > ---------------------------(end of broadcast)--------------------------- > TIP 2: you can get off all lists at once with the unregister command > (send "unregister YourEmailAddressHere" to majordomo@postgresql.org) > -- 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 Tue Oct 19 22:03:32 2004 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 3AC4F329FC6 for ; Wed, 13 Oct 2004 09:49:02 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 94681-09 for ; Wed, 13 Oct 2004 08:48:57 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id EA60032A2BC for ; Wed, 13 Oct 2004 09:48:58 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i9D8mtJ7097029 for ; Wed, 13 Oct 2004 08:48:55 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i9D8kxi8096514 for pgsql-performance@postgresql.org; Wed, 13 Oct 2004 08:46:59 GMT From: CoL X-Newsgroups: comp.databases.postgresql.performance Subject: Re: Which plattform do you recommend I run PostgreSQL for best Date: Wed, 13 Oct 2004 10:46:59 +0200 Organization: Hub.Org Networking Services Lines: 20 Message-ID: References: <3388.130.243.14.147.1097595901.squirrel@130.243.14.147> 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: Thunderbird CoL X-Accept-Language: en-us, en In-Reply-To: <3388.130.243.14.147.1097595901.squirrel@130.243.14.147> To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/287 X-Sequence-Number: 8763 hi, nd02tsk@student.hig.se wrote: > Hello > > I am doing a comparison between MySQL and PostgreSQL. > > In the MySQL manual it says that MySQL performs best with Linux 2.4 with > ReiserFS on x86. Can anyone official, or in the know, give similar > information regarding PostgreSQL? > > Also, any links to benchmarking tests available on the internet between > MySQL and PostgreSQL would be appreciated. http://www.potentialtech.com/wmoran/postgresql.php http://www.varlena.com/varlena/GeneralBits/Tidbits/perf.html http://candle.pha.pa.us/main/writings/pgsql/hw_performance/ http://database.sarang.net/database/postgres/optimizing_postgresql.html C. From pgsql-performance-owner@postgresql.org Wed Oct 13 10:14:30 2004 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 6549D32C61D for ; Wed, 13 Oct 2004 10:14:27 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 02925-05 for ; Wed, 13 Oct 2004 09:14:20 +0000 (GMT) Received: from zaphod.profecta.se (zaphod.profecta.se [193.235.206.49]) by svr1.postgresql.org (Postfix) with ESMTP id DD92232C609 for ; Wed, 13 Oct 2004 10:14:20 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by zaphod.profecta.se (Postfix) with ESMTP id 9D0241140 for ; Wed, 13 Oct 2004 11:14:13 +0200 (CEST) Received: from [192.168.128.80] (ip050.noname4us.com [212.247.87.50]) by zaphod.profecta.se (Postfix) with ESMTP id 310A91124 for ; Wed, 13 Oct 2004 11:14:13 +0200 (CEST) Subject: query problem From: Robin Ericsson To: pgsql-performance@postgresql.org Content-Type: multipart/mixed; boundary="=-o5nK7lU1Fuj/F/D20jNX" Organization: Profecta HB Date: Wed, 13 Oct 2004 11:21:11 +0200 Message-Id: <1097659271.24018.68.camel@pylver.localhost.nu.> Mime-Version: 1.0 X-Mailer: Evolution 2.0.0 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/172 X-Sequence-Number: 8648 --=-o5nK7lU1Fuj/F/D20jNX Content-Type: text/plain Content-Transfer-Encoding: 7bit Hi, I sent this to general earlier but I was redirected to performance. The query have been running ok for quite some time, but after I did a vacuum on the database, it's very very slow. This IN-query is only 2 ids. Before the problem that in was a subselect-query returning around 6-7 ids. The tables included in the query are described in database.txt. status=# select count(id) from data; count --------- 1577621 (1 row) status=# select count(data_id) from data_values; count --------- 9680931 (1 row) I did run a new explain analyze on the query and found the attached result. The obvious problem I see is a full index scan in "idx_dv_data_id". I tried dropping and adding the index again, thats why is't called "idx_data_values_data_id" in the dump. status=# EXPLAIN ANALYZE status-# SELECT status-# data.entered, status-# data.machine_id, status-# datatemplate_intervals.template_id, status-# data_values.value status-# FROM status-# data, data_values, datatemplate_intervals status-# WHERE status-# datatemplate_intervals.id = data_values.template_id AND status-# data_values.data_id = data.id AND status-# data.machine_id IN (2,3) AND status-# current_timestamp::timestamp - interval '60 seconds' < data.entered; Regards, Robin -- Robin Ericsson Profecta HB --=-o5nK7lU1Fuj/F/D20jNX Content-Disposition: attachment; filename=explain2.txt Content-Type: text/plain; name=explain2.txt; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Hash Join (cost=28646.01..274260.15 rows=555706 width=24) (actual time=102323.087..102323.196 rows=5 loops=1) Hash Cond: ("outer".template_id = "inner".id) -> Merge Join (cost=28644.09..265922.62 rows=555706 width=24) (actual time=102322.632..102322.709 rows=5 loops=1) Merge Cond: ("outer".data_id = "inner".id) -> Index Scan using idx_dv_data_id on data_values (cost=0.00..205034.19 rows=9580032 width=16) (actual time=17.503..86263.130 rows=9596747 loops=1) -> Sort (cost=28644.09..28870.83 rows=90697 width=16) (actual time=0.829..0.835 rows=1 loops=1) Sort Key: data.id -> Index Scan using idx_d_entered on data (cost=0.00..20202.81 rows=90697 width=16) (actual time=0.146..0.185 rows=1 loops=1) Index Cond: (((('now'::text)::timestamp(6) with time zone)::timestamp without time zone - '00:01:00'::interval) < entered) Filter: ((machine_id = 2) OR (machine_id = 3)) -> Hash (cost=1.74..1.74 rows=74 width=8) (actual time=0.382..0.382 rows=0 loops=1) -> Seq Scan on datatemplate_intervals (cost=0.00..1.74 rows=74 width=8) (actual time=0.024..0.250 rows=74 loops=1) Total runtime: 102323.491 ms (13 rows) --=-o5nK7lU1Fuj/F/D20jNX Content-Disposition: attachment; filename=database.txt Content-Type: text/plain; name=database.txt; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit status=# \d data Table "public.data" Column | Type | Modifiers ------------+-----------------------------+-------------------------------------------- id | integer | not null default nextval('data_seq'::text) updated | timestamp without time zone | entered | timestamp without time zone | machine_id | integer | datas | character varying(512)[] | Indexes: "data_pkey" primary key, btree (id) "idx_d_entered" btree (entered) "idx_d_machine_id" btree (machine_id) Foreign-key constraints: "machine_id" FOREIGN KEY (machine_id) REFERENCES machines(id) Triggers: data_datestamp BEFORE INSERT OR UPDATE ON data FOR EACH ROW EXECUTE PROCEDURE datestamp_e() status=# \d data_values Table "public.data_values" Column | Type | Modifiers -------------+-----------------------------+----------- updated | timestamp without time zone | entered | timestamp without time zone | data_id | integer | template_id | integer | value | character varying(512) | Indexes: "idx_data_values_data_id" btree (data_id) "idx_dv_template_id" btree (template_id) Foreign-key constraints: "data_id" FOREIGN KEY (data_id) REFERENCES data(id) "template_id" FOREIGN KEY (template_id) REFERENCES datatemplate_intervals(id) Triggers: data_values_datestamp BEFORE INSERT OR UPDATE ON data_values FOR EACH ROW EXECUTE PROCEDURE datestamp_e() status=# \d datatemplate_intervals Table "public.datatemplate_intervals" Column | Type | Modifiers -------------+-----------------------------+-------------------------------------------------------------- id | integer | not null default nextval('datatemplate_intervals_seq'::text) updated | timestamp without time zone | entered | timestamp without time zone | machine_id | integer | template_id | integer | interval | integer | Indexes: "datatemplate_intervals_pkey" primary key, btree (id) "idx_di_machine_id" btree (machine_id) "idx_di_template_id" btree (template_id) Foreign-key constraints: "machine_id" FOREIGN KEY (machine_id) REFERENCES machines(id) "template_id" FOREIGN KEY (template_id) REFERENCES datatemplates(id) Triggers: datatemplate_intervals_datestamp BEFORE INSERT OR UPDATE ON datatemplate_intervals FOR EACH ROW EXECUTE PROCEDURE datestamp_e() --=-o5nK7lU1Fuj/F/D20jNX-- From pgsql-performance-owner@postgresql.org Wed Oct 13 15:38:02 2004 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 CDF5A32B4DE for ; Wed, 13 Oct 2004 15:38:00 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 08358-05 for ; Wed, 13 Oct 2004 14:37:46 +0000 (GMT) Received: from server272.com (server272.com [64.14.68.49]) by svr1.postgresql.org (Postfix) with SMTP id 51FCB32B922 for ; Wed, 13 Oct 2004 15:37:49 +0100 (BST) Received: (qmail 14280 invoked from network); 13 Oct 2004 14:38:08 -0000 Received: from unknown (HELO pesky) (63.170.214.187) by server272.com with SMTP; 13 Oct 2004 14:38:08 -0000 Subject: Re: query problem From: ken To: Robin Ericsson Cc: pgsql-performance@postgresql.org In-Reply-To: <1097659271.24018.68.camel@pylver.localhost.nu.> References: <1097659271.24018.68.camel@pylver.localhost.nu.> Content-Type: text/plain Message-Id: <1097678263.9676.275.camel@pesky> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.5 Date: Wed, 13 Oct 2004 07:37:43 -0700 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/173 X-Sequence-Number: 8649 On Wed, 2004-10-13 at 02:21, Robin Ericsson wrote: > Hi, > > I sent this to general earlier but I was redirected to performance. > > The query have been running ok for quite some time, but after I did a > vacuum on the database, it's very very slow. Did you do a VACUUM FULL ANALYZE on the database or just a VACUUM? It looks like your statistics in your query are all off which ANALYZE should fix. > This IN-query is only 2 > ids. Before the problem that in was a subselect-query returning around > 6-7 ids. The tables included in the query are described in database.txt. > > status=# select count(id) from data; > count > --------- > 1577621 > (1 row) > > status=# select count(data_id) from data_values; > count > --------- > 9680931 > (1 row) > > I did run a new explain analyze on the query and found the attached > result. The obvious problem I see is a full index scan in > "idx_dv_data_id". I tried dropping and adding the index again, thats why > is't called "idx_data_values_data_id" in the dump. > > status=# EXPLAIN ANALYZE > status-# SELECT > status-# data.entered, > status-# data.machine_id, > status-# datatemplate_intervals.template_id, > status-# data_values.value > status-# FROM > status-# data, data_values, datatemplate_intervals > status-# WHERE > status-# datatemplate_intervals.id = data_values.template_id AND > status-# data_values.data_id = data.id AND > status-# data.machine_id IN (2,3) AND > status-# current_timestamp::timestamp - interval '60 seconds' < > data.entered; > > > > Regards, > Robin > From pgsql-performance-owner@postgresql.org Wed Oct 13 16:03:21 2004 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 AB89A329F4D for ; Wed, 13 Oct 2004 16:03:19 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 15894-07 for ; Wed, 13 Oct 2004 15:03:15 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 39B34329F3E for ; Wed, 13 Oct 2004 16:03:14 +0100 (BST) 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 i9DF3DDD014685; Wed, 13 Oct 2004 11:03:13 -0400 (EDT) To: Robin Ericsson Cc: pgsql-performance@postgresql.org Subject: Re: query problem In-reply-to: <1097659271.24018.68.camel@pylver.localhost.nu.> References: <1097659271.24018.68.camel@pylver.localhost.nu.> Comments: In-reply-to Robin Ericsson message dated "Wed, 13 Oct 2004 11:21:11 +0200" Date: Wed, 13 Oct 2004 11:03:13 -0400 Message-ID: <14684.1097679793@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/174 X-Sequence-Number: 8650 Robin Ericsson writes: > I sent this to general earlier but I was redirected to performance. Actually, I think I suggested that you consult the pgsql-performance archives, where this type of problem has been hashed out before. See for instance this thread: http://archives.postgresql.org/pgsql-performance/2004-07/msg00169.php particularly http://archives.postgresql.org/pgsql-performance/2004-07/msg00175.php http://archives.postgresql.org/pgsql-performance/2004-07/msg00184.php http://archives.postgresql.org/pgsql-performance/2004-07/msg00185.php which show three different ways of getting the planner to do something sane with an index range bound like "now() - interval". regards, tom lane From pgsql-performance-owner@postgresql.org Wed Oct 13 17:19:13 2004 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 048EC32A4C4 for ; Wed, 13 Oct 2004 17:19:10 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 47143-06 for ; Wed, 13 Oct 2004 16:19:04 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id 12FF332B980 for ; Wed, 13 Oct 2004 17:19:02 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i9DGJ1J9048775 for ; Wed, 13 Oct 2004 16:19:01 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i9DG43Dc043313 for pgsql-performance@postgresql.org; Wed, 13 Oct 2004 16:04:03 GMT From: Chris Browne X-Newsgroups: comp.databases.postgresql.performance Subject: Opteron vs RHAT Date: Wed, 13 Oct 2004 11:52:28 -0400 Organization: cbbrowne Computing Inc Lines: 24 Message-ID: <60fz4ivcn7.fsf_-_@dba2.int.libertyrms.com> References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <20041011173819.GB22076@phlogiston.dyndns.org> <1097518852.54644.20.camel@home> <416B7F4C.5070600@ymogen.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Complaints-To: usenet@news.hub.org User-Agent: Gnus/5.1006 (Gnus v5.10.6) XEmacs/21.4 (Security Through Obscurity, linux) Cancel-Lock: sha1:/FtYmNcpjHRMUzCxI5qunzsWgXs= To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/177 X-Sequence-Number: 8653 matt@ymogen.net (Matt Clark) writes: >>As for "vendor support" for Opteron, that sure looks like a >>trainwreck... If you're going through IBM, then they won't want to >>respond to any issues if you're not running a "bog-standard" RHAS/RHES >>release from Red Hat. And that, on Opteron, is preposterous, because >>there's plenty of the bits of Opteron support that only ever got put >>in Linux 2.6, whilst RHAT is still back in the 2.4 days. > > To be fair, they have backported a boatload of 2.6 features to their kernel: > http://www.redhat.com/software/rhel/kernel26/ > > And that page certainly isn't an exhaustive list... To be fair, we keep on actually running into things that _can't_ be backported, like fibrechannel drivers that were written to take advantage of changes in the SCSI support in 2.6. This sort of thing will be particularly problematic with Opteron, where the porting efforts for AMD64 have taken place alongside the creation of 2.6. -- let name="cbbrowne" and tld="cbbrowne.com" in String.concat "@" [name;tld];; http://www.ntlug.org/~cbbrowne/linuxxian.html A VAX is virtually a computer, but not quite. From pgsql-performance-owner@postgresql.org Wed Oct 13 16:54:32 2004 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 AFA8432A684 for ; Wed, 13 Oct 2004 16:54:29 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 37055-06 for ; Wed, 13 Oct 2004 15:54:24 +0000 (GMT) Received: from zaphod.profecta.se (profecta1.cust.morotsmedia.se [193.235.206.49]) by svr1.postgresql.org (Postfix) with ESMTP id A62A332A465 for ; Wed, 13 Oct 2004 16:54:22 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by zaphod.profecta.se (Postfix) with ESMTP id 60E7A10F4; Wed, 13 Oct 2004 17:54:08 +0200 (CEST) Received: from [192.168.128.80] (ip050.noname4us.com [212.247.87.50]) by zaphod.profecta.se (Postfix) with ESMTP id 06659A57; Wed, 13 Oct 2004 17:54:08 +0200 (CEST) Subject: Re: query problem From: Robin Ericsson To: Tom Lane Cc: pgsql-performance@postgresql.org In-Reply-To: <14684.1097679793@sss.pgh.pa.us> References: <1097659271.24018.68.camel@pylver.localhost.nu.> <14684.1097679793@sss.pgh.pa.us> Content-Type: text/plain Organization: Profecta HB Date: Wed, 13 Oct 2004 18:01:10 +0200 Message-Id: <1097683270.24018.124.camel@pylver.localhost.nu.> Mime-Version: 1.0 X-Mailer: Evolution 2.0.0 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/175 X-Sequence-Number: 8651 On Wed, 2004-10-13 at 11:03 -0400, Tom Lane wrote: > Robin Ericsson writes: > > I sent this to general earlier but I was redirected to performance. > > Actually, I think I suggested that you consult the pgsql-performance > archives, where this type of problem has been hashed out before. > See for instance this thread: > http://archives.postgresql.org/pgsql-performance/2004-07/msg00169.php > particularly > http://archives.postgresql.org/pgsql-performance/2004-07/msg00175.php > http://archives.postgresql.org/pgsql-performance/2004-07/msg00184.php > http://archives.postgresql.org/pgsql-performance/2004-07/msg00185.php > which show three different ways of getting the planner to do something > sane with an index range bound like "now() - interval". Using exact timestamp makes the query go back as it should in speed (see explain below). However I still have the problem using a stored procedure or even using the "ago"-example from above. regards, Robin status=# explain analyse status-# SELECT status-# data.entered, status-# data.machine_id, status-# datatemplate_intervals.template_id, status-# data_values.value status-# FROM status-# data, data_values, datatemplate_intervals status-# WHERE status-# datatemplate_intervals.id = data_values.template_id AND status-# data_values.data_id = data.id AND status-# data.machine_id IN (SELECT machine_id FROM machine_group_xref WHERE group_id = 1) AND status-# '2004-10-13 17:47:36.902062' < data.entered status-# ; QUERY PLAN -------------------------------------------------------------------------------------------------------------------------------------------------- Hash Join (cost=3.09..481.28 rows=777 width=24) (actual time=0.637..1.804 rows=57 loops=1) Hash Cond: ("outer".template_id = "inner".id) -> Nested Loop (cost=1.17..467.71 rows=776 width=24) (actual time=0.212..1.012 rows=57 loops=1) -> Hash IN Join (cost=1.17..9.56 rows=146 width=16) (actual time=0.165..0.265 rows=9 loops=1) Hash Cond: ("outer".machine_id = "inner".machine_id) -> Index Scan using idx_d_entered on data (cost=0.00..6.14 rows=159 width=16) (actual time=0.051..0.097 rows=10 loops=1) Index Cond: ('2004-10-13 17:47:36.902062'::timestamp without time zone < entered) -> Hash (cost=1.14..1.14 rows=11 width=4) (actual time=0.076..0.076 rows=0 loops=1) -> Seq Scan on machine_group_xref (cost=0.00..1.14 rows=11 width=4) (actual time=0.017..0.054 rows=11 loops=1) Filter: (group_id = 1) -> Index Scan using idx_data_values_data_id on data_values (cost=0.00..3.07 rows=5 width=16) (actual time=0.018..0.047 rows=6 loops=9) Index Cond: (data_values.data_id = "outer".id) -> Hash (cost=1.74..1.74 rows=74 width=8) (actual time=0.382..0.382 rows=0 loops=1) -> Seq Scan on datatemplate_intervals (cost=0.00..1.74 rows=74 width=8) (actual time=0.024..0.248 rows=74 loops=1) Total runtime: 2.145 ms (15 rows) From pgsql-performance-owner@postgresql.org Wed Oct 13 17:09:22 2004 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 C9FD532B163 for ; Wed, 13 Oct 2004 17:09:20 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 41523-08 for ; Wed, 13 Oct 2004 16:09:15 +0000 (GMT) Received: from saturn.opentools.org (saturn.opentools.org [66.250.40.202]) by svr1.postgresql.org (Postfix) with ESMTP id ADE7A32A684 for ; Wed, 13 Oct 2004 17:09:15 +0100 (BST) Received: by saturn.opentools.org (Postfix, from userid 500) id 934FD3ECC; Wed, 13 Oct 2004 12:21:27 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by saturn.opentools.org (Postfix) with ESMTP id 8C0E5F58D for ; Wed, 13 Oct 2004 12:21:27 -0400 (EDT) Date: Wed, 13 Oct 2004 12:21:27 -0400 (EDT) From: Aaron Mulder X-X-Sender: ammulder@saturn.opentools.org To: pgsql-performance@postgresql.org Subject: Free PostgreSQL Training, Philadelphia, Oct 30 Message-ID: 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=0.5 tagged_above=0.0 required=5.0 tests=SUB_FREE_OFFER X-Spam-Level: X-Archive-Number: 200410/176 X-Sequence-Number: 8652 All, My company (Chariot Solutions) is sponsoring a day of free PostgreSQL training by Bruce Momjian (one of the core PostgreSQL developers). The day is split into 2 sessions (plus a Q&A session): * Mastering PostgreSQL Administration * PostgreSQL Performance Tuning Registration is required, and space is limited. The location is Malvern, PA (suburb of Philadelphia) and it's on Saturday Oct 30. For more information or to register, see http://chariotsolutions.com/postgresql.jsp Thanks, Aaron From pgsql-performance-owner@postgresql.org Wed Oct 13 17:23:51 2004 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 DD1ED32B439 for ; Wed, 13 Oct 2004 17:23:49 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49724-05 for ; Wed, 13 Oct 2004 16:23:44 +0000 (GMT) Received: from adsl-63-196-151-90.dsl.lsan03.pacbell.net (mail.valleypres.org [63.196.151.90]) by svr1.postgresql.org (Postfix) with ESMTP id 912D932A526 for ; Wed, 13 Oct 2004 17:23:43 +0100 (BST) Received: from clamav.valleypres.org ([172.16.14.58]) by adsl-63-196-151-90.dsl.lsan03.pacbell.net with esmtp (Exim 3.13 #5) id 1CHluf-0008MM-00; Wed, 13 Oct 2004 09:23:41 -0700 Received: from AT10111 (unknown [172.16.31.147]) by clamav.valleypres.org (Postfix) with SMTP id 9F14717C3BC; Wed, 13 Oct 2004 09:23:37 -0700 (PDT) Reply-To: From: "Bryan Encina" To: "'Aaron Mulder'" , Subject: Re: Free PostgreSQL Training, Philadelphia, Oct 30 Date: Wed, 13 Oct 2004 09:23:47 -0700 Message-ID: <036401c4b141$04defe10$931f10ac@AT10111> MIME-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook CWS, Build 9.0.2416 (9.0.2911.0) In-Reply-To: X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000 Importance: Normal X-yoursite-MailScanner-Information: Please contact the ISP for more information X-yoursite-MailScanner: Found to be clean X-MailScanner-From: bryan.encina@valleypres.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/178 X-Sequence-Number: 8654 > All, > My company (Chariot Solutions) is sponsoring a day of free > PostgreSQL training by Bruce Momjian (one of the core PostgreSQL > developers). The day is split into 2 sessions (plus a Q&A session): > > * Mastering PostgreSQL Administration > * PostgreSQL Performance Tuning > > Registration is required, and space is limited. The location is > Malvern, PA (suburb of Philadelphia) and it's on Saturday Oct 30. For > more information or to register, see > > http://chariotsolutions.com/postgresql.jsp > > Thanks, > Aaron Wow, that's good stuff, too bad there's no one doing stuff like that in the Los Angeles area. -b From pgsql-general-owner@postgresql.org Wed Oct 13 17:20:44 2004 X-Original-To: pgsql-general-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B620432C231 for ; Wed, 13 Oct 2004 17:20:41 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49116-01 for ; Wed, 13 Oct 2004 16:20:31 +0000 (GMT) Received: from zaphod.profecta.se (profecta1.cust.morotsmedia.se [193.235.206.49]) by svr1.postgresql.org (Postfix) with ESMTP id 265F232C223 for ; Wed, 13 Oct 2004 17:20:29 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by zaphod.profecta.se (Postfix) with ESMTP id A8C9011A4 for ; Wed, 13 Oct 2004 18:20:17 +0200 (CEST) Received: from [192.168.128.80] (ip050.noname4us.com [212.247.87.50]) by zaphod.profecta.se (Postfix) with ESMTP id 67E7611A3 for ; Wed, 13 Oct 2004 18:20:17 +0200 (CEST) Subject: Re: [PERFORM] query problem From: Robin Ericsson To: pgsql-general@postgresql.org In-Reply-To: <1097683270.24018.124.camel@pylver.localhost.nu.> References: <1097659271.24018.68.camel@pylver.localhost.nu.> <14684.1097679793@sss.pgh.pa.us> <1097683270.24018.124.camel@pylver.localhost.nu.> Content-Type: text/plain Organization: Profecta HB Date: Wed, 13 Oct 2004 18:27:20 +0200 Message-Id: <1097684840.24018.133.camel@pylver.localhost.nu.> Mime-Version: 1.0 X-Mailer: Evolution 2.0.0 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/550 X-Sequence-Number: 67074 On Wed, 2004-10-13 at 18:01 +0200, Robin Ericsson wrote: > Using exact timestamp makes the query go back as it should in speed (see > explain below). However I still have the problem using a stored > procedure or even using the "ago"-example from above. Well, changing ago() to use timestamp without time zone it goes ok in the query. This query now takes ~2ms. SELECT data.entered, data.machine_id, datatemplate_intervals.template_id, data_values.value FROM data, data_values, datatemplate_intervals WHERE datatemplate_intervals.id = data_values.template_id AND data_values.data_id = data.id AND data.machine_id IN (SELECT machine_id FROM machine_group_xref WHERE group_id = 1) AND ago('60 seconds') < data.entered Using it in this procedure. select * from get_current_machine_status('60 seconds', 1); takes ~100s. Maybe there's some obvious wrong I do about it? CREATE TYPE public.mstatus_holder AS (entered timestamp, machine_id int4, template_id int4, value varchar); CREATE OR REPLACE FUNCTION public.get_current_machine_status(interval, int4) RETURNS SETOF mstatus_holder AS ' SELECT data.entered, data.machine_id, datatemplate_intervals.template_id, data_values.value FROM data, data_values, datatemplate_intervals WHERE datatemplate_intervals.id = data_values.template_id AND data_values.data_id = data.id AND data.machine_id IN (SELECT machine_id FROM machine_group_xref WHERE group_id = $2) AND ago($1) < data.entered ' LANGUAGE 'sql' VOLATILE; Regards, Robin From pgsql-general-owner@postgresql.org Wed Oct 13 17:29:42 2004 X-Original-To: pgsql-general-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 31CE232A115 for ; Wed, 13 Oct 2004 17:29:40 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 52004-03 for ; Wed, 13 Oct 2004 16:29:34 +0000 (GMT) Received: from zaphod.profecta.se (profecta1.cust.morotsmedia.se [193.235.206.49]) by svr1.postgresql.org (Postfix) with ESMTP id C5FE232A0AC for ; Wed, 13 Oct 2004 17:29:34 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by zaphod.profecta.se (Postfix) with ESMTP id A2ABB1111 for ; Wed, 13 Oct 2004 18:29:23 +0200 (CEST) Received: from [192.168.128.80] (ip050.noname4us.com [212.247.87.50]) by zaphod.profecta.se (Postfix) with ESMTP id 49F37981 for ; Wed, 13 Oct 2004 18:29:23 +0200 (CEST) Subject: Re: [PERFORM] query problem From: Robin Ericsson To: Postgres general mailing list In-Reply-To: <1097684840.24018.133.camel@pylver.localhost.nu.> References: <1097659271.24018.68.camel@pylver.localhost.nu.> <14684.1097679793@sss.pgh.pa.us> <1097683270.24018.124.camel@pylver.localhost.nu.> <1097684840.24018.133.camel@pylver.localhost.nu.> Content-Type: text/plain Organization: Profecta HB Date: Wed, 13 Oct 2004 18:36:26 +0200 Message-Id: <1097685386.24018.134.camel@pylver.localhost.nu.> Mime-Version: 1.0 X-Mailer: Evolution 2.0.0 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/551 X-Sequence-Number: 67075 Sorry, this should have been going to performance. Regards, Robin From pgsql-performance-owner@postgresql.org Wed Oct 13 17:30:13 2004 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 C362032A892 for ; Wed, 13 Oct 2004 17:30:03 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49846-09 for ; Wed, 13 Oct 2004 16:29:58 +0000 (GMT) Received: from zaphod.profecta.se (zaphod.profecta.se [193.235.206.49]) by svr1.postgresql.org (Postfix) with ESMTP id 32D8332A6B1 for ; Wed, 13 Oct 2004 17:29:58 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by zaphod.profecta.se (Postfix) with ESMTP id E187611A3 for ; Wed, 13 Oct 2004 18:29:46 +0200 (CEST) Received: from [192.168.128.80] (ip050.noname4us.com [212.247.87.50]) by zaphod.profecta.se (Postfix) with ESMTP id A3BB71194 for ; Wed, 13 Oct 2004 18:29:46 +0200 (CEST) Subject: [Fwd: Re: [GENERAL] query problem] From: Robin Ericsson To: pgsql-performance@postgresql.org Content-Type: text/plain Organization: Profecta HB Date: Wed, 13 Oct 2004 18:36:50 +0200 Message-Id: <1097685410.24018.136.camel@pylver.localhost.nu.> Mime-Version: 1.0 X-Mailer: Evolution 2.0.0 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/179 X-Sequence-Number: 8655 Sent this to wrong list. -------- Forwarded Message -------- From: Robin Ericsson To: pgsql-general@postgresql.org Subject: Re: [GENERAL] [PERFORM] query problem Date: Wed, 13 Oct 2004 18:27:20 +0200 On Wed, 2004-10-13 at 18:01 +0200, Robin Ericsson wrote: > Using exact timestamp makes the query go back as it should in speed (see > explain below). However I still have the problem using a stored > procedure or even using the "ago"-example from above. Well, changing ago() to use timestamp without time zone it goes ok in the query. This query now takes ~2ms. SELECT data.entered, data.machine_id, datatemplate_intervals.template_id, data_values.value FROM data, data_values, datatemplate_intervals WHERE datatemplate_intervals.id = data_values.template_id AND data_values.data_id = data.id AND data.machine_id IN (SELECT machine_id FROM machine_group_xref WHERE group_id = 1) AND ago('60 seconds') < data.entered Using it in this procedure. select * from get_current_machine_status('60 seconds', 1); takes ~100s. Maybe there's some obvious wrong I do about it? CREATE TYPE public.mstatus_holder AS (entered timestamp, machine_id int4, template_id int4, value varchar); CREATE OR REPLACE FUNCTION public.get_current_machine_status(interval, int4) RETURNS SETOF mstatus_holder AS ' SELECT data.entered, data.machine_id, datatemplate_intervals.template_id, data_values.value FROM data, data_values, datatemplate_intervals WHERE datatemplate_intervals.id = data_values.template_id AND data_values.data_id = data.id AND data.machine_id IN (SELECT machine_id FROM machine_group_xref WHERE group_id = $2) AND ago($1) < data.entered ' LANGUAGE 'sql' VOLATILE; Regards, Robin ---------------------------(end of broadcast)--------------------------- TIP 7: don't forget to increase your free space map settings -- Robin Ericsson Profecta HB From pgsql-performance-owner@postgresql.org Wed Oct 13 17:47:31 2004 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 DC1C532C1B0 for ; Wed, 13 Oct 2004 17:47:28 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 57066-07 for ; Wed, 13 Oct 2004 16:47:23 +0000 (GMT) Received: from bayswater1.ymogen.net (host-154-240-27-217.pobox.net.uk [217.27.240.154]) by svr1.postgresql.org (Postfix) with ESMTP id 7EDDB32C117 for ; Wed, 13 Oct 2004 17:47:22 +0100 (BST) Received: from solent (unknown [213.165.136.10]) by bayswater1.ymogen.net (Postfix) with ESMTP id BD3F7A3105; Wed, 13 Oct 2004 17:47:20 +0100 (BST) Reply-To: From: "Matt Clark" To: "'Chris Browne'" , Subject: Re: Opteron vs RHAT Date: Wed, 13 Oct 2004 17:47:20 +0100 Organization: Ymogen Ltd Message-ID: <021b01c4b144$4e822120$8300a8c0@solent> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook, Build 10.0.6626 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 Importance: Normal In-Reply-To: <60fz4ivcn7.fsf_-_@dba2.int.libertyrms.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/180 X-Sequence-Number: 8656 > >>trainwreck... If you're going through IBM, then they won't want to=20 > >>respond to any issues if you're not running a=20 > "bog-standard" RHAS/RHES=20 > >>release from Red Hat.=20=20 ...> To be fair, we keep on actually running into things that=20 > _can't_ be backported, like fibrechannel drivers that were=20 > written to take advantage of changes in the SCSI support in 2.6. I thought IBM had good support for SUSE? I don't know why I thought that... From pgsql-performance-owner@postgresql.org Wed Oct 13 19:57:55 2004 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 D770C329F4D for ; Wed, 13 Oct 2004 19:57:53 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 98492-05 for ; Wed, 13 Oct 2004 18:57:45 +0000 (GMT) Received: from mail.osdl.org (fw.osdl.org [65.172.181.6]) by svr1.postgresql.org (Postfix) with ESMTP id 9176532AA03 for ; Wed, 13 Oct 2004 19:57:45 +0100 (BST) Received: (from markw@localhost) by mail.osdl.org (8.11.6/8.11.6) id i9DIviQ13537; Wed, 13 Oct 2004 11:57:44 -0700 Date: Wed, 13 Oct 2004 11:57:44 -0700 From: Mark Wong To: pgsql-performance@postgresql.org Cc: neilc@samurai.com, swm@alcove.com.au, Stephen Hemminger Subject: futex results with dbt-3 Message-ID: <20041013115744.A11278@osdl.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/181 X-Sequence-Number: 8657 Hi guys, I have some DBT-3 (decision support) results using Gavin's original futex patch fix. It's on out 8-way Pentium III Xeon systems in our STP environment. Definitely see some overall throughput performance on the tests, about 15% increase, but not change with respect to the number of context switches. Perhaps it doesn't really address what's causing the incredible number of context switches, but still helps. I think I'm seeing what Gavin has, that it seems to solves some concurrency problems on x86 platforms. Here's results without futexes: http://khack.osdl.org/stp/298114/ Results with futexes: http://khack.osdl.org/stp/298115/ Mark From pgsql-performance-owner@postgresql.org Wed Oct 13 21:12:16 2004 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 411F132ABB7 for ; Wed, 13 Oct 2004 21:12:13 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 23310-05 for ; Wed, 13 Oct 2004 20:11:59 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.196]) by svr1.postgresql.org (Postfix) with ESMTP id D7B73329FB1 for ; Wed, 13 Oct 2004 21:11:15 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 73so388935rnk for ; Wed, 13 Oct 2004 13:10:48 -0700 (PDT) Received: by 10.38.165.75 with SMTP id n75mr2611842rne; Wed, 13 Oct 2004 13:10:48 -0700 (PDT) Received: by 10.38.151.56 with HTTP; Wed, 13 Oct 2004 13:10:48 -0700 (PDT) Message-ID: <18f60194041013131067b6afd9@mail.gmail.com> Date: Wed, 13 Oct 2004 13:10:48 -0700 From: Aaron Glenn Reply-To: Aaron Glenn To: pgsql-performance@postgresql.org Subject: Re: Free PostgreSQL Training, Philadelphia, Oct 30 In-Reply-To: <036401c4b141$04defe10$931f10ac@AT10111> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: <036401c4b141$04defe10$931f10ac@AT10111> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/182 X-Sequence-Number: 8658 On Wed, 13 Oct 2004 09:23:47 -0700, Bryan Encina wrote: > > Wow, that's good stuff, too bad there's no one doing stuff like that in the > Los Angeles area. > > -b That makes two of us. Hanging out with Tom, Bruce, and others at OSCON 2002 was one of the most informative and fun times I've had. That and I could really stand to brush up on my Postgres basics aaron.glenn From pgsql-performance-owner@postgresql.org Wed Oct 13 23:34:43 2004 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 1C1DE32ACAB for ; Wed, 13 Oct 2004 23:34:40 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 58277-08 for ; Wed, 13 Oct 2004 22:34:36 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 9BB6B32A219 for ; Wed, 13 Oct 2004 23:34:37 +0100 (BST) 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 6500372; Wed, 13 Oct 2004 15:36:01 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org, Aaron Glenn Subject: Re: Free PostgreSQL Training, Philadelphia, Oct 30 Date: Wed, 13 Oct 2004 15:36:33 -0700 User-Agent: KMail/1.6.2 References: <036401c4b141$04defe10$931f10ac@AT10111> <18f60194041013131067b6afd9@mail.gmail.com> In-Reply-To: <18f60194041013131067b6afd9@mail.gmail.com> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410131536.33286.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/183 X-Sequence-Number: 8659 Aaron, > That makes two of us. Hanging out with Tom, Bruce, and others at OSCON > 2002 was one of the most informative and fun times I've had. That and > I could really stand to brush up on my Postgres basics You're thinking of Jan. Tom wasn't at OSCON. ;-) -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Thu Oct 14 01:49:31 2004 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 0EC3432BBB2 for ; Thu, 14 Oct 2004 01:49:30 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 90454-01 for ; Thu, 14 Oct 2004 00:49:12 +0000 (GMT) Received: from smtp108.mail.sc5.yahoo.com (smtp108.mail.sc5.yahoo.com [66.163.170.6]) by svr1.postgresql.org (Postfix) with SMTP id B717F32BB9E for ; Thu, 14 Oct 2004 01:49:14 +0100 (BST) Received: from unknown (HELO jupiter.black-lion.info) (janwieck@68.80.245.191 with login) by smtp108.mail.sc5.yahoo.com with SMTP; 14 Oct 2004 00:49:14 -0000 Received: from [172.21.8.18] ([192.168.192.101]) (authenticated bits=0) by jupiter.black-lion.info (8.12.10/8.12.9) with ESMTP id i9E0n294074029; Wed, 13 Oct 2004 20:49:07 -0400 (EDT) (envelope-from JanWieck@Yahoo.com) Message-ID: <416DCCFF.70304@Yahoo.com> Date: Wed, 13 Oct 2004 20:49:03 -0400 From: Jan Wieck User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.1) Gecko/20040707 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Christopher Browne Cc: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some References: <200410081443.16109.josh@agliodbs.com> In-Reply-To: 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/184 X-Sequence-Number: 8660 On 10/8/2004 10:10 PM, Christopher Browne wrote: > josh@agliodbs.com (Josh Berkus) wrote: >> I've been trying to peg the "sweet spot" for shared memory using >> OSDL's equipment. With Jan's new ARC patch, I was expecting that >> the desired amount of shared_buffers to be greatly increased. This >> has not turned out to be the case. > > That doesn't surprise me. Neither does it surprise me. > > My primary expectation would be that ARC would be able to make small > buffers much more effective alongside vacuums and seq scans than they > used to be. That does not establish anything about the value of > increasing the size buffer caches... The primary goal of ARC is to prevent total cache eviction caused by sequential scans. Which means it is designed to avoid the catastrophic impact of a pg_dump or other, similar access in parallel to the OLTP traffic. It would be much more interesting to see how a half way into a 2 hour measurement interval started pg_dump affects the response times. One also has to take a closer look at the data of the DBT2. What amount of that 32GB is high-frequently accessed, and therefore a good thing to live in the PG shared cache? A cache significantly larger than that doesn't make sense to me, under no cache strategy. Jan > >> This result is so surprising that I want people to take a look at it >> and tell me if there's something wrong with the tests or some >> bottlenecking factor that I've not seen. > > I'm aware of two conspicuous scenarios where ARC would be expected to > _substantially_ improve performance: > > 1. When it allows a VACUUM not to throw useful data out of > the shared cache in that VACUUM now only 'chews' on one > page of the cache; > > 2. When it allows a Seq Scan to not push useful data out of > the shared cache, for much the same reason. > > I don't imagine either scenario are prominent in the OSDL tests. > > Increasing the number of cache buffers _is_ likely to lead to some > slowdowns: > > - Data that passes through the cache also passes through kernel > cache, so it's recorded twice, and read twice... > > - The more cache pages there are, the more work is needed for > PostgreSQL to manage them. That will notably happen anywhere > that there is a need to scan the cache. > > - If there are any inefficiencies in how the OS kernel manages shared > memory, as their size scales, well, that will obviously cause a > slowdown. -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com # From pgsql-performance-owner@postgresql.org Thu Oct 14 01:52:56 2004 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 2A12732BBB6 for ; Thu, 14 Oct 2004 01:52:55 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 90645-02 for ; Thu, 14 Oct 2004 00:52:46 +0000 (GMT) Received: from smtp110.mail.sc5.yahoo.com (smtp110.mail.sc5.yahoo.com [66.163.170.8]) by svr1.postgresql.org (Postfix) with SMTP id 417B232BBB2 for ; Thu, 14 Oct 2004 01:52:46 +0100 (BST) Received: from unknown (HELO jupiter.black-lion.info) (janwieck@68.80.245.191 with login) by smtp110.mail.sc5.yahoo.com with SMTP; 14 Oct 2004 00:52:46 -0000 Received: from [172.21.8.18] ([192.168.192.101]) (authenticated bits=0) by jupiter.black-lion.info (8.12.10/8.12.9) with ESMTP id i9E0qg94074054; Wed, 13 Oct 2004 20:52:45 -0400 (EDT) (envelope-from JanWieck@Yahoo.com) Message-ID: <416DCDDB.1010902@Yahoo.com> Date: Wed, 13 Oct 2004 20:52:43 -0400 From: Jan Wieck User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.1) Gecko/20040707 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Kevin Brown Cc: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> In-Reply-To: <20041009112048.GA665@filer> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/185 X-Sequence-Number: 8661 On 10/9/2004 7:20 AM, Kevin Brown wrote: > Christopher Browne wrote: >> Increasing the number of cache buffers _is_ likely to lead to some >> slowdowns: >> >> - Data that passes through the cache also passes through kernel >> cache, so it's recorded twice, and read twice... > > Even worse, memory that's used for the PG cache is memory that's not > available to the kernel's page cache. Even if the overall memory Which underlines my previous statement, that a PG shared cache much larger than the high-frequently accessed data portion of the DB is counterproductive. Double buffering (kernel-disk-buffer plus shared buffer) only makes sense for data that would otherwise cause excessive memory copies in and out of the shared buffer. After that, in only lowers the memory available for disk buffers. Jan > usage in the system isn't enough to cause some paging to disk, most > modern kernels will adjust the page/disk cache size dynamically to fit > the memory demands of the system, which in this case means it'll be > smaller if running programs need more memory for their own use. > > This is why I sometimes wonder whether or not it would be a win to use > mmap() to access the data and index files -- doing so under a truly > modern OS would surely at the very least save a buffer copy (from the > page/disk cache to program memory) because the OS could instead > direcly map the buffer cache pages directly to the program's memory > space. > > Since PG often has to have multiple files open at the same time, and > in a production database many of those files will be rather large, PG > would have to limit the size of the mmap()ed region on 32-bit > platforms, which means that things like the order of mmap() operations > to access various parts of the file can become just as important in > the mmap()ed case as it is in the read()/write() case (if not more > so!). I would imagine that the use of mmap() on a 64-bit platform > would be a much, much larger win because PG would most likely be able > to mmap() entire files and let the OS work out how to order disk reads > and writes. > > The biggest problem as I see it is that (I think) mmap() would have to > be made to cooperate with malloc() for virtual address space. I > suspect issues like this have already been worked out by others, > however... > > > -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com # From pgsql-performance-owner@postgresql.org Thu Oct 14 03:30:24 2004 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 6784C32BDD6 for ; Thu, 14 Oct 2004 03:30:22 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 11081-06 for ; Thu, 14 Oct 2004 02:30:09 +0000 (GMT) Received: from sue.samurai.com (sue.samurai.com [205.207.28.74]) by svr1.postgresql.org (Postfix) with ESMTP id 8939732BDD1 for ; Thu, 14 Oct 2004 03:30:11 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by sue.samurai.com (Postfix) with ESMTP id DECF1197E8; Wed, 13 Oct 2004 22:30:11 -0400 (EDT) Received: from sue.samurai.com ([127.0.0.1]) by localhost (sue.samurai.com [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 94238-01-9; Wed, 13 Oct 2004 22:30:09 -0400 (EDT) Received: from [61.88.101.19] (unknown [61.88.101.19]) by sue.samurai.com (Postfix) with ESMTP id D1CAC197D0; Wed, 13 Oct 2004 22:30:07 -0400 (EDT) Subject: Re: futex results with dbt-3 From: Neil Conway To: Mark Wong Cc: pgsql-performance@postgresql.org, Gavin Sherry , Stephen Hemminger In-Reply-To: <20041013115744.A11278@osdl.org> References: <20041013115744.A11278@osdl.org> Content-Type: multipart/mixed; boundary="=-HuItn3AfDAyHfOdEZ/QL" Message-Id: <1097721005.4682.2.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Thu, 14 Oct 2004 12:30:05 +1000 X-Virus-Scanned: by amavisd-new at mailbox.samurai.com X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/186 X-Sequence-Number: 8662 --=-HuItn3AfDAyHfOdEZ/QL Content-Type: text/plain Content-Transfer-Encoding: 7bit On Thu, 2004-10-14 at 04:57, Mark Wong wrote: > I have some DBT-3 (decision support) results using Gavin's original > futex patch fix. I sent an initial description of the futex patch to the mailing lists last week, but it never appeared (from talking to Marc I believe it exceeded the size limit on -performance). In any case, the "futex patch" uses the Linux 2.6 futex API to implement PostgreSQL spinlocks. The hope is that using futexes will lead to better performance when there is contention for spinlocks (e.g. on a busy SMP system). The original patch was written by Stephen Hemminger at OSDL; Gavin and myself have done a bunch of additional bugfixing and optimization, as well as added IA64 support. I've attached a WIP copy of the patch to this email (it supports x86, x86-64 (untested) and IA64 -- more architectures can be added at request). I'll post a longer writeup when I submit the patch to -patches. > Definitely see some overall throughput performance on the tests, about > 15% increase, but not change with respect to the number of context > switches. I'm glad to see that there is a performance improvement; in my own testing on an 8-way P3 system provided by OSDL, I saw a similar improvement in pgbench performance (50 concurrent clients, 1000 transactions each, scale factor 75; without the patch, TPS/sec was between 180 and 185, with the patch TPS/sec was between 200 and 215). As for context switching, there was some earlier speculation that the patch might improve or even resolve the "CS storm" issue that some people have experienced with SMP Xeon P4 systems. I don't think we have enough evidence to answer this one way or the other at this point. -Neil --=-HuItn3AfDAyHfOdEZ/QL Content-Disposition: attachment; filename=futex-spin-16.patch Content-Type: text/x-patch; name=futex-spin-16.patch; charset=iso-8859-1 Content-Transfer-Encoding: 7bit Index: src/backend/storage/lmgr/s_lock.c =================================================================== RCS file: /var/lib/cvs/pgsql/src/backend/storage/lmgr/s_lock.c,v retrieving revision 1.32 diff -c -r1.32 s_lock.c *** src/backend/storage/lmgr/s_lock.c 30 Aug 2004 23:47:20 -0000 1.32 --- src/backend/storage/lmgr/s_lock.c 13 Oct 2004 06:23:26 -0000 *************** *** 15,26 **** */ #include "postgres.h" #include - #include #include "storage/s_lock.h" #include "miscadmin.h" /* * s_lock_stuck() - complain about a stuck spinlock */ --- 15,49 ---- */ #include "postgres.h" + #ifdef S_LOCK_TEST + #undef Assert + #define Assert(cond) DoAssert(cond, #cond, __FILE__, __LINE__) + + #define DoAssert(cond, text, file, line) \ + if (!(cond)) \ + { \ + printf("ASSERTION FAILED! [%s], file = %s, line = %d\n", \ + text, file, line); \ + abort(); \ + } + #endif + #include #include "storage/s_lock.h" #include "miscadmin.h" + #ifdef S_LOCK_TEST + #define LOCK_TEST_MSG() \ + do \ + { \ + fprintf(stdout, "*"); \ + fflush(stdout); \ + } while (0); + #else + #define LOCK_TEST_MSG() + #endif + /* * s_lock_stuck() - complain about a stuck spinlock */ *************** *** 38,43 **** --- 61,131 ---- #endif } + #ifdef HAVE_FUTEX + /* + * futex_lock_contended() is similar to s_lock() for the normal TAS + * implementation of spinlocks. When this function is invoked, we have + * failed to immediately acquire the spinlock, so we should spin some + * number of times attempting to acquire the lock before invoking + * sys_futex() to have the kernel wake us up later. "val" is the + * current value of the mutex we saw when we tried to acquire it; it + * may have changed since then, of course. + */ + void + futex_lock_contended(volatile slock_t *lock, slock_t val, + const char *file, int line) + { + int loop_count = 0; + + #define MAX_LOCK_WAIT 30 + #define SPINS_BEFORE_WAIT 100 + + Assert(val != FUTEX_UNLOCKED); + + if (val == FUTEX_LOCKED_NOWAITER) + val = atomic_exchange(lock, FUTEX_LOCKED_WAITER); + + while (val != FUTEX_UNLOCKED) + { + static struct timespec delay = { .tv_sec = MAX_LOCK_WAIT, + .tv_nsec = 0 }; + + LOCK_TEST_MSG(); + + #if defined(__i386__) || defined(__x86_64__) + /* See spin_delay() */ + __asm__ __volatile__(" rep; nop\n"); + #endif + + /* + * XXX: This code is derived from the Drepper algorithm, which + * doesn't spin (why, I'm not sure). We should actually change + * the lock status to "lock, with waiters" just before we wait + * on the futex, not before we begin looping (that avoids a + * system call when the lock is released). + */ + + /* XXX: worth using __builtin_expect() here? */ + if (++loop_count >= SPINS_BEFORE_WAIT) + { + loop_count = 0; + if (sys_futex(lock, FUTEX_OP_WAIT, + FUTEX_LOCKED_WAITER, &delay)) + { + if (errno == ETIMEDOUT) + s_lock_stuck(lock, file, line); + } + } + + /* + * Do a non-locking test before asserting the bus lock. + */ + if (*lock == FUTEX_UNLOCKED) + val = atomic_exchange(lock, FUTEX_LOCKED_WAITER); + } + } + + #else /* * s_lock(lock) - platform-independent portion of waiting for a spinlock. *************** *** 98,107 **** pg_usleep(cur_delay * 10000L); ! #if defined(S_LOCK_TEST) ! fprintf(stdout, "*"); ! fflush(stdout); ! #endif /* increase delay by a random fraction between 1X and 2X */ cur_delay += (int) (cur_delay * --- 186,192 ---- pg_usleep(cur_delay * 10000L); ! LOCK_TEST_MSG(); /* increase delay by a random fraction between 1X and 2X */ cur_delay += (int) (cur_delay * *************** *** 114,119 **** --- 199,205 ---- } } } + #endif /* HAVE_FUTEX */ /* * Various TAS implementations that cannot live in s_lock.h as no inline *************** *** 134,140 **** * All the gcc flavors that are not inlined */ - #if defined(__m68k__) static void tas_dummy() /* really means: extern int tas(slock_t --- 220,225 ---- *************** *** 265,272 **** --- 350,361 ---- test_lock.pad1 = test_lock.pad2 = 0x44; + printf("Address of lock variable: %p\n", &test_lock.lock); + S_INIT_LOCK(&test_lock.lock); + printf("<1> Value of lock variable: %d\n", test_lock.lock); + if (test_lock.pad1 != 0x44 || test_lock.pad2 != 0x44) { printf("S_LOCK_TEST: failed, declared datatype is wrong size\n"); *************** *** 280,285 **** --- 369,375 ---- } S_LOCK(&test_lock.lock); + printf("<2> Value of lock variable: %d\n", test_lock.lock); if (test_lock.pad1 != 0x44 || test_lock.pad2 != 0x44) { *************** *** 289,299 **** --- 379,391 ---- if (S_LOCK_FREE(&test_lock.lock)) { + printf("<3> Value of lock variable: %d\n", test_lock.lock); printf("S_LOCK_TEST: failed, lock not locked\n"); return 1; } S_UNLOCK(&test_lock.lock); + printf("<4> Value of lock variable: %d\n", test_lock.lock); if (test_lock.pad1 != 0x44 || test_lock.pad2 != 0x44) { *************** *** 326,332 **** printf(" if S_LOCK() and TAS() are working.\n"); fflush(stdout); ! s_lock(&test_lock.lock, __FILE__, __LINE__); printf("S_LOCK_TEST: failed, lock not locked\n"); return 1; --- 418,424 ---- printf(" if S_LOCK() and TAS() are working.\n"); fflush(stdout); ! S_LOCK(&test_lock.lock); printf("S_LOCK_TEST: failed, lock not locked\n"); return 1; Index: src/include/storage/s_lock.h =================================================================== RCS file: /var/lib/cvs/pgsql/src/include/storage/s_lock.h,v retrieving revision 1.132 diff -c -r1.132 s_lock.h *** src/include/storage/s_lock.h 6 Oct 2004 23:41:59 -0000 1.132 --- src/include/storage/s_lock.h 13 Oct 2004 06:22:54 -0000 *************** *** 46,55 **** * will "fail" if interrupted. Therefore TAS() should always be invoked * in a retry loop, even if you are certain the lock is free. * ! * ANOTHER CAUTION: be sure that TAS() and S_UNLOCK() represent sequence ! * points, ie, loads and stores of other values must not be moved across ! * a lock or unlock. In most cases it suffices to make the operation be ! * done through a "volatile" pointer. * * On most supported platforms, TAS() uses a tas() function written * in assembly language to execute a hardware atomic-test-and-set --- 46,57 ---- * will "fail" if interrupted. Therefore TAS() should always be invoked * in a retry loop, even if you are certain the lock is free. * ! * ANOTHER CAUTION: be sure that TAS() and S_UNLOCK() represent ! * sequence points, ie, loads and stores of other values must not be ! * moved across a lock or unlock. In most cases it suffices to make ! * the operation be done through a "volatile" pointer; also, adding ! * "memory" to the list of clobbered registers (for GCC inline asm) ! * may also be necessary. * * On most supported platforms, TAS() uses a tas() function written * in assembly language to execute a hardware atomic-test-and-set *************** *** 73,83 **** #ifndef S_LOCK_H #define S_LOCK_H #include "storage/pg_sema.h" #ifdef HAVE_SPINLOCKS /* skip spinlocks if requested */ - #if defined(__GNUC__) || defined(__ICC) /************************************************************************* * All the gcc inlines --- 75,86 ---- #ifndef S_LOCK_H #define S_LOCK_H + #include + #include "storage/pg_sema.h" #ifdef HAVE_SPINLOCKS /* skip spinlocks if requested */ #if defined(__GNUC__) || defined(__ICC) /************************************************************************* * All the gcc inlines *************** *** 107,116 **** *---------- */ ! #if defined(__i386__) || defined(__x86_64__) /* AMD Opteron */ ! #define HAS_TEST_AND_SET typedef unsigned char slock_t; #define TAS(lock) tas(lock) --- 110,322 ---- *---------- */ + /* + * Linux 2.6 introduced the "futex" API. This provides an efficient + * building block for implementing concurrency primitives, via the + * sys_futex system call. The gist is that a futex is an aligned + * integer. It is decremented and incremented via atomic instructions, + * like a normal spinlock. In the contended case, the sys_futex() + * system call allows us to enter the kernel and have it awaken us + * when the futex is free. The best reference for understanding + * futexes is "Futexes are Tricky" by Ulrich Drepper; the + * implementation of spinlocks using mutexes is a fairly close + * transcription of the example code provided in that paper. + * + * If futexes are available at compile-time, we attempt to use them at + * run-time. It is possible that they aren't available (e.g. we're + * compiled on a Linux 2.6 system and run on a Linux 2.4 system), so + * we fallback to using normal TAS in that case. + */ + #if defined(__linux__) ! #include ! ! #ifdef SYS_futex ! /* ! * Additional atomic operations are required to use futexes. We ! * consider futexes to be available on a system iff the sys_futex ! * system call is available AND we have implemented the necessary ! * assembly routines for the CPU architecture. If both conditions do ! * not hold, we fall back to using the normal TAS technique. ! */ ! #if defined(__i386__) || defined(__x86_64__) ! #define HAVE_FUTEX ! ! typedef int slock_t; ! ! static __inline__ slock_t ! atomic_cmp_exchange(volatile slock_t *ptr, slock_t old, slock_t new) ! { ! slock_t prev; ! __asm__ __volatile__( ! " lock \n" ! " cmpxchgl %1,%2 \n" ! : "=a"(prev) ! : "q"(new), "m"(*ptr), "0"(old) ! : "memory", "cc"); ! ! return prev; ! } ! ! static __inline__ slock_t ! atomic_exchange(volatile slock_t *ptr, slock_t x) ! { ! /* no lock prefix needed, xchgl is always locked */ ! __asm__ __volatile__( ! " xchgl %0,%1 \n" ! :"=q"(x) ! :"m"(*ptr), "0"(x) ! :"memory"); ! ! return x; ! } ! ! /* ! * XXX: is there a more efficient way to write this? Perhaps using ! * decl...? ! */ ! static __inline__ slock_t ! atomic_dec(volatile slock_t *ptr) ! { ! slock_t prev = -1; ! ! __asm__ __volatile__( ! " lock \n" ! " xadd %0,%1 \n" ! :"=q"(prev) ! :"m"(*ptr), "0"(prev) ! :"memory", "cc"); ! ! return prev; ! } ! #endif /* defined(__i386__) || defined(__x86_64__) */ ! ! #if defined(__ia64__) || defined(__ia64) ! /* Intel Itanium */ ! #define HAVE_FUTEX ! ! typedef int slock_t; ! ! static __inline__ slock_t ! atomic_cmp_exchange(volatile slock_t *ptr, slock_t old, slock_t new) ! { ! long int prev; ! ! __asm__ __volatile__( ! " mov ar.ccv=%2;;" ! " cmpxchg4.acq %0 = %4,%3,ar.ccv" ! : "=r" (prev), "=m" (*ptr) ! : "r" (old), "r" (new), "m" (*ptr) ! : "memory"); ! ! return (slock_t) prev; ! } ! ! static __inline__ slock_t ! atomic_exchange(volatile slock_t *ptr, slock_t x) ! { ! /* Note that xchg has "acquire" semantics implicitely */ ! __asm__ __volatile__( ! " xchg4 %0=%1,%2 \n" ! : "+r"(x), "+m"(*ptr) ! : ! : "memory"); ! ! return x; ! } ! ! static __inline__ slock_t ! atomic_dec(volatile slock_t *ptr) ! { ! slock_t result; ! ! /* ! * Note that IA64 does _not_ guarantee that fetchadd is always ! * atomic. However, it should always be atomic when the operand is ! * -1; see the IA64 docs for details. ! */ ! __asm__ __volatile__( ! "fetchadd4.rel %0=[%1],%2" ! : "=r"(result) ! : "r"(ptr), "i"(-1) ! : "memory"); ! ! return result; ! } ! ! #endif /* __ia64__ || __ia64 */ ! ! #endif /* SYS_futex */ ! #endif /* defined(__linux__) */ ! ! #ifdef HAVE_FUTEX ! ! /* ! * Legal states of a futex. ! */ ! #define FUTEX_UNLOCKED 0 ! #define FUTEX_LOCKED_NOWAITER 1 ! #define FUTEX_LOCKED_WAITER 2 ! ! /* ! * Futex operations (for sys_futex()). ! */ ! #define FUTEX_OP_WAIT 0 ! #define FUTEX_OP_WAKE 1 ! ! #define S_LOCK(lock) futex_lock((lock), __FILE__, __LINE__) ! #define S_UNLOCK(lock) futex_unlock(lock) ! #define S_LOCK_FREE(lock) (*(lock) == FUTEX_UNLOCKED) ! #define S_INIT_LOCK(lock) (*(lock) = FUTEX_UNLOCKED) ! #define SPIN_DELAY() ! ! #define NUM_DELAYS 100 + extern void + futex_lock_contended(volatile slock_t *lock, slock_t val, + const char *file, int line); + + static __inline__ int + sys_futex(volatile int *ptr, int op, int val, struct timespec *timeout) + { + return syscall(SYS_futex, ptr, op, val, timeout); + } + + static __inline__ void + futex_lock(volatile slock_t *lock, const char *file, int line) + { + /* + * In the common case, there is no contention for the lock, so we + * can acquire it with a single compare-exchange instruction. Thus, + * it is important to avoid the overhead of a function call. If we + * need to contend for the futex we need to spin or make a kernel + * call anyway, so a function call doesn't hurt. + */ + slock_t val = atomic_cmp_exchange(lock, + FUTEX_UNLOCKED, + FUTEX_LOCKED_NOWAITER); + + if (val != FUTEX_UNLOCKED) + futex_lock_contended(lock, val, file, line); + } + + static __inline__ void + futex_unlock(volatile slock_t *lock) + { + Assert(*lock == FUTEX_LOCKED_NOWAITER || + *lock == FUTEX_LOCKED_WAITER); + if (atomic_dec(lock) != FUTEX_LOCKED_NOWAITER) + { + *lock = FUTEX_UNLOCKED; + sys_futex(lock, FUTEX_OP_WAKE, 1, NULL); + } + } + + #else + + #if defined(__i386__) || defined(__x86_64__) + + #define HAS_TEST_AND_SET typedef unsigned char slock_t; #define TAS(lock) tas(lock) *************** *** 120,130 **** { register slock_t _res = 1; ! /* Use a non-locking test before asserting the bus lock */ __asm__ __volatile__( " cmpb $0,%1 \n" " jne 1f \n" - " lock \n" " xchgb %0,%1 \n" "1: \n" : "+q"(_res), "+m"(*lock) --- 326,336 ---- { register slock_t _res = 1; ! /* Use a non-locking test before doing xchgb, which implicitely ! * asserts the bus lock */ __asm__ __volatile__( " cmpb $0,%1 \n" " jne 1f \n" " xchgb %0,%1 \n" "1: \n" : "+q"(_res), "+m"(*lock) *************** *** 167,173 **** #endif /* __i386__ || __x86_64__ */ - #if defined(__ia64__) || defined(__ia64) /* __ia64 used by ICC compiler? */ /* Intel Itanium */ #define HAS_TEST_AND_SET --- 373,378 ---- *************** *** 453,459 **** typedef unsigned int slock_t; #endif ! #endif /* __GNUC__ */ --- 658,664 ---- typedef unsigned int slock_t; #endif ! #endif /* ! HAVE_FUTEX */ #endif /* __GNUC__ */ *************** *** 697,703 **** /* Blow up if we didn't have any way to do spinlocks */ ! #ifndef HAS_TEST_AND_SET #error PostgreSQL does not have native spinlock support on this platform. To continue the compilation, rerun configure using --disable-spinlocks. However, performance will be poor. Please report this to pgsql-bugs@postgresql.org. #endif --- 902,908 ---- /* Blow up if we didn't have any way to do spinlocks */ ! #if !defined(HAS_TEST_AND_SET) && !defined(HAVE_FUTEX) #error PostgreSQL does not have native spinlock support on this platform. To continue the compilation, rerun configure using --disable-spinlocks. However, performance will be poor. Please report this to pgsql-bugs@postgresql.org. #endif --=-HuItn3AfDAyHfOdEZ/QL-- From pgsql-performance-owner@postgresql.org Thu Oct 14 04:47:31 2004 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 D813D32C6D4 for ; Thu, 14 Oct 2004 04:47:29 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 34447-03 for ; Thu, 14 Oct 2004 03:47:16 +0000 (GMT) Received: from candle.pha.pa.us (candle.pha.pa.us [207.106.42.251]) by svr1.postgresql.org (Postfix) with ESMTP id 8CF8932C59B for ; Thu, 14 Oct 2004 04:47:13 +0100 (BST) Received: (from pgman@localhost) by candle.pha.pa.us (8.11.6/8.11.6) id i9E3l5910978; Wed, 13 Oct 2004 23:47:05 -0400 (EDT) From: Bruce Momjian Message-Id: <200410140347.i9E3l5910978@candle.pha.pa.us> Subject: Re: Free PostgreSQL Training, Philadelphia, Oct 30 In-Reply-To: <200410131536.33286.josh@agliodbs.com> To: josh@agliodbs.com Date: Wed, 13 Oct 2004 23:47:05 -0400 (EDT) Cc: pgsql-performance@postgresql.org, Aaron Glenn X-Mailer: ELM [version 2.4ME+ PL108 (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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/187 X-Sequence-Number: 8663 Josh Berkus wrote: > Aaron, > > > That makes two of us. Hanging out with Tom, Bruce, and others at OSCON > > 2002 was one of the most informative and fun times I've had. That and > > I could really stand to brush up on my Postgres basics > > You're thinking of Jan. Tom wasn't at OSCON. ;-) Ah, but he said 2002 and I think Tom was there that year. -- 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 Thu Oct 14 04:52:46 2004 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 33E6A32C6E1 for ; Thu, 14 Oct 2004 04:52:44 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 34642-05 for ; Thu, 14 Oct 2004 03:52:40 +0000 (GMT) Received: from stark.xeocode.com (gsstark.mtl.istop.com [66.11.160.162]) by svr1.postgresql.org (Postfix) with ESMTP id BB7D832C6DE for ; Thu, 14 Oct 2004 04:52:40 +0100 (BST) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1CHwfD-0005tv-00; Wed, 13 Oct 2004 23:52:27 -0400 To: Jan Wieck Cc: Christopher Browne , pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some References: <200410081443.16109.josh@agliodbs.com> <416DCCFF.70304@Yahoo.com> In-Reply-To: <416DCCFF.70304@Yahoo.com> From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 13 Oct 2004 23:52:27 -0400 Message-ID: <871xg25538.fsf@stark.xeocode.com> Lines: 34 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/188 X-Sequence-Number: 8664 Jan Wieck writes: > On 10/8/2004 10:10 PM, Christopher Browne wrote: > > > josh@agliodbs.com (Josh Berkus) wrote: > >> I've been trying to peg the "sweet spot" for shared memory using > >> OSDL's equipment. With Jan's new ARC patch, I was expecting that > >> the desired amount of shared_buffers to be greatly increased. This > >> has not turned out to be the case. > > That doesn't surprise me. > > Neither does it surprise me. There's been some speculation that having a large shared buffers be about 50% of your RAM is pessimal as it guarantees the OS cache is merely doubling up on all the buffers postgres is keeping. I wonder whether there's a second sweet spot where the postgres cache is closer to the total amount of RAM. That configuration would have disadvantages for servers running other jobs besides postgres. And I was led to believe earlier that postgres starts each backend with a fairly fresh slate as far as the ARC algorithm, so it wouldn't work well for a postgres server that had lots of short to moderate life sessions. But if it were even close it could be interesting. Reading the data with O_DIRECT and having a single global cache could be interesting experiments. I know there are arguments against each of these, but ... I'm still pulling for an mmap approach to eliminate postgres's buffer cache entirely in the long term, but it seems like slim odds now. But one way or the other having two layers of buffering seems like a waste. -- greg From pgsql-performance-owner@postgresql.org Thu Oct 14 05:17:56 2004 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 81E8A32C59B for ; Thu, 14 Oct 2004 05:17:51 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 42297-04 for ; Thu, 14 Oct 2004 04:17:42 +0000 (GMT) Received: from smtp108.mail.sc5.yahoo.com (smtp108.mail.sc5.yahoo.com [66.163.170.6]) by svr1.postgresql.org (Postfix) with SMTP id D2F9E32A60E for ; Thu, 14 Oct 2004 05:17:41 +0100 (BST) Received: from unknown (HELO jupiter.black-lion.info) (janwieck@68.80.245.191 with login) by smtp108.mail.sc5.yahoo.com with SMTP; 14 Oct 2004 04:17:40 -0000 Received: from [172.21.8.18] ([192.168.192.101]) (authenticated bits=0) by jupiter.black-lion.info (8.12.10/8.12.9) with ESMTP id i9E4HX94074613; Thu, 14 Oct 2004 00:17:34 -0400 (EDT) (envelope-from JanWieck@Yahoo.com) Message-ID: <416DFDDE.5070600@Yahoo.com> Date: Thu, 14 Oct 2004 00:17:34 -0400 From: Jan Wieck User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.1) Gecko/20040707 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Greg Stark Cc: Christopher Browne , pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some References: <200410081443.16109.josh@agliodbs.com> <416DCCFF.70304@Yahoo.com> <871xg25538.fsf@stark.xeocode.com> In-Reply-To: <871xg25538.fsf@stark.xeocode.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/189 X-Sequence-Number: 8665 On 10/13/2004 11:52 PM, Greg Stark wrote: > Jan Wieck writes: > >> On 10/8/2004 10:10 PM, Christopher Browne wrote: >> >> > josh@agliodbs.com (Josh Berkus) wrote: >> >> I've been trying to peg the "sweet spot" for shared memory using >> >> OSDL's equipment. With Jan's new ARC patch, I was expecting that >> >> the desired amount of shared_buffers to be greatly increased. This >> >> has not turned out to be the case. >> > That doesn't surprise me. >> >> Neither does it surprise me. > > There's been some speculation that having a large shared buffers be about 50% > of your RAM is pessimal as it guarantees the OS cache is merely doubling up on > all the buffers postgres is keeping. I wonder whether there's a second sweet > spot where the postgres cache is closer to the total amount of RAM. Which would require that shared memory is not allowed to be swapped out, and that is allowed in Linux by default IIRC, not to completely distort the entire test. Jan -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com # From pgsql-performance-owner@postgresql.org Thu Oct 14 05:22:55 2004 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 E7F1732A43F for ; Thu, 14 Oct 2004 05:22:53 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 42515-05 for ; Thu, 14 Oct 2004 04:22:44 +0000 (GMT) Received: from stark.xeocode.com (gsstark.mtl.istop.com [66.11.160.162]) by svr1.postgresql.org (Postfix) with ESMTP id 0347132BEED for ; Thu, 14 Oct 2004 05:22:44 +0100 (BST) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1CHx8T-0005zx-00; Thu, 14 Oct 2004 00:22:41 -0400 To: Jan Wieck Cc: Greg Stark , Christopher Browne , pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some References: <200410081443.16109.josh@agliodbs.com> <416DCCFF.70304@Yahoo.com> <871xg25538.fsf@stark.xeocode.com> <416DFDDE.5070600@Yahoo.com> In-Reply-To: <416DFDDE.5070600@Yahoo.com> From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 14 Oct 2004 00:22:40 -0400 Message-ID: <87sm8i3p4f.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/190 X-Sequence-Number: 8666 Jan Wieck writes: > Which would require that shared memory is not allowed to be swapped out, and > that is allowed in Linux by default IIRC, not to completely distort the entire > test. Well if it's getting swapped out then it's clearly not being used effectively. There are APIs to bar swapping out pages and the tests could be run without swap. I suggested it only as an experiment though, there are lots of details between here and having it be a good configuration for production use. -- greg From pgsql-performance-owner@postgresql.org Thu Oct 14 05:29:26 2004 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 72BB532BD7F for ; Thu, 14 Oct 2004 05:29:24 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 44858-04 for ; Thu, 14 Oct 2004 04:29:22 +0000 (GMT) Received: from smtp111.mail.sc5.yahoo.com (smtp111.mail.sc5.yahoo.com [66.163.170.9]) by svr1.postgresql.org (Postfix) with SMTP id 680A732A9C1 for ; Thu, 14 Oct 2004 05:29:21 +0100 (BST) Received: from unknown (HELO jupiter.black-lion.info) (janwieck@68.80.245.191 with login) by smtp111.mail.sc5.yahoo.com with SMTP; 14 Oct 2004 04:29:20 -0000 Received: from [172.21.8.18] ([192.168.192.101]) (authenticated bits=0) by jupiter.black-lion.info (8.12.10/8.12.9) with ESMTP id i9E4TI94074678; Thu, 14 Oct 2004 00:29:19 -0400 (EDT) (envelope-from JanWieck@Yahoo.com) Message-ID: <416E009F.3010709@Yahoo.com> Date: Thu, 14 Oct 2004 00:29:19 -0400 From: Jan Wieck User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.1) Gecko/20040707 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Greg Stark Cc: Christopher Browne , pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some References: <200410081443.16109.josh@agliodbs.com> <416DCCFF.70304@Yahoo.com> <871xg25538.fsf@stark.xeocode.com> <416DFDDE.5070600@Yahoo.com> <87sm8i3p4f.fsf@stark.xeocode.com> In-Reply-To: <87sm8i3p4f.fsf@stark.xeocode.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/191 X-Sequence-Number: 8667 On 10/14/2004 12:22 AM, Greg Stark wrote: > Jan Wieck writes: > >> Which would require that shared memory is not allowed to be swapped out, and >> that is allowed in Linux by default IIRC, not to completely distort the entire >> test. > > Well if it's getting swapped out then it's clearly not being used effectively. Is it really that easy if 3 different cache algorithms (PG cache, kernel buffers and swapping) are competing for the same chips? Jan > > There are APIs to bar swapping out pages and the tests could be run without > swap. I suggested it only as an experiment though, there are lots of details > between here and having it be a good configuration for production use. > -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com # From pgsql-hackers-win32-owner@postgresql.org Thu Oct 14 18:06:23 2004 X-Original-To: pgsql-hackers-win32-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 05CE332A4FC; Thu, 14 Oct 2004 18:02:06 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 76102-09; Thu, 14 Oct 2004 17:02:03 +0000 (GMT) Received: from hotmail.com (bay10-dav14.bay10.hotmail.com [64.4.37.188]) by svr1.postgresql.org (Postfix) with ESMTP id 79BAA32A273; Thu, 14 Oct 2004 18:02:02 +0100 (BST) Received: from mail pickup service by hotmail.com with Microsoft SMTPSVC; Thu, 14 Oct 2004 10:02:01 -0700 Received: from 199.249.167.148 by BAY10-DAV14.phx.gbl with DAV; Thu, 14 Oct 2004 17:01:43 +0000 X-Originating-IP: [199.249.167.148] X-Originating-Email: [mikesmialek2@hotmail.com] X-Sender: mikesmialek2@hotmail.com From: "MikeSmialek2@Hotmail.com" To: , Subject: Performance on Win32 vs Cygwin Date: Thu, 14 Oct 2004 12:01:38 -0500 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 6.00.2800.1437 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 Message-ID: X-OriginalArrivalTime: 14 Oct 2004 17:02:01.0458 (UTC) FILETIME=[85F02D20:01C4B20F] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/64 X-Sequence-Number: 2354 Hi, We are experiencing slow performance on 8 Beta 2 Dev3 on Win32 and are trying to determine why. Any info is appreciated. We have a Web Server and a DB server both running Win2KServer with all service packs and critical updates. An ASP page on the Web Server hits the DB Server with a simple query that returns 205 rows and makes the ASP page delivered to the user about 350K. On an ethernet lan a client pc perceives just under 1 sec performance with the following DB Server configuration: PIII 550Mhz 256MB RAM 7200 RPM HD cygwin Postgresql 7.1.3 PGODBC 7.3.2 We set up another DB Server with 8 beta (same Web Server, same network, same client pc) and now the client pc perceives response of just over 3 sec with the following DB server config: PIII 700 Mhz 448MB RAM 7200 RPM HD 8 Beta 2 Dev3 on Win32 running as a service Is the speed decrease because it's a beta? Is the speed decrease because it's running on Win instead of cygwin? We did not install cygwin on the new DB Server. Thanks, Mike From pgsql-hackers-win32-owner@postgresql.org Thu Oct 14 18:29:46 2004 X-Original-To: pgsql-hackers-win32-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CC97B32A821; Thu, 14 Oct 2004 18:29:41 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 90314-01; Thu, 14 Oct 2004 17:29:29 +0000 (GMT) Received: from mx-2.sollentuna.net (mx-2.sollentuna.net [195.84.163.199]) by svr1.postgresql.org (Postfix) with ESMTP id E02DE32A084; Thu, 14 Oct 2004 18:29:28 +0100 (BST) Received: from ALGOL.sollentuna.se (janus.sollentuna.se [62.65.68.67]) by mx-2.sollentuna.net (Postfix) with ESMTP id 3A50A8F38F; Thu, 14 Oct 2004 19:29:26 +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 on Win32 vs Cygwin X-MimeOLE: Produced By Microsoft Exchange V6.0.6249.0 Date: Thu, 14 Oct 2004 19:29:25 +0200 Message-ID: <6BCB9D8A16AC4241919521715F4D8BCE475EE5@algol.sollentuna.se> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [pgsql-hackers-win32] Performance on Win32 vs Cygwin Thread-Index: AcSyEDdpszD9R4IDSBicK3E0DhHuuAAAu0PA From: "Magnus Hagander" To: "MikeSmialek2@Hotmail.com" , , X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/65 X-Sequence-Number: 2355 >Hi, > >We are experiencing slow performance on 8 Beta 2 Dev3 on Win32 and are >trying to determine why. Any info is appreciated. > >We have a Web Server and a DB server both running Win2KServer with all >service packs and critical updates. > >An ASP page on the Web Server hits the DB Server with a simple=20 >query that >returns 205 rows and makes the ASP page delivered to the user=20 >about 350K. > >On an ethernet lan a client pc perceives just under 1 sec=20 >performance with >the following DB Server configuration: > PIII 550Mhz > 256MB RAM > 7200 RPM HD > cygwin > Postgresql 7.1.3 > PGODBC 7.3.2 > >We set up another DB Server with 8 beta (same Web Server, same=20 >network, same >client pc) and now the client pc perceives response of just=20 >over 3 sec with >the following DB server config: > PIII 700 Mhz > 448MB RAM > 7200 RPM HD > 8 Beta 2 Dev3 on Win32 running as a service > >Is the speed decrease because it's a beta? >Is the speed decrease because it's running on Win instead of cygwin? > >We did not install cygwin on the new DB Server. IIRC, previous versions of postgresql (< 8.0) did not correctly sync disks when running on Cygwin. I'm not 100% sure, can someone confirm? 8.0 does, and I beleive it does both under native win32 and cygwin. It's been my experience that the native version is slightly faster than the cygwin one, but I've only compared 8.0 to 8.0. //Magnus From pgsql-performance-owner@postgresql.org Thu Oct 14 19:33:27 2004 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 013A332B1B3 for ; Thu, 14 Oct 2004 19:33:25 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 09484-07 for ; Thu, 14 Oct 2004 18:33:21 +0000 (GMT) Received: from providerst.com.br (unknown [200.253.138.61]) by svr1.postgresql.org (Postfix) with ESMTP id 93E4832B1AA for ; Thu, 14 Oct 2004 19:33:21 +0100 (BST) Brma: igor [192.168.1.104] Received: from sistemas (igor [192.168.1.104]) by providerst.com.br (8.11.6/8.11.6) with SMTP id i9EIQGp11506 for ; Thu, 14 Oct 2004 16:26:16 -0200 Message-ID: <003101c4b21d$10f00e60$6801a8c0@providerst.local> From: "Igor Maciel Macaubas" To: Subject: Performance vs Schemas Date: Thu, 14 Oct 2004 15:38:52 -0300 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_002E_01C4B203.E88590D0" 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=0.9 tagged_above=0.0 required=5.0 tests=HTML_30_40, HTML_FONTCOLOR_BLUE, HTML_MESSAGE X-Spam-Level: X-Archive-Number: 200410/193 X-Sequence-Number: 8669 This is a multi-part message in MIME format. ------=_NextPart_000_002E_01C4B203.E88590D0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Hi all, I recently migrated my database from schema 'public' to multiple schema. I have around 100 tables, and divided them in 14 different schemas, and the= n adapted my application to use schemas as well. I could percept that the query / insert / update times get pretty much fast= er then when I was using the old unique schema, and I'd just like to confir= m with you if using schemas speed up the things. Is that true ? What else I can do to speed up the query processing, best pratices, recomme= ndations ... ? What about indexed views, does postgresql supports it? Regards, Igor -- igor@providerst.com.br ------=_NextPart_000_002E_01C4B203.E88590D0 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Hi all,
 
I recently migrated my database from sc= hema=20 'public' to multiple schema.
I have around 100 tables, and divided t= hem in=20 14 different schemas, and then adapted my application to use schemas a= s=20 well.
I could percept that the query / insert=  /=20 update times get pretty much faster then when I was using the old unique sc= hema,=20 and I'd just like to confirm with you if using schemas speed up the things.= Is=20 that true ?
 
What else I can do to speed up the quer= y=20 processing, best pratices, recommendations ... ? What about indexed views, = does=20 postgresql supports it?
 
Regards,
Igor
--
igor@providerst.com.br
 
------=_NextPart_000_002E_01C4B203.E88590D0-- From pgsql-performance-owner@postgresql.org Thu Oct 14 19:45:24 2004 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 7C0BD32A589 for ; Thu, 14 Oct 2004 19:45:22 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 14773-04 for ; Thu, 14 Oct 2004 18:45:14 +0000 (GMT) Received: from loki.globexplorer.com (unknown [208.35.14.101]) by svr1.postgresql.org (Postfix) with ESMTP id BF20732A200 for ; Thu, 14 Oct 2004 19:45:11 +0100 (BST) 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="Windows-1252" Content-Transfer-Encoding: quoted-printable Subject: Re: Performance vs Schemas Date: Thu, 14 Oct 2004 11:45:10 -0700 Message-ID: <71E37EF6B7DCC1499CEA0316A256832801D4B992@loki.wc.globexplorer.net> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Performance vs Schemas Thread-Index: AcSyHJ3fb3RT00EXRvSCzjKrxBWKgwAAGqen From: "Gregory S. Williamson" To: "Igor Maciel Macaubas" , X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/194 X-Sequence-Number: 8670 Igor, I'm not sure if it is proper to state that schemas are themselves speeding = things up. As an example, we have data that is usually accessed by county; when we put= all of the data into one big table and select from it using a code for a c= ounty of interest, the process is fairly slow as there are several hundred = thousand candidate rows from that county in a table with many millions of r= ows. When we broke out certain aspects of the data into schemas (one per co= unty) the searches become very fast indeed because we can skip the searchin= g for a specific county code with the relevant tables and there is less (un= needed) data in the table being searched.=20=20 As always, "EXPLAIN ANALYZE ..." is your friend in understanding what the p= lanner is doing with a given query. See for some useful i= nformation, especially under the performance tips section. HTH, Greg Williamson DBA GlobeXplorer LLC -----Original Message----- From: Igor Maciel Macaubas [mailto:igor@providerst.com.br] Sent: Thu 10/14/2004 11:38 AM To: pgsql-performance@postgresql.org Cc:=09 Subject: [PERFORM] Performance vs Schemas Hi all, I recently migrated my database from schema 'public' to multiple schema. I have around 100 tables, and divided them in 14 different schemas, and the= n adapted my application to use schemas as well. I could percept that the query / insert / update times get pretty much fast= er then when I was using the old unique schema, and I'd just like to confir= m with you if using schemas speed up the things. Is that true ? What else I can do to speed up the query processing, best pratices, recomme= ndations ... ? What about indexed views, does postgresql supports it? Regards, Igor -- igor@providerst.com.br From pgsql-performance-owner@postgresql.org Tue Oct 19 22:03:49 2004 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 1B15F32B42D for ; Thu, 14 Oct 2004 20:28:17 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 28139-03 for ; Thu, 14 Oct 2004 19:28:02 +0000 (GMT) Received: from hotmail.com (bay9-dav25.bay9.hotmail.com [64.4.46.82]) by svr1.postgresql.org (Postfix) with ESMTP id 6837A32B3F1 for ; Thu, 14 Oct 2004 20:28:02 +0100 (BST) Received: from mail pickup service by hotmail.com with Microsoft SMTPSVC; Thu, 14 Oct 2004 12:28:01 -0700 Received: from 67.81.98.198 by bay9-dav25.bay9.hotmail.com with DAV; Thu, 14 Oct 2004 19:27:36 +0000 X-Originating-IP: [67.81.98.198] X-Originating-Email: [awerman@hotmail.com] X-Sender: awerman@hotmail.com From: "Aaron Werman" To: "Gregory S. Williamson" , "Igor Maciel Macaubas" , References: <71E37EF6B7DCC1499CEA0316A256832801D4B992@loki.wc.globexplorer.net> Subject: Re: Performance vs Schemas Date: Thu, 14 Oct 2004 15:27:51 -0400 MIME-Version: 1.0 Content-Type: text/plain; charset="Windows-1252" Content-Transfer-Encoding: 7bit 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 Message-ID: X-OriginalArrivalTime: 14 Oct 2004 19:28:01.0699 (UTC) FILETIME=[EB72DF30:01C4B223] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/289 X-Sequence-Number: 8765 Right - if you split a table to a lot of more selective tables, it can often dramatically change the plan options (e.g. - in a single table, selectivity for a query may be 1% and require an expensive nested loop while in the more restrictive table it may match 14% of the data and do a cheaper scan). Also - don't forget that just rebuilding a database cleanly can dramatically improve performance. The only dbms I know that indexes views is MS SQL Server 2000, where it is a limited form of materialized queries. pg doesn't do MQs, but check out functional indices. /Aaron ----- Original Message ----- From: "Gregory S. Williamson" To: "Igor Maciel Macaubas" ; Sent: Thursday, October 14, 2004 2:45 PM Subject: Re: [PERFORM] Performance vs Schemas Igor, I'm not sure if it is proper to state that schemas are themselves speeding things up. As an example, we have data that is usually accessed by county; when we put all of the data into one big table and select from it using a code for a county of interest, the process is fairly slow as there are several hundred thousand candidate rows from that county in a table with many millions of rows. When we broke out certain aspects of the data into schemas (one per county) the searches become very fast indeed because we can skip the searching for a specific county code with the relevant tables and there is less (unneeded) data in the table being searched. As always, "EXPLAIN ANALYZE ..." is your friend in understanding what the planner is doing with a given query. See for some useful information, especially under the performance tips section. HTH, Greg Williamson DBA GlobeXplorer LLC -----Original Message----- From: Igor Maciel Macaubas [mailto:igor@providerst.com.br] Sent: Thu 10/14/2004 11:38 AM To: pgsql-performance@postgresql.org Cc: Subject: [PERFORM] Performance vs Schemas Hi all, I recently migrated my database from schema 'public' to multiple schema. I have around 100 tables, and divided them in 14 different schemas, and then adapted my application to use schemas as well. I could percept that the query / insert / update times get pretty much faster then when I was using the old unique schema, and I'd just like to confirm with you if using schemas speed up the things. Is that true ? What else I can do to speed up the query processing, best pratices, recommendations ... ? What about indexed views, does postgresql supports it? Regards, Igor -- igor@providerst.com.br ---------------------------(end of broadcast)--------------------------- TIP 6: Have you searched our list archives? http://archives.postgresql.org From pgsql-hackers-win32-owner@postgresql.org Thu Oct 14 20:41:21 2004 X-Original-To: pgsql-hackers-win32-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8425732B2F2; Thu, 14 Oct 2004 20:41:13 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 32120-03; Thu, 14 Oct 2004 19:41:02 +0000 (GMT) Received: from smartmx-06.inode.at (smartmx-06.inode.at [213.229.60.38]) by svr1.postgresql.org (Postfix) with ESMTP id 5B17732B2B4; Thu, 14 Oct 2004 20:41:02 +0100 (BST) Received: from [62.99.252.218] (port=61998 helo=[192.168.0.2]) by smartmx-06.inode.at with esmtp (Exim 4.30) id 1CIBTB-0001OO-99; Thu, 14 Oct 2004 21:41:01 +0200 Message-ID: <416ED64B.3070000@x-ray.at> Date: Thu, 14 Oct 2004 21:40:59 +0200 From: Reini Urban User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; de-AT; rv:1.8a3) Gecko/20040817 X-Accept-Language: de, en MIME-Version: 1.0 To: pgsql-hackers-win32@postgresql.org Cc: pgsql-performance@postgresql.org Subject: Re: Performance on Win32 vs Cygwin References: <6BCB9D8A16AC4241919521715F4D8BCE475EE5@algol.sollentuna.se> In-Reply-To: <6BCB9D8A16AC4241919521715F4D8BCE475EE5@algol.sollentuna.se> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/67 X-Sequence-Number: 2357 Magnus Hagander schrieb: > IIRC, previous versions of postgresql (< 8.0) did not correctly sync > disks when running on Cygwin. I'm not 100% sure, can someone confirm? > 8.0 does, and I beleive it does both under native win32 and cygwin. yes, sync is a NOOP on cygwin. > It's been my experience that the native version is slightly faster than > the cygwin one, but I've only compared 8.0 to 8.0. Sure. This is expected. Cygwin's interim's layer costs a lot of time. (process handling, path resolution) -- Reini Urban http://xarch.tu-graz.ac.at/home/rurban/ From pgsql-performance-owner@postgresql.org Thu Oct 14 20:59:17 2004 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 8BECF32BA38 for ; Thu, 14 Oct 2004 20:59:11 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 37592-10 for ; Thu, 14 Oct 2004 19:58:58 +0000 (GMT) Received: from net2.micro-automation.com (net2.micro-automation.com [64.7.141.29]) by svr1.postgresql.org (Postfix) with SMTP id F0C3F32AE2C for ; Thu, 14 Oct 2004 20:58:58 +0100 (BST) Received: (qmail 17950 invoked from network); 14 Oct 2004 20:00:23 -0000 Received: from dcdsl.ebox.com (HELO ?192.168.1.36?) (davec@64.7.143.116) by net2.micro-automation.com with SMTP; 14 Oct 2004 20:00:23 -0000 Subject: Re: Excessive context switching on SMP Xeons From: Dave Cramer Reply-To: pg@fastcrypt.com To: Bill Montgomery Cc: Michael Adler , perform In-Reply-To: <41658F8D.9090204@lulu.com> References: <4162CA14.6080206@lulu.com> <200410050947.36174.josh@agliodbs.com> <41630D50.3020308@lulu.com> <200410051538.51664.josh@agliodbs.com> <41636D8E.1090403@rentec.com> <41648446.5020102@bigfoot.com> <4164B48C.9060202@rentec.com> <41656559.70100@lulu.com> <20041007181537.GA1693@pobox.com> <41658F8D.9090204@lulu.com> Content-Type: text/plain Organization: Cramer Consulting Message-Id: <1097783954.7978.501.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Thu, 14 Oct 2004 15:59:14 -0400 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/195 X-Sequence-Number: 8671 Bill, In order to manifest the context switch problem you will definitely require clients to be set to more than one in pgbench. It only occurs when 2 or more backends need access to shared memory. If you want help backpatching Gavin's patch I'll be glad to do it for you, but you do need a recent kernel. Dave On Thu, 2004-10-07 at 14:48, Bill Montgomery wrote: > Michael Adler wrote: > > >On Thu, Oct 07, 2004 at 11:48:41AM -0400, Bill Montgomery wrote: > > > > > >>Alan Stange wrote: > >> > >>The same test on a Dell PowerEdge 1750, Dual Xeon 3.2 GHz, 512k cache, > >>HT on, Linux 2.4.21-20.ELsmp (RHEL 3), 4GB memory, pg 7.4.5: > >> > >>Far less performance that the Dual Opterons with a low number of > >>clients, but the gap narrows as the number of clients goes up. Anyone > >>smarter than me care to explain? > >> > >> > > > >You'll have to wait for someone smarter than you, but I will posit > >this: Did you use a tmpfs filesystem like Alan? You didn't mention > >either way. Alan did that as an attempt remove IO as a variable. > > > >-Mike > > > > > > Yes, I should have been more explicit. My goal was to replicate his > experiment as closely as possible in my environment, so I did run my > postgres data directory on a tmpfs. > > -Bill Montgomery > > ---------------------------(end of broadcast)--------------------------- > TIP 4: Don't 'kill -9' the postmaster -- Dave Cramer 519 939 0336 ICQ # 14675561 www.postgresintl.com From pgsql-performance-owner@postgresql.org Thu Oct 14 21:26:11 2004 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 D9E0D32C3D2 for ; Thu, 14 Oct 2004 21:25:45 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 45862-04 for ; Thu, 14 Oct 2004 20:25:34 +0000 (GMT) Received: from sccrmhc12.comcast.net (sccrmhc12.comcast.net [204.127.202.56]) by svr1.postgresql.org (Postfix) with ESMTP id 2204532BB53 for ; Thu, 14 Oct 2004 21:25:34 +0100 (BST) Received: from sysexperts.com (c-24-6-183-218.client.comcast.net[24.6.183.218]) by comcast.net (sccrmhc12) with ESMTP id <2004101420253201200fshgme>; Thu, 14 Oct 2004 20:25:34 +0000 Received: from localhost (localhost [127.0.0.1]) (uid 1000) by filer with local; Thu, 14 Oct 2004 13:25:31 -0700 id 000063D2.416EE0BB.00000CC1 Date: Thu, 14 Oct 2004 13:25:31 -0700 From: Kevin Brown To: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... Message-ID: <20041014202531.GD665@filer> Mail-Followup-To: Kevin Brown , pgsql-performance@postgresql.org References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> <4859.1097363137@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline In-Reply-To: <4859.1097363137@sss.pgh.pa.us> Organization: Frobozzco International User-Agent: Mutt/1.5.6+20040818i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/196 X-Sequence-Number: 8672 Tom Lane wrote: > Kevin Brown writes: > > Tom Lane wrote: > >> mmap() is Right Out because it does not afford us sufficient control > >> over when changes to the in-memory data will propagate to disk. > > > ... that's especially true if we simply cannot > > have the page written to disk in a partially-modified state (something > > I can easily see being an issue for the WAL -- would the same hold > > true of the index/data files?). > > You're almost there. Remember the fundamental WAL rule: log entries > must hit disk before the data changes they describe. That means that we > need not only a way of forcing changes to disk (fsync) but a way of > being sure that changes have *not* gone to disk yet. In the existing > implementation we get that by just not issuing write() for a given page > until we know that the relevant WAL log entries are fsync'd down to > disk. (BTW, this is what the LSN field on every page is for: it tells > the buffer manager the latest WAL offset that has to be flushed before > it can safely write the page.) > > mmap provides msync which is comparable to fsync, but AFAICS it > provides no way to prevent an in-memory change from reaching disk too > soon. This would mean that WAL entries would have to be written *and > flushed* before we could make the data change at all, which would > convert multiple updates of a single page into a series of write-and- > wait-for-WAL-fsync steps. Not good. fsync'ing WAL once per transaction > is bad enough, once per atomic action is intolerable. Hmm...something just occurred to me about this. Would a hybrid approach be possible? That is, use mmap() to handle reads, and use write() to handle writes? Any code that wishes to write to a page would have to recognize that it's doing so and fetch a copy from the storage manager (or something), which would look to see if the page already exists as a writeable buffer. If it doesn't, it creates it by allocating the memory and then copying the page from the mmap()ed area to the new buffer, and returning it. If it does, it just returns a pointer to the buffer. There would obviously have to be some bookkeeping involved: the storage manager would have to know how to map a mmap()ed page back to a writeable buffer and vice-versa, so that once it decides to write the buffer it can determine which page in the original file the buffer corresponds to (so it can do the appropriate seek()). In a write-heavy database, you'll end up with a lot of memory copy operations, but with the scheme we currently use you get that anyway (it just happens in kernel code instead of user code), so I don't see that as much of a loss, if any. Where you win is in a read-heavy database: you end up being able to read directly from the pages in the kernel's page cache and thus save a memory copy from kernel space to user space, not to mention the context switch that happens due to issuing the read(). Obviously you'd want to mmap() the file read-only in order to prevent the issues you mention regarding an errant backend, and then reopen the file read-write for the purpose of writing to it. In fact, you could decouple the two: mmap() the file, then close the file -- the mmap()ed region will remain mapped. Then, as long as the file remains mapped, you need to open the file again only when you want to write to it. -- Kevin Brown kevin@sysexperts.com From pgsql-performance-owner@postgresql.org Thu Oct 14 21:57:18 2004 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 B17C032C143 for ; Thu, 14 Oct 2004 21:57:14 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 58792-01 for ; Thu, 14 Oct 2004 20:57:10 +0000 (GMT) Received: from providerst.com.br (unknown [200.253.138.61]) by svr1.postgresql.org (Postfix) with ESMTP id B925C32C12B for ; Thu, 14 Oct 2004 21:57:06 +0100 (BST) Brma: igor [192.168.1.104] Received: from sistemas (igor [192.168.1.104]) by providerst.com.br (8.11.6/8.11.6) with SMTP id i9EKo2p12490 for ; Thu, 14 Oct 2004 18:50:02 -0200 Message-ID: <001301c4b231$280c4190$6801a8c0@providerst.local> From: "Igor Maciel Macaubas" To: Subject: View & Query Performance Date: Thu, 14 Oct 2004 18:02:41 -0300 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0010_01C4B217.FFA54670" 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=0.2 tagged_above=0.0 required=5.0 tests=HTML_60_70, HTML_FONTCOLOR_BLUE, HTML_MESSAGE X-Spam-Level: X-Archive-Number: 200410/197 X-Sequence-Number: 8673 This is a multi-part message in MIME format. ------=_NextPart_000_0010_01C4B217.FFA54670 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Hi all, I'm trying to find smarter ways to dig data from my database, and have the = following scenario: table1 -- id -- name =2E =2E =2E =2E =2E =2E table2 -- id -- number =2E =2E =2E =2E =2E =2E I want to create a view to give me back just what I want: The id, the name and the number. I tought in doing the following: create view my_view as select t1.id, t1.name, t2.number from table1 as t1, = table2 as t2 where t1.id =3D t2.id; Will this be enough fast ? Are there a faster way to make it work ?! This table is mid-big, around 100K registers ..=20 Regards, Igor -- igor@providerst.com.br ------=_NextPart_000_0010_01C4B217.FFA54670 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Hi all,
 
I'm trying to find smarter ways to dig = data from=20 my database, and have the following scenario:
 
table1
-- id
-- name
.
.
.
.
.
.
 
table2
-- id
-- number
.
.
.
.
.
.
 
I want to create a view to give me back= just=20 what I want:
The id, the name and the number.=
I tought in doing the following:=
create view my_view as select= t1.id,=20 t1.name, t2.number from table1 as t1, table2 as t2 where t1.id = =3D=20 t2.id;
 
Will this be enough fast ? Are there a = faster=20 way to make it work ?!
This table is mid-big, around 100K regi= sters ..=20
 
Regards,
Igor
--
igor@providerst.com.br
 
 
 
------=_NextPart_000_0010_01C4B217.FFA54670-- From pgsql-hackers-win32-owner@postgresql.org Thu Oct 14 22:49:49 2004 X-Original-To: pgsql-hackers-win32-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D507632B010; Thu, 14 Oct 2004 22:46:11 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 72230-08; Thu, 14 Oct 2004 21:46:02 +0000 (GMT) Received: from hotmail.com (bay10-dav18.bay10.hotmail.com [64.4.37.192]) by svr1.postgresql.org (Postfix) with ESMTP id 2AF6D32B00A; Thu, 14 Oct 2004 22:46:03 +0100 (BST) Received: from mail pickup service by hotmail.com with Microsoft SMTPSVC; Thu, 14 Oct 2004 14:46:03 -0700 Received: from 199.249.167.148 by BAY10-DAV18.phx.gbl with DAV; Thu, 14 Oct 2004 21:45:55 +0000 X-Originating-IP: [199.249.167.148] X-Originating-Email: [mikesmialek2@hotmail.com] X-Sender: mikesmialek2@hotmail.com From: "MikeSmialek2@Hotmail.com" To: "Magnus Hagander" , , References: <6BCB9D8A16AC4241919521715F4D8BCE475EE5@algol.sollentuna.se> Subject: Re: Performance on Win32 vs Cygwin Date: Thu, 14 Oct 2004 16:45:51 -0500 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 6.00.2800.1437 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 Message-ID: X-OriginalArrivalTime: 14 Oct 2004 21:46:03.0135 (UTC) FILETIME=[3391B8F0:01C4B237] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/70 X-Sequence-Number: 2360 Thanks Magnus, So are we correct to rely on - 8 being slower than 7.x in general and - 8 on Win32 being a little faster than 8 on Cygwin? Will the final release of 8 be faster than the beta? Thanks, Mike ----- Original Message ----- From: "Magnus Hagander" To: "MikeSmialek2@Hotmail.com" ; ; Sent: Thursday, October 14, 2004 12:29 PM Subject: SV: [pgsql-hackers-win32] Performance on Win32 vs Cygwin >Hi, > >We are experiencing slow performance on 8 Beta 2 Dev3 on Win32 and are >trying to determine why. Any info is appreciated. > >We have a Web Server and a DB server both running Win2KServer with all >service packs and critical updates. > >An ASP page on the Web Server hits the DB Server with a simple >query that >returns 205 rows and makes the ASP page delivered to the user >about 350K. > >On an ethernet lan a client pc perceives just under 1 sec >performance with >the following DB Server configuration: > PIII 550Mhz > 256MB RAM > 7200 RPM HD > cygwin > Postgresql 7.1.3 > PGODBC 7.3.2 > >We set up another DB Server with 8 beta (same Web Server, same >network, same >client pc) and now the client pc perceives response of just >over 3 sec with >the following DB server config: > PIII 700 Mhz > 448MB RAM > 7200 RPM HD > 8 Beta 2 Dev3 on Win32 running as a service > >Is the speed decrease because it's a beta? >Is the speed decrease because it's running on Win instead of cygwin? > >We did not install cygwin on the new DB Server. IIRC, previous versions of postgresql (< 8.0) did not correctly sync disks when running on Cygwin. I'm not 100% sure, can someone confirm? 8.0 does, and I beleive it does both under native win32 and cygwin. It's been my experience that the native version is slightly faster than the cygwin one, but I've only compared 8.0 to 8.0. //Magnus From pgsql-performance-owner@postgresql.org Thu Oct 14 23:18:18 2004 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 8A3D332B1AA for ; Thu, 14 Oct 2004 23:18:09 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 81361-06 for ; Thu, 14 Oct 2004 22:17:59 +0000 (GMT) Received: from cmailm2.svr.pol.co.uk (cmailm2.svr.pol.co.uk [195.92.193.210]) by svr1.postgresql.org (Postfix) with ESMTP id 739D032B1A4 for ; Thu, 14 Oct 2004 23:17:59 +0100 (BST) Received: from modem-649.lynx.dialup.pol.co.uk ([217.135.194.137] helo=happyplace) by cmailm2.svr.pol.co.uk with smtp (Exim 4.14) id 1CIDux-0002Vd-I8; Thu, 14 Oct 2004 23:17:51 +0100 From: "Simon Riggs" To: , Cc: Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... Date: Thu, 14 Oct 2004 23:36:22 +0100 Message-ID: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" 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) Importance: Normal In-Reply-To: <200410081443.16109.josh@agliodbs.com> X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.3 tagged_above=0.0 required=5.0 tests=LOTS_OF_STUFF X-Spam-Level: X-Archive-Number: 200410/198 X-Sequence-Number: 8674 First off, I'd like to get involved with these tests - pressure of other work only has prevented me. Here's my take on the results so far: I think taking the ratio of the memory allocated to shared_buffers against the total memory available on the server is completely fallacious. That is why they cannnot be explained - IMHO the ratio has no real theoretical basis. The important ratio for me is the amount of shared_buffers against the total size of the database in the benchmark test. Every database workload has a differing percentage of the total database size that represents the "working set", or the memory that can be beneficially cached. For the tests that DBT-2 is performing, I say that there is only so many blocks that are worth the trouble caching. If you cache more than this, you are wasting your time. For me, these tests don't show that there is a "sweet spot" that you should set your shared_buffers to, only that for that specific test, you have located the correct size for shared_buffers. For me, it would be an incorrect inference that this could then be interpreted that this was the percentage of the available RAM where the "sweet spot" lies for all workloads. The theoretical basis for my comments is this: DBT-2 is essentially a static workload. That means, for a long test, we can work out with reasonable certainty the probability that a block will be requested, for every single block in the database. Given a particular size of cache, you can work out what your overall cache hit ratio is and therfore what your speed up is compared with retrieving every single block from disk (the no cache scenario). If you draw a graph of speedup (y) against cache size as a % of total database size, the graph looks like an upside-down "L" - i.e. the graph rises steeply as you give it more memory, then turns sharply at a particular point, after which it flattens out. The "turning point" is the "sweet spot" we all seek - the optimum amount of cache memory to allocate - but this spot depends upon the worklaod and database size, not on available RAM on the system under test. Clearly, the presence of the OS disk cache complicates this. Since we have two caches both allocated from the same pot of memory, it should be clear that if we overallocate one cache beyond its optimium effectiveness, while the second cache is still in its "more is better" stage, then we will get reduced performance. That seems to be the case here. I wouldn't accept that a fixed ratio between the two caches exists for ALL, or even the majority of workloads - though clearly broad brush workloads such as "OLTP" and "Data Warehousing" do have similar-ish requirements. As an example, lets look at an example: An application with two tables: SmallTab has 10,000 rows of 100 bytes each (so table is ~1 Mb)- one row per photo in a photo gallery web site. LargeTab has large objects within it and has 10,000 photos, average size 10 Mb (so table is ~100Gb). Assuming all photos are requested randomly, you can see that an optimum cache size for this workload is 1Mb RAM, 100Gb disk. Trying to up the cache doesn't have much effect on the probability that a photo (from LargeTab) will be in cache, unless you have a large % of 100Gb of RAM, when you do start to make gains. (Please don't be picky about indexes, catalog, block size etc). That clearly has absolutely nothing at all to do with the RAM of the system on which it is running. I think Jan has said this also in far fewer words, but I'll leave that to Jan to agree/disagree... I say this: ARC in 8.0 PostgreSQL allows us to sensibly allocate as large a shared_buffers cache as is required by the database workload, and this should not be constrained to a small percentage of server RAM. Best Regards, Simon Riggs > -----Original Message----- > From: pgsql-performance-owner@postgresql.org > [mailto:pgsql-performance-owner@postgresql.org]On Behalf Of Josh Berkus > Sent: 08 October 2004 22:43 > To: pgsql-performance@postgresql.org > Cc: testperf-general@pgfoundry.org > Subject: [PERFORM] First set of OSDL Shared Mem scalability results, > some wierdness ... > > > Folks, > > I'm hoping that some of you can shed some light on this. > > I've been trying to peg the "sweet spot" for shared memory using OSDL's > equipment. With Jan's new ARC patch, I was expecting that the desired > amount of shared_buffers to be greatly increased. This has not > turned out to > be the case. > > The first test series was using OSDL's DBT2 (OLTP) test, with 150 > "warehouses". All tests were run on a 4-way Pentium III 700mhz > 3.8GB RAM > system hooked up to a rather high-end storage device (14 > spindles). Tests > were on PostgreSQL 8.0b3, Linux 2.6.7. > > Here's a top-level summary: > > shared_buffers % RAM NOTPM20* > 1000 0.2% 1287 > 23000 5% 1507 > 46000 10% 1481 > 69000 15% 1382 > 92000 20% 1375 > 115000 25% 1380 > 138000 30% 1344 > > * = New Order Transactions Per Minute, last 20 Minutes > Higher is better. The maximum possible is 1800. > > As you can see, the "sweet spot" appears to be between 5% and 10% of RAM, > which is if anything *lower* than recommendations for 7.4! > > This result is so surprising that I want people to take a look at > it and tell > me if there's something wrong with the tests or some bottlenecking factor > that I've not seen. > > in order above: > http://khack.osdl.org/stp/297959/ > http://khack.osdl.org/stp/297960/ > http://khack.osdl.org/stp/297961/ > http://khack.osdl.org/stp/297962/ > http://khack.osdl.org/stp/297963/ > http://khack.osdl.org/stp/297964/ > http://khack.osdl.org/stp/297965/ > > Please note that many of the Graphs in these reports are broken. For one > thing, some aren't recorded (flat lines) and the CPU usage graph has > mislabeled lines. > > -- > --Josh > > Josh Berkus > Aglio Database Solutions > San Francisco > > > ---------------------------(end of broadcast)--------------------------- > TIP 8: explain analyze is your friend From pgsql-performance-owner@postgresql.org Fri Oct 15 00:55:56 2004 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 0FB6632AC23 for ; Fri, 15 Oct 2004 00:55:54 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 02518-10 for ; Thu, 14 Oct 2004 23:55:48 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id A502932A9F8 for ; Fri, 15 Oct 2004 00:55:50 +0100 (BST) 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 6504705; Thu, 14 Oct 2004 16:57:14 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... Date: Thu, 14 Oct 2004 16:57:45 -0700 User-Agent: KMail/1.6.2 Cc: "Simon Riggs" , References: In-Reply-To: MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410141657.45415.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/199 X-Sequence-Number: 8675 Simon, > If you draw a graph of speedup (y) against cache size as a > % of total database size, the graph looks like an upside-down "L" - i.e. > the graph rises steeply as you give it more memory, then turns sharply at a > particular point, after which it flattens out. The "turning point" is the > "sweet spot" we all seek - the optimum amount of cache memory to allocate - > but this spot depends upon the worklaod and database size, not on available > RAM on the system under test. Hmmm ... how do you explain, then the "camel hump" nature of the real performance? That is, when we allocated even a few MB more than the "optimum" ~190MB, overall performance stated to drop quickly. The result is that allocating 2x optimum RAM is nearly as bad as allocating too little (e.g. 8MB). The only explanation I've heard of this so far is that there is a significant loss of efficiency with larger caches. Or do you see the loss of 200MB out of 3500MB would actually affect the Kernel cache that much? Anyway, one test of your theory that I can run immediately is to run the exact same workload on a bigger, faster server and see if the desired quantity of shared_buffers is roughly the same. I'm hoping that you're wrong -- not because I don't find your argument persuasive, but because if you're right it leaves us without any reasonable ability to recommend shared_buffer settings. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Fri Oct 15 01:09:09 2004 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 9198C32A90A for ; Fri, 15 Oct 2004 01:09:03 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 08770-03 for ; Fri, 15 Oct 2004 00:08:56 +0000 (GMT) Received: from mail.osdl.org (fw.osdl.org [65.172.181.6]) by svr1.postgresql.org (Postfix) with ESMTP id 3589332A828 for ; Fri, 15 Oct 2004 01:08:57 +0100 (BST) Received: from localhost (osdlab.pdx.osdl.net [172.20.1.28]) by mail.osdl.org (8.11.6/8.11.6) with ESMTP id i9F08t916459; Thu, 14 Oct 2004 17:08:55 -0700 Subject: Re: [Testperf-general] Re: First set of OSDL Shared Mem From: "Timothy D. Witham" To: josh@agliodbs.com Cc: pgsql-performance@postgresql.org, Simon Riggs , testperf-general@pgfoundry.org In-Reply-To: <200410141657.45415.josh@agliodbs.com> References: <200410141657.45415.josh@agliodbs.com> Content-Type: text/plain Organization: Open Source Development Lab, Inc. Date: Thu, 14 Oct 2004 17:09:05 -0700 Message-Id: <1097798945.6387.86.camel@wookie-zd7> Mime-Version: 1.0 X-Mailer: Evolution 2.0.0 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/200 X-Sequence-Number: 8676 On Thu, 2004-10-14 at 16:57 -0700, Josh Berkus wrote: > Simon, > > > > > If you draw a graph of speedup (y) against cache size as a > > % of total database size, the graph looks like an upside-down "L" - i.e. > > the graph rises steeply as you give it more memory, then turns sharply at a > > particular point, after which it flattens out. The "turning point" is the > > "sweet spot" we all seek - the optimum amount of cache memory to allocate - > > but this spot depends upon the worklaod and database size, not on available > > RAM on the system under test. > > Hmmm ... how do you explain, then the "camel hump" nature of the real > performance? That is, when we allocated even a few MB more than the > "optimum" ~190MB, overall performance stated to drop quickly. The result is > that allocating 2x optimum RAM is nearly as bad as allocating too little > (e.g. 8MB). > > The only explanation I've heard of this so far is that there is a significant > loss of efficiency with larger caches. Or do you see the loss of 200MB out > of 3500MB would actually affect the Kernel cache that much? > In a past life there seemed to be a sweet spot around the applications working set. Performance went up until you got just a little larger than the cache needed to hold the working set and then went down. Most of the time a nice looking hump. It seems to have to do with the additional pages not increasing your hit ratio but increasing the amount of work to get a hit in cache. This seemed to be independent of the actual database software being used. (I observed this running Oracle, Informix, Sybase and Ingres.) > Anyway, one test of your theory that I can run immediately is to run the exact > same workload on a bigger, faster server and see if the desired quantity of > shared_buffers is roughly the same. I'm hoping that you're wrong -- not > because I don't find your argument persuasive, but because if you're right it > leaves us without any reasonable ability to recommend shared_buffer settings. > -- Timothy D. Witham - Chief Technology Officer - wookie@osdl.org Open Source Development Lab Inc - A non-profit corporation 12725 SW Millikan Way - Suite 400 - Beaverton OR, 97005 (503)-626-2455 x11 (office) (503)-702-2871 (cell) (503)-626-2436 (fax) From pgsql-performance-owner@postgresql.org Fri Oct 15 02:19:26 2004 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 05A7932BA82 for ; Fri, 15 Oct 2004 02:19:23 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 25035-05 for ; Fri, 15 Oct 2004 01:19:16 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id E259A32B3AB for ; Fri, 15 Oct 2004 02:19:19 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i9F1JFJ9025861 for ; Fri, 15 Oct 2004 01:19:15 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i9F0o1pt018278 for pgsql-performance@postgresql.org; Fri, 15 Oct 2004 00:50:01 GMT From: Christopher Browne X-Newsgroups: comp.databases.postgresql.performance Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... Date: Thu, 14 Oct 2004 20:10:59 -0400 Organization: cbbrowne Computing Inc Lines: 37 Message-ID: References: <200410081443.16109.josh@agliodbs.com> 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 (Security Through Obscurity, linux) Cancel-Lock: sha1:XQr6mfAZWXpVgi6qa+zMNUctOcE= To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/204 X-Sequence-Number: 8680 Quoth simon@2ndquadrant.com ("Simon Riggs"): > I say this: ARC in 8.0 PostgreSQL allows us to sensibly allocate as > large a shared_buffers cache as is required by the database > workload, and this should not be constrained to a small percentage > of server RAM. I don't think that this particularly follows from "what ARC does." "What ARC does" is to prevent certain conspicuous patterns of sequential accesses from essentially trashing the contents of the cache. If a particular benchmark does not include conspicuous vacuums or sequential scans on large tables, then there is little reason to expect ARC to have a noticeable impact on performance. It _could_ be that this implies that ARC allows you to get some use out of a larger shared cache, as it won't get blown away by vacuums and Seq Scans. But it is _not_ obvious that this is a necessary truth. _Other_ truths we know about are: a) If you increase the shared cache, that means more data that is represented in both the shared cache and the OS buffer cache, which seems rather a waste; b) The larger the shared cache, the more pages there are for the backend to rummage through before it looks to the filesystem, and therefore the more expensive cache misses get. Cache hits get more expensive, too. Searching through memory is not costless. -- (format nil "~S@~S" "cbbrowne" "acm.org") http://linuxfinances.info/info/linuxdistributions.html "The X-Files are too optimistic. The truth is *not* out there..." -- Anthony Ord From pgsql-performance-owner@postgresql.org Fri Oct 15 01:25:44 2004 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 F3DF632B1BD for ; Fri, 15 Oct 2004 01:25:40 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 10722-06 for ; Fri, 15 Oct 2004 00:25:34 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.204]) by svr1.postgresql.org (Postfix) with ESMTP id 4203332B0E9 for ; Fri, 15 Oct 2004 01:25:36 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 74so123281rnk for ; Thu, 14 Oct 2004 17:25:37 -0700 (PDT) Received: by 10.38.102.60 with SMTP id z60mr3324377rnb; Thu, 14 Oct 2004 17:25:36 -0700 (PDT) Received: by 10.38.151.75 with HTTP; Thu, 14 Oct 2004 17:25:36 -0700 (PDT) Message-ID: <157f64840410141725162e43b5@mail.gmail.com> Date: Thu, 14 Oct 2004 20:25:36 -0400 From: Aaron Werman Reply-To: Aaron Werman To: pgsql-performance@postgresql.org, Kevin Brown Subject: mmap (was First set of OSDL Shared Mem scalability results, some wierdness ... 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/201 X-Sequence-Number: 8677 pg to my mind is unique in not trying to avoid OS buffering. Other dbmses spend a substantial effort to create a virtual OS (task management, I/O drivers, etc.) both in code and support. Choosing mmap seems such a limiting an option - it adds OS dependency and limits kernel developer options (2G limits, global mlock serializations, porting problems, inability to schedule or parallelize I/O, still having to coordinate writers and readers). More to the point, I think it is very hard to effectively coordinate multithreaded I/O, and mmap seems used mostly to manage relatively simple scenarios. If the I/O options are: - OS (which has enormous investment and is stable, but is general purpose with overhead) - pg (direct I/O would be costly and potentially destabilizing, but with big possible performance rewards) - mmap (a feature mostly used to reduce buffer copies in less concurrent apps such as image processing that has major architectural risk including an order of magnitude more semaphores, but can reduce some extra block copies) mmap doesn't look that promising. /Aaron ----- Original Message ----- From: "Kevin Brown" To: Sent: Thursday, October 14, 2004 4:25 PM Subject: Re: [PERFORM] First set of OSDL Shared Mem scalability results, some wierdness ... > Tom Lane wrote: > > Kevin Brown writes: > > > Tom Lane wrote: > > >> mmap() is Right Out because it does not afford us sufficient control > > >> over when changes to the in-memory data will propagate to disk. > > > > > ... that's especially true if we simply cannot > > > have the page written to disk in a partially-modified state (something > > > I can easily see being an issue for the WAL -- would the same hold > > > true of the index/data files?). > > > > You're almost there. Remember the fundamental WAL rule: log entries > > must hit disk before the data changes they describe. That means that we > > need not only a way of forcing changes to disk (fsync) but a way of > > being sure that changes have *not* gone to disk yet. In the existing > > implementation we get that by just not issuing write() for a given page > > until we know that the relevant WAL log entries are fsync'd down to > > disk. (BTW, this is what the LSN field on every page is for: it tells > > the buffer manager the latest WAL offset that has to be flushed before > > it can safely write the page.) > > > > mmap provides msync which is comparable to fsync, but AFAICS it > > provides no way to prevent an in-memory change from reaching disk too > > soon. This would mean that WAL entries would have to be written *and > > flushed* before we could make the data change at all, which would > > convert multiple updates of a single page into a series of write-and- > > wait-for-WAL-fsync steps. Not good. fsync'ing WAL once per transaction > > is bad enough, once per atomic action is intolerable. > > Hmm...something just occurred to me about this. > > Would a hybrid approach be possible? That is, use mmap() to handle > reads, and use write() to handle writes? > > Any code that wishes to write to a page would have to recognize that > it's doing so and fetch a copy from the storage manager (or > something), which would look to see if the page already exists as a > writeable buffer. If it doesn't, it creates it by allocating the > memory and then copying the page from the mmap()ed area to the new > buffer, and returning it. If it does, it just returns a pointer to > the buffer. There would obviously have to be some bookkeeping > involved: the storage manager would have to know how to map a mmap()ed > page back to a writeable buffer and vice-versa, so that once it > decides to write the buffer it can determine which page in the > original file the buffer corresponds to (so it can do the appropriate > seek()). > > In a write-heavy database, you'll end up with a lot of memory copy > operations, but with the scheme we currently use you get that anyway > (it just happens in kernel code instead of user code), so I don't see > that as much of a loss, if any. Where you win is in a read-heavy > database: you end up being able to read directly from the pages in the > kernel's page cache and thus save a memory copy from kernel space to > user space, not to mention the context switch that happens due to > issuing the read(). > > > Obviously you'd want to mmap() the file read-only in order to prevent > the issues you mention regarding an errant backend, and then reopen > the file read-write for the purpose of writing to it. In fact, you > could decouple the two: mmap() the file, then close the file -- the > mmap()ed region will remain mapped. Then, as long as the file remains > mapped, you need to open the file again only when you want to write to > it. > > > -- > Kevin Brown kevin@sysexperts.com > > ---------------------------(end of broadcast)--------------------------- > TIP 8: explain analyze is your friend > -- Regards, /Aaron From pgsql-performance-owner@postgresql.org Fri Oct 15 01:52:47 2004 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 4BBEE32A805 for ; Fri, 15 Oct 2004 01:52:44 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 18222-02 for ; Fri, 15 Oct 2004 00:52:38 +0000 (GMT) Received: from sue.samurai.com (sue.samurai.com [205.207.28.74]) by svr1.postgresql.org (Postfix) with ESMTP id 8096C32A407 for ; Fri, 15 Oct 2004 01:52:40 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by sue.samurai.com (Postfix) with ESMTP id D706D197A4; Thu, 14 Oct 2004 20:52:40 -0400 (EDT) Received: from sue.samurai.com ([127.0.0.1]) by localhost (sue.samurai.com [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 42357-01-6; Thu, 14 Oct 2004 20:52:40 -0400 (EDT) Received: from [61.88.101.19] (unknown [61.88.101.19]) by sue.samurai.com (Postfix) with ESMTP id 309B519792; Thu, 14 Oct 2004 20:52:38 -0400 (EDT) Subject: Re: Performance vs Schemas From: Neil Conway To: Igor Maciel Macaubas Cc: pgsql-performance@postgresql.org In-Reply-To: <003101c4b21d$10f00e60$6801a8c0@providerst.local> References: <003101c4b21d$10f00e60$6801a8c0@providerst.local> Content-Type: text/plain Message-Id: <1097801555.29932.11.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Fri, 15 Oct 2004 10:52:35 +1000 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/202 X-Sequence-Number: 8678 On Fri, 2004-10-15 at 04:38, Igor Maciel Macaubas wrote: > I have around 100 tables, and divided them in 14 different schemas, > and then adapted my application to use schemas as well. > I could percept that the query / insert / update times get pretty much > faster then when I was using the old unique schema, and I'd just like > to confirm with you if using schemas speed up the things. Is that true > ? Schemas are a namespacing technique; AFAIK they shouldn't significantly affect performance (either positively or negatively). > What about indexed views, does postgresql supports it? No, you'll need to create indexes on the view's base tables. -Neil From pgsql-performance-owner@postgresql.org Fri Oct 15 02:08:49 2004 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 9974732A61F for ; Fri, 15 Oct 2004 02:08:47 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 20828-05 for ; Fri, 15 Oct 2004 01:08:41 +0000 (GMT) Received: from sccrmhc13.comcast.net (sccrmhc13.comcast.net [204.127.202.64]) by svr1.postgresql.org (Postfix) with ESMTP id 6781B32A518 for ; Fri, 15 Oct 2004 02:08:42 +0100 (BST) Received: from sysexperts.com (c-24-6-183-218.client.comcast.net[24.6.183.218]) by comcast.net (sccrmhc13) with ESMTP id <20041015010842016000kbste>; Fri, 15 Oct 2004 01:08:43 +0000 Received: from localhost (localhost [127.0.0.1]) (uid 1000) by filer with local; Thu, 14 Oct 2004 18:08:41 -0700 id 000071AD.416F2319.00001B0A Date: Thu, 14 Oct 2004 18:08:41 -0700 From: Kevin Brown To: pgsql-performance@postgresql.org Subject: Re: mmap (was First set of OSDL Shared Mem scalability results, some wierdness ... Message-ID: <20041015010841.GE665@filer> Mail-Followup-To: Kevin Brown , pgsql-performance@postgresql.org References: <157f64840410141725162e43b5@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline In-Reply-To: <157f64840410141725162e43b5@mail.gmail.com> Organization: Frobozzco International User-Agent: Mutt/1.5.6+20040818i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/203 X-Sequence-Number: 8679 Aaron Werman wrote: > pg to my mind is unique in not trying to avoid OS buffering. Other > dbmses spend a substantial effort to create a virtual OS (task > management, I/O drivers, etc.) both in code and support. Choosing mmap > seems such a limiting an option - it adds OS dependency and limits > kernel developer options (2G limits, global mlock serializations, > porting problems, inability to schedule or parallelize I/O, still > having to coordinate writers and readers). I'm not sure I entirely agree with this. Whether you access a file via mmap() or via read(), the end result is that you still have to access it, and since PG has significant chunks of system-dependent code that it heavily relies on as it is (e.g., locking mechanisms, shared memory), writing the I/O subsystem in a similar way doesn't seem to me to be that much of a stretch (especially since PG already has the storage manager), though it might involve quite a bit of work. As for parallelization of I/O, the use of mmap() for reads should signficantly improve parallelization -- now instead of issuing read() system calls, possibly for the same set of blocks, all the backends would essentially be examining the same data directly. The performance improvements as a result of accessing the kernel's cache pages directly instead of having it do buffer copies to process-local memory should increase as concurrency goes up. But see below. > More to the point, I think it is very hard to effectively coordinate > multithreaded I/O, and mmap seems used mostly to manage relatively > simple scenarios. PG already manages and coordinates multithreaded I/O. The mechanisms used to coordinate writes needn't change at all. But the way reads are done relative to writes might have to be rethought, since an mmap()ed buffer always reflects what's actually in kernel space at the time the buffer is accessed, while a buffer retrieved via read() reflects the state of the file at the time of the read(). If it's necessary for the state of the buffers to be fixed at examination time, then mmap() will be at best a draw, not a win. > mmap doesn't look that promising. This ultimately depends on two things: how much time is spent copying buffers around in kernel memory, and how much advantage can be gained by freeing up the memory used by the backends to store the backend-local copies of the disk pages they use (and thus making that memory available to the kernel to use for additional disk buffering). The gains from the former are likely small. The gains from the latter are probably also small, but harder to estimate. The use of mmap() is probably one of those optimizations that should be done when there's little else left to optimize, because the potential gains are possibly (if not probably) relatively small and the amount of work involved may be quite large. So I agree -- compared with other, much lower-hanging fruit, mmap() doesn't look promising. -- Kevin Brown kevin@sysexperts.com From pgsql-performance-owner@postgresql.org Fri Oct 15 02:41:46 2004 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 65FA932A273 for ; Fri, 15 Oct 2004 02:41:45 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 30856-01 for ; Fri, 15 Oct 2004 01:41:38 +0000 (GMT) Received: from millenium.mst.co.jp (unknown [210.230.185.241]) by svr1.postgresql.org (Postfix) with ESMTP id 5964F32BB51 for ; Fri, 15 Oct 2004 02:41:40 +0100 (BST) Received: from mst1x5r347kymb (lc12114 [192.168.1.114]) by millenium.mst.co.jp (8.11.6p2/3.7W) with SMTP id i9F1fWq08516; Fri, 15 Oct 2004 10:41:32 +0900 Message-ID: <00cb01c4b258$1e51a5b0$7201a8c0@mst1x5r347kymb> From: "Iain" To: "Igor Maciel Macaubas" , References: <003101c4b21d$10f00e60$6801a8c0@providerst.local> Subject: Re: Performance vs Schemas Date: Fri, 15 Oct 2004 10:41:35 +0900 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_00C3_01C4B2A3.8B373140" 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=0.6 tagged_above=0.0 required=5.0 tests=HTML_40_50, HTML_FONTCOLOR_BLUE, HTML_MESSAGE X-Spam-Level: X-Archive-Number: 200410/205 X-Sequence-Number: 8681 This is a multi-part message in MIME format. ------=_NextPart_000_00C3_01C4B2A3.8B373140 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Hi Igor, I expect that when you moved your tables to different schemas that you effectively did a physical re-organization (ie unload/reload of the tables). It's nothing to do with the use of schemas as such. If you had reloaded your tables into the same system schema you would have experienced the same speedup as the data tables would be more compact. regards Iain ----- Original Message ----- From: Igor Maciel Macaubas To: pgsql-performance@postgresql.org Sent: Friday, October 15, 2004 3:38 AM Subject: [PERFORM] Performance vs Schemas Hi all, I recently migrated my database from schema 'public' to multiple schema. I have around 100 tables, and divided them in 14 different schemas, and then adapted my application to use schemas as well. I could percept that the query / insert / update times get pretty much faster then when I was using the old unique schema, and I'd just like to confirm with you if using schemas speed up the things. Is that true ? What else I can do to speed up the query processing, best pratices, recommendations ... ? What about indexed views, does postgresql supports it? Regards, Igor -- igor@providerst.com.br ------=_NextPart_000_00C3_01C4B2A3.8B373140 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Hi Igor,
 
I expect that when you moved your= tables=20 to different schemas that you effectively did a physical re-organization (i= e=20 unload/reload of the tables). It's nothing to do with the use of= =20 schemas as such. If you had reloaded your tables into the same system= =20 schema you would have experienced the same speedup as the data tables = would=20 be more compact.
 
regards
Iain
----- Original Message -----
From:=20 = Igor=20 Maciel Macaubas
To: pgsql-performance@postgr= esql.org=20
Sent: Friday, October 15, 2= 004=20 3:38 AM
Subject: [PERFORM] Performa= nce vs=20 Schemas

Hi all,
 
I recently migrated my database from = schema=20 'public' to multiple schema.
I have around 100 tables, and divided= them in=20 14 different schemas, and then adapted my application to use schemas= as=20 well.
I could percept that the query / inse= rt /=20 update times get pretty much faster then when I was using the old unique= =20 schema, and I'd just like to confirm with you if using schemas speed up t= he=20 things. Is that true ?
 
What else I can do to speed up the qu= ery=20 processing, best pratices, recommendations ... ? What about indexed views= ,=20 does postgresql supports it?
 
Regards,
Igor
--
igor@providerst.com.br<= /DIV>
 
------=_NextPart_000_00C3_01C4B2A3.8B373140-- From pgsql-performance-owner@postgresql.org Fri Oct 15 06:13:42 2004 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 92FB432C91F for ; Fri, 15 Oct 2004 06:13:40 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 76884-10 for ; Fri, 15 Oct 2004 05:13:37 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 8672D32C91E for ; Fri, 15 Oct 2004 06:13:37 +0100 (BST) 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 i9F5DaWu005236; Fri, 15 Oct 2004 01:13:36 -0400 (EDT) To: Kevin Brown Cc: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... In-reply-to: <20041014202531.GD665@filer> References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> <4859.1097363137@sss.pgh.pa.us> <20041014202531.GD665@filer> Comments: In-reply-to Kevin Brown message dated "Thu, 14 Oct 2004 13:25:31 -0700" Date: Fri, 15 Oct 2004 01:13:36 -0400 Message-ID: <5235.1097817216@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/206 X-Sequence-Number: 8682 Kevin Brown writes: > Hmm...something just occurred to me about this. > Would a hybrid approach be possible? That is, use mmap() to handle > reads, and use write() to handle writes? Nope. Have you read the specs regarding mmap-vs-stdio synchronization? Basically it says that there are no guarantees whatsoever if you try this. The SUS text is a bit weaselly ("the application must ensure correct synchronization") but the HPUX mmap man page, among others, lays it on the line: It is also unspecified whether write references to a memory region mapped with MAP_SHARED are visible to processes reading the file and whether writes to a file are visible to processes that have mapped the modified portion of that file, except for the effect of msync(). It might work on particular OSes but I think depending on such behavior would be folly... regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 15 08:28:31 2004 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 6C7DB32C2F3 for ; Fri, 15 Oct 2004 08:28:27 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 11608-02 for ; Fri, 15 Oct 2004 07:28:22 +0000 (GMT) Received: from mailer.unicite.fr.netcentrex.net (unknown [62.161.167.249]) by svr1.postgresql.org (Postfix) with ESMTP id B60AB32C2FC for ; Fri, 15 Oct 2004 08:28:20 +0100 (BST) Received: from akira ([192.168.101.228]) by mailer.unicite.fr.netcentrex.net with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2657.72) id 42CYABK1; Fri, 15 Oct 2004 09:28:18 +0200 From: "Alban Medici (NetCentrex)" To: "'Igor Maciel Macaubas'" , Subject: Re: View & Query Performance Date: Fri, 15 Oct 2004 09:28:18 +0200 Organization: NetCentrex MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0050_01C4B299.4E97AD00" X-Mailer: Microsoft Office Outlook, Build 11.0.6353 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 Thread-Index: AcSyMMsrEu7NNwBtSRmj3nO3Tb/vEQAV2chA In-Reply-To: <001301c4b231$280c4190$6801a8c0@providerst.local> Message-Id: <20041015072820.B60AB32C2FC@svr1.postgresql.org> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.2 tagged_above=0.0 required=5.0 tests=HTML_60_70, HTML_FONTCOLOR_BLUE, HTML_MESSAGE X-Spam-Level: X-Archive-Number: 200410/207 X-Sequence-Number: 8683 This is a multi-part message in MIME format. ------=_NextPart_000_0050_01C4B299.4E97AD00 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Can you tell us more about the structure of your tables, witch sort of index did you set on witch fields ? =20 Did you really need to get ALL records at once, instead you may be could use paging (cursor or SELECT LIMIT OFFSET ) ? =20 And did you well configure your .conf ? =20 Regards =20 Alban M=E9dici _____=20=20 From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of Igor Maciel Macaubas Sent: jeudi 14 octobre 2004 23:03 To: pgsql-performance@postgresql.org Subject: [PERFORM] View & Query Performance Hi all, =20 I'm trying to find smarter ways to dig data from my database, and have the following scenario: =20 table1 -- id -- name =2E =2E =2E =2E =2E =2E =20 table2 -- id -- number =2E =2E =2E =2E =2E =2E =20 I want to create a view to give me back just what I want: The id, the name and the number. I tought in doing the following: create view my_view as select t1.id, t1.name, t2.number from table1 as t1, table2 as t2 where t1.id =3D t2.id; =20 Will this be enough fast ? Are there a faster way to make it work ?! This table is mid-big, around 100K registers ..=20 =20 Regards, Igor -- igor@providerst.com.br =20 =20 =20 ------=_NextPart_000_0050_01C4B299.4E97AD00 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Can you tell us more about the structure of your= =20 tables,
witch sort of index did you set on witch fields=20 ?
 
Did you really need to get ALL records at=20 once, instead you may be could use pa= ging=20 (cursor or SELECT LIMIT OFFSET ) ?
 
And did you= well=20 configure your .conf ?
 
Regards
 
Alban=20 M=E9dici

From: pgsql-performance-owner@postgresq= l.org=20 [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of Igor Ma= ciel=20 Macaubas
Sent: jeudi 14 octobre 2004 23:03
To:=20 pgsql-performance@postgresql.org
Subject: [PERFORM] View & Qu= ery=20 Performance

Hi all,
 
I'm trying to find smarter ways to dig = data from=20 my database, and have the following scenario:
 
table1
-- id
-- name
.
.
.
.
.
.
 
table2
-- id
-- number
.
.
.
.
.
.
 
I want to create a view to give me back= just=20 what I want:
The id, the name and the number.=
I tought in doing the following:=
create view my_view as select= t1.id,=20 t1.name, t2.number from table1 as t1, table2 as t2 where t1.id = =3D=20 t2.id;
 
Will this be enough fast ? Are there a = faster=20 way to make it work ?!
This table is mid-big, around 100K regi= sters ..=20
 
Regards,
Igor
--
igor@providerst.com.br
 
 
 
------=_NextPart_000_0050_01C4B299.4E97AD00-- From pgsql-performance-owner@postgresql.org Fri Oct 15 08:37:38 2004 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 1BE1432C302 for ; Fri, 15 Oct 2004 08:37:37 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 13616-05 for ; Fri, 15 Oct 2004 07:37:32 +0000 (GMT) Received: from anchor-post-33.mail.demon.net (anchor-post-33.mail.demon.net [194.217.242.91]) by svr1.postgresql.org (Postfix) with ESMTP id 2BCF132C30B for ; Fri, 15 Oct 2004 08:37:33 +0100 (BST) Received: from mwynhau.demon.co.uk ([193.237.186.96] helo=mainbox.archonet.com) by anchor-post-33.mail.demon.net with esmtp (Exim 4.42) id 1CIMeu-0004g7-9m; Fri, 15 Oct 2004 07:38:06 +0000 Received: from [192.168.1.17] (client17.archonet.com [192.168.1.17]) by mainbox.archonet.com (Postfix) with ESMTP id 5F20C15A0F; Fri, 15 Oct 2004 08:37:07 +0100 (BST) Message-ID: <416F7E22.6000702@archonet.com> Date: Fri, 15 Oct 2004 08:37:06 +0100 From: Richard Huxton User-Agent: Mozilla Thunderbird 0.7 (X11/20040615) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Igor Maciel Macaubas Cc: pgsql-performance@postgresql.org Subject: Re: View & Query Performance References: <001301c4b231$280c4190$6801a8c0@providerst.local> In-Reply-To: <001301c4b231$280c4190$6801a8c0@providerst.local> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/208 X-Sequence-Number: 8684 Igor Maciel Macaubas wrote: > Hi all, > > I'm trying to find smarter ways to dig data from my database, and > have the following scenario: > > table1 -- id -- name . . . . . . > > table2 -- id -- number . . . . . . > > I want to create a view to give me back just what I want: The id, the > name and the number. I tought in doing the following: create view > my_view as select t1.id, t1.name, t2.number from table1 as t1, table2 > as t2 where t1.id = t2.id; > > Will this be enough fast ? Are there a faster way to make it work ?! > This table is mid-big, around 100K registers .. That's as simple a way as you will find. If you apply further conditions, e.g. SELECT * FROM my_view WHERE id = 123; then you should see any index on "id" being used. -- Richard Huxton Archonet Ltd From pgsql-performance-owner@postgresql.org Fri Oct 15 08:37:54 2004 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 C949132C304 for ; Fri, 15 Oct 2004 08:37:52 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 12291-07 for ; Fri, 15 Oct 2004 07:37:48 +0000 (GMT) Received: from cmailg1.svr.pol.co.uk (cmailg1.svr.pol.co.uk [195.92.195.171]) by svr1.postgresql.org (Postfix) with ESMTP id 8A26C32C302 for ; Fri, 15 Oct 2004 08:37:45 +0100 (BST) Received: from modem-141.leopard.dialup.pol.co.uk ([217.135.144.141] helo=happyplace) by cmailg1.svr.pol.co.uk with smtp (Exim 4.14) id 1CIMef-0003L8-1L; Fri, 15 Oct 2004 08:37:37 +0100 From: "Simon Riggs" To: "Timothy D. Witham" , Cc: , Subject: Re: [Testperf-general] Re: First set of OSDL Shared Memscalability results, some wierdness ... Date: Fri, 15 Oct 2004 08:55:59 +0100 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) Importance: Normal In-Reply-To: <1097798945.6387.86.camel@wookie-zd7> X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/209 X-Sequence-Number: 8685 >Timothy D. Witham > On Thu, 2004-10-14 at 16:57 -0700, Josh Berkus wrote: > > Simon, > > > > > > > > > If you draw a graph of speedup (y) against cache size as a > > > % of total database size, the graph looks like an upside-down > "L" - i.e. > > > the graph rises steeply as you give it more memory, then > turns sharply at a > > > particular point, after which it flattens out. The "turning > point" is the > > > "sweet spot" we all seek - the optimum amount of cache memory > to allocate - > > > but this spot depends upon the worklaod and database size, > not on available > > > RAM on the system under test. > > > > Hmmm ... how do you explain, then the "camel hump" nature of the real > > performance? That is, when we allocated even a few MB more than the > > "optimum" ~190MB, overall performance stated to drop quickly. > The result is > > that allocating 2x optimum RAM is nearly as bad as allocating > too little > > (e.g. 8MB). Two ways of explaining this: 1. Once you've hit the optimum size of shared_buffers, you may not yet have hit the optimum size of the OS cache. If that is true, every extra block given to shared_buffers is wasted, yet detracts from the beneficial effect of the OS cache. I don't see how the small drop in size of the OS cache could have the effect you have measured, so I suggest that this possible explanation doesn't fit the results well. 2. There is some algorithmic effect within PostgreSQL that makes larger shared_buffers much worse than smaller ones. Imagine that each extra block we hold in cache has the positive benefit from caching, minus a postulated negative drag effect. With that model we would get: Once the optimal size of the cache has been reached the positive benefit tails off to almost zero and we are just left with the situation that each new block added to shared_buffers acts as a further drag on performance. That model would fit the results, so we can begin to look at what the drag effect might be. Speculating wildly because I don't know that portion of the code this might be: CONJECTURE 1: the act of searching for a block in cache is an O(n) operation, not an O(1) or O(log n) operation - so searching a larger cache has an additional slowing effect on the application, via a buffer cache lock that is held while the cache is searched - larger caches are locked for longer than smaller caches, so this causes additional contention in the system, which then slows down performance. The effect might show up by examining the oprofile results for the test cases. What we would be looking for is something that is being called more frequently with larger shared_buffers - this could be anything....but my guess is the oprofile results won't be similar and could lead us to a better understanding. > > > > The only explanation I've heard of this so far is that there is > a significant > > loss of efficiency with larger caches. Or do you see the loss > of 200MB out > > of 3500MB would actually affect the Kernel cache that much? > > > In a past life there seemed to be a sweet spot around the > applications > working set. Performance went up until you got just a little larger > than > the cache needed to hold the working set and then went down. Most of > the time a nice looking hump. It seems to have to do with the > additional pages > not increasing your hit ratio but increasing the amount of work to get a > hit in cache. This seemed to be independent of the actual database > software being used. (I observed this running Oracle, Informix, Sybase > and Ingres.) Good, our experiences seems to be similar. > > > Anyway, one test of your theory that I can run immediately is > to run the exact > > same workload on a bigger, faster server and see if the desired > quantity of > > shared_buffers is roughly the same. I agree that you could test this by running on a bigger or smaller server, i.e. one with more or less RAM. Running on a faster/slower server at the same time might alter the results and confuse the situation. > I'm hoping that you're wrong -- not > because I don't find your argument persuasive, but because if > you're right it > > leaves us without any reasonable ability to recommend > shared_buffer settings. > For the record, what I think we need is dynamically resizable shared_buffers, not a-priori knowledge of what you should set shared_buffers to. I've been thinking about implementing a scheme that helps you decide how big the shared_buffers SHOULD BE, by making the LRU list bigger than the cache itself, so you'd be able to see whether there is beneficial effect in increasing shared_buffers. ...remember that this applies to other databases too, and with those we find that they have dynamically resizable memory. Having said all that, there are still a great many other performance tests to run so that we CAN recommend other settings, such as the optimizer cost parameters, bg writer defaults etc. Best Regards, Simon Riggs 2nd Quadrant From pgsql-performance-owner@postgresql.org Fri Oct 15 10:19:58 2004 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 B6DB432C26C for ; Fri, 15 Oct 2004 10:19:48 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 43938-01 for ; Fri, 15 Oct 2004 09:19:43 +0000 (GMT) Received: from rwcrmhc11.comcast.net (rwcrmhc11.comcast.net [204.127.198.35]) by svr1.postgresql.org (Postfix) with ESMTP id 49EB432C24C for ; Fri, 15 Oct 2004 10:19:44 +0100 (BST) Received: from sysexperts.com (c-24-6-183-218.client.comcast.net[24.6.183.218]) by comcast.net (rwcrmhc11) with ESMTP id <2004101509194201300t513se>; Fri, 15 Oct 2004 09:19:42 +0000 Received: from localhost (localhost [127.0.0.1]) (uid 1000) by filer with local; Fri, 15 Oct 2004 02:19:41 -0700 id 0000AFFE.416F962D.0000335A Date: Fri, 15 Oct 2004 02:19:40 -0700 From: Kevin Brown To: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... Message-ID: <20041015091940.GF665@filer> Mail-Followup-To: Kevin Brown , pgsql-performance@postgresql.org References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> <4859.1097363137@sss.pgh.pa.us> <20041014202531.GD665@filer> <5235.1097817216@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline In-Reply-To: <5235.1097817216@sss.pgh.pa.us> Organization: Frobozzco International User-Agent: Mutt/1.5.6+20040818i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/210 X-Sequence-Number: 8686 Tom Lane wrote: > Kevin Brown writes: > > Hmm...something just occurred to me about this. > > > Would a hybrid approach be possible? That is, use mmap() to handle > > reads, and use write() to handle writes? > > Nope. Have you read the specs regarding mmap-vs-stdio synchronization? > Basically it says that there are no guarantees whatsoever if you try > this. The SUS text is a bit weaselly ("the application must ensure > correct synchronization") but the HPUX mmap man page, among others, > lays it on the line: > > It is also unspecified whether write references to a memory region > mapped with MAP_SHARED are visible to processes reading the file and > whether writes to a file are visible to processes that have mapped the > modified portion of that file, except for the effect of msync(). > > It might work on particular OSes but I think depending on such behavior > would be folly... Yeah, and at this point it can't be considered portable in any real way because of this. Thanks for the perspective. I should have expected the general specification to be quite broken in this regard, not to mention certain implementations. :-) Good thing there's a lot of lower-hanging fruit than this... -- Kevin Brown kevin@sysexperts.com From pgsql-performance-owner@postgresql.org Fri Oct 15 11:25:40 2004 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 9314332C28B for ; Fri, 15 Oct 2004 11:25:36 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 60815-04 for ; Fri, 15 Oct 2004 10:25:30 +0000 (GMT) Received: from mail.core.genedata.com (mail.core.genedata.com [157.161.173.16]) by svr1.postgresql.org (Postfix) with ESMTP id 07AA332C2B1 for ; Fri, 15 Oct 2004 11:25:29 +0100 (BST) Received: from relay.core.genedata.com (root@nila-e0.core.genedata.com [172.20.16.64]) (authenticated bits=128) by mail.core.genedata.com (8.13.1/8.13.1) with ESMTP id i9FAPRZW002208 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK) for ; Fri, 15 Oct 2004 12:25:27 +0200 Received: from relay.ch.genedata.com (root@vesuvio-e0.ch.genedata.com [172.20.16.80]) (authenticated bits=128) by relay.core.genedata.com (8.13.1/8.13.1) with ESMTP id i9FAPQE0013485 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK) for ; Fri, 15 Oct 2004 12:25:26 +0200 X-DomainKeys: Sendmail dk v0.2.2 relay.core.genedata.com i9FAPQE0013485 DomainKey-Signature: a=rsa-sha1; s=genedata; d=genedata.com; c=simple; q=dns; h=received:from:organization:to:subject:date:user-agent:mime-version:content-disposition:content-type:content-transfer-encoding:message-id:x-filter-version; b=rLjItBje6hzXRWSPVoPiE/m6NHQuFKUWfa5vLIEXX6VDHU0MkDJvQ23VtgwHY7Zc Received: from rotokaua.ch.genedata.com (rotokaua.ch.genedata.com [172.20.39.133]) by relay.ch.genedata.com (8.13.1/8.13.1) with ESMTP id i9FAPQX7019139 for ; Fri, 15 Oct 2004 12:25:26 +0200 From: Bernd Organization: Genedata To: pgsql-performance@postgresql.org Subject: Select with qualified join condition / Batch inserts Date: Fri, 15 Oct 2004 12:25:26 +0200 User-Agent: KMail/1.6.2 MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Message-Id: <200410151225.26083.bernd_pg@genedata.com> X-Filter-Version: 1.15 (nila) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/211 X-Sequence-Number: 8687 Hi, we are working on a product which was originally developed against an Oracle database and which should be changed to also work with postgres. Overall the changes we had to make are very small and we are very pleased with the good performance of postgres - but we also found queries which execute much faster on Oracle. Since I am not yet familiar with tuning queries for postgres, it would be great if someone could give me a hint on the following two issues. (We are using PG 8.0.0beta3 on Linux kernel 2.4.27): 1/ The following query takes about 5 sec. with postrgres whereas on Oracle it executes in about 30 ms (although both tables only contain 200 k records in the postgres version). SQL: SELECT cmp.WELL_INDEX, cmp.COMPOUND, con.CONCENTRATION FROM SCR_WELL_COMPOUND cmp, SCR_WELL_CONCENTRATION con WHERE cmp.BARCODE=con.BARCODE AND cmp.WELL_INDEX=con.WELL_INDEX AND cmp.MAT_ID=con.MAT_ID AND cmp.MAT_ID = 3 AND cmp.BARCODE='910125864' AND cmp.ID_LEVEL = 1; Table-def: Table "public.scr_well_compound" Column | Type | Modifiers ------------+------------------------+----------- mat_id | numeric(10,0) | not null barcode | character varying(240) | not null well_index | numeric(5,0) | not null id_level | numeric(3,0) | not null compound | character varying(240) | not null Indexes: "scr_wcm_pk" PRIMARY KEY, btree (id_level, mat_id, barcode, well_index) Foreign-key constraints: "scr_wcm_mat_fk" FOREIGN KEY (mat_id) REFERENCES scr_mapping_table(mat_id) ON DELETE CASCADE Table "public.scr_well_concentration" Column | Type | Modifiers ---------------+------------------------+----------- mat_id | numeric(10,0) | not null barcode | character varying(240) | not null well_index | numeric(5,0) | not null concentration | numeric(20,10) | not null Indexes: "scr_wco_pk" PRIMARY KEY, btree (mat_id, barcode, well_index) Foreign-key constraints: "scr_wco_mat_fk" FOREIGN KEY (mat_id) REFERENCES scr_mapping_table(mat_id) ON DELETE CASCADE I tried several variants of the query (including the SQL 92 JOIN ON syntax) but with no success. I have also rebuilt the underlying indices. A strange observation is that the same query runs pretty fast without the restriction to a certain MAT_ID, i. e. omitting the MAT_ID=3 part. Also fetching the data for both tables separately is pretty fast and a possible fallback would be to do the actual join in the application (which is of course not as beautiful as doing it using SQL ;-) 2/ Batch-inserts using jdbc (maybe this should go to the jdbc-mailing list - but it is also performance related ...): Performing many inserts using a PreparedStatement and batch execution makes a significant performance improvement in Oracle. In postgres, I did not observe any performance improvement using batch execution. Are there any special caveats when using batch execution with postgres? Thanks and regards Bernd From pgsql-performance-owner@postgresql.org Fri Oct 15 11:36:43 2004 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 DA9F832C0CE for ; Fri, 15 Oct 2004 11:36:41 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 64632-07 for ; Fri, 15 Oct 2004 10:36:35 +0000 (GMT) Received: from bayswater1.ymogen.net (host-154-240-27-217.pobox.net.uk [217.27.240.154]) by svr1.postgresql.org (Postfix) with ESMTP id 68EF8329FB1 for ; Fri, 15 Oct 2004 11:36:36 +0100 (BST) Received: from solent (unknown [213.165.136.10]) by bayswater1.ymogen.net (Postfix) with ESMTP id 22D4BA3244; Fri, 15 Oct 2004 11:36:35 +0100 (BST) Reply-To: From: "Matt Clark" To: "'Bernd'" , Subject: Re: Select with qualified join condition / Batch inserts Date: Fri, 15 Oct 2004 11:36:35 +0100 Organization: Ymogen Ltd Message-ID: <001e01c4b2a2$d82b77f0$8300a8c0@solent> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook, Build 10.0.6626 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 In-Reply-To: <200410151225.26083.bernd_pg@genedata.com> Importance: Normal X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.1 tagged_above=0.0 required=5.0 tests=UPPERCASE_50_75 X-Spam-Level: * X-Archive-Number: 200410/212 X-Sequence-Number: 8688 > SELECT cmp.WELL_INDEX, cmp.COMPOUND, con.CONCENTRATION > FROM SCR_WELL_COMPOUND cmp, SCR_WELL_CONCENTRATION con > WHERE cmp.BARCODE=con.BARCODE > AND cmp.WELL_INDEX=con.WELL_INDEX > AND cmp.MAT_ID=con.MAT_ID > AND cmp.MAT_ID = 3 > AND cmp.BARCODE='910125864' > AND cmp.ID_LEVEL = 1; Quick guess - type mismatch forcing sequential scan. Try some quotes: AND cmp.MAT_ID = '3' AND cmp.BARCODE='910125864' AND cmp.ID_LEVEL = '1'; M From pgsql-performance-owner@postgresql.org Fri Oct 15 11:45:09 2004 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 1B9EC32C20B for ; Fri, 15 Oct 2004 11:45:08 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 64245-09 for ; Fri, 15 Oct 2004 10:45:02 +0000 (GMT) Received: from usbb-lacimss1.unisys.com (usbb-lacimss1.unisys.com [192.63.108.51]) by svr1.postgresql.org (Postfix) with ESMTP id 7081232C12F for ; Fri, 15 Oct 2004 11:45:02 +0100 (BST) Received: from usbb-lacgw1.lac.uis.unisys.com ([129.226.160.21]unverified) by usbb-lacimss1 with InterScan Messaging Security Suite; Fri, 15 Oct 2004 06:48:01 -0400 Received: from usbb-lacgw1.lac.uis.unisys.com ([129.226.160.23]) by usbb-lacgw1.lac.uis.unisys.com with Microsoft SMTPSVC(6.0.3790.0); Fri, 15 Oct 2004 06:44:42 -0400 Received: from gbmk-eugw2.eu.uis.unisys.com ([129.221.133.27]) by usbb-lacgw1.lac.uis.unisys.com with Microsoft SMTPSVC(6.0.3790.0); Fri, 15 Oct 2004 06:44:41 -0400 Received: from nlshl-exch1.eu.uis.unisys.com ([192.39.239.20]) by gbmk-eugw2.eu.uis.unisys.com with Microsoft SMTPSVC(6.0.3790.47); Fri, 15 Oct 2004 11:44:39 +0100 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: Select with qualified join condition / Batch inserts Date: Fri, 15 Oct 2004 12:44:37 +0200 Message-ID: X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Select with qualified join condition / Batch inserts Thread-Index: AcSyownVZObCmInqSUWgaP1oAebrewAAMkPA From: "Leeuw van der, Tim" To: , "Bernd" , X-OriginalArrivalTime: 15 Oct 2004 10:44:39.0048 (UTC) FILETIME=[F86C5880:01C4B2A3] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/213 X-Sequence-Number: 8689 But he's testing with v8 beta3, so you'd expect the typecast problem not to= appear? Are all tables fully vacuumed? Should the statistics-target be raised for s= ome columns, perhaps? What about the config file? --Tim -----Original Message----- From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owne= r@postgresql.org]On Behalf Of Matt Clark Sent: Friday, October 15, 2004 12:37 PM To: 'Bernd'; pgsql-performance@postgresql.org Subject: Re: [PERFORM] Select with qualified join condition / Batch inserts > SELECT cmp.WELL_INDEX, cmp.COMPOUND, con.CONCENTRATION=20 > FROM SCR_WELL_COMPOUND cmp, SCR_WELL_CONCENTRATION con=20 > WHERE cmp.BARCODE=3Dcon.BARCODE=20 > AND cmp.WELL_INDEX=3Dcon.WELL_INDEX=20 > AND cmp.MAT_ID=3Dcon.MAT_ID=20 > AND cmp.MAT_ID =3D 3=20 > AND cmp.BARCODE=3D'910125864'=20 > AND cmp.ID_LEVEL =3D 1; Quick guess - type mismatch forcing sequential scan. Try some quotes: AND cmp.MAT_ID =3D '3'=20 AND cmp.BARCODE=3D'910125864'=20 AND cmp.ID_LEVEL =3D '1'; M ---------------------------(end of broadcast)--------------------------- TIP 3: 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 Oct 19 22:15:15 2004 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 D1E5B32C2CC for ; Fri, 15 Oct 2004 11:48:15 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 68737-05 for ; Fri, 15 Oct 2004 10:48:09 +0000 (GMT) Received: from linuxworld.com.au (unknown [203.34.46.50]) by svr1.postgresql.org (Postfix) with ESMTP id 4EFA732C2BA for ; Fri, 15 Oct 2004 11:48:10 +0100 (BST) Received: from localhost (swm@localhost) by linuxworld.com.au (8.11.6/8.11.4) with ESMTP id i9FAlua30359; Fri, 15 Oct 2004 20:47:56 +1000 Date: Fri, 15 Oct 2004 20:47:56 +1000 (EST) From: Gavin Sherry To: Bernd Cc: pgsql-performance@postgresql.org Subject: Re: Select with qualified join condition / Batch inserts In-Reply-To: <200410151225.26083.bernd_pg@genedata.com> Message-ID: References: <200410151225.26083.bernd_pg@genedata.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/292 X-Sequence-Number: 8768 On Fri, 15 Oct 2004, Bernd wrote: > Hi, > > we are working on a product which was originally developed against an Oracle > database and which should be changed to also work with postgres. > > Overall the changes we had to make are very small and we are very pleased with > the good performance of postgres - but we also found queries which execute > much faster on Oracle. Since I am not yet familiar with tuning queries for > postgres, it would be great if someone could give me a hint on the following > two issues. (We are using PG 8.0.0beta3 on Linux kernel 2.4.27): > > 1/ The following query takes about 5 sec. with postrgres whereas on Oracle it > executes in about 30 ms (although both tables only contain 200 k records in > the postgres version). > > SQL: > > SELECT cmp.WELL_INDEX, cmp.COMPOUND, con.CONCENTRATION > FROM SCR_WELL_COMPOUND cmp, SCR_WELL_CONCENTRATION con > WHERE cmp.BARCODE=con.BARCODE > AND cmp.WELL_INDEX=con.WELL_INDEX > AND cmp.MAT_ID=con.MAT_ID > AND cmp.MAT_ID = 3 > AND cmp.BARCODE='910125864' > AND cmp.ID_LEVEL = 1; > > Table-def: > Table "public.scr_well_compound" > Column | Type | Modifiers > ------------+------------------------+----------- > mat_id | numeric(10,0) | not null > barcode | character varying(240) | not null > well_index | numeric(5,0) | not null > id_level | numeric(3,0) | not null > compound | character varying(240) | not null > Indexes: > "scr_wcm_pk" PRIMARY KEY, btree (id_level, mat_id, barcode, well_index) I presume you've VACUUM FULL'd and ANALYZE'd? Can we also see a plan? EXPLAIN ANALYZE . http://www.postgresql.org/docs/7.4/static/sql-explain.html. You may need to create indexes with other primary columns. Ie, on mat_id or barcode. > 2/ Batch-inserts using jdbc (maybe this should go to the jdbc-mailing list - > but it is also performance related ...): > Performing many inserts using a PreparedStatement and batch execution makes a > significant performance improvement in Oracle. In postgres, I did not observe > any performance improvement using batch execution. Are there any special > caveats when using batch execution with postgres? The JDBC people should be able to help with that. Gavin From pgsql-performance-owner@postgresql.org Fri Oct 15 13:53:28 2004 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 698D232C0D9 for ; Fri, 15 Oct 2004 13:53:25 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 05648-03 for ; Fri, 15 Oct 2004 12:53:17 +0000 (GMT) Received: from ram.rentec.com (ram.rentec.com [192.5.35.66]) by svr1.postgresql.org (Postfix) with ESMTP id D0A6232C0CE for ; Fri, 15 Oct 2004 13:53:19 +0100 (BST) Received: from [172.26.132.145] (hoopoe.rentec.com [172.26.132.145]) by ram.rentec.com (8.12.9/8.12.1) with ESMTP id i9FCrJnK001841; Fri, 15 Oct 2004 08:53:19 -0400 (EDT) Message-ID: <416FC83F.3020900@rentec.com> Date: Fri, 15 Oct 2004 08:53:19 -0400 From: Alan Stange Reply-To: stange@rentec.com Organization: Renaissance Technologies Corp. User-Agent: Mozilla Thunderbird 0.8 (X11/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: Kevin Brown , pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> <4859.1097363137@sss.pgh.pa.us> <20041014202531.GD665@filer> <5235.1097817216@sss.pgh.pa.us> In-Reply-To: <5235.1097817216@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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/214 X-Sequence-Number: 8690 Tom Lane wrote: >Kevin Brown writes: > > >>Hmm...something just occurred to me about this. >> >> >>Would a hybrid approach be possible? That is, use mmap() to handle >>reads, and use write() to handle writes? >> >> > >Nope. Have you read the specs regarding mmap-vs-stdio synchronization? >Basically it says that there are no guarantees whatsoever if you try >this. The SUS text is a bit weaselly ("the application must ensure >correct synchronization") but the HPUX mmap man page, among others, >lays it on the line: > > It is also unspecified whether write references to a memory region > mapped with MAP_SHARED are visible to processes reading the file and > whether writes to a file are visible to processes that have mapped the > modified portion of that file, except for the effect of msync(). > >It might work on particular OSes but I think depending on such behavior >would be folly... > We have some anecdotal experience along these lines: There was a set of kernel bugs in Solaris 2.6 or 7 related to this as well. We had several kernel panics and it took a bit to chase down, but the basic feedback was "oops. we're screwed". I've forgotten most of the details right now; the basic problem was a file was being read+written via mmap and read()/write() at (essentially) the same time from the same pid. It would panic the system quite reliably. I believe the bugs related to this have been resolved in Solaris, but it was unpleasant to chase that problem down... -- Alan From pgsql-performance-owner@postgresql.org Fri Oct 15 15:17:15 2004 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 C1A4232C61A for ; Fri, 15 Oct 2004 15:17:12 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 31954-08 for ; Fri, 15 Oct 2004 14:17:04 +0000 (GMT) Received: from auscorpex-1.austin.messageone.com (auscorpex-1.austin.messageone.com [66.219.55.10]) by svr1.postgresql.org (Postfix) with ESMTP id 670C932C60E for ; Fri, 15 Oct 2004 15:17:06 +0100 (BST) X-MimeOLE: Produced By Microsoft Exchange V6.5.6944.0 Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Subject: Re: Select with qualified join condition / Batch inserts Date: Fri, 15 Oct 2004 09:17:06 -0500 Message-ID: X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Select with qualified join condition / Batch inserts Thread-Index: AcSyoZgjV0sYiqBoQ/ytF3TSIWTHdwAH2UIw From: "Michael Nonemacher" To: "Bernd" , X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/215 X-Sequence-Number: 8691 > 2/ Batch-inserts using jdbc (maybe this should go to the jdbc-mailing list -=20 > but it is also performance related ...): > Performing many inserts using a PreparedStatement and batch execution makes a=20 > significant performance improvement in Oracle. In postgres, I did not observe=20 > any performance improvement using batch execution. Are there any special=20 > caveats when using batch execution with postgres? When you call executeBatch(), it doesn't send all the queries in a single round-trip; it just iterates through the batched queries and executes them one by one. In my own applications, I've done simulated-batch queries like this: insert into T (a, b, c) select 1,2,3 union all select 2,3,4 union all select 3,4,5 It's ugly, and you have to structure your code in such a way that the query can't get too large, but it provides a similar performance benefit to batching. You probably don't save nearly as much parse time as using a batched PreparedStatement, but you at least get rid of the network roundtrips. (Of course, it'd be much nicer if statement-batching worked. There have been rumblings about doing this, and some discussion on how to do it, but I haven't heard about any progress. Anyone?) mike -----Original Message----- From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of Bernd Sent: Friday, October 15, 2004 5:25 AM To: pgsql-performance@postgresql.org Subject: [PERFORM] Select with qualified join condition / Batch inserts Hi, we are working on a product which was originally developed against an Oracle=20 database and which should be changed to also work with postgres.=20 Overall the changes we had to make are very small and we are very pleased with=20 the good performance of postgres - but we also found queries which execute=20 much faster on Oracle. Since I am not yet familiar with tuning queries for=20 postgres, it would be great if someone could give me a hint on the following=20 two issues. (We are using PG 8.0.0beta3 on Linux kernel 2.4.27): 1/ The following query takes about 5 sec. with postrgres whereas on Oracle it=20 executes in about 30 ms (although both tables only contain 200 k records in=20 the postgres version). SQL: SELECT cmp.WELL_INDEX, cmp.COMPOUND, con.CONCENTRATION=20 FROM SCR_WELL_COMPOUND cmp, SCR_WELL_CONCENTRATION con=20 WHERE cmp.BARCODE=3Dcon.BARCODE=20 AND cmp.WELL_INDEX=3Dcon.WELL_INDEX=20 AND cmp.MAT_ID=3Dcon.MAT_ID=20 AND cmp.MAT_ID =3D 3=20 AND cmp.BARCODE=3D'910125864'=20 AND cmp.ID_LEVEL =3D 1; Table-def: Table "public.scr_well_compound" Column | Type | Modifiers ------------+------------------------+----------- mat_id | numeric(10,0) | not null barcode | character varying(240) | not null well_index | numeric(5,0) | not null id_level | numeric(3,0) | not null compound | character varying(240) | not null Indexes: "scr_wcm_pk" PRIMARY KEY, btree (id_level, mat_id, barcode, well_index) Foreign-key constraints: "scr_wcm_mat_fk" FOREIGN KEY (mat_id) REFERENCES scr_mapping_table(mat_id)=20 ON DELETE CASCADE Table "public.scr_well_concentration" Column | Type | Modifiers ---------------+------------------------+----------- mat_id | numeric(10,0) | not null barcode | character varying(240) | not null well_index | numeric(5,0) | not null concentration | numeric(20,10) | not null Indexes: "scr_wco_pk" PRIMARY KEY, btree (mat_id, barcode, well_index) Foreign-key constraints: "scr_wco_mat_fk" FOREIGN KEY (mat_id) REFERENCES scr_mapping_table(mat_id)=20 ON DELETE CASCADE I tried several variants of the query (including the SQL 92 JOIN ON syntax)=20 but with no success. I have also rebuilt the underlying indices. A strange observation is that the same query runs pretty fast without the=20 restriction to a certain MAT_ID, i. e. omitting the MAT_ID=3D3 part. Also fetching the data for both tables separately is pretty fast and a=20 possible fallback would be to do the actual join in the application (which is=20 of course not as beautiful as doing it using SQL ;-) 2/ Batch-inserts using jdbc (maybe this should go to the jdbc-mailing list -=20 but it is also performance related ...): Performing many inserts using a PreparedStatement and batch execution makes a=20 significant performance improvement in Oracle. In postgres, I did not observe=20 any performance improvement using batch execution. Are there any special caveats when using batch execution with postgres? Thanks and regards Bernd ---------------------------(end of broadcast)--------------------------- TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org From pgsql-hackers-win32-owner@postgresql.org Fri Oct 15 16:37:52 2004 X-Original-To: pgsql-hackers-win32-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C014B32BEB1 for ; Fri, 15 Oct 2004 16:37:50 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 65540-04 for ; Fri, 15 Oct 2004 15:37:45 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id B4C4232BE65 for ; Fri, 15 Oct 2004 16:37:44 +0100 (BST) 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 i9FFbeBe009632; Fri, 15 Oct 2004 11:37:40 -0400 (EDT) To: "MikeSmialek2@Hotmail.com" Cc: "Magnus Hagander" , pgsql-performance@postgresql.org, pgsql-hackers-win32@postgresql.org Subject: Re: Performance on Win32 vs Cygwin In-reply-to: References: <6BCB9D8A16AC4241919521715F4D8BCE475EE5@algol.sollentuna.se> Comments: In-reply-to "MikeSmialek2@Hotmail.com" message dated "Thu, 14 Oct 2004 16:45:51 -0500" Date: Fri, 15 Oct 2004 11:37:40 -0400 Message-ID: <9631.1097854660@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests=TO_ADDRESS_EQ_REAL X-Spam-Level: X-Archive-Number: 200410/71 X-Sequence-Number: 2361 "MikeSmialek2@Hotmail.com" writes: > So are we correct to rely on > - 8 being slower than 7.x in general and I think this is a highly unlikely claim ... *especially* if you are comparing against 7.1. The point about sync() being a no-op is real, but offhand I think it would only come into play at checkpoints. We have never issued sync() during regular queries. What seems more likely to me is that you have neglected to do any performance tuning on the new installation. Have you vacuumed/analyzed all your tables? Checked the postgresql.conf settings for sanity? If you'd like to do an apples-to-apples comparison to prove whether 7.1's failure to sync() is relevant, then turn off fsync in the 8.0 configuration and see how much difference that makes. If you can identify specific queries that are slower in 8.0 than 7.1, I'd be interested to see the EXPLAIN ANALYZE details from each. (Actually, I'm not sure 7.1 had EXPLAIN ANALYZE; you may have to settle for EXPLAIN from it.) regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 15 17:07:39 2004 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 D8C2932BEC2 for ; Fri, 15 Oct 2004 17:07:37 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 75775-04 for ; Fri, 15 Oct 2004 16:07:32 +0000 (GMT) Received: from neomail07.traderonline.com (email.traderonline.com [65.213.231.10]) by svr1.postgresql.org (Postfix) with ESMTP id 8871D32BE0C for ; Fri, 15 Oct 2004 17:07:30 +0100 (BST) Received: from [192.168.1.68] (64-139-89-109-ubr02b-epensb01-pa.hfc.comcastbusiness.net [64.139.89.109]) by neomail07.traderonline.com (8.12.10/8.12.8) with ESMTP id i9FG7TOY018965 for ; Fri, 15 Oct 2004 12:07:29 -0400 Message-ID: <416FF5BF.8060200@ptd.net> Date: Fri, 15 Oct 2004 12:07:27 -0400 From: Doug Y User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Tuning shared_buffers with ipcs ? 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/217 X-Sequence-Number: 8693 Hello, I've seen a couple references to using ipcs to help properly size shared_buffers. I don't claim to be a SA guru, so could someone help explain how to interpret the output of ipcs and how that relates to shared_buffers? How does one determine the size of the segment arrays? I see the total size using ipcs -m which is roughly shared_buffers * 8k. I tried all of the dash commands in the ipcs man page, and the only one that might give a clue is ipcs -t which shows the time the semaphores were last used. If you look at the example I give below, it appears as if I'm only using 4 of the 17 semaphores (PG was started on Oct 8). Am I correct in assuming that if the arrays are all the same size then I should only need about 1/4 of my currently allocated shared_buffers? ------ Shared Memory Operation/Change Times -------- shmid owner last-op last-changed 847183872 postgres Fri Oct 8 11:03:31 2004 Fri Oct 8 11:03:31 2004 847216641 postgres Fri Oct 8 11:03:31 2004 Fri Oct 8 11:03:31 2004 847249410 postgres Fri Oct 8 11:03:31 2004 Fri Oct 8 11:03:31 2004 847282179 postgres Fri Oct 8 11:03:31 2004 Fri Oct 8 11:03:31 2004 847314948 postgres Fri Oct 8 11:03:31 2004 Fri Oct 8 11:03:31 2004 847347717 postgres Fri Oct 8 11:03:31 2004 Fri Oct 8 11:03:31 2004 847380486 postgres Fri Oct 8 11:03:31 2004 Fri Oct 8 11:03:31 2004 847413255 postgres Fri Oct 8 11:03:31 2004 Fri Oct 8 11:03:31 2004 847446024 postgres Fri Oct 8 11:03:31 2004 Fri Oct 8 11:03:31 2004 847478793 postgres Fri Oct 8 11:03:31 2004 Fri Oct 8 11:03:31 2004 847511562 postgres Fri Oct 8 11:03:31 2004 Fri Oct 8 11:03:31 2004 847544331 postgres Fri Oct 8 11:03:31 2004 Fri Oct 8 11:03:31 2004 847577100 postgres Fri Oct 8 11:03:31 2004 Fri Oct 8 11:03:31 2004 847609869 postgres Fri Oct 15 11:34:28 2004 Fri Oct 15 11:34:29 2004 847642638 postgres Fri Oct 15 11:33:35 2004 Fri Oct 15 11:33:35 2004 847675407 postgres Fri Oct 15 11:34:28 2004 Fri Oct 15 11:34:29 2004 847708176 postgres Fri Oct 15 11:27:17 2004 Fri Oct 15 11:32:20 2004 Also, isn't the shared memory supposed to show up in free? Its always showing as 0: # free total used free shared buffers cached Mem: 3896928 3868424 28504 0 59788 3605548 -/+ buffers/cache: 203088 3693840 Swap: 1052216 16 1052200 Thanks! From pgsql-hackers-win32-owner@postgresql.org Fri Oct 15 17:22:51 2004 X-Original-To: pgsql-hackers-win32-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3F87B32BEDE; Fri, 15 Oct 2004 17:22:49 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 79806-09; Fri, 15 Oct 2004 16:22:42 +0000 (GMT) Received: from Herge.rcsinc.local (mail.rcsonline.com [205.217.85.91]) by svr1.postgresql.org (Postfix) with ESMTP id B4AFF32BEC5; Fri, 15 Oct 2004 17:22:41 +0100 (BST) 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 V6.5.6944.0 Subject: Re: Performance on Win32 vs Cygwin Date: Fri, 15 Oct 2004 12:22:40 -0400 Message-ID: <6EE64EF3AB31D5448D0007DD34EEB3412A7506@Herge.rcsinc.local> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [pgsql-hackers-win32] Performance on Win32 vs Cygwin Thread-Index: AcSyN8X/LYt2ZT6+S2Kt3p2J/oCEXwAmmvmw From: "Merlin Moncure" To: "MikeSmialek2@Hotmail.com" Cc: , , "Magnus Hagander" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests=TO_ADDRESS_EQ_REAL X-Spam-Level: X-Archive-Number: 200410/72 X-Sequence-Number: 2362 > Thanks Magnus, >=20 > So are we correct to rely on > - 8 being slower than 7.x in general and > - 8 on Win32 being a little faster than 8 on Cygwin? >=20 > Will the final release of 8 be faster than the beta? I'm pretty certain that previous to 8.0 no win32 based postgesql properly sync()ed the files. Win32 does not have sync(), and it is impossible to emulate it without relying on the application to track which files to sync. 8.0 does this because it fsync()s the files individually. Therefore, benchmarking fsync=3Don on 8.0 to a <8.0 version of windows is not apples to apples. This includes, by the way, the SFU based port of postgresql because they didn't implement sync() there, either. Other than the sync() issue, the cygwin/win32 i/o performance should be roughly equal. Unless I'm terribly mistaken about things, all the i/o calls should boil down to win32 api calls. The cygwin IPC stack is implemented differently...pg 8.0 win32 native version does all the ipc stuff by hand, so you might get slightly different behavior there. Merln From pgsql-performance-owner@postgresql.org Fri Oct 15 17:48:31 2004 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 18C76329F33 for ; Fri, 15 Oct 2004 17:48:28 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 89548-03 for ; Fri, 15 Oct 2004 16:48:22 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id E74C932BF01 for ; Fri, 15 Oct 2004 17:48:22 +0100 (BST) 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 i9FGmFl0010203; Fri, 15 Oct 2004 12:48:15 -0400 (EDT) To: "Simon Riggs" Cc: "Timothy D. Witham" , josh@agliodbs.com, pgsql-performance@postgresql.org, testperf-general@pgfoundry.org Subject: Re: [Testperf-general] Re: First set of OSDL Shared Memscalability results, some wierdness ... In-reply-to: References: Comments: In-reply-to "Simon Riggs" message dated "Fri, 15 Oct 2004 08:55:59 +0100" Date: Fri, 15 Oct 2004 12:48:14 -0400 Message-ID: <10202.1097858894@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/219 X-Sequence-Number: 8695 "Simon Riggs" writes: > Speculating wildly because I don't know that portion of the code this might > be: > CONJECTURE 1: the act of searching for a block in cache is an O(n) > operation, not an O(1) or O(log n) operation I'm not sure how this meme got into circulation, but I've seen a couple of people recently either conjecturing or asserting that. Let me remind people of the actual facts: 1. We use a hashtable to keep track of which blocks are currently in shared buffers. Either a cache hit or a cache miss should be O(1), because the hashtable size is scaled proportionally to shared_buffers, and so the number of hash entries examined should remain constant. 2. There are some allegedly-not-performance-critical operations that do scan through all the buffers, and therefore are O(N) in shared_buffers. I just eyeballed all the latter, and came up with this list of O(N) operations and their call points: AtEOXact_Buffers transaction commit or abort UnlockBuffers transaction abort, backend exit StrategyDirtyBufferList background writer's idle loop FlushRelationBuffers VACUUM DROP TABLE, DROP INDEX TRUNCATE, CLUSTER, REINDEX ALTER TABLE SET TABLESPACE DropRelFileNodeBuffers TRUNCATE (only for ON COMMIT TRUNC temp tables) REINDEX (inplace case only) smgr_internal_unlink (ie, the tail end of DROP TABLE/INDEX) DropBuffers DROP DATABASE The fact that the first two are called during transaction commit/abort is mildly alarming. The constant factors are going to be very tiny though, because what these routines actually do is scan backend-local status arrays looking for locked buffers, which they're not going to find very many of. For instance AtEOXact_Buffers looks like int i; for (i = 0; i < NBuffers; i++) { if (PrivateRefCount[i] != 0) { // some code that should never be executed at all in the commit // case, and not that much in the abort case either } } I suppose with hundreds of thousands of shared buffers this might get to the point of being noticeable, but I've never seen it show up at all in profiling with more-normal buffer counts. Not sure if it's worth devising a more complex data structure to aid in finding locked buffers. (To some extent this code is intended to be belt-and-suspenders stuff for catching omissions elsewhere, and so a more complex data structure that could have its own bugs is not especially attractive.) The one that's bothering me at the moment is StrategyDirtyBufferList, which is a new overhead in 8.0. It wouldn't directly affect foreground query performance, but indirectly it would hurt by causing the bgwriter to suck more CPU cycles than one would like (and it holds the BufMgrLock while it's doing it, too :-(). One easy way you could see whether this is an issue in the OSDL test is to see what happens if you double all three bgwriter parameters (delay, percent, maxpages). This should result in about the same net I/O demand from the bgwriter, but StrategyDirtyBufferList will be executed half as often. I doubt that the other ones are issues. We could improve them by devising a way to quickly find all buffers for a given relation, but I am just about sure that complicating the buffer management to do so would be a net loss for normal workloads. > For the record, what I think we need is dynamically resizable > shared_buffers, not a-priori knowledge of what you should set > shared_buffers to. This isn't likely to happen because the SysV shared memory API isn't conducive to it. Absent some amazingly convincing demonstration that we have to have it, the effort of making it happen in a portable way isn't going to get spent. > I've been thinking about implementing a scheme that helps you decide how > big the shared_buffers SHOULD BE, by making the LRU list bigger than the > cache itself, so you'd be able to see whether there is beneficial effect in > increasing shared_buffers. ARC already keeps such a list --- couldn't you learn what you want to know from the existing data structure? It'd be fairly cool if we could put out warnings "you ought to increase shared_buffers" analogous to the existing facility for noting excessive checkpointing. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 15 17:57:38 2004 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 AC05132BEE1 for ; Fri, 15 Oct 2004 17:57:36 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91708-09 for ; Fri, 15 Oct 2004 16:57:28 +0000 (GMT) Received: from lax-gate3.raytheon.com (lax-gate3.raytheon.com [199.46.200.232]) by svr1.postgresql.org (Postfix) with ESMTP id 6219D32BEDB for ; Fri, 15 Oct 2004 17:57:22 +0100 (BST) Received: from ds02w00.directory.ray.com (ds02w00.directory.ray.com [147.25.146.118]) by lax-gate3.raytheon.com (8.12.10/8.12.10) with ESMTP id i9FGupdw003025 for ; Fri, 15 Oct 2004 09:56:56 -0700 (PDT) Received: from ds02w00 (localhost [127.0.0.1]) by ds02w00.directory.ray.com (Switch-3.1.4/Switch-3.1.0) with ESMTP id i9FGuekD027050 for ; Fri, 15 Oct 2004 16:56:44 GMT Received: from ds02w00.directory.ray.com with LMTP by ds02w00 (2.0.6/sieved-2-0-build-559) for ; Fri, 15 Oct 2004 16:56:40 +0000 Received: from notesserver5.ftw.us.ray.com (notesserver5.ftw.us.ray.com [151.168.145.35]) by ds02w00.directory.ray.com (Switch-3.1.4/Switch-3.1.0) with ESMTP id i9FGsolF026508 sender Richard_D_Levine@raytheon.com for ; Fri, 15 Oct 2004 16:54:51 GMT Subject: Does PostgreSQL run with Oracle? To: X-Mailer: Lotus Notes Release 5.0.8 June 18, 2001 Message-ID: From: Richard_D_Levine@raytheon.com Date: Fri, 15 Oct 2004 11:54:44 -0500 X-MIMETrack: Serialize by Router on NotesServer5/HDC(Release 5.0.13a |April 8, 2004) at 10/15/2004 11:54:51 AM MIME-Version: 1.0 Content-type: text/plain; charset=us-ascii X-SPAM: 0.00 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.3 tagged_above=0.0 required=5.0 tests=NO_REAL_NAME X-Spam-Level: X-Archive-Number: 200410/220 X-Sequence-Number: 8696 My basic question to the community is "is PostgreSQL approximately as fast as Oracle?" I don't want benchmarks, they're BS. I want a gut feel from this community because I know many of you are in mixed shops that run both products, or have had experience with both. I fully intend to tune, vacuum, analyze, size buffers, etc. I've read what people have written on the topic, and from that my gut feel is that using PostgreSQL will not adversely affect performance of my application versus Oracle. I know it won't adversely affect my pocket book. I also know that requests for help will be quick, clear, and multifaceted. I'm currently running single processor UltraSPARC workstations, and intend to use Intel Arch laptops and Linux. The application is a big turnkey workstation app. I know the hardware switch alone will enhance performance, and may do so to the point where even a slower database will still be adequate. Whadyall think? Thanks, Rick From pgsql-performance-owner@postgresql.org Fri Oct 15 17:58:13 2004 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 7C81532BEF9 for ; Fri, 15 Oct 2004 17:58:11 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 92621-03 for ; Fri, 15 Oct 2004 16:57:52 +0000 (GMT) Received: from candle.pha.pa.us (candle.pha.pa.us [207.106.42.251]) by svr1.postgresql.org (Postfix) with ESMTP id E561E32BEF3 for ; Fri, 15 Oct 2004 17:57:51 +0100 (BST) Received: (from pgman@localhost) by candle.pha.pa.us (8.11.6/8.11.6) id i9FGvZQ13745; Fri, 15 Oct 2004 12:57:35 -0400 (EDT) From: Bruce Momjian Message-Id: <200410151657.i9FGvZQ13745@candle.pha.pa.us> Subject: Re: [Testperf-general] Re: First set of OSDL Shared Memscalability In-Reply-To: <10202.1097858894@sss.pgh.pa.us> To: Tom Lane Date: Fri, 15 Oct 2004 12:57:35 -0400 (EDT) Cc: Simon Riggs , "Timothy D. Witham" , josh@agliodbs.com, pgsql-performance@postgresql.org, testperf-general@pgfoundry.org X-Mailer: ELM [version 2.4ME+ PL108 (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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/221 X-Sequence-Number: 8697 Tom Lane wrote: > > I've been thinking about implementing a scheme that helps you decide how > > big the shared_buffers SHOULD BE, by making the LRU list bigger than the > > cache itself, so you'd be able to see whether there is beneficial effect in > > increasing shared_buffers. > > ARC already keeps such a list --- couldn't you learn what you want to > know from the existing data structure? It'd be fairly cool if we could > put out warnings "you ought to increase shared_buffers" analogous to the > existing facility for noting excessive checkpointing. Agreed. ARC already keeps a list of buffers it had to push out recently so if it needs them again soon it knows its sizing of recent/frequent might be off (I think). Anyway, such a log report would be super-cool, say if you pushed out a buffer and needed it very soon, and the ARC buffers are already at their maximum for that buffer pool. -- 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 Fri Oct 15 18:02:39 2004 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 5EE0332BFD4 for ; Fri, 15 Oct 2004 18:02:38 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 93458-08 for ; Fri, 15 Oct 2004 17:02:26 +0000 (GMT) Received: from candle.pha.pa.us (candle.pha.pa.us [207.106.42.251]) by svr1.postgresql.org (Postfix) with ESMTP id 81327329F39 for ; Fri, 15 Oct 2004 18:02:24 +0100 (BST) Received: (from pgman@localhost) by candle.pha.pa.us (8.11.6/8.11.6) id i9FH2Kj19214; Fri, 15 Oct 2004 13:02:20 -0400 (EDT) From: Bruce Momjian Message-Id: <200410151702.i9FH2Kj19214@candle.pha.pa.us> Subject: Re: Does PostgreSQL run with Oracle? In-Reply-To: To: Richard_D_Levine@raytheon.com Date: Fri, 15 Oct 2004 13:02:20 -0400 (EDT) Cc: pgsql-performance@postgresql.org X-Mailer: ELM [version 2.4ME+ PL108 (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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/222 X-Sequence-Number: 8698 Richard_D_Levine@raytheon.com wrote: > My basic question to the community is "is PostgreSQL approximately as fast > as Oracle?" > > I don't want benchmarks, they're BS. I want a gut feel from this community > because I know many of you are in mixed shops that run both products, or > have had experience with both. > > I fully intend to tune, vacuum, analyze, size buffers, etc. I've read what > people have written on the topic, and from that my gut feel is that using > PostgreSQL will not adversely affect performance of my application versus > Oracle. I know it won't adversely affect my pocket book. I also know that > requests for help will be quick, clear, and multifaceted. > > I'm currently running single processor UltraSPARC workstations, and intend > to use Intel Arch laptops and Linux. The application is a big turnkey > workstation app. I know the hardware switch alone will enhance > performance, and may do so to the point where even a slower database will > still be adequate. I have always been told we are +/- 10% of Oracle. That's what I say at talks and no one has disputed that. We are +10-30% faster than Informix. -- 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 Fri Oct 15 18:08:23 2004 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 A3F7F32C05E for ; Fri, 15 Oct 2004 18:08:21 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 94863-09 for ; Fri, 15 Oct 2004 17:08:15 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 21D5F32C03F for ; Fri, 15 Oct 2004 18:08:13 +0100 (BST) 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 i9FH81KD010424; Fri, 15 Oct 2004 13:08:01 -0400 (EDT) To: Bernd Cc: pgsql-performance@postgresql.org Subject: Re: Select with qualified join condition / Batch inserts In-reply-to: <200410151225.26083.bernd_pg@genedata.com> References: <200410151225.26083.bernd_pg@genedata.com> Comments: In-reply-to Bernd message dated "Fri, 15 Oct 2004 12:25:26 +0200" Date: Fri, 15 Oct 2004 13:08:01 -0400 Message-ID: <10423.1097860081@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/223 X-Sequence-Number: 8699 Bernd writes: > 1/ The following query takes about 5 sec. with postrgres whereas on Oracle it > executes in about 30 ms (although both tables only contain 200 k records in > the postgres version). What does EXPLAIN ANALYZE have to say about it? Have you ANALYZEd the tables involved in the query? You would in any case be very well advised to change the "numeric" columns to integer, bigint, or smallint when appropriate. There is a substantial performance advantage to using the simple integral datatypes instead of the general numeric type. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 15 18:13:35 2004 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 345D332BEEA for ; Fri, 15 Oct 2004 18:13:34 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 97879-06 for ; Fri, 15 Oct 2004 17:13:22 +0000 (GMT) Received: from gp.word-to-the-wise.com (gp.word-to-the-wise.com [64.71.176.18]) by svr1.postgresql.org (Postfix) with ESMTP id 205C932BC0E for ; Fri, 15 Oct 2004 18:13:21 +0100 (BST) Received: by gp.word-to-the-wise.com (Postfix, from userid 500) id E694A900015; Fri, 15 Oct 2004 10:19:48 -0700 (PDT) Date: Fri, 15 Oct 2004 10:19:48 -0700 From: Steve Atkins To: pgsql-performance@postgresql.org Subject: Re: Does PostgreSQL run with Oracle? Message-ID: <20041015171948.GA14759@gp.word-to-the-wise.com> References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.4.1i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/224 X-Sequence-Number: 8700 On Fri, Oct 15, 2004 at 11:54:44AM -0500, Richard_D_Levine@raytheon.com wrote: > My basic question to the community is "is PostgreSQL approximately as fast > as Oracle?" > I'm currently running single processor UltraSPARC workstations, and intend > to use Intel Arch laptops and Linux. The application is a big turnkey > workstation app. I know the hardware switch alone will enhance > performance, and may do so to the point where even a slower database will > still be adequate. I have found that PostgreSQL seems to perform poorly on Solaris/SPARC (less so after recent improvements, but still...) compared to x86 systems - more so than the delta between Oracle on the two platforms. Just a gut impression, but it might mean that comparing the two databases on SPARC may not be that useful comparison if you're planning to move to x86. Cheers, Steve From pgsql-performance-owner@postgresql.org Fri Oct 15 18:51:18 2004 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 50680329FBD for ; Fri, 15 Oct 2004 18:51:12 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 10339-01 for ; Fri, 15 Oct 2004 17:51:06 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id E97EE32A07C for ; Fri, 15 Oct 2004 18:51:05 +0100 (BST) 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 i9FHp50c010906; Fri, 15 Oct 2004 13:51:05 -0400 (EDT) To: Doug Y Cc: pgsql-performance@postgresql.org Subject: Re: Tuning shared_buffers with ipcs ? In-reply-to: <416FF5BF.8060200@ptd.net> References: <416FF5BF.8060200@ptd.net> Comments: In-reply-to Doug Y message dated "Fri, 15 Oct 2004 12:07:27 -0400" Date: Fri, 15 Oct 2004 13:51:05 -0400 Message-ID: <10905.1097862665@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/225 X-Sequence-Number: 8701 Doug Y writes: > I've seen a couple references to using ipcs to help properly size > shared_buffers. I have not seen any such claim, and I do not see any way offhand that ipcs could help. > I tried all of the dash commands in the ipcs man page, and the only one > that might give a clue is ipcs -t which shows the time the semaphores > were last used. If you look at the example I give below, it appears as > if I'm only using 4 of the 17 semaphores (PG was started on Oct 8). This might tell you something about how many concurrent backends you've used, but nothing about how many shared buffers you need. regards, tom lane From pgsql-hackers-win32-owner@postgresql.org Fri Oct 15 18:55:22 2004 X-Original-To: pgsql-hackers-win32-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7B07432C0D3 for ; Fri, 15 Oct 2004 18:55:19 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 09572-05 for ; Fri, 15 Oct 2004 17:55:13 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 430A232C037 for ; Fri, 15 Oct 2004 18:55:13 +0100 (BST) 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 i9FHtCE1011491; Fri, 15 Oct 2004 13:55:12 -0400 (EDT) To: "Merlin Moncure" Cc: "MikeSmialek2@Hotmail.com" , pgsql-hackers-win32@postgresql.org, pgsql-performance@postgresql.org, "Magnus Hagander" Subject: Re: [PERFORM] Performance on Win32 vs Cygwin In-reply-to: <6EE64EF3AB31D5448D0007DD34EEB3412A7506@Herge.rcsinc.local> References: <6EE64EF3AB31D5448D0007DD34EEB3412A7506@Herge.rcsinc.local> Comments: In-reply-to "Merlin Moncure" message dated "Fri, 15 Oct 2004 12:22:40 -0400" Date: Fri, 15 Oct 2004 13:55:11 -0400 Message-ID: <11490.1097862911@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/73 X-Sequence-Number: 2363 "Merlin Moncure" writes: > I'm pretty certain that previous to 8.0 no win32 based postgesql > properly sync()ed the files. Win32 does not have sync(), and it is > impossible to emulate it without relying on the application to track > which files to sync. 8.0 does this because it fsync()s the files > individually. Therefore, benchmarking fsync=on on 8.0 to a <8.0 version > of windows is not apples to apples. This includes, by the way, the SFU > based port of postgresql because they didn't implement sync() there, > either. This is all true, but for performance testing I am not sure that you'd notice much difference, because the sync or lack of it only happens within checkpoints. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 15 19:06:33 2004 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 226BC32C0F7 for ; Fri, 15 Oct 2004 19:06:31 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 14102-04 for ; Fri, 15 Oct 2004 18:06:20 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id ECA3732C0D5 for ; Fri, 15 Oct 2004 19:06:19 +0100 (BST) 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 i9FI6Iu9011591; Fri, 15 Oct 2004 14:06:18 -0400 (EDT) To: Richard_D_Levine@raytheon.com Cc: pgsql-performance@postgresql.org Subject: Re: Does PostgreSQL run with Oracle? In-reply-to: References: Comments: In-reply-to Richard_D_Levine@raytheon.com message dated "Fri, 15 Oct 2004 11:54:44 -0500" Date: Fri, 15 Oct 2004 14:06:18 -0400 Message-ID: <11590.1097863578@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/227 X-Sequence-Number: 8703 Richard_D_Levine@raytheon.com writes: > My basic question to the community is "is PostgreSQL approximately as fast > as Oracle?" The anecdotal evidence I've seen leaves me with the impression that when you first take an Oracle-based app and drop it into Postgres, it won't perform particularly well, but with tuning and tweaking you can roughly equal and often exceed the original performance. The two DBs are enough unalike that a database schema that's been tuned for Oracle is probably mistuned for Postgres. You will certainly find "some things are faster, some are slower" at the end of the day, but we've had lots of satisfied switchers ... regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 15 19:16:13 2004 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 2B0DF329EEB for ; Fri, 15 Oct 2004 19:16:11 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 16384-10 for ; Fri, 15 Oct 2004 18:16:03 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.199]) by svr1.postgresql.org (Postfix) with ESMTP id 0B2C7329E7E for ; Fri, 15 Oct 2004 19:15:59 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 73so6290rnk for ; Fri, 15 Oct 2004 11:15:57 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; 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:references; b=o+W9CsWbxlsYzGJhpP8is7mB6XloZkGGB4zvKlb+ttVSS0dhwq8ktoZE1Eqkuyi+uX53Fz5+9LBTL1n4XGt5c8ktWbaZ+Eo8RG69nj5YVEUxjFpvZQOxTJxlP07FkbCAc1Gc4mT1ln0In8W4N8KCL0YJT4EJGewbpahpoT6F+FU Received: by 10.38.72.80 with SMTP id u80mr354382rna; Fri, 15 Oct 2004 11:15:25 -0700 (PDT) Received: by 10.38.208.68 with HTTP; Fri, 15 Oct 2004 11:15:25 -0700 (PDT) Message-ID: <9662496504101511152592fe4f@mail.gmail.com> Date: Fri, 15 Oct 2004 11:15:25 -0700 From: Marc Slemko Reply-To: Marc Slemko To: "richard_d_levine@raytheon.com" Subject: Re: Does PostgreSQL run with Oracle? Cc: pgsql-performance@postgresql.org In-Reply-To: Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests=TO_ADDRESS_EQ_REAL X-Spam-Level: X-Archive-Number: 200410/228 X-Sequence-Number: 8704 On Fri, 15 Oct 2004 11:54:44 -0500, richard_d_levine@raytheon.com wrote: > My basic question to the community is "is PostgreSQL approximately as fast > as Oracle?" > > I don't want benchmarks, they're BS. I want a gut feel from this community > because I know many of you are in mixed shops that run both products, or > have had experience with both. That all depends on exactly what your application needs to do. There are many more features that Oracle has and postgres doesn't than vice versa. If you need to do something with your data that isn't possible to do as efficiently without one of those features, then yes postgresql can be much slower. If you don't need any such features, it can be ballpark, until you start getting into fairly hefty hardware. From pgsql-performance-owner@postgresql.org Fri Oct 15 19:51:45 2004 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 828B032BDCC for ; Fri, 15 Oct 2004 19:51:44 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 25036-08 for ; Fri, 15 Oct 2004 18:51:37 +0000 (GMT) Received: from neomail07.traderonline.com (email.traderonline.com [65.213.231.10]) by svr1.postgresql.org (Postfix) with ESMTP id EB82932BD81 for ; Fri, 15 Oct 2004 19:51:36 +0100 (BST) Received: from [192.168.1.68] (64-139-89-109-ubr02b-epensb01-pa.hfc.comcastbusiness.net [64.139.89.109]) by neomail07.traderonline.com (8.12.10/8.12.8) with ESMTP id i9FIpZOY004409; Fri, 15 Oct 2004 14:51:35 -0400 Message-ID: <41701C35.4060808@ptd.net> Date: Fri, 15 Oct 2004 14:51:33 -0400 From: Doug Y User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: pgsql-performance@postgresql.org Subject: Re: Tuning shared_buffers with ipcs ? References: <416FF5BF.8060200@ptd.net> <10905.1097862665@sss.pgh.pa.us> In-Reply-To: <10905.1097862665@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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/229 X-Sequence-Number: 8705 Tom Lane wrote: >Doug Y writes: > > >>I've seen a couple references to using ipcs to help properly size >>shared_buffers. >> >> > >I have not seen any such claim, and I do not see any way offhand that >ipcs could help. > > Directly from: http://www.varlena.com/varlena/GeneralBits/Tidbits/annotated_conf_e.html "As a rule of thumb, observe shared memory usage of PostgreSQL with tools like ipcs and determine the setting." I've seen references in the admin >>I tried all of the dash commands in the ipcs man page, and the only one >>that might give a clue is ipcs -t which shows the time the semaphores >>were last used. If you look at the example I give below, it appears as >>if I'm only using 4 of the 17 semaphores (PG was started on Oct 8). >> >> > >This might tell you something about how many concurrent backends you've >used, but nothing about how many shared buffers you need. > > Thats strange, I know I've had more than 4 concurrent connections on that box... (I just checked and there were at least a dozen). A mirror DB with the same config also has the same basic output from ipcs, except that it has times for 11 of the 17 arrays slots and most of them are the time when we do our backup dump (which makes sense that it would require more memory at that time.) > regards, tom lane > > > > I'm not saying you're wrong, because I don't know how the nitty gritty stuff works, I'm just trying to find something to work with, since presently there isn't anything other than anecdotal evidence. From what I've inferred, there seems to be some circumstantial evidence supporting my theory. Thanks. From pgsql-performance-owner@postgresql.org Fri Oct 15 20:24:34 2004 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 CD1E7329EEB for ; Fri, 15 Oct 2004 20:24:32 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 35839-08 for ; Fri, 15 Oct 2004 19:24:25 +0000 (GMT) Received: from cmailg3.svr.pol.co.uk (cmailg3.svr.pol.co.uk [195.92.195.173]) by svr1.postgresql.org (Postfix) with ESMTP id 0C0C4329E4A for ; Fri, 15 Oct 2004 20:24:26 +0100 (BST) Received: from modem-3812.rhino.dialup.pol.co.uk ([62.137.110.228] helo=happyplace) by cmailg3.svr.pol.co.uk with smtp (Exim 4.14) id 1CIXga-0008Eq-2P; Fri, 15 Oct 2004 20:24:20 +0100 From: "Simon Riggs" To: "Bruce Momjian" , "Tom Lane" Cc: "Timothy D. Witham" , , , Subject: Re: [Testperf-general] Re: First set of OSDL Shared Memscalability results, some wierdness ... Date: Fri, 15 Oct 2004 20:42:39 +0100 Message-ID: MIME-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" 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) Importance: Normal In-Reply-To: <200410151657.i9FGvZQ13745@candle.pha.pa.us> X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/230 X-Sequence-Number: 8706 > Bruce Momjian > Tom Lane wrote: > > > I've been thinking about implementing a scheme that helps you > decide how > > > big the shared_buffers SHOULD BE, by making the LRU list > bigger than the > > > cache itself, so you'd be able to see whether there is > beneficial effect in > > > increasing shared_buffers. > > > > ARC already keeps such a list --- couldn't you learn what you want to > > know from the existing data structure? It'd be fairly cool if we could > > put out warnings "you ought to increase shared_buffers" analogous to the > > existing facility for noting excessive checkpointing. First off, many thanks for taking the time to provide the real detail on the code. That gives us some much needed direction in interpreting the oprofile output. > > Agreed. ARC already keeps a list of buffers it had to push out recently > so if it needs them again soon it knows its sizing of recent/frequent > might be off (I think). Anyway, such a log report would be super-cool, > say if you pushed out a buffer and needed it very soon, and the ARC > buffers are already at their maximum for that buffer pool. > OK, I guess I hadn't realised we were half-way there. The "increase shared_buffers" warning would be useful, but it would be much cooler to have some guidance as to how big to set it, especially since this requires a restart of the server. What I had in mind was a way of keeping track of how the buffer cache hit ratio would look at various sizes of shared_buffers, for example 50%, 80%, 120%, 150%, 200% and 400% say. That way you'd stand a chance of plotting the curve and thereby assessing how much memory could be allocated. I've got a few ideas, but I need to check out the code first. I'll investigate both simple/complex options as an 8.1 feature. Best Regards, Simon Riggs From pgsql-performance-owner@postgresql.org Fri Oct 15 20:43:05 2004 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 D345A32B18D for ; Fri, 15 Oct 2004 20:43:00 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 42641-09 for ; Fri, 15 Oct 2004 19:42:54 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 0C46432BDF1 for ; Fri, 15 Oct 2004 20:42:54 +0100 (BST) 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 6507959; Fri, 15 Oct 2004 12:44:18 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: [Testperf-general] Re: First set of OSDL Shared Memscalability results, some wierdness ... Date: Fri, 15 Oct 2004 12:44:52 -0700 User-Agent: KMail/1.6.2 Cc: "Simon Riggs" , "Bruce Momjian" , "Tom Lane" , "Timothy D. Witham" , References: In-Reply-To: MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410151244.52735.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/231 X-Sequence-Number: 8707 People: > First off, many thanks for taking the time to provide the real detail on > the code. > > That gives us some much needed direction in interpreting the oprofile > output. I have some oProfile output; however, it's in 2 out of 20 tests I ran recently and I need to get them sorted out. --Josh -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Fri Oct 15 20:45:57 2004 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 8EDDD32BD54 for ; Fri, 15 Oct 2004 20:45:55 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 43116-10 for ; Fri, 15 Oct 2004 19:45:47 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.203]) by svr1.postgresql.org (Postfix) with ESMTP id B139132A1B4 for ; Fri, 15 Oct 2004 20:45:46 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 77so12909rnk for ; Fri, 15 Oct 2004 12:45:46 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; 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:references; b=TuQkRQd9H2CsEZl+pfcz0aXQeEExeYhscgjB88Z9Vc2Be/LTmX33qRVS54roYQWhikhkHiilFOj+pJmy03rKmTae7AJkWUyLwOt2BoA7qVnw+yHMjl80ZMXp3XJHwHyQF2oW7loybIDzkTQC9xrilfgQ02/S4WwWdpwcx7XO1Ck Received: by 10.38.99.68 with SMTP id w68mr98078rnb; Fri, 15 Oct 2004 12:45:11 -0700 (PDT) Received: by 10.38.126.17 with HTTP; Fri, 15 Oct 2004 12:45:11 -0700 (PDT) Message-ID: Date: Fri, 15 Oct 2004 15:45:11 -0400 From: Mike Rylander Reply-To: Mike Rylander To: "richard_d_levine@raytheon.com" Subject: Re: Does PostgreSQL run with Oracle? Cc: pgsql-performance@postgresql.org In-Reply-To: Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests=TO_ADDRESS_EQ_REAL X-Spam-Level: X-Archive-Number: 200410/232 X-Sequence-Number: 8708 On Fri, 15 Oct 2004 11:54:44 -0500, richard_d_levine@raytheon.com wrote: > My basic question to the community is "is PostgreSQL approximately as fast > as Oracle?" My personal experience comparing PG to Oracle is across platforms, Oracle on Sun/Solaris (2.7, quad-proc R440) and PG on Intel/Linux (2.6 kernel, dual P3/1GHz). When both were tuned for the specific app I saw a 45% speedup after switching to PG. This was with a customized CRM and System Monitoring application serving ~40,000 trouble tickets and monitoring 5,000 metric datapoints every 5-30 minutes. The hardware was definitely not comparable (the Intel boxes have more horsepower and faster disks), but dollar for dollar, including support costs, PG is the winner by a BIG margin. YMMV, of course, and my results are apparently above average. Another big plus I found was that PG is much easier to admin as long as you turn on pg_autovacuum. --miker From pgsql-performance-owner@postgresql.org Fri Oct 15 20:53:38 2004 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 6B80B32BE02 for ; Fri, 15 Oct 2004 20:53:36 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 47213-07 for ; Fri, 15 Oct 2004 19:53:28 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id C0DE932BDF7 for ; Fri, 15 Oct 2004 20:53:29 +0100 (BST) 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 i9FJrSJ0020132; Fri, 15 Oct 2004 15:53:28 -0400 (EDT) To: Doug Y Cc: pgsql-performance@postgresql.org Subject: Re: Tuning shared_buffers with ipcs ? In-reply-to: <41701C35.4060808@ptd.net> References: <416FF5BF.8060200@ptd.net> <10905.1097862665@sss.pgh.pa.us> <41701C35.4060808@ptd.net> Comments: In-reply-to Doug Y message dated "Fri, 15 Oct 2004 14:51:33 -0400" Date: Fri, 15 Oct 2004 15:53:28 -0400 Message-ID: <20131.1097870008@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/233 X-Sequence-Number: 8709 Doug Y writes: > Tom Lane wrote: >> I have not seen any such claim, and I do not see any way offhand that >> ipcs could help. >> > Directly from: > http://www.varlena.com/varlena/GeneralBits/Tidbits/annotated_conf_e.html > "As a rule of thumb, observe shared memory usage of PostgreSQL with > tools like ipcs and determine the setting." [ shrug ... ] So ask elein why she thinks that will help. >> This might tell you something about how many concurrent backends you've >> used, but nothing about how many shared buffers you need. >> > Thats strange, I know I've had more than 4 concurrent connections on > that box... (I just checked and there were at least a dozen). There is more than one per-backend semaphore per semaphore set, 16 per set if memory serves; so the ipcs evidence points to a maximum of between 49 and 64 concurrently active backends. It's not telling you a darn thing about appropriate shared_buffers settings, however. > A mirror DB with the same config also has the same basic output from > ipcs, except that it has times for 11 of the 17 arrays slots and most > of them are the time when we do our backup dump (which makes sense > that it would require more memory at that time.) That doesn't follow either. I think you may have some bottleneck that causes client requests to pile up during a backup dump. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 15 21:09:21 2004 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 6B80632BDA7 for ; Fri, 15 Oct 2004 21:09:16 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 52686-06 for ; Fri, 15 Oct 2004 20:09:08 +0000 (GMT) Received: from mail.trippynames.com (unknown [38.113.223.19]) by svr1.postgresql.org (Postfix) with ESMTP id B9FCD32BD54 for ; Fri, 15 Oct 2004 21:09:08 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by mail.trippynames.com (Postfix) with ESMTP id C00F1A10C6; Fri, 15 Oct 2004 13:09:07 -0700 (PDT) Received: from mail.trippynames.com ([127.0.0.1]) by localhost (rand.nxad.com [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 32974-05; Fri, 15 Oct 2004 13:09:04 -0700 (PDT) Received: from [192.168.1.104] (wbar4.sjo1-4.28.216.220.sjo1.dsl-verizon.net [4.28.216.220]) by mail.trippynames.com (Postfix) with ESMTP id C0F51A10C0; Fri, 15 Oct 2004 13:09:03 -0700 (PDT) In-Reply-To: <20041015010841.GE665@filer> References: <157f64840410141725162e43b5@mail.gmail.com> <20041015010841.GE665@filer> Mime-Version: 1.0 (Apple Message framework v619) Content-Type: text/plain; charset=US-ASCII; format=flowed Message-Id: <0E3BB9CB-1EE6-11D9-A0BB-000A95C705DC@chittenden.org> Content-Transfer-Encoding: 7bit Cc: pgsql-performance@postgresql.org From: Sean Chittenden Subject: Re: mmap (was First set of OSDL Shared Mem scalability results, some wierdness ... Date: Fri, 15 Oct 2004 13:09:01 -0700 To: Kevin Brown X-Mailer: Apple Mail (2.619) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/234 X-Sequence-Number: 8710 >> pg to my mind is unique in not trying to avoid OS buffering. Other >> dbmses spend a substantial effort to create a virtual OS (task >> management, I/O drivers, etc.) both in code and support. Choosing mmap >> seems such a limiting an option - it adds OS dependency and limits >> kernel developer options (2G limits, global mlock serializations, >> porting problems, inability to schedule or parallelize I/O, still >> having to coordinate writers and readers). 2G limits? That must be a Linux limitation, not a limitation with mmap(2). On OS-X and FreeBSD it's anywhere from 4GB to ... well, whatever the 64bit limit is (which is bigger than any data file in $PGDATA). An mlock(2) serialization problem is going to be cheaper than hitting the disk in nearly all cases and should be no worse than a context switch or semaphore (what we use for the current locking scheme), of which PostgreSQL causes plenty of 'em because it's multi-process, not multi-threaded. Coordination of data isn't necessary if you mmap(2) data as a private block, which takes a snapshot of the page at the time you make the mmap(2) call and gets copied only when the page is written to. More on that later. > I'm not sure I entirely agree with this. Whether you access a file > via mmap() or via read(), the end result is that you still have to > access it, and since PG has significant chunks of system-dependent > code that it heavily relies on as it is (e.g., locking mechanisms, > shared memory), writing the I/O subsystem in a similar way doesn't > seem to me to be that much of a stretch (especially since PG already > has the storage manager), though it might involve quite a bit of work. Obviously you have to access the file on the hard drive, but you're forgetting an enormous advantage of mmap(2). With a read(2) system call, the program has to allocate space for the read(2), then it copies data from the kernel into the allocated memory in the userland's newly allocated memory location. With mmap(2) there is no second copy. Let's look at what happens with a read(2) call. To read(2) data you have to have a block of memory to copy data into. Assume your OS of choice has a good malloc(3) implementation and it only needs to call brk(2) once to extend the process's memory address after the first malloc(3) call. There's your first system call, which guarantees one context switch. The second hit, a much larger hit, is the actual read(2) call itself, wherein the kernel has to copy the data twice: once into a kernel buffer, then from the kernel buffer into the userland's memory space. Yuk. Webserver's figured this out long ago that read(2) is slow and evil in terms of performance. Apache uses mmap(2) to send static files at performance levels that don't suck and is actually quite fast (in terms of responsiveness, I'm not talking about Apache's parallelism/concurrency performance levels... which in 1.X aren't great). mmap(2) is a totally different animal in that you don't ever need to make calls to read(2): mmap(2) is used in place of those calls (With #ifdef and a good abstraction, the rest of PostgreSQL wouldn't know it was working with a page of mmap(2)'ed data or need to know that it is). Instead you mmap(2) a file descriptor and the kernel does some heavy lifting/optimized magic in its VM. The kernel reads the file descriptor and places the data it reads into its buffer (exactly the same as what happens with read(2)), but, instead of copying the data to the userspace, mmap(2) adjusts the process's address space and maps the address of the kernel buffer into the process's address space. No copying necessary. The savings here are *huge*! Depending on the mmap(2) implementation, the VM may not even get a page from disk until its actually needed. So, lets say you mmap(2) a 16M file. The address space picks up an extra 16M of bits that the process *can* use, but doesn't necessarily use. So if a user reads only ten pages out of a 16MB file, only 10 pages (10 * getpagesize()), or usually 40,960K, which is 0.24% the amount of disk access (((4096 * 10) / (16 *1024 * 1024)) * 100). Did I forget to mention that if the file is already in the kernel's buffers, there's no need for the kernel to access the hard drive? Another big win for data that's hot/frequently accessed. There's another large savings if the machine is doing network IO too... > As for parallelization of I/O, the use of mmap() for reads should > signficantly improve parallelization -- now instead of issuing read() > system calls, possibly for the same set of blocks, all the backends > would essentially be examining the same data directly. The > performance improvements as a result of accessing the kernel's cache > pages directly instead of having it do buffer copies to process-local > memory should increase as concurrency goes up. But see below. That's kinda true... though not quite correct. The improvement in IO concurrency comes from zero-socket-copy operations from the disk to the network controller. If a write(2) system call is issued on a page of mmap(2)'ed data (and your operating system supports it, I know FreeBSD does, but don't think Linux does), then the page of data is DMA'ed by the network controller and sent out without the data needing to be copied into the network controller's buffer. So, instead of the CPU copying data from the OS's buffer to a kernel buffer, the network card grabs the chunk of data in one interrupt because of the DMA (direct memory access). This is a pretty big deal for web serving, but if you've got a database sending large sets of data over the network, assuming the network isn't the bottle neck, this results in a heafty performance boost (that won't be noticed by most until they're running huge, very busy installations). This optimization comes for free and without needing to add one line of code to an application once mmap(2) has been added to an application. >> More to the point, I think it is very hard to effectively coordinate >> multithreaded I/O, and mmap seems used mostly to manage relatively >> simple scenarios. > > PG already manages and coordinates multithreaded I/O. The mechanisms > used to coordinate writes needn't change at all. But the way reads > are done relative to writes might have to be rethought, since an > mmap()ed buffer always reflects what's actually in kernel space at the > time the buffer is accessed, while a buffer retrieved via read() > reflects the state of the file at the time of the read(). If it's > necessary for the state of the buffers to be fixed at examination > time, then mmap() will be at best a draw, not a win. Here's where things can get interesting from a transaction stand point. Your statement is correct up until you make the assertion that a page needs to be fixed. If you're doing a read(2) transaction, mmap(2) a region and set the MAP_PRIVATE flag so the ground won't change underneath you. No copying of this page is done by the kernel unless it gets written to. If you're doing a write(2) or are directly scribbling on an mmap(2)'ed page[1], you need to grab some kind of an exclusive lock on the page/file (mlock(2) is going to be no more expensive than a semaphore, but probably less expensive). We already do that with semaphores, however. So for databases that don't have high contention for the same page/file of data, there are no additional copies made. When a piece of data is written, a page is duplicated before it gets scribbled on, but the application never knows this happens. The next time a process mmap(2)'s a region of memory that's been written to, it'll get the updated data without any need to flush a cache or mark pages as dirty: the operating system does all of this for us (and probably faster too). mmap(2) implementations are, IMHO, more optimized that shared memory implementations (mmap(2) is a VM function, which gets many eyes to look it over and is always being tuned, whereas shared mem is a bastardized subsystem that works, but isn't integral to any performance areas in the kernel so it gets neglected. Just my observations from the *BSD commit lists. Linux it may be different). [1] I forgot to mention earlier, you don't have to write(2) data to a file if it's mmap(2)'ed, you can change the contents of an mmap(2)'ed region, then msync(2) it back to disk (to ensure it gets written out) or let the last munmap(2) call do that for you (which would be just as dangerous as running without fsync... but would result in an additional performance boost). >> mmap doesn't look that promising. > > This ultimately depends on two things: how much time is spent copying > buffers around in kernel memory, and how much advantage can be gained > by freeing up the memory used by the backends to store the > backend-local copies of the disk pages they use (and thus making that > memory available to the kernel to use for additional disk buffering). Someone on IRC pointed me to some OSDL benchmarks, which broke down where time is being spent. Want to know what the most expensive part of PostgreSQL is? *drum roll* http://khack.osdl.org/stp/297960/profile/DBT_2_Profile-tick.sort 3967393 total 1.7735 2331284 default_idle 36426.3125 825716 do_sigaction 1290.1813 133126 __copy_from_user_ll 1040.0469 97780 __copy_to_user_ll 763.9062 43135 finish_task_switch 269.5938 30973 do_anonymous_page 62.4456 24175 scsi_request_fn 22.2197 23355 __do_softirq 121.6406 17039 __wake_up 133.1172 16527 __make_request 10.8730 9823 try_to_wake_up 13.6431 9525 generic_unplug_device 66.1458 8799 find_get_page 78.5625 7878 scsi_end_request 30.7734 Copying data to/from userspace and signal handling!!!! Let's hear it for the need for mmap(2)!!! *crowd goes wild* > The gains from the former are likely small. The gains from the latter > are probably also small, but harder to estimate. I disagree. > The use of mmap() is probably one of those optimizations that should > be done when there's little else left to optimize, because the > potential gains are possibly (if not probably) relatively small and > the amount of work involved may be quite large. If system/kernel time is where most of your database spends its time, then mmap(2) is a huge optimization that is very much worth pursuing. It's stable (nearly all webservers use it, notably Apache), widely deployed, POSIX specified (granted not all implementations are 100% consistent, but that's an OS bug and mmap(2) doesn't have to be turned on for those platforms: it's no worse than where we are now), and well optimized by operating system hackers. I guarantee that your operating system of choice has a faster VM and disk cache than PostgreSQL's userland cache, nevermind using the OSs buffers leads to many performance boosts as the OS can short-circuit common pathways that would require data copying (ex: zero-socket-copy operations and copying data to/from userland). mmap(2) isn't a panacea or replacement for good software design, but it certainly does make IO operations vastly faster, which is what PostgreSQL does a lot of (hence its need for a userland cache). Remember, back when PostgreSQL had its architecture thunk up, mmap(2) hardly existed in anyone's eyes, nevermind it being widely used or a POSIX function. It wasn't until Apache started using it that Operating System vendors felt the need to implement it or make it work well. Now it's integral to nearly all virtual memory implementations and a modern OS can't live without it or have it broken in any way. It would be largely beneficial to PostgreSQL to heavily utilize mmap(2). A few places it should be used include: *) Storage. It is a good idea to mmap(2) all files instead of read(2)'ing files. mmap(2) doesn't fetch a page from disk until its actually needed, which is a nifty savings. Sure it causes a fault in the kernel, but it won't the second time that page is accessed. Changes are necessary to src/backend/storage/file/, possibly src/backend/storage/freespace/ (why is it using fread(3) and not read(2)?), src/backend/storage/large_object/ can remain gimpy since people should use BYTEA instead (IMHO), src/backend/storage/page/ doesn't need changes (I don't think), src/backend/storage/smgr/ shouldn't need any modifications either. *) ARC. Why unmmap(2) data if you don't need to? With ARC, it's possible for the database to coach the operating system in what pages should be persistent. ARC's a smart algorithm for handling the needs of a database. Instead of having a cache of pages in userland, PostgreSQL would have a cache of mmap(2)'ed pages. It's shared between processes, the changes are public to external programs read(2)'ing data, and its quick. The needs for shared memory by the kernel drops to nearly nothing. The needs for mmap(2)'able space in the kernel, however, does go up. Unlike SysV shared mem, this can normally be changed on the fly. The end result would be, if a page is needed, it checks to see if its in the cache. If it is, the mmap(2)'ed page is returned. If it isn't, the page gets read(2)/mmap(2) like it currently is loaded (except in the mmap(2) case where after the data has been loaded, the page gets munmap(2)'ed). If ARC decides to keep the page, the page doesn't get munmap(2)'ed. I don't think any changes need to be made though to take advantage of mmap(2) if the changes are made in the places mentioned above in the Storage point. A few other perks: *) DIRECTIO can be used without much of a cache coherency headache since the cache of data is in the kernel, not userland. *) NFS. I'm not suggesting multiple clients use the same data directory via NFS (unless read only), but if there were a single client accessing a data directory over NFS, performance would be much better than it is today because data consistency is handled by the kernel so in flight packets for writes that get dropped or lost won't cause a slow down (mmap(2) behaves differently with NFS pages) or corruption. *) mmap(2) is conditional on the operating system's abilities, but doesn't require any architectural changes. It does change the location of the cache, from being in the userland, down in to the kernel. This is a change for database administrators, but a good one, IMHO. Previously, the operating system would be split 25% kernel, 75% user because PostgreSQL would need the available RAM for its cache. Now, that can be moved closer to the opposite, 75% kernel, 25% user because most of the memory is mmap(2)'ed pages instead of actual memory in the userland. *) Pages can be protected via PROT_(EXEC|READ|WRITE). For backends that aren't making changes to the DDL or system catalogs (permissions, etc.), pages that are loaded from the catalogs could be loaded with the protection PROT_READ, which would prevent changes to the catalogs. All DDL and permission altering commands (anything that touches the system catalogs) would then load the page with the PROT_WRITE bit set, make their changes, then PROT_READ the page again. This would provide a first line of defense against buggy programs or exploits. *) Eliminates the double caching done currently (caching in PostgreSQL and the kernel) by pushing the cache into the kernel... but without PostgreSQL knowing it's working on a page that's in the kernel. Please ask questions if you have them. -sc -- Sean Chittenden From pgsql-performance-owner@postgresql.org Fri Oct 15 21:11:23 2004 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 936A432BEBB for ; Fri, 15 Oct 2004 21:11:21 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 52512-07 for ; Fri, 15 Oct 2004 20:11:14 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 9B9F532BDFB for ; Fri, 15 Oct 2004 21:11:15 +0100 (BST) 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 6508067; Fri, 15 Oct 2004 13:12:39 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: [Testperf-general] Re: First set of OSDL Shared Memscalability results, some wierdness ... Date: Fri, 15 Oct 2004 13:13:13 -0700 User-Agent: KMail/1.6.2 Cc: Tom Lane , "Simon Riggs" , "Timothy D. Witham" , testperf-general@pgfoundry.org References: <10202.1097858894@sss.pgh.pa.us> In-Reply-To: <10202.1097858894@sss.pgh.pa.us> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable Message-Id: <200410151313.13624.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/235 X-Sequence-Number: 8711 Tom, Simon: First off, two test runs with OProfile are available at: http://khack.osdl.org/stp/298124/ http://khack.osdl.org/stp/298121/ > AtEOXact_Buffers > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0transaction commit or abo= rt > UnlockBuffers > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0transaction abort, backen= d exit Actually, this might explain the "hump" shape of the curve for this test.= =20=20 DBT2 is an OLTP test, which means that (at this scale level) it's attemptin= g=20 to do approximately 30 COMMITs per second as well as one ROLLBACK every 3= =20 seconds. When I get the tests on DBT3 running, if we see a more gentle= =20 dropoff on overallocated memory, it would indicate that the above may be a= =20 factor. --=20 --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Fri Oct 15 21:16:39 2004 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 1EFFC32BDDD for ; Fri, 15 Oct 2004 21:16:38 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 55674-01 for ; Fri, 15 Oct 2004 20:16:30 +0000 (GMT) Received: from mail.trippynames.com (unknown [38.113.223.19]) by svr1.postgresql.org (Postfix) with ESMTP id 5A26732BD81 for ; Fri, 15 Oct 2004 21:16:32 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by mail.trippynames.com (Postfix) with ESMTP id EB232A1004; Fri, 15 Oct 2004 13:16:30 -0700 (PDT) Received: from mail.trippynames.com ([127.0.0.1]) by localhost (rand.nxad.com [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 33979-04; Fri, 15 Oct 2004 13:16:29 -0700 (PDT) Received: from [192.168.1.104] (wbar4.sjo1-4.28.216.220.sjo1.dsl-verizon.net [4.28.216.220]) by mail.trippynames.com (Postfix) with ESMTP id A3633A10C0; Fri, 15 Oct 2004 13:16:29 -0700 (PDT) In-Reply-To: <5235.1097817216@sss.pgh.pa.us> References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> <4859.1097363137@sss.pgh.pa.us> <20041014202531.GD665@filer> <5235.1097817216@sss.pgh.pa.us> Mime-Version: 1.0 (Apple Message framework v619) Content-Type: text/plain; charset=US-ASCII; format=flowed Message-Id: <17EBD8C6-1EE7-11D9-A0BB-000A95C705DC@chittenden.org> Content-Transfer-Encoding: 7bit Cc: Kevin Brown , pgsql-performance@postgresql.org From: Sean Chittenden Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... Date: Fri, 15 Oct 2004 13:16:27 -0700 To: Tom Lane X-Mailer: Apple Mail (2.619) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/236 X-Sequence-Number: 8712 > this. The SUS text is a bit weaselly ("the application must ensure > correct synchronization") but the HPUX mmap man page, among others, > lays it on the line: > > It is also unspecified whether write references to a memory region > mapped with MAP_SHARED are visible to processes reading the file > and > whether writes to a file are visible to processes that have > mapped the > modified portion of that file, except for the effect of msync(). > > It might work on particular OSes but I think depending on such behavior > would be folly... Agreed. Only OSes with a coherent file system buffer cache should ever use mmap(2). In order for this to work on HPUX, msync(2) would need to be used. -sc -- Sean Chittenden From pgsql-performance-owner@postgresql.org Fri Oct 15 21:34:13 2004 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 025AE32BEE6 for ; Fri, 15 Oct 2004 21:34:11 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 59343-07 for ; Fri, 15 Oct 2004 20:34:07 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 2EBDB32BECD for ; Fri, 15 Oct 2004 21:34:08 +0100 (BST) 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 i9FKY25J026069; Fri, 15 Oct 2004 16:34:02 -0400 (EDT) To: josh@agliodbs.com Cc: pgsql-performance@postgresql.org, "Simon Riggs" , "Timothy D. Witham" , testperf-general@pgfoundry.org Subject: Re: [Testperf-general] Re: First set of OSDL Shared Memscalability results, some wierdness ... In-reply-to: <200410151313.13624.josh@agliodbs.com> References: <10202.1097858894@sss.pgh.pa.us> <200410151313.13624.josh@agliodbs.com> Comments: In-reply-to Josh Berkus message dated "Fri, 15 Oct 2004 13:13:13 -0700" Date: Fri, 15 Oct 2004 16:34:02 -0400 Message-ID: <26068.1097872442@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/237 X-Sequence-Number: 8713 Josh Berkus writes: > First off, two test runs with OProfile are available at: > http://khack.osdl.org/stp/298124/ > http://khack.osdl.org/stp/298121/ Hmm. The stuff above 1% in the first of these is Counted CPU_CLK_UNHALTED events (clocks processor is not halted) with a unit mask of 0x00 (No unit mask) count 100000 samples % app name symbol name 8522858 19.7539 vmlinux default_idle 3510225 8.1359 vmlinux recalc_sigpending_tsk 1874601 4.3449 vmlinux .text.lock.signal 1653816 3.8331 postgres SearchCatCache 1080908 2.5053 postgres AllocSetAlloc 920369 2.1332 postgres AtEOXact_Buffers 806218 1.8686 postgres OpernameGetCandidates 803125 1.8614 postgres StrategyDirtyBufferList 746123 1.7293 vmlinux __copy_from_user_ll 651978 1.5111 vmlinux __copy_to_user_ll 640511 1.4845 postgres XLogInsert 630797 1.4620 vmlinux rm_from_queue 607833 1.4088 vmlinux next_thread 436682 1.0121 postgres LWLockAcquire 419672 0.9727 postgres yyparse In the second test AtEOXact_Buffers is much lower (down around 0.57 percent) but the other suspects are similar. Since the only difference in parameters is shared_buffers (36000 vs 9000), it does look like we are approaching the point where AtEOXact_Buffers is a problem, but so far it's only a 2% drag. I suspect the reason recalc_sigpending_tsk is so high is that the original coding of PG_TRY involved saving and restoring the signal mask, which led to a whole lot of sigsetmask-type kernel calls. Is this test with beta3, or something older? Another interesting item here is the costs of __copy_from_user_ll/ __copy_to_user_ll: 36000 buffers: 746123 1.7293 vmlinux __copy_from_user_ll 651978 1.5111 vmlinux __copy_to_user_ll 9000 buffers: 866414 2.0810 vmlinux __copy_from_user_ll 852620 2.0479 vmlinux __copy_to_user_ll Presumably the higher costs for 9000 buffers reflect an increased amount of shuffling of data between kernel and user space. So 36000 is not enough to make the working set totally memory-resident, but even if we drove this cost to zero we'd only be buying a couple percent. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 15 21:36:27 2004 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 DE7A832C4CF for ; Fri, 15 Oct 2004 21:36:25 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 59307-10 for ; Fri, 15 Oct 2004 20:36:18 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 72E8E32C279 for ; Fri, 15 Oct 2004 21:36:17 +0100 (BST) 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 6508228; Fri, 15 Oct 2004 13:37:42 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: [Testperf-general] Re: First set of OSDL Shared Memscalability results, some wierdness ... Date: Fri, 15 Oct 2004 13:38:17 -0700 User-Agent: KMail/1.6.2 Cc: Tom Lane , "Simon Riggs" , "Timothy D. Witham" , testperf-general@pgfoundry.org References: <200410151313.13624.josh@agliodbs.com> <26068.1097872442@sss.pgh.pa.us> In-Reply-To: <26068.1097872442@sss.pgh.pa.us> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Message-Id: <200410151338.17277.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/238 X-Sequence-Number: 8714 Tom, > I suspect the reason recalc_sigpending_tsk is so high is that the > original coding of PG_TRY involved saving and restoring the signal mask, > which led to a whole lot of sigsetmask-type kernel calls. Is this test > with beta3, or something older? Beta3, *without* Gavin or Neil's Futex patch. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Fri Oct 15 21:40:04 2004 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 7C04132ACB1 for ; Fri, 15 Oct 2004 21:39:58 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 61786-04 for ; Fri, 15 Oct 2004 20:39:55 +0000 (GMT) Received: from mail.osdl.org (fw.osdl.org [65.172.181.6]) by svr1.postgresql.org (Postfix) with ESMTP id 49A9032A47F for ; Fri, 15 Oct 2004 21:39:54 +0100 (BST) Received: (from markw@localhost) by mail.osdl.org (8.11.6/8.11.6) id i9FKdm611366; Fri, 15 Oct 2004 13:39:48 -0700 Date: Fri, 15 Oct 2004 13:39:47 -0700 From: Mark Wong To: Sean Chittenden Cc: Kevin Brown , pgsql-performance@postgresql.org Subject: Re: mmap (was First set of OSDL Shared Mem scalability results, some wierdness ... Message-ID: <20041015133947.A9225@osdl.org> References: <157f64840410141725162e43b5@mail.gmail.com> <20041015010841.GE665@filer> <0E3BB9CB-1EE6-11D9-A0BB-000A95C705DC@chittenden.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5i In-Reply-To: <0E3BB9CB-1EE6-11D9-A0BB-000A95C705DC@chittenden.org>; from sean@chittenden.org on Fri, Oct 15, 2004 at 01:09:01PM -0700 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/239 X-Sequence-Number: 8715 On Fri, Oct 15, 2004 at 01:09:01PM -0700, Sean Chittenden wrote: [snip] > > > > This ultimately depends on two things: how much time is spent copying > > buffers around in kernel memory, and how much advantage can be gained > > by freeing up the memory used by the backends to store the > > backend-local copies of the disk pages they use (and thus making that > > memory available to the kernel to use for additional disk buffering). > > Someone on IRC pointed me to some OSDL benchmarks, which broke down > where time is being spent. Want to know what the most expensive part > of PostgreSQL is? *drum roll* > > http://khack.osdl.org/stp/297960/profile/DBT_2_Profile-tick.sort > > 3967393 total 1.7735 > 2331284 default_idle 36426.3125 > 825716 do_sigaction 1290.1813 > 133126 __copy_from_user_ll 1040.0469 > 97780 __copy_to_user_ll 763.9062 > 43135 finish_task_switch 269.5938 > 30973 do_anonymous_page 62.4456 > 24175 scsi_request_fn 22.2197 > 23355 __do_softirq 121.6406 > 17039 __wake_up 133.1172 > 16527 __make_request 10.8730 > 9823 try_to_wake_up 13.6431 > 9525 generic_unplug_device 66.1458 > 8799 find_get_page 78.5625 > 7878 scsi_end_request 30.7734 > > Copying data to/from userspace and signal handling!!!! Let's hear it > for the need for mmap(2)!!! *crowd goes wild* > [snip] I know where the do_sigaction is coming from in this particular case. Manfred Spraul tracked it to a pair of pgsignal calls in libpq. Commenting out those two calls out virtually eliminates do_sigaction from the kernel profile for this workload. I've lost track of the discussion over the past year, but I heard a rumor that it was finally addressed to some degree. I did understand it touched on a lot of other things, but can anyone summarize where that discussion has gone? Mark From pgsql-performance-owner@postgresql.org Fri Oct 15 22:18:18 2004 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 BC81A32A232 for ; Fri, 15 Oct 2004 22:18:16 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 71353-06 for ; Fri, 15 Oct 2004 21:18:13 +0000 (GMT) Received: from neomail03.traderonline.com (email.traderonline.com [65.213.231.10]) by svr1.postgresql.org (Postfix) with ESMTP id E16D632A205 for ; Fri, 15 Oct 2004 22:18:13 +0100 (BST) Received: from [192.168.1.68] (64-139-89-109-ubr02b-epensb01-pa.hfc.comcastbusiness.net [64.139.89.109]) by neomail03.traderonline.com (8.12.10/8.12.8) with ESMTP id i9FKffPG025327; Fri, 15 Oct 2004 16:41:41 -0400 Message-ID: <41703E94.3030800@ptd.net> Date: Fri, 15 Oct 2004 17:18:12 -0400 From: Doug Y User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: pgsql-performance@postgresql.org Subject: Re: Tuning shared_buffers with ipcs ? References: <416FF5BF.8060200@ptd.net> <10905.1097862665@sss.pgh.pa.us> <41701C35.4060808@ptd.net> <20131.1097870008@sss.pgh.pa.us> In-Reply-To: <20131.1097870008@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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/240 X-Sequence-Number: 8716 Tom Lane wrote: >Doug Y writes: > > >>Tom Lane wrote: >> >> >>>This might tell you something about how many concurrent backends you've >>>used, but nothing about how many shared buffers you need. >>> >>> >>> >>Thats strange, I know I've had more than 4 concurrent connections on >>that box... (I just checked and there were at least a dozen). >> >> > >There is more than one per-backend semaphore per semaphore set, 16 per >set if memory serves; so the ipcs evidence points to a maximum of >between 49 and 64 concurrently active backends. It's not telling you a >darn thing about appropriate shared_buffers settings, however. > > > >>A mirror DB with the same config also has the same basic output from >>ipcs, except that it has times for 11 of the 17 arrays slots and most >>of them are the time when we do our backup dump (which makes sense >>that it would require more memory at that time.) >> >> > >That doesn't follow either. I think you may have some bottleneck that >causes client requests to pile up during a backup dump. > > regards, tom lane > > Ok, that explains the number of arrays... max_connections / 16. Thanks... my mind works better when I can associate actual settings to effects like that. And I'm sure that performance takes a hit during out back-up dump. We're in the process of migrating them to dedicated mirror machine to run dumps/reports etc from crons so that it won't negatively affect the DB servers that get queries from the web applications. Thanks again for clarification. From pgsql-performance-owner@postgresql.org Fri Oct 15 22:22:36 2004 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 011AD329EAC for ; Fri, 15 Oct 2004 22:22:35 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 73448-04 for ; Fri, 15 Oct 2004 21:22:31 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 13A0F329E86 for ; Fri, 15 Oct 2004 22:22:32 +0100 (BST) 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 i9FLMQMP026391; Fri, 15 Oct 2004 17:22:26 -0400 (EDT) To: Sean Chittenden Cc: Kevin Brown , pgsql-performance@postgresql.org Subject: Re: mmap (was First set of OSDL Shared Mem scalability results, some wierdness ... In-reply-to: <0E3BB9CB-1EE6-11D9-A0BB-000A95C705DC@chittenden.org> References: <157f64840410141725162e43b5@mail.gmail.com> <20041015010841.GE665@filer> <0E3BB9CB-1EE6-11D9-A0BB-000A95C705DC@chittenden.org> Comments: In-reply-to Sean Chittenden message dated "Fri, 15 Oct 2004 13:09:01 -0700" Date: Fri, 15 Oct 2004 17:22:26 -0400 Message-ID: <26390.1097875346@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/241 X-Sequence-Number: 8717 Sean Chittenden writes: > Coordination of data isn't > necessary if you mmap(2) data as a private block, which takes a > snapshot of the page at the time you make the mmap(2) call and gets > copied only when the page is written to. More on that later. We cannot move to a model where different backends have different views of the same page, which seems to me to be inherent in the idea of using MAP_PRIVATE for anything. To take just one example, a backend that had mapped one btree index page some time ago could get completely confused if that page splits, because it might see the effects of the split in nearby index pages but not in the one that was split. Or it could follow an index link to a heap entry that isn't there anymore, or miss an entry it should have seen. MVCC doesn't save you from this because btree adjustments happen below the level of transactions. However the really major difficulty with using mmap is that it breaks the scheme we are currently using for WAL, because you don't have any way to restrict how soon a change in an mmap'd page will go to disk. (No, I don't believe that mlock guarantees this. It says that the page will not be removed from main memory; it does not specify that, say, the syncer won't write the contents out anyway.) > Let's look at what happens with a read(2) call. To read(2) data you > have to have a block of memory to copy data into. Assume your OS of > choice has a good malloc(3) implementation and it only needs to call > brk(2) once to extend the process's memory address after the first > malloc(3) call. There's your first system call, which guarantees one > context switch. Wrong. Our reads occur into shared memory allocated at postmaster startup, remember? > mmap(2) is a totally different animal in that you don't ever need to > make calls to read(2): mmap(2) is used in place of those calls (With > #ifdef and a good abstraction, the rest of PostgreSQL wouldn't know it > was working with a page of mmap(2)'ed data or need to know that it is). Instead, you have to worry about address space management and keeping a consistent view of the data. > ... If a write(2) system call is issued on a page of > mmap(2)'ed data (and your operating system supports it, I know FreeBSD > does, but don't think Linux does), then the page of data is DMA'ed by > the network controller and sent out without the data needing to be > copied into the network controller's buffer. Perfectly irrelevant to Postgres, since there is no situation where we'd ever write directly from a disk buffer to a socket; in the present implementation there are at least two levels of copy needed in between (datatype-specific output function and protocol message assembly). And that's not even counting the fact that any data item large enough to make the savings interesting would have been sliced, diced, and compressed by TOAST. > ... If you're doing a write(2) or are directly > scribbling on an mmap(2)'ed page[1], you need to grab some kind of an > exclusive lock on the page/file (mlock(2) is going to be no more > expensive than a semaphore, but probably less expensive). More incorrect information. The locking involved here is done by LWLockAcquire, which is significantly *less* expensive than a kernel call in the case where there is no need to block. (If you have to block, any kernel call to do so is probably about as bad as any other.) Switching over to mlock would likely make things considerably slower. In any case, you didn't actually mean to say mlock did you? It doesn't lock pages against writes by other processes AFAICS. > shared mem is a bastardized subsystem that works, but isn't integral to > any performance areas in the kernel so it gets neglected. What performance issues do you think shared memory needs to have fixed? We don't issue any shmem kernel calls after the initial shmget, so comparing the level of kernel tenseness about shmget to the level of tenseness about mmap is simply irrelevant. Perhaps the reason you don't see any traffic about this on the kernel lists is that shared memory already works fine and doesn't need any fixing. > Please ask questions if you have them. Do you have any arguments that are actually convincing? What I just read was a proposal to essentially throw away not only the entire low-level data access model, but the entire low-level locking model, and start from scratch. There is no possible way we could support both this approach and the current one, which means that we'd be permanently dropping support for all platforms without high-quality mmap implementations; and despite your enthusiasm I don't think that that category includes every interesting platform. Furthermore, you didn't give any really convincing reasons to think that the enormous effort involved would be repaid. Those oprofile reports Josh just put up showed 3% of the CPU time going into userspace/kernelspace copying. Even assuming that that number consists entirely of reads and writes of shared buffers (and of course no other kernel call ever transfers any data across that boundary ;-)), there's no way we are going to buy into this sort of project in hopes of a 3% win. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 15 22:27:53 2004 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 9CAF7329E76 for ; Fri, 15 Oct 2004 22:27:52 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 73448-10 for ; Fri, 15 Oct 2004 21:27:33 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 1350F32A750 for ; Fri, 15 Oct 2004 22:27:34 +0100 (BST) 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 i9FLRUKS026452; Fri, 15 Oct 2004 17:27:30 -0400 (EDT) To: josh@agliodbs.com Cc: pgsql-performance@postgresql.org, "Simon Riggs" , "Timothy D. Witham" , testperf-general@pgfoundry.org Subject: Re: [Testperf-general] Re: First set of OSDL Shared Memscalability results, some wierdness ... In-reply-to: <200410151338.17277.josh@agliodbs.com> References: <200410151313.13624.josh@agliodbs.com> <26068.1097872442@sss.pgh.pa.us> <200410151338.17277.josh@agliodbs.com> Comments: In-reply-to Josh Berkus message dated "Fri, 15 Oct 2004 13:38:17 -0700" Date: Fri, 15 Oct 2004 17:27:29 -0400 Message-ID: <26451.1097875649@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/242 X-Sequence-Number: 8718 Josh Berkus writes: >> I suspect the reason recalc_sigpending_tsk is so high is that the >> original coding of PG_TRY involved saving and restoring the signal mask, >> which led to a whole lot of sigsetmask-type kernel calls. Is this test >> with beta3, or something older? > Beta3, *without* Gavin or Neil's Futex patch. Hmm, in that case the cost deserves some further investigation. Can we find out just what that routine does and where it's being called from? regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 15 22:32:24 2004 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 C938532A4F1 for ; Fri, 15 Oct 2004 22:32:22 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 76961-02 for ; Fri, 15 Oct 2004 21:32:17 +0000 (GMT) Received: from mail.osdl.org (fw.osdl.org [65.172.181.6]) by svr1.postgresql.org (Postfix) with ESMTP id D431032A388 for ; Fri, 15 Oct 2004 22:32:17 +0100 (BST) Received: (from markw@localhost) by mail.osdl.org (8.11.6/8.11.6) id i9FLW6X21560; Fri, 15 Oct 2004 14:32:06 -0700 Date: Fri, 15 Oct 2004 14:32:06 -0700 From: Mark Wong To: Tom Lane Cc: josh@agliodbs.com, Simon Riggs , testperf-general@pgfoundry.org, pgsql-performance@postgresql.org Subject: Re: [Testperf-general] Re: First set of OSDL Shared Memscalability results, some wierdness ... Message-ID: <20041015143206.A17724@osdl.org> References: <200410151313.13624.josh@agliodbs.com> <26068.1097872442@sss.pgh.pa.us> <200410151338.17277.josh@agliodbs.com> <26451.1097875649@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5i In-Reply-To: <26451.1097875649@sss.pgh.pa.us>; from tgl@sss.pgh.pa.us on Fri, Oct 15, 2004 at 05:27:29PM -0400 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/243 X-Sequence-Number: 8719 On Fri, Oct 15, 2004 at 05:27:29PM -0400, Tom Lane wrote: > Josh Berkus writes: > >> I suspect the reason recalc_sigpending_tsk is so high is that the > >> original coding of PG_TRY involved saving and restoring the signal mask, > >> which led to a whole lot of sigsetmask-type kernel calls. Is this test > >> with beta3, or something older? > > > Beta3, *without* Gavin or Neil's Futex patch. > > Hmm, in that case the cost deserves some further investigation. Can we > find out just what that routine does and where it's being called from? > There's a call-graph feature with oprofile as of version 0.8 with the opstack tool, but I'm having a terrible time figuring out why the output isn't doing the graphing part. Otherwise, I'd have that available already... Mark From pgsql-performance-owner@postgresql.org Fri Oct 15 22:38:17 2004 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 A5C7732A772 for ; Fri, 15 Oct 2004 22:38:15 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 77045-06 for ; Fri, 15 Oct 2004 21:37:56 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 3D20E32A5CC for ; Fri, 15 Oct 2004 22:37:55 +0100 (BST) 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 i9FLbohv026543; Fri, 15 Oct 2004 17:37:50 -0400 (EDT) To: Mark Wong Cc: Sean Chittenden , Kevin Brown , pgsql-performance@postgresql.org Subject: Re: mmap (was First set of OSDL Shared Mem scalability results, some wierdness ... In-reply-to: <20041015133947.A9225@osdl.org> References: <157f64840410141725162e43b5@mail.gmail.com> <20041015010841.GE665@filer> <0E3BB9CB-1EE6-11D9-A0BB-000A95C705DC@chittenden.org> <20041015133947.A9225@osdl.org> Comments: In-reply-to Mark Wong message dated "Fri, 15 Oct 2004 13:39:47 -0700" Date: Fri, 15 Oct 2004 17:37:50 -0400 Message-ID: <26542.1097876270@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/244 X-Sequence-Number: 8720 Mark Wong writes: > I know where the do_sigaction is coming from in this particular case. > Manfred Spraul tracked it to a pair of pgsignal calls in libpq. > Commenting out those two calls out virtually eliminates do_sigaction from > the kernel profile for this workload. Hmm, I suppose those are the ones associated with suppressing SIGPIPE during send(). It looks to me like those should go away in 8.0 if you have compiled with ENABLE_THREAD_SAFETY ... exactly how is PG being built in the current round of tests? regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 15 22:45:03 2004 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 87BF932A750 for ; Fri, 15 Oct 2004 22:44:59 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 80109-02 for ; Fri, 15 Oct 2004 21:44:40 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 31BA732A602 for ; Fri, 15 Oct 2004 22:44:41 +0100 (BST) 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 i9FLiYHQ026630; Fri, 15 Oct 2004 17:44:34 -0400 (EDT) To: Mark Wong Cc: josh@agliodbs.com, Simon Riggs , testperf-general@pgfoundry.org, pgsql-performance@postgresql.org Subject: Re: [Testperf-general] Re: First set of OSDL Shared Memscalability results, some wierdness ... In-reply-to: <20041015143206.A17724@osdl.org> References: <200410151313.13624.josh@agliodbs.com> <26068.1097872442@sss.pgh.pa.us> <200410151338.17277.josh@agliodbs.com> <26451.1097875649@sss.pgh.pa.us> <20041015143206.A17724@osdl.org> Comments: In-reply-to Mark Wong message dated "Fri, 15 Oct 2004 14:32:06 -0700" Date: Fri, 15 Oct 2004 17:44:34 -0400 Message-ID: <26629.1097876674@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/245 X-Sequence-Number: 8721 Mark Wong writes: > On Fri, Oct 15, 2004 at 05:27:29PM -0400, Tom Lane wrote: >> Hmm, in that case the cost deserves some further investigation. Can we >> find out just what that routine does and where it's being called from? > There's a call-graph feature with oprofile as of version 0.8 with > the opstack tool, but I'm having a terrible time figuring out why the > output isn't doing the graphing part. Otherwise, I'd have that > available already... I was wondering if this might be associated with do_sigaction. do_sigaction is only 0.23 percent of the runtime according to the oprofile results: http://khack.osdl.org/stp/298124/oprofile/DBT_2_Profile-all.oprofile.txt but the profile results for the same run: http://khack.osdl.org/stp/298124/profile/DBT_2_Profile-tick.sort show do_sigaction very high and recalc_sigpending_tsk nowhere at all. Something funny there. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 15 22:56:51 2004 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 9B64632A99B for ; Fri, 15 Oct 2004 22:56:48 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83776-03 for ; Fri, 15 Oct 2004 21:56:45 +0000 (GMT) Received: from mail.osdl.org (fw.osdl.org [65.172.181.6]) by svr1.postgresql.org (Postfix) with ESMTP id DD67032A920 for ; Fri, 15 Oct 2004 22:56:46 +0100 (BST) Received: (from markw@localhost) by mail.osdl.org (8.11.6/8.11.6) id i9FLuU927102; Fri, 15 Oct 2004 14:56:30 -0700 Date: Fri, 15 Oct 2004 14:56:30 -0700 From: Mark Wong To: Tom Lane Cc: Sean Chittenden , Kevin Brown , pgsql-performance@postgresql.org Subject: Re: mmap (was First set of OSDL Shared Mem scalability results, some wierdness ... Message-ID: <20041015145630.A21918@osdl.org> References: <157f64840410141725162e43b5@mail.gmail.com> <20041015010841.GE665@filer> <0E3BB9CB-1EE6-11D9-A0BB-000A95C705DC@chittenden.org> <20041015133947.A9225@osdl.org> <26542.1097876270@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5i In-Reply-To: <26542.1097876270@sss.pgh.pa.us>; from tgl@sss.pgh.pa.us on Fri, Oct 15, 2004 at 05:37:50PM -0400 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/246 X-Sequence-Number: 8722 On Fri, Oct 15, 2004 at 05:37:50PM -0400, Tom Lane wrote: > Mark Wong writes: > > I know where the do_sigaction is coming from in this particular case. > > Manfred Spraul tracked it to a pair of pgsignal calls in libpq. > > Commenting out those two calls out virtually eliminates do_sigaction from > > the kernel profile for this workload. > > Hmm, I suppose those are the ones associated with suppressing SIGPIPE > during send(). It looks to me like those should go away in 8.0 if you > have compiled with ENABLE_THREAD_SAFETY ... exactly how is PG being > built in the current round of tests? > Ah, yes. Ok. It's not being configured with any options. That'll be easy to rememdy though. I'll get that change made and we can try again. Mark From pgsql-performance-owner@postgresql.org Fri Oct 15 23:11:02 2004 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 5E51332A115 for ; Fri, 15 Oct 2004 23:10:59 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 87453-04 for ; Fri, 15 Oct 2004 22:10:39 +0000 (GMT) Received: from mail.osdl.org (fw.osdl.org [65.172.181.6]) by svr1.postgresql.org (Postfix) with ESMTP id CF2AA32A750 for ; Fri, 15 Oct 2004 23:10:37 +0100 (BST) Received: (from markw@localhost) by mail.osdl.org (8.11.6/8.11.6) id i9FMAMO29943; Fri, 15 Oct 2004 15:10:22 -0700 Date: Fri, 15 Oct 2004 15:10:22 -0700 From: Mark Wong To: Tom Lane Cc: josh@agliodbs.com, Simon Riggs , testperf-general@pgfoundry.org, pgsql-performance@postgresql.org Subject: Re: [Testperf-general] Re: First set of OSDL Shared Memscalability results, some wierdness ... Message-ID: <20041015151022.A27429@osdl.org> References: <200410151313.13624.josh@agliodbs.com> <26068.1097872442@sss.pgh.pa.us> <200410151338.17277.josh@agliodbs.com> <26451.1097875649@sss.pgh.pa.us> <20041015143206.A17724@osdl.org> <26629.1097876674@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5i In-Reply-To: <26629.1097876674@sss.pgh.pa.us>; from tgl@sss.pgh.pa.us on Fri, Oct 15, 2004 at 05:44:34PM -0400 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/247 X-Sequence-Number: 8723 On Fri, Oct 15, 2004 at 05:44:34PM -0400, Tom Lane wrote: > Mark Wong writes: > > On Fri, Oct 15, 2004 at 05:27:29PM -0400, Tom Lane wrote: > >> Hmm, in that case the cost deserves some further investigation. Can we > >> find out just what that routine does and where it's being called from? > > > There's a call-graph feature with oprofile as of version 0.8 with > > the opstack tool, but I'm having a terrible time figuring out why the > > output isn't doing the graphing part. Otherwise, I'd have that > > available already... > > I was wondering if this might be associated with do_sigaction. > do_sigaction is only 0.23 percent of the runtime according to the > oprofile results: > http://khack.osdl.org/stp/298124/oprofile/DBT_2_Profile-all.oprofile.txt > but the profile results for the same run: > http://khack.osdl.org/stp/298124/profile/DBT_2_Profile-tick.sort > show do_sigaction very high and recalc_sigpending_tsk nowhere at all. > Something funny there. > I have always attributed those kind of differences based on how readprofile and oprofile collect their data. Granted I don't exactly understand it. Anyone familiar with the two differences? Mark From pgsql-performance-owner@postgresql.org Fri Oct 15 23:32:52 2004 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 5FD04329F42 for ; Fri, 15 Oct 2004 23:32:35 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 90333-08 for ; Fri, 15 Oct 2004 22:32:29 +0000 (GMT) Received: from mail63.csoft.net (leary3.csoft.net [63.111.22.74]) by svr1.postgresql.org (Postfix) with SMTP id 68A8D32A07C for ; Fri, 15 Oct 2004 23:30:59 +0100 (BST) Received: (qmail 28085 invoked by uid 1112); 15 Oct 2004 22:30:17 -0000 Date: Fri, 15 Oct 2004 17:30:17 -0500 (EST) From: Kris Jurka X-X-Sender: books@leary.csoft.net To: Stef Cc: pgsql-performance@postgresql.org Subject: Re: execute cursor fetch In-Reply-To: <20041012150515.167c5e5e@svb.ucs.co.za> Message-ID: References: <20041012114343.69531.qmail@web54110.mail.yahoo.com> <20041012150515.167c5e5e@svb.ucs.co.za> MIME-Version: 1.0 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/248 X-Sequence-Number: 8724 On Tue, 12 Oct 2004, Stef wrote: > Pierre-Fr=E9d=E9ric Caillaud mentioned : > =3D> http://www.postgresql.org/docs/7.4/static/jdbc-query.html#AEN24298 >=20 > My question is : > Is this only true for postgres versions >=3D 7.4 ? >=20 > I see the same section about "Setting fetch size to turn cursors on and o= ff" > is not in the postgres 7.3.7 docs. Does this mean 7.3 the JDBC driver > for postgres < 7.4 doesn't support this ? >=20 You need the 7.4 JDBC driver, but can run it against a 7.3 (or 7.2)=20 database. Also note the 8.0 JDBC driver can only do this against a 7.4 or= =20 8.0 database and not older versions. Kris Jurka From pgsql-performance-owner@postgresql.org Sat Oct 16 02:22:51 2004 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 9C93B329E7E for ; Sat, 16 Oct 2004 02:22:47 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 24672-08 for ; Sat, 16 Oct 2004 01:22:35 +0000 (GMT) Received: from candle.pha.pa.us (candle.pha.pa.us [207.106.42.251]) by svr1.postgresql.org (Postfix) with ESMTP id 38162329E43 for ; Sat, 16 Oct 2004 02:22:36 +0100 (BST) Received: (from pgman@localhost) by candle.pha.pa.us (8.11.6/8.11.6) id i9G1M3h03028; Fri, 15 Oct 2004 21:22:03 -0400 (EDT) From: Bruce Momjian Message-Id: <200410160122.i9G1M3h03028@candle.pha.pa.us> Subject: Re: mmap (was First set of OSDL Shared Mem scalability results, In-Reply-To: <26542.1097876270@sss.pgh.pa.us> To: Tom Lane Date: Fri, 15 Oct 2004 21:22:03 -0400 (EDT) Cc: Mark Wong , Sean Chittenden , Kevin Brown , pgsql-performance@postgresql.org X-Mailer: ELM [version 2.4ME+ PL108 (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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/249 X-Sequence-Number: 8725 Tom Lane wrote: > Mark Wong writes: > > I know where the do_sigaction is coming from in this particular case. > > Manfred Spraul tracked it to a pair of pgsignal calls in libpq. > > Commenting out those two calls out virtually eliminates do_sigaction from > > the kernel profile for this workload. > > Hmm, I suppose those are the ones associated with suppressing SIGPIPE > during send(). It looks to me like those should go away in 8.0 if you > have compiled with ENABLE_THREAD_SAFETY ... exactly how is PG being > built in the current round of tests? Yes, those calls are gone in 8.0 with --enable-thread-safety and were added specifically because of Manfred's reports. -- 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-hackers-owner@postgresql.org Sat Oct 16 17:54:26 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6733A32A15D; Sat, 16 Oct 2004 17:54:24 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 10850-03; Sat, 16 Oct 2004 16:54:18 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id E1F7F32A2C6; Sat, 16 Oct 2004 17:54:18 +0100 (BST) 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 i9GGsHSA004665; Sat, 16 Oct 2004 12:54:17 -0400 (EDT) To: pgsql-hackers@postgreSQL.org, pgsql-performance@postgreSQL.org Subject: Getting rid of AtEOXact_Buffers (was Re: [Testperf-general] Re: [PERFORM] First set of OSDL Shared Memscalability results, some wierdness ...) In-reply-to: <26068.1097872442@sss.pgh.pa.us> References: <10202.1097858894@sss.pgh.pa.us> <200410151313.13624.josh@agliodbs.com> <26068.1097872442@sss.pgh.pa.us> Comments: In-reply-to Tom Lane message dated "Fri, 15 Oct 2004 16:34:02 -0400" Date: Sat, 16 Oct 2004 12:54:17 -0400 Message-ID: <4664.1097945657@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/525 X-Sequence-Number: 60028 I wrote: > Josh Berkus writes: >> First off, two test runs with OProfile are available at: >> http://khack.osdl.org/stp/298124/ >> http://khack.osdl.org/stp/298121/ > Hmm. The stuff above 1% in the first of these is > Counted CPU_CLK_UNHALTED events (clocks processor is not halted) with a unit mask of 0x00 (No unit mask) count 100000 > samples % app name symbol name > ... > 920369 2.1332 postgres AtEOXact_Buffers > ... > In the second test AtEOXact_Buffers is much lower (down around 0.57 > percent) but the other suspects are similar. Since the only difference > in parameters is shared_buffers (36000 vs 9000), it does look like we > are approaching the point where AtEOXact_Buffers is a problem, but so > far it's only a 2% drag. It occurs to me that given the 8.0 resource manager mechanism, we could in fact dispense with AtEOXact_Buffers, or perhaps better turn it into a no-op unless #ifdef USE_ASSERT_CHECKING. We'd just get rid of the special case for transaction termination in resowner.c and let the resource owner be responsible for releasing locked buffers always. The OSDL results suggest that this won't matter much at the level of 10000 or so shared buffers, but for 100000 or more buffers the linear scan in AtEOXact_Buffers is going to become a problem. We could also get rid of the linear search in UnlockBuffers(). The only thing it's for anymore is to release a BM_PIN_COUNT_WAITER flag, and since a backend could not be doing more than one of those at a time, we don't really need an array of flags for that, only a single variable. This does not show in the OSDL results, which I presume means that their test case is not exercising transaction aborts; but I think we need to zap both routines to make the world safe for large shared_buffers values. (See also http://archives.postgresql.org/pgsql-performance/2004-10/msg00218.php) Any objection to doing this for 8.0? regards, tom lane From pgsql-hackers-owner@postgresql.org Sat Oct 16 22:16:41 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5B755329D20; Sat, 16 Oct 2004 22:16:33 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 60907-08; Sat, 16 Oct 2004 21:16:26 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id E2C78329D22; Sat, 16 Oct 2004 22:16:27 +0100 (BST) 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 6518021; Sat, 16 Oct 2004 14:17:50 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Getting rid of AtEOXact_Buffers (was Re: [Testperf-general] Re: [PERFORM] First set of OSDL Shared Memscalability results, some wierdness ...) Date: Sat, 16 Oct 2004 14:18:26 -0700 User-Agent: KMail/1.6.2 Cc: Tom Lane , pgsql-hackers@postgreSQL.org References: <26068.1097872442@sss.pgh.pa.us> <4664.1097945657@sss.pgh.pa.us> In-Reply-To: <4664.1097945657@sss.pgh.pa.us> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable Message-Id: <200410161418.26557.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/529 X-Sequence-Number: 60032 Tom, > We could also get rid of the linear search in UnlockBuffers(). =C2=A0The = only > thing it's for anymore is to release a BM_PIN_COUNT_WAITER flag, and > since a backend could not be doing more than one of those at a time, > we don't really need an array of flags for that, only a single variable. > This does not show in the OSDL results, which I presume means that their > test case is not exercising transaction aborts; In the test, one out of every 100 new order transactions is aborted (about = 1=20 out of 150 transactions overall). --=20 --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-hackers-owner@postgresql.org Sat Oct 16 22:19:25 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 83029329D28; Sat, 16 Oct 2004 22:19:23 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 63801-02; Sat, 16 Oct 2004 21:19:16 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 0E9A2329D20; Sat, 16 Oct 2004 22:19:15 +0100 (BST) 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 i9GLJCkp000179; Sat, 16 Oct 2004 17:19:12 -0400 (EDT) To: josh@agliodbs.com Cc: pgsql-performance@postgresql.org, pgsql-hackers@postgresql.org Subject: Re: Getting rid of AtEOXact_Buffers (was Re: [Testperf-general] Re: [PERFORM] First set of OSDL Shared Memscalability results, some wierdness ...) In-reply-to: <200410161418.26557.josh@agliodbs.com> References: <26068.1097872442@sss.pgh.pa.us> <4664.1097945657@sss.pgh.pa.us> <200410161418.26557.josh@agliodbs.com> Comments: In-reply-to Josh Berkus message dated "Sat, 16 Oct 2004 14:18:26 -0700" Date: Sat, 16 Oct 2004 17:19:11 -0400 Message-ID: <178.1097961551@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/530 X-Sequence-Number: 60033 Josh Berkus writes: >> This does not show in the OSDL results, which I presume means that their >> test case is not exercising transaction aborts; > In the test, one out of every 100 new order transactions is aborted (about 1 > out of 150 transactions overall). Okay, but that just ensures that any bottlenecks in xact abort will be down in the noise in this test case ... In any case, those changes are in CVS now if you want to try them. regards, tom lane From pgsql-hackers-owner@postgresql.org Sat Oct 16 22:35:07 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 55143329D22; Sat, 16 Oct 2004 22:35:03 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 64460-09; Sat, 16 Oct 2004 21:34:44 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id DCC4C329DB3; Sat, 16 Oct 2004 22:34:45 +0100 (BST) 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 6518089; Sat, 16 Oct 2004 14:36:09 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Getting rid of AtEOXact_Buffers (was Re: [Testperf-general] Re: [PERFORM] First set of OSDL Shared Memscalability results, some wierdness ...) Date: Sat, 16 Oct 2004 14:36:45 -0700 User-Agent: KMail/1.6.2 Cc: Tom Lane , pgsql-hackers@postgresql.org References: <200410161418.26557.josh@agliodbs.com> <178.1097961551@sss.pgh.pa.us> In-Reply-To: <178.1097961551@sss.pgh.pa.us> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Message-Id: <200410161436.45310.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/532 X-Sequence-Number: 60035 Tom, > In any case, those changes are in CVS now if you want to try them. OK. Will have to wait until OSDL gives me a dedicated testing machine sometime mon/tues/wed. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Tue Oct 19 22:22:46 2004 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 E59B932A334 for ; Sun, 17 Oct 2004 08:53:55 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78403-07 for ; Sun, 17 Oct 2004 07:53:45 +0000 (GMT) Received: from dbl.q-ag.de (dbl.q-ag.de [213.172.117.3]) by svr1.postgresql.org (Postfix) with ESMTP id C8017329E65 for ; Sun, 17 Oct 2004 08:53:42 +0100 (BST) Received: from [127.0.0.2] (dbl [127.0.0.1]) by dbl.q-ag.de (8.12.3/8.12.3/Debian-6.6) with ESMTP id i9H7rBSL028958; Sun, 17 Oct 2004 09:53:13 +0200 Message-ID: <417221B5.1050704@colorfullife.com> Date: Sun, 17 Oct 2004 09:39:33 +0200 From: Manfred Spraul User-Agent: Mozilla/5.0 (X11; U; Linux i686; fr-FR; rv:1.7.3) Gecko/20040922 X-Accept-Language: en-us, en MIME-Version: 1.0 To: neilc@samurai.com Cc: markw@osdl.org, pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 Content-Type: multipart/mixed; boundary="------------090401080301040305060201" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/293 X-Sequence-Number: 8769 This is a multi-part message in MIME format. --------------090401080301040305060201 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Neil wrote: >. In any case, the "futex patch" >uses the Linux 2.6 futex API to implement PostgreSQL spinlocks. > Has anyone tried to replace the whole lwlock implementation with pthread_rwlock? At least for Linux with recent glibcs, pthread_rwlock is implemented with futexes, i.e. we would get a fast lock handling without os specific hacks. Perhaps other os contain user space pthread locks, too. Attached is an old patch. I tested it on an uniprocessor system a year ago and it didn't provide much difference, but perhaps the scalability is better. You'll have to add -lpthread to the library list for linking. Regarding Neil's patch: >! /* >! * XXX: is there a more efficient way to write this? Perhaps using >! * decl...? >! */ >! static __inline__ slock_t >! atomic_dec(volatile slock_t *ptr) >! { >! slock_t prev = -1; >! >! __asm__ __volatile__( >! " lock \n" >! " xadd %0,%1 \n" >! :"=q"(prev) >! :"m"(*ptr), "0"(prev) >! :"memory", "cc"); >! >! return prev; >! } > xadd is not supported by original 80386 cpus, it was added for 80486 cpus. There is no instruction in the 80386 cpu that allows to atomically decrement and retrieve the old value of an integer. The only option are atomic_dec_test_zero or atomic_dec_test_negative - that can be implemented by looking at the sign/zero flag. Depending on what you want this may be enough. Or make the futex code conditional for > 80386 cpus. -- Manfred --------------090401080301040305060201 Content-Type: text/plain; name="patch-pthread-force" Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="patch-pthread-force" --- p7.3.3.orig/src/backend/storage/lmgr/lwlock.c 2002-09-25 22:31:40.000000000 +0200 +++ postgresql-7.3.3/src/backend/storage/lmgr/lwlock.c 2003-09-06 14:15:01.000000000 +0200 @@ -26,6 +26,28 @@ #include "storage/proc.h" #include "storage/spin.h" +#define USE_PTHREAD_LOCKS + +#ifdef USE_PTHREAD_LOCKS + +#include +#include +typedef pthread_rwlock_t LWLock; + +inline static void +InitLWLock(LWLock *p) +{ + pthread_rwlockattr_t rwattr; + int i; + + pthread_rwlockattr_init(&rwattr); + pthread_rwlockattr_setpshared(&rwattr, PTHREAD_PROCESS_SHARED); + i=pthread_rwlock_init(p, &rwattr); + pthread_rwlockattr_destroy(&rwattr); + if (i) + elog(FATAL, "pthread_rwlock_init failed"); +} +#else typedef struct LWLock { @@ -38,6 +60,17 @@ /* tail is undefined when head is NULL */ } LWLock; +inline static void +InitLWLock(LWLock *lock) +{ + SpinLockInit(&lock->mutex); + lock->releaseOK = true; + lock->exclusive = 0; + lock->shared = 0; + lock->head = NULL; + lock->tail = NULL; +} +#endif /* * This points to the array of LWLocks in shared memory. Backends inherit * the pointer by fork from the postmaster. LWLockIds are indexes into @@ -61,7 +94,7 @@ static LWLockId held_lwlocks[MAX_SIMUL_LWLOCKS]; -#ifdef LOCK_DEBUG +#if defined(LOCK_DEBUG) && !defined(USE_PTHREAD_LOCKS) bool Trace_lwlocks = false; inline static void @@ -153,12 +186,7 @@ */ for (id = 0, lock = LWLockArray; id < numLocks; id++, lock++) { - SpinLockInit(&lock->mutex); - lock->releaseOK = true; - lock->exclusive = 0; - lock->shared = 0; - lock->head = NULL; - lock->tail = NULL; + InitLWLock(lock); } /* @@ -185,7 +213,116 @@ return (LWLockId) (LWLockCounter[0]++); } +#ifdef USE_PTHREAD_LOCKS +/* + * LWLockAcquire - acquire a lightweight lock in the specified mode + * + * If the lock is not available, sleep until it is. + * + * Side effect: cancel/die interrupts are held off until lock release. + */ +void +LWLockAcquire(LWLockId lockid, LWLockMode mode) +{ + int i; + PRINT_LWDEBUG("LWLockAcquire", lockid, &LWLockArray[lockid]); + + /* + * We can't wait if we haven't got a PGPROC. This should only occur + * during bootstrap or shared memory initialization. Put an Assert + * here to catch unsafe coding practices. + */ + Assert(!(proc == NULL && IsUnderPostmaster)); + + /* + * Lock out cancel/die interrupts until we exit the code section + * protected by the LWLock. This ensures that interrupts will not + * interfere with manipulations of data structures in shared memory. + */ + HOLD_INTERRUPTS(); + + if (mode == LW_EXCLUSIVE) { + i = pthread_rwlock_wrlock(&LWLockArray[lockid]); + } else { + i = pthread_rwlock_rdlock(&LWLockArray[lockid]); + } + if (i) + elog(FATAL, "Unexpected error from pthread_rwlock."); + + /* Add lock to list of locks held by this backend */ + Assert(num_held_lwlocks < MAX_SIMUL_LWLOCKS); + held_lwlocks[num_held_lwlocks++] = lockid; +} + +/* + * LWLockConditionalAcquire - acquire a lightweight lock in the specified mode + * + * If the lock is not available, return FALSE with no side-effects. + * + * If successful, cancel/die interrupts are held off until lock release. + */ +bool +LWLockConditionalAcquire(LWLockId lockid, LWLockMode mode) +{ + int i; + PRINT_LWDEBUG("LWLockConditionalAcquire", lockid, &LWLockArray[lockid]); + + HOLD_INTERRUPTS(); + + if (mode == LW_EXCLUSIVE) { + i = pthread_rwlock_trywrlock(&LWLockArray[lockid]); + } else { + i = pthread_rwlock_tryrdlock(&LWLockArray[lockid]); + } + switch(i) { + case 0: + /* Add lock to list of locks held by this backend */ + Assert(num_held_lwlocks < MAX_SIMUL_LWLOCKS); + held_lwlocks[num_held_lwlocks++] = lockid; + return true; + case EBUSY: + RESUME_INTERRUPTS(); + return false; + default: + elog(FATAL, "Unexpected error from pthread_rwlock_try."); + return false; + } +} + +/* + * LWLockRelease - release a previously acquired lock + */ +void +LWLockRelease(LWLockId lockid) +{ + int i; + + /* + * Remove lock from list of locks held. Usually, but not always, it + * will be the latest-acquired lock; so search array backwards. + */ + for (i = num_held_lwlocks; --i >= 0;) + { + if (lockid == held_lwlocks[i]) + break; + } + if (i < 0) + elog(ERROR, "LWLockRelease: lock %d is not held", (int) lockid); + num_held_lwlocks--; + for (; i < num_held_lwlocks; i++) + held_lwlocks[i] = held_lwlocks[i + 1]; + + i = pthread_rwlock_unlock(&LWLockArray[lockid]); + if (i) + elog(FATAL, "Unexpected error from pthread_rwlock_unlock."); + + /* + * Now okay to allow cancel/die interrupts. + */ + RESUME_INTERRUPTS(); +} +#else /* * LWLockAcquire - acquire a lightweight lock in the specified mode * @@ -499,6 +636,7 @@ RESUME_INTERRUPTS(); } +#endif /* * LWLockReleaseAll - release all currently-held locks --------------090401080301040305060201-- From pgsql-hackers-owner@postgresql.org Sun Oct 17 20:42:21 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 89DFE32A2A1; Sun, 17 Oct 2004 20:42:14 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 15654-04; Sun, 17 Oct 2004 19:42:11 +0000 (GMT) Received: from moutng.kundenserver.de (moutng.kundenserver.de [212.227.126.184]) by svr1.postgresql.org (Postfix) with ESMTP id C9A4D32A1BC; Sun, 17 Oct 2004 20:42:08 +0100 (BST) Received: from [212.227.126.202] (helo=mrvnet.kundenserver.de) by moutng.kundenserver.de with esmtp (Exim 3.35 #1) id 1CJGuo-00014X-00; Sun, 17 Oct 2004 21:42:02 +0200 Received: from [172.23.4.143] (helo=config16.kundenserver.de) by mrvnet.kundenserver.de with esmtp (Exim 3.35 #1) id 1CJGun-00032J-00; Sun, 17 Oct 2004 21:42:01 +0200 Received: from www-data by config16.kundenserver.de with local (Exim 3.35 #1 (Debian)) id 1CJGun-00087B-00; Sun, 17 Oct 2004 21:42:01 +0200 To: =?iso-8859-1?Q?Tom_Lane?= , Subject: =?iso-8859-1?Q?Re:__Getting_rid_of_AtEOXact_Buffers_(was_Re:_[Testperf-general]_Re:_[PERFORM]_First_set_of_OSDL_Shared_Memscalability_results, _some_wierdness_=2E=2E=2E)?= From: Cc: , Message-Id: <28292295$10980398474172c227e88c75.95384212@config16.schlund.de> X-Binford: 6100 (more power) X-Originating-From: 28292295 X-Mailer: Webmail X-Routing: UK X-Received: from config16 by 217.135.203.211 with HTTP id 28292295 for tgl@sss.pgh.pa.us; Sun, 17 Oct 2004 21:40:01 +0200 Content-Type: text/plain; charset="iso-8859-1" Mime-Version: 1.0 Content-Transfer-Encoding: 8bit X-Priority: 3 Date: Sun, 17 Oct 2004 21:40:01 +0200 X-Provags-ID: kundenserver.de abuse@kundenserver.de ident:@172.23.4.143 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.3 tagged_above=0.0 required=5.0 tests=NO_REAL_NAME X-Spam-Level: X-Archive-Number: 200410/552 X-Sequence-Number: 60055 Seeing as I've missed the last N messages... I'll just reply to this one, rather than each of them in turn... Tom Lane wrote on 16.10.2004, 18:54:17: > I wrote: > > Josh Berkus writes: > >> First off, two test runs with OProfile are available at: > >> http://khack.osdl.org/stp/298124/ > >> http://khack.osdl.org/stp/298121/ > > > Hmm. The stuff above 1% in the first of these is > > > Counted CPU_CLK_UNHALTED events (clocks processor is not halted) with a unit mask of 0x00 (No unit mask) count 100000 > > samples % app name symbol name > > ... > > 920369 2.1332 postgres AtEOXact_Buffers > > ... > > > In the second test AtEOXact_Buffers is much lower (down around 0.57 > > percent) but the other suspects are similar. Since the only difference > > in parameters is shared_buffers (36000 vs 9000), it does look like we > > are approaching the point where AtEOXact_Buffers is a problem, but so > > far it's only a 2% drag. Yes... as soon as you first mentioned AtEOXact_Buffers, I realised I'd seen it near the top of the oprofile results on previous tests. Although you don't say this, I presume you're acting on the thought that a 2% drag would soon become a much larger contention point with more users and/or smaller transactions - since these things are highly non-linear. > > It occurs to me that given the 8.0 resource manager mechanism, we could > in fact dispense with AtEOXact_Buffers, or perhaps better turn it into a > no-op unless #ifdef USE_ASSERT_CHECKING. We'd just get rid of the > special case for transaction termination in resowner.c and let the > resource owner be responsible for releasing locked buffers always. The > OSDL results suggest that this won't matter much at the level of 10000 > or so shared buffers, but for 100000 or more buffers the linear scan in > AtEOXact_Buffers is going to become a problem. If the resource owner is always responsible for releasing locked buffers, who releases the locks if the backend crashes? Do we need some additional code in bgwriter (or?) to clean up buffer locks? > > We could also get rid of the linear search in UnlockBuffers(). The only > thing it's for anymore is to release a BM_PIN_COUNT_WAITER flag, and > since a backend could not be doing more than one of those at a time, > we don't really need an array of flags for that, only a single variable. > This does not show in the OSDL results, which I presume means that their > test case is not exercising transaction aborts; but I think we need to > zap both routines to make the world safe for large shared_buffers > values. (See also > http://archives.postgresql.org/pgsql-performance/2004-10/msg00218.php) Yes, that's important. > Any objection to doing this for 8.0? > As you say, if these issues are definitely kicking in at 100000 shared_buffers - there's a good few people out there with 800Mb shared_buffers already. Could I also suggest that we adopt your earlier suggestion of raising the bgwriter parameters as a permanent measure - i.e. changing the defaults in postgresql.conf. That way, StrategyDirtyBufferList won't immediately show itself as a problem when using the default parameter set. It would be a shame to remove one obstacle only to leave another one following so close behind. [...and that also argues against an earlier thought to introduce more fine grained values for the bgwriter's parameters, ISTM] Also, what will Vacuum delay do to the O(N) effect of FlushRelationBuffers when called by VACUUM? Will the locks be held for longer? I think we should do some tests while running a VACUUM in the background also, which isn't part of the DBT-2 set-up, but perhaps we might argue *it should be for the PostgreSQL version*? Dare we hope for a scalability increase in 8.0 after all.... Best Regards, Simon Riggs From pgsql-hackers-owner@postgresql.org Sun Oct 17 21:12:55 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 04380329E4C; Sun, 17 Oct 2004 21:12:48 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 22889-01; Sun, 17 Oct 2004 20:12:37 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id E4FA1329D3B; Sun, 17 Oct 2004 21:12:38 +0100 (BST) 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 i9HKCZMF010585; Sun, 17 Oct 2004 16:12:35 -0400 (EDT) To: simon@2ndquadrant.com Cc: josh@agliodbs.com, pgsql-hackers@postgresql.org, pgsql-performance@postgresql.org Subject: Re: =?iso-8859-1?Q?Re:__Getting_rid_of_AtEOXact_Buffers_(was_Re:_[Testperf-general]_Re:_[PERFORM]_First_set_of_OSDL_Shared_Memscalability_results, _some_wierdness_=2E=2E=2E)?= In-reply-to: <28292295$10980398474172c227e88c75.95384212@config16.schlund.de> References: <28292295$10980398474172c227e88c75.95384212@config16.schlund.de> Comments: In-reply-to message dated "Sun, 17 Oct 2004 21:40:01 +0200" Date: Sun, 17 Oct 2004 16:12:35 -0400 Message-ID: <10584.1098043955@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/554 X-Sequence-Number: 60057 writes: > If the resource owner is always responsible for releasing locked > buffers, who releases the locks if the backend crashes? The ensuing system reset takes care of that. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Oct 18 01:50:10 2004 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 63C6E329FF6 for ; Mon, 18 Oct 2004 01:50:01 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78993-04 for ; Mon, 18 Oct 2004 00:49:55 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id 71AC0329EA8 for ; Mon, 18 Oct 2004 01:49:58 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i9I0nsJ9079812 for ; Mon, 18 Oct 2004 00:49:54 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i9I0M7C6074762 for pgsql-performance@postgresql.org; Mon, 18 Oct 2004 00:22:07 GMT From: Gaetano Mendola X-Newsgroups: comp.databases.postgresql.performance Subject: Re: [HACKERS] Getting rid of AtEOXact Buffers (was Re: [Testperf-general] Date: Mon, 18 Oct 2004 02:22:08 +0200 Organization: PYRENET Midi-pyrenees Provider Lines: 11 Message-ID: <41730CB0.60903@bigfoot.com> References: <28292295$10980398474172c227e88c75.95384212@config16.schlund.de> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Complaints-To: abuse@pyrenet.fr To: simon@2ndquadrant.com User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en In-Reply-To: <28292295$10980398474172c227e88c75.95384212@config16.schlund.de> X-Enigmail-Version: 0.86.1.0 X-Enigmail-Supports: pgp-inline, pgp-mime To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/256 X-Sequence-Number: 8732 simon@2ndquadrant.com wrote: > If the resource owner is always responsible for releasing locked > buffers, who releases the locks if the backend crashes? The semaphore "undo" I hope. Regards Gaetano Mendola From pgsql-performance-owner@postgresql.org Mon Oct 18 09:50:31 2004 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 A4AF3329E6E for ; Mon, 18 Oct 2004 09:50:29 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 94421-01 for ; Mon, 18 Oct 2004 08:50:23 +0000 (GMT) Received: from audi.ibcp.fr (audi-ghost.ibcp.fr [193.55.43.240]) by svr1.postgresql.org (Postfix) with ESMTP id 65A3A32A18A for ; Mon, 18 Oct 2004 09:50:22 +0100 (BST) Received: from ibcp.fr (gdeleage15.ibcp.fr [193.51.160.66]) by audi.ibcp.fr (Postfix) with ESMTP id 8FC841AF369 for ; Mon, 18 Oct 2004 10:50:21 +0200 (CEST) Message-ID: <417383CD.3050802@ibcp.fr> Date: Mon, 18 Oct 2004 10:50:21 +0200 From: charavay User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.4.2) Gecko/20040308 X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Indexes performance 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/257 X-Sequence-Number: 8733 Hello ! We have difficulties with the use of indexes. For example, we have two tables : * table lnk : Table "public.lnk" Column | Type | Modifiers --------+-----------------------+----------- index | integer | not null sgaccn | character varying(12) | not null Indexes: "pkey1" primary key, btree ("index", sgaccn) Foreign-key constraints: "fk_sgaccn1" FOREIGN KEY (sgaccn) REFERENCES main_tbl(sgaccn) ON UPDATE CASCADE ON DELETE CASCADE * table dic : Table "public.dic" Column | Type | Modifiers --------+-----------------------+-------------------------------------------------------------------- index | integer | not null default nextval('public.dic_index_seq'::text) word | character varying(60) | not null Indexes: "dic_pkey" primary key, btree ("index") "dic_word_idx" unique, btree (word) "dic_word_key" unique, btree (word) The table lnk contains 33 000 000 tuples and table dic contains 303 000 tuples. When we try to execute a join between these two tables, the planner proposes to excute a hash-join plan : explain select sgaccn from dic, lnk where dic.index=lnk.index; QUERY PLAN ----------------------------------------------------------------------------------- Hash Join (cost=6793.29..1716853.80 rows=33743101 width=11) Hash Cond: ("outer"."index" = "inner"."index") -> Seq Scan on lnk (cost=0.00..535920.00 rows=33743100 width=15) -> Hash (cost=4994.83..4994.83 rows=303783 width=4) -> Seq Scan dic (cost=0.00..4994.83 rows=303783 width=4) (5 rows) So the planner decides to scan 33 000 000 of tuples and we would like to force it to scan the table dic (303 000 tuples) and to use the index on the integer index to execute the join. So we have set the parameters enable_hashjoin and enable_mergejoin to off. So the planner proposes the following query : QUERY PLAN -------------------------------------------------------------------------------------------------- Nested Loop (cost=0.00..102642540.60 rows=33743101 width=11) -> Seq Scan on refs_ra_lnk1 (cost=0.00..535920.00 rows=33743100 width=15) -> Index Scan using refs_ra_dic_new_pkey on refs_ra_dic_new (cost=0.00..3.01 rows=1 width=4) Index Cond: (refs_ra_dic_new."index" = "outer"."index") (4 rows) We were surprised of this response because the planner continues to propose us to scan the 33 000 000 of tuples instead of the smaller table. Is there any way to force it to scan the smaller table ? Thanks Celine Charavay From pgsql-performance-owner@postgresql.org Mon Oct 18 16:28:48 2004 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 80293329F8C for ; Mon, 18 Oct 2004 16:28:46 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 14799-02 for ; Mon, 18 Oct 2004 15:28:41 +0000 (GMT) Received: from mail.osdl.org (fw.osdl.org [65.172.181.6]) by svr1.postgresql.org (Postfix) with ESMTP id 9C7BA329DB0 for ; Mon, 18 Oct 2004 16:28:40 +0100 (BST) Received: (from markw@localhost) by mail.osdl.org (8.11.6/8.11.6) id i9IFSAJ12704; Mon, 18 Oct 2004 08:28:10 -0700 Date: Mon, 18 Oct 2004 08:28:10 -0700 From: Mark Wong To: Bruce Momjian Cc: Tom Lane , Sean Chittenden , Kevin Brown , pgsql-performance@postgresql.org Subject: Re: mmap (was First set of OSDL Shared Mem scalability results, some wierdness ... Message-ID: <20041018082810.A11062@osdl.org> References: <26542.1097876270@sss.pgh.pa.us> <200410160122.i9G1M3h03028@candle.pha.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5i In-Reply-To: <200410160122.i9G1M3h03028@candle.pha.pa.us>; from pgman@candle.pha.pa.us on Fri, Oct 15, 2004 at 09:22:03PM -0400 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/258 X-Sequence-Number: 8734 On Fri, Oct 15, 2004 at 09:22:03PM -0400, Bruce Momjian wrote: > Tom Lane wrote: > > Mark Wong writes: > > > I know where the do_sigaction is coming from in this particular case. > > > Manfred Spraul tracked it to a pair of pgsignal calls in libpq. > > > Commenting out those two calls out virtually eliminates do_sigaction from > > > the kernel profile for this workload. > > > > Hmm, I suppose those are the ones associated with suppressing SIGPIPE > > during send(). It looks to me like those should go away in 8.0 if you > > have compiled with ENABLE_THREAD_SAFETY ... exactly how is PG being > > built in the current round of tests? > > Yes, those calls are gone in 8.0 with --enable-thread-safety and were > added specifically because of Manfred's reports. > Ok, I had the build commands changed for installing PostgreSQL in STP. The do_sigaction call isn't at the top of the profile anymore, here's a reference for those who are interested; it should have the same test parameters as the one Tom referenced a little earlier: http://khack.osdl.org/stp/298230/ Mark From pgsql-performance-owner@postgresql.org Mon Oct 18 19:01:23 2004 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 B3E54329F90 for ; Mon, 18 Oct 2004 19:01:21 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 68545-01 for ; Mon, 18 Oct 2004 18:01:13 +0000 (GMT) Received: from shenley.e-rm.co.uk (shenley.e-rm.co.uk [217.154.196.234]) by svr1.postgresql.org (Postfix) with ESMTP id E2D1632A94E for ; Mon, 18 Oct 2004 19:00:12 +0100 (BST) Received: from gator.e-rm.internal ([192.168.2.1] helo=dev1) by shenley.e-rm.co.uk with esmtp (Exim 3.35 #1 (Debian)) id 1CJbnc-0003pu-00 for ; Mon, 18 Oct 2004 19:00:00 +0100 From: "Rod Dutton" To: Subject: Queries slow using stored procedures Date: Mon, 18 Oct 2004 19:01:25 +0100 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0028_01C4B544.DFB19FA0" X-Mailer: Microsoft Office Outlook, Build 11.0.5510 Thread-Index: AcS1PHuwz718w8JtTi2M+MPIPiou1g== X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1106 Message-Id: X-MailScanner: Found to be clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.2 tagged_above=0.0 required=5.0 tests=HTML_50_60, HTML_MESSAGE X-Spam-Level: X-Archive-Number: 200410/259 X-Sequence-Number: 8735 This is a multi-part message in MIME format. ------=_NextPart_000_0028_01C4B544.DFB19FA0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Hi, I have a problem where a query inside a function is up to 100 times slower inside a function than as a stand alone query run in psql. The column 'botnumber' is a character(10), is indexed and there are 125000 rows in the table. Help please! This query is fast:- explain analyze SELECT batchserial FROM transbatch WHERE botnumber = '1-7' LIMIT 1; QUERY PLAN ---------------------------------------------------------------------------- ---------------------------------------------------- Limit (cost=0.00..0.42 rows=1 width=4) (actual time=0.73..148.23 rows=1 loops=1) -> Index Scan using ind_tbatchx on transbatch (cost=0.00..18.73 rows=45 width=4) (actual time=0.73..148.22 rows=1 loops=1) Index Cond: (botnumber = '1-7'::bpchar) Total runtime: 148.29 msec (4 rows) This function is slow:- CREATE OR REPLACE FUNCTION sp_test_rod3 ( ) returns integer as ' DECLARE bot char(10); oldbatch INTEGER; BEGIN bot := ''1-7''; SELECT INTO oldbatch batchserial FROM transbatch WHERE botnumber = bot LIMIT 1; IF FOUND THEN RETURN 1; ELSE RETURN 0; END IF; END; ' language plpgsql ; explain analyze SELECT sp_test_rod3(); QUERY PLAN ---------------------------------------------------------------------------- ------------ Result (cost=0.00..0.01 rows=1 width=0) (actual time=1452.39..1452.40 rows=1 loops=1) Total runtime: 1452.42 msec (2 rows) ------=_NextPart_000_0028_01C4B544.DFB19FA0 Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable
Hi,
 
I have a = problem=20 where a query inside a function is up to 100 times slower inside a function= than=20 as a stand alone query run in psql.
 
The colum= n=20 'botnumber' is a character(10), is indexed and there are 125000 rows i= n the=20 table.
 
Help=20 please!
 
This quer= y is=20 fast:-
 
explain= =20 analyze  
&nb= sp;=20 SELECT batchserial
  FROM transbatch
  WHERE botnumb= er =3D=20 '1-7'
  LIMIT 1;
           =             &nb= sp;            =             &nb= sp;         =20 QUERY=20 PLAN            = ;            &n= bsp;            = ;            &n= bsp;        =20
-----------------------------------------------------------------------= ---------------------------------------------------------
 Limit&nb= sp;=20 (cost=3D0.00..0.42 rows=3D1 width=3D4) (actual time=3D0.73..148.23 rows=3D1= =20 loops=3D1)
   ->  Index Scan using ind_tbatchx on=20 transbatch  (cost=3D0.00..18.73 rows=3D45 width=3D4) (actual time=3D0.= 73..148.22=20 rows=3D1 loops=3D1)
         Ind= ex Cond:=20 (botnumber =3D '1-7'::bpchar)
 Total runtime: 148.29 msec
(4=20 rows)
 
 
Thi= s function=20 is slow:-
 
CRE= ATE OR=20 REPLACE FUNCTION  sp_test_rod3 ( ) returns=20 integer         
as=20 '
DECLARE
  bot char(10);
  oldbatch=20 INTEGER;
BEGIN
 
&nb= sp; bot :=3D=20 ''1-7'';
 
&nb= sp; SELECT=20 INTO oldbatch batchserial
  FROM transbatch
  WHERE botnumb= er =3D=20 bot
  LIMIT 1;
 
&nb= sp; IF=20 FOUND THEN
    RETURN 1;
  ELSE
  &n= bsp;=20 RETURN 0;
  END IF;
 
END;
'
language plpgsql  ;
 
exp= lain=20 analyze SELECT sp_test_rod3();
           =             &nb= sp;            =   =20 QUERY=20 PLAN            = ;            &n= bsp;            = ; =20
-----------------------------------------------------------------------= -----------------
 Result =20 (cost=3D0.00..0.01 rows=3D1 width=3D0) (actual time=3D1452.39..1452.40 rows= =3D1=20 loops=3D1)
 Total runtime: 1452.42 msec
(2=20 rows)
------=_NextPart_000_0028_01C4B544.DFB19FA0-- From pgsql-performance-owner@postgresql.org Mon Oct 18 19:13:40 2004 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 54086329E57 for ; Mon, 18 Oct 2004 19:13:39 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 69404-09 for ; Mon, 18 Oct 2004 18:13:29 +0000 (GMT) Received: from hosting.commandprompt.com (128.commandprompt.com [207.173.200.128]) by svr1.postgresql.org (Postfix) with ESMTP id A7CAC329DFC for ; Mon, 18 Oct 2004 19:13:28 +0100 (BST) Received: from [192.168.1.20] (clbb-248.saw.net [64.146.135.248]) (authenticated) by hosting.commandprompt.com (8.11.6/8.11.6) with ESMTP id i9IIDOu16699; Mon, 18 Oct 2004 11:13:24 -0700 Message-ID: <4174074F.5010607@commandprompt.com> Date: Mon, 18 Oct 2004 11:11:27 -0700 From: "Joshua D. Drake" Organization: Command Prompt, Inc. User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1) X-Accept-Language: en-us, en MIME-Version: 1.0 To: nd02tsk@student.hig.se Cc: pgsql-performance@postgresql.org Subject: Re: How to time several queries? References: <2612.130.243.12.107.1098124104.squirrel@130.243.12.107> In-Reply-To: <2612.130.243.12.107.1098124104.squirrel@130.243.12.107> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/261 X-Sequence-Number: 8737 nd02tsk@student.hig.se wrote: >Hello > >I posted this on the general list but think it would be more appropriate >here. Sorry. > >I know it is possible to time isolated queries through the settting of the >\timing option in psql. This makes PgSQL report the time it took to >perform one operation. > >I would like to know how one can get a time summary of many operations, if >it is at all possible. > > > Hello, You can turn on statement and duration logging in the postgresql.conf >Thank you. > >Tim > > > >---------------------------(end of broadcast)--------------------------- >TIP 4: Don't 'kill -9' the postmaster > > -- Command Prompt, Inc., home of Mammoth PostgreSQL - S/ODBC and S/JDBC Postgresql support, programming shared hosting and dedicated hosting. +1-503-667-4564 - jd@commandprompt.com - http://www.commandprompt.com PostgreSQL Replicator -- production quality replication for PostgreSQL From pgsql-performance-owner@postgresql.org Mon Oct 18 19:09:24 2004 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 812C7329E57 for ; Mon, 18 Oct 2004 19:09:23 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 70667-05 for ; Mon, 18 Oct 2004 18:09:07 +0000 (GMT) Received: from heimdall.hig.se (heimdall.hig.se [130.243.8.180]) by svr1.postgresql.org (Postfix) with SMTP id 8F21E329E7F for ; Mon, 18 Oct 2004 19:09:04 +0100 (BST) Received: from webmail.student.hig.se ([130.243.8.161]) by heimdall.hig.se (SAVSMTP 3.1.0.29) with SMTP id M2004101820045716856 for ; Mon, 18 Oct 2004 20:04:57 +0200 Received: from 130.243.12.107 (SquirrelMail authenticated user nd02tsk); by webmail.student.hig.se with HTTP; Mon, 18 Oct 2004 20:28:24 +0200 (CEST) Message-ID: <2612.130.243.12.107.1098124104.squirrel@130.243.12.107> Date: Mon, 18 Oct 2004 20:28:24 +0200 (CEST) Subject: How to time several queries? From: nd02tsk@student.hig.se To: pgsql-performance@postgresql.org User-Agent: SquirrelMail/1.4.3 X-Mailer: SquirrelMail/1.4.3 MIME-Version: 1.0 Content-Type: text/plain;charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Priority: 3 (Normal) Importance: Normal X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.3 tagged_above=0.0 required=5.0 tests=NO_REAL_NAME X-Spam-Level: X-Archive-Number: 200410/260 X-Sequence-Number: 8736 Hello I posted this on the general list but think it would be more appropriate here. Sorry. I know it is possible to time isolated queries through the settting of the \timing option in psql. This makes PgSQL report the time it took to perform one operation. I would like to know how one can get a time summary of many operations, if it is at all possible. Thank you. Tim From pgsql-performance-owner@postgresql.org Mon Oct 18 19:44:56 2004 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 D60D2329C89 for ; Mon, 18 Oct 2004 19:44:54 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 82034-05 for ; Mon, 18 Oct 2004 18:44:45 +0000 (GMT) Received: from smtp110.mail.sc5.yahoo.com (smtp110.mail.sc5.yahoo.com [66.163.170.8]) by svr1.postgresql.org (Postfix) with SMTP id 42B79329F9B for ; Mon, 18 Oct 2004 19:44:45 +0100 (BST) Received: from unknown (HELO jupiter.black-lion.info) (janwieck@68.80.245.191 with login) by smtp110.mail.sc5.yahoo.com with SMTP; 18 Oct 2004 18:44:43 -0000 Received: from [172.21.8.18] (ismtp.afilias.com [216.217.55.254]) (authenticated bits=0) by jupiter.black-lion.info (8.12.10/8.12.9) with ESMTP id i9IIid94027894; Mon, 18 Oct 2004 14:44:40 -0400 (EDT) (envelope-from JanWieck@Yahoo.com) Message-ID: <41740F11.1070902@Yahoo.com> Date: Mon, 18 Oct 2004 14:44:33 -0400 From: Jan Wieck User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.1) Gecko/20040707 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Bruce Momjian Cc: josh@agliodbs.com, pgsql-performance@postgresql.org, Aaron Glenn Subject: Re: Free PostgreSQL Training, Philadelphia, Oct 30 References: <200410140347.i9E3l5910978@candle.pha.pa.us> In-Reply-To: <200410140347.i9E3l5910978@candle.pha.pa.us> 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=1.6 tagged_above=0.0 required=5.0 tests=RCVD_IN_NJABL_DUL, RCVD_IN_SORBS_DUL X-Spam-Level: * X-Archive-Number: 200410/262 X-Sequence-Number: 8738 On 10/13/2004 11:47 PM, Bruce Momjian wrote: > Josh Berkus wrote: >> Aaron, >> >> > That makes two of us. Hanging out with Tom, Bruce, and others at OSCON >> > 2002 was one of the most informative and fun times I've had. That and >> > I could really stand to brush up on my Postgres basics >> >> You're thinking of Jan. Tom wasn't at OSCON. ;-) > > Ah, but he said 2002 and I think Tom was there that year. > And I wasn't, which makes it rather difficult to hang out with me. I will however be in Malvern too, since it's just around the corner for me. Jan -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com # From pgsql-performance-owner@postgresql.org Mon Oct 18 20:17:44 2004 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 F329032A892 for ; Mon, 18 Oct 2004 20:17:41 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 90817-06 for ; Mon, 18 Oct 2004 19:17:38 +0000 (GMT) Received: from smtp109.mail.sc5.yahoo.com (smtp109.mail.sc5.yahoo.com [66.163.170.7]) by svr1.postgresql.org (Postfix) with SMTP id 309F632A887 for ; Mon, 18 Oct 2004 20:17:35 +0100 (BST) Received: from unknown (HELO jupiter.black-lion.info) (janwieck@68.80.245.191 with login) by smtp109.mail.sc5.yahoo.com with SMTP; 18 Oct 2004 19:17:33 -0000 Received: from [172.21.8.18] (ismtp.afilias.com [216.217.55.254]) (authenticated bits=0) by jupiter.black-lion.info (8.12.10/8.12.9) with ESMTP id i9IJHH94028842; Mon, 18 Oct 2004 15:17:23 -0400 (EDT) (envelope-from JanWieck@Yahoo.com) Message-ID: <417416B7.4060209@Yahoo.com> Date: Mon, 18 Oct 2004 15:17:11 -0400 From: Jan Wieck User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.1) Gecko/20040707 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Simon Riggs Cc: josh@agliodbs.com, pgsql-performance@postgresql.org, testperf-general@pgfoundry.org Subject: Re: First set of OSDL Shared Mem scalability results, some References: In-Reply-To: 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=1.6 tagged_above=0.0 required=5.0 tests=RCVD_IN_NJABL_DUL, RCVD_IN_SORBS_DUL X-Spam-Level: * X-Archive-Number: 200410/263 X-Sequence-Number: 8739 On 10/14/2004 6:36 PM, Simon Riggs wrote: > [...] > I think Jan has said this also in far fewer words, but I'll leave that to > Jan to agree/disagree... I do agree. The total DB size has as little to do with the optimum shared buffer cache size as the total available RAM of the machine. After reading your comments it appears more clear to me. All what those tests did show is the amount of high frequently accessed data in this database population and workload combination. > > I say this: ARC in 8.0 PostgreSQL allows us to sensibly allocate as large a > shared_buffers cache as is required by the database workload, and this > should not be constrained to a small percentage of server RAM. Right. Jan -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com # From pgsql-performance-owner@postgresql.org Mon Oct 18 20:37:59 2004 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 8F77032A887 for ; Mon, 18 Oct 2004 20:37:58 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 99517-03 for ; Mon, 18 Oct 2004 19:37:55 +0000 (GMT) Received: from smtp111.mail.sc5.yahoo.com (smtp111.mail.sc5.yahoo.com [66.163.170.9]) by svr1.postgresql.org (Postfix) with SMTP id 8065F32A34A for ; Mon, 18 Oct 2004 20:37:55 +0100 (BST) Received: from unknown (HELO jupiter.black-lion.info) (janwieck@68.80.245.191 with login) by smtp111.mail.sc5.yahoo.com with SMTP; 18 Oct 2004 19:37:54 -0000 Received: from [172.21.8.18] (ismtp.afilias.com [216.217.55.254]) (authenticated bits=0) by jupiter.black-lion.info (8.12.10/8.12.9) with ESMTP id i9IJbo94029432; Mon, 18 Oct 2004 15:37:52 -0400 (EDT) (envelope-from JanWieck@Yahoo.com) Message-ID: <41741B87.4090104@Yahoo.com> Date: Mon, 18 Oct 2004 15:37:43 -0400 From: Jan Wieck User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.1) Gecko/20040707 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Christopher Browne Cc: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some References: <200410081443.16109.josh@agliodbs.com> In-Reply-To: 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=1.6 tagged_above=0.0 required=5.0 tests=RCVD_IN_NJABL_DUL, RCVD_IN_SORBS_DUL X-Spam-Level: * X-Archive-Number: 200410/264 X-Sequence-Number: 8740 On 10/14/2004 8:10 PM, Christopher Browne wrote: > Quoth simon@2ndquadrant.com ("Simon Riggs"): >> I say this: ARC in 8.0 PostgreSQL allows us to sensibly allocate as >> large a shared_buffers cache as is required by the database >> workload, and this should not be constrained to a small percentage >> of server RAM. > > I don't think that this particularly follows from "what ARC does." The combination of ARC together with the background writer is supposed to allow us to allocate the optimum even if that is large. The former implementation of the LRU without background writer would just hang the server for a long time during a checkpoint, which is absolutely inacceptable for any OLTP system. Jan > > "What ARC does" is to prevent certain conspicuous patterns of > sequential accesses from essentially trashing the contents of the > cache. > > If a particular benchmark does not include conspicuous vacuums or > sequential scans on large tables, then there is little reason to > expect ARC to have a noticeable impact on performance. > > It _could_ be that this implies that ARC allows you to get some use > out of a larger shared cache, as it won't get blown away by vacuums > and Seq Scans. But it is _not_ obvious that this is a necessary > truth. > > _Other_ truths we know about are: > > a) If you increase the shared cache, that means more data that is > represented in both the shared cache and the OS buffer cache, > which seems rather a waste; > > b) The larger the shared cache, the more pages there are for the > backend to rummage through before it looks to the filesystem, > and therefore the more expensive cache misses get. Cache hits > get more expensive, too. Searching through memory is not > costless. -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com # From pgsql-performance-owner@postgresql.org Mon Oct 18 20:43:25 2004 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 C6B9C32A97D for ; Mon, 18 Oct 2004 20:43:23 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 01054-05 for ; Mon, 18 Oct 2004 19:43:12 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 5376F32A887 for ; Mon, 18 Oct 2004 20:43:10 +0100 (BST) 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 6524695; Mon, 18 Oct 2004 12:44:34 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Indexes performance Date: Mon, 18 Oct 2004 12:45:10 -0700 User-Agent: KMail/1.6.2 Cc: charavay References: <417383CD.3050802@ibcp.fr> In-Reply-To: <417383CD.3050802@ibcp.fr> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Message-Id: <200410181245.10810.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/265 X-Sequence-Number: 8741 Charavay, > -------------------------------------------------------------------------= -- >-------- Hash Join =A0(cost=3D6793.29..1716853.80 rows=3D33743101 width=3D= 11) > =A0 =A0Hash Cond: ("outer"."index" =3D "inner"."index") > =A0 =A0-> =A0Seq Scan on lnk =A0(cost=3D0.00..535920.00 rows=3D33743100 w= idth=3D15) > =A0 =A0-> =A0Hash =A0(cost=3D4994.83..4994.83 rows=3D303783 width=3D4) > =A0 =A0 =A0 =A0 =A0-> =A0Seq Scan dic =A0(cost=3D0.00..4994.83 rows=3D303= 783 width=3D4) > (5 rows) According to the estimate, you are selecting all of the rows in the databas= e.=20=20=20 This is going to require a Seq Scan no matter what. --=20 --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-hackers-owner@postgresql.org Mon Oct 18 21:36:43 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id C7D6D32A507 for ; Mon, 18 Oct 2004 21:36:40 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 17568-09 for ; Mon, 18 Oct 2004 20:36:33 +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 251EF32A0C8 for ; Mon, 18 Oct 2004 21:36:34 +0100 (BST) Received: from unknown (HELO jupiter.black-lion.info) (janwieck@68.80.245.191 with login) by smtp104.mail.sc5.yahoo.com with SMTP; 18 Oct 2004 20:36:33 -0000 Received: from [172.21.8.18] (ismtp.afilias.com [216.217.55.254]) (authenticated bits=0) by jupiter.black-lion.info (8.12.10/8.12.9) with ESMTP id i9IKaQ94031018; Mon, 18 Oct 2004 16:36:26 -0400 (EDT) (envelope-from JanWieck@Yahoo.com) Message-ID: <41742943.4050603@Yahoo.com> Date: Mon, 18 Oct 2004 16:36:19 -0400 From: Jan Wieck User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.1) Gecko/20040707 X-Accept-Language: en-us, en MIME-Version: 1.0 To: simon@2ndquadrant.com Cc: Tom Lane , josh@agliodbs.com, pgsql-hackers@postgresql.org, pgsql-performance@postgresql.org Subject: Re: Getting rid of AtEOXact Buffers (was Re: [Testperf-general] References: <28292295$10980398474172c227e88c75.95384212@config16.schlund.de> In-Reply-To: <28292295$10980398474172c227e88c75.95384212@config16.schlund.de> 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=1.6 tagged_above=0.0 required=5.0 tests=RCVD_IN_NJABL_DUL, RCVD_IN_SORBS_DUL X-Spam-Level: * X-Archive-Number: 200410/612 X-Sequence-Number: 60115 On 10/17/2004 3:40 PM, simon@2ndquadrant.com wrote: > Seeing as I've missed the last N messages... I'll just reply to this > one, rather than each of them in turn... > > Tom Lane wrote on 16.10.2004, 18:54:17: >> I wrote: >> > Josh Berkus writes: >> >> First off, two test runs with OProfile are available at: >> >> http://khack.osdl.org/stp/298124/ >> >> http://khack.osdl.org/stp/298121/ >> >> > Hmm. The stuff above 1% in the first of these is >> >> > Counted CPU_CLK_UNHALTED events (clocks processor is not halted) with a unit mask of 0x00 (No unit mask) count 100000 >> > samples % app name symbol name >> > ... >> > 920369 2.1332 postgres AtEOXact_Buffers >> > ... >> >> > In the second test AtEOXact_Buffers is much lower (down around 0.57 >> > percent) but the other suspects are similar. Since the only difference >> > in parameters is shared_buffers (36000 vs 9000), it does look like we >> > are approaching the point where AtEOXact_Buffers is a problem, but so >> > far it's only a 2% drag. > > Yes... as soon as you first mentioned AtEOXact_Buffers, I realised I'd > seen it near the top of the oprofile results on previous tests. > > Although you don't say this, I presume you're acting on the thought that > a 2% drag would soon become a much larger contention point with more > users and/or smaller transactions - since these things are highly > non-linear. > >> >> It occurs to me that given the 8.0 resource manager mechanism, we could >> in fact dispense with AtEOXact_Buffers, or perhaps better turn it into a >> no-op unless #ifdef USE_ASSERT_CHECKING. We'd just get rid of the >> special case for transaction termination in resowner.c and let the >> resource owner be responsible for releasing locked buffers always. The >> OSDL results suggest that this won't matter much at the level of 10000 >> or so shared buffers, but for 100000 or more buffers the linear scan in >> AtEOXact_Buffers is going to become a problem. > > If the resource owner is always responsible for releasing locked > buffers, who releases the locks if the backend crashes? Do we need some > additional code in bgwriter (or?) to clean up buffer locks? If the backend crashes, the postmaster (assuming a possibly corrupted shared memory) restarts the whole lot ... so why bother? > >> >> We could also get rid of the linear search in UnlockBuffers(). The only >> thing it's for anymore is to release a BM_PIN_COUNT_WAITER flag, and >> since a backend could not be doing more than one of those at a time, >> we don't really need an array of flags for that, only a single variable. >> This does not show in the OSDL results, which I presume means that their >> test case is not exercising transaction aborts; but I think we need to >> zap both routines to make the world safe for large shared_buffers >> values. (See also >> http://archives.postgresql.org/pgsql-performance/2004-10/msg00218.php) > > Yes, that's important. > >> Any objection to doing this for 8.0? >> > > As you say, if these issues are definitely kicking in at 100000 > shared_buffers - there's a good few people out there with 800Mb > shared_buffers already. > > Could I also suggest that we adopt your earlier suggestion of raising > the bgwriter parameters as a permanent measure - i.e. changing the > defaults in postgresql.conf. That way, StrategyDirtyBufferList won't > immediately show itself as a problem when using the default parameter > set. It would be a shame to remove one obstacle only to leave another > one following so close behind. [...and that also argues against an > earlier thought to introduce more fine grained values for the > bgwriter's parameters, ISTM] I realized that StrategyDirtyBufferList currently wasts a lot of time by first scanning over all the buffers that haven't even been hit since it's last call and neither have been dirty last time (and thus, are at the beginning of the list and can't be dirty anyway). If we would have a way to give it a smart "point in the list to start scanning" ... > > Also, what will Vacuum delay do to the O(N) effect of > FlushRelationBuffers when called by VACUUM? Will the locks be held for > longer? Vacuum only naps at the points where it checks for interrupts, and at that time it isn't supposed to hold any critical locks. Jan -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com # From pgsql-performance-owner@postgresql.org Mon Oct 18 21:59:08 2004 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 78D0132A585 for ; Mon, 18 Oct 2004 21:59:03 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 27745-01 for ; Mon, 18 Oct 2004 20:58:53 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id B4E8532A552 for ; Mon, 18 Oct 2004 21:58:51 +0100 (BST) 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 6524976 for pgsql-performance@postgresql.org; Mon, 18 Oct 2004 14:00:15 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: [Testperf-general] Re: First set of OSDL Shared Memscalability results, some wierdness ... Date: Mon, 18 Oct 2004 14:00:45 -0700 User-Agent: KMail/1.6.2 References: In-Reply-To: MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410181400.45005.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/267 X-Sequence-Number: 8743 Simon, > I agree that you could test this by running on a bigger or smaller server, > i.e. one with more or less RAM. Running on a faster/slower server at the > same time might alter the results and confuse the situation. Unfortunately, a faster server is the only option I have that also has more RAM. If I double the RAM and double the processors at the same time, what would you expect to happen to the shared_buffers curve? -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-hackers-owner@postgresql.org Mon Oct 18 22:01:36 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 80D9B329F5D for ; Mon, 18 Oct 2004 22:01:35 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 29152-03 for ; Mon, 18 Oct 2004 21:01:31 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 29574329F4C for ; Mon, 18 Oct 2004 22:01:30 +0100 (BST) 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 i9IL1PkW006909; Mon, 18 Oct 2004 17:01:26 -0400 (EDT) To: Jan Wieck Cc: simon@2ndquadrant.com, josh@agliodbs.com, pgsql-hackers@postgresql.org, pgsql-performance@postgresql.org Subject: Re: Getting rid of AtEOXact Buffers (was Re: [Testperf-general] Re: [PERFORM] First set of OSDL Shared Memscalability results, some wierdness ...) In-reply-to: <41742943.4050603@Yahoo.com> References: <28292295$10980398474172c227e88c75.95384212@config16.schlund.de> <41742943.4050603@Yahoo.com> Comments: In-reply-to Jan Wieck message dated "Mon, 18 Oct 2004 16:36:19 -0400" Date: Mon, 18 Oct 2004 17:01:25 -0400 Message-ID: <6908.1098133285@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/613 X-Sequence-Number: 60116 Jan Wieck writes: > I realized that StrategyDirtyBufferList currently wasts a lot of time by > first scanning over all the buffers that haven't even been hit since > it's last call and neither have been dirty last time (and thus, are at > the beginning of the list and can't be dirty anyway). If we would have a > way to give it a smart "point in the list to start scanning" ... I don't think it's true that they *can't* be dirty. (1) Buffers are marked dirty when released, whereas they are moved to the fronts of the lists when acquired. (2) the cntxDirty bit can be set asynchronously to any BufMgrLock'd operation. But it sure seems like we are doing more work than we really need to. One idea I had was for the bgwriter to collect all the dirty pages up to say halfway on the LRU lists, and then write *all* of these, not just the first N, over as many rounds as are needed. Then go back and call StrategyDirtyBufferList again to get a new list. (We don't want it to write every dirty buffer this way, because the ones near the front of the list are likely to be dirtied again right away. But certainly we could write more than 1% of the dirty buffers without getting into the area of the recently-used buffers.) There isn't any particularly good reason for this to share code with checkpoint-style BufferSync, btw. BufferSync could just as easily scan the buffers linearly, since it doesn't matter what order it writes them in. So we could change StrategyDirtyBufferList to stop as soon as it's halfway up the LRU lists, which would save at least a few cycles. regards, tom lane From pgsql-hackers-owner@postgresql.org Mon Oct 18 22:19:56 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 017FE329FAF for ; Mon, 18 Oct 2004 22:19:49 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 31757-06 for ; Mon, 18 Oct 2004 21:19:38 +0000 (GMT) Received: from smtp018.mail.yahoo.com (smtp018.mail.yahoo.com [216.136.174.115]) by svr1.postgresql.org (Postfix) with SMTP id 53ADB329EBB for ; Mon, 18 Oct 2004 22:19:38 +0100 (BST) Received: from unknown (HELO jupiter.black-lion.info) (janwieck@68.80.245.191 with login) by smtp018.mail.yahoo.com with SMTP; 18 Oct 2004 21:19:37 -0000 Received: from [172.21.8.18] (ismtp.afilias.com [216.217.55.254]) (authenticated bits=0) by jupiter.black-lion.info (8.12.10/8.12.9) with ESMTP id i9ILJT94032117; Mon, 18 Oct 2004 17:19:30 -0400 (EDT) (envelope-from JanWieck@Yahoo.com) Message-ID: <41743355.9060005@Yahoo.com> Date: Mon, 18 Oct 2004 17:19:17 -0400 From: Jan Wieck User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.1) Gecko/20040707 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Jan Wieck Cc: simon@2ndquadrant.com, Tom Lane , josh@agliodbs.com, pgsql-hackers@postgresql.org, pgsql-performance@postgresql.org Subject: Autotuning of shared buffer size (was: Re: Getting rid of AtEOXact Buffers (was Re: [Testperf-general] Re: [PERFORM] First set of OSDL Shared Memscalability results, some wierdness ...)) References: <28292295$10980398474172c227e88c75.95384212@config16.schlund.de> <41742943.4050603@Yahoo.com> In-Reply-To: <41742943.4050603@Yahoo.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=1.6 tagged_above=0.0 required=5.0 tests=RCVD_IN_NJABL_DUL, RCVD_IN_SORBS_DUL X-Spam-Level: * X-Archive-Number: 200410/614 X-Sequence-Number: 60117 Trying to think a little out of the box, how "common" is it in modern operating systems to be able to swap out shared memory? Maybe we're not using the ARC algorithm correctly after all. The ARC algorithm does not consider the second level OS buffer cache in it's design. Maybe the total size of the ARC cache directory should not be 2x the size of what is configured as the shared buffer cache, but rather 2x the size of the effective cache size (in 8k pages). If we assume that the job of the T1 queue is better done by the OS buffers anyway (and this is what this discussion seems to point out), we shouldn't hold them in shared buffers (only very few of them and evict them ASAP). We just account for them and assume that the OS will have those cached that we find in our T1 directory. I think with the right configuration for effective cache size, this is a fair assumption. The T2 queue represents the frequently used blocks. If our implementation would cause the unused/used portions of the buffers not to move around, the OS will swap out currently unused portions of the shared buffer cache and utilize those as OS buffers. To verify this theory it would be interesting what the ARC strategy after a long DBT run with a "large" buffer cache thinks a good T2 size would be. Enabling the strategy debug message and running postmaster with -d1 will show that. In theory, this size should be anywhere near the sweet spot. Jan -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com # From pgsql-performance-owner@postgresql.org Tue Oct 19 01:02:23 2004 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 0C098329E48 for ; Tue, 19 Oct 2004 01:02:16 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 71236-10 for ; Tue, 19 Oct 2004 00:02:03 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 2265D329E23 for ; Tue, 19 Oct 2004 01:02:05 +0100 (BST) 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 i9J023jU009791; Mon, 18 Oct 2004 20:02:03 -0400 (EDT) To: charavay Cc: pgsql-performance@postgresql.org Subject: Re: Indexes performance In-reply-to: <417383CD.3050802@ibcp.fr> References: <417383CD.3050802@ibcp.fr> Comments: In-reply-to charavay message dated "Mon, 18 Oct 2004 10:50:21 +0200" Date: Mon, 18 Oct 2004 20:02:03 -0400 Message-ID: <9790.1098144123@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/270 X-Sequence-Number: 8746 charavay writes: > ... So the planner decides to scan 33 000 000 of tuples and we would like to > force it to scan the table dic (303 000 tuples) and to use > the index on the integer index to execute the join. I'm mystified why you think that that will be a superior plan. It still requires visiting every row of the larger table (I assume that all of the larger table's rows do join to some row of the smaller table). All that it accomplishes is to force those visits to occur in a quasi-random order; which not only loses any chance of kernel read-ahead optimizations, but very likely causes each page of the table to be read more than once. AFAICT the planner made exactly the right choice by picking a hashjoin. Have you tried comparing its estimates against actual runtimes for the different plans? (See EXPLAIN ANALYZE.) Offhand the only way I can think of to force it to do the nestloop the other way around from what it wants to is to temporarily drop the index it wants to use. You can do that conveniently like so: begin; alter table dic drop constraint dic_pkey; explain analyze select ...; rollback; which of course would be no good for production, but it should at least serve to destroy your illusions about wanting to do it in production. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Oct 19 08:25:37 2004 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 3351132A0E7 for ; Tue, 19 Oct 2004 08:25:34 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 82887-04 for ; Tue, 19 Oct 2004 07:25:23 +0000 (GMT) Received: from mailer.unicite.fr.netcentrex.net (unknown [62.161.167.249]) by svr1.postgresql.org (Postfix) with ESMTP id D191332A06D for ; Tue, 19 Oct 2004 08:25:18 +0100 (BST) Received: from akira ([192.168.101.228]) by mailer.unicite.fr.netcentrex.net with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2657.72) id 42CYALCL; Tue, 19 Oct 2004 09:25:17 +0200 From: "Alban Medici (NetCentrex)" To: Subject: Re: Queries slow using stored procedures Date: Tue, 19 Oct 2004 09:25:08 +0200 Organization: NetCentrex MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_007C_01C4B5BD.8C390510" X-Mailer: Microsoft Office Outlook, Build 11.0.6353 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 Thread-Index: AcS1PHuwz718w8JtTi2M+MPIPiou1gAbm7fg In-Reply-To: Message-Id: <20041019072518.D191332A06D@svr1.postgresql.org> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.3 tagged_above=0.0 required=5.0 tests=HTML_50_60, HTML_FONTCOLOR_BLUE, HTML_MESSAGE X-Spam-Level: X-Archive-Number: 200410/271 X-Sequence-Number: 8747 This is a multi-part message in MIME format. ------=_NextPart_000_007C_01C4B5BD.8C390510 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit You seem to not have index on botnumber, but in your query bot number is the clause. I don't explain you why the same query is so long. but have your try procedure with a loop structure (witch create cursor) ? you could try CREATE OR REPLACE FUNCTION sp_test_Alban1 ( ) returns integer as ' DECLARE bot char(10); oldbatch INTEGER; rec RECORD; query VARCHAR; BEGIN -- initialisation bot := ''1-7''; query := '' SELECT batchserial FROM transbatch WHERE botnumber = ' || quote_ident(bot) || '' ;''; FOR rec IN EXECUTE var_query LOOP return rec."batchserial ".; END LOOP; --else return 0; END; ' language plpgsql ; does it return the same results in the same time ? _____ From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of Rod Dutton Sent: lundi 18 octobre 2004 20:01 To: pgsql-performance@postgresql.org Subject: [PERFORM] Queries slow using stored procedures Hi, I have a problem where a query inside a function is up to 100 times slower inside a function than as a stand alone query run in psql. The column 'botnumber' is a character(10), is indexed and there are 125000 rows in the table. Help please! This query is fast:- explain analyze SELECT batchserial FROM transbatch WHERE botnumber = '1-7' LIMIT 1; QUERY PLAN ---------------------------------------------------------------------------- ---------------------------------------------------- Limit (cost=0.00..0.42 rows=1 width=4) (actual time=0.73..148.23 rows=1 loops=1) -> Index Scan using ind_tbatchx on transbatch (cost=0.00..18.73 rows=45 width=4) (actual time=0.73..148.22 rows=1 loops=1) Index Cond: (botnumber = '1-7'::bpchar) Total runtime: 148.29 msec (4 rows) This function is slow:- CREATE OR REPLACE FUNCTION sp_test_rod3 ( ) returns integer as ' DECLARE bot char(10); oldbatch INTEGER; BEGIN bot := ''1-7''; SELECT INTO oldbatch batchserial FROM transbatch WHERE botnumber = bot LIMIT 1; IF FOUND THEN RETURN 1; ELSE RETURN 0; END IF; END; ' language plpgsql ; explain analyze SELECT sp_test_rod3(); QUERY PLAN ---------------------------------------------------------------------------- ------------ Result (cost=0.00..0.01 rows=1 width=0) (actual time=1452.39..1452.40 rows=1 loops=1) Total runtime: 1452.42 msec (2 rows) ------=_NextPart_000_007C_01C4B5BD.8C390510 Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable
You seem to not have index on botnumber,  b= ut in=20 your query bot number is the clause.
 
I don't explain you why the same query is so=20 long.
but have your try procedure with a loop structur= e=20 (witch create cursor) ?
 
you could try
 
 
CR= EATE OR=20 REPLACE FUNCTION  sp_test_Alban1 ( )=20 returns integer         
a= s=20 '
DECLARE
  bot char(10);
  oldbatch=20 INTEGER;
  rec=20 RECORD;
=   query=20 VARCHAR;
=
BEGIN
 
  -- initialisation
  bot :=3D ''1-7'';
  query  :=3D ''= =20 SELECT  batchserial FROM transbatch WHERE=20 botnumber  =3D ' || quote_ident(bot) || '' <optionaly your limit=20 clause> ;'';
 
 
   FOR rec = IN=20 EXECUTE var_query  LOOP
        = return=20 rec."batchserial ".;    
   END=20 LOOP;
    <= /DIV>
   =20 --else
    return=20 0;
 
END;
'
language plpgsql =20 ;
does it return the same results in the same time= =20 ? 

From: pgsql-performance-owner@postgresq= l.org=20 [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of Rod=20 Dutton
Sent: lundi 18 octobre 2004 20:01
To:=20 pgsql-performance@postgresql.org
Subject: [PERFORM] Queries slow = using=20 stored procedures

Hi,
 
I have a = problem=20 where a query inside a function is up to 100 times slower inside a function= than=20 as a stand alone query run in psql.
 
The colum= n=20 'botnumber' is a character(10), is indexed and there are 125000 rows i= n the=20 table.
 
Help=20 please!
 
This quer= y is=20 fast:-
 
explain= =20 analyze  
&nb= sp;=20 SELECT batchserial
  FROM transbatch
  WHERE botnumb= er =3D=20 '1-7'
  LIMIT 1;
           =             &nb= sp;            =             &nb= sp;         =20 QUERY=20 PLAN            = ;            &n= bsp;            = ;            &n= bsp;        =20
-----------------------------------------------------------------------= ---------------------------------------------------------
 Limit&nb= sp;=20 (cost=3D0.00..0.42 rows=3D1 width=3D4) (actual time=3D0.73..148.23 rows=3D1= =20 loops=3D1)
   ->  Index Scan using ind_tbatchx on=20 transbatch  (cost=3D0.00..18.73 rows=3D45 width=3D4) (actual time=3D0.= 73..148.22=20 rows=3D1 loops=3D1)
         Ind= ex Cond:=20 (botnumber =3D '1-7'::bpchar)
 Total runtime: 148.29 msec
(4=20 rows)
 
 
This=20 function is slow:-
 
CREATE=20 OR REPLACE FUNCTION  sp_test_rod3 ( ) returns=20 integer         
as=20 '
DECLARE
  bot char(10);
  oldbatch=20 INTEGER;
BEGIN
 
 =20 bot :=3D ''1-7'';
 
 =20 SELECT INTO oldbatch batchserial
  FROM transbatch
  WHERE= =20 botnumber =3D bot
  LIMIT 1;
 
 =20 IF FOUND THEN
    RETURN 1;
 =20 ELSE
    RETURN 0;
  END=20 IF;
 
END;
'
language plpgsql  ;
 
explain analyze SELECT sp_test_rod3();
           =             &nb= sp;            =   =20 QUERY=20 PLAN            = ;            &n= bsp;            = ; =20
-----------------------------------------------------------------------= -----------------
 Result =20 (cost=3D0.00..0.01 rows=3D1 width=3D0) (actual time=3D1452.39..1452.40 rows= =3D1=20 loops=3D1)
 Total runtime: 1452.42 msec
(2=20 rows)
------=_NextPart_000_007C_01C4B5BD.8C390510-- From pgsql-performance-owner@postgresql.org Tue Oct 19 16:15:21 2004 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 EBE35329F3E for ; Tue, 19 Oct 2004 16:15:19 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 26376-05 for ; Tue, 19 Oct 2004 15:15:15 +0000 (GMT) Received: from MER-TSM1.NASDAQ.COM (mer-tsm1.nasdaq.com [206.200.253.23]) by svr1.postgresql.org (Postfix) with ESMTP id 7B46D329F28 for ; Tue, 19 Oct 2004 16:15:12 +0100 (BST) Received: from 206.200.98.133 (mer-exb1.corp.nasdaq.com) by MER-TSM1.NASDAQ.COM with ESMTP (Nasdaq SMTP Gateway and will be monitored for compliance. (MMS v5.6.3)); Tue, 19 Oct 2004 11:14:55 -0400 X-Server-Uuid: EECCD282-DD20-463F-80F0-CC7552CC1D9A Received: from MER-EXCH1.CORP.NASDAQ.COM ([206.200.99.75]) by MER-EXB1.CORP.NASDAQ.COM with Microsoft SMTPSVC(5.0.2195.6713); Tue, 19 Oct 2004 11:14:55 -0400 X-MimeOLE: Produced By Microsoft Exchange V6.0.6487.1 content-class: urn:content-classes:message MIME-Version: 1.0 Subject: Why isn't this index being used? Date: Tue, 19 Oct 2004 11:14:55 -0400 Message-ID: X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: Why isn't this index being used? Thread-Index: AcS17mM1apA4pmIcRmWqLEsoOHDDYw== From: "Knutsen, Mark" To: pgsql-performance@postgresql.org X-OriginalArrivalTime: 19 Oct 2004 15:14:55.0288 (UTC) FILETIME=[63B53380:01C4B5EE] X-WSS-ID: 6D6BF0E51V85192542-01-01 Content-Type: multipart/alternative; boundary="----_=_NextPart_001_01C4B5EE.638E0C40" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.1 tagged_above=0.0 required=5.0 tests=HTML_60_70, HTML_MESSAGE X-Spam-Level: X-Archive-Number: 200410/272 X-Sequence-Number: 8748 This is a multi-part message in MIME format. ------_=_NextPart_001_01C4B5EE.638E0C40 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: quoted-printable The following is from a database of several hundred million rows of real data that has been VACUUM ANALYZEd. =20 Why isn't the index being used for a query that seems tailor-made for it? The results (6,300 rows) take about ten minutes to retrieve with a sequential scan. =20 A copy of this database with "integer" in place of "smallint", a primary key in column order (date, time, type, subtype) and a secondary index in the required order (type, subtype, date, time) correctly uses the secondary index to return results in under a second. =20 Actually, the integer version is the first one I made, and the smallint is the copy, but that shouldn't matter. =20 Postgres is version "postgresql-server-7.3.4-3.rhl9" from Red Hat Linux 9. =20 =3D=3D=3D=3D=3D =20 testdb2=3D# \d db Table "public.db" Column | Type | Modifiers ---------+------------------------+----------- date | date | not null time | time without time zone | not null type | smallint | not null subtype | smallint | not null value | integer | Indexes: db_pkey primary key btree ("type", subtype, date, "time") =20 testdb2=3D# set enable_seqscan to off; SET =20 testdb2=3D# explain select * from db where type=3D90 and subtype=3D70 and date=3D'7/1/2004'; QUERY PLAN ------------------------------------------------------------------------ ------ Seq Scan on db (cost=3D100000000.00..107455603.76 rows=3D178 width=3D20) Filter: (("type" =3D 90) AND (subtype =3D 70) AND (date =3D '2004-07-01'::date)) (2 rows) ------_=_NextPart_001_01C4B5EE.638E0C40 Content-Type: text/html; charset=us-ascii Content-Transfer-Encoding: quoted-printable

The following is from a database of several hund= red million rows of real data that has been VACUUM ANALYZEd.

 

Why isn't the index being used for a query that = seems tailor-made for it? The results (6,300 rows) take about ten minutes to retr= ieve with a sequential scan.

 

A copy of this database with "integer"= in place of "smallint", a primary key in column order (date, time, t= ype, subtype) and a secondary index in the required order (type, subtype, date, time) correctly uses the secondary index to return results in under a secon= d.

 

Actually, the integer version is the first one I made, and the smallint is the copy, but that shouldn't matter.

 

Postgres is version "postgresql-server-7.3.= 4-3.rhl9" from Red Hat Linux 9.

 

=3D=3D=3D=3D=3D

 

testdb2=3D# \d db

        =       Table "public.db"

 Column  |          Type          | Modifiers=

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

 date    | date            = ;       | not null

 time    | time without time zone | not null

 type    | smallint           &= nbsp;   | not null

 subtype | smallint           &= nbsp;   | not null

 value   | integer           &n= bsp;    |

Indexes: db_pkey primary key btree ("type", subtype, date, "time")

 

testdb2=3D# set enable_seqscan to off;

SET

 

testdb2=3D# explain select * from db where type= =3D90 and subtype=3D70 and date=3D'7/1/2004';

        =             &nb= sp;             QUERY PLAN

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

 Seq Scan on db  (cost=3D100000000.00..107455603.76 rows=3D178 width=3D20)=

   Filter: (("type" =3D 90) = AND (subtype =3D 70) AND (date =3D '2004-07-01'::date))

(2 rows)

------_=_NextPart_001_01C4B5EE.638E0C40-- From pgsql-performance-owner@postgresql.org Tue Oct 19 16:28:37 2004 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 63382329F08 for ; Tue, 19 Oct 2004 16:28:32 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 29928-09 for ; Tue, 19 Oct 2004 15:28:22 +0000 (GMT) Received: from neomail10.traderonline.com (email.traderonline.com [65.213.231.10]) by svr1.postgresql.org (Postfix) with ESMTP id A8FC2329EC2 for ; Tue, 19 Oct 2004 16:28:21 +0100 (BST) Received: from [192.168.1.68] (64-139-89-109-ubr02b-epensb01-pa.hfc.comcastbusiness.net [64.139.89.109]) by neomail10.traderonline.com (8.12.10/8.12.8) with ESMTP id i9JFSGB1009630; Tue, 19 Oct 2004 11:28:17 -0400 Message-ID: <41753290.8010709@ptd.net> Date: Tue, 19 Oct 2004 11:28:16 -0400 From: Doug Y User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: "Knutsen, Mark" Cc: pgsql-performance@postgresql.org Subject: Re: Why isn't this index being used? 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/273 X-Sequence-Number: 8749 Hi, I ran into a similar problem using bigints... See: http://www.postgresql.org/docs/7.3/static/datatype.html#DATATYPE-INT small & big int have to be cast when used in querries... try: explain select * from db where type=90::smallint and subtype=70::smallint and date='7/1/2004'; or explain select * from db where type='90' and subtype='70' and date='7/1/2004'; Knutsen, Mark wrote: > The following is from a database of several hundred million rows of > real data that has been VACUUM ANALYZEd. > > > > Why isn't the index being used for a query that seems tailor-made for > it? The results (6,300 rows) take about ten minutes to retrieve with a > sequential scan. > > > > A copy of this database with "integer" in place of "smallint", a > primary key in column order (date, time, type, subtype) and a > secondary index in the required order (type, subtype, date, time) > correctly uses the secondary index to return results in under a second. > > > > Actually, the integer version is the first one I made, and the > smallint is the copy, but that shouldn't matter. > > > > Postgres is version "postgresql-server-7.3.4-3.rhl9" from Red Hat Linux 9. > > > > ===== > > > > testdb2=# \d db > > Table "public.db" > > Column | Type | Modifiers > > ---------+------------------------+----------- > > date | date | not null > > time | time without time zone | not null > > type | smallint | not null > > subtype | smallint | not null > > value | integer | > > Indexes: db_pkey primary key btree ("type", subtype, date, "time") > > > > testdb2=# set enable_seqscan to off; > > SET > > > > testdb2=# explain select * from db where type=90 and subtype=70 and > date='7/1/2004'; > > QUERY PLAN > > ------------------------------------------------------------------------------ > > Seq Scan on db (cost=100000000.00..107455603.76 rows=178 width=20) > > Filter: (("type" = 90) AND (subtype = 70) AND (date = > '2004-07-01'::date)) > > (2 rows) > From pgsql-performance-owner@postgresql.org Tue Oct 19 16:34:10 2004 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 22CE432A15B for ; Tue, 19 Oct 2004 16:34:07 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 33090-10 for ; Tue, 19 Oct 2004 15:34:03 +0000 (GMT) Received: from MER-TSM1.NASDAQ.COM (mer-tsm1.nasdaq.com [206.200.253.23]) by svr1.postgresql.org (Postfix) with ESMTP id 500E9329F5F for ; Tue, 19 Oct 2004 16:34:02 +0100 (BST) Received: from 206.200.98.133 (mer-exb1.corp.nasdaq.com) by MER-TSM1.NASDAQ.COM with ESMTP (Nasdaq SMTP Gateway and will be monitored for compliance. (MMS v5.6.3)); Tue, 19 Oct 2004 11:33:51 -0400 X-Server-Uuid: EECCD282-DD20-463F-80F0-CC7552CC1D9A Received: from MER-EXCH1.CORP.NASDAQ.COM ([206.200.99.75]) by MER-EXB1.CORP.NASDAQ.COM with Microsoft SMTPSVC(5.0.2195.6713); Tue, 19 Oct 2004 11:33:51 -0400 X-MimeOLE: Produced By Microsoft Exchange V6.0.6487.1 content-class: urn:content-classes:message MIME-Version: 1.0 Subject: Re: Why isn't this index being used? Date: Tue, 19 Oct 2004 11:33:50 -0400 Message-ID: X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Why isn't this index being used? Thread-Index: AcS18Enzx69p7zdlRmm1ga2GD0WIcgAAG2lQ From: "Knutsen, Mark" To: "Doug Y" Cc: pgsql-performance@postgresql.org X-OriginalArrivalTime: 19 Oct 2004 15:33:51.0030 (UTC) FILETIME=[08A9AD60:01C4B5F1] X-WSS-ID: 6D6BEC551V85193690-01-01 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: quoted-printable X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/274 X-Sequence-Number: 8750 (Why don't replies automatically go to the list?) Sure enough, quoting the constants fixes the problem. Is it a best practice to always quote constants? > -----Original Message----- > From: Doug Y [mailto:dylists@ptd.net] > Sent: Tuesday, October 19, 2004 11:28 AM > To: Knutsen, Mark > Cc: pgsql-performance@postgresql.org > Subject: Re: [PERFORM] Why isn't this index being used? >=20 > Hi, I ran into a similar problem using bigints... >=20 > See: > http://www.postgresql.org/docs/7.3/static/datatype.html#DATATYPE-INT >=20 > small & big int have to be cast when used in querries... try: > explain select * from db where type=3D90::smallint and > subtype=3D70::smallint and date=3D'7/1/2004'; > or > explain select * from db where type=3D'90' and subtype=3D'70' and > date=3D'7/1/2004'; >=20 > Knutsen, Mark wrote: >=20 > > The following is from a database of several hundred million rows of > > real data that has been VACUUM ANALYZEd. > > > > > > > > Why isn't the index being used for a query that seems tailor-made for > > it? The results (6,300 rows) take about ten minutes to retrieve with a > > sequential scan. > > > > > > > > A copy of this database with "integer" in place of "smallint", a > > primary key in column order (date, time, type, subtype) and a > > secondary index in the required order (type, subtype, date, time) > > correctly uses the secondary index to return results in under a second. > > > > > > > > Actually, the integer version is the first one I made, and the > > smallint is the copy, but that shouldn't matter. > > > > > > > > Postgres is version "postgresql-server-7.3.4-3.rhl9" from Red Hat Linux > 9. > > > > > > > > =3D=3D=3D=3D=3D > > > > > > > > testdb2=3D# \d db > > > > Table "public.db" > > > > Column | Type | Modifiers > > > > ---------+------------------------+----------- > > > > date | date | not null > > > > time | time without time zone | not null > > > > type | smallint | not null > > > > subtype | smallint | not null > > > > value | integer | > > > > Indexes: db_pkey primary key btree ("type", subtype, date, "time") > > > > > > > > testdb2=3D# set enable_seqscan to off; > > > > SET > > > > > > > > testdb2=3D# explain select * from db where type=3D90 and subtype=3D70 a= nd > > date=3D'7/1/2004'; > > > > QUERY PLAN > > > > ------------------------------------------------------------------------ > ------ > > > > Seq Scan on db (cost=3D100000000.00..107455603.76 rows=3D178 width=3D= 20) > > > > Filter: (("type" =3D 90) AND (subtype =3D 70) AND (date =3D > > '2004-07-01'::date)) > > > > (2 rows) From pgsql-performance-owner@postgresql.org Tue Oct 19 16:38:27 2004 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 25B5C32A0A2 for ; Tue, 19 Oct 2004 16:38:26 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 34760-08 for ; Tue, 19 Oct 2004 15:38:24 +0000 (GMT) Received: from roue.portalpotty.net (roue.portalpotty.net [69.44.62.206]) by svr1.postgresql.org (Postfix) with ESMTP id 30A3332A0A1 for ; Tue, 19 Oct 2004 16:38:23 +0100 (BST) Received: from max by roue.portalpotty.net with local (Exim 4.34) id 1CJw45-0005YA-F7 for pgsql-performance@postgresql.org; Tue, 19 Oct 2004 08:38:21 -0700 Date: Tue, 19 Oct 2004 11:38:21 -0400 From: Max Baker To: pgsql-performance@postgresql.org Subject: Vacuum takes a really long time, vacuum full required Message-ID: <20041019153821.GA21258@warped.org> Mail-Followup-To: pgsql-performance@postgresql.org Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.4.1i X-AntiAbuse: This header was added to track abuse, please include it with any abuse report X-AntiAbuse: Primary Hostname - roue.portalpotty.net X-AntiAbuse: Original Domain - postgresql.org X-AntiAbuse: Originator/Caller UID/GID - [514 32003] / [47 12] X-AntiAbuse: Sender Address Domain - roue.portalpotty.net X-Source: /usr/bin/mutt X-Source-Args: mutt X-Source-Dir: /home/max X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/275 X-Sequence-Number: 8751 Hi Folks, This is my _4th_ time trying to post this, me and the mailing list software are fighting. I think it's because of the attachments so I'll just put links to them instead. All apologies if this gets duplicated. I've been having problems maintaining the speed of the database in the long run. VACUUMs of the main tables happen a few times a day after maybe 50,000 or less rows are added and deleted (say 6 times a day). I have a whole lot (probably too much) indexing going on to try to speed things up. Whatever the case, the database still slows down to a halt after a month or so, and I have to go in and shut everything down and do a VACUUM FULL by hand. One index (of many many) takes 2000 seconds to vacuum. The whole process takes a few hours. I would love suggestions on what I can do either inside my application, or from a dba point of view to keep the database maintained without having to inflict downtime. This is for 'Netdisco' -- an open source network management software by the way. I'ld like to fix this for everyone who uses it. Sys Info : $ uname -a FreeBSD xxxx.ucsc.edu 4.10-STABLE FreeBSD 4.10-STABLE #0: Mon Aug 16 14:56:19 PDT 2004 root@xxxx.ucsc.edu:/usr/src/sys/compile/xxxx i386 $ pg_config --version PostgreSQL 7.3.2 $ cat postgresql.conf max_connections = 32 shared_buffers = 3900 # 30Mb - Bsd current kernel limit max_fsm_relations = 1000 # min 10, fsm is free space map, ~40 bytes max_fsm_pages = 10000 # min 1000, fsm is free space map, ~6 bytes max_locks_per_transaction = 64 # min 10 wal_buffers = 8 # min 4, typically 8KB each The log of the vacuum and the db schema could not be attached, so they are at : http://netdisco.net/db_vacuum.txt http://netdisco.net/pg_all.input Thanks for any help! -m From pgsql-performance-owner@postgresql.org Tue Oct 19 16:44:48 2004 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 6A7A532A138 for ; Tue, 19 Oct 2004 16:44:45 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 37627-09 for ; Tue, 19 Oct 2004 15:44:43 +0000 (GMT) Received: from tht.net (vista.tht.net [216.126.88.2]) by svr1.postgresql.org (Postfix) with ESMTP id 91A0B32A11E for ; Tue, 19 Oct 2004 16:44:42 +0100 (BST) Received: from [134.22.69.212] (dyn-69-212.tor.dsl.tht.net [134.22.69.212]) by tht.net (Postfix) with ESMTP id 4824176AD0; Tue, 19 Oct 2004 11:41:07 -0400 (EDT) Subject: Re: Vacuum takes a really long time, vacuum full required From: Rod Taylor To: Max Baker Cc: Postgresql Performance In-Reply-To: <20041019153821.GA21258@warped.org> References: <20041019153821.GA21258@warped.org> Content-Type: text/plain Message-Id: <1098200416.750.238.camel@home> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Tue, 19 Oct 2004 11:40:17 -0400 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/276 X-Sequence-Number: 8752 > Whatever the case, the database still slows down to a halt after a month or > so, and I have to go in and shut everything down and do a VACUUM FULL by > hand. One index (of many many) takes 2000 seconds to vacuum. The whole > process takes a few hours. Do a REINDEX on that table instead, and regular vacuum more frequently. > $ pg_config --version > PostgreSQL 7.3.2 7.4.x deals with index growth a little better 7.3 and older did. From pgsql-performance-owner@postgresql.org Tue Oct 19 16:47:27 2004 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 9D1BA32A0D8 for ; Tue, 19 Oct 2004 16:47:25 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 40473-02 for ; Tue, 19 Oct 2004 15:47:22 +0000 (GMT) Received: from smtp100.rog.mail.re2.yahoo.com (smtp100.rog.mail.re2.yahoo.com [206.190.36.78]) by svr1.postgresql.org (Postfix) with SMTP id 1D86B32A2A7 for ; Tue, 19 Oct 2004 16:47:20 +0100 (BST) Received: from unknown (HELO phlogiston.dydns.org) (a.sullivan@rogers.com@65.49.125.184 with login) by smtp100.rog.mail.re2.yahoo.com with SMTP; 19 Oct 2004 15:47:19 -0000 Received: by phlogiston.dydns.org (Postfix, from userid 1000) id 541F04058; Tue, 19 Oct 2004 11:47:18 -0400 (EDT) Date: Tue, 19 Oct 2004 11:47:18 -0400 From: Andrew Sullivan To: pgsql-performance@postgresql.org Subject: Re: Why isn't this index being used? Message-ID: <20041019154718.GG3359@phlogiston.dyndns.org> Mail-Followup-To: pgsql-performance@postgresql.org References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.5.6+20040722i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/277 X-Sequence-Number: 8753 On Tue, Oct 19, 2004 at 11:33:50AM -0400, Knutsen, Mark wrote: > (Why don't replies automatically go to the list?) Because sometimes you don't want them to. There's been dozens of discussions about this. BTW, mutt has a nice feature which allows you to reply to lists -- I imagine other MUAs have such a feature too. > Sure enough, quoting the constants fixes the problem. > > Is it a best practice to always quote constants? No, but it's very useful in these cases. The problem is I believe this is fixed in 8.0, BTW. See the FAQ, question 4.8 A -- Andrew Sullivan | ajs@crankycanuck.ca I remember when computers were frustrating because they *did* exactly what you told them to. That actually seems sort of quaint now. --J.D. Baldwin From pgsql-performance-owner@postgresql.org Tue Oct 19 17:21:12 2004 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 6920B32A53B for ; Tue, 19 Oct 2004 17:21:08 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 50862-08 for ; Tue, 19 Oct 2004 16:21:06 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id F36EF32A540 for ; Tue, 19 Oct 2004 17:21:05 +0100 (BST) 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 i9JGL5nA016630; Tue, 19 Oct 2004 12:21:05 -0400 (EDT) To: Max Baker Cc: pgsql-performance@postgresql.org Subject: Re: Vacuum takes a really long time, vacuum full required In-reply-to: <20041019153821.GA21258@warped.org> References: <20041019153821.GA21258@warped.org> Comments: In-reply-to Max Baker message dated "Tue, 19 Oct 2004 11:38:21 -0400" Date: Tue, 19 Oct 2004 12:21:04 -0400 Message-ID: <16629.1098202864@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/278 X-Sequence-Number: 8754 Max Baker writes: > I've been having problems maintaining the speed of the database in the > long run. VACUUMs of the main tables happen a few times a day after maybe > 50,000 or less rows are added and deleted (say 6 times a day). > I have a whole lot (probably too much) indexing going on to try to speed > things up. > Whatever the case, the database still slows down to a halt after a month or > so, and I have to go in and shut everything down and do a VACUUM FULL by > hand. One index (of many many) takes 2000 seconds to vacuum. The whole > process takes a few hours. The first and foremost recommendation is to increase your FSM settings; you seem to be using the defaults, which are pegged for a database size of not more than about 100Mb. Second is to update to PG 7.4. I think you are probably suffering from index bloat to some extent, and 7.4 should help. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Oct 19 17:27:00 2004 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 982E432A091 for ; Tue, 19 Oct 2004 17:26:58 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 52031-10 for ; Tue, 19 Oct 2004 16:26:48 +0000 (GMT) Received: from ar-sd.net (unknown [82.77.155.72]) by svr1.postgresql.org (Postfix) with ESMTP id 287E3329F7A for ; Tue, 19 Oct 2004 17:26:46 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by ar-sd.net (Postfix) with ESMTP id D122420491 for ; Tue, 19 Oct 2004 19:30:35 +0300 (EEST) Received: from ar-sd.net ([127.0.0.1]) by localhost (linz [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 30316-01 for ; Tue, 19 Oct 2004 19:30:34 +0300 (EEST) Received: from forge (unknown [192.168.0.11]) by ar-sd.net (Postfix) with SMTP id 2199D203AA for ; Tue, 19 Oct 2004 19:30:34 +0300 (EEST) Message-ID: <0bef01c4b5f8$70026ae0$0b00a8c0@forge> From: "Andrei Bintintan" To: Subject: Index not used in query. Why? Date: Tue, 19 Oct 2004 19:26:49 +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 6.00.2720.3000 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2727.1300 X-Virus-Scanned: by amavisd-new at ar-sd.net X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/279 X-Sequence-Number: 8755 Hi to all! I have the following query. The execution time is very big, it doesn't use the indexes and I don't understand why... SELECT count(o.id) FROM orders o INNER JOIN report r ON o.id=r.id_order INNER JOIN status s ON o.id_status=s.id INNER JOIN contact c ON o.id_ag=c.id INNER JOIN endkunde e ON o.id_endkunde=e.id INNER JOIN zufriden z ON r.id_zufriden=z.id INNER JOIN plannung v ON v.id=o.id_plannung INNER JOIN mpsworker w ON v.id_worker=w.id INNER JOIN person p ON p.id = w.id_person WHERE o.id_status > 3 The query explain: Aggregate (cost=32027.38..32027.38 rows=1 width=4) -> Hash Join (cost=23182.06..31944.82 rows=33022 width=4) Hash Cond: ("outer".id_person = "inner".id) -> Hash Join (cost=23179.42..31446.85 rows=33022 width=8) Hash Cond: ("outer".id_endkunde = "inner".id) -> Hash Join (cost=21873.54..28891.42 rows=33022 width=12) Hash Cond: ("outer".id_ag = "inner".id) -> Hash Join (cost=21710.05..28067.50 rows=33021 width=16) Hash Cond: ("outer".id_status = "inner".id) -> Hash Join (cost=21708.97..27571.11 rows=33021 width=20) Hash Cond: ("outer".id_worker = "inner".id) -> Hash Join (cost=21707.49..27074.31 rows=33021 width=20) Hash Cond: ("outer".id_zufriden = "inner".id) -> Hash Join (cost=21706.34..26564.09 rows=35772 width=24) Hash Cond: ("outer".id_plannung = "inner".id) -> Hash Join (cost=20447.15..23674.04 rows=35771 width=24) Hash Cond: ("outer".id = "inner".id_order) -> Seq Scan on orders o (cost=0.00..1770.67 rows=36967 width=20) Filter: (id_status > 3) -> Hash (cost=20208.32..20208.32 rows=37132 width=8) -> Seq Scan on report r (cost=0.00..20208.32 rows=37132 width=8) -> Hash (cost=913.15..913.15 rows=54015 width=8) -> Seq Scan on plannung v (cost=0.00..913.15 rows=54015 width=8) -> Hash (cost=1.12..1.12 rows=12 width=4) -> Seq Scan on zufriden z (cost=0.00..1.12 rows=12 width=4) -> Hash (cost=1.39..1.39 rows=39 width=8) -> Seq Scan on mpsworker w (cost=0.00..1.39 rows=39 width=8) -> Hash (cost=1.06..1.06 rows=6 width=4) -> Seq Scan on status s (cost=0.00..1.06 rows=6 width=4) -> Hash (cost=153.19..153.19 rows=4119 width=4) -> Seq Scan on contact c (cost=0.00..153.19 rows=4119 width=4) -> Hash (cost=1077.91..1077.91 rows=38391 width=4) -> Seq Scan on endkunde e (cost=0.00..1077.91 rows=38391 width=4) -> Hash (cost=2.51..2.51 rows=51 width=4) -> Seq Scan on person p (cost=0.00..2.51 rows=51 width=4) As you can see, no index is used.I made everywhere indexes where the jons are made. If I use the following query the indexes are used: SELECT count(o.id) FROM orders o INNER JOIN report r ON o.id=r.id_order INNER JOIN status s ON o.id_status=s.id INNER JOIN contact c ON o.id_ag=c.id INNER JOIN endkunde e ON o.id_endkunde=e.id INNER JOIN zufriden z ON r.id_zufriden=z.id INNER JOIN plannung v ON v.id=o.id_plannung INNER JOIN mpsworker w ON v.id_worker=w.id INNER JOIN person p ON p.id = w.id_person WHERE o.id_status =4 Aggregate (cost=985.55..985.55 rows=1 width=4) -> Hash Join (cost=5.28..985.42 rows=50 width=4) Hash Cond: ("outer".id_person = "inner".id) -> Hash Join (cost=2.64..982.03 rows=50 width=8) Hash Cond: ("outer".id_worker = "inner".id) -> Nested Loop (cost=1.15..979.79 rows=50 width=8) -> Nested Loop (cost=1.15..769.64 rows=49 width=8) -> Nested Loop (cost=1.15..535.57 rows=48 width=12) -> Seq Scan on status s (cost=0.00..1.07 rows=1 width=4) Filter: (4 = id) -> Nested Loop (cost=1.15..534.01 rows=48 width=16) -> Hash Join (cost=1.15..366.37 rows=47 width=20) Hash Cond: ("outer".id_zufriden = "inner".id) -> Nested Loop (cost=0.00..364.48 rows=51 width=24) -> Index Scan using orders_id_status_idx on orders o (cost=0.00..69.55 rows=52 width=20) Index Cond: (id_status = 4) -> Index Scan using report_id_order_idx on report r (cost=0.00..5.66 rows=1 width=8) Index Cond: ("outer".id = r.id_order) -> Hash (cost=1.12..1.12 rows=12 width=4) -> Seq Scan on zufriden z (cost=0.00..1.12 rows=12 width=4) -> Index Scan using endkunde_pkey on endkunde e (cost=0.00..3.55 rows=1 width=4) Index Cond: ("outer".id_endkunde = e.id) -> Index Scan using contact_pkey on contact c (cost=0.00..4.86 rows=1 width=4) Index Cond: ("outer".id_ag = c.id) -> Index Scan using plannung_pkey on plannung v (cost=0.00..4.28 rows=1 width=8) Index Cond: (v.id = "outer".id_plannung) -> Hash (cost=1.39..1.39 rows=39 width=8) -> Seq Scan on mpsworker w (cost=0.00..1.39 rows=39 width=8) -> Hash (cost=2.51..2.51 rows=51 width=4) -> Seq Scan on person p (cost=0.00..2.51 rows=51 width=4) Best regards, Andy. From pgsql-performance-owner@postgresql.org Tue Oct 19 17:53:10 2004 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 C843032A0EC for ; Tue, 19 Oct 2004 17:53:07 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 64106-02 for ; Tue, 19 Oct 2004 16:53:05 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id DE92E32A0A2 for ; Tue, 19 Oct 2004 17:53:04 +0100 (BST) 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 i9JGqnaR016866; Tue, 19 Oct 2004 12:52:50 -0400 (EDT) To: "Andrei Bintintan" Cc: pgsql-performance@postgresql.org Subject: Re: Index not used in query. Why? In-reply-to: <0bef01c4b5f8$70026ae0$0b00a8c0@forge> References: <0bef01c4b5f8$70026ae0$0b00a8c0@forge> Comments: In-reply-to "Andrei Bintintan" message dated "Tue, 19 Oct 2004 19:26:49 +0300" Date: Tue, 19 Oct 2004 12:52:49 -0400 Message-ID: <16865.1098204769@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/280 X-Sequence-Number: 8756 "Andrei Bintintan" writes: > Hi to all! I have the following query. The execution time is very big, it > doesn't use the indexes and I don't understand why... Indexes are not necessarily the best way to do a large join. > If I use the following query the indexes are used: The key reason this wins seems to be that the id_status = 4 condition is far more selective than id_status > 3 (the estimates are 52 and 36967 rows respectively ... is that accurate?) which means that the second query is inherently about 1/700th as much work. This, and not the use of indexes, is the fundamental reason why it's faster. regards, tom lane From pgsql-performance-owner@postgresql.org Wed Oct 20 16:08:52 2004 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 359EF32A00B for ; Tue, 19 Oct 2004 18:49:59 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 82656-04 for ; Tue, 19 Oct 2004 17:49:44 +0000 (GMT) Received: from ar-sd.net (unknown [82.77.155.72]) by svr1.postgresql.org (Postfix) with ESMTP id 4D5CF329FA4 for ; Tue, 19 Oct 2004 18:49:42 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by ar-sd.net (Postfix) with ESMTP id 7BD5020491; Tue, 19 Oct 2004 20:53:32 +0300 (EEST) Received: from ar-sd.net ([127.0.0.1]) by localhost (linz [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 30470-07; Tue, 19 Oct 2004 20:53:31 +0300 (EEST) Received: from forge (unknown [192.168.0.11]) by ar-sd.net (Postfix) with SMTP id 4E26D203AA; Tue, 19 Oct 2004 20:53:31 +0300 (EEST) Message-ID: <0c4501c4b604$06f15460$0b00a8c0@forge> From: "Contact AR-SD.NET" To: "Tom Lane" Cc: References: <0bef01c4b5f8$70026ae0$0b00a8c0@forge> <16865.1098204769@sss.pgh.pa.us> Subject: Re: Index not used in query. Why? Date: Tue, 19 Oct 2004 20:49:45 +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 6.00.2720.3000 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2727.1300 X-Virus-Scanned: by amavisd-new at ar-sd.net X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/318 X-Sequence-Number: 8794 Is there a solution to make it faster? At the end I need only in the query the id_status =4 and 6, but if I write in the sql query (where condition) where id_status in (4,6), the explain says the same(the slow version). For example: SELECT count(o.id) FROM orders o INNER JOIN report r ON o.id=r.id_order INNER JOIN status s ON o.id_status=s.id INNER JOIN contact c ON o.id_ag=c.id INNER JOIN endkunde e ON o.id_endkunde=e.id INNER JOIN zufriden z ON r.id_zufriden=z.id INNER JOIN plannung v ON v.id=o.id_plannung INNER JOIN mpsworker w ON v.id_worker=w.id INNER JOIN person p ON p.id = w.id_person WHERE o.id_status in (4,6); The result for this query is also without index searches. I really have to make this query a little more faster. Suggestions? Regards, Andy. ----- Original Message ----- From: "Tom Lane" To: "Andrei Bintintan" Cc: Sent: Tuesday, October 19, 2004 7:52 PM Subject: Re: [PERFORM] Index not used in query. Why? > "Andrei Bintintan" writes: > > Hi to all! I have the following query. The execution time is very big, it > > doesn't use the indexes and I don't understand why... > > Indexes are not necessarily the best way to do a large join. > > > If I use the following query the indexes are used: > > The key reason this wins seems to be that the id_status = 4 condition > is far more selective than id_status > 3 (the estimates are 52 and 36967 > rows respectively ... is that accurate?) which means that the second > query is inherently about 1/700th as much work. This, and not the use > of indexes, is the fundamental reason why it's faster. > > regards, tom lane > > ---------------------------(end of broadcast)--------------------------- > TIP 4: Don't 'kill -9' the postmaster > From pgsql-performance-owner@postgresql.org Tue Oct 19 19:12:56 2004 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 57C4A32A401 for ; Tue, 19 Oct 2004 19:12:54 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 90732-07 for ; Tue, 19 Oct 2004 18:12:47 +0000 (GMT) Received: from roue.portalpotty.net (roue.portalpotty.net [69.44.62.206]) by svr1.postgresql.org (Postfix) with ESMTP id 107E432A42F for ; Tue, 19 Oct 2004 19:12:47 +0100 (BST) Received: from max by roue.portalpotty.net with local (Exim 4.34) id 1CJyTV-0006EW-Ix; Tue, 19 Oct 2004 11:12:45 -0700 Date: Tue, 19 Oct 2004 14:12:45 -0400 From: Max Baker To: Rod Taylor Cc: Postgresql Performance Subject: Re: Vacuum takes a really long time, vacuum full required Message-ID: <20041019181245.GJ21258@warped.org> Mail-Followup-To: Rod Taylor , Postgresql Performance References: <20041019153821.GA21258@warped.org> <1098200416.750.238.camel@home> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1098200416.750.238.camel@home> User-Agent: Mutt/1.4.1i X-AntiAbuse: This header was added to track abuse, please include it with any abuse report X-AntiAbuse: Primary Hostname - roue.portalpotty.net X-AntiAbuse: Original Domain - postgresql.org X-AntiAbuse: Originator/Caller UID/GID - [514 32003] / [47 12] X-AntiAbuse: Sender Address Domain - roue.portalpotty.net X-Source: /usr/bin/mutt X-Source-Args: mutt X-Source-Dir: /home/max X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/281 X-Sequence-Number: 8757 Hi Rod, On Tue, Oct 19, 2004 at 11:40:17AM -0400, Rod Taylor wrote: > > Whatever the case, the database still slows down to a halt after a month or > > so, and I have to go in and shut everything down and do a VACUUM FULL by > > hand. One index (of many many) takes 2000 seconds to vacuum. The whole > > process takes a few hours. > > Do a REINDEX on that table instead, and regular vacuum more frequently. Great, this is exactly what I think it needs. Meanwhile, I was checking out http://www.postgresql.org/docs/7.3/static/sql-reindex.html Which suggests I might be able to do a drop/add on each index with the database 'live'. However, the DROP INDEX command was taking an awfully long time to complete and it hung my app in the mean time. Does anyone know if the DROP INDEX causes an exclusive lock, or is it just a lengthy process? > > $ pg_config --version > > PostgreSQL 7.3.2 > > 7.4.x deals with index growth a little better 7.3 and older did. Will do. Meanwhile I'm stuck supporting older 7.x versions, so I'm still looking for a solution for them. Thanks! -m From pgsql-performance-owner@postgresql.org Tue Oct 19 19:43:42 2004 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 379D832A0C2 for ; Tue, 19 Oct 2004 19:43:41 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 00929-06 for ; Tue, 19 Oct 2004 18:43:30 +0000 (GMT) Received: from roue.portalpotty.net (roue.portalpotty.net [69.44.62.206]) by svr1.postgresql.org (Postfix) with ESMTP id AD10F32A015 for ; Tue, 19 Oct 2004 19:43:30 +0100 (BST) Received: from max by roue.portalpotty.net with local (Exim 4.34) id 1CJyxF-0006Mi-Ht for pgsql-performance@postgresql.org; Tue, 19 Oct 2004 11:43:29 -0700 Date: Tue, 19 Oct 2004 14:43:29 -0400 From: Max Baker To: pgsql-performance@postgresql.org Subject: Re: Free PostgreSQL Training, Philadelphia, Oct 30 Message-ID: <20041019184329.GO21258@warped.org> Mail-Followup-To: pgsql-performance@postgresql.org References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.4.1i X-AntiAbuse: This header was added to track abuse, please include it with any abuse report X-AntiAbuse: Primary Hostname - roue.portalpotty.net X-AntiAbuse: Original Domain - postgresql.org X-AntiAbuse: Originator/Caller UID/GID - [514 32003] / [47 12] X-AntiAbuse: Sender Address Domain - roue.portalpotty.net X-Source: /usr/bin/mutt X-Source-Args: mutt X-Source-Dir: /home/max X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/282 X-Sequence-Number: 8758 On Wed, Oct 13, 2004 at 12:21:27PM -0400, Aaron Mulder wrote: > All, > My company (Chariot Solutions) is sponsoring a day of free > PostgreSQL training by Bruce Momjian (one of the core PostgreSQL > developers). The day is split into 2 sessions (plus a Q&A session): > > * Mastering PostgreSQL Administration > * PostgreSQL Performance Tuning > > Registration is required, and space is limited. The location is > Malvern, PA (suburb of Philadelphia) and it's on Saturday Oct 30. For > more information or to register, see > > http://chariotsolutions.com/postgresql.jsp I'm up in New York City and would be taking the train down to Philly. Is anyone coming from Philly or New York that would be able to give me a lift to/from the train station? Sounds like a great event. Cheers, -m From pgsql-performance-owner@postgresql.org Tue Oct 19 20:35:34 2004 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 6758F32ABEE for ; Tue, 19 Oct 2004 20:35:27 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 18226-08 for ; Tue, 19 Oct 2004 19:35:23 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.193]) by svr1.postgresql.org (Postfix) with ESMTP id 565E932A484 for ; Tue, 19 Oct 2004 20:35:24 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 74so369270rnk for ; Tue, 19 Oct 2004 12:35:24 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:mime-version:content-type:content-transfer-encoding; b=SVcP5nOyK7w3LPPrzsyDb7Vt+tvWvW4NtKpTAtEfx5o1hXOVBH1lQUFHt44Z9aATULohkdZbPFjNaXck65pekayN87p7aYrzPWAe6HOoc7kJ4YB7bjzv48ieGhrlnAsRFAD4YMiTD19cw9eUEo1z1vWK24hQuVeah+se84mmLjk Received: by 10.38.24.12 with SMTP id 12mr2139150rnx; Tue, 19 Oct 2004 12:35:24 -0700 (PDT) Received: by 10.38.165.32 with HTTP; Tue, 19 Oct 2004 12:35:24 -0700 (PDT) Message-ID: <27c475ec04101912351c451e8@mail.gmail.com> Date: Tue, 19 Oct 2004 15:35:24 -0400 From: Matt Nuzum Reply-To: matt@followers.net To: pgsql-performance Subject: Speeding up this function 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/283 X-Sequence-Number: 8759 Hello, I've thought it would be nice to index certain aspects of my apache log files for analysis. I've used several different techniques and have something usable now, but I'd like to tweak it one step further. My first performance optimization was to change the logformat into a CSV format. I processed the logfiles with PHP and plsql stored procedures. Unfortunately, it took more than 24 hours to process 1 days worth of log files. I've now switched to using C# (using mono) to create hash-tables to do almost all of the pre-processing. This has brought the time down to about 3 hours. Actually, if I take out one step it brought the process down to about 6 minutes, which is a tremendous improvement. The one step that is adding 2.5+ hours to the job is not easily done in C#, as far as I know. Once the mostly-normalized data has been put into a table called usage_raw_access I then use this query: insert into usage_access select * , usage_normalize_session(accountid,client,atime) as sessionid from usage_raw_access; All it does is try to "link" pageviews together into a session. here's the function: create or replace function usage_normalize_session (varchar(12), inet, timestamptz) returns integer as ' DECLARE -- $1 = Account ID, $2 = IP Address, $3 = Time RecordSet record; BEGIN SELECT INTO RecordSet DISTINCT sessionid FROM usage_access ua WHERE ua.accountid = $1 AND ua.client = $2 AND ua.atime <= ($3 - ''20 min''::interval)::timestamptz; if found then return RecordSet.sessionid; end if; return nextval(''usage_session_ids''); END;' language plpgsql; And the table usage_access looks like this: Table "public.usage_access" Column | Type | Modifiers -------------+--------------------------+----------- [snip] client | inet | atime | timestamp with time zone | accountid | character varying(12) | sessionid | integer | Indexes: usage_acccess_req_url btree (req_url), usage_access_accountid btree (accountid), usage_access_atime btree (atime), usage_access_hostid btree (hostid), usage_access_sessionid btree (sessionid) usage_access_sessionlookup btree (accountid,client,atime); As you can see, this looks for clients who have visited the same site within 20 min. If there is no match, a unique sessionid is assigned from a sequence. If there is a visit, the session id assigned to them is used. I'm only able to process about 25 records per second with my setup. My window to do this job is 3-4 hours and the shorter the better. Here is an explain analyze of the query I do (note I limited it to 1000): EXPLAIN ANALYZE insert into usage_access select * , usage_normalize_session(accountid,client,atime) as sessionid from usage_raw_access limit 1000; QUERY PLAN ---------------------------------------------------------------------------------------------------------------------------------- Subquery Scan "*SELECT*" (cost=0.00..20.00 rows=1000 width=196) (actual time=51.63..47634.22 rows=1000 loops=1) -> Limit (cost=0.00..20.00 rows=1000 width=196) (actual time=51.59..47610.23 rows=1000 loops=1) -> Seq Scan on usage_raw_access (cost=0.00..20.00 rows=1000 width=196) (actual time=51.58..47606.14 rows=1001 loops=1) Total runtime: 48980.54 msec I also did an explain of the query that's performed inside the function: EXPLAIN ANALYZE select sessionid from usage_access ua where ua.accountid = 'XYZ' and ua.client = '64.68.88.45'::inet and ua.atime <= '2003-11-02 04:50:01-05'::timestamptz; QUERY PLAN ---------------------------------------------------------------------------------------------------------------------------------------------------------------------- Index Scan using usage_access_sessionlookup on usage_access ua (cost=0.00..6.02 rows=1 width=4) (actual time=0.29..0.29 rows=0 loops=1) Index Cond: ((accountid = 'XYZ'::character varying) AND (client = '64.68.88.45'::inet) AND (atime <= '2003-11-02 04:50:01-05'::timestamp with time zone)) Total runtime: 0.35 msec (3 rows) What I'd really like to know is if someone knows a way to do any of the following: a: Make the INSERT into ... SELECT *,usage_access_sessionlookup().. work faster b: Make the usage_access_sessionlookup() smarter,better,etc. c: Do this in C# using a hash-table or some other procedure that would be quicker. d: Find an algorithm to create the sessionid without having to do any database or hash-table lookups. As the dataset gets bigger, it won't fit in RAM and the lookup queries will become I/O bound, drastically slowing things down. d: is my first choice. For some reason I just can't seem to get my mind around the data. I wonder if there's someway to create a unique value from client ip address, the accountid and the period of time so that all visits by the IP for the account in that period would match. I thought of using the date_part function to create a unique period, but it would be a hack because if someone visits at 11:50 pm and continues to browse for an hour they would be counted as two sessions. That's not the end of the world, but some of my customers in drastically different time zones would always have skewed results. I tried and tried to get C# to turn the apache date string into a usable time but could not. I just leave the date intact and let postgresql handle it when I do the copy. Therefore, though I'd like to do it in my C# program, I'll likely have to do the sessionid code in a stored procedure. I'd really love some feedback on ways to optimize this. Any suggestions are greatly appreciated. -- Matthew Nuzum | Makers of "Elite Content Management System" www.followers.net | View samples of Elite CMS in action matt@followers.net | http://www.followers.net/portfolio/ From pgsql-performance-owner@postgresql.org Tue Oct 19 20:49:57 2004 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 D80FF32ABEE for ; Tue, 19 Oct 2004 20:49:55 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 24565-03 for ; Tue, 19 Oct 2004 19:49:45 +0000 (GMT) Received: from mail.autorevenue.com (ip-208.178.167.106.gblx.net [208.178.167.106]) by svr1.postgresql.org (Postfix) with ESMTP id 60A2232AA22 for ; Tue, 19 Oct 2004 20:49:43 +0100 (BST) Received: from jeremydunn ([192.168.1.79]) (authenticated) by mail.autorevenue.com (8.11.6/8.11.6) with ESMTP id i9JJnaU12950; Tue, 19 Oct 2004 15:49:36 -0400 Reply-To: From: "Jeremy Dunn" To: , "Postgresql Performance" Subject: Re: Speeding up this function Date: Tue, 19 Oct 2004 15:49:45 -0400 Organization: AutoRevenue Message-ID: <008e01c4b614$c86aa550$4f01a8c0@jeremydunn> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook, Build 10.0.2627 In-Reply-To: <27c475ec04101912351c451e8@mail.gmail.com> X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 Importance: Normal X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/284 X-Sequence-Number: 8760 > -----Original Message----- > From: pgsql-performance-owner@postgresql.org > [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of > Matt Nuzum > Sent: Tuesday, October 19, 2004 3:35 PM > To: pgsql-performance > Subject: [PERFORM] Speeding up this function > > > All it does is try to "link" pageviews together into a session. > here's the function: > create or replace function usage_normalize_session > (varchar(12), inet, timestamptz) returns integer as ' DECLARE > -- $1 = Account ID, $2 = IP Address, $3 = Time > RecordSet record; > BEGIN > SELECT INTO RecordSet DISTINCT sessionid FROM usage_access ua > WHERE ua.accountid = $1 > AND ua.client = $2 > AND ua.atime <= ($3 - ''20 > min''::interval)::timestamptz; > > if found > then return RecordSet.sessionid; > end if; > > return nextval(''usage_session_ids''); > END;' > language plpgsql; > This is probably a stupid question, but why are you trying to create sessions after the fact? Since it appears that users of your site must login, why not just assign a sessionID to them at login time, and keep it in the URL for the duration of the session? Then it would be easy to track where they've been. - Jeremy From pgsql-performance-owner@postgresql.org Tue Oct 19 20:55:13 2004 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 C4EDF32A096 for ; Tue, 19 Oct 2004 20:55:11 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 26815-01 for ; Tue, 19 Oct 2004 19:55:04 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.192]) by svr1.postgresql.org (Postfix) with ESMTP id 71C5932A084 for ; Tue, 19 Oct 2004 20:55:04 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 73so375340rnk for ; Tue, 19 Oct 2004 12:55:04 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; 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:references; b=LTXl/v2hYcx/LR6tVuX4NrgsLuM1QEcpxnR7TeKGP+ylldNzzlkUXOI1PFrGHd0J+BExNbZqWbm2IxADaw+ZY7B2YpXn1hhntu9F+P9N5kIFwf63YBTozYm0cPQVkqpOZEDesbi7PzocR4wcYxE5Y/ZWoSehD2bHS7cezNr8WtE Received: by 10.38.152.19 with SMTP id z19mr2147997rnd; Tue, 19 Oct 2004 12:55:04 -0700 (PDT) Received: by 10.38.165.32 with HTTP; Tue, 19 Oct 2004 12:55:04 -0700 (PDT) Message-ID: <27c475ec04101912556af1d77e@mail.gmail.com> Date: Tue, 19 Oct 2004 15:55:04 -0400 From: Matt Nuzum Reply-To: matt@followers.net To: jdunn@autorevenue.com Subject: Re: Speeding up this function Cc: Postgresql Performance In-Reply-To: <008e01c4b614$c86aa550$4f01a8c0@jeremydunn> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: <27c475ec04101912351c451e8@mail.gmail.com> <008e01c4b614$c86aa550$4f01a8c0@jeremydunn> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/285 X-Sequence-Number: 8761 On Tue, 19 Oct 2004 15:49:45 -0400, Jeremy Dunn wrote: > > -----Original Message----- > > From: pgsql-performance-owner@postgresql.org > > [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of > > Matt Nuzum > > Sent: Tuesday, October 19, 2004 3:35 PM > > To: pgsql-performance > > Subject: [PERFORM] Speeding up this function > > > > > This is probably a stupid question, but why are you trying to create > sessions after the fact? Since it appears that users of your site must > login, why not just assign a sessionID to them at login time, and keep > it in the URL for the duration of the session? Then it would be easy to > track where they've been. > > - Jeremy > > You don't have to log in to visit the sites. These log files are actually for many domains. Right now, we do logging with a web-bug and it does handle the sessions, but it relies on javascript and we want to track a lot more than we are now. Plus, that code is in JavaScript and one of our primary motiviations is to ditch MySQL completely. -- Matthew Nuzum | Makers of "Elite Content Management System" www.followers.net | View samples of Elite CMS in action matt@followers.net | http://www.followers.net/portfolio/ From pgsql-performance-owner@postgresql.org Tue Oct 19 22:52:31 2004 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 65F9F329E89 for ; Tue, 19 Oct 2004 22:52:29 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 13797-01 for ; Tue, 19 Oct 2004 21:52:18 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 44F82329E6E for ; Tue, 19 Oct 2004 22:52:19 +0100 (BST) 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 i9JLqDkc020152; Tue, 19 Oct 2004 17:52:13 -0400 (EDT) To: Manfred Spraul Cc: neilc@samurai.com, markw@osdl.org, pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 In-reply-to: <417221B5.1050704@colorfullife.com> References: <417221B5.1050704@colorfullife.com> Comments: In-reply-to Manfred Spraul message dated "Sun, 17 Oct 2004 09:39:33 +0200" Date: Tue, 19 Oct 2004 17:52:13 -0400 Message-ID: <20151.1098222733@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/294 X-Sequence-Number: 8770 Manfred Spraul writes: > Has anyone tried to replace the whole lwlock implementation with > pthread_rwlock? At least for Linux with recent glibcs, pthread_rwlock is > implemented with futexes, i.e. we would get a fast lock handling without > os specific hacks. "At least for Linux" does not strike me as equivalent to "without OS-specific hacks". The bigger problem here is that the SMP locking bottlenecks we are currently seeing are *hardware* issues (AFAICT anyway). The only way that futexes can offer a performance win is if they have a smarter way of executing the basic atomic-test-and-set sequence than we do; and if so, we could read their code and adopt that method without having to buy into any large reorganization of our code. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Oct 19 22:57:59 2004 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 E084A329E4F for ; Tue, 19 Oct 2004 22:57:57 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 18052-09 for ; Tue, 19 Oct 2004 21:57:47 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id C17AA32A37B for ; Tue, 19 Oct 2004 22:57:45 +0100 (BST) 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 6529810; Tue, 19 Oct 2004 14:59:11 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 Date: Tue, 19 Oct 2004 14:59:48 -0700 User-Agent: KMail/1.6.2 Cc: Tom Lane , Manfred Spraul , neilc@samurai.com, markw@osdl.org References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> In-Reply-To: <20151.1098222733@sss.pgh.pa.us> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable Message-Id: <200410191459.48080.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/295 X-Sequence-Number: 8771 Tom, > The bigger problem here is that the SMP locking bottlenecks we are > currently seeing are *hardware* issues (AFAICT anyway). =C2=A0The only way > that futexes can offer a performance win is if they have a smarter way > of executing the basic atomic-test-and-set sequence than we do; > and if so, we could read their code and adopt that method without having > to buy into any large reorganization of our code. Well, initial results from Gavin/Neil's patch seem to indicate that, while= =20 futexes do not cure the CSStorm bug, they do lessen its effects in terms of= =20 real performance loss. --=20 --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-general-owner@postgresql.org Tue Oct 19 23:09:32 2004 X-Original-To: pgsql-general-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id EE3A1329E77 for ; Tue, 19 Oct 2004 23:09:26 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 30186-05 for ; Tue, 19 Oct 2004 22:09:22 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.194]) by svr1.postgresql.org (Postfix) with ESMTP id 2DA60329E6F for ; Tue, 19 Oct 2004 23:09:24 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 73so393035rnk for ; Tue, 19 Oct 2004 15:09:22 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; 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:references; b=kPiOeanz1JVvEsqk8dIbi4hXLgfuVGjtSHHBc9Fct0Dh6w+sQviztFbsutbBCL+pBpvALABYgHDCArKFqqUJIwaWhkP8LO+0IgyICLPQpse0DSkuOdpVYjrmMFtoBywOBwTaQ9JL2QuXWgepnqNJjwecJC499gfW1b1r1kvroXs Received: by 10.38.152.19 with SMTP id z19mr2207450rnd; Tue, 19 Oct 2004 15:09:21 -0700 (PDT) Received: by 10.38.165.1 with HTTP; Tue, 19 Oct 2004 15:09:21 -0700 (PDT) Message-ID: Date: Tue, 19 Oct 2004 17:09:21 -0500 From: Kevin Barnard Reply-To: Kevin Barnard To: "MikeSmialek2@Hotmail.com" Subject: Re: [PERFORM] Performance on Win32 vs Cygwin Cc: PostgreSQL In-Reply-To: Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests=TO_ADDRESS_EQ_REAL X-Spam-Level: X-Archive-Number: 200410/861 X-Sequence-Number: 67384 Have you looked at the 7.3 configuration file vs. the 8.0. It's possible that the 7.3 file is tweakled better then the 8.0. Have you anaylzed the tables after loading the data into 8.0 On Thu, 14 Oct 2004 12:01:38 -0500, MikeSmialek2@Hotmail.com wrote: > Hi, > > We are experiencing slow performance on 8 Beta 2 Dev3 on Win32 and are > trying to determine why. Any info is appreciated. > > We have a Web Server and a DB server both running Win2KServer with all > service packs and critical updates. > > An ASP page on the Web Server hits the DB Server with a simple query that > returns 205 rows and makes the ASP page delivered to the user about 350K. > > On an ethernet lan a client pc perceives just under 1 sec performance with > the following DB Server configuration: > PIII 550Mhz > 256MB RAM > 7200 RPM HD > cygwin > Postgresql 7.1.3 > PGODBC 7.3.2 > > We set up another DB Server with 8 beta (same Web Server, same network, same > client pc) and now the client pc perceives response of just over 3 sec with > the following DB server config: > PIII 700 Mhz > 448MB RAM > 7200 RPM HD > 8 Beta 2 Dev3 on Win32 running as a service > > Is the speed decrease because it's a beta? > Is the speed decrease because it's running on Win instead of cygwin? > > We did not install cygwin on the new DB Server. > > Thanks, > > Mike > > ---------------------------(end of broadcast)--------------------------- > TIP 4: Don't 'kill -9' the postmaster > From pgsql-performance-owner@postgresql.org Tue Oct 19 23:32:11 2004 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 5257232A087 for ; Tue, 19 Oct 2004 23:32:07 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 40973-05 for ; Tue, 19 Oct 2004 22:31:59 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.195]) by svr1.postgresql.org (Postfix) with ESMTP id 1AA3132A0F4 for ; Tue, 19 Oct 2004 23:32:00 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 73so395398rnk for ; Tue, 19 Oct 2004 15:32:00 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:mime-version:content-type:content-transfer-encoding; b=jaqaVwwr0jWoW0C6unOf4Gk1VEACtIVDrSfW3+UHxuOwM2+COPo51IivOpOSVAyMRCWZwTLLdsrhoPB4OfwgDkW+5NoTfe/4uoZMh5kTSyor9vWGqBOs2asFyVycFtrBAPWARWrwq5nELo966bIqIL8VVFAt1REUathkn44xFfA Received: by 10.38.181.77 with SMTP id d77mr2218334rnf; Tue, 19 Oct 2004 15:32:00 -0700 (PDT) Received: by 10.38.76.42 with HTTP; Tue, 19 Oct 2004 15:31:59 -0700 (PDT) Message-ID: <4a0cafe2041019153133f854e9@mail.gmail.com> Date: Tue, 19 Oct 2004 17:31:59 -0500 From: Josh Close Reply-To: Josh Close To: POSTGRES-PERFORMANCE Subject: how much mem to give postgres? 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/296 X-Sequence-Number: 8772 I'm trying to figure out what I need to do to get my postgres server moving faster. It's just crawling right now. It's on a p4 HT with 2 gigs of mem. I was thinking I need to increase the amount of shared buffers, but I've been told "the sweet spot for shared_buffers is usually on the order of 10000 buffers". I already have it set at 21,078. If you have, say 100 gigs of ram, are you supposed to still only give postgres 10,000? Also, do I need to up the shmmax at all? I've used the formula "250 kB + 8.2 kB * shared_buffers + 14.2 kB * max_connections up to infinity" at http://www.postgresql.org/docs/7.4/interactive/kernel-resources.html#SYSVIPC but it's never quite high enough, so I just make sure it's above the amount that the postgres log says it needs. What else can I do to speed this server up? I'm running vacuum analyze on the heavily updated/inserted/deleted db's once an hour, and doing a full vacuum once a night. Should I change the vacuum mem setting at all? Are there any other settings I should be concerned with? I've heard about the effective_cache_size setting, but I haven't seen anything on what the size should be. Any help would be great. This server is very very slow at the moment. Also, I'm using a 2.6.8.1 kernel with high mem enabled, so all the ram is recognized. Thanks. -Josh From pgsql-performance-owner@postgresql.org Tue Oct 19 23:40:10 2004 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 3718332A179 for ; Tue, 19 Oct 2004 23:39:57 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 51510-05 for ; Tue, 19 Oct 2004 22:39:52 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 121CB32A10D for ; Tue, 19 Oct 2004 23:39:54 +0100 (BST) 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 i9JMdmQ8027698; Tue, 19 Oct 2004 18:39:49 -0400 (EDT) To: josh@agliodbs.com Cc: pgsql-performance@postgresql.org, Manfred Spraul , neilc@samurai.com, markw@osdl.org Subject: Re: futex results with dbt-3 In-reply-to: <200410191459.48080.josh@agliodbs.com> References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> <200410191459.48080.josh@agliodbs.com> Comments: In-reply-to Josh Berkus message dated "Tue, 19 Oct 2004 14:59:48 -0700" Date: Tue, 19 Oct 2004 18:39:48 -0400 Message-ID: <27697.1098225588@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/297 X-Sequence-Number: 8773 Josh Berkus writes: >> The bigger problem here is that the SMP locking bottlenecks we are >> currently seeing are *hardware* issues (AFAICT anyway). > Well, initial results from Gavin/Neil's patch seem to indicate that, while > futexes do not cure the CSStorm bug, they do lessen its effects in terms of > real performance loss. It would be reasonable to expect that futexes would have a somewhat more efficient code path in the case where you have to block (mainly because SysV semaphores have such a heavyweight API, much more complex than we really need). However, the code path that is killing us is the one where you *don't* actually need to block. If we had a proper fix for the problem then the context swap storm itself would go away, and whatever advantage you might be able to measure now for futexes likewise would go away. In other words, I'm not real excited about a wholesale replacement of code in order to speed up a code path that I don't want to be taking in the first place; especially not if that replacement puts a fence between me and working on the code path that I do care about. regards, tom lane From pgsql-performance-owner@postgresql.org Wed Oct 20 00:22:09 2004 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 B3A9C329F05 for ; Wed, 20 Oct 2004 00:21:32 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 61243-07 for ; Tue, 19 Oct 2004 23:21:20 +0000 (GMT) Received: from mail.pws.com.au (mail.pws.com.au [210.23.138.139]) by svr1.postgresql.org (Postfix) with SMTP id B0ADB329D86 for ; Wed, 20 Oct 2004 00:21:20 +0100 (BST) Received: (qmail 22105 invoked by uid 0); 20 Oct 2004 09:21:16 +1000 Received: from cpe-203-45-44-212.vic.bigpond.net.au (HELO wizzard.pws.com.au) (russell@pws.com.au@203.45.44.212) by mail.pws.com.au with SMTP; 20 Oct 2004 09:21:16 +1000 From: Russell Smith To: Gavin Sherry Subject: Re: Select with qualified join condition / Batch inserts Date: Wed, 20 Oct 2004 09:18:57 +1000 User-Agent: KMail/1.7 Cc: Bernd , pgsql-performance@postgresql.org References: <200410151225.26083.bernd_pg@genedata.com> In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200410200918.57761.mr-russ@pws.com.au> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/298 X-Sequence-Number: 8774 On Fri, 15 Oct 2004 08:47 pm, Gavin Sherry wrote: > On Fri, 15 Oct 2004, Bernd wrote: > > > Hi, [snip] > > Table-def: > > Table "public.scr_well_compound" > > Column | Type | Modifiers > > ------------+------------------------+----------- > > mat_id | numeric(10,0) | not null > > barcode | character varying(240) | not null > > well_index | numeric(5,0) | not null > > id_level | numeric(3,0) | not null > > compound | character varying(240) | not null > > Indexes: > > "scr_wcm_pk" PRIMARY KEY, btree (id_level, mat_id, barcode, well_index) > numeric is not optimized by postgresql like it is by Oracle. You will get much better performance by changing the numeric types to int, big int, or small int. That should get the query time down to somewhere near what Oracle is giving you. Regards Russell Smith. [snip] From pgsql-performance-owner@postgresql.org Wed Oct 20 00:28:41 2004 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 2134F32A7F9 for ; Wed, 20 Oct 2004 00:25:34 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 61903-07 for ; Tue, 19 Oct 2004 23:25:29 +0000 (GMT) Received: from biglumber.com (biglumber.com [207.228.252.42]) by svr1.postgresql.org (Postfix) with SMTP id 910C932A091 for ; Wed, 20 Oct 2004 00:25:27 +0100 (BST) Received: (qmail 16108 invoked from network); 19 Oct 2004 23:25:26 -0000 Received: from unknown (HELO localhost) (207.228.252.42) by 0 with SMTP; 19 Oct 2004 23:25:26 -0000 From: "Greg Sabino Mullane" To: pgsql-performance@postgresql.org Subject: Re: Does PostgreSQL run with Oracle? X-PGP-Key: 2529 DF6A B8F7 9407 E944 45B4 BC9B 9067 1496 4AC8 X-Request-PGP: http://www.biglumber.com/x/web?pk=2529DF6AB8F79407E94445B4BC9B906714964AC8 In-Reply-To: Date: Tue, 19 Oct 2004 23:25:26 -0000 X-Mailer: JoyMail 1.48 Message-ID: <1f817f45d6cda937b40fda765f490601@biglumber.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/299 X-Sequence-Number: 8775 -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 > My basic question to the community is "is PostgreSQL approximately > as fast as Oracle?" > > I don't want benchmarks, they're BS. I want a gut feel from this community > because I know many of you are in mixed shops that run both products, or > have had experience with both. My gut feeling is not just "as fast", but "often times faster." I've found very few cases in which Oracle was faster, and that was usually due to some easily spotted difference such as tablespace support. - -- Greg Sabino Mullane greg@turnstep.com PGP Key: 0x14964AC8 200410191925 -----BEGIN PGP SIGNATURE----- iD8DBQFBdaK0vJuQZxSWSsgRApPRAKDTjM+QybR2HnB1UNOao1RY7YDU9ACcDhnr zvH1gwn35Ah8mixo2XHOFr4= =NNZf -----END PGP SIGNATURE----- From pgsql-performance-owner@postgresql.org Wed Oct 20 01:07:44 2004 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 4071732A23E for ; Wed, 20 Oct 2004 01:07:43 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 73580-05 for ; Wed, 20 Oct 2004 00:07:36 +0000 (GMT) Received: from boutiquenumerique.com (boutiquenumerique.com [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 9E27A32A214 for ; Wed, 20 Oct 2004 01:07:37 +0100 (BST) Received: (qmail 24831 invoked from network); 20 Oct 2004 02:07:38 +0200 Received: from unknown (HELO musicbox) (boutiquenumerique-lists@192.168.0.2) by boutiquenumerique.com with SMTP; 20 Oct 2004 02:07:38 +0200 Date: Wed, 20 Oct 2004 02:08:16 +0200 To: pgsql-performance Subject: Re: Speeding up this function References: <27c475ec04101912351c451e8@mail.gmail.com> From: =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= 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: <27c475ec04101912351c451e8@mail.gmail.com> User-Agent: Opera M2/7.54 (Linux, build 751) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/300 X-Sequence-Number: 8776 How many lines do you have in your daily logfiles > As you can see, this looks for clients who have visited the same site > within 20 min. If there is no match, a unique sessionid is assigned > from a sequence. If there is a visit, the session id assigned to them > is used. I'm only able to process about 25 records per second with my > setup. My window to do this job is 3-4 hours and the shorter the > better. I'd say your function is flawed because if a client stays more than 20 minutes he'll get two sessions. I'd propose the following : * solution with postgres (variant #1): - insert everything into big table, - SELECT make_session(...) FROM big table GROUP BY account_id (you may or may not wish to use the ip address, using it will duplicate sessions for people using anonimyzing crowds-style proxies, not using it will merge sessions from the same user from two different ip's). I'd not use it. use an index-powered GroupAggregate maybe. Now it's well ordered, ie. all accesses from the same account are grouped, you just have to find 'gaps' of more than 20 minutes in the atimes to merge or make sessions. This is made by the aggregate 'make_session' which has an internal state consisting of a list of sessions of the form : - session : - session start time - session end time all the aggregate does is look if the atime of the incoming row is < (session end time + 20 min) if <, update session to mark session end time to atime if >, create a new session with session start time = session end time = atime and append it to the session list So you get a table of session arrays, you just have to assign them id's and trackback to the URLs to mark them. If an aggregate can issue INSERT or UPDATE queries, it can even generate session ids on the fly in a table, which simplifies its internal state. * solution with postgres (variant #2): - insert everything into raw_table, - CREATE TABLE sorted_table just like raw_table but with a "id SERIAL PRIMARY KEY" added. - INSERT INTO sorted_table SELECT * FROM raw_table ORDER by account_id, atime; the aggregate was basically comparing the atime's of two adjacent lines to detect a gap of more than 20 minutes, so you could also do a join between rows a and b where b.id = a.id+1 AND ( b.account_id != a.account_id OR (b.atime > a.atime+20 minutes) OR b does not exist ) this will give you the rows which mark a session start, then you have to join again to update all the rows in that session (BETWEEN id's) with the session id. * solution without postgres Take advantage of the fact that the sessions are created and then die to only use RAM for the active sessions. Read the logfile sequentially, you'll need to parse the date, if you can't do it use another language, change your apache date format output, or write a parser. Basically you're doing event-driven programming like in a logic simulator, where the events are session expirations. As you read the file, - keep a hash of sessions indexed on account_id, - and also a sorted (btree) list of sessions indexed on a the session expiry time. It's very important that this list has fast insertion even in the middle, which is why a tree structure would be better. Try a red-black tree. For each record do: - look in the hashtable for account_id, find expiry date for this session, if session still alive you're in that session, update session expiry date and btree index accordingly append url and infos to a list in the session if you want to keep them else expire session and start a new one, insert into hash and btree store the expired session on disk and remove it from memory, you dont need it anymore ! And, as you see the atime advancing, scan the btree for sessions to expire. It's ordered by expiry date, so that's fast. For all expired sessions found, expire session store the expired session on disk and remove it from memory, you dont need it anymore ! That'd be my preferred solution. You'll need a good implementation of a sorted tree, you can find that in opensource. * solution with postgres (variant #3) just like variant #2 but instead of an aggregate use a plpgsql procedure which reads the logs ordered by account_id, atime, while keeping a copy of the last row, and detecting session expirations on the fly. From pgsql-performance-owner@postgresql.org Wed Oct 20 01:33:27 2004 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 D34F1329F9B for ; Wed, 20 Oct 2004 01:33:22 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91479-03 for ; Wed, 20 Oct 2004 00:33:10 +0000 (GMT) Received: from cmailm1.svr.pol.co.uk (cmailm1.svr.pol.co.uk [195.92.193.18]) by svr1.postgresql.org (Postfix) with ESMTP id 29BB7329F0C for ; Wed, 20 Oct 2004 01:33:12 +0100 (BST) Received: from modem-1859.llama.dialup.pol.co.uk ([217.135.183.67] helo=Nightingale) by cmailm1.svr.pol.co.uk with smtp (Exim 4.14) id 1CK4Pg-0008QN-Hz; Wed, 20 Oct 2004 01:33:12 +0100 Message-ID: <01f301c4b63c$643340b0$6400a8c0@Nightingale> From: "Simon Riggs" To: "Josh Close" , "POSTGRES-PERFORMANCE" References: <4a0cafe2041019153133f854e9@mail.gmail.com> Subject: Re: how much mem to give postgres? Date: Wed, 20 Oct 2004 01:33:16 +0100 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 6.00.2800.1409 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1409 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/301 X-Sequence-Number: 8777 >Josh Close > I'm trying to figure out what I need to do to get my postgres server > moving faster. It's just crawling right now. It's on a p4 HT with 2 > gigs of mem. ....and using what version of PostgreSQL are you using? 8.0beta, I hope? > I was thinking I need to increase the amount of shared buffers, but > I've been told "the sweet spot for shared_buffers is usually on the > order of 10000 buffers". I already have it set at 21,078. If you have, > say 100 gigs of ram, are you supposed to still only give postgres > 10,000? Thats under test currently. My answer would be, "clearly not", others differ, for varying reasons. > Also, do I need to up the shmmax at all? I've used the formula "250 kB > + 8.2 kB * shared_buffers + 14.2 kB * max_connections up to infinity" > at http://www.postgresql.org/docs/7.4/interactive/kernel-resources.html#SYSVIPC > but it's never quite high enough, so I just make sure it's above the > amount that the postgres log says it needs. shmmax isn't a tuning parameter for PostgreSQL, its just a limit. If you get no error messages, then its high enough. > Are there any other settings I should be concerned with? I've heard > about the effective_cache_size setting, but I haven't seen anything on > what the size should be. wal_buffers if the databases are heavily updated. > Any help would be great. This server is very very slow at the moment. > Try *very fast disks*, especially for the logs. Best regards, Simon Riggs From pgsql-performance-owner@postgresql.org Wed Oct 20 02:53:40 2004 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 4922C32A0C2 for ; Wed, 20 Oct 2004 02:53:40 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 29089-01 for ; Wed, 20 Oct 2004 01:53:35 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.202]) by svr1.postgresql.org (Postfix) with ESMTP id B540D329F75 for ; Wed, 20 Oct 2004 02:53:37 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 73so414090rnk for ; Tue, 19 Oct 2004 18:53:38 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:mime-version:content-type:content-transfer-encoding; b=T3K9MaxSDJgKoe2EwE8ohedc9NHCV9PzhjXVWD/ofRJY599qu8qPG3zqGIIyf68WcSE8HO3BDb5uUzutW/X5muz4pwQ9vzwCVsQzaJ5cT8BuLEpks0x9TdO0OAumJyBkXqqmHDlAcQtnPydKsB3zYtRKDvOhBDZwiEQTnFj38Uw Received: by 10.38.72.80 with SMTP id u80mr2280650rna; Tue, 19 Oct 2004 18:53:37 -0700 (PDT) Received: by 10.38.165.15 with HTTP; Tue, 19 Oct 2004 18:53:37 -0700 (PDT) Message-ID: <97b3fe2041019185337a8c3d8@mail.gmail.com> Date: Wed, 20 Oct 2004 11:53:37 +1000 From: Brock Henry Reply-To: Brock Henry To: pgsql-performance@postgresql.org Subject: Insert performance, what should I expect? Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/302 X-Sequence-Number: 8778 Hi,=20 I've after some opinions about insert performance. I'm importing a file with 13,002 lines to a database that ends up with 75,703 records across 6 tables. This is a partial file =E2=80=93 the real d= ata is 4 files with total lines 95174. I'll be loading these files each morning, and then running a number of queries on them. The select queries run fast enough, (mostly - 2 queries are slow but I'll look into that later), but importing is slower than I'd like it to be, but I'm wondering what to expect? I've done some manual benchmarking running my script 'time script.pl' I realise my script uses some of the time, bench marking shows that %50 of the time is spent in dbd:execute. Test 1, For each import, I'm dropping all indexes and pkeys/fkeys, then importing, then adding keys and indexes. Then I've got successive runs. I figure the reindexing will get more expensive as the database grows? Successive Imports: 44,49,50,57,55,61,72 (seconds) =3D average 1051inserts/second (which now that I've written this seems fairly good) Test 2, no dropping etc of indexes, just INSERTs Import =E2=80=93 61, 62, 73, 68, 78, 74 (seconds) =3D average 1091 inserts/second Machine is Linux 2.6.4, 1GB RAM, 3.something GHz XEON processor, SCSI hdd's (raid1). PostgreSQL 7.4.2. Lightly loaded machine, not doing much other than my script. Script and DB on same machine. Sysctl =E2=80=93a | grep shm kernel.shmmni =3D 4096 kernel.shmall =3D 134217728 (pages or bytes? Anyway=E2=80=A6) kernel.shmmax =3D 134217728 postgresql.conf tcpip_socket =3D true max_connections =3D 32 superuser_reserved_connections =3D 2 shared_buffers =3D 8192=20=20=20=20=20=20=20 sort_mem =3D 4096=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20 vacuum_mem =3D 16384=20=20=20=20=20=20 max_fsm_relations =3D 300=20=20 fsync =3D true=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20 wal_buffers =3D 64=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20 checkpoint_segments =3D 10=20=20=20=20=20=20=20=20=20 effective_cache_size =3D 16000=20=20=20=20 syslog =3D 1=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20 silent_mode =3D false=20=20=20=20=20=20=20=20=20=20=20=20=20=20 log_connections =3D true log_pid =3D true log_timestamp =3D true stats_start_collector =3D true stats_row_level =3D true Can I expect it to go faster than this? I'll see where I can make my script itself go faster, but I don't think I'll be able to do much. I'll do some pre-prepare type stuff, but I don't expect significant gains, maybe 5-10%. I'd could happily turn off fsync for this job, but not for some other databases the server is hosting. Any comments/suggestions would be appreciated. Thanks :) Brock Henry From pgsql-performance-owner@postgresql.org Wed Oct 20 03:12:30 2004 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 4EC8132AEDA for ; Wed, 20 Oct 2004 03:12:29 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 62207-08 for ; Wed, 20 Oct 2004 02:12:22 +0000 (GMT) Received: from outbound.mailhop.org (outbound.mailhop.org [63.208.196.171]) by svr1.postgresql.org (Postfix) with ESMTP id 41859329E72 for ; Wed, 20 Oct 2004 03:12:21 +0100 (BST) Received: from ool-4350c7ad.dyn.optonline.net ([67.80.199.173] helo=[192.168.0.66]) by outbound.mailhop.org with esmtpsa (TLSv1:AES256-SHA:256) (Exim 4.42) id 1CK5xd-0000lz-EP; Tue, 19 Oct 2004 22:12:21 -0400 Message-ID: <4175C97A.2080300@zeut.net> Date: Tue, 19 Oct 2004 22:12:10 -0400 From: "Matthew T. O'Connor" Organization: Terrie O'Connor Realtors User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Brock Henry Cc: pgsql-performance@postgresql.org Subject: Re: Insert performance, what should I expect? References: <97b3fe2041019185337a8c3d8@mail.gmail.com> In-Reply-To: <97b3fe2041019185337a8c3d8@mail.gmail.com> Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit X-Mail-Handler: MailHop Outbound by DynDNS.org X-Originating-IP: 67.80.199.173 X-Report-Abuse-To: abuse@dyndns.org (see http://www.mailhop.org/outbound/abuse.html for abuse reporting information) X-MHO-User: zeut X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/303 X-Sequence-Number: 8779 Brock Henry wrote: >Hi, > >I've after some opinions about insert performance. > Have you looked into using the copy command instead of inserts? For bulk loading of data it can be significantly faster. From pgsql-performance-owner@postgresql.org Wed Oct 20 03:17:28 2004 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 2036C32AEA9 for ; Wed, 20 Oct 2004 03:17:27 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 93420-09 for ; Wed, 20 Oct 2004 02:17:22 +0000 (GMT) Received: from tht.net (vista.tht.net [216.126.88.2]) by svr1.postgresql.org (Postfix) with ESMTP id E2EAE32AEAC for ; Wed, 20 Oct 2004 03:17:25 +0100 (BST) Received: from [134.22.69.212] (dyn-69-212.tor.dsl.tht.net [134.22.69.212]) by tht.net (Postfix) with ESMTP id 326DD76B52; Tue, 19 Oct 2004 22:13:27 -0400 (EDT) Subject: Re: Insert performance, what should I expect? From: Rod Taylor To: Brock Henry Cc: Postgresql Performance In-Reply-To: <97b3fe2041019185337a8c3d8@mail.gmail.com> References: <97b3fe2041019185337a8c3d8@mail.gmail.com> Content-Type: text/plain Message-Id: <1098238348.747.20.camel@home> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Tue, 19 Oct 2004 22:12:28 -0400 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/304 X-Sequence-Number: 8780 > I've done some manual benchmarking running my script 'time script.pl' > I realise my script uses some of the time, bench marking shows that > %50 of the time is spent in dbd:execute. The perl drivers don't currently use database level prepared statements which would give a small boost. But your best bet is to switch to using COPY instead of INSERT. Two ways to do this. 1) Drop DBD::Pg and switch to the Pg driver for Perl instead (non-DBI compliant) which has functions similar to putline() that allow COPY to be used. 2) Have your perl script output a .sql file with the data prepared (COPY statements) which you feed into the database via psql. You can probably achieve a 50% increase in throughput. From pgsql-performance-owner@postgresql.org Wed Oct 20 05:02:43 2004 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 CAC0A32AF4A for ; Wed, 20 Oct 2004 05:02:39 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 73174-08 for ; Wed, 20 Oct 2004 04:02:34 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.194]) by svr1.postgresql.org (Postfix) with ESMTP id 42E6332AF0C for ; Wed, 20 Oct 2004 05:02:31 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 73so425702rnk for ; Tue, 19 Oct 2004 21:02:30 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; 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:references; b=cBu7ohHjSXpGh5QXF7aMIn9pFQEOXe1q5wR7VskXABmRTrgV80luQs2r3svAcDxSTZqdgKn4VbeHcx/7h2MV1pMiu7rww2i1MviEtJLceXE3iJkHwkEiBDof3h6ghiv1Ve9ybl7zNMOOiZPoRbOxe8KrPQFXP8bZkC3t1eTgP68 Received: by 10.38.181.77 with SMTP id d77mr2339362rnf; Tue, 19 Oct 2004 21:02:30 -0700 (PDT) Received: by 10.38.76.42 with HTTP; Tue, 19 Oct 2004 21:02:30 -0700 (PDT) Message-ID: <4a0cafe2041019210265656d4e@mail.gmail.com> Date: Tue, 19 Oct 2004 23:02:30 -0500 From: Josh Close Reply-To: Josh Close To: POSTGRES-PERFORMANCE Subject: Re: how much mem to give postgres? In-Reply-To: <01f301c4b63c$643340b0$6400a8c0@Nightingale> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: <4a0cafe2041019153133f854e9@mail.gmail.com> <01f301c4b63c$643340b0$6400a8c0@Nightingale> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/305 X-Sequence-Number: 8781 On Wed, 20 Oct 2004 01:33:16 +0100, Simon Riggs wrote: > ....and using what version of PostgreSQL are you using? 8.0beta, I hope? I'm using version 7.4.5. > > I was thinking I need to increase the amount of shared buffers, but > > I've been told "the sweet spot for shared_buffers is usually on the > > order of 10000 buffers". I already have it set at 21,078. If you have, > > say 100 gigs of ram, are you supposed to still only give postgres > > 10,000? > > Thats under test currently. My answer would be, "clearly not", others > differ, for varying reasons. Should I stick the rule of around 15% of mem then? I haven't found any information on why you should use certain settings at all. I read somewhere on the postgres site about using as much memory as possible, but leave a little room for other processes. Whould that be an ok theory? I'd kinda like to know why I should or shouldn't do something like this. -Josh From pgsql-performance-owner@postgresql.org Wed Oct 20 05:35:38 2004 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 EA1F432AEF3 for ; Wed, 20 Oct 2004 05:35:34 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 81047-03 for ; Wed, 20 Oct 2004 04:35:33 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id CB93632AEFF for ; Wed, 20 Oct 2004 05:35:32 +0100 (BST) 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 i9K4ZVb6029977; Wed, 20 Oct 2004 00:35:31 -0400 (EDT) To: Josh Close Cc: POSTGRES-PERFORMANCE Subject: Re: how much mem to give postgres? In-reply-to: <4a0cafe2041019153133f854e9@mail.gmail.com> References: <4a0cafe2041019153133f854e9@mail.gmail.com> Comments: In-reply-to Josh Close message dated "Tue, 19 Oct 2004 17:31:59 -0500" Date: Wed, 20 Oct 2004 00:35:31 -0400 Message-ID: <29976.1098246931@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/306 X-Sequence-Number: 8782 Josh Close writes: > I'm trying to figure out what I need to do to get my postgres server > moving faster. It's just crawling right now. I suspect that fooling with shared_buffers is entirely the wrong tree for you to be barking up. My suggestion is to be looking at individual queries that are slow, and seeing how to speed those up. This might involve adding indexes, or tweaking the query source, or adjusting planner parameters, or several other things. EXPLAIN ANALYZE is your friend ... regards, tom lane From pgsql-performance-owner@postgresql.org Wed Oct 20 06:23:51 2004 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 724A632A8DD for ; Wed, 20 Oct 2004 06:23:49 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91232-05 for ; Wed, 20 Oct 2004 05:23:39 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 240DF329F9F for ; Wed, 20 Oct 2004 06:23:39 +0100 (BST) 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 6531652; Tue, 19 Oct 2004 22:25:03 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org, Josh Close Subject: Re: how much mem to give postgres? Date: Tue, 19 Oct 2004 22:23:24 -0700 User-Agent: KMail/1.6.2 References: <4a0cafe2041019153133f854e9@mail.gmail.com> In-Reply-To: <4a0cafe2041019153133f854e9@mail.gmail.com> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410192223.24285.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/307 X-Sequence-Number: 8783 JJosh, > I'm trying to figure out what I need to do to get my postgres server > moving faster. It's just crawling right now. It's on a p4 HT with 2 > gigs of mem. There have been issues with Postgres+HT, especially on Linux 2.4. Try turning HT off if other tuning doesn't solve things. Otherwise, see: http://www.varlena.com/varlena/GeneralBits/Tidbits/perf.html -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Wed Oct 20 08:51:09 2004 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 A766632A8ED for ; Wed, 20 Oct 2004 08:51:08 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 31150-03 for ; Wed, 20 Oct 2004 07:51:01 +0000 (GMT) Received: from mail1.catalyst.net.nz (godel.catalyst.net.nz [202.49.159.12]) by svr1.postgresql.org (Postfix) with ESMTP id 4C5A232A892 for ; Wed, 20 Oct 2004 08:51:00 +0100 (BST) Received: from 222-152-142-149.jetstream.xtra.co.nz ([222.152.142.149] helo=lamb.mcmillan.net.nz) by mail1.catalyst.net.nz with asmtp (TLS-1.0:DHE_RSA_3DES_EDE_CBC_SHA:24) (Exim 4.34) id 1CKBFH-00032E-Rv; Wed, 20 Oct 2004 20:50:56 +1300 Received: from lamb.mcmillan.net.nz (lamb.mcmillan.net.nz [127.0.0.1]) by lamb.mcmillan.net.nz (Postfix) with ESMTP id 59E66AD97BBE; Wed, 20 Oct 2004 20:50:45 +1300 (NZDT) Subject: Re: Insert performance, what should I expect? From: Andrew McMillan To: Brock Henry Cc: pgsql-performance@postgresql.org In-Reply-To: <97b3fe2041019185337a8c3d8@mail.gmail.com> References: <97b3fe2041019185337a8c3d8@mail.gmail.com> Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="=-fDpxLEsYDsVSux/w1NKM" Date: Wed, 20 Oct 2004 20:50:44 +1300 Message-Id: <1098258644.22373.161.camel@lamb.mcmillan.net.nz> Mime-Version: 1.0 X-Mailer: Evolution 2.0.1 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/308 X-Sequence-Number: 8784 --=-fDpxLEsYDsVSux/w1NKM Content-Type: text/plain Content-Transfer-Encoding: quoted-printable On Wed, 2004-10-20 at 11:53 +1000, Brock Henry wrote: >=20 > Test 1, For each import, I'm dropping all indexes and pkeys/fkeys, > then importing, then adding keys and indexes. Then I've got successive > runs. I figure the reindexing will get more expensive as the database > grows? Sounds like the right approach to me, if the tables are empty before the import. > Successive Imports: 44,49,50,57,55,61,72 (seconds) > =3D average 1051inserts/second (which now that I've written this seems > fairly good) (A) Are you doing the whole thing inside a transaction? This will be significantly quicker. COPY would probably be quicker still, but the biggest difference will be a single transaction. (B) If you are starting with empty files, are you ensuring that the dead records are vacuumed before you start? I would recommend a "vacuum full" on the affected tables prior to the first import run (i.e. when the tables are empty). This is likely to be the reason that the timing on your successive imports increases so much. > sort_mem =3D 4096=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20 You probably want to increase this - if you have 1G of RAM then there is probably some spare. But if you actually expect to use 32 connections then 32 * 4M =3D 128M might mean a careful calculation is needed. If you are really only likely to have 1-2 connections running concurrently then increase it to (e.g.) 32768. > max_fsm_relations =3D 300=20=20 If you do a "vacuum full verbose;" the last line will give you some clues as to what to set this (and max_fsm_pages) too. > effective_cache_size =3D 16000=20=20=20=20 16000 * 8k =3D 128M seems low for a 1G machine - probably you could say 64000 without fear of being wrong. What does "free" show as "cached"? Depending on how dedicated the machine is to the database, the effective cache size may be as much as 80-90% of that. > Can I expect it to go faster than this? I'll see where I can make my > script itself go faster, but I don't think I'll be able to do much. > I'll do some pre-prepare type stuff, but I don't expect significant > gains, maybe 5-10%. I'd could happily turn off fsync for this job, but > not for some other databases the server is hosting. You can probably double the speed - maybe more. Cheers, Andrew, ------------------------------------------------------------------------- Andrew @ Catalyst .Net .NZ Ltd, PO Box 11-053, Manners St, Wellington WEB: http://catalyst.net.nz/ PHYS: Level 2, 150-154 Willis St DDI: +64(4)803-2201 MOB: +64(272)DEBIAN OFFICE: +64(4)499-2267 How many things I can do without! -- Socrates ------------------------------------------------------------------------- --=-fDpxLEsYDsVSux/w1NKM Content-Type: application/pgp-signature; name=signature.asc Content-Description: This is a digitally signed message part -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) iD8DBQBBdhjUjJA0f48GgBIRAqjyAKC8ABSJoxFHklEVFE5KethNR5j5ZQCdHmMP mSFjmB/Yjs3HNcSYWoJdeeM= =mKj3 -----END PGP SIGNATURE----- --=-fDpxLEsYDsVSux/w1NKM-- From pgsql-performance-owner@postgresql.org Wed Oct 20 12:12:32 2004 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 64B0732A75E for ; Wed, 20 Oct 2004 12:12:31 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 74330-10 for ; Wed, 20 Oct 2004 11:12:23 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.206]) by svr1.postgresql.org (Postfix) with ESMTP id 242FF32A9B5 for ; Wed, 20 Oct 2004 12:12:25 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 73so455941rnk for ; Wed, 20 Oct 2004 04:12:25 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; 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:references; b=pxTN5mK5QQHkviWwOfU0fA5j7l5TKkxnVJCF+gb0JGc88siR2Jz8q0ef1y1eZx19i3SX1KYdQE4VtW6wSwgcFZQo1Fofm8IrLshTZbVzxGU5/EFdA4NyOUn0zvzFptFXq6JtXdXPpJHNRdyPZj1I4pETU6l+DEeWzU4nqe6UBGI Received: by 10.38.152.46 with SMTP id z46mr524849rnd; Wed, 20 Oct 2004 04:10:35 -0700 (PDT) Received: by 10.38.151.75 with HTTP; Wed, 20 Oct 2004 04:10:35 -0700 (PDT) Message-ID: <157f648404102004103a126b27@mail.gmail.com> Date: Wed, 20 Oct 2004 07:10:35 -0400 From: Aaron Werman Reply-To: Aaron Werman To: pgsql-performance@postgresql.org Subject: Re: Free PostgreSQL Training, Philadelphia, Oct 30 In-Reply-To: <20041019184329.GO21258@warped.org> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: <20041019184329.GO21258@warped.org> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/309 X-Sequence-Number: 8785 I'm driving from Tenafly NJ and going to both sessions. If you're able to get to the George Washington Bridge (A train to 178th Street [Port Authority North] and a bus over the bridge), I can drive you down. I'm not sure right now about the return because I have confused plans to meet someone. /Aaron On Tue, 19 Oct 2004 14:43:29 -0400, Max Baker wrote: > On Wed, Oct 13, 2004 at 12:21:27PM -0400, Aaron Mulder wrote: > > All, > > My company (Chariot Solutions) is sponsoring a day of free > > PostgreSQL training by Bruce Momjian (one of the core PostgreSQL > > developers). The day is split into 2 sessions (plus a Q&A session): > > > > * Mastering PostgreSQL Administration > > * PostgreSQL Performance Tuning > > > > Registration is required, and space is limited. The location is > > Malvern, PA (suburb of Philadelphia) and it's on Saturday Oct 30. For > > more information or to register, see > > > > http://chariotsolutions.com/postgresql.jsp > > I'm up in New York City and would be taking the train down to Philly. Is > anyone coming from Philly or New York that would be able to give me a lift > to/from the train station? Sounds like a great event. > > Cheers, > -m > > ---------------------------(end of broadcast)--------------------------- > TIP 3: 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 > -- Regards, /Aaron From pgsql-performance-owner@postgresql.org Wed Oct 20 12:31:30 2004 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 384C432A0AE for ; Wed, 20 Oct 2004 12:31:29 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 80084-06 for ; Wed, 20 Oct 2004 11:31:20 +0000 (GMT) Received: from heimdall.hig.se (heimdall.hig.se [130.243.8.180]) by svr1.postgresql.org (Postfix) with SMTP id 6FBAB32A093 for ; Wed, 20 Oct 2004 12:31:16 +0100 (BST) Received: from webmail.student.hig.se ([130.243.8.161]) by heimdall.hig.se (SAVSMTP 3.1.0.29) with SMTP id M2004102013270611230 for ; Wed, 20 Oct 2004 13:27:06 +0200 Received: from 130.243.12.121 (SquirrelMail authenticated user nd02tsk); by webmail.student.hig.se with HTTP; Wed, 20 Oct 2004 13:50:38 +0200 (CEST) Message-ID: <3879.130.243.12.121.1098273038.squirrel@130.243.12.121> Date: Wed, 20 Oct 2004 13:50:38 +0200 (CEST) Subject: Re: How to time several queries? From: nd02tsk@student.hig.se To: pgsql-performance@postgresql.org User-Agent: SquirrelMail/1.4.3 X-Mailer: SquirrelMail/1.4.3 MIME-Version: 1.0 Content-Type: text/plain;charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Priority: 3 (Normal) Importance: Normal X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.5 tagged_above=0.0 required=5.0 tests=NO_REAL_NAME, UPPERCASE_25_50 X-Spam-Level: X-Archive-Number: 200410/310 X-Sequence-Number: 8786 It doesn't seem to work. I want a time summary at the end. I am inserting insert queries from a file with the \i option. This is the outcome: [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.672 ms [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.730 ms [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.698 ms [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.805 ms [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.670 ms [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.831 ms [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.815 ms [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.793 ms [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.660 ms [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.667 ms [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.754 ms [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.668 ms [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.688 ms [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.671 ms [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.787 ms [7259] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27'); [7259] LOG: duration: 1.722 ms [7309] LOG: statement: DELETE FROM weather; [7309] LOG: duration: 11.314 ms [7330] LOG: statement: INSERT INTO weather VALUES ('San Francisco', 46, 50, 0.25, '1994-11-27') Tim From pgsql-performance-owner@postgresql.org Wed Oct 20 16:08:57 2004 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 21790329F1F for ; Wed, 20 Oct 2004 13:51:06 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 02911-05 for ; Wed, 20 Oct 2004 12:50:57 +0000 (GMT) Received: from ms-smtp-02.tampabay.rr.com (ms-smtp-02-smtplb.tampabay.rr.com [65.32.5.132]) by svr1.postgresql.org (Postfix) with ESMTP id 86638329E72 for ; Wed, 20 Oct 2004 13:50:57 +0100 (BST) Received: from MATTSPC (222-60.26-24.tampabay.rr.com [24.26.60.222]) by ms-smtp-02.tampabay.rr.com (8.12.10/8.12.7) with ESMTP id i9KConZn016806; Wed, 20 Oct 2004 08:50:50 -0400 (EDT) Message-Id: <200410201250.i9KConZn016806@ms-smtp-02.tampabay.rr.com> From: "Matthew Nuzum" To: , Subject: Re: How to time several queries? Date: Wed, 20 Oct 2004 08:50:42 -0400 MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit X-Mailer: Microsoft Office Outlook, Build 11.0.6353 X-MIMEOLE: Produced By Microsoft MimeOLE V6.00.2900.2180 In-reply-to: <2612.130.243.12.107.1098124104.squirrel@130.243.12.107> Thread-index: AcS1QaijScudGnrvSma4ma/DI8PrLgBYV7gg X-Virus-Scanned: Symantec AntiVirus Scan Engine X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/319 X-Sequence-Number: 8795 When I'm using psql and I want to time queries, which is what I've been doing for a little over a day now, I do the following: Select now(); query 1; query 2; query 3; select now(); This works fine unless you're doing selects with a lot of rows which will cause your first timestamp to scroll off the screen. -- Matthew Nuzum + "Man was born free, and everywhere www.bearfruit.org : he is in chains," Rousseau +~~~~~~~~~~~~~~~~~~+ "Then you will know the truth, and the TRUTH will set you free," Jesus Christ (John 8:32 NIV) -----Original Message----- From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of nd02tsk@student.hig.se Sent: Monday, October 18, 2004 2:28 PM To: pgsql-performance@postgresql.org Subject: [PERFORM] How to time several queries? Hello I posted this on the general list but think it would be more appropriate here. Sorry. I know it is possible to time isolated queries through the settting of the \timing option in psql. This makes PgSQL report the time it took to perform one operation. I would like to know how one can get a time summary of many operations, if it is at all possible. Thank you. Tim ---------------------------(end of broadcast)--------------------------- TIP 4: Don't 'kill -9' the postmaster From pgsql-performance-owner@postgresql.org Wed Oct 20 13:34:55 2004 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 4B64432A12E for ; Wed, 20 Oct 2004 13:34:54 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 00190-03 for ; Wed, 20 Oct 2004 12:34:45 +0000 (GMT) Received: from heimdall.hig.se (heimdall.hig.se [130.243.8.180]) by svr1.postgresql.org (Postfix) with SMTP id ABC4932A11E for ; Wed, 20 Oct 2004 13:34:46 +0100 (BST) Received: from webmail.student.hig.se ([130.243.8.161]) by heimdall.hig.se (SAVSMTP 3.1.0.29) with SMTP id M2004102014304115757 for ; Wed, 20 Oct 2004 14:30:41 +0200 Received: from 130.243.12.121 (SquirrelMail authenticated user nd02tsk); by webmail.student.hig.se with HTTP; Wed, 20 Oct 2004 14:54:12 +0200 (CEST) Message-ID: <3992.130.243.12.121.1098276852.squirrel@130.243.12.121> In-Reply-To: References: <3388.130.243.14.147.1097595901.squirrel@130.243.14.147> Date: Wed, 20 Oct 2004 14:54:12 +0200 (CEST) Subject: Re: Which plattform do you recommend I run PostgreSQL From: nd02tsk@student.hig.se To: pgsql-performance@postgresql.org User-Agent: SquirrelMail/1.4.3 X-Mailer: SquirrelMail/1.4.3 MIME-Version: 1.0 Content-Type: text/plain;charset=iso-8859-1 Content-Transfer-Encoding: 8bit X-Priority: 3 (Normal) Importance: Normal X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.3 tagged_above=0.0 required=5.0 tests=NO_REAL_NAME X-Spam-Level: X-Archive-Number: 200410/311 X-Sequence-Number: 8787 Thank you. Tim > hi, > > nd02tsk@student.hig.se wrote: >> Hello >> >> I am doing a comparison between MySQL and PostgreSQL. >> >> In the MySQL manual it says that MySQL performs best with Linux 2.4 with >> ReiserFS on x86. Can anyone official, or in the know, give similar >> information regarding PostgreSQL? >> >> Also, any links to benchmarking tests available on the internet between >> MySQL and PostgreSQL would be appreciated. > > http://www.potentialtech.com/wmoran/postgresql.php > http://www.varlena.com/varlena/GeneralBits/Tidbits/perf.html > http://candle.pha.pa.us/main/writings/pgsql/hw_performance/ > http://database.sarang.net/database/postgres/optimizing_postgresql.html > > C. > > ---------------------------(end of broadcast)--------------------------- > TIP 3: 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 Oct 20 14:06:26 2004 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 E65CD32A15F for ; Wed, 20 Oct 2004 14:06:25 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 08428-10 for ; Wed, 20 Oct 2004 13:06:15 +0000 (GMT) Received: from mail.ebuz.de (web01.ebuz.de [62.146.49.2]) by svr1.postgresql.org (Postfix) with ESMTP id 47F3532A124 for ; Wed, 20 Oct 2004 14:06:17 +0100 (BST) Received: from [217.232.163.185] (helo=nixe) by mail.ebuz.de with asmtp (Exim 3.33 #1) id 1CKGAQ-000Epr-00 for pgsql-performance@postgresql.org; Wed, 20 Oct 2004 15:06:14 +0200 Date: Wed, 20 Oct 2004 15:10:15 +0200 From: Tom Fischer To: pgsql-performance@postgresql.org Subject: OS desicion Message-ID: <20041020151015.766bb00c@nixe> Organization: eBuz Internetdienste GbR X-Mailer: Sylpheed-Claws 0.9.12b (GTK+ 1.2.10; i686-pc-linux-gnu) Mime-Version: 1.0 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/313 X-Sequence-Number: 8789 Hi List, I have a Dual-Xeon 3Ghz System with with GB RAM and an Adaptec 212=DF SCSI RAID with 4 SCA Harddiscs. Our customer wants to have the Machine tuned for best Database performance. Which OS should we used? We are tending between Linux 2.6 or FreeBSD. The Database Size is 5GB and ascending. Most SQL-Queries are Selects, the Tablesizes are beetween 300k and up to 10 MB. I've read the Hardware Performance Guide and the result was to take FreeBSD in the Decision too :) And what is on this Context Switiching Bug i have read in the Archive?=20 Hope you can help me Regards Tom From pgsql-performance-owner@postgresql.org Wed Oct 20 13:57:24 2004 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 3554A32A11C for ; Wed, 20 Oct 2004 13:57:23 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 06547-07 for ; Wed, 20 Oct 2004 12:57:13 +0000 (GMT) Received: from saturn.opentools.org (saturn.opentools.org [66.250.40.202]) by svr1.postgresql.org (Postfix) with ESMTP id 122D432A16E for ; Wed, 20 Oct 2004 13:57:16 +0100 (BST) Received: by saturn.opentools.org (Postfix, from userid 500) id D22EC3E99; Wed, 20 Oct 2004 09:11:25 -0400 (EDT) Received: from localhost (localhost [127.0.0.1]) by saturn.opentools.org (Postfix) with ESMTP id CBB37F5A0 for ; Wed, 20 Oct 2004 09:11:25 -0400 (EDT) Date: Wed, 20 Oct 2004 09:11:25 -0400 (EDT) From: Aaron Mulder X-X-Sender: ammulder@saturn.opentools.org To: pgsql-performance@postgresql.org Subject: Re: Free PostgreSQL Training, Philadelphia, Oct 30 In-Reply-To: <157f648404102004103a126b27@mail.gmail.com> Message-ID: References: <20041019184329.GO21258@warped.org> <157f648404102004103a126b27@mail.gmail.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/312 X-Sequence-Number: 8788 If anyone is going to take the train all the way, please e-mail me offline. There is a train station relatively close to the event (NY to Philly then the R5 to Malvern), but it's not within walking distance, so we'll figure out some way to pick people up from there. Thanks, Aaron On Wed, 20 Oct 2004, Aaron Werman wrote: > I'm driving from Tenafly NJ and going to both sessions. If you're able > to get to the George Washington Bridge (A train to 178th Street [Port > Authority North] and a bus over the bridge), I can drive you down. I'm > not sure right now about the return because I have confused plans to > meet someone. > > /Aaron > > > On Tue, 19 Oct 2004 14:43:29 -0400, Max Baker wrote: > > On Wed, Oct 13, 2004 at 12:21:27PM -0400, Aaron Mulder wrote: > > > All, > > > My company (Chariot Solutions) is sponsoring a day of free > > > PostgreSQL training by Bruce Momjian (one of the core PostgreSQL > > > developers). The day is split into 2 sessions (plus a Q&A session): > > > > > > * Mastering PostgreSQL Administration > > > * PostgreSQL Performance Tuning > > > > > > Registration is required, and space is limited. The location is > > > Malvern, PA (suburb of Philadelphia) and it's on Saturday Oct 30. For > > > more information or to register, see > > > > > > http://chariotsolutions.com/postgresql.jsp > > > > I'm up in New York City and would be taking the train down to Philly. Is > > anyone coming from Philly or New York that would be able to give me a lift > > to/from the train station? Sounds like a great event. > > > > Cheers, > > -m > > > > ---------------------------(end of broadcast)--------------------------- > > TIP 3: 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 > > > > > -- > > Regards, > /Aaron > > ---------------------------(end of broadcast)--------------------------- > TIP 5: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faqs/FAQ.html > From pgsql-performance-owner@postgresql.org Wed Oct 20 14:23:39 2004 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 EA41F32A1FF for ; Wed, 20 Oct 2004 14:23:38 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 13084-09 for ; Wed, 20 Oct 2004 13:23:23 +0000 (GMT) Received: from bayswater1.ymogen.net (host-154-240-27-217.pobox.net.uk [217.27.240.154]) by svr1.postgresql.org (Postfix) with ESMTP id 940E232A11E for ; Wed, 20 Oct 2004 14:23:24 +0100 (BST) Received: from [82.68.132.233] (82-68-132-233.dsl.in-addr.zen.co.uk [82.68.132.233]) by bayswater1.ymogen.net (Postfix) with ESMTP id 458A8A3A8E; Wed, 20 Oct 2004 14:23:22 +0100 (BST) Message-ID: <417666C9.80008@ymogen.net> Date: Wed, 20 Oct 2004 14:23:21 +0100 From: Matt Clark User-Agent: Mozilla Thunderbird 0.7.3 (Windows/20040803) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Fischer Cc: pgsql-performance@postgresql.org Subject: Re: OS desicion References: <20041020151015.766bb00c@nixe> In-Reply-To: <20041020151015.766bb00c@nixe> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/314 X-Sequence-Number: 8790 You are asking the wrong question. The best OS is the OS you (and/or the customer) knows and can administer competently. The real performance differences between unices are so small as to be ignorable in this context. The context switching bug is not OS-dependent, but varys in severity across machine architectures (I understand it to be mostly P4/Athlon related, but don't take my word for it). M Tom Fischer wrote: >Hi List, > >I have a Dual-Xeon 3Ghz System with with GB RAM and an Adaptec 212� SCSI >RAID with 4 SCA Harddiscs. Our customer wants to have the Machine tuned >for best Database performance. Which OS should we used? We are tending >between Linux 2.6 or FreeBSD. The Database Size is 5GB and ascending. >Most SQL-Queries are Selects, the Tablesizes are beetween 300k and up to >10 MB. I've read the Hardware Performance Guide and the result was to >take FreeBSD in the Decision too :) > >And what is on this Context Switiching Bug i have read in the Archive? > >Hope you can help me > >Regards > >Tom > >---------------------------(end of broadcast)--------------------------- >TIP 3: 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 Oct 20 14:36:58 2004 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 D21CD32A1FF for ; Wed, 20 Oct 2004 14:36:57 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 17237-09 for ; Wed, 20 Oct 2004 13:36:46 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.194]) by svr1.postgresql.org (Postfix) with ESMTP id B2670329F90 for ; Wed, 20 Oct 2004 14:36:48 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 74so454090rnk for ; Wed, 20 Oct 2004 06:36:49 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; 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:references; b=W37flJuiyWw8N8vsFdmSEhpLS1CCwtm6PzuC7TKPyA4aDHp6nitkJdRIYbgxDR5mGzM+L8a9S7MISuNB2bgPMMchEbB50ikYB1uhhRDfEM0VzkFf3ng6VBLtqzOznHQoSp5pwDMal7aowkYCdZz2y0P5YrVO1BiQky7QgfgO5qI Received: by 10.38.152.39 with SMTP id z39mr2542808rnd; Wed, 20 Oct 2004 06:36:49 -0700 (PDT) Received: by 10.38.76.42 with HTTP; Wed, 20 Oct 2004 06:36:49 -0700 (PDT) Message-ID: <4a0cafe204102006361e372071@mail.gmail.com> Date: Wed, 20 Oct 2004 08:36:49 -0500 From: Josh Close Reply-To: Josh Close To: POSTGRES-PERFORMANCE Subject: Re: how much mem to give postgres? In-Reply-To: <29976.1098246931@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: <4a0cafe2041019153133f854e9@mail.gmail.com> <29976.1098246931@sss.pgh.pa.us> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/315 X-Sequence-Number: 8791 On Wed, 20 Oct 2004 00:35:31 -0400, Tom Lane wrote: > I suspect that fooling with shared_buffers is entirely the wrong tree > for you to be barking up. My suggestion is to be looking at individual > queries that are slow, and seeing how to speed those up. This might > involve adding indexes, or tweaking the query source, or adjusting > planner parameters, or several other things. EXPLAIN ANALYZE is your > friend ... > > regards, tom lane Only problem is, a "select count(1)" is taking a long time. Indexes shouldn't matter with this since it's counting every row, right? The tables are fairly well indexed also, I could probably add a few more. If shared_buffers isn't the way to go ( you said 10k is the sweetspot ), then what about the effective_cache_size? I was suggested on the general list about possibly setting that to 75% of ram. Thanks. -Josh From pgsql-performance-owner@postgresql.org Wed Oct 20 14:40:06 2004 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 BCDA232A20C for ; Wed, 20 Oct 2004 14:40:05 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 21520-01 for ; Wed, 20 Oct 2004 13:39:50 +0000 (GMT) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.193]) by svr1.postgresql.org (Postfix) with ESMTP id D198132A245 for ; Wed, 20 Oct 2004 14:39:53 +0100 (BST) Received: by mproxy.gmail.com with SMTP id 74so454395rnk for ; Wed, 20 Oct 2004 06:39:53 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; 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:references; b=JarKR8W2Nk6BbYEST7obGQAqKM5OGGUuTNxuC6dZjrGXmdYrhrxaKDXRuXT2n4Ayjf1A0EmqYZmMqzP8KLX3MelxoNWf7tsO/oKeO/izWwEMQYzjbm2QNojaH5fmo1kLZFPaC6Gjivm5Gl+dJ8sYI908E2Aygcr1fIpMFOi6HQ0 Received: by 10.38.24.12 with SMTP id 12mr2533446rnx; Wed, 20 Oct 2004 06:39:53 -0700 (PDT) Received: by 10.38.76.42 with HTTP; Wed, 20 Oct 2004 06:39:53 -0700 (PDT) Message-ID: <4a0cafe204102006391f726a4f@mail.gmail.com> Date: Wed, 20 Oct 2004 08:39:53 -0500 From: Josh Close Reply-To: Josh Close To: POSTGRES-PERFORMANCE Subject: Re: how much mem to give postgres? In-Reply-To: <200410192223.24285.josh@agliodbs.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: <4a0cafe2041019153133f854e9@mail.gmail.com> <200410192223.24285.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/316 X-Sequence-Number: 8792 On Tue, 19 Oct 2004 22:23:24 -0700, Josh Berkus wrote: > There have been issues with Postgres+HT, especially on Linux 2.4. Try > turning HT off if other tuning doesn't solve things. > > Otherwise, see: > http://www.varlena.com/varlena/GeneralBits/Tidbits/perf.html How would I turn that off? In the kernel config? Not too familiar with that. I have a 2 proc xeon with 4 gigs of mem on the way for postgres, so I hope HT isn't a problem. If HT is turned off, does it just not use the other "half" of the processor? Or does the processor just work as one unit? Also, I'm taking a look at that site right now :) -Josh From pgsql-performance-owner@postgresql.org Wed Oct 20 15:07:08 2004 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 427F132A2AE for ; Wed, 20 Oct 2004 15:07:06 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 28376-04 for ; Wed, 20 Oct 2004 14:06:59 +0000 (GMT) Received: from bayswater1.ymogen.net (host-154-240-27-217.pobox.net.uk [217.27.240.154]) by svr1.postgresql.org (Postfix) with ESMTP id 429F932A298 for ; Wed, 20 Oct 2004 15:07:01 +0100 (BST) Received: from [82.68.132.233] (82-68-132-233.dsl.in-addr.zen.co.uk [82.68.132.233]) by bayswater1.ymogen.net (Postfix) with ESMTP id 453E4A338F; Wed, 20 Oct 2004 15:07:01 +0100 (BST) Message-ID: <41767104.10702@ymogen.net> Date: Wed, 20 Oct 2004 15:07:00 +0100 From: Matt Clark User-Agent: Mozilla Thunderbird 0.7.3 (Windows/20040803) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Josh Close Cc: POSTGRES-PERFORMANCE Subject: Re: how much mem to give postgres? References: <4a0cafe2041019153133f854e9@mail.gmail.com> <200410192223.24285.josh@agliodbs.com> <4a0cafe204102006391f726a4f@mail.gmail.com> In-Reply-To: <4a0cafe204102006391f726a4f@mail.gmail.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/317 X-Sequence-Number: 8793 >How would I turn that off? In the kernel config? Not too familiar with >that. I have a 2 proc xeon with 4 gigs of mem on the way for postgres, >so I hope HT isn't a problem. If HT is turned off, does it just not >use the other "half" of the processor? Or does the processor just work >as one unit? > > You turn it off in the BIOS. There is no 'other half', the processor is just pretending to have two cores by shuffling registers around, which gives maybe a 5-10% performance gain in certain multithreaded situations. A hack to overcome marchitactural limitations due to the overly long pipeline in the Prescott core.. Really of most use for desktop interactivity rather than actual throughput. M From pgsql-performance-owner@postgresql.org Wed Oct 20 17:14:04 2004 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 A37C432B11B for ; Wed, 20 Oct 2004 17:14:02 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 14348-01 for ; Wed, 20 Oct 2004 16:13:57 +0000 (GMT) Received: from window.monsterlabs.com (window.monsterlabs.com [216.183.105.176]) by svr1.postgresql.org (Postfix) with SMTP id 8F05B32B10F for ; Wed, 20 Oct 2004 17:13:56 +0100 (BST) Received: (qmail 9379 invoked from network); 20 Oct 2004 16:13:53 -0000 Received: from pcp02680586pcs.nash01.tn.comcast.net (HELO ?192.168.0.114?) (68.52.123.224) by 0 with SMTP; 20 Oct 2004 16:13:53 -0000 In-Reply-To: <0c4501c4b604$06f15460$0b00a8c0@forge> References: <0bef01c4b5f8$70026ae0$0b00a8c0@forge> <16865.1098204769@sss.pgh.pa.us> <0c4501c4b604$06f15460$0b00a8c0@forge> Mime-Version: 1.0 (Apple Message framework v619) Content-Type: text/plain; charset=US-ASCII; format=flowed Message-Id: <0736EED9-22B3-11D9-A5E5-000D93AE0944@sitening.com> Content-Transfer-Encoding: 7bit Cc: From: Thomas F.O'Connell Subject: Re: Index not used in query. Why? Date: Wed, 20 Oct 2004 11:13:49 -0500 To: "Contact AR-SD.NET" X-Mailer: Apple Mail (2.619) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/320 X-Sequence-Number: 8796 There's a chance that you could gain from quoting the '4' and '6' if those orders.id_status isn't a pure int column and is indexed. See http://www.postgresql.org/docs/7.4/static/datatype.html#DATATYPE-INT -tfo -- Thomas F. O'Connell Co-Founder, Information Architect Sitening, LLC http://www.sitening.com/ 110 30th Avenue North, Suite 6 Nashville, TN 37203-6320 615-260-0005 On Oct 19, 2004, at 12:49 PM, Contact AR-SD.NET wrote: > Is there a solution to make it faster? > At the end I need only in the query the id_status =4 and 6, but if I > write > in the sql query (where condition) where id_status in (4,6), the > explain > says the same(the slow version). > > For example: > SELECT count(o.id) FROM orders o > INNER JOIN report r ON > o.id=r.id_order > INNER JOIN status s ON > o.id_status=s.id > INNER JOIN contact c ON o.id_ag=c.id > INNER JOIN endkunde e ON > o.id_endkunde=e.id > INNER JOIN zufriden z ON > r.id_zufriden=z.id > INNER JOIN plannung v ON > v.id=o.id_plannung > INNER JOIN mpsworker w ON > v.id_worker=w.id > INNER JOIN person p ON p.id = > w.id_person > WHERE o.id_status in (4,6); > > The result for this query is also without index searches. > > I really have to make this query a little more faster. Suggestions? > > Regards, > Andy. > > ----- Original Message ----- > From: "Tom Lane" > To: "Andrei Bintintan" > Cc: > Sent: Tuesday, October 19, 2004 7:52 PM > Subject: Re: [PERFORM] Index not used in query. Why? > > >> "Andrei Bintintan" writes: >>> Hi to all! I have the following query. The execution time is very >>> big, > it >>> doesn't use the indexes and I don't understand why... >> >> Indexes are not necessarily the best way to do a large join. >> >>> If I use the following query the indexes are used: >> >> The key reason this wins seems to be that the id_status = 4 condition >> is far more selective than id_status > 3 (the estimates are 52 and >> 36967 >> rows respectively ... is that accurate?) which means that the second >> query is inherently about 1/700th as much work. This, and not the use >> of indexes, is the fundamental reason why it's faster. >> >> regards, tom lane >> >> ---------------------------(end of >> broadcast)--------------------------- >> TIP 4: Don't 'kill -9' the postmaster >> > > > ---------------------------(end of > broadcast)--------------------------- > TIP 2: you can get off all lists at once with the unregister command > (send "unregister YourEmailAddressHere" to > majordomo@postgresql.org) From pgsql-performance-owner@postgresql.org Wed Oct 20 17:39:17 2004 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 0341C32A12E for ; Wed, 20 Oct 2004 17:39:16 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 31312-08 for ; Wed, 20 Oct 2004 16:39:10 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 3B26332A04B for ; Wed, 20 Oct 2004 17:39:09 +0100 (BST) 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 6533609; Wed, 20 Oct 2004 09:40:33 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: OS desicion Date: Wed, 20 Oct 2004 09:38:51 -0700 User-Agent: KMail/1.6.2 Cc: Matt Clark , Tom Fischer References: <20041020151015.766bb00c@nixe> <417666C9.80008@ymogen.net> In-Reply-To: <417666C9.80008@ymogen.net> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410200938.51391.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/321 X-Sequence-Number: 8797 Tom, > You are asking the wrong question. The best OS is the OS you (and/or > the customer) knows and can administer competently. I'll have to 2nd this. > The real > performance differences between unices are so small as to be ignorable > in this context. Well, at least the difference between Linux and BSD. There are substantial tradeoffs should you chose to use Solaris or UnixWare. > The context switching bug is not OS-dependent, but > varys in severity across machine architectures (I understand it to be > mostly P4/Athlon related, but don't take my word for it). The bug is at its apparent worst on multi-processor HT Xeons and weak northbridges running Linux 2.4. However, it has been demonstrated (with lesser impact) on Solaris/Sparc, PentiumIII, and Athalon. Primarily it seems to affect data warehousing applications. Your choice of OS is not affected by this bug. -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Wed Oct 20 18:12:10 2004 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 78A9832A791 for ; Wed, 20 Oct 2004 18:12:07 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 53445-06 for ; Wed, 20 Oct 2004 17:11:58 +0000 (GMT) Received: from rambo.iniquinet.com (rambo.iniquinet.com [69.39.89.10]) by svr1.postgresql.org (Postfix) with ESMTP id 7176E32A1BB for ; Wed, 20 Oct 2004 18:11:54 +0100 (BST) Received: (qmail 6077 invoked by uid 1010); 20 Oct 2004 12:45:37 -0400 Received: from Robert_Creager@LogicalChaos.org by rambo.iniquinet.com by uid 1002 with qmail-scanner-1.20 (clamscan: 0.70. spamassassin: 2.64. Clear:RC:0(63.147.78.131):SA:0(-4.6/6.0):. Processed in 1.281667 secs); 20 Oct 2004 16:45:37 -0000 Received: from unknown (HELO thunder.mshome.net) (perl?test@logicalchaos.org@63.147.78.131) by rambo.iniquinet.com with AES256-SHA encrypted SMTP; 20 Oct 2004 12:45:36 -0400 Received: from localhost.localdomain (thunder.mshome.net [192.168.0.250]) by thunder.mshome.net (Postfix) with ESMTP id C1C2639C9; Wed, 20 Oct 2004 10:45:31 -0600 (MDT) Received: from logicalchaos.org (thunder.mshome.net [192.168.0.250]) by thunder.mshome.net (Postfix) with SMTP id 4303931CF; Wed, 20 Oct 2004 10:45:28 -0600 (MDT) Date: Wed, 20 Oct 2004 10:45:27 -0600 From: Robert Creager To: Rod Taylor Cc: Brock Henry , Postgresql Performance Subject: Re: Insert performance, what should I expect? Message-ID: <20041020104527.14927c06@thunder.mshome.net> In-Reply-To: <1098238348.747.20.camel@home> References: <97b3fe2041019185337a8c3d8@mail.gmail.com> <1098238348.747.20.camel@home> Organization: Starlight Vision, LLC. X-Mailer: Sylpheed-Claws 0.9.12a (GTK+ 1.2.10; i586-mandrake-linux-gnu) Mime-Version: 1.0 Content-Type: multipart/signed; protocol="application/pgp-signature"; micalg="pgp-sha1"; boundary="Signature=_Wed__20_Oct_2004_10_45_27_-0600_FSwak.CH9e/r+FHK" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/323 X-Sequence-Number: 8799 --Signature=_Wed__20_Oct_2004_10_45_27_-0600_FSwak.CH9e/r+FHK Content-Type: text/plain; charset=US-ASCII Content-Disposition: inline Content-Transfer-Encoding: 7bit When grilled further on (Tue, 19 Oct 2004 22:12:28 -0400), Rod Taylor confessed: > > I've done some manual benchmarking running my script 'time script.pl' > > I realise my script uses some of the time, bench marking shows that > > %50 of the time is spent in dbd:execute. > > > 1) Drop DBD::Pg and switch to the Pg driver for Perl instead (non-DBI > compliant) which has functions similar to putline() that allow COPY to > be used. COPY can be used with DBD::Pg, per a script I use: $dbh->do( "COPY temp_obs_$band ( $col_list ) FROM stdin" ); $dbh->func( join ( "\t", @data ) . "\n", 'putline' ); $dbh->func( "\\.\n", 'putline' ); $dbh->func( 'endcopy' ); With sets of data from 1000 to 8000 records, my COPY performance is consistent at ~10000 records per second. Cheers, Rob -- 10:39:31 up 2 days, 16:25, 2 users, load average: 2.15, 2.77, 3.06 Linux 2.6.5-02 #8 SMP Mon Jul 12 21:34:44 MDT 2004 --Signature=_Wed__20_Oct_2004_10_45_27_-0600_FSwak.CH9e/r+FHK Content-Type: application/pgp-signature -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (GNU/Linux) iEYEARECAAYFAkF2licACgkQLQ/DKuwDYzkCTQCfTAdnx0jDOUE3N2PyTesJcIgQ AvIAnitYXOsWZpmcQJGXM1rOFCJ+4p8y =ysQM -----END PGP SIGNATURE----- --Signature=_Wed__20_Oct_2004_10_45_27_-0600_FSwak.CH9e/r+FHK-- From pgsql-performance-owner@postgresql.org Mon Oct 25 16:57:46 2004 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 8A09232A178 for ; Wed, 20 Oct 2004 17:52:27 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 34483-10 for ; Wed, 20 Oct 2004 16:52:12 +0000 (GMT) Received: from dbl.q-ag.de (dbl.q-ag.de [213.172.117.3]) by svr1.postgresql.org (Postfix) with ESMTP id AA73032A134 for ; Wed, 20 Oct 2004 17:52:08 +0100 (BST) Received: from [127.0.0.2] (dbl [127.0.0.1]) by dbl.q-ag.de (8.12.3/8.12.3/Debian-6.6) with ESMTP id i9KGpoSL003977; Wed, 20 Oct 2004 18:51:52 +0200 Message-ID: <417697A5.1050600@colorfullife.com> Date: Wed, 20 Oct 2004 18:51:49 +0200 From: Manfred Spraul User-Agent: Mozilla/5.0 (X11; U; Linux i686; fr-FR; rv:1.7.3) Gecko/20040922 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: neilc@samurai.com, markw@osdl.org, pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> In-Reply-To: <20151.1098222733@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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/411 X-Sequence-Number: 8887 Tom Lane wrote: >Manfred Spraul writes: > > >>Has anyone tried to replace the whole lwlock implementation with >>pthread_rwlock? At least for Linux with recent glibcs, pthread_rwlock is >>implemented with futexes, i.e. we would get a fast lock handling without >>os specific hacks. >> >> > >"At least for Linux" does not strike me as equivalent to "without >OS-specific hacks". > > > For me, "at least for Linux" means that I have tested the patch with Linux. I'd expect that the patch works on most recent unices (pthread_rwlock_t is probably mandatory for Unix98 compatibility). You and others on this mailing list have access to other systems - my patch should be seen as a call for testers, not as a proposal for merging. I expect that Linux is not the only OS with fast user space semaphores, and if an OS has such objects, then the pthread_ locking functions are hopefully implemented by using them. IMHO it's better to support the standard function instead of trying to use the native (and OS specific) fast semaphore functions. >The bigger problem here is that the SMP locking bottlenecks we are >currently seeing are *hardware* issues (AFAICT anyway). The only way >that futexes can offer a performance win is if they have a smarter way >of executing the basic atomic-test-and-set sequence than we do; > > lwlocks operations are not a basic atomic-test-and-set sequence. They are spinlock, several nonatomic operations, spin_unlock. -- Manfred From pgsql-performance-owner@postgresql.org Wed Oct 20 18:10:24 2004 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 3C96D32B0DF for ; Wed, 20 Oct 2004 18:10:22 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49924-06 for ; Wed, 20 Oct 2004 17:10:11 +0000 (GMT) Received: from mail.osdl.org (fw.osdl.org [65.172.181.6]) by svr1.postgresql.org (Postfix) with ESMTP id D2E7832AEB0 for ; Wed, 20 Oct 2004 18:10:10 +0100 (BST) Received: (from markw@localhost) by mail.osdl.org (8.11.6/8.11.6) id i9KHA1g13777; Wed, 20 Oct 2004 10:10:01 -0700 Date: Wed, 20 Oct 2004 10:10:01 -0700 From: Mark Wong To: Manfred Spraul Cc: neilc@samurai.com, pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 Message-ID: <20041020101001.A13359@osdl.org> References: <417221B5.1050704@colorfullife.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5i In-Reply-To: <417221B5.1050704@colorfullife.com>; from manfred@colorfullife.com on Sun, Oct 17, 2004 at 09:39:33AM +0200 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/322 X-Sequence-Number: 8798 On Sun, Oct 17, 2004 at 09:39:33AM +0200, Manfred Spraul wrote: > Neil wrote: > > >. In any case, the "futex patch" > >uses the Linux 2.6 futex API to implement PostgreSQL spinlocks. > > > Has anyone tried to replace the whole lwlock implementation with > pthread_rwlock? At least for Linux with recent glibcs, pthread_rwlock is > implemented with futexes, i.e. we would get a fast lock handling without > os specific hacks. Perhaps other os contain user space pthread locks, too. > Attached is an old patch. I tested it on an uniprocessor system a year > ago and it didn't provide much difference, but perhaps the scalability > is better. You'll have to add -lpthread to the library list for linking. I've heard that simply linking to the pthreads libraries, regardless of whether you're using them or not creates a significant overhead. Has anyone tried it for kicks? Mark From pgsql-performance-owner@postgresql.org Wed Oct 20 18:12:54 2004 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 26E0C329FED for ; Wed, 20 Oct 2004 18:12:53 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 55793-04 for ; Wed, 20 Oct 2004 17:12:49 +0000 (GMT) Received: from bayswater1.ymogen.net (host-154-240-27-217.pobox.net.uk [217.27.240.154]) by svr1.postgresql.org (Postfix) with ESMTP id 66175329FB2 for ; Wed, 20 Oct 2004 18:12:45 +0100 (BST) Received: from [82.68.132.233] (82-68-132-233.dsl.in-addr.zen.co.uk [82.68.132.233]) by bayswater1.ymogen.net (Postfix) with ESMTP id 33139A2EB2; Wed, 20 Oct 2004 18:12:39 +0100 (BST) Message-ID: <41769C86.5090208@ymogen.net> Date: Wed, 20 Oct 2004 18:12:38 +0100 From: Matt Clark User-Agent: Mozilla Thunderbird 0.7.3 (Windows/20040803) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Josh Berkus Cc: pgsql-performance@postgresql.org, Tom Fischer Subject: Re: OS desicion References: <20041020151015.766bb00c@nixe> <417666C9.80008@ymogen.net> <200410200938.51391.josh@agliodbs.com> In-Reply-To: <200410200938.51391.josh@agliodbs.com> Content-Type: multipart/alternative; boundary="------------050501040502070906030603" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.4 tagged_above=0.0 required=5.0 tests=HTML_30_40, HTML_MESSAGE, HTML_TITLE_EMPTY X-Spam-Level: * X-Archive-Number: 200410/324 X-Sequence-Number: 8800 This is a multi-part message in MIME format. --------------050501040502070906030603 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit >>The real >>performance differences between unices are so small as to be ignorable >>in this context. >> >> > <> > Well, at least the difference between Linux and BSD. There are > substantial > tradeoffs should you chose to use Solaris or UnixWare. Yes, quite right, I should have said 'popular x86-based unices'. --------------050501040502070906030603 Content-Type: text/html; charset=us-ascii Content-Transfer-Encoding: 7bit 
The real 
performance differences between unices are so small as to be ignorable
in this context.
<>
Well, at least the difference between Linux and BSD. There are substantial
tradeoffs should you chose to use Solaris or UnixWare.
Yes, quite right, I should have said 'popular x86-based unices'. 
--------------050501040502070906030603-- From pgsql-performance-owner@postgresql.org Mon Oct 25 16:58:29 2004 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 9864A32B16F for ; Wed, 20 Oct 2004 18:15:09 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 66618-01 for ; Wed, 20 Oct 2004 17:14:55 +0000 (GMT) Received: from dbl.q-ag.de (dbl.q-ag.de [213.172.117.3]) by svr1.postgresql.org (Postfix) with ESMTP id 2858632A2FD for ; Wed, 20 Oct 2004 18:14:54 +0100 (BST) Received: from [127.0.0.2] (dbl [127.0.0.1]) by dbl.q-ag.de (8.12.3/8.12.3/Debian-6.6) with ESMTP id i9KHEkSL004225; Wed, 20 Oct 2004 19:14:47 +0200 Message-ID: <41769D06.7080407@colorfullife.com> Date: Wed, 20 Oct 2004 19:14:46 +0200 From: Manfred Spraul User-Agent: Mozilla/5.0 (X11; U; Linux i686; fr-FR; rv:1.7.3) Gecko/20040922 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Mark Wong Cc: neilc@samurai.com, pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 References: <417221B5.1050704@colorfullife.com> <20041020101001.A13359@osdl.org> In-Reply-To: <20041020101001.A13359@osdl.org> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/419 X-Sequence-Number: 8895 Mark Wong wrote: >I've heard that simply linking to the pthreads libraries, regardless of >whether you're using them or not creates a significant overhead. Has >anyone tried it for kicks? > > > That depends on the OS and the functions that are used. The typical worst case is buffered IO of single characters: The single threaded implementation is just copy and update buffer status, the multi threaded implementation contains full locking. For most other functions there is no difference at all. -- Manfred From pgsql-performance-owner@postgresql.org Wed Oct 20 18:15:58 2004 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 9266832A3B9 for ; Wed, 20 Oct 2004 18:15:56 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 54367-10 for ; Wed, 20 Oct 2004 17:15:42 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 9C65732A37C for ; Wed, 20 Oct 2004 18:15:41 +0100 (BST) 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 i9KHFZKW007179; Wed, 20 Oct 2004 13:15:35 -0400 (EDT) To: Manfred Spraul Cc: neilc@samurai.com, markw@osdl.org, pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 In-reply-to: <417697A5.1050600@colorfullife.com> References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> <417697A5.1050600@colorfullife.com> Comments: In-reply-to Manfred Spraul message dated "Wed, 20 Oct 2004 18:51:49 +0200" Date: Wed, 20 Oct 2004 13:15:35 -0400 Message-ID: <7178.1098292535@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/325 X-Sequence-Number: 8801 Manfred Spraul writes: > Tom Lane wrote: >> The bigger problem here is that the SMP locking bottlenecks we are >> currently seeing are *hardware* issues (AFAICT anyway). The only way >> that futexes can offer a performance win is if they have a smarter way >> of executing the basic atomic-test-and-set sequence than we do; >> > lwlocks operations are not a basic atomic-test-and-set sequence. They > are spinlock, several nonatomic operations, spin_unlock. Right, and it is the spinlock that is the problem. See discussions a few months back: at least on Intel SMP machines, most of the problem seems to have to do with trading the spinlock's cache line back and forth between CPUs. It's difficult to see how a futex is going to avoid that. regards, tom lane From pgsql-performance-owner@postgresql.org Wed Oct 20 18:21:17 2004 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 028B132A341 for ; Wed, 20 Oct 2004 18:21:16 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 66940-04 for ; Wed, 20 Oct 2004 17:21:11 +0000 (GMT) Received: from tht.net (vista.tht.net [216.126.88.2]) by svr1.postgresql.org (Postfix) with ESMTP id F400E32A31C for ; Wed, 20 Oct 2004 18:21:11 +0100 (BST) Received: from [134.22.69.212] (dyn-69-212.tor.dsl.tht.net [134.22.69.212]) by tht.net (Postfix) with ESMTP id 96D0B76BA8; Wed, 20 Oct 2004 13:21:13 -0400 (EDT) Subject: Re: Insert performance, what should I expect? From: Rod Taylor To: Robert Creager Cc: Brock Henry , Postgresql Performance In-Reply-To: <20041020104527.14927c06@thunder.mshome.net> References: <97b3fe2041019185337a8c3d8@mail.gmail.com> <1098238348.747.20.camel@home> <20041020104527.14927c06@thunder.mshome.net> Content-Type: text/plain Message-Id: <1098292819.747.188.camel@home> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Wed, 20 Oct 2004 13:20:19 -0400 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/326 X-Sequence-Number: 8802 On Wed, 2004-10-20 at 12:45, Robert Creager wrote: > When grilled further on (Tue, 19 Oct 2004 22:12:28 -0400), > Rod Taylor confessed: > > > > I've done some manual benchmarking running my script 'time script.pl' > > > I realise my script uses some of the time, bench marking shows that > > > %50 of the time is spent in dbd:execute. > > > > > 1) Drop DBD::Pg and switch to the Pg driver for Perl instead (non-DBI > > compliant) which has functions similar to putline() that allow COPY to > > be used. > > COPY can be used with DBD::Pg, per a script I use: > > $dbh->do( "COPY temp_obs_$band ( $col_list ) FROM stdin" ); > $dbh->func( join ( "\t", @data ) . "\n", 'putline' ); > $dbh->func( "\\.\n", 'putline' ); > $dbh->func( 'endcopy' ); Thanks for that. All of the conversations I've seen on the subject stated that DBD::Pg only supported standard DB features -- copy not amongst them. > With sets of data from 1000 to 8000 records, my COPY performance is consistent > at ~10000 records per second. Well done. From pgsql-performance-owner@postgresql.org Mon Oct 25 16:57:57 2004 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 C431E329EC2 for ; Wed, 20 Oct 2004 18:39:32 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 94125-09 for ; Wed, 20 Oct 2004 17:39:22 +0000 (GMT) Received: from dbl.q-ag.de (dbl.q-ag.de [213.172.117.3]) by svr1.postgresql.org (Postfix) with ESMTP id 85BF832A17A for ; Wed, 20 Oct 2004 18:39:22 +0100 (BST) Received: from [127.0.0.2] (dbl [127.0.0.1]) by dbl.q-ag.de (8.12.3/8.12.3/Debian-6.6) with ESMTP id i9KHdDSL005028; Wed, 20 Oct 2004 19:39:14 +0200 Message-ID: <4176A2C1.3070205@colorfullife.com> Date: Wed, 20 Oct 2004 19:39:13 +0200 From: Manfred Spraul User-Agent: Mozilla/5.0 (X11; U; Linux i686; fr-FR; rv:1.7.3) Gecko/20040922 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: neilc@samurai.com, markw@osdl.org, pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> <417697A5.1050600@colorfullife.com> <7178.1098292535@sss.pgh.pa.us> In-Reply-To: <7178.1098292535@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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/413 X-Sequence-Number: 8889 Tom Lane wrote: >Manfred Spraul writes: > > >>Tom Lane wrote: >> >> >>>The bigger problem here is that the SMP locking bottlenecks we are >>>currently seeing are *hardware* issues (AFAICT anyway). The only way >>>that futexes can offer a performance win is if they have a smarter way >>>of executing the basic atomic-test-and-set sequence than we do; >>> >>> >>> >>lwlocks operations are not a basic atomic-test-and-set sequence. They >>are spinlock, several nonatomic operations, spin_unlock. >> >> > >Right, and it is the spinlock that is the problem. See discussions a >few months back: at least on Intel SMP machines, most of the problem >seems to have to do with trading the spinlock's cache line back and >forth between CPUs. > I'd disagree: cache line bouncing is one problem. If this happens then there is only one solution: The number of changes to that cacheline must be reduced. The tools that are used in the linux kernel are: - hashing. An emergency approach if there is no other solution. I think RedHat used it for the buffer cache RH AS: Instead of one buffer cache, there were lots of smaller buffer caches with individual locks. The cache was chosen based on the file position (probably mixed with some pointers to avoid overloading cache 0). - For read-heavy loads: sequence locks. A reader reads a counter value and then accesses the data structure. At the end it checks if the counter was modified. If it's still the same value then it can continue, otherwise it must retry. Writers acquire a normal spinlock and then modify the counter value. RCU is the second option, but there are patents - please be careful before using that tool. - complete rewrites that avoid the global lock. I think the global buffer cache is now gone, everything is handled per-file. I think there is a global list for buffer replacement, but the at the top of the buffer replacement strategy is a simple clock algorithm. That means that simple lookups/accesses just set a (local) referenced bit and don't have to acquire a global lock. I know that this is the total opposite of ARC, but perhaps it's the only scalable solution. ARC could be used as the second level strategy. But: According to the descriptions the problem is a context switch storm. I don't see that cache line bouncing can cause a context switch storm. What causes the context switch storm? If it's the pg_usleep in s_lock, then my patch should help a lot: with pthread_rwlock locks, this line doesn't exist anymore. -- Manfred From pgsql-performance-owner@postgresql.org Wed Oct 20 18:44:24 2004 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 2EDBB329F1E for ; Wed, 20 Oct 2004 18:44:24 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 36063-03 for ; Wed, 20 Oct 2004 17:44:14 +0000 (GMT) Received: from gp.word-to-the-wise.com (gp.word-to-the-wise.com [64.71.176.18]) by svr1.postgresql.org (Postfix) with ESMTP id B12F2329F0C for ; Wed, 20 Oct 2004 18:44:13 +0100 (BST) Received: by gp.word-to-the-wise.com (Postfix, from userid 500) id 0EB529029FB; Wed, 20 Oct 2004 10:50:39 -0700 (PDT) Date: Wed, 20 Oct 2004 10:50:39 -0700 From: Steve Atkins To: POSTGRES-PERFORMANCE Subject: Re: how much mem to give postgres? Message-ID: <20041020175038.GA1339@gp.word-to-the-wise.com> References: <4a0cafe2041019153133f854e9@mail.gmail.com> <200410192223.24285.josh@agliodbs.com> <4a0cafe204102006391f726a4f@mail.gmail.com> <41767104.10702@ymogen.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <41767104.10702@ymogen.net> User-Agent: Mutt/1.4.1i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/327 X-Sequence-Number: 8803 On Wed, Oct 20, 2004 at 03:07:00PM +0100, Matt Clark wrote: > You turn it off in the BIOS. There is no 'other half', the processor is > just pretending to have two cores by shuffling registers around, which > gives maybe a 5-10% performance gain in certain multithreaded > situations. > A hack to overcome marchitactural limitations due > to the overly long pipeline in the Prescott core.. Really of > most use for desktop interactivity rather than actual throughput. Hyperthreading is actually an excellent architectural feature that can give significant performance gains when implemented well and used for an appropriate workload under a decently HT aware OS. IMO, typical RDBMS streams are not an obviously appropriate workload, Intel didn't implement it particularly well and I don't think there are any OSes that support it particularly well. But don't write off using it in the future, when it's been improved at both the OS and the silicon levels. Cheers, Steve From pgsql-performance-owner@postgresql.org Wed Oct 20 19:16:33 2004 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 EB0EC329F19 for ; Wed, 20 Oct 2004 19:16:32 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 58685-01 for ; Wed, 20 Oct 2004 18:16:22 +0000 (GMT) Received: from bayswater1.ymogen.net (host-154-240-27-217.pobox.net.uk [217.27.240.154]) by svr1.postgresql.org (Postfix) with ESMTP id 15AD8329EC2 for ; Wed, 20 Oct 2004 19:16:19 +0100 (BST) Received: from [82.68.132.233] (82-68-132-233.dsl.in-addr.zen.co.uk [82.68.132.233]) by bayswater1.ymogen.net (Postfix) with ESMTP id 4B10BA2EB2; Wed, 20 Oct 2004 19:16:19 +0100 (BST) Message-ID: <4176AB72.6040101@ymogen.net> Date: Wed, 20 Oct 2004 19:16:18 +0100 From: Matt Clark User-Agent: Mozilla Thunderbird 0.7.3 (Windows/20040803) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Steve Atkins Cc: POSTGRES-PERFORMANCE Subject: Re: how much mem to give postgres? References: <4a0cafe2041019153133f854e9@mail.gmail.com> <200410192223.24285.josh@agliodbs.com> <4a0cafe204102006391f726a4f@mail.gmail.com> <41767104.10702@ymogen.net> <20041020175038.GA1339@gp.word-to-the-wise.com> In-Reply-To: <20041020175038.GA1339@gp.word-to-the-wise.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/328 X-Sequence-Number: 8804 > >Hyperthreading is actually an excellent architectural feature that >can give significant performance gains when implemented well and used >for an appropriate workload under a decently HT aware OS. > >IMO, typical RDBMS streams are not an obviously appropriate workload, >Intel didn't implement it particularly well and I don't think there >are any OSes that support it particularly well. > > >But don't write off using it in the future, when it's been improved >at both the OS and the silicon levels. > > > You are quite right of course - unfortunately the current Intel implementation meets nearly none of these criteria! As Rod Taylor pointed out off-list, IBM's SMT implementation on the Power5 is vastly superior. Though he's also just told me that Sun is beating IBM on price/performance for his workload, so who knows how reliable a chap he is... ;-) M From pgsql-performance-owner@postgresql.org Wed Oct 20 19:32:58 2004 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 A5F9F32A341 for ; Wed, 20 Oct 2004 19:32:57 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 63024-09 for ; Wed, 20 Oct 2004 18:32:55 +0000 (GMT) Received: from gp.word-to-the-wise.com (gp.word-to-the-wise.com [64.71.176.18]) by svr1.postgresql.org (Postfix) with ESMTP id 0BB0D32A1C9 for ; Wed, 20 Oct 2004 19:32:55 +0100 (BST) Received: by gp.word-to-the-wise.com (Postfix, from userid 500) id 65E599029FB; Wed, 20 Oct 2004 11:39:22 -0700 (PDT) Date: Wed, 20 Oct 2004 11:39:22 -0700 From: Steve Atkins To: POSTGRES-PERFORMANCE Subject: Re: how much mem to give postgres? Message-ID: <20041020183922.GB2120@gp.word-to-the-wise.com> References: <4a0cafe2041019153133f854e9@mail.gmail.com> <200410192223.24285.josh@agliodbs.com> <4a0cafe204102006391f726a4f@mail.gmail.com> <41767104.10702@ymogen.net> <20041020175038.GA1339@gp.word-to-the-wise.com> <4176AB72.6040101@ymogen.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <4176AB72.6040101@ymogen.net> User-Agent: Mutt/1.4.1i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/329 X-Sequence-Number: 8805 On Wed, Oct 20, 2004 at 07:16:18PM +0100, Matt Clark wrote: > > > >Hyperthreading is actually an excellent architectural feature that > >can give significant performance gains when implemented well and used > >for an appropriate workload under a decently HT aware OS. > > > >IMO, typical RDBMS streams are not an obviously appropriate workload, > >Intel didn't implement it particularly well and I don't think there > >are any OSes that support it particularly well. > > > > > >But don't write off using it in the future, when it's been improved > >at both the OS and the silicon levels. > > > > > > > You are quite right of course - unfortunately the current Intel > implementation meets nearly none of these criteria! Indeed. And when I said "no OSes support it particularly well" I meant the x86 SMT implementation, rather than SMT in general. As Rod pointed out, AIX seems to have decent support and Power has a very nice implementation, and the same is probably true for at least one other OS/architecture implementation. > As Rod Taylor pointed out off-list, IBM's SMT implementation on the > Power5 is vastly superior. Though he's also just told me that Sun > is beating IBM on price/performance for his workload, so who knows > how reliable a chap he is... ;-) :) Cheers, Steve From pgsql-performance-owner@postgresql.org Wed Oct 20 23:06:13 2004 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 7DD0C32A165 for ; Wed, 20 Oct 2004 23:06:12 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 56436-03 for ; Wed, 20 Oct 2004 22:05:59 +0000 (GMT) Received: from mail.osdl.org (fw.osdl.org [65.172.181.6]) by svr1.postgresql.org (Postfix) with ESMTP id D88F432A155 for ; Wed, 20 Oct 2004 23:05:57 +0100 (BST) Received: (from markw@localhost) by mail.osdl.org (8.11.6/8.11.6) id i9KM5Sc10745; Wed, 20 Oct 2004 15:05:28 -0700 Date: Wed, 20 Oct 2004 15:05:28 -0700 From: Mark Wong To: Manfred Spraul Cc: Tom Lane , neilc@samurai.com, pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 Message-ID: <20041020150528.B7838@osdl.org> References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> <417697A5.1050600@colorfullife.com> <7178.1098292535@sss.pgh.pa.us> <4176A2C1.3070205@colorfullife.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5i In-Reply-To: <4176A2C1.3070205@colorfullife.com>; from manfred@colorfullife.com on Wed, Oct 20, 2004 at 07:39:13PM +0200 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/330 X-Sequence-Number: 8806 On Wed, Oct 20, 2004 at 07:39:13PM +0200, Manfred Spraul wrote: > > But: According to the descriptions the problem is a context switch > storm. I don't see that cache line bouncing can cause a context switch > storm. What causes the context switch storm? If it's the pg_usleep in > s_lock, then my patch should help a lot: with pthread_rwlock locks, this > line doesn't exist anymore. > I gave Manfred's patch a try on my 4-way Xeon system with Tom's test_script.sql files. I ran 4 processes of test_script.sql against 8.0beta3 (without any patches) and from my observations with top, the cpu utilization between processors was pretty erratic. They'd jump anywhere from 30% - 70%. With the futex patches that Neil and Gavin have been working on, I'd see the processors evenly utilized at about 50% each. With just Manfred's patch I think there might be a problem somewhere with the patch, or something else, as only one processor is doing anything at a time and 100% utilized. Here are some other details, per Manfred's request: Linux 2.6.8.1 (on a gentoo distro) gcc 3.3.4 glibc 2.3.3.20040420 Mark From pgsql-performance-owner@postgresql.org Thu Oct 21 01:27:37 2004 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 3BAE432A64E for ; Thu, 21 Oct 2004 01:27:34 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 56381-10 for ; Thu, 21 Oct 2004 00:27:25 +0000 (GMT) Received: from net2.micro-automation.com (net2.micro-automation.com [64.7.141.29]) by svr1.postgresql.org (Postfix) with SMTP id 84D7732A569 for ; Thu, 21 Oct 2004 01:27:24 +0100 (BST) Received: (qmail 17092 invoked from network); 21 Oct 2004 00:28:47 -0000 Received: from dcdsl.ebox.com (HELO ?192.168.1.36?) (davec@64.7.143.116) by net2.micro-automation.com with SMTP; 21 Oct 2004 00:28:47 -0000 Message-ID: <41770285.5090604@fastcrypt.com> Date: Wed, 20 Oct 2004 20:27:49 -0400 From: Dave Cramer Reply-To: pg@fastcrypt.com Organization: Postgres International User-Agent: Mozilla Thunderbird 0.8 (X11/20041001) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: Manfred Spraul , neilc@samurai.com, markw@osdl.org, pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> <417697A5.1050600@colorfullife.com> <7178.1098292535@sss.pgh.pa.us> In-Reply-To: <7178.1098292535@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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/331 X-Sequence-Number: 8807 Forgive my naivete, but do futex's implement some priority algorithm for which process gets control. One of the problems as I understand it is that linux does (did ) not implement a priority algorithm, so it is possible for the context which just gave up control to be the next context woken up, which of course is a complete waste of time. --dc-- Tom Lane wrote: >Manfred Spraul writes: > > >>Tom Lane wrote: >> >> >>>The bigger problem here is that the SMP locking bottlenecks we are >>>currently seeing are *hardware* issues (AFAICT anyway). The only way >>>that futexes can offer a performance win is if they have a smarter way >>>of executing the basic atomic-test-and-set sequence than we do; >>> >>> >>> >>lwlocks operations are not a basic atomic-test-and-set sequence. They >>are spinlock, several nonatomic operations, spin_unlock. >> >> > >Right, and it is the spinlock that is the problem. See discussions a >few months back: at least on Intel SMP machines, most of the problem >seems to have to do with trading the spinlock's cache line back and >forth between CPUs. It's difficult to see how a futex is going to avoid >that. > > regards, tom lane > >---------------------------(end of broadcast)--------------------------- >TIP 3: 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 > > > > -- Dave Cramer www.postgresintl.com 519 939 0336 ICQ#14675561 From pgsql-performance-owner@postgresql.org Thu Oct 21 01:28:28 2004 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 20BCE32AEC8 for ; Thu, 21 Oct 2004 01:28:23 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 57552-07 for ; Thu, 21 Oct 2004 00:28:13 +0000 (GMT) Received: from smtp812.mail.sc5.yahoo.com (smtp812.mail.sc5.yahoo.com [66.163.170.82]) by svr1.postgresql.org (Postfix) with SMTP id B780432AE2C for ; Thu, 21 Oct 2004 01:28:11 +0100 (BST) Received: from unknown (HELO ?192.168.0.2?) (jellej@pacbell.net@67.127.86.222 with login) by smtp812.mail.sc5.yahoo.com with SMTP; 21 Oct 2004 00:28:07 -0000 Date: Wed, 20 Oct 2004 17:28:08 -0700 (PDT) From: jelle X-X-Sender: jelle@localhost.localdomain Reply-To: jellej@pacbell.net To: Postgresql Performance Subject: iostat question Message-ID: 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/332 X-Sequence-Number: 8808 Hello All, I have an iostat question in that one of the raid arrays seems to act differently than the other 3. Is this reasonable behavior for the database or should I suspect a hardware or configuration problem? But first some background: Postgresql 7.4.2 Linux 2.4.20, 2GB RAM, 1-Xeon 2.4ghz with HT turned off 3Ware SATA RAID controller with 8 identical drives configured as 4 RAID-1 spindles 64MB RAM disk postgresql.conf differences to postgresql.conf.sample: tcpip_socket = true max_connections = 128 shared_buffers = 2048 vacuum_mem = 16384 max_fsm_pages = 50000 wal_buffers = 128 checkpoint_segments = 64 effective_cache_size = 196000 random_page_cost = 1 default_statistics_target = 100 stats_command_string = true stats_block_level = true stats_row_level = true The database is spread over 5 spindles: /ram0 holds the busiest insert/update/delete table and assoc. indexes for temporary session data /sda5 holds the OS and most of the tables and indexes /sdb2 holds the WAL /sdc1 holds the 2nd busiest i/u/d table (70% of the writes) /sdd1 holds the single index for that busy table on/sdc1 Lately we have 45 connections open from a python/psycopg connection pool. 99% of the reads are cached. No swapping. And finally iostat reports: Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s rkB/s wkB/s avgrq-sz avgqu-sz await svctm %util /dev/sda5 0.01 3.32 0.01 0.68 0.16 32.96 0.08 16.48 48.61 0.09 12.16 2.01 0.14 /dev/sdb2 0.00 6.38 0.00 3.54 0.01 79.36 0.00 39.68 22.39 0.12 3.52 1.02 0.36 /dev/sdc1 0.03 0.13 0.00 0.08 0.27 1.69 0.13 0.84 24.06 0.13 163.28 13.75 0.11 /dev/sdd1 0.01 8.67 0.00 0.77 0.06 82.35 0.03 41.18 107.54 0.09 10.51 2.76 0.21 The /sdc1's await seems awfully long compared to the rest to the stats. Jelle -- http://www.sv650.org/audiovisual/loading_a_bike.mpeg Osama-in-October office pool. From pgsql-performance-owner@postgresql.org Thu Oct 21 03:25:25 2004 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 B35CC32ADC5 for ; Thu, 21 Oct 2004 03:25:24 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 07802-02 for ; Thu, 21 Oct 2004 02:25:16 +0000 (GMT) Received: from mail.ge-ts.com.hk (unknown [202.66.115.131]) by svr1.postgresql.org (Postfix) with ESMTP id 5B05932A4B0 for ; Thu, 21 Oct 2004 03:25:17 +0100 (BST) Received: from raysiu ([172.16.96.15]) by mail.ge-ts.com.hk (8.12.8/8.12.8) with SMTP id i9L2PGCh013233 for ; Thu, 21 Oct 2004 10:25:17 +0800 Message-ID: <069201c4b715$34851670$0f6010ac@raysiu> From: "Ray" To: Subject: create index with substr function Date: Thu, 21 Oct 2004 10:25:17 +0800 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_068F_01C4B758.4260C620" 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=0.5 tagged_above=0.0 required=5.0 tests=HTML_40_50, HTML_MESSAGE X-Spam-Level: X-Archive-Number: 200410/333 X-Sequence-Number: 8809 This is a multi-part message in MIME format. ------=_NextPart_000_068F_01C4B758.4260C620 Content-Type: text/plain; charset="big5" Content-Transfer-Encoding: quoted-printable Hi All, I have a table in my postgres: Table: doc Column | Type | Modifiers=20 ---------------+-----------------------------+----------- doc_id | bigint | not null comp_grp_id | bigint | not null doc_type | character varying(10)| not null doc_urn | character varying(20)| not null I want to create an index on doc_urn column with using substr function like= this: CREATE INDEX idx_doc_substr_doc_urn ON doc USING btree (SUBSTR(doc_urn,10)); but there is an error: ERROR: parser: parse error at or near "10" at character 68 what's wrong for this SQL? As I have found some reference on the internet, = I can't find anything wrong in this SQL. Thanks Ray= ------=_NextPart_000_068F_01C4B758.4260C620 Content-Type: text/html; charset="big5" Content-Transfer-Encoding: quoted-printable
Hi All,
 
I have a table in my postgres:
Table: doc
    =20 Column    =20 |           =20 Type            = ; |=20 Modifiers
    =20 ---------------+-----------------------------+-----------
 doc_id&n= bsp;         |=20 bigint           &nb= sp;         =20 | not null
 comp_grp_id |=20 bigint           &nb= sp;         =20 | not null
 doc_type      | character= =20 varying(10)| not null
 doc_urn      &= nbsp;=20 | character varying(20)| not null
I want to create an index on doc_urn colum= n with=20 using substr function like this:
CREATE INDEX idx_doc_substr_doc_urn ON doc= USING=20 btree (SUBSTR(doc_urn,10));
 
but there is an error:

ERROR:  parser: parse error at or near "10" at character 68
 
what's wrong for this SQL? As I have found some reference on the inter= net,=20 I can't find anything wrong in this SQL.
 
Thanks
Ray
 ------=_NextPart_000_068F_01C4B758.4260C620-- From pgsql-performance-owner@postgresql.org Thu Oct 21 06:12:22 2004 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 7757D32B1E5 for ; Thu, 21 Oct 2004 06:12:20 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 61307-02 for ; Thu, 21 Oct 2004 05:12:16 +0000 (GMT) Received: from svr4.postgresql.org (svr4.postgresql.org [66.98.251.159]) by svr1.postgresql.org (Postfix) with ESMTP id 4C3FB32B18C for ; Thu, 21 Oct 2004 06:11:39 +0100 (BST) Received: from megazone.bigpanda.com (megazone.bigpanda.com [64.147.171.210]) by svr4.postgresql.org (Postfix) with ESMTP id 857515AF563 for ; Thu, 21 Oct 2004 03:05:30 +0000 (GMT) Received: by megazone.bigpanda.com (Postfix, from userid 1001) id 2B4F235B1F; Wed, 20 Oct 2004 19:57:54 -0700 (PDT) Received: from localhost (localhost [127.0.0.1]) by megazone.bigpanda.com (Postfix) with ESMTP id 29D1C35AF8; Wed, 20 Oct 2004 19:57:54 -0700 (PDT) Date: Wed, 20 Oct 2004 19:57:54 -0700 (PDT) From: Stephan Szabo To: Ray Cc: pgsql-performance@postgresql.org Subject: Re: create index with substr function In-Reply-To: <069201c4b715$34851670$0f6010ac@raysiu> Message-ID: <20041020195604.T32588@megazone.bigpanda.com> References: <069201c4b715$34851670$0f6010ac@raysiu> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/335 X-Sequence-Number: 8811 On Thu, 21 Oct 2004, Ray wrote: > Hi All, > > I have a table in my postgres: > Table: doc > Column | Type | Modifiers > ---------------+-----------------------------+----------- > doc_id | bigint | not null > comp_grp_id | bigint | not null > doc_type | character varying(10)| not null > doc_urn | character varying(20)| not null > > I want to create an index on doc_urn column with using substr function like this: > CREATE INDEX idx_doc_substr_doc_urn ON doc USING btree (SUBSTR(doc_urn,10)); > > but there is an error: > > ERROR: parser: parse error at or near "10" at character 68 > > what's wrong for this SQL? As I have found some reference on the > internet, I can't find anything wrong in this SQL. What version are you using? If you're using anything previous to 7.4 then the above definately won't work and the only work around I know of is to make another function which takes only the column argument and calls substr with the 10 constant. From pgsql-performance-owner@postgresql.org Thu Oct 21 06:11:40 2004 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 4393632B10F for ; Thu, 21 Oct 2004 06:11:38 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 58206-07 for ; Thu, 21 Oct 2004 05:11:35 +0000 (GMT) Received: from svr4.postgresql.org (svr4.postgresql.org [66.98.251.159]) by svr1.postgresql.org (Postfix) with ESMTP id 3C14C32B0DF for ; Thu, 21 Oct 2004 06:11:31 +0100 (BST) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr4.postgresql.org (Postfix) with ESMTP id 119D65AF423 for ; Thu, 21 Oct 2004 03:01:49 +0000 (GMT) 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 i9L2xEMF010775; Wed, 20 Oct 2004 22:59:14 -0400 (EDT) To: "Ray" Cc: pgsql-performance@postgresql.org Subject: Re: create index with substr function In-reply-to: <069201c4b715$34851670$0f6010ac@raysiu> References: <069201c4b715$34851670$0f6010ac@raysiu> Comments: In-reply-to "Ray" message dated "Thu, 21 Oct 2004 10:25:17 +0800" Date: Wed, 20 Oct 2004 22:59:14 -0400 Message-ID: <10774.1098327554@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/334 X-Sequence-Number: 8810 "Ray" writes: > CREATE INDEX idx_doc_substr_doc_urn ON doc USING btree (SUBSTR(doc_urn,10)); > ERROR: parser: parse error at or near "10" at character 68 This will work in 7.4, but not older releases. regards, tom lane From pgsql-performance-owner@postgresql.org Thu Oct 21 06:12:22 2004 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 44D0432B1EA for ; Thu, 21 Oct 2004 06:12:21 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 61057-03 for ; Thu, 21 Oct 2004 05:12:15 +0000 (GMT) Received: from svr4.postgresql.org (svr4.postgresql.org [66.98.251.159]) by svr1.postgresql.org (Postfix) with ESMTP id 3769E32B1FB for ; Thu, 21 Oct 2004 06:11:41 +0100 (BST) Received: from mail.ge-ts.com.hk (unknown [202.66.115.131]) by svr4.postgresql.org (Postfix) with ESMTP id 9E8135AF5FC for ; Thu, 21 Oct 2004 03:13:08 +0000 (GMT) Received: from raysiu ([172.16.96.15]) by mail.ge-ts.com.hk (8.12.8/8.12.8) with SMTP id i9L3B5Ch013400 for ; Thu, 21 Oct 2004 11:11:05 +0800 Message-ID: <06e201c4b71b$9ae9cc70$0f6010ac@raysiu> From: "Ray" To: References: <069201c4b715$34851670$0f6010ac@raysiu> <20041020195604.T32588@megazone.bigpanda.com> Subject: Re: create index with substr function Date: Thu, 21 Oct 2004 11:11:05 +0800 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 6.00.2800.1437 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/336 X-Sequence-Number: 8812 Thank you all kindly response..... : ) I am currently using postgres 7.3, so any example or solution for version after 7.4 if i want to create an index with substr function??? Thanks, Ray ----- Original Message ----- From: "Stephan Szabo" To: "Ray" Cc: Sent: Thursday, October 21, 2004 10:57 AM Subject: Re: [PERFORM] create index with substr function > > On Thu, 21 Oct 2004, Ray wrote: > > > Hi All, > > > > I have a table in my postgres: > > Table: doc > > Column | Type | Modifiers > > ---------------+-----------------------------+----------- > > doc_id | bigint | not null > > comp_grp_id | bigint | not null > > doc_type | character varying(10)| not null > > doc_urn | character varying(20)| not null > > > > I want to create an index on doc_urn column with using substr function like this: > > CREATE INDEX idx_doc_substr_doc_urn ON doc USING btree (SUBSTR(doc_urn,10)); > > > > but there is an error: > > > > ERROR: parser: parse error at or near "10" at character 68 > > > > what's wrong for this SQL? As I have found some reference on the > > internet, I can't find anything wrong in this SQL. > > What version are you using? If you're using anything previous to 7.4 then > the above definately won't work and the only work around I know of is to > make another function which takes only the column argument and calls > substr with the 10 constant. > From pgsql-performance-owner@postgresql.org Thu Oct 21 06:44:53 2004 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 B6FFF32AE15 for ; Thu, 21 Oct 2004 06:44:51 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 70432-03 for ; Thu, 21 Oct 2004 05:44:48 +0000 (GMT) Received: from svr4.postgresql.org (svr4.postgresql.org [66.98.251.159]) by svr1.postgresql.org (Postfix) with ESMTP id EFCF932A7AD for ; Thu, 21 Oct 2004 06:44:48 +0100 (BST) Received: from mproxy.gmail.com (rproxy.gmail.com [64.233.170.203]) by svr4.postgresql.org (Postfix) with ESMTP id 081775AF51C for ; Thu, 21 Oct 2004 03:36:55 +0000 (GMT) Received: by mproxy.gmail.com with SMTP id 73so586619rnk for ; Wed, 20 Oct 2004 20:34:55 -0700 (PDT) DomainKey-Signature: a=rsa-sha1; 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:references; b=fMzchdGnZQLA9E1hUQqXxc18YROGwM8N3BApSyw2dQh6JOQNwBUPc/WY9wj2HPlq0ZL6ITUORNQ01zbjvnU9eM+jaQbm8KgSstH0vWquZb+XrJQ8zC/39P57MW3d1i57LuDsROQjoMDTfRuKJ+LrqVxT9vkN2r8egqjnWlkrWsg Received: by 10.39.1.26 with SMTP id d26mr2874209rni; Wed, 20 Oct 2004 20:34:55 -0700 (PDT) Received: by 10.38.76.43 with HTTP; Wed, 20 Oct 2004 20:34:55 -0700 (PDT) Message-ID: <37d451f704102020346fdbb855@mail.gmail.com> Date: Wed, 20 Oct 2004 22:34:55 -0500 From: Rosser Schwarz Reply-To: Rosser Schwarz To: Ray Subject: Re: create index with substr function Cc: pgsql-performance@postgresql.org In-Reply-To: <069201c4b715$34851670$0f6010ac@raysiu> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: <069201c4b715$34851670$0f6010ac@raysiu> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.3 tagged_above=0.0 required=5.0 tests=UPPERCASE_25_50 X-Spam-Level: X-Archive-Number: 200410/337 X-Sequence-Number: 8813 while you weren't looking, Ray wrote: > CREATE INDEX idx_doc_substr_doc_urn ON doc USING btree (SUBSTR(doc_urn,10)); CREATE INDEX idx_doc_substr_doc_urn ON doc USING btree ((SUBSTR(doc_urn,10))); You need an additional set of parens around the SUBSTR() call. /rls -- :wq From pgsql-performance-owner@postgresql.org Thu Oct 21 06:45:24 2004 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 DA1AC32AE59 for ; Thu, 21 Oct 2004 06:45:14 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 71154-02 for ; Thu, 21 Oct 2004 05:45:09 +0000 (GMT) Received: from svr4.postgresql.org (svr4.postgresql.org [66.98.251.159]) by svr1.postgresql.org (Postfix) with ESMTP id EAA3E32AE41 for ; Thu, 21 Oct 2004 06:44:56 +0100 (BST) Received: from mail.ge-ts.com.hk (unknown [202.66.115.131]) by svr4.postgresql.org (Postfix) with ESMTP id B08555AF5D5 for ; Thu, 21 Oct 2004 03:40:56 +0000 (GMT) Received: from raysiu ([172.16.96.15]) by mail.ge-ts.com.hk (8.12.8/8.12.8) with SMTP id i9L3bPCh013589; Thu, 21 Oct 2004 11:37:26 +0800 Message-ID: <078b01c4b71f$4936e4e0$0f6010ac@raysiu> From: "Ray" To: "Rosser Schwarz" Cc: References: <069201c4b715$34851670$0f6010ac@raysiu> <37d451f704102020346fdbb855@mail.gmail.com> Subject: Re: create index with substr function Date: Thu, 21 Oct 2004 11:37:26 +0800 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 6.00.2800.1437 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/338 X-Sequence-Number: 8814 sorry it doesn't works, as my postgres is 7.3 not 7.4. any other alternative solution for version after 7.4?? Thank Ray : ) ----- Original Message ----- From: "Rosser Schwarz" To: "Ray" Cc: Sent: Thursday, October 21, 2004 11:34 AM Subject: Re: [PERFORM] create index with substr function > while you weren't looking, Ray wrote: > > > CREATE INDEX idx_doc_substr_doc_urn ON doc USING btree (SUBSTR(doc_urn,10)); > > CREATE INDEX idx_doc_substr_doc_urn ON doc USING btree ((SUBSTR(doc_urn,10))); > > You need an additional set of parens around the SUBSTR() call. > > /rls > > -- > :wq > From pgsql-performance-owner@postgresql.org Mon Oct 25 16:57:49 2004 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 E752032AEF5 for ; Thu, 21 Oct 2004 06:46:18 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 71477-07 for ; Thu, 21 Oct 2004 05:46:14 +0000 (GMT) Received: from dbl.q-ag.de (dbl.q-ag.de [213.172.117.3]) by svr1.postgresql.org (Postfix) with ESMTP id D716732AE2A for ; Thu, 21 Oct 2004 06:46:09 +0100 (BST) Received: from [127.0.0.2] (dbl [127.0.0.1]) by dbl.q-ag.de (8.12.3/8.12.3/Debian-6.6) with ESMTP id i9L5jsSL029502; Thu, 21 Oct 2004 07:45:55 +0200 Message-ID: <41774D11.1030906@colorfullife.com> Date: Thu, 21 Oct 2004 07:45:53 +0200 From: Manfred Spraul User-Agent: Mozilla/5.0 (X11; U; Linux i686; fr-FR; rv:1.7.3) Gecko/20040922 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Mark Wong Cc: Tom Lane , neilc@samurai.com, pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> <417697A5.1050600@colorfullife.com> <7178.1098292535@sss.pgh.pa.us> <4176A2C1.3070205@colorfullife.com> <20041020150528.B7838@osdl.org> In-Reply-To: <20041020150528.B7838@osdl.org> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/412 X-Sequence-Number: 8888 Mark Wong wrote: >Here are some other details, per Manfred's request: > >Linux 2.6.8.1 (on a gentoo distro) > > How complicated are Tom's test scripts? His immediate reply was that I should retest with Fedora, to rule out any gentoo bugs. I have a dual-cpu system with RH FC, I could use it for testing. -- Manfred From pgsql-performance-owner@postgresql.org Thu Oct 21 07:19:05 2004 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 6E5E832AD91 for ; Thu, 21 Oct 2004 07:19:04 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 80409-02 for ; Thu, 21 Oct 2004 06:18:59 +0000 (GMT) Received: from hosting.commandprompt.com (128.commandprompt.com [207.173.200.128]) by svr1.postgresql.org (Postfix) with ESMTP id DE07132A849 for ; Thu, 21 Oct 2004 07:18:59 +0100 (BST) Received: from [192.168.1.20] (clbb-248.saw.net [64.146.135.248]) (authenticated) by hosting.commandprompt.com (8.11.6/8.11.6) with ESMTP id i9L6FRD07936; Wed, 20 Oct 2004 23:15:27 -0700 Message-ID: <4177538C.7010900@commandprompt.com> Date: Wed, 20 Oct 2004 23:13:32 -0700 From: "Joshua D. Drake" Organization: Command Prompt, Inc. User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: Ray , pgsql-performance@postgresql.org Subject: Re: create index with substr function References: <069201c4b715$34851670$0f6010ac@raysiu> <10774.1098327554@sss.pgh.pa.us> In-Reply-To: <10774.1098327554@sss.pgh.pa.us> Content-Type: multipart/alternative; boundary="------------080507080403000504090501" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.0 tagged_above=0.0 required=5.0 tests=HTML_20_30, HTML_MESSAGE, HTML_TITLE_EMPTY X-Spam-Level: * X-Archive-Number: 200410/339 X-Sequence-Number: 8815 This is a multi-part message in MIME format. --------------080507080403000504090501 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Tom Lane wrote: >"Ray" writes: > > >>CREATE INDEX idx_doc_substr_doc_urn ON doc USING btree (SUBSTR(doc_urn,10)); >>ERROR: parser: parse error at or near "10" at character 68 >> >> > >This will work in 7.4, but not older releases. > > > Can't you just use a SQL function that calls the substr function? I have done that with date functions before like: CREATE OR REPLACE FUNCTION get_month(text) returns double precision AS ' SELECT date_part('month',$1); ' LANGUAGE 'SQL' IMMUTABLE; CREATE INDEX get_month_idx on foo(get_month(date_field)); Or in this case: CREATE OR REPLACE FUNCTION sub_text(text) returns text AS ' SELECT SUBSTR($1,10) from foo; ' LANGUAGE 'SQL' IMMUTABLE; CREATE INDEX sub_text_idx ON foo(sub_text(doc_urn)); This works on 7.3.6??? Sincerely, Joshua D. Drake > regards, tom lane > >---------------------------(end of broadcast)--------------------------- >TIP 7: don't forget to increase your free space map settings > > -- Command Prompt, Inc., home of Mammoth PostgreSQL - S/ODBC and S/JDBC Postgresql support, programming shared hosting and dedicated hosting. +1-503-667-4564 - jd@commandprompt.com - http://www.commandprompt.com PostgreSQL Replicator -- production quality replication for PostgreSQL --------------080507080403000504090501 Content-Type: text/html; charset=us-ascii Content-Transfer-Encoding: 7bit Tom Lane wrote:
"Ray" <ray_siu@ge-ts.com.hk> writes:
  
CREATE INDEX idx_doc_substr_doc_urn ON doc USING btree (SUBSTR(doc_urn,10));
ERROR:  parser: parse error at or near "10" at character 68
    

This will work in 7.4, but not older releases.
Can't you just use a SQL function that calls the substr function? I have done that with date functions before
like:
CREATE OR REPLACE FUNCTION get_month(text) returns double precision AS '
   SELECT date_part('month',$1);
' LANGUAGE 'SQL' IMMUTABLE;

CREATE INDEX get_month_idx on foo(get_month(date_field));
Or in this case:

CREATE OR REPLACE FUNCTION sub_text(text) returns text AS '
      SELECT SUBSTR($1,10) from foo;
' LANGUAGE 'SQL' IMMUTABLE;

CREATE INDEX sub_text_idx ON foo(sub_text(doc_urn));

This works on 7.3.6???

Sincerely,

Joshua D. Drake




			regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 7: don't forget to increase your free space map settings
  


-- 
Command Prompt, Inc., home of Mammoth PostgreSQL - S/ODBC and S/JDBC
Postgresql support, programming shared hosting and dedicated hosting.
+1-503-667-4564 - jd@commandprompt.com - http://www.commandprompt.com
PostgreSQL Replicator -- production quality replication for PostgreSQL
--------------080507080403000504090501-- From pgsql-performance-owner@postgresql.org Thu Oct 21 08:58:23 2004 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 A292B32AE3A for ; Thu, 21 Oct 2004 08:58:22 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 12997-07 for ; Thu, 21 Oct 2004 07:58:03 +0000 (GMT) Received: from bayswater1.ymogen.net (host-154-240-27-217.pobox.net.uk [217.27.240.154]) by svr1.postgresql.org (Postfix) with ESMTP id 8FA0332A0C3 for ; Thu, 21 Oct 2004 08:58:03 +0100 (BST) Received: from solent (unknown [213.165.136.10]) by bayswater1.ymogen.net (Postfix) with ESMTP id 80325A39CA for ; Thu, 21 Oct 2004 08:58:01 +0100 (BST) Reply-To: From: "Matt Clark" To: Subject: Anything to be gained from a 'Postgres Filesystem'? Date: Thu, 21 Oct 2004 08:58:01 +0100 Organization: Ymogen Ltd Message-ID: <008601c4b743$aff57f10$8300a8c0@solent> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook, Build 10.0.6626 X-MIMEOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 Importance: Normal X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/340 X-Sequence-Number: 8816 I suppose I'm just idly wondering really. Clearly it's against PG philosophy to build an FS or direct IO management into PG, but now it's so relatively easy to plug filesystems into the main open-source Oses, It struck me that there might be some useful changes to, say, XFS or ext3, that could be made that would help PG out. I'm thinking along the lines of an FS that's aware of PG's strategies and requirements and therefore optimised to make those activities as efiicient as possible - possibly even being aware of PG's disk layout and treating files differently on that basis. Not being an FS guru I'm not really clear on whether this would help much (enough to be worth it anyway) or not - any thoughts? And if there were useful gains to be had, would it need a whole new FS or could an existing one be modified? So there might be (as I said, I'm not an FS guru...): * great append performance for the WAL? * optimised scattered writes for checkpointing? * Knowledge that FSYNC is being used for preserving ordering a lot of the time, rather than requiring actual writes to disk (so long as the writes eventually happen in order...)? Matt Matt Clark Ymogen Ltd P: 0845 130 4531 W: https://ymogen.net/ M: 0774 870 1584 =20 From pgsql-performance-owner@postgresql.org Thu Oct 21 09:24:31 2004 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 5F221329E7D for ; Thu, 21 Oct 2004 09:24:26 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 20410-10 for ; Thu, 21 Oct 2004 08:24:05 +0000 (GMT) Received: from mailer.unicite.fr.netcentrex.net (unknown [62.161.167.249]) by svr1.postgresql.org (Postfix) with ESMTP id 7275F329E5C for ; Thu, 21 Oct 2004 09:24:05 +0100 (BST) Received: from akira ([192.168.101.228]) by mailer.unicite.fr.netcentrex.net with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2657.72) id 42CYASGM; Thu, 21 Oct 2004 10:24:04 +0200 From: "Alban Medici (NetCentrex)" To: Subject: Re: Free PostgreSQL Training, Philadelphia, Oct 30 Date: Thu, 21 Oct 2004 10:24:04 +0200 Organization: NetCentrex MIME-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit X-Mailer: Microsoft Office Outlook, Build 11.0.6353 In-Reply-To: X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 Thread-Index: AcS2pMduY5IzJ7+sSdSL6SqR98shiQAomPOg Message-Id: <20041021082405.7275F329E5C@svr1.postgresql.org> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/341 X-Sequence-Number: 8817 Nobody got a plane to came from europe :-) ??? As a poor frenchie I will not come ... Have a good time Alban -----Original Message----- From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of Aaron Mulder Sent: mercredi 20 octobre 2004 15:11 To: pgsql-performance@postgresql.org Subject: Re: [PERFORM] Free PostgreSQL Training, Philadelphia, Oct 30 If anyone is going to take the train all the way, please e-mail me offline. There is a train station relatively close to the event (NY to Philly then the R5 to Malvern), but it's not within walking distance, so we'll figure out some way to pick people up from there. Thanks, Aaron On Wed, 20 Oct 2004, Aaron Werman wrote: > I'm driving from Tenafly NJ and going to both sessions. If you're able > to get to the George Washington Bridge (A train to 178th Street [Port > Authority North] and a bus over the bridge), I can drive you down. I'm > not sure right now about the return because I have confused plans to > meet someone. > > /Aaron > > > On Tue, 19 Oct 2004 14:43:29 -0400, Max Baker wrote: > > On Wed, Oct 13, 2004 at 12:21:27PM -0400, Aaron Mulder wrote: > > > All, > > > My company (Chariot Solutions) is sponsoring a day of free > > > PostgreSQL training by Bruce Momjian (one of the core PostgreSQL > > > developers). The day is split into 2 sessions (plus a Q&A session): > > > > > > * Mastering PostgreSQL Administration > > > * PostgreSQL Performance Tuning > > > > > > Registration is required, and space is limited. The > > > location is Malvern, PA (suburb of Philadelphia) and it's on > > > Saturday Oct 30. For more information or to register, see > > > > > > http://chariotsolutions.com/postgresql.jsp > > > > I'm up in New York City and would be taking the train down to > > Philly. Is anyone coming from Philly or New York that would be able > > to give me a lift to/from the train station? Sounds like a great event. > > > > Cheers, > > -m > > > > ---------------------------(end of > > broadcast)--------------------------- > > TIP 3: 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 > > > > > -- > > Regards, > /Aaron > > ---------------------------(end of > broadcast)--------------------------- > TIP 5: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faqs/FAQ.html > ---------------------------(end of broadcast)--------------------------- TIP 6: Have you searched our list archives? http://archives.postgresql.org From pgsql-performance-owner@postgresql.org Thu Oct 21 09:27:46 2004 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 ACBB032A056 for ; Thu, 21 Oct 2004 09:27:44 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 24856-02 for ; Thu, 21 Oct 2004 08:27:41 +0000 (GMT) Received: from usbb-lacimss2.unisys.com (usbb-lacimss2.unisys.com [192.63.108.52]) by svr1.postgresql.org (Postfix) with ESMTP id 09B81329FB2 for ; Thu, 21 Oct 2004 09:27:41 +0100 (BST) Received: from usbb-lacgw2.lac.uis.unisys.com ([129.226.160.22]unverified) by usbb-lacimss2 with InterScan Messaging Security Suite; Thu, 21 Oct 2004 04:30:55 -0400 Received: from usbb-lacgw2.lac.uis.unisys.com ([129.226.160.25]) by usbb-lacgw2.lac.uis.unisys.com with Microsoft SMTPSVC(6.0.3790.211); Thu, 21 Oct 2004 04:27:40 -0400 Received: from gbmk-eugw2.eu.uis.unisys.com ([129.221.133.27]) by usbb-lacgw2.lac.uis.unisys.com with Microsoft SMTPSVC(6.0.3790.211); Thu, 21 Oct 2004 04:27:39 -0400 Received: from nlshl-exch1.eu.uis.unisys.com ([192.39.239.20]) by gbmk-eugw2.eu.uis.unisys.com with Microsoft SMTPSVC(6.0.3790.47); Thu, 21 Oct 2004 09:27:24 +0100 Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: Anything to be gained from a 'Postgres Filesystem'? Date: Thu, 21 Oct 2004 10:27:22 +0200 Message-ID: X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Anything to be gained from a 'Postgres Filesystem'? Thread-Index: AcS3RRCYZb82zgd2TRuDst9A9MylTQAAmKLw From: "Leeuw van der, Tim" To: , X-OriginalArrivalTime: 21 Oct 2004 08:27:24.0229 (UTC) FILETIME=[CA912F50:01C4B747] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/342 X-Sequence-Number: 8818 Hiya, Looking at that list, I got the feeling that you'd want to push that PG-awa= reness down into the block-io layer as well, then, so as to be able to opti= mise for (perhaps) conflicting goals depending on what the app does; for th= e IO system to be able to read the apps mind it needs to have some knowledg= e of what the app is / needs / wants and I get the impression that this awa= reness needs to go deeper than the FS only. --Tim (But you might have time to rewrite Linux/BSD as a PG-OS? just kidding!) -----Original Message----- From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owne= r@postgresql.org]On Behalf Of Matt Clark Sent: Thursday, October 21, 2004 9:58 AM To: pgsql-performance@postgresql.org Subject: [PERFORM] Anything to be gained from a 'Postgres Filesystem'? I suppose I'm just idly wondering really. Clearly it's against PG philosophy to build an FS or direct IO management into PG, but now it's so relatively easy to plug filesystems into the main open-source Oses, It struck me that there might be some useful changes to, say, XFS or ext3, that could be made that would help PG out. I'm thinking along the lines of an FS that's aware of PG's strategies and requirements and therefore optimised to make those activities as efiicient as possible - possibly even being aware of PG's disk layout and treating files differently on that basis. Not being an FS guru I'm not really clear on whether this would help much (enough to be worth it anyway) or not - any thoughts? And if there were useful gains to be had, would it need a whole new FS or could an existing one be modified? So there might be (as I said, I'm not an FS guru...): * great append performance for the WAL? * optimised scattered writes for checkpointing? * Knowledge that FSYNC is being used for preserving ordering a lot of the time, rather than requiring actual writes to disk (so long as the writes eventually happen in order...)? Matt Matt Clark Ymogen Ltd P: 0845 130 4531 W: https://ymogen.net/ M: 0774 870 1584 =20 ---------------------------(end of broadcast)--------------------------- TIP 2: you can get off all lists at once with the unregister command (send "unregister YourEmailAddressHere" to majordomo@postgresql.org) From pgsql-performance-owner@postgresql.org Thu Oct 21 09:32:52 2004 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 EACCA32A03C for ; Thu, 21 Oct 2004 09:32:50 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 27321-03 for ; Thu, 21 Oct 2004 08:32:45 +0000 (GMT) Received: from boutiquenumerique.com (boutiquenumerique.com [82.67.9.10]) by svr1.postgresql.org (Postfix) with ESMTP id 0497732A026 for ; Thu, 21 Oct 2004 09:32:46 +0100 (BST) Received: (qmail 6706 invoked from network); 21 Oct 2004 10:32:47 +0200 Received: from unknown (HELO musicbox) (boutiquenumerique-lists@192.168.0.2) by boutiquenumerique.com with SMTP; 21 Oct 2004 10:32:47 +0200 To: matt@ymogen.net, pgsql-performance@postgresql.org Subject: Re: Anything to be gained from a 'Postgres Filesystem'? References: <008601c4b743$aff57f10$8300a8c0@solent> Message-ID: From: =?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= Organization: =?iso-8859-15?Q?La_Boutique_Num=E9rique?= Content-Type: text/plain; format=flowed; delsp=yes; charset=iso-8859-15 MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Date: Thu, 21 Oct 2004 10:33:31 +0200 In-Reply-To: <008601c4b743$aff57f10$8300a8c0@solent> User-Agent: Opera M2/7.54 (Linux, build 751) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/343 X-Sequence-Number: 8819 Reiser4 ? On Thu, 21 Oct 2004 08:58:01 +0100, Matt Clark wrote: > I suppose I'm just idly wondering really. Clearly it's against PG > philosophy to build an FS or direct IO management into PG, but now it's > so > relatively easy to plug filesystems into the main open-source Oses, It > struck me that there might be some useful changes to, say, XFS or ext3, > that > could be made that would help PG out. > > I'm thinking along the lines of an FS that's aware of PG's strategies and > requirements and therefore optimised to make those activities as > efiicient > as possible - possibly even being aware of PG's disk layout and treating > files differently on that basis. > > Not being an FS guru I'm not really clear on whether this would help much > (enough to be worth it anyway) or not - any thoughts? And if there were > useful gains to be had, would it need a whole new FS or could an existing > one be modified? > > So there might be (as I said, I'm not an FS guru...): > * great append performance for the WAL? > * optimised scattered writes for checkpointing? > * Knowledge that FSYNC is being used for preserving ordering a lot of the > time, rather than requiring actual writes to disk (so long as the writes > eventually happen in order...)? > > > Matt > > > > Matt Clark > Ymogen Ltd > P: 0845 130 4531 > W: https://ymogen.net/ > M: 0774 870 1584 > > > ---------------------------(end of broadcast)--------------------------- > TIP 2: you can get off all lists at once with the unregister command > (send "unregister YourEmailAddressHere" to majordomo@postgresql.org) > From pgsql-performance-owner@postgresql.org Thu Oct 21 09:38:53 2004 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 7BE3D329E65 for ; Thu, 21 Oct 2004 09:38:52 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 25984-08 for ; Thu, 21 Oct 2004 08:38:41 +0000 (GMT) Received: from bayswater1.ymogen.net (host-154-240-27-217.pobox.net.uk [217.27.240.154]) by svr1.postgresql.org (Postfix) with ESMTP id 9B53F32A0F7 for ; Thu, 21 Oct 2004 09:38:41 +0100 (BST) Received: from solent (unknown [213.165.136.10]) by bayswater1.ymogen.net (Postfix) with ESMTP id 029CBA37B2; Thu, 21 Oct 2004 09:38:41 +0100 (BST) Reply-To: From: "Matt Clark" To: "'Leeuw van der, Tim'" , Subject: Re: Anything to be gained from a 'Postgres Filesystem'? Date: Thu, 21 Oct 2004 09:38:40 +0100 Organization: Ymogen Ltd Message-ID: <008b01c4b749$5e0862c0$8300a8c0@solent> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook, Build 10.0.6626 X-MIMEOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 In-Reply-To: Importance: Normal X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/344 X-Sequence-Number: 8820 > Looking at that list, I got the feeling that you'd want to > push that PG-awareness down into the block-io layer as well, > then, so as to be able to optimise for (perhaps) conflicting > goals depending on what the app does; for the IO system to be > able to read the apps mind it needs to have some knowledge of > what the app is / needs / wants and I get the impression that > this awareness needs to go deeper than the FS only. That's a fair point, it would need be a kernel patch really, although not necessarily a very big one, more a case of looking at FDs and if they're flagged in some way then get the PGfs to do the job instead of/as well as the normal code path. From pgsql-performance-owner@postgresql.org Thu Oct 21 11:27:32 2004 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 40ACA32A10E for ; Thu, 21 Oct 2004 11:27:32 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 55557-09 for ; Thu, 21 Oct 2004 10:27:28 +0000 (GMT) Received: from trofast.sesse.net (trofast.sesse.net [129.241.93.32]) by svr1.postgresql.org (Postfix) with ESMTP id C442932A0A1 for ; Thu, 21 Oct 2004 11:27:29 +0100 (BST) Received: from sesse by trofast.sesse.net with local (Exim 3.36 #1 (Debian)) id 1CKaAJ-0000BO-00 for ; Thu, 21 Oct 2004 12:27:27 +0200 Date: Thu, 21 Oct 2004 12:27:27 +0200 From: "Steinar H. Gunderson" To: pgsql-performance@postgresql.org Subject: Re: Anything to be gained from a 'Postgres Filesystem'? Message-ID: <20041021102727.GB586@uio.no> Mail-Followup-To: pgsql-performance@postgresql.org References: <008601c4b743$aff57f10$8300a8c0@solent> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <008601c4b743$aff57f10$8300a8c0@solent> X-Operating-System: Linux 2.6.8.1 on a i686 User-Agent: Mutt/1.5.6+20040907i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/345 X-Sequence-Number: 8821 On Thu, Oct 21, 2004 at 08:58:01AM +0100, Matt Clark wrote: > I suppose I'm just idly wondering really. Clearly it's against PG > philosophy to build an FS or direct IO management into PG, but now it's so > relatively easy to plug filesystems into the main open-source Oses, It > struck me that there might be some useful changes to, say, XFS or ext3, that > could be made that would help PG out. This really sounds like a poor replacement for just making PostgreSQL use raw devices to me. (I have no idea why that isn't done already, but presumably it isn't all that easy to get right. :-) ) /* Steinar */ -- Homepage: http://www.sesse.net/ From pgsql-performance-owner@postgresql.org Thu Oct 21 11:44:29 2004 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 6B3A632A10E for ; Thu, 21 Oct 2004 11:44:29 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 58400-10 for ; Thu, 21 Oct 2004 10:44:25 +0000 (GMT) Received: from usbb-lacimss3.unisys.com (usbb-lacimss3.unisys.com [192.63.108.53]) by svr1.postgresql.org (Postfix) with ESMTP id C2BF932A0E5 for ; Thu, 21 Oct 2004 11:44:24 +0100 (BST) Received: from USBB-LACGW3.na.uis.unisys.com ([129.224.98.43]unverified) by usbb-lacimss3 with InterScan Messaging Security Suite; Thu, 21 Oct 2004 06:48:08 -0400 Received: from USBB-LACGW3.na.uis.unisys.com ([129.224.98.44]) by USBB-LACGW3.na.uis.unisys.com with Microsoft SMTPSVC(6.0.3790.211); Thu, 21 Oct 2004 06:44:13 -0400 Received: from gbmk-eugw1.eu.uis.unisys.com ([129.221.133.28]) by USBB-LACGW3.na.uis.unisys.com with Microsoft SMTPSVC(6.0.3790.211); Thu, 21 Oct 2004 06:44:13 -0400 Received: from nlshl-exch1.eu.uis.unisys.com ([192.39.239.20]) by gbmk-eugw1.eu.uis.unisys.com with Microsoft SMTPSVC(6.0.3790.47); Thu, 21 Oct 2004 11:44:11 +0100 Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.5.7226.0 Subject: Re: Anything to be gained from a 'Postgres Filesystem'? Date: Thu, 21 Oct 2004 12:44:10 +0200 Message-ID: X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Anything to be gained from a 'Postgres Filesystem'? Thread-Index: AcS3WWQJaLJXjRM2RjW9o2B6un6TMAAAFA8g From: "Leeuw van der, Tim" To: "Steinar H. Gunderson" , X-OriginalArrivalTime: 21 Oct 2004 10:44:11.0592 (UTC) FILETIME=[E6896480:01C4B75A] X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/346 X-Sequence-Number: 8822 Hi, I guess the difference is in 'severe hacking inside PG' vs. 'some unknown a= mount of hacking that doesn't touch PG code'. Hacking PG internally to handle raw devices will meet with strong resistanc= e from large portions of the development team. I don't expect (m)any core d= evs of PG will be excited about rewriting the entire I/O architecture of PG= and duplicating large amounts of OS type of code inside the application, j= ust to try to attain an unknown performance benefit. PG doesn't use one big file, as some databases do, but many small files. No= w PG would need to be able to do file-management, if you put the PG databas= e on a raw disk partition! That's icky stuff, and you'll find much resistan= ce against putting such code inside PG. So why not try to have the external FS know a bit about PG and it's directo= ry-layout, and it's IO requirements? Then such type of code can at least be= maintained outside the application, and will not be as much of a burden to= the rest of the application. (I'm not sure if it's a good idea to create a PG-specific FS in your OS of = choice, but it's certainly gonna be easier than getting FS code inside of P= G) cheers, --Tim -----Original Message----- From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owne= r@postgresql.org]On Behalf Of Steinar H. Gunderson Sent: Thursday, October 21, 2004 12:27 PM To: pgsql-performance@postgresql.org Subject: Re: [PERFORM] Anything to be gained from a 'Postgres Filesystem'? On Thu, Oct 21, 2004 at 08:58:01AM +0100, Matt Clark wrote: > I suppose I'm just idly wondering really. Clearly it's against PG > philosophy to build an FS or direct IO management into PG, but now it's so > relatively easy to plug filesystems into the main open-source Oses, It > struck me that there might be some useful changes to, say, XFS or ext3, t= hat > could be made that would help PG out. This really sounds like a poor replacement for just making PostgreSQL use r= aw devices to me. (I have no idea why that isn't done already, but presumably = it isn't all that easy to get right. :-) ) /* Steinar */ --=20 Homepage: http://www.sesse.net/ ---------------------------(end of broadcast)--------------------------- TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org From pgsql-performance-owner@postgresql.org Thu Oct 21 12:47:18 2004 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 05C3B32A165 for ; Thu, 21 Oct 2004 12:47:18 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78512-07 for ; Thu, 21 Oct 2004 11:47:13 +0000 (GMT) Received: from rproxy.gmail.com (rproxy.gmail.com [64.233.170.200]) by svr1.postgresql.org (Postfix) with ESMTP id EC72532A15F for ; Thu, 21 Oct 2004 12:47:15 +0100 (BST) Received: by rproxy.gmail.com with SMTP id 73so9646rnk for ; Thu, 21 Oct 2004 04:47:15 -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:references; b=a4IEXvhcEt0imSl9ZsbCTf8czV2q5PhOPRzhQea50sB2IOftrF7HeZjxJXhFlJqZN0uRess89681a6MFLMIDGBdc5by13poSIeAa8RVpn4R6wiRpk2dlx3PUc0NU5pi0QITqfymm+rZaTICkf5fHqUCGerAptjLoI4dh/3AiN8M= Received: by 10.39.1.26 with SMTP id d26mr3033814rni; Thu, 21 Oct 2004 04:47:14 -0700 (PDT) Received: by 10.38.151.75 with HTTP; Thu, 21 Oct 2004 04:47:14 -0700 (PDT) Message-ID: <157f6484041021044720eb7520@mail.gmail.com> Date: Thu, 21 Oct 2004 07:47:14 -0400 From: Aaron Werman Reply-To: Aaron Werman To: "Leeuw van der, Tim" Subject: Re: Anything to be gained from a 'Postgres Filesystem'? Cc: "Steinar H. Gunderson" , pgsql-performance@postgresql.org In-Reply-To: Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/347 X-Sequence-Number: 8823 The intuitive thing would be to put pg into a file system. /Aaron On Thu, 21 Oct 2004 12:44:10 +0200, Leeuw van der, Tim wrote: > Hi, > > I guess the difference is in 'severe hacking inside PG' vs. 'some unknown amount of hacking that doesn't touch PG code'. > > Hacking PG internally to handle raw devices will meet with strong resistance from large portions of the development team. I don't expect (m)any core devs of PG will be excited about rewriting the entire I/O architecture of PG and duplicating large amounts of OS type of code inside the application, just to try to attain an unknown performance benefit. > > PG doesn't use one big file, as some databases do, but many small files. Now PG would need to be able to do file-management, if you put the PG database on a raw disk partition! That's icky stuff, and you'll find much resistance against putting such code inside PG. > So why not try to have the external FS know a bit about PG and it's directory-layout, and it's IO requirements? Then such type of code can at least be maintained outside the application, and will not be as much of a burden to the rest of the application. > > (I'm not sure if it's a good idea to create a PG-specific FS in your OS of choice, but it's certainly gonna be easier than getting FS code inside of PG) > > cheers, > > --Tim > > > > -----Original Message----- > From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owner@postgresql.org]On Behalf Of Steinar H. Gunderson > Sent: Thursday, October 21, 2004 12:27 PM > To: pgsql-performance@postgresql.org > Subject: Re: [PERFORM] Anything to be gained from a 'Postgres Filesystem'? > > On Thu, Oct 21, 2004 at 08:58:01AM +0100, Matt Clark wrote: > > I suppose I'm just idly wondering really. Clearly it's against PG > > philosophy to build an FS or direct IO management into PG, but now it's so > > relatively easy to plug filesystems into the main open-source Oses, It > > struck me that there might be some useful changes to, say, XFS or ext3, that > > could be made that would help PG out. > > This really sounds like a poor replacement for just making PostgreSQL use raw > devices to me. (I have no idea why that isn't done already, but presumably it > isn't all that easy to get right. :-) ) > > /* Steinar */ > -- > Homepage: http://www.sesse.net/ > > ---------------------------(end of broadcast)--------------------------- > TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org > > ---------------------------(end of broadcast)--------------------------- > TIP 7: don't forget to increase your free space map settings > -- Regards, /Aaron From pgsql-performance-owner@postgresql.org Thu Oct 21 13:02:14 2004 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 7974A32B15F for ; Thu, 21 Oct 2004 13:02:11 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 85403-09 for ; Thu, 21 Oct 2004 12:01:58 +0000 (GMT) Received: from sue.samurai.com (sue.samurai.com [205.207.28.74]) by svr1.postgresql.org (Postfix) with ESMTP id D7C7F32B181 for ; Thu, 21 Oct 2004 13:01:58 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by sue.samurai.com (Postfix) with ESMTP id 13140197FB; Thu, 21 Oct 2004 08:01:58 -0400 (EDT) Received: from sue.samurai.com ([127.0.0.1]) by localhost (sue.samurai.com [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 56250-01-6; Thu, 21 Oct 2004 08:01:57 -0400 (EDT) Received: from [220.101.3.10] (r220-101-3-10.cpe.unwired.net.au [220.101.3.10]) by sue.samurai.com (Postfix) with ESMTP id EC516197F0; Thu, 21 Oct 2004 08:01:55 -0400 (EDT) Message-ID: <4177A538.5070002@samurai.com> Date: Thu, 21 Oct 2004 22:02:00 +1000 From: Neil Conway User-Agent: Mozilla Thunderbird 0.8 (Macintosh/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: matt@ymogen.net Cc: pgsql-performance@postgresql.org Subject: Re: Anything to be gained from a 'Postgres Filesystem'? References: <008601c4b743$aff57f10$8300a8c0@solent> In-Reply-To: <008601c4b743$aff57f10$8300a8c0@solent> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/348 X-Sequence-Number: 8824 Matt Clark wrote: > I'm thinking along the lines of an FS that's aware of PG's strategies and > requirements and therefore optimised to make those activities as efiicient > as possible - possibly even being aware of PG's disk layout and treating > files differently on that basis. As someone else noted, this doesn't belong in the filesystem (rather the kernel's block I/O layer/buffer cache). But I agree, an API by which we can tell the kernel what kind of I/O behavior to expect would be good. The kernel needs to provide good behavior for a wide range of applications, but the DBMS can take advantage of a lot of domain-specific information. In theory, being able to pass that domain-specific information on to the kernel would mean we could get better performance without needing to reimplement large chunks of functionality that really ought to be done by the kernel anyway (as implementing raw I/O would require, for example). On the other hand, it would probably mean adding a fair bit of OS-specific hackery, which we've largely managed to avoid in the past. The closest API to what you're describing that I'm aware of is posix_fadvise(). While that is technically-speaking a POSIX standard, it is not widely implemented (I know Linux 2.6 implements it; based on some quick googling, it looks like AIX does too). Using posix_fadvise() has been discussed in the past, so you might want to search the archives. We could use FADV_SEQUENTIAL to request more aggressive readahead on a file that we know we're about to sequentially scan. We might be able to use FADV_NOREUSE on the WAL. We might be able to get away with specifying FADV_RANDOM for indexes all of the time, or at least most of the time. One question is how this would interact with concurrent access (AFAICS there is no way to fetch the "current advice" on an fd...) Also, I would imagine Win32 provides some means to inform the kernel about your expected I/O pattern, but I haven't checked. Does anyone know of any other relevant APIs? -Neil From pgsql-performance-owner@postgresql.org Thu Oct 21 14:45:21 2004 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 DA92432A138 for ; Thu, 21 Oct 2004 14:45:20 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 19314-05 for ; Thu, 21 Oct 2004 13:45:07 +0000 (GMT) Received: from trofast.sesse.net (trofast.sesse.net [129.241.93.32]) by svr1.postgresql.org (Postfix) with ESMTP id A009332A11C for ; Thu, 21 Oct 2004 14:45:08 +0100 (BST) Received: from sesse by trofast.sesse.net with local (Exim 3.36 #1 (Debian)) id 1CKdFa-0000Rs-00 for ; Thu, 21 Oct 2004 15:45:06 +0200 Date: Thu, 21 Oct 2004 15:45:06 +0200 From: "Steinar H. Gunderson" To: pgsql-performance@postgresql.org Subject: Re: Anything to be gained from a 'Postgres Filesystem'? Message-ID: <20041021134506.GA1667@uio.no> Mail-Followup-To: pgsql-performance@postgresql.org References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: X-Operating-System: Linux 2.6.8.1 on a i686 User-Agent: Mutt/1.5.6+20040907i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/349 X-Sequence-Number: 8825 On Thu, Oct 21, 2004 at 12:44:10PM +0200, Leeuw van der, Tim wrote: > Hacking PG internally to handle raw devices will meet with strong > resistance from large portions of the development team. I don't expect > (m)any core devs of PG will be excited about rewriting the entire I/O > architecture of PG and duplicating large amounts of OS type of code inside > the application, just to try to attain an unknown performance benefit. Well, at least I see people claiming >30% difference between different file systems, but no, I'm not shouting "bah, you'd better do this or I'll warez Oracle" :-) I have no idea how much you can improve over the "best" filesystems out there, but having two layers of journalling (both WAL _and_ FS journalling) on top of each other don't make all that much sense to me. :-) /* Steinar */ -- Homepage: http://www.sesse.net/ From pgsql-performance-owner@postgresql.org Thu Oct 21 14:55:20 2004 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 1879232B17C for ; Thu, 21 Oct 2004 14:55:20 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 23358-04 for ; Thu, 21 Oct 2004 13:55:14 +0000 (GMT) Received: from mail.osdl.org (fw.osdl.org [65.172.181.6]) by svr1.postgresql.org (Postfix) with ESMTP id 2E2D032A263 for ; Thu, 21 Oct 2004 14:55:15 +0100 (BST) Received: (from markw@localhost) by mail.osdl.org (8.11.6/8.11.6) id i9LDswG02355; Thu, 21 Oct 2004 06:54:58 -0700 Date: Thu, 21 Oct 2004 06:54:58 -0700 From: Mark Wong To: Manfred Spraul Cc: Tom Lane , neilc@samurai.com, pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 Message-ID: <20041021065458.A5580@osdl.org> References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> <417697A5.1050600@colorfullife.com> <7178.1098292535@sss.pgh.pa.us> <4176A2C1.3070205@colorfullife.com> <20041020150528.B7838@osdl.org> <41774D11.1030906@colorfullife.com> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="sdtB3X0nJg68CQEu" Content-Disposition: inline User-Agent: Mutt/1.2.5i In-Reply-To: <41774D11.1030906@colorfullife.com>; from manfred@colorfullife.com on Thu, Oct 21, 2004 at 07:45:53AM +0200 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/350 X-Sequence-Number: 8826 --sdtB3X0nJg68CQEu Content-Type: text/plain; charset=us-ascii Content-Disposition: inline On Thu, Oct 21, 2004 at 07:45:53AM +0200, Manfred Spraul wrote: > Mark Wong wrote: > > >Here are some other details, per Manfred's request: > > > >Linux 2.6.8.1 (on a gentoo distro) > > > > > How complicated are Tom's test scripts? His immediate reply was that I > should retest with Fedora, to rule out any gentoo bugs. > > I have a dual-cpu system with RH FC, I could use it for testing. > Pretty, simple. One to load the database, and 1 to query it. I'll attach them. Mark --sdtB3X0nJg68CQEu Content-Type: text/plain; charset=us-ascii Content-Disposition: attachment; filename="test_createscript.sql" drop table test_data; create table test_data(f1 int); insert into test_data values (random() * 100); insert into test_data select random() * 100 from test_data; insert into test_data select random() * 100 from test_data; insert into test_data select random() * 100 from test_data; insert into test_data select random() * 100 from test_data; insert into test_data select random() * 100 from test_data; insert into test_data select random() * 100 from test_data; insert into test_data select random() * 100 from test_data; insert into test_data select random() * 100 from test_data; insert into test_data select random() * 100 from test_data; insert into test_data select random() * 100 from test_data; insert into test_data select random() * 100 from test_data; insert into test_data select random() * 100 from test_data; insert into test_data select random() * 100 from test_data; insert into test_data select random() * 100 from test_data; insert into test_data select random() * 100 from test_data; insert into test_data select random() * 100 from test_data; create index test_index on test_data(f1); vacuum verbose analyze test_data; checkpoint; --sdtB3X0nJg68CQEu Content-Type: text/plain; charset=us-ascii Content-Disposition: attachment; filename="test_script.sql" -- force nestloop indexscan plan set enable_seqscan to 0; set enable_mergejoin to 0; set enable_hashjoin to 0; explain select count(*) from test_data a, test_data b, test_data c where a.f1 = b.f1 and b.f1 = c.f1; select count(*) from test_data a, test_data b, test_data c where a.f1 = b.f1 and b.f1 = c.f1; --sdtB3X0nJg68CQEu-- From pgsql-performance-owner@postgresql.org Thu Oct 21 15:21:12 2004 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 942F232A17C for ; Thu, 21 Oct 2004 15:21:11 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 32287-05 for ; Thu, 21 Oct 2004 14:21:05 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id C780332A11C for ; Thu, 21 Oct 2004 15:21:05 +0100 (BST) 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 i9LEKtIM015150; Thu, 21 Oct 2004 10:20:55 -0400 (EDT) To: "Steinar H. Gunderson" Cc: pgsql-performance@postgresql.org Subject: Re: Anything to be gained from a 'Postgres Filesystem'? In-reply-to: <20041021134506.GA1667@uio.no> References: <20041021134506.GA1667@uio.no> Comments: In-reply-to "Steinar H. Gunderson" message dated "Thu, 21 Oct 2004 15:45:06 +0200" Date: Thu, 21 Oct 2004 10:20:55 -0400 Message-ID: <15149.1098368455@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/351 X-Sequence-Number: 8827 "Steinar H. Gunderson" writes: > ... I have no idea how much you can improve over the "best" > filesystems out there, but having two layers of journalling (both WAL _and_ > FS journalling) on top of each other don't make all that much sense to me. Which is why setting the FS to journal metadata but not file contents is often suggested as best practice for a PG-only filesystem. regards, tom lane From pgsql-performance-owner@postgresql.org Thu Oct 21 15:23:31 2004 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 B34A332A219 for ; Thu, 21 Oct 2004 15:23:29 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 33848-04 for ; Thu, 21 Oct 2004 14:23:23 +0000 (GMT) Received: from ll.mit.edu (LLMAIL.LL.MIT.EDU [129.55.12.40]) by svr1.postgresql.org (Postfix) with ESMTP id 2E72732A263 for ; Thu, 21 Oct 2004 15:23:20 +0100 (BST) Received: (from smtp@localhost) by ll.mit.edu (8.12.10/8.8.8) id i9LEMxRs000661; Thu, 21 Oct 2004 10:22:59 -0400 (EDT) Received: from sty.llan.ll.mit.edu( ), claiming to be "sty.llan" via SMTP by llpost, id smtpdAAA8mayq6; Thu Oct 21 10:22:37 2004 Date: Thu, 21 Oct 2004 10:22:37 -0400 From: george young To: "Ray" Cc: pgsql-performance@postgresql.org Subject: Re: create index with substr function Message-Id: <20041021102237.5ae1d14e.gry@ll.mit.edu> In-Reply-To: <078b01c4b71f$4936e4e0$0f6010ac@raysiu> References: <069201c4b715$34851670$0f6010ac@raysiu> <37d451f704102020346fdbb855@mail.gmail.com> <078b01c4b71f$4936e4e0$0f6010ac@raysiu> Reply-To: gry@ll.mit.edu Organization: MIT Lincoln Laboratory X-Mailer: Sylpheed version 0.9.10 (GTK+ 1.2.10; 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/352 X-Sequence-Number: 8828 As previously suggested by Stephan Szabo, you need to create a helper function, e.g.: create or replace function after9(text)returns text language plpgsql immutable as ' begin return substr($1, 10); end; '; You may need the "immutable" specification is to allow the function's use in an index. Then use this function in the index creation: CREATE INDEX idx_doc_substr_doc_urn ON doc USING btree (after9(doc_urn)); I think that should do it. -- George > On Thu, 21 Oct 2004 11:37:26 +0800 "Ray" threw this fish to the penguins: > sorry it doesn't works, as my postgres is 7.3 not 7.4. any other alternative > solution for version after 7.4?? > > Thank > Ray : ) > > ----- Original Message ----- > From: "Rosser Schwarz" > To: "Ray" > Cc: > Sent: Thursday, October 21, 2004 11:34 AM > Subject: Re: [PERFORM] create index with substr function > > > > while you weren't looking, Ray wrote: > > > > > CREATE INDEX idx_doc_substr_doc_urn ON doc USING btree > (SUBSTR(doc_urn,10)); > > > > CREATE INDEX idx_doc_substr_doc_urn ON doc USING btree > ((SUBSTR(doc_urn,10))); > > > > You need an additional set of parens around the SUBSTR() call. > > > > /rls > > > > -- > > :wq > > > > > ---------------------------(end of broadcast)--------------------------- > TIP 2: you can get off all lists at once with the unregister command > (send "unregister YourEmailAddressHere" to majordomo@postgresql.org) > -- "Are the gods not just?" "Oh no, child. What would become of us if they were?" (CSL) From pgsql-performance-owner@postgresql.org Thu Oct 21 15:34:24 2004 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 6A22732AA2B for ; Thu, 21 Oct 2004 15:34:23 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 38578-03 for ; Thu, 21 Oct 2004 14:34:15 +0000 (GMT) Received: from rproxy.gmail.com (rproxy.gmail.com [64.233.170.192]) by svr1.postgresql.org (Postfix) with ESMTP id 8B07D32AA12 for ; Thu, 21 Oct 2004 15:34:17 +0100 (BST) Received: by rproxy.gmail.com with SMTP id 74so24405rnk for ; Thu, 21 Oct 2004 07:34:17 -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; b=A2pYI9jfZtpzE9euxgct9C1w28Hn+BYYMAIbeDGbU/EsaErtcADCTPmR3Ksyr2k4hjUDP/L95xj0M3INrOgFJaVBBLsL2CtiFOE9MMODLGiLNtkrw4MgSleYG7RCqhJR00w6kb2QsktwM0JoCS3vKqa2u6hBswrj3wSHy2rsNc0= Received: by 10.38.152.39 with SMTP id z39mr3124662rnd; Thu, 21 Oct 2004 07:34:17 -0700 (PDT) Received: by 10.38.125.26 with HTTP; Thu, 21 Oct 2004 07:34:17 -0700 (PDT) Message-ID: Date: Thu, 21 Oct 2004 17:34:17 +0300 From: Victor Ciurus Reply-To: Victor Ciurus To: pgsql-performance@postgresql.org Subject: Simple machine-killing query! 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/353 X-Sequence-Number: 8829 Hi all, I'm writing this because I've reached the limit of my imagination and patience! So here is it... 2 tables: 1 containing 27 million variable lenght, alpha-numeric records (strings) in 1 (one) field. (10 - 145 char lenght per record) 1 containing 2.5 million variable lenght, alpha-numeric records (strings) in 1 (one) field. table wehere created using: CREATE TABLE "public"."BIGMA" ("string" VARCHAR(255) NOT NULL) WITH OIDS; + CREATE INDEX "BIGMA_INDEX" ON "public"."BIGMA" USING btree ("string"); and CREATE TABLE "public"."DIRTY" ("string" VARCHAR(128) NOT NULL) WITH OIDS; + CREATE INDEX "DIRTY_INDEX" ON "public"."DIRTY" USING btree ("string"); What I am requested to do is to keep all records from 'BIGMA' that do not apear in 'DIRTY' So far I have tried solving this by going for: [explain] select * from BIGMA where string not in (select * from DIRTY); QUERY PLAN ------------------------------------------------------------------------ Seq Scan on bigma (cost=0.00..24582291.25 rows=500 width=145) Filter: (NOT (subplan)) SubPlan -> Seq Scan on dirty (cost=0.00..42904.63 rows=2503963 width=82) (4 rows) AND [explain] select * from bigma,dirty where bigma.email!=dirty.email; QUERY PLAN ----------------------------------------------------------------------- Nested Loop (cost=20.00..56382092.13 rows=2491443185 width=227) Join Filter: (("inner".email)::text <> ("outer".email)::text) -> Seq Scan on dirty (cost=0.00..42904.63 rows=2503963 width=82) -> Materialize (cost=20.00..30.00 rows=1000 width=145) -> Seq Scan on bigma (cost=0.00..20.00 rows=1000 width=145) (5 rows) Now the problem is that both of my previous tries seem to last forever! I'm not a pqsql guru so that's why I'm asking you fellas to guide mw right! I've tried this on mysql previosly but there seems to be no way mysql can handle this large query. QUESTIONS: What can I do in order to make this work? Where do I make mistakes? Is there a way I can improve the performance in table design, query style, server setting so that I can get this monster going and producing a result? Thanks all for your preciuos time and answers! Victor C. From pgsql-performance-owner@postgresql.org Thu Oct 21 16:02:32 2004 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 6A61D32AB6A for ; Thu, 21 Oct 2004 16:02:26 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 47942-05 for ; Thu, 21 Oct 2004 15:02:16 +0000 (GMT) Received: from zoidberg.portrix.net (port-212-202-157-208.static.qsc.de [212.202.157.208]) by svr1.postgresql.org (Postfix) with ESMTP id 3C8D832A0E5 for ; Thu, 21 Oct 2004 16:02:09 +0100 (BST) Received: from [127.0.0.1] (port-212-202-157-213.static.qsc.de [212.202.157.213]) (authenticated bits=0) by zoidberg.portrix.net (8.12.3/8.12.3/Debian-6.6) with ESMTP id i9LF20Dd027395 (version=TLSv1/SSLv3 cipher=RC4-MD5 bits=128 verify=NO); Thu, 21 Oct 2004 17:02:01 +0200 Message-ID: <4177CF68.2050108@ppp0.net> Date: Thu, 21 Oct 2004 17:02:00 +0200 From: Jan Dittmer User-Agent: Mozilla/5.0 (X11; U; Linux i686; rv:1.7.3) Gecko/20040926 Thunderbird/0.8 Mnenhy/0.6.0.104 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Neil Conway Cc: matt@ymogen.net, pgsql-performance@postgresql.org Subject: Re: Anything to be gained from a 'Postgres Filesystem'? References: <008601c4b743$aff57f10$8300a8c0@solent> <4177A538.5070002@samurai.com> In-Reply-To: <4177A538.5070002@samurai.com> X-Enigmail-Version: 0.86.1.0 X-Enigmail-Supports: pgp-inline, pgp-mime 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=0.3 tagged_above=0.0 required=5.0 tests=UPPERCASE_25_50 X-Spam-Level: X-Archive-Number: 200410/354 X-Sequence-Number: 8830 Neil Conway wrote: > Also, I would imagine Win32 provides some means to inform the kernel > about your expected I/O pattern, but I haven't checked. Does anyone know > of any other relevant APIs? See CreateFile, Parameter dwFlagsAndAttributes http://msdn.microsoft.com/library/default.asp?url=/library/en-us/fileio/base/createfile.asp There is FILE_FLAG_NO_BUFFERING, FILE_FLAG_OPEN_NO_RECALL, FILE_FLAG_RANDOM_ACCESS and even FILE_FLAG_POSIX_SEMANTICS Jan From pgsql-performance-owner@postgresql.org Thu Oct 21 16:06:12 2004 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 64EBE329F6D for ; Thu, 21 Oct 2004 16:06:11 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49763-05 for ; Thu, 21 Oct 2004 15:06:06 +0000 (GMT) Received: from megazone.bigpanda.com (megazone.bigpanda.com [64.147.171.210]) by svr1.postgresql.org (Postfix) with ESMTP id 9EE4232AA47 for ; Thu, 21 Oct 2004 16:06:05 +0100 (BST) Received: by megazone.bigpanda.com (Postfix, from userid 1001) id 788B2359E0; Thu, 21 Oct 2004 08:05:59 -0700 (PDT) Received: from localhost (localhost [127.0.0.1]) by megazone.bigpanda.com (Postfix) with ESMTP id 770653541C; Thu, 21 Oct 2004 08:05:59 -0700 (PDT) Date: Thu, 21 Oct 2004 08:05:59 -0700 (PDT) From: Stephan Szabo To: Victor Ciurus Cc: pgsql-performance@postgresql.org Subject: Re: Simple machine-killing query! In-Reply-To: Message-ID: <20041021075757.A54616@megazone.bigpanda.com> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/355 X-Sequence-Number: 8831 On Thu, 21 Oct 2004, Victor Ciurus wrote: > Hi all, > > I'm writing this because I've reached the limit of my imagination and > patience! So here is it... > > 2 tables: > 1 containing 27 million variable lenght, alpha-numeric records > (strings) in 1 (one) field. (10 - 145 char lenght per record) > 1 containing 2.5 million variable lenght, alpha-numeric records > (strings) in 1 (one) field. > > table wehere created using: > CREATE TABLE "public"."BIGMA" ("string" VARCHAR(255) NOT NULL) WITH OIDS; + > CREATE INDEX "BIGMA_INDEX" ON "public"."BIGMA" USING btree ("string"); > and > CREATE TABLE "public"."DIRTY" ("string" VARCHAR(128) NOT NULL) WITH OIDS; + > CREATE INDEX "DIRTY_INDEX" ON "public"."DIRTY" USING btree ("string"); > > What I am requested to do is to keep all records from 'BIGMA' that do > not apear in 'DIRTY' > So far I have tried solving this by going for: > > [explain] select * from BIGMA where string not in (select * from DIRTY); > QUERY PLAN > ------------------------------------------------------------------------ > Seq Scan on bigma (cost=0.00..24582291.25 rows=500 width=145) > Filter: (NOT (subplan)) > SubPlan > -> Seq Scan on dirty (cost=0.00..42904.63 rows=2503963 width=82) > (4 rows) Have you analyzed bigma? The number of rows from the two explains for that table look suspiciously like default values. Also, what version are you using, because there are some differences from 7.3 to 7.4 that change possible suggestions. The first is that on 7.4, you may be able to do better with a higher sort_mem which could possible switch over to the hashed implementation, although I think it's probably going to take a pretty high value given the size. The second is that you might get better results (even on older versions) from an exists or left join solution, something like (assuming no nulls in bigma.email): select * from bigma where not exists(select 1 from dirty where dirty.email != bigma.email); select bigma.* from bigma left outer join dirty on (dirty.email = bigma.email) where dirty.email is null; If you've got nulls in bigma.email you have to be a little more careful. > [explain] select * from bigma,dirty where bigma.email!=dirty.email; This *almost* certainly does not do what you want. For most data sets this is going to give you a number of rows very close to # of rows in dirty * # of rows in bigma. Needless to say, this is going to take a long time. From pgsql-performance-owner@postgresql.org Thu Oct 21 16:14:39 2004 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 377E432A014 for ; Thu, 21 Oct 2004 16:14:31 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 53130-06 for ; Thu, 21 Oct 2004 15:14:21 +0000 (GMT) Received: from rproxy.gmail.com (rproxy.gmail.com [64.233.170.202]) by svr1.postgresql.org (Postfix) with ESMTP id 916BA329F45 for ; Thu, 21 Oct 2004 16:14:17 +0100 (BST) Received: by rproxy.gmail.com with SMTP id 73so37559rnk for ; Thu, 21 Oct 2004 08:14:14 -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:references; b=nzdYqxeVZqXLTp49IeT5waWE7IqaiMcSXaEWBbhi3+l442omzMhES7SDZd5u3IixvlD9uixNOw82duTkyKzvOwtYAdm7MmN9pynAr5eGnaBTVc4cu/507mSIMv7r4hRCjKAHfV347BP34uFxSTR3I2p0pMFX4pKInBZ/sP77/6M= Received: by 10.38.181.77 with SMTP id d77mr3130083rnf; Thu, 21 Oct 2004 08:14:14 -0700 (PDT) Received: by 10.38.151.75 with HTTP; Thu, 21 Oct 2004 08:14:14 -0700 (PDT) Message-ID: <157f648404102108141836a5ac@mail.gmail.com> Date: Thu, 21 Oct 2004 11:14:14 -0400 From: Aaron Werman Reply-To: Aaron Werman To: Victor Ciurus Subject: Re: Simple machine-killing query! Cc: pgsql-performance@postgresql.org In-Reply-To: Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/356 X-Sequence-Number: 8832 Sounds like you need some way to match a subset of the data first, rather than try indices that are bigger than the data. Can you add operation indices, perhaps on the first 10 bytes of the keys in both tables or on a integer hash of all of the strings? If so you could join on the exact set difference over the set difference of the operation match. /Aaron On Thu, 21 Oct 2004 17:34:17 +0300, Victor Ciurus wrote: > Hi all, > > I'm writing this because I've reached the limit of my imagination and > patience! So here is it... > > 2 tables: > 1 containing 27 million variable lenght, alpha-numeric records > (strings) in 1 (one) field. (10 - 145 char lenght per record) > 1 containing 2.5 million variable lenght, alpha-numeric records > (strings) in 1 (one) field. > > table wehere created using: > CREATE TABLE "public"."BIGMA" ("string" VARCHAR(255) NOT NULL) WITH OIDS; + > CREATE INDEX "BIGMA_INDEX" ON "public"."BIGMA" USING btree ("string"); > and > CREATE TABLE "public"."DIRTY" ("string" VARCHAR(128) NOT NULL) WITH OIDS; + > CREATE INDEX "DIRTY_INDEX" ON "public"."DIRTY" USING btree ("string"); > > What I am requested to do is to keep all records from 'BIGMA' that do > not apear in 'DIRTY' > So far I have tried solving this by going for: > > [explain] select * from BIGMA where string not in (select * from DIRTY); > QUERY PLAN > ------------------------------------------------------------------------ > Seq Scan on bigma (cost=0.00..24582291.25 rows=500 width=145) > Filter: (NOT (subplan)) > SubPlan > -> Seq Scan on dirty (cost=0.00..42904.63 rows=2503963 width=82) > (4 rows) > > AND > > [explain] select * from bigma,dirty where bigma.email!=dirty.email; > QUERY PLAN > ----------------------------------------------------------------------- > Nested Loop (cost=20.00..56382092.13 rows=2491443185 width=227) > Join Filter: (("inner".email)::text <> ("outer".email)::text) > -> Seq Scan on dirty (cost=0.00..42904.63 rows=2503963 width=82) > -> Materialize (cost=20.00..30.00 rows=1000 width=145) > -> Seq Scan on bigma (cost=0.00..20.00 rows=1000 width=145) > (5 rows) > > Now the problem is that both of my previous tries seem to last > forever! I'm not a pqsql guru so that's why I'm asking you fellas to > guide mw right! I've tried this on mysql previosly but there seems to > be no way mysql can handle this large query. > > QUESTIONS: > What can I do in order to make this work? > Where do I make mistakes? Is there a way I can improve the performance > in table design, query style, server setting so that I can get this > monster going and producing a result? > > Thanks all for your preciuos time and answers! > > Victor C. > > ---------------------------(end of broadcast)--------------------------- > TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org > -- Regards, /Aaron From pgsql-performance-owner@postgresql.org Thu Oct 21 16:42:05 2004 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 9204832ABC7 for ; Thu, 21 Oct 2004 16:42:04 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 62336-08 for ; Thu, 21 Oct 2004 15:41:55 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 7C82332A452 for ; Thu, 21 Oct 2004 16:41:51 +0100 (BST) 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 i9LFfnUk016018; Thu, 21 Oct 2004 11:41:49 -0400 (EDT) To: Victor Ciurus Cc: pgsql-performance@postgresql.org Subject: Re: Simple machine-killing query! In-reply-to: References: Comments: In-reply-to Victor Ciurus message dated "Thu, 21 Oct 2004 17:34:17 +0300" Date: Thu, 21 Oct 2004 11:41:48 -0400 Message-ID: <16017.1098373308@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/357 X-Sequence-Number: 8833 Victor Ciurus writes: > What I am requested to do is to keep all records from 'BIGMA' that do > not apear in 'DIRTY' > So far I have tried solving this by going for: > [explain] select * from BIGMA where string not in (select * from DIRTY); > QUERY PLAN > ------------------------------------------------------------------------ > Seq Scan on bigma (cost=0.00..24582291.25 rows=500 width=145) > Filter: (NOT (subplan)) > SubPlan > -> Seq Scan on dirty (cost=0.00..42904.63 rows=2503963 width=82) > (4 rows) If you are using PG 7.4, you can get reasonable performance out of this approach, but you need to jack sort_mem up to the point where the whole DIRTY table will fit into sort_mem (so that you get a hashed-subplan plan and not a plain subplan). If you find yourself setting sort_mem to more than say half of your machine's available RAM, you should probably forget that idea. > [explain] select * from bigma,dirty where bigma.email!=dirty.email; This of course does not give the right answer at all. A trick that people sometimes use is an outer join: select * from bigma left join dirty on (bigma.email=dirty.email) where dirty.email is null; Understanding why this works is left as an exercise for the reader ... but it does work, and pretty well too. If you're using pre-7.4 PG then this is about the only effective solution AFAIR. regards, tom lane From pgsql-performance-owner@postgresql.org Thu Oct 21 16:49:33 2004 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 CC3C232A154 for ; Thu, 21 Oct 2004 16:49:31 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 65023-10 for ; Thu, 21 Oct 2004 15:49:26 +0000 (GMT) Received: from trofast.sesse.net (trofast.sesse.net [129.241.93.32]) by svr1.postgresql.org (Postfix) with ESMTP id 3587B329E61 for ; Thu, 21 Oct 2004 16:49:24 +0100 (BST) Received: from sesse by trofast.sesse.net with local (Exim 3.36 #1 (Debian)) id 1CKfBr-0000gY-00 for ; Thu, 21 Oct 2004 17:49:23 +0200 Date: Thu, 21 Oct 2004 17:49:23 +0200 From: "Steinar H. Gunderson" To: pgsql-performance@postgresql.org Subject: Re: Anything to be gained from a 'Postgres Filesystem'? Message-ID: <20041021154923.GA2574@uio.no> Mail-Followup-To: pgsql-performance@postgresql.org References: <20041021134506.GA1667@uio.no> <15149.1098368455@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <15149.1098368455@sss.pgh.pa.us> X-Operating-System: Linux 2.6.8.1 on a i686 User-Agent: Mutt/1.5.6+20040907i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/358 X-Sequence-Number: 8834 On Thu, Oct 21, 2004 at 10:20:55AM -0400, Tom Lane wrote: >> ... I have no idea how much you can improve over the "best" >> filesystems out there, but having two layers of journalling (both WAL _and_ >> FS journalling) on top of each other don't make all that much sense to me. > Which is why setting the FS to journal metadata but not file contents is > often suggested as best practice for a PG-only filesystem. Mm, but you still journal the metadata. Oh well, noatime etc.. :-) By the way, I'm probably hitting a FAQ here, but would O_DIRECT help PostgreSQL any, given large enough shared_buffers? /* Steinar */ -- Homepage: http://www.sesse.net/ From pgsql-performance-owner@postgresql.org Thu Oct 21 18:03:48 2004 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 C9C0832A15F for ; Thu, 21 Oct 2004 18:03:47 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 89977-03 for ; Thu, 21 Oct 2004 17:03:45 +0000 (GMT) Received: from rproxy.gmail.com (rproxy.gmail.com [64.233.170.206]) by svr1.postgresql.org (Postfix) with ESMTP id 01E7F32A095 for ; Thu, 21 Oct 2004 18:03:45 +0100 (BST) Received: by rproxy.gmail.com with SMTP id 73so54722rnk for ; Thu, 21 Oct 2004 10:03:41 -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:references; b=IQ9dpCYO4Dn87UkWt+DaueTvl0uZ9UpNxzs0DafKjH8V9qTrq9lgc7ClqoVdp09Q+fVYEKNkAe9GU1ZPA2GkqediNQInZdYrNRcaVa7vc1pM06h3OF8Q7xtg6oXNkYvlCxMiLsr3RpTJ+gjxMvntCAPN2nY028swawfUfzAw6Zg= Received: by 10.38.151.68 with SMTP id y68mr3190700rnd; Thu, 21 Oct 2004 10:03:41 -0700 (PDT) Received: by 10.38.125.26 with HTTP; Thu, 21 Oct 2004 10:03:41 -0700 (PDT) Message-ID: Date: Thu, 21 Oct 2004 20:03:41 +0300 From: Victor Ciurus Reply-To: Victor Ciurus To: pgsql-performance@postgresql.org Subject: Re: Simple machine-killing query! Cc: Tom Lane In-Reply-To: <16017.1098373308@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: <16017.1098373308@sss.pgh.pa.us> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/359 X-Sequence-Number: 8835 Well guys, Your replies have been more than helpful to me, showing me both the learning stuff I still have to get in my mind about real SQL and the wonder called PostgreSQL and a very good solution from Tom Lane (thanks a lot sir!)! Indeed, changing mem_sort and other server parmeters along with the quite strange (to me!) outer join Tom mentioned finally got me to finalize the cleaning task and indeed in warp speed (some 5 mintues or less!). I am running PG v7.4.5 on a PIV Celeron 1,7Ghz with 1,5Gb ram so talking about the time performance I might say that I'm more than pleased with the result. As with the amazement PG "caused" me through its reliability so far I am decided to go even deeper in learning it! Thanks again all for your precious help! Regards, Victor > > If you are using PG 7.4, you can get reasonable performance out of this > approach, but you need to jack sort_mem up to the point where the whole > DIRTY table will fit into sort_mem (so that you get a hashed-subplan > plan and not a plain subplan). If you find yourself setting sort_mem to > more than say half of your machine's available RAM, you should probably > forget that idea. > > > [explain] select * from bigma,dirty where bigma.email!=dirty.email; > > This of course does not give the right answer at all. > > A trick that people sometimes use is an outer join: > > select * from bigma left join dirty on (bigma.email=dirty.email) > where dirty.email is null; > > Understanding why this works is left as an exercise for the reader > ... but it does work, and pretty well too. If you're using pre-7.4 > PG then this is about the only effective solution AFAIR. > > regards, tom lane > From pgsql-performance-owner@postgresql.org Thu Oct 21 18:15:09 2004 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 09B1E32AB6A for ; Thu, 21 Oct 2004 18:15:07 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 92384-08 for ; Thu, 21 Oct 2004 17:15:04 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 0A60B32AA72 for ; Thu, 21 Oct 2004 18:15:04 +0100 (BST) 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 6538430; Thu, 21 Oct 2004 10:16:27 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org, Victor Ciurus Subject: Re: Simple machine-killing query! Date: Thu, 21 Oct 2004 10:14:39 -0700 User-Agent: KMail/1.6.2 References: In-Reply-To: MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Message-Id: <200410211014.39916.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/360 X-Sequence-Number: 8836 Victor, > [explain] select * from BIGMA where string not in (select * from DIRTY); > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0QUERY PLAN > ------------------------------------------------------------------------ > =A0Seq Scan on bigma =A0(cost=3D0.00..24582291.25 rows=3D500 width=3D145) > =A0 =A0Filter: (NOT (subplan)) > =A0 =A0SubPlan > =A0 =A0 =A0-> =A0Seq Scan on dirty =A0(cost=3D0.00..42904.63 rows=3D25039= 63 width=3D82) This is what you call an "evil query". I'm not surprised it takes forever= ;=20 you're telling the database "Compare every value in 2.7 million rows of tex= t=20 against 2.5 million rows of text and give me those that don't match." The= re=20 is simply no way, on ANY RDBMS, for this query to execute and not eat all o= f=20 your RAM and CPU for a long time. You're simply going to have to allocate shared_buffers and sort_mem (about = 2GB=20 of sort_mem would be good) to the query, and turn the computer over to the= =20 task until it's done. And, for the sake of sanity, when you find the=20 200,000 rows that don't match, flag them so that you don't have to do this= =20 again. --=20 Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Thu Oct 21 19:40:49 2004 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 DA80B329E6D for ; Thu, 21 Oct 2004 19:40:47 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 20382-04 for ; Thu, 21 Oct 2004 18:40:35 +0000 (GMT) Received: from mail.trippynames.com (mail.trippynames.com [38.113.223.19]) by svr1.postgresql.org (Postfix) with ESMTP id 803C732A938 for ; Thu, 21 Oct 2004 19:40:35 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by mail.trippynames.com (Postfix) with ESMTP id 3EE08A3232; Thu, 21 Oct 2004 11:40:34 -0700 (PDT) Received: from mail.trippynames.com ([127.0.0.1]) by localhost (rand.nxad.com [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 84941-07; Thu, 21 Oct 2004 11:40:33 -0700 (PDT) Received: from [192.168.1.102] (wbar4.sjo1-4.28.216.220.sjo1.dsl-verizon.net [4.28.216.220]) by mail.trippynames.com (Postfix) with ESMTP id C4E90A323B; Thu, 21 Oct 2004 11:40:20 -0700 (PDT) In-Reply-To: <4177A538.5070002@samurai.com> References: <008601c4b743$aff57f10$8300a8c0@solent> <4177A538.5070002@samurai.com> Mime-Version: 1.0 (Apple Message framework v619) Content-Type: text/plain; charset=US-ASCII; format=flowed Message-Id: Content-Transfer-Encoding: 7bit Cc: matt@ymogen.net, pgsql-performance@postgresql.org From: Sean Chittenden Subject: Re: Anything to be gained from a 'Postgres Filesystem'? Date: Thu, 21 Oct 2004 11:40:08 -0700 To: Neil Conway X-Mailer: Apple Mail (2.619) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/361 X-Sequence-Number: 8837 > As someone else noted, this doesn't belong in the filesystem (rather > the kernel's block I/O layer/buffer cache). But I agree, an API by > which we can tell the kernel what kind of I/O behavior to expect would > be good. [snip] > The closest API to what you're describing that I'm aware of is > posix_fadvise(). While that is technically-speaking a POSIX standard, > it is not widely implemented (I know Linux 2.6 implements it; based on > some quick googling, it looks like AIX does too). Don't forget about the existence/usefulness/widely implemented madvise(2)/posix_madvise(2) call, which can give the OS the following hints: MADV_NORMAL, MADV_SEQUENTIAL, MADV_RANDOM, MADV_WILLNEED, MADV_DONTNEED, and MADV_FREE. :) -sc -- Sean Chittenden From pgsql-performance-owner@postgresql.org Thu Oct 21 21:26:49 2004 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 4C34C32A8D5 for ; Thu, 21 Oct 2004 21:26:47 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 53966-01 for ; Thu, 21 Oct 2004 20:26:39 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id D150D32A0DC for ; Thu, 21 Oct 2004 21:26:39 +0100 (BST) 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 6539044; Thu, 21 Oct 2004 13:28:04 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Links to OSDL test results up Date: Thu, 21 Oct 2004 13:28:46 -0700 User-Agent: KMail/1.6.2 Cc: "Simon Riggs" , References: In-Reply-To: MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410211328.46068.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/362 X-Sequence-Number: 8838 Simon, Folks, I've put links to all of my OSDL-STP test results up on the TestPerf project: http://pgfoundry.org/forum/forum.php?thread_id=164&forum_id=160 SHare&Enjoy! -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Thu Oct 21 21:29:52 2004 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 026C432ABD5 for ; Thu, 21 Oct 2004 21:29:52 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 52533-10 for ; Thu, 21 Oct 2004 20:29:41 +0000 (GMT) Received: from mail.trippynames.com (mail.trippynames.com [38.113.223.19]) by svr1.postgresql.org (Postfix) with ESMTP id B228F32ABB2 for ; Thu, 21 Oct 2004 21:29:40 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by mail.trippynames.com (Postfix) with ESMTP id EA4C4A4694; Thu, 21 Oct 2004 13:29:39 -0700 (PDT) Received: from mail.trippynames.com ([127.0.0.1]) by localhost (rand.nxad.com [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 98075-08; Thu, 21 Oct 2004 13:29:37 -0700 (PDT) Received: from [192.168.1.102] (wbar4.sjo1-4.28.216.220.sjo1.dsl-verizon.net [4.28.216.220]) by mail.trippynames.com (Postfix) with ESMTP id 18BD1A464B; Thu, 21 Oct 2004 13:29:36 -0700 (PDT) In-Reply-To: <26390.1097875346@sss.pgh.pa.us> References: <157f64840410141725162e43b5@mail.gmail.com> <20041015010841.GE665@filer> <0E3BB9CB-1EE6-11D9-A0BB-000A95C705DC@chittenden.org> <26390.1097875346@sss.pgh.pa.us> Mime-Version: 1.0 (Apple Message framework v619) Content-Type: text/plain; charset=US-ASCII; format=flowed Message-Id: Content-Transfer-Encoding: 7bit Cc: Kevin Brown , pgsql-performance@postgresql.org From: Sean Chittenden Subject: Re: mmap (was First set of OSDL Shared Mem scalability results, some wierdness ... Date: Thu, 21 Oct 2004 13:29:34 -0700 To: Tom Lane X-Mailer: Apple Mail (2.619) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/363 X-Sequence-Number: 8839 > However the really major difficulty with using mmap is that it breaks > the scheme we are currently using for WAL, because you don't have any > way to restrict how soon a change in an mmap'd page will go to disk. > (No, I don't believe that mlock guarantees this. It says that the > page will not be removed from main memory; it does not specify that, > say, the syncer won't write the contents out anyway.) I had to think about this for a minute (now nearly a week) and reread the docs on WAL before I groked what could happen here. You're absolutely right in that WAL needs to be taken into account first. How does this execution path sound to you? By default, all mmap(2)'ed pages are MAP_SHARED. There are no complications with regards to reads. When a backend wishes to write a page, the following steps are taken: 1) Backend grabs a lock from the lockmgr to write to the page (exactly as it does now) 2) Backend mmap(2)'s a second copy of the page(s) being written to, this time with the MAP_PRIVATE flag set. Mapping a copy of the page again is wasteful in terms of address space, but does not require any more memory than our current scheme. The re-mapping of the page with MAP_PRIVATE prevents changes to the data that other backends are viewing. 3) The writing backend, can then scribble on its private copy of the page(s) as it sees fit. 4) Once completed making changes and a transaction is to be committed, the backend WAL logs its changes. 5) Once the WAL logging is complete and it has hit the disk, the backend msync(2)'s its private copy of the pages to disk (ASYNC or SYNC, it doesn't really matter too much to me). 6) Optional(?). I'm not sure whether or not the backend would need to also issues an msync(2) MS_INVALIDATE, but, I suspect it would not need to on systems with unified buffer caches such as FreeBSD or OS-X. On HPUX, or other older *NIX'es, it may be necessary. *shrug* I could be trying to be overly protective here. 7) Backend munmap(2)'s its private copy of the written on page(s). 8) Backend releases its lock from the lockmgr. At this point, the remaining backends now are able to see the updated pages of data. >> Let's look at what happens with a read(2) call. To read(2) data you >> have to have a block of memory to copy data into. Assume your OS of >> choice has a good malloc(3) implementation and it only needs to call >> brk(2) once to extend the process's memory address after the first >> malloc(3) call. There's your first system call, which guarantees one >> context switch. > > Wrong. Our reads occur into shared memory allocated at postmaster > startup, remember? Doh. Fair enough. In most programs that involve read(2), a call to alloc(3) needs to be made. >> mmap(2) is a totally different animal in that you don't ever need to >> make calls to read(2): mmap(2) is used in place of those calls (With >> #ifdef and a good abstraction, the rest of PostgreSQL wouldn't know it >> was working with a page of mmap(2)'ed data or need to know that it >> is). > > Instead, you have to worry about address space management and keeping a > consistent view of the data. Which is largely handled by mmap() and the VM. >> ... If a write(2) system call is issued on a page of >> mmap(2)'ed data (and your operating system supports it, I know FreeBSD >> does, but don't think Linux does), then the page of data is DMA'ed by >> the network controller and sent out without the data needing to be >> copied into the network controller's buffer. > > Perfectly irrelevant to Postgres, since there is no situation where > we'd > ever write directly from a disk buffer to a socket; in the present > implementation there are at least two levels of copy needed in between > (datatype-specific output function and protocol message assembly). And > that's not even counting the fact that any data item large enough to > make the savings interesting would have been sliced, diced, and > compressed by TOAST. The biggest winners will be columns whos storage type is PLAIN or EXTERNAL. writev(2) from mmap(2)'ed pages and non-mmap(2)'ed pages would be a nice perk too (not sure if PostgreSQL uses this or not). Since compression isn't happening on most tuples under 1K in size and most tuples in a database are going to be under that, most tuples are going to be uncompressed. Total pages for the database, however, is likely a different story. For large tuples that are uncompressed and larger than a page, it is probably beneficial to use sendfile(2) instead of mmap(2) + write(2)'ing the page/file. If a large tuple is compressed, it'd be interesting to see if it'd be worthwhile to have the data uncompressed onto an anonymously mmap(2)'ed page(s) that way the benefits of zero-socket-copies could be used. >> shared mem is a bastardized subsystem that works, but isn't integral >> to >> any performance areas in the kernel so it gets neglected. > > What performance issues do you think shared memory needs to have fixed? > We don't issue any shmem kernel calls after the initial shmget, so > comparing the level of kernel tenseness about shmget to the level of > tenseness about mmap is simply irrelevant. Perhaps the reason you > don't > see any traffic about this on the kernel lists is that shared memory > already works fine and doesn't need any fixing. I'm gunna get flamed for this, but I think its improperly used as a second level cache on top of the operating system's cache. mmap(2) would consolidate all caching into the kernel. >> Please ask questions if you have them. > > Do you have any arguments that are actually convincing? Three things come to mind. 1) A single cache for pages 2) Ability to give access hints to the kernel regarding future IO 3) On the fly memory use for a cache. There would be no need to preallocate slabs of shared memory on startup. And a more minor point would be: 4) Not having shared pages get lost when the backend dies (mmap(2) uses refcounts and cleans itself up, no need for ipcs/ipcrm/ipcclean). This isn't too practical in production though, but it sucks doing PostgreSQL development on OS-X because there is no ipcs/ipcrm command. > What I just read was a proposal to essentially throw away not only the > entire > low-level data access model, but the entire low-level locking model, > and start from scratch. From the above list, steps 2, 3, 5, 6, and 7 would be different than our current approach, all of which could be safely handled with some #ifdef's on platforms that don't have mmap(2). > There is no possible way we could support both > this approach and the current one, which means that we'd be permanently > dropping support for all platforms without high-quality mmap > implementations; Architecturally, I don't see anything different or incompatibilities that aren't solved with an #ifdef USE_MMAP/#else/#endif. > Furthermore, you didn't > give any really convincing reasons to think that the enormous effort > involved would be repaid. Steven's has a great reimplementaion of cat(1) that uses mmap(1) and benchmarks the two. I did my own version of that here: http://people.freebsd.org/~seanc/mmap_test/ When read(2)'ing/write(2)'ing /etc/services 100,000 times without mmap(2), it takes 82 seconds. With mmap(2), it takes anywhere from 1.1 to 18 seconds. Worst case scenario with mmap(2) yields a speedup by a factor of four. Best case scenario... *shrug* something better than 4x. I doubt PostgreSQL would see 4x speedups in the IO department, but I do think it would be vastly greater than the 3% suggested. > Those oprofile reports Josh just put up > showed 3% of the CPU time going into userspace/kernelspace copying. > Even assuming that that number consists entirely of reads and writes of > shared buffers (and of course no other kernel call ever transfers any > data across that boundary ;-)), there's no way we are going to buy into > this sort of project in hopes of a 3% win. Would it be helpful if I created a test program that demonstrated that the execution path for writing mmap(2)'ed pages as outlined above? -sc -- Sean Chittenden From pgsql-performance-owner@postgresql.org Thu Oct 21 21:36:31 2004 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 0B67A32A017 for ; Thu, 21 Oct 2004 21:36:29 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 55162-07 for ; Thu, 21 Oct 2004 20:36:21 +0000 (GMT) Received: from window.monsterlabs.com (window.monsterlabs.com [216.183.105.176]) by svr1.postgresql.org (Postfix) with SMTP id 7F28632A002 for ; Thu, 21 Oct 2004 21:36:20 +0100 (BST) Received: (qmail 16834 invoked from network); 21 Oct 2004 20:36:18 -0000 Received: from w080.z064003242.bna-tn.dsl.cnc.net (HELO ?192.168.1.22?) (64.3.242.80) by 0 with SMTP; 21 Oct 2004 20:36:18 -0000 Mime-Version: 1.0 (Apple Message framework v619) Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: Content-Transfer-Encoding: 7bit From: Thomas F.O'Connell Subject: Performance Anomalies in 7.4.5 Date: Thu, 21 Oct 2004 15:36:02 -0500 To: PgSQL - Performance X-Mailer: Apple Mail (2.619) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/364 X-Sequence-Number: 8840 I'm seeing some weird behavior on a repurposed server that was wiped clean and set up to run as a database and application server with postgres and Apache, as well as some command-line PHP scripts. The box itself is a quad processor (2.4 GHz Intel Xeons) Debian woody GNU/Linux (2.6.2) system. postgres is crawling on some fairly routine queries. I'm wondering if this could somehow be related to the fact that this isn't a database-only server, but Apache is not really using any resources when postgres slows to a crawl. Here's an example of analysis of a recent query: EXPLAIN ANALYZE SELECT COUNT(DISTINCT u.id) FROM userdata as u, userdata_history as h WHERE h.id = '18181' AND h.id = u.id; QUERY PLAN ------------------------------------------------------------------------ ------------------------------------------------------------------------ Aggregate (cost=0.02..0.02 rows=1 width=8) (actual time=298321.421..298321.422 rows=1 loops=1) -> Nested Loop (cost=0.00..0.01 rows=1 width=8) (actual time=1.771..298305.531 rows=2452 loops=1) Join Filter: ("inner".id = "outer".id) -> Seq Scan on userdata u (cost=0.00..0.00 rows=1 width=8) (actual time=0.026..11.869 rows=2452 loops=1) -> Seq Scan on userdata_history h (cost=0.00..0.00 rows=1 width=8) (actual time=0.005..70.519 rows=41631 loops=2452) Filter: (id = 18181::bigint) Total runtime: 298321.926 ms (7 rows) userdata has a primary/foreign key on id, which references userdata_history.id, which is a primary key. At the time of analysis, the userdata table had < 2,500 rows. userdata_history had < 50,000 rows. I can't imagine how even a seq scan could result in a runtime of nearly 5 minutes in these circumstances. Also, doing a count( * ) from each table individually returns nearly instantly. I can provide details of postgresql.conf and kernel settings if necessary, but I'm using some pretty well tested settings that I use any time I admin a postgres installation these days based on box resources and database size. I'm more interested in knowing if there are any bird's eye details I should be checking immediately. Thanks. -tfo -- Thomas F. O'Connell Co-Founder, Information Architect Sitening, LLC http://www.sitening.com/ 110 30th Avenue North, Suite 6 Nashville, TN 37203-6320 615-260-0005 From pgsql-performance-owner@postgresql.org Thu Oct 21 21:50:38 2004 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 A73DB32A518 for ; Thu, 21 Oct 2004 21:50:35 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 57362-07 for ; Thu, 21 Oct 2004 20:50:24 +0000 (GMT) Received: from window.monsterlabs.com (window.monsterlabs.com [216.183.105.176]) by svr1.postgresql.org (Postfix) with SMTP id 7BE5F32A002 for ; Thu, 21 Oct 2004 21:50:24 +0100 (BST) Received: (qmail 21915 invoked from network); 21 Oct 2004 20:50:24 -0000 Received: from w080.z064003242.bna-tn.dsl.cnc.net (HELO ?192.168.1.22?) (64.3.242.80) by 0 with SMTP; 21 Oct 2004 20:50:24 -0000 Mime-Version: 1.0 (Apple Message framework v619) In-Reply-To: References: Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: Content-Transfer-Encoding: 7bit From: Thomas F.O'Connell Subject: Re: Performance Anomalies in 7.4.5 Date: Thu, 21 Oct 2004 15:50:20 -0500 To: PgSQL - Performance X-Mailer: Apple Mail (2.619) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/365 X-Sequence-Number: 8841 I know, I know: I should've done this before I posted. REINDEXing and VACUUMing mostly fixed this problem. Which gets me back to where I was yesterday, reviewing an import process (that existed previously) that populates tables in this system that seems to allow small data sets to cause simple queries like this to crawl. Is there anything about general COPY/INSERT activity that can cause small data sets to become so severely slow in postgres that can be prevented other than being diligent about VACUUMing? I was hoping that pg_autovacuum along with post-import manual VACUUMs would be sufficient, but it doesn't seem to be the case necessarily. Granted, I haven't done a methodical and complete review of the process, but I'm still surprised at how quickly it seems able to debilitate postgres with even small amounts of data. I had a similar situation crawl yesterday based on a series of COPYs involving 5 rows! As in, can I look for something to treat the cause rather than the symptoms? If not, should I be REINDEXing manually, as well as VACUUMing manually after large data imports (whether via COPY or INSERT)? Or will a VACUUM FULL ANALYZE be enough? Thanks! -tfo -- Thomas F. O'Connell Co-Founder, Information Architect Sitening, LLC http://www.sitening.com/ 110 30th Avenue North, Suite 6 Nashville, TN 37203-6320 615-260-0005 On Oct 21, 2004, at 3:36 PM, Thomas F.O'Connell wrote: > I'm seeing some weird behavior on a repurposed server that was wiped > clean and set up to run as a database and application server with > postgres and Apache, as well as some command-line PHP scripts. > > The box itself is a quad processor (2.4 GHz Intel Xeons) Debian woody > GNU/Linux (2.6.2) system. > > postgres is crawling on some fairly routine queries. I'm wondering if > this could somehow be related to the fact that this isn't a > database-only server, but Apache is not really using any resources > when postgres slows to a crawl. > > Here's an example of analysis of a recent query: > > EXPLAIN ANALYZE SELECT COUNT(DISTINCT u.id) > FROM userdata as u, userdata_history as h > WHERE h.id = '18181' > AND h.id = u.id; > > QUERY PLAN > ----------------------------------------------------------------------- > ----------------------------------------------------------------------- > -- > Aggregate (cost=0.02..0.02 rows=1 width=8) (actual > time=298321.421..298321.422 rows=1 loops=1) > -> Nested Loop (cost=0.00..0.01 rows=1 width=8) (actual > time=1.771..298305.531 rows=2452 loops=1) > Join Filter: ("inner".id = "outer".id) > -> Seq Scan on userdata u (cost=0.00..0.00 rows=1 width=8) > (actual time=0.026..11.869 rows=2452 loops=1) > -> Seq Scan on userdata_history h (cost=0.00..0.00 rows=1 > width=8) (actual time=0.005..70.519 rows=41631 loops=2452) > Filter: (id = 18181::bigint) > Total runtime: 298321.926 ms > (7 rows) > > userdata has a primary/foreign key on id, which references > userdata_history.id, which is a primary key. > > At the time of analysis, the userdata table had < 2,500 rows. > userdata_history had < 50,000 rows. I can't imagine how even a seq > scan could result in a runtime of nearly 5 minutes in these > circumstances. > > Also, doing a count( * ) from each table individually returns nearly > instantly. > > I can provide details of postgresql.conf and kernel settings if > necessary, but I'm using some pretty well tested settings that I use > any time I admin a postgres installation these days based on box > resources and database size. I'm more interested in knowing if there > are any bird's eye details I should be checking immediately. > > Thanks. > > -tfo > > -- > Thomas F. O'Connell > Co-Founder, Information Architect > Sitening, LLC > http://www.sitening.com/ > 110 30th Avenue North, Suite 6 > Nashville, TN 37203-6320 > 615-260-0005 > > > ---------------------------(end of > broadcast)--------------------------- > TIP 8: explain analyze is your friend From pgsql-performance-owner@postgresql.org Mon Oct 25 16:58:07 2004 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 4C28C32A06A for ; Thu, 21 Oct 2004 22:21:04 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 69877-05 for ; Thu, 21 Oct 2004 21:20:47 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id DBBA532A5A2 for ; Thu, 21 Oct 2004 22:20:48 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i9LLKjJ7070808 for ; Thu, 21 Oct 2004 21:20:45 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i9LL4ul7065956 for pgsql-performance@postgresql.org; Thu, 21 Oct 2004 21:04:56 GMT From: Bricklen User-Agent: Mozilla Thunderbird 0.8 (X11/20040926) X-Accept-Language: en-us, en MIME-Version: 1.0 X-Newsgroups: comp.databases.postgresql.performance Subject: Re: Performance Anomalies in 7.4.5 References: In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Lines: 69 Message-ID: <_vVdd.31479$z96.8087@clgrps12> Date: Thu, 21 Oct 2004 21:04:58 GMT To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.1 tagged_above=0.0 required=5.0 tests=NO_DNS_FOR_FROM X-Spam-Level: * X-Archive-Number: 200410/415 X-Sequence-Number: 8891 Thomas F.O'Connell wrote: > I'm seeing some weird behavior on a repurposed server that was wiped > clean and set up to run as a database and application server with > postgres and Apache, as well as some command-line PHP scripts. > > The box itself is a quad processor (2.4 GHz Intel Xeons) Debian woody > GNU/Linux (2.6.2) system. > > postgres is crawling on some fairly routine queries. I'm wondering if > this could somehow be related to the fact that this isn't a > database-only server, but Apache is not really using any resources when > postgres slows to a crawl. > > Here's an example of analysis of a recent query: > > EXPLAIN ANALYZE SELECT COUNT(DISTINCT u.id) > FROM userdata as u, userdata_history as h > WHERE h.id = '18181' > AND h.id = u.id; > > QUERY PLAN > ------------------------------------------------------------------------ > ------------------------------------------------------------------------ > Aggregate (cost=0.02..0.02 rows=1 width=8) (actual > time=298321.421..298321.422 rows=1 loops=1) > -> Nested Loop (cost=0.00..0.01 rows=1 width=8) (actual > time=1.771..298305.531 rows=2452 loops=1) > Join Filter: ("inner".id = "outer".id) > -> Seq Scan on userdata u (cost=0.00..0.00 rows=1 width=8) > (actual time=0.026..11.869 rows=2452 loops=1) > -> Seq Scan on userdata_history h (cost=0.00..0.00 rows=1 > width=8) (actual time=0.005..70.519 rows=41631 loops=2452) > Filter: (id = 18181::bigint) > Total runtime: 298321.926 ms > (7 rows) > > userdata has a primary/foreign key on id, which references > userdata_history.id, which is a primary key. > > At the time of analysis, the userdata table had < 2,500 rows. > userdata_history had < 50,000 rows. I can't imagine how even a seq scan > could result in a runtime of nearly 5 minutes in these circumstances. > > Also, doing a count( * ) from each table individually returns nearly > instantly. > > I can provide details of postgresql.conf and kernel settings if > necessary, but I'm using some pretty well tested settings that I use > any time I admin a postgres installation these days based on box > resources and database size. I'm more interested in knowing if there > are any bird's eye details I should be checking immediately. > > Thanks. > > -tfo > > -- > Thomas F. O'Connell > Co-Founder, Information Architect > Sitening, LLC > http://www.sitening.com/ > 110 30th Avenue North, Suite 6 > Nashville, TN 37203-6320 > 615-260-0005 Is your enable_seqscan set to true? Try it after issuing set enable_seqscan to off; From pgsql-performance-owner@postgresql.org Thu Oct 21 22:05:27 2004 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 BDDDC32A540 for ; Thu, 21 Oct 2004 22:05:26 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 65708-01 for ; Thu, 21 Oct 2004 21:05:22 +0000 (GMT) Received: from zigo.dhs.org (as2-4-3.an.g.bonet.se [194.236.34.191]) by svr1.postgresql.org (Postfix) with ESMTP id D8A2A329F05 for ; Thu, 21 Oct 2004 22:05:22 +0100 (BST) Received: from zigo.zigo.dhs.org (zigo.zigo.dhs.org [192.168.0.1]) by zigo.dhs.org (Postfix) with ESMTP id D74E38467; Thu, 21 Oct 2004 23:05:20 +0200 (CEST) Date: Thu, 21 Oct 2004 23:05:20 +0200 (CEST) From: Dennis Bjorklund To: "Thomas F.O'Connell" Cc: PgSQL - Performance Subject: Re: Performance Anomalies in 7.4.5 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/366 X-Sequence-Number: 8842 On Thu, 21 Oct 2004, Thomas F.O'Connell wrote: > Aggregate (cost=0.02..0.02 rows=1 width=8) (actual > time=298321.421..298321.422 rows=1 loops=1) > -> Nested Loop (cost=0.00..0.01 rows=1 width=8) (actual > time=1.771..298305.531 rows=2452 loops=1) > Join Filter: ("inner".id = "outer".id) > -> Seq Scan on userdata u (cost=0.00..0.00 rows=1 width=8) > (actual time=0.026..11.869 rows=2452 loops=1) > -> Seq Scan on userdata_history h (cost=0.00..0.00 rows=1 > width=8) (actual time=0.005..70.519 rows=41631 loops=2452) > Filter: (id = 18181::bigint) It looks like you have not run ANALYZE recently. Most people run VACUUM ANALYZE every night (or similar) in a cron job. -- /Dennis Bj�rklund From pgsql-performance-owner@postgresql.org Thu Oct 21 22:11:58 2004 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 A3D71329F05 for ; Thu, 21 Oct 2004 22:11:56 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 67153-02 for ; Thu, 21 Oct 2004 21:11:44 +0000 (GMT) Received: from window.monsterlabs.com (window.monsterlabs.com [216.183.105.176]) by svr1.postgresql.org (Postfix) with SMTP id CBABA32A2BF for ; Thu, 21 Oct 2004 22:11:44 +0100 (BST) Received: (qmail 29794 invoked from network); 21 Oct 2004 21:11:43 -0000 Received: from w080.z064003242.bna-tn.dsl.cnc.net (HELO ?192.168.1.22?) (64.3.242.80) by 0 with SMTP; 21 Oct 2004 21:11:43 -0000 In-Reply-To: References: Mime-Version: 1.0 (Apple Message framework v619) Content-Type: text/plain; charset=ISO-8859-1; format=flowed Message-Id: Content-Transfer-Encoding: quoted-printable Cc: PgSQL - Performance From: Thomas F.O'Connell Subject: Re: Performance Anomalies in 7.4.5 Date: Thu, 21 Oct 2004 16:11:39 -0500 To: Dennis Bjorklund X-Mailer: Apple Mail (2.619) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/367 X-Sequence-Number: 8843 The irony is that I had just disabled pg_autovacuum the previous day=20 during analysis of a wider issue affecting imports of data into the=20 system. -tfo -- Thomas F. O'Connell Co-Founder, Information Architect Sitening, LLC http://www.sitening.com/ 110 30th Avenue North, Suite 6 Nashville, TN 37203-6320 615-260-0005 On Oct 21, 2004, at 4:05 PM, Dennis Bjorklund wrote: > On Thu, 21 Oct 2004, Thomas F.O'Connell wrote: > >> Aggregate (cost=3D0.02..0.02 rows=3D1 width=3D8) (actual >> time=3D298321.421..298321.422 rows=3D1 loops=3D1) >> -> Nested Loop (cost=3D0.00..0.01 rows=3D1 width=3D8) (actual >> time=3D1.771..298305.531 rows=3D2452 loops=3D1) >> Join Filter: ("inner".id =3D "outer".id) >> -> Seq Scan on userdata u (cost=3D0.00..0.00 rows=3D1 width= =3D8) >> (actual time=3D0.026..11.869 rows=3D2452 loops=3D1) >> -> Seq Scan on userdata_history h (cost=3D0.00..0.00 rows=3D1 >> width=3D8) (actual time=3D0.005..70.519 rows=3D41631 loops=3D2452) >> Filter: (id =3D 18181::bigint) > > It looks like you have not run ANALYZE recently. Most people run VACUUM > ANALYZE every night (or similar) in a cron job. > > --=20 > /Dennis Bj=F6rklund From pgsql-performance-owner@postgresql.org Thu Oct 21 22:53:29 2004 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 94C3B32A6C8 for ; Thu, 21 Oct 2004 22:53:28 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 79473-04 for ; Thu, 21 Oct 2004 21:53:18 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id CF0B832A65B for ; Thu, 21 Oct 2004 22:53:19 +0100 (BST) 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 i9LLrJ1b025229; Thu, 21 Oct 2004 17:53:19 -0400 (EDT) To: "Thomas F.O'Connell" Cc: PgSQL - Performance Subject: Re: Performance Anomalies in 7.4.5 In-reply-to: References: Comments: In-reply-to "Thomas F.O'Connell" message dated "Thu, 21 Oct 2004 15:36:02 -0500" Date: Thu, 21 Oct 2004 17:53:19 -0400 Message-ID: <25228.1098395599@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/368 X-Sequence-Number: 8844 "Thomas F.O'Connell" writes: > -> Nested Loop (cost=0.00..0.01 rows=1 width=8) (actual > time=1.771..298305.531 rows=2452 loops=1) > Join Filter: ("inner".id = "outer".id) > -> Seq Scan on userdata u (cost=0.00..0.00 rows=1 width=8) > (actual time=0.026..11.869 rows=2452 loops=1) > -> Seq Scan on userdata_history h (cost=0.00..0.00 rows=1 > width=8) (actual time=0.005..70.519 rows=41631 loops=2452) > Filter: (id = 18181::bigint) > Total runtime: 298321.926 ms > (7 rows) What's killing you here is that the planner thinks these tables are completely empty (notice the zero cost estimates, which implies the table has zero blocks --- the fact that the rows estimate is 1 and not 0 is the result of sanity-check clamping inside costsize.c). This leads it to choose a nestloop, which would be the best plan if there were only a few rows involved, but it degenerates rapidly when there are not. It's easy to fall into this trap when truncating and reloading tables; all you need is an "analyze" while the table is empty. The rule of thumb is to analyze just after you reload the table, not just before. I'm getting more and more convinced that we need to drop the reltuples and relpages entries in pg_class, in favor of checking the physical table size whenever we make a plan. We could derive the tuple count estimate by having ANALYZE store a tuples-per-page estimate in pg_class and then multiply by the current table size; tuples-per-page should be a much more stable figure than total tuple count. One drawback to this is that it would require an additional lseek per table while planning, but that doesn't seem like a huge penalty. Probably the most severe objection to doing things this way is that the selected plan could change unexpectedly as a result of the physical table size changing. Right now the DBA can keep tight rein on actions that might affect plan selection (ie, VACUUM and ANALYZE), but that would go by the board with this. OTOH, we seem to be moving towards autovacuum, which also takes away any guarantees in this department. In any case this is speculation for 8.1; I think it's too late for 8.0. regards, tom lane From pgsql-performance-owner@postgresql.org Thu Oct 21 23:11:15 2004 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 758A432A773 for ; Thu, 21 Oct 2004 23:11:14 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83591-09 for ; Thu, 21 Oct 2004 22:11:02 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [66.143.173.58]) by svr1.postgresql.org (Postfix) with ESMTP id 95CA032A774 for ; Thu, 21 Oct 2004 23:11:03 +0100 (BST) Received: by flake.decibel.org (Postfix, from userid 1001) id 9DCE61C8F3; Thu, 21 Oct 2004 22:11:00 +0000 (GMT) Date: Thu, 21 Oct 2004 17:11:00 -0500 From: "Jim C. Nasby" To: Steve Atkins Cc: pgsql-performance@postgresql.org Subject: Re: Does PostgreSQL run with Oracle? Message-ID: <20041021221100.GI68407@decibel.org> References: <20041015171948.GA14759@gp.word-to-the-wise.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20041015171948.GA14759@gp.word-to-the-wise.com> X-Operating-System: FreeBSD 4.10-RELEASE-p3 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/369 X-Sequence-Number: 8845 On Fri, Oct 15, 2004 at 10:19:48AM -0700, Steve Atkins wrote: > On Fri, Oct 15, 2004 at 11:54:44AM -0500, Richard_D_Levine@raytheon.com wrote: > > My basic question to the community is "is PostgreSQL approximately as fast > > as Oracle?" > > > I'm currently running single processor UltraSPARC workstations, and intend > > to use Intel Arch laptops and Linux. The application is a big turnkey > > workstation app. I know the hardware switch alone will enhance > > performance, and may do so to the point where even a slower database will > > still be adequate. > > I have found that PostgreSQL seems to perform poorly on Solaris/SPARC > (less so after recent improvements, but still...) compared to x86 > systems - more so than the delta between Oracle on the two platforms. > Just a gut impression, but it might mean that comparing the two > databases on SPARC may not be that useful comparison if you're > planning to move to x86. As a point of reference, an IBM hardware sales rep I worked with a few years ago told me that he got a lot of sales from Oracle shops that were running Sun and switched to RS/6000. Basically, for a given workload, it would take 2x the number of Sun CPUs as RS/6000 CPUs. The difference in Oracle licensing costs was usually enough to pay for the new hardware in one year. -- 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 Thu Oct 21 23:15:11 2004 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 F355932A7B8 for ; Thu, 21 Oct 2004 23:15:08 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 85537-04 for ; Thu, 21 Oct 2004 22:14:56 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [66.143.173.58]) by svr1.postgresql.org (Postfix) with ESMTP id C07BD32A7A8 for ; Thu, 21 Oct 2004 23:14:57 +0100 (BST) Received: by flake.decibel.org (Postfix, from userid 1001) id 3D6851C8F3; Thu, 21 Oct 2004 22:14:57 +0000 (GMT) Date: Thu, 21 Oct 2004 17:14:57 -0500 From: "Jim C. Nasby" To: Josh Berkus Cc: pgsql-performance@postgresql.org, Matt Clark , Tom Fischer Subject: Re: OS desicion Message-ID: <20041021221457.GJ68407@decibel.org> References: <20041020151015.766bb00c@nixe> <417666C9.80008@ymogen.net> <200410200938.51391.josh@agliodbs.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <200410200938.51391.josh@agliodbs.com> X-Operating-System: FreeBSD 4.10-RELEASE-p3 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/370 X-Sequence-Number: 8846 On Wed, Oct 20, 2004 at 09:38:51AM -0700, Josh Berkus wrote: > Tom, > > > You are asking the wrong question. The best OS is the OS you (and/or > > the customer) knows and can administer competently. > > I'll have to 2nd this. I'll 3rd but add one tidbit: FreeBSD will schedule disk I/O based on process priority, while linux won't. This can be very handy for things like vacuum. -- 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 Thu Oct 21 23:20:19 2004 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 73BDC32A014 for ; Thu, 21 Oct 2004 23:18:41 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 86773-05 for ; Thu, 21 Oct 2004 22:18:29 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [66.143.173.58]) by svr1.postgresql.org (Postfix) with ESMTP id BEBB1329E15 for ; Thu, 21 Oct 2004 23:18:31 +0100 (BST) Received: by flake.decibel.org (Postfix, from userid 1001) id A12511C8F7; Thu, 21 Oct 2004 22:18:30 +0000 (GMT) Date: Thu, 21 Oct 2004 17:18:30 -0500 From: "Jim C. Nasby" To: pgsql-performance@postgresql.org Subject: Re: Anything to be gained from a 'Postgres Filesystem'? Message-ID: <20041021221830.GK68407@decibel.org> References: <008601c4b743$aff57f10$8300a8c0@solent> <20041021102727.GB586@uio.no> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20041021102727.GB586@uio.no> X-Operating-System: FreeBSD 4.10-RELEASE-p3 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/371 X-Sequence-Number: 8847 Note that most people are now moving away from raw devices for databases in most applicaitons. The relatively small performance gain isn't worth the hassles. On Thu, Oct 21, 2004 at 12:27:27PM +0200, Steinar H. Gunderson wrote: > On Thu, Oct 21, 2004 at 08:58:01AM +0100, Matt Clark wrote: > > I suppose I'm just idly wondering really. Clearly it's against PG > > philosophy to build an FS or direct IO management into PG, but now it's so > > relatively easy to plug filesystems into the main open-source Oses, It > > struck me that there might be some useful changes to, say, XFS or ext3, that > > could be made that would help PG out. > > This really sounds like a poor replacement for just making PostgreSQL use raw > devices to me. (I have no idea why that isn't done already, but presumably it > isn't all that easy to get right. :-) ) > > /* Steinar */ > -- > Homepage: http://www.sesse.net/ > > ---------------------------(end of broadcast)--------------------------- > TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org > -- 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 Thu Oct 21 23:21:43 2004 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 2DE3232A7BC for ; Thu, 21 Oct 2004 23:21:39 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 86184-10 for ; Thu, 21 Oct 2004 22:21:34 +0000 (GMT) Received: from server.gpdnet.co.uk (gpdnet.plus.com [212.56.100.243]) by svr1.postgresql.org (Postfix) with ESMTP id 859C532A79B for ; Thu, 21 Oct 2004 23:21:35 +0100 (BST) Received: from gary (unknown [192.168.1.2]) by server.gpdnet.co.uk (Postfix) with ESMTP id 38221FBBB7 for ; Thu, 21 Oct 2004 23:22:11 +0100 (BST) From: "Gary Doades" To: pgsql-performance@postgresql.org Date: Thu, 21 Oct 2004 23:21:35 +0100 MIME-Version: 1.0 Subject: Re: Performance Anomalies in 7.4.5 Message-ID: <4178447F.961.6D7BBB5B@localhost> In-reply-to: References: X-mailer: Pegasus Mail for Windows (v4.12a) Content-type: text/plain; charset=US-ASCII Content-transfer-encoding: 7BIT Content-description: Mail message body X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/372 X-Sequence-Number: 8848 On 21 Oct 2004 at 15:50, Thomas F.O'Connell wrote: > If not, should I be REINDEXing manually, as well as VACUUMing manually > after large data imports (whether via COPY or INSERT)? Or will a VACUUM > FULL ANALYZE be enough? > It's not the vacuuming that's important here, just the analyze. If you import any data into a table, Postgres often does not *know* that until you gather the statistics on the table. You are simply running into the problem of the planner not knowing how much data/distribution of data in your tables. If you have large imports it may be faster overall to drop the indexes first, then insert the data, then put the indexes back on, then analyze. Cheers, Gary. From pgsql-performance-owner@postgresql.org Fri Oct 22 04:15:10 2004 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 48BDD32B22A for ; Fri, 22 Oct 2004 04:15:08 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 61668-02 for ; Fri, 22 Oct 2004 03:14:55 +0000 (GMT) Received: from rproxy.gmail.com (rproxy.gmail.com [64.233.170.198]) by svr1.postgresql.org (Postfix) with ESMTP id 5C5E132A33D for ; Fri, 22 Oct 2004 04:14:58 +0100 (BST) Received: by rproxy.gmail.com with SMTP id 77so140208rnk for ; Thu, 21 Oct 2004 20:14:59 -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; b=W68+CTdasbZBZnZb+NTQQjE3pIm8JZ2N0KCS1SZO5xLQl9uJ8kR8d8s4tirp+9VXBYtG7gpTUxXOUrNAGPChLqTJZMa3jZLSLEGYDWcbW3isDE1h1O8COe6foYIGBXS8L5CmAZksehXRO9mJkwDxRLZ9NwkhFir/UzU1CVBRxAA= Received: by 10.38.14.9 with SMTP id 9mr3188075rnn; Thu, 21 Oct 2004 20:14:59 -0700 (PDT) Received: by 10.38.77.73 with HTTP; Thu, 21 Oct 2004 20:14:59 -0700 (PDT) Message-ID: <38242de904102120143c55ba34@mail.gmail.com> Date: Thu, 21 Oct 2004 21:14:59 -0600 From: Joshua Marsh Reply-To: Joshua Marsh To: pgsql-performance@postgresql.org Subject: Large Database Performance suggestions 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/373 X-Sequence-Number: 8849 Hello everyone, I am currently working on a data project that uses PostgreSQL extensively to store, manage and maintain the data. We haven't had any problems regarding database size until recently. The three major tables we use never get bigger than 10 million records. With this size, we can do things like storing the indexes or even the tables in memory to allow faster access. Recently, we have found customers who are wanting to use our service with data files between 100 million and 300 million records. At that size, each of the three major tables will hold between 150 million and 700 million records. At this size, I can't expect it to run queries in 10-15 seconds (what we can do with 10 million records), but would prefer to keep them all under a minute. We did some original testing and with a server with 8GB or RAM and found we can do operations on data file up to 50 million fairly well, but performance drop dramatically after that. Does anyone have any suggestions on a good way to improve performance for these extra large tables? Things that have come to mind are Replication and Beowulf clusters, but from what I have recently studied, these don't do so wel with singular processes. We will have parallel process running, but it's more important that the speed of each process be faster than several parallel processes at once. Any help would be greatly appreciated! Thanks, Joshua Marsh P.S. Off-topic, I have a few invitations to gmail. If anyone would like one, let me know. From pgsql-performance-owner@postgresql.org Fri Oct 22 04:30:21 2004 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 3F1A132B22A for ; Fri, 22 Oct 2004 04:30:20 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 63835-05 for ; Fri, 22 Oct 2004 03:30:07 +0000 (GMT) Received: from linuxworld.com.au (unknown [203.34.46.50]) by svr1.postgresql.org (Postfix) with ESMTP id 60C9A32A681 for ; Fri, 22 Oct 2004 04:30:08 +0100 (BST) Received: from localhost (swm@localhost) by linuxworld.com.au (8.11.6/8.11.4) with ESMTP id i9M3TvN31573; Fri, 22 Oct 2004 13:29:57 +1000 Date: Fri, 22 Oct 2004 13:29:57 +1000 (EST) From: Gavin Sherry X-X-Sender: swm@linuxworld.com.au To: Joshua Marsh Cc: pgsql-performance@postgresql.org Subject: Re: Large Database Performance suggestions In-Reply-To: <38242de904102120143c55ba34@mail.gmail.com> Message-ID: References: <38242de904102120143c55ba34@mail.gmail.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/374 X-Sequence-Number: 8850 On Thu, 21 Oct 2004, Joshua Marsh wrote: > Recently, we have found customers who are wanting to use our service > with data files between 100 million and 300 million records. At that > size, each of the three major tables will hold between 150 million and > 700 million records. At this size, I can't expect it to run queries > in 10-15 seconds (what we can do with 10 million records), but would > prefer to keep them all under a minute. To provide any useful information, we'd need to look at your table schemas and sample queries. The values for sort_mem and shared_buffers will also be useful. Are you VACUUMing and ANALYZEing? (or is the data read only?)) gavin From pgsql-performance-owner@postgresql.org Fri Oct 22 04:37:05 2004 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 6EF9932B22F for ; Fri, 22 Oct 2004 04:37:04 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 65995-04 for ; Fri, 22 Oct 2004 03:36:57 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id D460832A33D for ; Fri, 22 Oct 2004 04:37:00 +0100 (BST) 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 i9M3b1CX002569; Thu, 21 Oct 2004 23:37:01 -0400 (EDT) To: Joshua Marsh Cc: pgsql-performance@postgresql.org Subject: Re: Large Database Performance suggestions In-reply-to: <38242de904102120143c55ba34@mail.gmail.com> References: <38242de904102120143c55ba34@mail.gmail.com> Comments: In-reply-to Joshua Marsh message dated "Thu, 21 Oct 2004 21:14:59 -0600" Date: Thu, 21 Oct 2004 23:37:00 -0400 Message-ID: <2568.1098416220@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/375 X-Sequence-Number: 8851 Joshua Marsh writes: > ... We did some original testing and with a server with 8GB or RAM and > found we can do operations on data file up to 50 million fairly well, > but performance drop dramatically after that. What you have to ask is *why* does it drop dramatically? There aren't any inherent limits in Postgres that are going to kick in at that level. I'm suspicious that you could improve the situation by adjusting sort_mem and/or other configuration parameters; but there's not enough info here to make specific recommendations. I would suggest posting EXPLAIN ANALYZE results for your most important queries both in the size range where you are getting good results, and the range where you are not. Then we'd have something to chew on. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 22 05:13:04 2004 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 BF70A329E73 for ; Fri, 22 Oct 2004 05:13:01 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 74172-03 for ; Fri, 22 Oct 2004 04:12:48 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 1B57A32A154 for ; Fri, 22 Oct 2004 05:12:48 +0100 (BST) 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 i9M4CiNN002824; Fri, 22 Oct 2004 00:12:44 -0400 (EDT) To: Sean Chittenden Cc: Kevin Brown , pgsql-performance@postgresql.org Subject: Re: mmap (was First set of OSDL Shared Mem scalability results, some wierdness ... In-reply-to: References: <157f64840410141725162e43b5@mail.gmail.com> <20041015010841.GE665@filer> <0E3BB9CB-1EE6-11D9-A0BB-000A95C705DC@chittenden.org> <26390.1097875346@sss.pgh.pa.us> Comments: In-reply-to Sean Chittenden message dated "Thu, 21 Oct 2004 13:29:34 -0700" Date: Fri, 22 Oct 2004 00:12:44 -0400 Message-ID: <2823.1098418364@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/376 X-Sequence-Number: 8852 Sean Chittenden writes: > When a backend wishes to write a page, the following steps are taken: > ... > 2) Backend mmap(2)'s a second copy of the page(s) being written to, > this time with the MAP_PRIVATE flag set. > ... > 5) Once the WAL logging is complete and it has hit the disk, the > backend msync(2)'s its private copy of the pages to disk (ASYNC or > SYNC, it doesn't really matter too much to me). My man page for mmap says that changes in a MAP_PRIVATE region are private; they do not affect the file at all, msync or no. So I don't think the above actually works. In any case, this scheme still forces you to flush WAL records to disk before making the changed page visible to other backends, so I don't see how it improves the situation. In the existing scheme we only have to fsync WAL at (1) transaction commit, (2) when we are forced to write a page out from shared buffers because we are short of buffers, or (3) checkpoint. Anything that implies an fsync per atomic action is going to be a loser. It does not matter how great your kernel API is if you only get to perform one atomic action per disk rotation :-( The important point here is that you can't postpone making changes at the page level visible to other backends; there's no MVCC at this level. Consider for example two backends wanting to insert a new row. If they both MAP_PRIVATE the same page, they'll probably choose the same tuple slot on the page to insert into (certainly there is nothing to stop that from happening). Now you have conflicting definitions for the same CTID, not to mention probably conflicting uses of the page's physical free space; disaster ensues. So "atomic action" really means "lock page, make changes, add WAL record to in-memory WAL buffers, unlock page" with the understanding that as soon as you unlock the page the changes you've made in it are visible to all other backends. You *can't* afford to put a WAL fsync in this sequence. You could possibly buy back most of the lossage in this scenario if there were some efficient way for a backend to hold the low-level lock on a page just until some other backend wanted to modify the page; whereupon the previous owner would have to do what's needed to make his changes visible before releasing the lock. Given the right access patterns you don't have to fsync very often (though given the wrong access patterns you're still in deep trouble). But we don't have any such mechanism and I think the communication costs of one would be forbidding. > [ much snipped ] > 4) Not having shared pages get lost when the backend dies (mmap(2) uses > refcounts and cleans itself up, no need for ipcs/ipcrm/ipcclean). Actually, that is not a bug that's a feature. One of the things that scares me about mmap is that a crashing backend is able to scribble all over live disk buffers before it finally SEGV's (think about memcpy gone wrong and similar cases). In our existing scheme there's a pretty good chance that we will be able to commit hara-kiri before any of the trashed data gets written out. In an mmap scheme, it's time to dig out your backup tapes, because there simply is no distinction between transient and permanent data --- the kernel has no way to know that you didn't mean it. In short, I remain entirely unconvinced that mmap is of any interest to us. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 22 06:01:35 2004 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 6BC2932B2B3 for ; Fri, 22 Oct 2004 06:01:34 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 85741-02 for ; Fri, 22 Oct 2004 05:01:15 +0000 (GMT) Received: from mpls-qmqp-03.inet.qwest.net (mpls-qmqp-03.inet.qwest.net [63.231.195.114]) by svr1.postgresql.org (Postfix) with SMTP id 0DC0232B2A9 for ; Fri, 22 Oct 2004 06:01:18 +0100 (BST) Received: (qmail 61471 invoked by uid 0); 22 Oct 2004 05:01:19 -0000 Received: from mpls-pop-14.inet.qwest.net (63.231.195.14) by mpls-qmqp-03.inet.qwest.net with QMQP; 22 Oct 2004 05:01:19 -0000 Received: from 63-227-127-37.dnvr.qwest.net (HELO ?10.0.0.2?) (63.227.127.37) by mpls-pop-14.inet.qwest.net with SMTP; 22 Oct 2004 05:01:18 -0000 Date: Thu, 21 Oct 2004 23:03:34 -0600 Message-Id: <1098421413.21035.78.camel@localhost.localdomain> From: "Scott Marlowe" To: "Joshua Marsh" Cc: pgsql-performance@postgresql.org Subject: Re: Large Database Performance suggestions In-Reply-To: <38242de904102120143c55ba34@mail.gmail.com> References: <38242de904102120143c55ba34@mail.gmail.com> Content-Type: text/plain Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/377 X-Sequence-Number: 8853 On Thu, 2004-10-21 at 21:14, Joshua Marsh wrote: > Hello everyone, > > I am currently working on a data project that uses PostgreSQL > extensively to store, manage and maintain the data. We haven't had > any problems regarding database size until recently. The three major > tables we use never get bigger than 10 million records. With this > size, we can do things like storing the indexes or even the tables in > memory to allow faster access. > > Recently, we have found customers who are wanting to use our service > with data files between 100 million and 300 million records. At that > size, each of the three major tables will hold between 150 million and > 700 million records. At this size, I can't expect it to run queries > in 10-15 seconds (what we can do with 10 million records), but would > prefer to keep them all under a minute. > > We did some original testing and with a server with 8GB or RAM and > found we can do operations on data file up to 50 million fairly well, > but performance drop dramatically after that. Does anyone have any > suggestions on a good way to improve performance for these extra large > tables? Things that have come to mind are Replication and Beowulf > clusters, but from what I have recently studied, these don't do so wel > with singular processes. We will have parallel process running, but > it's more important that the speed of each process be faster than > several parallel processes at once. I'd assume that what's happening is that up to a certain data set size, it all fits in memory, and you're going from CPU/memory bandwidth limited to I/O limited. If this is the case, then a faster storage subsystem is the only real answer. If the database is mostly read, then a large RAID5 or RAID 1+0 array should help quite a bit. You might wanna post some explain analyze of the queries that are going slower at some point in size, along with schema for those tables etc... From pgsql-performance-owner@postgresql.org Fri Oct 22 12:38:37 2004 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 8BBA6329E55 for ; Fri, 22 Oct 2004 12:38:36 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 83997-10 for ; Fri, 22 Oct 2004 11:38:20 +0000 (GMT) Received: from net2.micro-automation.com (net2.micro-automation.com [64.7.141.29]) by svr1.postgresql.org (Postfix) with SMTP id A1CE532A2F9 for ; Fri, 22 Oct 2004 12:38:26 +0100 (BST) Received: (qmail 28191 invoked from network); 22 Oct 2004 11:39:50 -0000 Received: from dcdsl.ebox.com (HELO ?192.168.1.36?) (davec@64.7.143.116) by net2.micro-automation.com with SMTP; 22 Oct 2004 11:39:50 -0000 Message-ID: <4178F149.9090408@fastcrypt.com> Date: Fri, 22 Oct 2004 07:38:49 -0400 From: Dave Cramer Reply-To: pg@fastcrypt.com Organization: Postgres International User-Agent: Mozilla Thunderbird 0.8 (X11/20041001) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Joshua Marsh Cc: pgsql-performance@postgresql.org Subject: Re: Large Database Performance suggestions References: <38242de904102120143c55ba34@mail.gmail.com> In-Reply-To: <38242de904102120143c55ba34@mail.gmail.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/378 X-Sequence-Number: 8854 Josh, Your hardware setup would be useful too. It's surprising how slow some big name servers really are. If you are seriously considering memory sizes over 4G you may want to look at an opteron. Dave Joshua Marsh wrote: >Hello everyone, > >I am currently working on a data project that uses PostgreSQL >extensively to store, manage and maintain the data. We haven't had >any problems regarding database size until recently. The three major >tables we use never get bigger than 10 million records. With this >size, we can do things like storing the indexes or even the tables in >memory to allow faster access. > >Recently, we have found customers who are wanting to use our service >with data files between 100 million and 300 million records. At that >size, each of the three major tables will hold between 150 million and >700 million records. At this size, I can't expect it to run queries >in 10-15 seconds (what we can do with 10 million records), but would >prefer to keep them all under a minute. > >We did some original testing and with a server with 8GB or RAM and >found we can do operations on data file up to 50 million fairly well, >but performance drop dramatically after that. Does anyone have any >suggestions on a good way to improve performance for these extra large >tables? Things that have come to mind are Replication and Beowulf >clusters, but from what I have recently studied, these don't do so wel >with singular processes. We will have parallel process running, but >it's more important that the speed of each process be faster than >several parallel processes at once. > >Any help would be greatly appreciated! > >Thanks, > >Joshua Marsh > >P.S. Off-topic, I have a few invitations to gmail. If anyone would >like one, let me know. > >---------------------------(end of broadcast)--------------------------- >TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org > > > > -- Dave Cramer http://www.postgresintl.com 519 939 0336 ICQ#14675561 From pgsql-hackers-owner@postgresql.org Fri Oct 22 19:53:02 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7ED0DEAEDAB; Fri, 22 Oct 2004 19:52:57 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 25645-02; Fri, 22 Oct 2004 18:53:00 +0000 (GMT) Received: from cmailm3.svr.pol.co.uk (cmailm3.svr.pol.co.uk [195.92.193.19]) by svr1.postgresql.org (Postfix) with ESMTP id DAF64EAED89; Fri, 22 Oct 2004 19:52:51 +0100 (BST) Received: from modem-910.orangutan.dialup.pol.co.uk ([217.135.227.142] helo=[192.168.0.102]) by cmailm3.svr.pol.co.uk with esmtp (Exim 4.41) id 1CL4XE-0003h7-N3; Fri, 22 Oct 2004 19:53:09 +0100 Subject: ARC Memory Usage analysis From: Simon Riggs To: pgsql-patches@postgresql.org, pgsql-hackers@postgresql.org Cc: josh@agliodbs.com Content-Type: multipart/mixed; boundary="=-yt9apKN4oP+msuzo0upS" Organization: 2nd Quadrant Message-Id: <1098471059.20926.49.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Date: Fri, 22 Oct 2004 19:50:59 +0100 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/752 X-Sequence-Number: 60255 --=-yt9apKN4oP+msuzo0upS Content-Type: text/plain Content-Transfer-Encoding: 7bit I've been using the ARC debug options to analyse memory usage on the PostgreSQL 8.0 server. This is a precursor to more complex performance analysis work on the OSDL test suite. I've simplified some of the ARC reporting into a single log line, which is enclosed here as a patch on freelist.c. This includes reporting of: - the total memory in use, which wasn't previously reported - the cache hit ratio, which was slightly incorrectly calculated - a useful-ish value for looking at the "B" lists in ARC (This is a patch against cvstip, but I'm not sure whether this has potential for inclusion in 8.0...) The total memory in use is useful because it allows you to tell whether shared_buffers is set too high. If it is set too high, then memory usage will continue to grow slowly up to the max, without any corresponding increase in cache hit ratio. If shared_buffers is too small, then memory usage will climb quickly and linearly to its maximum. The last one I've called "turbulence" in an attempt to ascribe some useful meaning to B1/B2 hits - I've tried a few other measures though without much success. Turbulence is the hit ratio of B1+B2 lists added together. By observation, this is zero when ARC gives smooth operation, and goes above zero otherwise. Typically, turbulence occurs when shared_buffers is too small for the working set of the database/workload combination and ARC repeatedly re-balances the lengths of T1/T2 as a result of "near-misses" on the B1/B2 lists. Turbulence doesn't usually cut in until the cache is fully utilized, so there is usually some delay after startup. We also recently discussed that I would add some further memory analysis features for 8.1, so I've been trying to figure out how. The idea that B1, B2 represent something really useful doesn't seem to have been borne out - though I'm open to persuasion there. I originally envisaged a "shadow list" operating in extension of the main ARC list. This will require some re-coding, since the variables and macros are all hard-coded to a single set of lists. No complaints, just it will take a little longer than we all thought (for me, that is...) My proposal is to alter the code to allow an array of memory linked lists. The actual list would be [0] - other additional lists would be created dynamically as required i.e. not using IFDEFs, since I want this to be controlled by a SIGHUP GUC to allow on-site tuning, not just lab work. This will then allow reporting against the additional lists, so that cache hit ratios can be seen with various other "prototype" shared_buffer settings. Any thoughts? -- Best Regards, Simon Riggs --=-yt9apKN4oP+msuzo0upS Content-Disposition: attachment; filename=freelist.patch Content-Type: text/x-patch; name=freelist.patch; charset=UTF-8 Content-Transfer-Encoding: 7bit Index: freelist.c =================================================================== RCS file: /projects/cvsroot/pgsql/src/backend/storage/buffer/freelist.c,v retrieving revision 1.48 diff -d -c -r1.48 freelist.c *** freelist.c 16 Sep 2004 16:58:31 -0000 1.48 --- freelist.c 22 Oct 2004 18:15:38 -0000 *************** *** 126,131 **** --- 126,133 ---- if (StrategyControl->stat_report + DebugSharedBuffers < now) { long all_hit, + buf_used, + b_hit, b1_hit, t1_hit, t2_hit, *************** *** 155,161 **** } if (StrategyControl->num_lookup == 0) ! all_hit = b1_hit = t1_hit = t2_hit = b2_hit = 0; else { b1_hit = (StrategyControl->num_hit[STRAT_LIST_B1] * 100 / --- 157,163 ---- } if (StrategyControl->num_lookup == 0) ! all_hit = buf_used = b_hit = b1_hit = t1_hit = t2_hit = b2_hit = 0; else { b1_hit = (StrategyControl->num_hit[STRAT_LIST_B1] * 100 / *************** *** 166,181 **** StrategyControl->num_lookup); b2_hit = (StrategyControl->num_hit[STRAT_LIST_B2] * 100 / StrategyControl->num_lookup); ! all_hit = b1_hit + t1_hit + t2_hit + b2_hit; } errcxtold = error_context_stack; error_context_stack = NULL; ! elog(DEBUG1, "ARC T1target=%5d B1len=%5d T1len=%5d T2len=%5d B2len=%5d", T1_TARGET, B1_LENGTH, T1_LENGTH, T2_LENGTH, B2_LENGTH); ! elog(DEBUG1, "ARC total =%4ld%% B1hit=%4ld%% T1hit=%4ld%% T2hit=%4ld%% B2hit=%4ld%%", all_hit, b1_hit, t1_hit, t2_hit, b2_hit); ! elog(DEBUG1, "ARC clean buffers at LRU T1= %5d T2= %5d", t1_clean, t2_clean); error_context_stack = errcxtold; --- 168,187 ---- StrategyControl->num_lookup); b2_hit = (StrategyControl->num_hit[STRAT_LIST_B2] * 100 / StrategyControl->num_lookup); ! all_hit = t1_hit + t2_hit; ! b_hit = b1_hit + b2_hit; ! buf_used = T1_LENGTH + T2_LENGTH; } errcxtold = error_context_stack; error_context_stack = NULL; ! elog(DEBUG1, "shared_buffers used=%8ld cache hits=%4ld%% turbulence=%4ld%%", ! buf_used, all_hit, b_hit); ! elog(DEBUG2, "ARC T1target=%5d B1len=%5d T1len=%5d T2len=%5d B2len=%5d", T1_TARGET, B1_LENGTH, T1_LENGTH, T2_LENGTH, B2_LENGTH); ! elog(DEBUG2, "ARC total =%4ld%% B1hit=%4ld%% T1hit=%4ld%% T2hit=%4ld%% B2hit=%4ld%%", all_hit, b1_hit, t1_hit, t2_hit, b2_hit); ! elog(DEBUG2, "ARC clean buffers at LRU T1= %5d T2= %5d", t1_clean, t2_clean); error_context_stack = errcxtold; --=-yt9apKN4oP+msuzo0upS-- From pgsql-performance-owner@postgresql.org Fri Oct 22 20:06:26 2004 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 7A36FEAEDBA for ; Fri, 22 Oct 2004 20:06:25 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 29194-05 for ; Fri, 22 Oct 2004 19:06:24 +0000 (GMT) Received: from exchange.musicreports.com (mail.musicreports.com [64.161.179.34]) by svr1.postgresql.org (Postfix) with ESMTP id 75110EAEDA4 for ; Fri, 22 Oct 2004 20:06:14 +0100 (BST) Received: from [192.168.20.70] (RG-HP-XP [192.168.20.70]) by exchange.musicreports.com with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2653.13) id SSYKJPJ2; Fri, 22 Oct 2004 12:07:11 -0700 Message-ID: <41795A35.4060802@paccomsys.com> Date: Fri, 22 Oct 2004 12:06:30 -0700 From: Roger Ging User-Agent: Mozilla Thunderbird 0.7.2 (Windows/20040707) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Slow query 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/379 X-Sequence-Number: 8855 The following query has never finished. I have let it run for over 24 hours. This is a one time update that is part of a conversion script from MSSQL data. All of the tables are freshly built and inserted into. I have not run explain analyze because it does not return in a reasonable time. Explain output is posted below. Any suggestions on syntax changes or anything else to improve this would be appreciated. Dual PIII 1Ghz 4 GB RAM 4 spindle IDE RAID 0 on LSI controller. Postgres 7.4.5 Linux version 2.6.3-7mdk-p3-smp-64GB postgresql.cong snip: tcpip_socket = true max_connections = 40 shared_buffers = 1000 sort_mem = 65536 fsync = true source_song_title +-10,500,000 rows source_song +-9,500,000 rows source_system 10 rows source_title +- 5,600,000 Code run right before this query: create index ssa_source_song_id on source_song_artist (source_song_id); analyze source_song_artist; create index sa_artist_id on source_artist (artist_id); analyze source_artist; create index ss_source_song_id on source_song (source_song_id); analyze source_song; create index st_title_id on source_title (title_id); analyze source_title; source_song.source_song_id = int4 source_song_title.source_song_id = int4 source_title.title_id = int4 source_song_title.title_id = int4 update source_song_title set source_song_title_id = nextval('source_song_title_seq') ,licensing_match_order = (select licensing_match_order from source_system where source_system_id = ss.source_system_id) ,affiliation_match_order = (select affiliation_match_order from source_system where source_system_id = ss.source_system_id) ,title = st.title from source_song_title sst join source_song ss on ss.source_song_id = sst.source_song_id join source_title st on st.title_id = sst.title_id where source_song_title.source_song_id = sst.source_song_id; Explain output: "Hash Join (cost=168589.60..16651072.43 rows=6386404 width=335)" " Hash Cond: ("outer".title_id = "inner".title_id)" " -> Merge Join (cost=0.00..1168310.61 rows=6386403 width=311)" " Merge Cond: ("outer".source_song_id = "inner".source_song_id)" " -> Merge Join (cost=0.00..679279.40 rows=6386403 width=16)" " Merge Cond: ("outer".source_song_id = "inner".source_song_id)" " -> Index Scan using source_song_title_pkey on source_song_title sst (cost=0.00..381779.37 rows=10968719 width=8)" " -> Index Scan using ss_source_song_id on source_song ss (cost=0.00..190583.36 rows=6386403 width=8)" " -> Index Scan using source_song_title_pkey on source_song_title (cost=0.00..381779.37 rows=10968719 width=303)" " -> Hash (cost=117112.08..117112.08 rows=5513808 width=32)" " -> Seq Scan on source_title st (cost=0.00..117112.08 rows=5513808 width=32)" " SubPlan" " -> Seq Scan on source_system (cost=0.00..1.14 rows=2 width=4)" " Filter: (source_system_id = $0)" " -> Seq Scan on source_system (cost=0.00..1.14 rows=2 width=2)" " Filter: (source_system_id = $0)" From pgsql-hackers-owner@postgresql.org Fri Oct 22 20:37:09 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 6E616EAE4A0 for ; Fri, 22 Oct 2004 20:37:07 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 39963-04 for ; Fri, 22 Oct 2004 19:37:05 +0000 (GMT) Received: from smtp016.mail.yahoo.com (smtp016.mail.yahoo.com [216.136.174.113]) by svr1.postgresql.org (Postfix) with SMTP id 4489EEAD274 for ; Fri, 22 Oct 2004 20:36:56 +0100 (BST) Received: from unknown (HELO jupiter.black-lion.info) (janwieck@68.80.245.191 with login) by smtp016.mail.yahoo.com with SMTP; 22 Oct 2004 19:37:15 -0000 Received: from [172.21.8.18] ([192.168.192.101]) (authenticated bits=0) by jupiter.black-lion.info (8.12.10/8.12.9) with ESMTP id i9MJbB94016993; Fri, 22 Oct 2004 15:37:12 -0400 (EDT) (envelope-from JanWieck@Yahoo.com) Message-ID: <41796115.2020501@Yahoo.com> Date: Fri, 22 Oct 2004 15:35:49 -0400 From: Jan Wieck User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.1) Gecko/20040707 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Simon Riggs Cc: pgsql-patches@postgresql.org, pgsql-hackers@postgresql.org, josh@agliodbs.com Subject: Re: ARC Memory Usage analysis References: <1098471059.20926.49.camel@localhost.localdomain> In-Reply-To: <1098471059.20926.49.camel@localhost.localdomain> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/755 X-Sequence-Number: 60258 On 10/22/2004 2:50 PM, Simon Riggs wrote: > I've been using the ARC debug options to analyse memory usage on the > PostgreSQL 8.0 server. This is a precursor to more complex performance > analysis work on the OSDL test suite. > > I've simplified some of the ARC reporting into a single log line, which > is enclosed here as a patch on freelist.c. This includes reporting of: > - the total memory in use, which wasn't previously reported > - the cache hit ratio, which was slightly incorrectly calculated > - a useful-ish value for looking at the "B" lists in ARC > (This is a patch against cvstip, but I'm not sure whether this has > potential for inclusion in 8.0...) > > The total memory in use is useful because it allows you to tell whether > shared_buffers is set too high. If it is set too high, then memory usage > will continue to grow slowly up to the max, without any corresponding > increase in cache hit ratio. If shared_buffers is too small, then memory > usage will climb quickly and linearly to its maximum. > > The last one I've called "turbulence" in an attempt to ascribe some > useful meaning to B1/B2 hits - I've tried a few other measures though > without much success. Turbulence is the hit ratio of B1+B2 lists added > together. By observation, this is zero when ARC gives smooth operation, > and goes above zero otherwise. Typically, turbulence occurs when > shared_buffers is too small for the working set of the database/workload > combination and ARC repeatedly re-balances the lengths of T1/T2 as a > result of "near-misses" on the B1/B2 lists. Turbulence doesn't usually > cut in until the cache is fully utilized, so there is usually some delay > after startup. > > We also recently discussed that I would add some further memory analysis > features for 8.1, so I've been trying to figure out how. > > The idea that B1, B2 represent something really useful doesn't seem to > have been borne out - though I'm open to persuasion there. > > I originally envisaged a "shadow list" operating in extension of the > main ARC list. This will require some re-coding, since the variables and > macros are all hard-coded to a single set of lists. No complaints, just > it will take a little longer than we all thought (for me, that is...) > > My proposal is to alter the code to allow an array of memory linked > lists. The actual list would be [0] - other additional lists would be > created dynamically as required i.e. not using IFDEFs, since I want this > to be controlled by a SIGHUP GUC to allow on-site tuning, not just lab > work. This will then allow reporting against the additional lists, so > that cache hit ratios can be seen with various other "prototype" > shared_buffer settings. All the existing lists live in shared memory, so that dynamic approach suffers from the fact that the memory has to be allocated during ipc_init. What do you think about my other theory to make C actually 2x effective cache size and NOT to keep T1 in shared buffers but to assume T1 lives in the OS buffer cache? Jan > > Any thoughts? > > > > ------------------------------------------------------------------------ > > Index: freelist.c > =================================================================== > RCS file: /projects/cvsroot/pgsql/src/backend/storage/buffer/freelist.c,v > retrieving revision 1.48 > diff -d -c -r1.48 freelist.c > *** freelist.c 16 Sep 2004 16:58:31 -0000 1.48 > --- freelist.c 22 Oct 2004 18:15:38 -0000 > *************** > *** 126,131 **** > --- 126,133 ---- > if (StrategyControl->stat_report + DebugSharedBuffers < now) > { > long all_hit, > + buf_used, > + b_hit, > b1_hit, > t1_hit, > t2_hit, > *************** > *** 155,161 **** > } > > if (StrategyControl->num_lookup == 0) > ! all_hit = b1_hit = t1_hit = t2_hit = b2_hit = 0; > else > { > b1_hit = (StrategyControl->num_hit[STRAT_LIST_B1] * 100 / > --- 157,163 ---- > } > > if (StrategyControl->num_lookup == 0) > ! all_hit = buf_used = b_hit = b1_hit = t1_hit = t2_hit = b2_hit = 0; > else > { > b1_hit = (StrategyControl->num_hit[STRAT_LIST_B1] * 100 / > *************** > *** 166,181 **** > StrategyControl->num_lookup); > b2_hit = (StrategyControl->num_hit[STRAT_LIST_B2] * 100 / > StrategyControl->num_lookup); > ! all_hit = b1_hit + t1_hit + t2_hit + b2_hit; > } > > errcxtold = error_context_stack; > error_context_stack = NULL; > ! elog(DEBUG1, "ARC T1target=%5d B1len=%5d T1len=%5d T2len=%5d B2len=%5d", > T1_TARGET, B1_LENGTH, T1_LENGTH, T2_LENGTH, B2_LENGTH); > ! elog(DEBUG1, "ARC total =%4ld%% B1hit=%4ld%% T1hit=%4ld%% T2hit=%4ld%% B2hit=%4ld%%", > all_hit, b1_hit, t1_hit, t2_hit, b2_hit); > ! elog(DEBUG1, "ARC clean buffers at LRU T1= %5d T2= %5d", > t1_clean, t2_clean); > error_context_stack = errcxtold; > > --- 168,187 ---- > StrategyControl->num_lookup); > b2_hit = (StrategyControl->num_hit[STRAT_LIST_B2] * 100 / > StrategyControl->num_lookup); > ! all_hit = t1_hit + t2_hit; > ! b_hit = b1_hit + b2_hit; > ! buf_used = T1_LENGTH + T2_LENGTH; > } > > errcxtold = error_context_stack; > error_context_stack = NULL; > ! elog(DEBUG1, "shared_buffers used=%8ld cache hits=%4ld%% turbulence=%4ld%%", > ! buf_used, all_hit, b_hit); > ! elog(DEBUG2, "ARC T1target=%5d B1len=%5d T1len=%5d T2len=%5d B2len=%5d", > T1_TARGET, B1_LENGTH, T1_LENGTH, T2_LENGTH, B2_LENGTH); > ! elog(DEBUG2, "ARC total =%4ld%% B1hit=%4ld%% T1hit=%4ld%% T2hit=%4ld%% B2hit=%4ld%%", > all_hit, b1_hit, t1_hit, t2_hit, b2_hit); > ! elog(DEBUG2, "ARC clean buffers at LRU T1= %5d T2= %5d", > t1_clean, t2_clean); > error_context_stack = errcxtold; > > > > ------------------------------------------------------------------------ > > > ---------------------------(end of broadcast)--------------------------- > TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com # From pgsql-performance-owner@postgresql.org Fri Oct 22 20:40:56 2004 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 52611EAD274 for ; Fri, 22 Oct 2004 20:40:55 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 39736-10 for ; Fri, 22 Oct 2004 19:40:53 +0000 (GMT) Received: from smtp100.rog.mail.re2.yahoo.com (smtp100.rog.mail.re2.yahoo.com [206.190.36.78]) by svr1.postgresql.org (Postfix) with SMTP id 8284AEAE4C2 for ; Fri, 22 Oct 2004 20:40:41 +0100 (BST) Received: from unknown (HELO phlogiston.dydns.org) (a.sullivan@rogers.com@65.49.125.184 with login) by smtp100.rog.mail.re2.yahoo.com with SMTP; 22 Oct 2004 19:41:00 -0000 Received: by phlogiston.dydns.org (Postfix, from userid 1000) id E608D4058; Fri, 22 Oct 2004 15:40:58 -0400 (EDT) Date: Fri, 22 Oct 2004 15:40:58 -0400 From: Andrew Sullivan To: PgSQL - Performance Subject: Re: Performance Anomalies in 7.4.5 Message-ID: <20041022194058.GC22395@phlogiston.dyndns.org> Mail-Followup-To: PgSQL - Performance References: <25228.1098395599@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <25228.1098395599@sss.pgh.pa.us> User-Agent: Mutt/1.5.6+20040722i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/380 X-Sequence-Number: 8856 > Probably the most severe objection to doing things this way is that the > selected plan could change unexpectedly as a result of the physical > table size changing. Right now the DBA can keep tight rein on actions > that might affect plan selection (ie, VACUUM and ANALYZE), but that > would go by the board with this. OTOH, we seem to be moving towards > autovacuum, which also takes away any guarantees in this department. But aren't we requiring that we can disable autovacuum on some tables? I've actually used, more than once, the finger-on-the-scale method of thumping values in pg_class when I had a pretty good idea of how the table was changing, particularly when it would change in such a way as to confuse the planner. There are still enough cases where the planner doesn't quite get things right that I'd really prefer the ability to give it clues, at least indirectly. I can't imagine that there's going to be a lot of enthusiasm for hints, so anything that isn't a sure-fire planner helper is a potential loss, at least to me. A -- Andrew Sullivan | ajs@crankycanuck.ca I remember when computers were frustrating because they *did* exactly what you told them to. That actually seems sort of quaint now. --J.D. Baldwin From pgsql-hackers-owner@postgresql.org Mon Oct 25 01:19:33 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5DC1BEAD7E2; Fri, 22 Oct 2004 21:09:51 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 48953-07; Fri, 22 Oct 2004 20:09:50 +0000 (GMT) Received: from is.rice.edu (is.rice.edu [128.42.42.24]) by svr1.postgresql.org (Postfix) with ESMTP id C7DD1EAE493; Fri, 22 Oct 2004 21:09:41 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by localhost.is.rice.edu (Postfix) with ESMTP id 85679419ED; Fri, 22 Oct 2004 15:10:01 -0500 (CDT) Received: from is.rice.edu ([127.0.0.1]) by localhost (it.is.rice.edu [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 02747-02; Fri, 22 Oct 2004 15:09:55 -0500 (CDT) Received: by is.rice.edu (Postfix, from userid 18612) id C5AB6419C1; Fri, 22 Oct 2004 15:09:55 -0500 (CDT) Date: Fri, 22 Oct 2004 15:09:55 -0500 From: Kenneth Marshall To: Jan Wieck Cc: Simon Riggs , pgsql-patches@postgresql.org, pgsql-hackers@postgresql.org, josh@agliodbs.com Subject: Re: ARC Memory Usage analysis Message-ID: <20041022200955.GC4382@it.is.rice.edu> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <41796115.2020501@Yahoo.com> User-Agent: Mutt/1.4.2i X-Virus-Scanned: by amavis-20030314-p2 at is.rice.edu X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/802 X-Sequence-Number: 60305 On Fri, Oct 22, 2004 at 03:35:49PM -0400, Jan Wieck wrote: > On 10/22/2004 2:50 PM, Simon Riggs wrote: > > >I've been using the ARC debug options to analyse memory usage on the > >PostgreSQL 8.0 server. This is a precursor to more complex performance > >analysis work on the OSDL test suite. > > > >I've simplified some of the ARC reporting into a single log line, which > >is enclosed here as a patch on freelist.c. This includes reporting of: > >- the total memory in use, which wasn't previously reported > >- the cache hit ratio, which was slightly incorrectly calculated > >- a useful-ish value for looking at the "B" lists in ARC > >(This is a patch against cvstip, but I'm not sure whether this has > >potential for inclusion in 8.0...) > > > >The total memory in use is useful because it allows you to tell whether > >shared_buffers is set too high. If it is set too high, then memory usage > >will continue to grow slowly up to the max, without any corresponding > >increase in cache hit ratio. If shared_buffers is too small, then memory > >usage will climb quickly and linearly to its maximum. > > > >The last one I've called "turbulence" in an attempt to ascribe some > >useful meaning to B1/B2 hits - I've tried a few other measures though > >without much success. Turbulence is the hit ratio of B1+B2 lists added > >together. By observation, this is zero when ARC gives smooth operation, > >and goes above zero otherwise. Typically, turbulence occurs when > >shared_buffers is too small for the working set of the database/workload > >combination and ARC repeatedly re-balances the lengths of T1/T2 as a > >result of "near-misses" on the B1/B2 lists. Turbulence doesn't usually > >cut in until the cache is fully utilized, so there is usually some delay > >after startup. > > > >We also recently discussed that I would add some further memory analysis > >features for 8.1, so I've been trying to figure out how. > > > >The idea that B1, B2 represent something really useful doesn't seem to > >have been borne out - though I'm open to persuasion there. > > > >I originally envisaged a "shadow list" operating in extension of the > >main ARC list. This will require some re-coding, since the variables and > >macros are all hard-coded to a single set of lists. No complaints, just > >it will take a little longer than we all thought (for me, that is...) > > > >My proposal is to alter the code to allow an array of memory linked > >lists. The actual list would be [0] - other additional lists would be > >created dynamically as required i.e. not using IFDEFs, since I want this > >to be controlled by a SIGHUP GUC to allow on-site tuning, not just lab > >work. This will then allow reporting against the additional lists, so > >that cache hit ratios can be seen with various other "prototype" > >shared_buffer settings. > > All the existing lists live in shared memory, so that dynamic approach > suffers from the fact that the memory has to be allocated during ipc_init. > > What do you think about my other theory to make C actually 2x effective > cache size and NOT to keep T1 in shared buffers but to assume T1 lives > in the OS buffer cache? > > > Jan > Jan, From the articles that I have seen on the ARC algorithm, I do not think that using the effective cache size to set C would be a win. The design of the ARC process is to allow the cache to optimize its use in response to the actual workload. It may be the best use of the cache in some cases to have the entire cache allocated to T1 and similarly for T2. If fact, the ability to alter the behavior as needed is one of the key advantages. --Ken From pgsql-hackers-owner@postgresql.org Fri Oct 22 21:23:40 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 1A1E5EAE48B; Fri, 22 Oct 2004 21:23:39 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 51144-10; Fri, 22 Oct 2004 20:23:39 +0000 (GMT) Received: from cmailg4.svr.pol.co.uk (cmailg4.svr.pol.co.uk [195.92.195.174]) by svr1.postgresql.org (Postfix) with ESMTP id C6D2DEAD819; Fri, 22 Oct 2004 21:23:30 +0100 (BST) Received: from modem-1017.lemur.dialup.pol.co.uk ([217.135.131.249] helo=[192.168.0.102]) by cmailg4.svr.pol.co.uk with esmtp (Exim 4.41) id 1CL5wz-0001kY-HN; Fri, 22 Oct 2004 21:23:49 +0100 Subject: Re: ARC Memory Usage analysis From: Simon Riggs To: Jan Wieck Cc: pgsql-patches@postgresql.org, pgsql-hackers@postgresql.org, josh@agliodbs.com In-Reply-To: <41796115.2020501@Yahoo.com> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> Content-Type: text/plain Organization: 2nd Quadrant Message-Id: <1098476499.20926.100.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Date: Fri, 22 Oct 2004 21:21:39 +0100 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/757 X-Sequence-Number: 60260 On Fri, 2004-10-22 at 20:35, Jan Wieck wrote: > On 10/22/2004 2:50 PM, Simon Riggs wrote: > > > > > My proposal is to alter the code to allow an array of memory linked > > lists. The actual list would be [0] - other additional lists would be > > created dynamically as required i.e. not using IFDEFs, since I want this > > to be controlled by a SIGHUP GUC to allow on-site tuning, not just lab > > work. This will then allow reporting against the additional lists, so > > that cache hit ratios can be seen with various other "prototype" > > shared_buffer settings. > > All the existing lists live in shared memory, so that dynamic approach > suffers from the fact that the memory has to be allocated during ipc_init. > [doh] - dreaming again. Yes of course, server startup it is then. [That way, we can include the memory for it at server startup, then allow the GUC to be turned off after a while to avoid another restart?] > What do you think about my other theory to make C actually 2x effective > cache size and NOT to keep T1 in shared buffers but to assume T1 lives > in the OS buffer cache? Summarised like that, I understand it. My observation is that performance varies significantly between startups of the database, which does indicate that the OS cache is working well. So, yes it does seem as if we have a 3 tier cache. I understand you to be effectively suggesting that we go back to having just a 2-tier cache. I guess we've got two options: 1. Keep ARC as it is, but just allocate much of the available physical memory to shared_buffers, so you know that effective_cache_size is low and that its either in T1 or its on disk. 2. Alter ARC so that we experiment with the view that T1 is in the OS and T2 is in shared_buffers, we don't bother keeping T1. (as you say) Hmmm...I think I'll pass on trying to judge its effectiveness - simplifying things is likely to make it easier to understand and predict behaviour. It's well worth trying, and it seems simple enough to make a patch that keeps T1target at zero. i.e. Scientific method: conjecture + experimental validation = theory If you make up a patch, probably against BETA4, Josh and I can include it in the performance testing that I'm hoping we can do over the next few weeks. Whatever makes 8.0 a high performance release is well worth it. Best Regards, Simon Riggs From pgsql-hackers-owner@postgresql.org Fri Oct 22 21:45:30 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 11373EAE4A0 for ; Fri, 22 Oct 2004 21:45:29 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 59322-05 for ; Fri, 22 Oct 2004 20:45:27 +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 C0115EAE492 for ; Fri, 22 Oct 2004 21:45:18 +0100 (BST) Received: from unknown (HELO jupiter.black-lion.info) (janwieck@68.80.245.191 with login) by smtp106.mail.sc5.yahoo.com with SMTP; 22 Oct 2004 20:45:38 -0000 Received: from [172.21.8.18] ([192.168.192.101]) (authenticated bits=0) by jupiter.black-lion.info (8.12.10/8.12.9) with ESMTP id i9MKjT94018232; Fri, 22 Oct 2004 16:45:29 -0400 (EDT) (envelope-from JanWieck@Yahoo.com) Message-ID: <41796DA0.30108@Yahoo.com> Date: Fri, 22 Oct 2004 16:29:20 -0400 From: Jan Wieck User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.1) Gecko/20040707 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Simon Riggs Cc: pgsql-patches@postgresql.org, pgsql-hackers@postgresql.org, josh@agliodbs.com Subject: Re: ARC Memory Usage analysis References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <1098476499.20926.100.camel@localhost.localdomain> In-Reply-To: <1098476499.20926.100.camel@localhost.localdomain> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/759 X-Sequence-Number: 60262 On 10/22/2004 4:21 PM, Simon Riggs wrote: > On Fri, 2004-10-22 at 20:35, Jan Wieck wrote: >> On 10/22/2004 2:50 PM, Simon Riggs wrote: >> >> > >> > My proposal is to alter the code to allow an array of memory linked >> > lists. The actual list would be [0] - other additional lists would be >> > created dynamically as required i.e. not using IFDEFs, since I want this >> > to be controlled by a SIGHUP GUC to allow on-site tuning, not just lab >> > work. This will then allow reporting against the additional lists, so >> > that cache hit ratios can be seen with various other "prototype" >> > shared_buffer settings. >> >> All the existing lists live in shared memory, so that dynamic approach >> suffers from the fact that the memory has to be allocated during ipc_init. >> > > [doh] - dreaming again. Yes of course, server startup it is then. [That > way, we can include the memory for it at server startup, then allow the > GUC to be turned off after a while to avoid another restart?] > >> What do you think about my other theory to make C actually 2x effective >> cache size and NOT to keep T1 in shared buffers but to assume T1 lives >> in the OS buffer cache? > > Summarised like that, I understand it. > > My observation is that performance varies significantly between startups > of the database, which does indicate that the OS cache is working well. > So, yes it does seem as if we have a 3 tier cache. I understand you to > be effectively suggesting that we go back to having just a 2-tier cache. Effectively yes, just with the difference that we keep a pseudo T1 list and hope that what we are tracking there is what the OS is caching. As said before, if the effective cache size is set properly, that is what should happen. > > I guess we've got two options: > 1. Keep ARC as it is, but just allocate much of the available physical > memory to shared_buffers, so you know that effective_cache_size is low > and that its either in T1 or its on disk. > 2. Alter ARC so that we experiment with the view that T1 is in the OS > and T2 is in shared_buffers, we don't bother keeping T1. (as you say) > > Hmmm...I think I'll pass on trying to judge its effectiveness - > simplifying things is likely to make it easier to understand and predict > behaviour. It's well worth trying, and it seems simple enough to make a > patch that keeps T1target at zero. Not keeping T1target at zero, because that would keep T2 at the size of shared_buffers. What I suspect is that in the current calculation the T1target is underestimated. It is incremented on B1 hits, but B1 is only of T2 size. What it currently tells is what got pushed from T1 into the OS cache. It could well be that it would work much more effective if it would fuzzily tell what got pushed out of the OS cache to disk. Jan > > i.e. Scientific method: conjecture + experimental validation = theory > > If you make up a patch, probably against BETA4, Josh and I can include it in the performance testing that I'm hoping we can do over the next few weeks. > > Whatever makes 8.0 a high performance release is well worth it. > > Best Regards, > > Simon Riggs -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com # From pgsql-performance-owner@postgresql.org Fri Oct 22 21:37:10 2004 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 F00AAEAD7E2 for ; Fri, 22 Oct 2004 21:37:09 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 56489-04 for ; Fri, 22 Oct 2004 20:37:09 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 3CDFAEAC955 for ; Fri, 22 Oct 2004 21:36:59 +0100 (BST) 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 i9MKbEwk018708; Fri, 22 Oct 2004 16:37:14 -0400 (EDT) To: Roger Ging Cc: pgsql-performance@postgresql.org Subject: Re: Slow query In-reply-to: <41795A35.4060802@paccomsys.com> References: <41795A35.4060802@paccomsys.com> Comments: In-reply-to Roger Ging message dated "Fri, 22 Oct 2004 12:06:30 -0700" Date: Fri, 22 Oct 2004 16:37:14 -0400 Message-ID: <18707.1098477434@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/381 X-Sequence-Number: 8857 Roger Ging writes: > update source_song_title set > source_song_title_id = nextval('source_song_title_seq') > ,licensing_match_order = (select licensing_match_order from > source_system where source_system_id = ss.source_system_id) > ,affiliation_match_order = (select affiliation_match_order from > source_system where source_system_id = ss.source_system_id) > ,title = st.title > from source_song_title sst > join source_song ss on ss.source_song_id = sst.source_song_id > join source_title st on st.title_id = sst.title_id > where source_song_title.source_song_id = sst.source_song_id; Why is "source_song_title sst" in there? To the extent that source_song_id is not unique, you are multiply updating rows because of the self-join. regards, tom lane From pgsql-hackers-owner@postgresql.org Fri Oct 22 21:45:48 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id B921CEAD819 for ; Fri, 22 Oct 2004 21:45:45 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 58432-06 for ; Fri, 22 Oct 2004 20:45:44 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id CF4FAEAE4A2 for ; Fri, 22 Oct 2004 21:45:36 +0100 (BST) 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 i9MKjpQZ018765; Fri, 22 Oct 2004 16:45:51 -0400 (EDT) To: Jan Wieck Cc: Simon Riggs , pgsql-patches@postgresql.org, pgsql-hackers@postgresql.org, josh@agliodbs.com Subject: Re: [PATCHES] ARC Memory Usage analysis In-reply-to: <41796115.2020501@Yahoo.com> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> Comments: In-reply-to Jan Wieck message dated "Fri, 22 Oct 2004 15:35:49 -0400" Date: Fri, 22 Oct 2004 16:45:51 -0400 Message-ID: <18764.1098477951@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/760 X-Sequence-Number: 60263 Jan Wieck writes: > What do you think about my other theory to make C actually 2x effective > cache size and NOT to keep T1 in shared buffers but to assume T1 lives > in the OS buffer cache? What will you do when initially fetching a page? It's not supposed to go directly into T2 on first use, but we're going to have some difficulty accessing a page that's not in shared buffers. I don't think you can equate the T1/T2 dichotomy to "is in shared buffers or not". You could maybe have a T3 list of "pages that aren't in shared buffers anymore but we think are still in OS buffer cache", but what would be the point? It'd be a sufficiently bad model of reality as to be pretty much useless for stats gathering, I'd think. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Oct 25 16:58:04 2004 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 77955EAE4AB for ; Fri, 22 Oct 2004 21:55:40 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 63013-04 for ; Fri, 22 Oct 2004 20:55:39 +0000 (GMT) Received: from dbl.q-ag.de (dbl.q-ag.de [213.172.117.3]) by svr1.postgresql.org (Postfix) with ESMTP id 9813EEAC955 for ; Fri, 22 Oct 2004 21:55:28 +0100 (BST) Received: from [127.0.0.2] (dbl [127.0.0.1]) by dbl.q-ag.de (8.12.3/8.12.3/Debian-6.6) with ESMTP id i9MKtdSL004760; Fri, 22 Oct 2004 22:55:40 +0200 Message-ID: <417973CA.1090807@colorfullife.com> Date: Fri, 22 Oct 2004 22:55:38 +0200 From: Manfred Spraul User-Agent: Mozilla/5.0 (X11; U; Linux i686; fr-FR; rv:1.7.3) Gecko/20040922 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Mark Wong Cc: Tom Lane , neilc@samurai.com, pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> <417697A5.1050600@colorfullife.com> <7178.1098292535@sss.pgh.pa.us> <4176A2C1.3070205@colorfullife.com> <20041020150528.B7838@osdl.org> <41774D11.1030906@colorfullife.com> <20041021065458.A5580@osdl.org> In-Reply-To: <20041021065458.A5580@osdl.org> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/414 X-Sequence-Number: 8890 Mark Wong wrote: >Pretty, simple. One to load the database, and 1 to query it. I'll >attach them. > > > I've tested it on my dual-cpu computer: - it works, both cpus run within the postmaster. It seems something your gentoo setup is broken. - the number of context switch is down slightly, but not significantly: The glibc implementation is more or less identical to the implementation right now in lwlock.c: a spinlock that protects a few variables that are used to implement the actual mutex, several wait queues: one for spinlock busy, one or two for the actual mutex code. Around 25% of the context switches are from spinlock collisions, the rest are from actual mutex collisions. It might be possible to get rid of the spinlock collisions by writing a special, futex based semaphore function that only supports exclusive access [like sem_wait/sem_post], but I don't think that it's worth the effort: 75% of the context switches would remain. What's needed is a buffer manager that can do lookups without a global lock. -- Manfred From pgsql-performance-owner@postgresql.org Fri Oct 22 22:13:12 2004 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 04EEAEAE486 for ; Fri, 22 Oct 2004 22:13:12 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 67643-07 for ; Fri, 22 Oct 2004 21:13:11 +0000 (GMT) Received: from outbound.mailhop.org (outbound.mailhop.org [63.208.196.171]) by svr1.postgresql.org (Postfix) with ESMTP id 26798EAE4AC for ; Fri, 22 Oct 2004 22:13:02 +0100 (BST) Received: from ool-4350c7ad.dyn.optonline.net ([67.80.199.173] helo=[192.168.0.66]) by outbound.mailhop.org with esmtpsa (TLSv1:AES256-SHA:256) (Exim 4.42) id 1CL6iu-0001Bf-BH; Fri, 22 Oct 2004 17:13:20 -0400 Message-ID: <417977EE.8050207@zeut.net> Date: Fri, 22 Oct 2004 17:13:18 -0400 From: "Matthew T. O'Connor" Organization: Terrie O'Connor Realtors User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Andrew Sullivan Cc: PgSQL - Performance Subject: Re: Performance Anomalies in 7.4.5 References: <25228.1098395599@sss.pgh.pa.us> <20041022194058.GC22395@phlogiston.dyndns.org> In-Reply-To: <20041022194058.GC22395@phlogiston.dyndns.org> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Mail-Handler: MailHop Outbound by DynDNS.org X-Originating-IP: 67.80.199.173 X-Report-Abuse-To: abuse@dyndns.org (see http://www.mailhop.org/outbound/abuse.html for abuse reporting information) X-MHO-User: zeut X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/382 X-Sequence-Number: 8858 Andrew Sullivan wrote: >>Probably the most severe objection to doing things this way is that the >>selected plan could change unexpectedly as a result of the physical >>table size changing. Right now the DBA can keep tight rein on actions >>that might affect plan selection (ie, VACUUM and ANALYZE), but that >>would go by the board with this. OTOH, we seem to be moving towards >>autovacuum, which also takes away any guarantees in this department. >> >> > >But aren't we requiring that we can disable autovacuum on some >tables? > Yes that is the long term goal, but the autovac in 8.0 is still all or nothing. From pgsql-hackers-owner@postgresql.org Fri Oct 22 23:03:02 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8CB44EAE486; Fri, 22 Oct 2004 23:03:00 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 81138-05; Fri, 22 Oct 2004 22:03:05 +0000 (GMT) Received: from cmailg4.svr.pol.co.uk (cmailg4.svr.pol.co.uk [195.92.195.174]) by svr1.postgresql.org (Postfix) with ESMTP id B5A44EAE483; Fri, 22 Oct 2004 23:02:56 +0100 (BST) Received: from modem-3564.lion.dialup.pol.co.uk ([217.135.173.236] helo=[192.168.0.102]) by cmailg4.svr.pol.co.uk with esmtp (Exim 4.41) id 1CL7VD-0007Us-LU; Fri, 22 Oct 2004 23:03:15 +0100 Subject: Re: [PATCHES] ARC Memory Usage analysis From: Simon Riggs To: Tom Lane Cc: Jan Wieck , pgsql-patches@postgresql.org, pgsql-hackers@postgresql.org, josh@agliodbs.com In-Reply-To: <18764.1098477951@sss.pgh.pa.us> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <18764.1098477951@sss.pgh.pa.us> Content-Type: text/plain Organization: 2nd Quadrant Message-Id: <1098482465.20926.122.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Date: Fri, 22 Oct 2004 23:01:05 +0100 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/768 X-Sequence-Number: 60271 On Fri, 2004-10-22 at 21:45, Tom Lane wrote: > Jan Wieck writes: > > What do you think about my other theory to make C actually 2x effective > > cache size and NOT to keep T1 in shared buffers but to assume T1 lives > > in the OS buffer cache? > > What will you do when initially fetching a page? It's not supposed to > go directly into T2 on first use, but we're going to have some > difficulty accessing a page that's not in shared buffers. I don't think > you can equate the T1/T2 dichotomy to "is in shared buffers or not". > Yes, there are issues there. I want Jan to follow his thoughts through. This is important enough that its worth it - there's only a few even attempting this. > You could maybe have a T3 list of "pages that aren't in shared buffers > anymore but we think are still in OS buffer cache", but what would be > the point? It'd be a sufficiently bad model of reality as to be pretty > much useless for stats gathering, I'd think. > The OS cache is in many ways a wild horse, I agree. Jan is trying to think of ways to harness it, whereas I had mostly ignored it - but its there. Raw disk usage never allowed this opportunity. For high performance systems, we can assume that the OS cache is ours to play with - what will we do with it? We need to use it for some purposes, yet would like to ignore it for others. -- Best Regards, Simon Riggs From pgsql-performance-owner@postgresql.org Sat Oct 23 08:44:55 2004 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 C78BAEAE515 for ; Sat, 23 Oct 2004 08:44:53 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 92528-02 for ; Sat, 23 Oct 2004 07:44:58 +0000 (GMT) Received: from platonic.cynic.net (platonic.cynic.net [204.80.150.245]) by svr1.postgresql.org (Postfix) with ESMTP id B3CD3EAE514 for ; Sat, 23 Oct 2004 08:44:44 +0100 (BST) Received: from angelic.cynic.net (angelic-platonic.cvpn.cynic.net [198.73.220.226]) by platonic.cynic.net (Postfix) with ESMTP id 854FDC146; Sat, 23 Oct 2004 16:33:47 +0900 (JST) Received: from localhost (localhost [127.0.0.1]) by angelic.cynic.net (Postfix) with ESMTP id 208548736; Sat, 23 Oct 2004 16:33:40 +0900 (JST) Date: Sat, 23 Oct 2004 16:33:40 +0900 (JST) From: Curt Sampson X-X-Sender: cjs@angelic-vtfw.cvpn.cynic.net To: Tom Lane Cc: Kevin Brown , pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some In-Reply-To: <4859.1097363137@sss.pgh.pa.us> Message-ID: References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> <4859.1097363137@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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/383 X-Sequence-Number: 8859 On Sat, 9 Oct 2004, Tom Lane wrote: > mmap provides msync which is comparable to fsync, but AFAICS it > provides no way to prevent an in-memory change from reaching disk too > soon. This would mean that WAL entries would have to be written *and > flushed* before we could make the data change at all, which would > convert multiple updates of a single page into a series of write-and- > wait-for-WAL-fsync steps. Not good. fsync'ing WAL once per transaction > is bad enough, once per atomic action is intolerable. Back when I was working out how to do this, I reckoned that you could use mmap by keeping a write queue for each modified page. Reading, you'd have to read the datum from the page and then check the write queue for that page to see if that datum had been updated, using the new value if it's there. Writing, you'd add the modified datum to the write queue, but not apply the write queue to the page until you'd had confirmation that the corresponding transaction log entry had been written. So multiple writes are no big deal; they just all queue up in the write queue, and at any time you can apply as much of the write queue to the page itself as the current log entry will allow. There are several different strategies available for mapping and unmapping the pages, and in fact there might need to be several available to get the best performance out of different systems. Most OSes do not seem to be optimized for having thousands or tens of thousands of small mappings (certainly NetBSD isn't), but I've never done any performance tests to see what kind of strategies might work well or not. cjs -- Curt Sampson +81 90 7737 2974 http://www.NetBSD.org Make up enjoying your city life...produced by BIC CAMERA From pgsql-performance-owner@postgresql.org Sat Oct 23 11:50:44 2004 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 06039EAE539 for ; Sat, 23 Oct 2004 11:50:42 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 21961-07 for ; Sat, 23 Oct 2004 10:50:51 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id EB1A9EAE506 for ; Sat, 23 Oct 2004 11:50:38 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i9NAomJ7025300 for ; Sat, 23 Oct 2004 10:50:48 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i9NAVIRG021829 for pgsql-performance@postgresql.org; Sat, 23 Oct 2004 10:31:18 GMT From: Gaetano Mendola X-Newsgroups: comp.databases.postgresql.performance Subject: Re: Insert performance, what should I expect? Date: Sat, 23 Oct 2004 12:31:32 +0200 Organization: PYRENET Midi-pyrenees Provider Lines: 12 Message-ID: <417A3304.9000608@bigfoot.com> References: <97b3fe2041019185337a8c3d8@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit X-Complaints-To: abuse@pyrenet.fr To: Brock Henry User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en In-Reply-To: <97b3fe2041019185337a8c3d8@mail.gmail.com> X-Enigmail-Version: 0.86.1.0 X-Enigmail-Supports: pgp-inline, pgp-mime To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/384 X-Sequence-Number: 8860 Brock Henry wrote: > Any comments/suggestions would be appreciated. Tune also the disk I/O elevator. look at this: http://www.varlena.com/varlena/GeneralBits/49.php Regards Gaetano Mendola From pgsql-performance-owner@postgresql.org Sat Oct 23 12:20:41 2004 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 778AFEAE539 for ; Sat, 23 Oct 2004 12:20:40 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 29319-02 for ; Sat, 23 Oct 2004 11:20:50 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id 10039EAE50A for ; Sat, 23 Oct 2004 12:20:37 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i9NBKmJ7030022 for ; Sat, 23 Oct 2004 11:20:48 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i9NApQBV025377 for pgsql-performance@postgresql.org; Sat, 23 Oct 2004 10:51:26 GMT From: Gaetano Mendola X-Newsgroups: comp.databases.postgresql.performance Subject: Re: futex results with dbt-3 Date: Sat, 23 Oct 2004 12:51:39 +0200 Organization: PYRENET Midi-pyrenees Provider Lines: 30 Message-ID: <417A37BB.9050305@bigfoot.com> References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> <200410191459.48080.josh@agliodbs.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit X-Complaints-To: abuse@pyrenet.fr To: Josh Berkus User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en In-Reply-To: <200410191459.48080.josh@agliodbs.com> X-Enigmail-Version: 0.86.1.0 X-Enigmail-Supports: pgp-inline, pgp-mime To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/386 X-Sequence-Number: 8862 Josh Berkus wrote: > Tom, > > >>The bigger problem here is that the SMP locking bottlenecks we are >>currently seeing are *hardware* issues (AFAICT anyway). The only way >>that futexes can offer a performance win is if they have a smarter way >>of executing the basic atomic-test-and-set sequence than we do; >>and if so, we could read their code and adopt that method without having >>to buy into any large reorganization of our code. > > > Well, initial results from Gavin/Neil's patch seem to indicate that, while > futexes do not cure the CSStorm bug, they do lessen its effects in terms of > real performance loss. I proposed weeks ago to see how the CSStorm is affected by stick each backend in one processor ( where the process was born ) using the cpu-affinity capability ( kernel 2.6 ), is this proposal completely out of mind ? Regards Gaetano Mendola From pgsql-performance-owner@postgresql.org Sat Oct 23 12:15:10 2004 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 2AE46EAE55D for ; Sat, 23 Oct 2004 12:15:06 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 25127-10 for ; Sat, 23 Oct 2004 11:15:15 +0000 (GMT) Received: from trofast.sesse.net (trofast.sesse.net [129.241.93.32]) by svr1.postgresql.org (Postfix) with ESMTP id 9C8E1EAE558 for ; Sat, 23 Oct 2004 12:15:02 +0100 (BST) Received: from sesse by trofast.sesse.net with local (Exim 3.36 #1 (Debian)) id 1CLJru-0006YS-00 for ; Sat, 23 Oct 2004 13:15:30 +0200 Date: Sat, 23 Oct 2004 13:15:30 +0200 From: "Steinar H. Gunderson" To: pgsql-performance@postgresql.org Subject: Re: Insert performance, what should I expect? Message-ID: <20041023111530.GA24976@uio.no> Mail-Followup-To: pgsql-performance@postgresql.org References: <97b3fe2041019185337a8c3d8@mail.gmail.com> <417A3304.9000608@bigfoot.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <417A3304.9000608@bigfoot.com> X-Operating-System: Linux 2.6.8.1 on a i686 User-Agent: Mutt/1.5.6+20040907i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/385 X-Sequence-Number: 8861 On Sat, Oct 23, 2004 at 12:31:32PM +0200, Gaetano Mendola wrote: >> Any comments/suggestions would be appreciated. > Tune also the disk I/O elevator. > > look at this: http://www.varlena.com/varlena/GeneralBits/49.php Mm, interesting. I've heard somewhere that the best for database-like loads on Linux is to disable the anticipatory I/O scheduler (http://kerneltrap.org/node/view/567), which should probably influence the numbers for elvtune also -- anybody know whether this is true or not for PostgreSQL? /* Steinar */ -- Homepage: http://www.sesse.net/ From pgsql-performance-owner@postgresql.org Sat Oct 23 13:07:38 2004 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 71636EAE4D0 for ; Sat, 23 Oct 2004 13:07:37 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 36291-06 for ; Sat, 23 Oct 2004 12:07:46 +0000 (GMT) Received: from rproxy.gmail.com (rproxy.gmail.com [64.233.170.203]) by svr1.postgresql.org (Postfix) with ESMTP id 8BCFDEAE557 for ; Sat, 23 Oct 2004 13:07:32 +0100 (BST) Received: by rproxy.gmail.com with SMTP id 77so290028rnk for ; Sat, 23 Oct 2004 05:08:04 -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:references; b=i6aqtIxT3PNAbXNa7FLBAniwsZmT0LyTT1SDgy7eZq1Jk5i4r9wD+fjXs502V4vjtEJPAUsi6yHyOUQnPIPQ504D+X/g+b51+tqQZ3xMiUqLMaaaram4VN/k7I+QZTJM9vJfkJNY8154u/an6/hlA3kGQ8WuZezf572O6EuEqvI= Received: by 10.38.79.21 with SMTP id c21mr195711rnb; Sat, 23 Oct 2004 05:08:04 -0700 (PDT) Received: by 10.38.77.73 with HTTP; Sat, 23 Oct 2004 05:08:04 -0700 (PDT) Message-ID: <38242de9041023050832a57cbf@mail.gmail.com> Date: Sat, 23 Oct 2004 06:08:04 -0600 From: Joshua Marsh Reply-To: Joshua Marsh To: pgsql-performance@postgresql.org Subject: Re: Slow query In-Reply-To: <18707.1098477434@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: <41795A35.4060802@paccomsys.com> <18707.1098477434@sss.pgh.pa.us> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/387 X-Sequence-Number: 8863 Any time you run subqueries, it's going to slow down the update process a lot. Each record that is updated in source_song_title runs two additional queries. When I do large updates like this, I usualy Run a transaction that will select all the new data into a new table on a join. For example SELECT a.*, b.licensing_match_order, b.affiliation_match_order, d.title INTO updated_data FROM source_song_title AS a INNER JOIN source_system AS b ON b.id = d.id INNER JOIN source_song AS c ON a.id = c.id INNER JOIN source_title AS d ON a.id = d.id I'm not sure that query does what you want, but you get the idea. Then just drop the old table and rename the updated_data table. This way instead of doing a bunch of updates, you do one select and a rename. -Josh On Fri, 22 Oct 2004 16:37:14 -0400, Tom Lane wrote: > Roger Ging writes: > > update source_song_title set > > source_song_title_id = nextval('source_song_title_seq') > > ,licensing_match_order = (select licensing_match_order from > > source_system where source_system_id = ss.source_system_id) > > ,affiliation_match_order = (select affiliation_match_order from > > source_system where source_system_id = ss.source_system_id) > > ,title = st.title > > from source_song_title sst > > join source_song ss on ss.source_song_id = sst.source_song_id > > join source_title st on st.title_id = sst.title_id > > where source_song_title.source_song_id = sst.source_song_id; > > Why is "source_song_title sst" in there? To the extent that > source_song_id is not unique, you are multiply updating rows > because of the self-join. > > regards, tom lane > > ---------------------------(end of broadcast)--------------------------- > TIP 4: Don't 'kill -9' the postmaster > From pgsql-performance-owner@postgresql.org Mon Oct 25 16:58:12 2004 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 75C06EAE4CE for ; Sat, 23 Oct 2004 16:53:40 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 84567-01 for ; Sat, 23 Oct 2004 15:53:50 +0000 (GMT) Received: from manta.curalia.se (manta.curalia.se [213.115.149.212]) by svr1.postgresql.org (Postfix) with ESMTP id 10CFEEAE575 for ; Sat, 23 Oct 2004 16:53:35 +0100 (BST) Received: from 192.168.1.13 (c-a95b70d5.025-98-73746f7.cust.bredbandsbolaget.se [213.112.91.169]) (using SSLv3 with cipher RC4-MD5 (128/128 bits)) (No client certificate requested) by manta.curalia.se (Postfix) with ESMTP id 66DDBABC003 for ; Sat, 23 Oct 2004 17:54:07 +0200 (CEST) Subject: different io elevators in linux From: Bjorn Bength To: pgsql-performance@postgresql.org Content-Type: text/plain; charset=ISO-8859-15 Organization: Curalia AB Date: Sat, 23 Oct 2004 17:54:05 +0200 Message-Id: <1098546845.7754.8.camel@localhost> Mime-Version: 1.0 X-Mailer: Evolution 2.0.2 Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/417 X-Sequence-Number: 8893 I haven't read much FAQs but has anyone done some benchmarks with different io schedulers in linux with postgresql? Look at these: http://kerneltrap.org/node/view/843 http://simonraven.nuit.ca/blog/archives/2004/05/20/quikconf/ Maybe its well known knowledge but i just found this information and haven't had time to do my own testing yet. mvh -- Bj�rn Bength | Systems Designer -- Curalia AB | www.curalia.se Tj�rhovsgatan 21, SE - 116 28 Stockholm, Sweden Phone: +46 (0)8-410 064 40 -- From pgsql-performance-owner@postgresql.org Sat Oct 23 19:10:40 2004 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 3FE72EAE58B for ; Sat, 23 Oct 2004 19:10:34 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 11768-05 for ; Sat, 23 Oct 2004 18:10:43 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 59C6FEAE582 for ; Sat, 23 Oct 2004 19:10:29 +0100 (BST) 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 i9NIB4Fc026468; Sat, 23 Oct 2004 14:11:04 -0400 (EDT) To: Curt Sampson Cc: Kevin Brown , pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some wierdness ... In-reply-to: References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> <4859.1097363137@sss.pgh.pa.us> Comments: In-reply-to Curt Sampson message dated "Sat, 23 Oct 2004 16:33:40 +0900" Date: Sat, 23 Oct 2004 14:11:04 -0400 Message-ID: <26467.1098555064@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/388 X-Sequence-Number: 8864 Curt Sampson writes: > Back when I was working out how to do this, I reckoned that you could > use mmap by keeping a write queue for each modified page. Reading, > you'd have to read the datum from the page and then check the write > queue for that page to see if that datum had been updated, using the > new value if it's there. Writing, you'd add the modified datum to the > write queue, but not apply the write queue to the page until you'd had > confirmation that the corresponding transaction log entry had been > written. So multiple writes are no big deal; they just all queue up in > the write queue, and at any time you can apply as much of the write > queue to the page itself as the current log entry will allow. Seems to me the overhead of any such scheme would swamp the savings from avoiding kernel/userspace copies ... the locking issues alone would be painful. regards, tom lane From pgsql-performance-owner@postgresql.org Sat Oct 23 19:20:47 2004 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 39392EAE57E for ; Sat, 23 Oct 2004 19:20:46 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 12041-09 for ; Sat, 23 Oct 2004 18:20:55 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id A277FEAE589 for ; Sat, 23 Oct 2004 19:20:40 +0100 (BST) 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 i9NILACQ026544; Sat, 23 Oct 2004 14:21:10 -0400 (EDT) To: Gaetano Mendola Cc: Josh Berkus , pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 In-reply-to: <417A37BB.9050305@bigfoot.com> References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> <200410191459.48080.josh@agliodbs.com> <417A37BB.9050305@bigfoot.com> Comments: In-reply-to Gaetano Mendola message dated "Sat, 23 Oct 2004 12:51:39 +0200" Date: Sat, 23 Oct 2004 14:21:10 -0400 Message-ID: <26543.1098555670@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/389 X-Sequence-Number: 8865 Gaetano Mendola writes: > I proposed weeks ago to see how the CSStorm is affected by stick each > backend in one processor ( where the process was born ) using the > cpu-affinity capability ( kernel 2.6 ), is this proposal completely > out of mind ? That was investigated long ago. See for instance http://archives.postgresql.org/pgsql-performance/2004-04/msg00313.php regards, tom lane From pgsql-performance-owner@postgresql.org Sat Oct 23 20:20:46 2004 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 88731EACD47 for ; Sat, 23 Oct 2004 20:20:45 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 25797-03 for ; Sat, 23 Oct 2004 19:20:50 +0000 (GMT) Received: from news.hub.org (news.hub.org [200.46.204.72]) by svr1.postgresql.org (Postfix) with ESMTP id AD237EAC8F4 for ; Sat, 23 Oct 2004 20:20:35 +0100 (BST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.12.9/8.12.9) with ESMTP id i9NJKnJ7026417 for ; Sat, 23 Oct 2004 19:20:49 GMT (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.12.9/8.12.9/Submit) id i9NIuesG022004 for pgsql-performance@postgresql.org; Sat, 23 Oct 2004 18:56:40 GMT From: Gaetano Mendola X-Newsgroups: comp.databases.postgresql.performance Subject: Re: futex results with dbt-3 Date: Sat, 23 Oct 2004 20:56:58 +0200 Organization: PYRENET Midi-pyrenees Provider Lines: 21 Message-ID: <417AA97A.6080202@bigfoot.com> References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> <200410191459.48080.josh@agliodbs.com> <417A37BB.9050305@bigfoot.com> <26543.1098555670@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Complaints-To: abuse@pyrenet.fr To: Tom Lane User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en In-Reply-To: <26543.1098555670@sss.pgh.pa.us> X-Enigmail-Version: 0.86.1.0 X-Enigmail-Supports: pgp-inline, pgp-mime To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/390 X-Sequence-Number: 8866 Tom Lane wrote: > Gaetano Mendola writes: > >>I proposed weeks ago to see how the CSStorm is affected by stick each >>backend in one processor ( where the process was born ) using the >>cpu-affinity capability ( kernel 2.6 ), is this proposal completely >>out of mind ? > > > That was investigated long ago. See for instance > http://archives.postgresql.org/pgsql-performance/2004-04/msg00313.php > If I read correctly this help on the CSStorm, I guess also that this could also help the performances. Unfortunatelly I do not have any kernel 2.6 running on SMP to give it a try. Regards Gaetano Mendola From pgsql-performance-owner@postgresql.org Sun Oct 24 00:20:55 2004 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 5E8CD3A2B2F for ; Sun, 24 Oct 2004 00:20:55 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 77918-02 for ; Sat, 23 Oct 2004 23:20:34 +0000 (GMT) Received: from pns.mm.eutelsat.org (pns.mm.eutelsat.org [194.214.173.227]) by svr1.postgresql.org (Postfix) with ESMTP id A050E3A1D95 for ; Sun, 24 Oct 2004 00:20:52 +0100 (BST) Received: from nts-03.mm.eutelsat.org (localhost [127.0.0.1]) by pns.mm.eutelsat.org (8.11.6/linuxconf) with ESMTP id i9NNImb01527; Sun, 24 Oct 2004 01:18:49 +0200 Received: from [127.0.0.1] (accesspoint.mm.eutelsat.org [194.214.173.4]) by nts-03.mm.eutelsat.org (8.11.6/linuxconf) with ESMTP id i9NNKsf21405; Sun, 24 Oct 2004 01:20:54 +0200 Message-ID: <417AE74E.7030004@bigfoot.com> Date: Sun, 24 Oct 2004 01:20:46 +0200 From: Gaetano Mendola User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: josh@agliodbs.com Cc: "pgsql-performance@postgresql.org" Subject: Re: futex results with dbt-3 References: <417221B5.1050704@colorfullife.com> <200410191459.48080.josh@agliodbs.com> <417A37BB.9050305@bigfoot.com> <200410231548.42762.josh@agliodbs.com> In-Reply-To: <200410231548.42762.josh@agliodbs.com> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/391 X-Sequence-Number: 8867 -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Josh Berkus wrote: | Gaetano, | | |>I proposed weeks ago to see how the CSStorm is affected by stick each |>backend in one processor ( where the process was born ) using the |>cpu-affinity capability ( kernel 2.6 ), is this proposal completely out of |>mind ? | | | I don't see how that would help. The problem is not backends switching | processors, it's the buffermgrlock needing to be swapped between processors. This is not clear to me. What happen if during a spinlock a backend is moved away from one processor to another one ? Regards Gaetano Mendola -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (MingW32) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFBeudN7UpzwH2SGd4RAkL9AKCUY9vsw1CPmBV1kC7BKxUtuneN2wCfXaYr E8utuJI34MAIP8jUm6By09M= =oRvU -----END PGP SIGNATURE----- From pgsql-performance-owner@postgresql.org Sun Oct 24 06:08:17 2004 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 510493A3B7C for ; Sun, 24 Oct 2004 06:08:16 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 41941-10 for ; Sun, 24 Oct 2004 05:07:46 +0000 (GMT) Received: from roue.portalpotty.net (roue.portalpotty.net [69.44.62.206]) by svr1.postgresql.org (Postfix) with ESMTP id 58B8C3A3B09 for ; Sun, 24 Oct 2004 06:08:07 +0100 (BST) Received: from max by roue.portalpotty.net with local (Exim 4.34) id 1CLabz-00039o-Q2; Sat, 23 Oct 2004 22:08:11 -0700 Date: Sun, 24 Oct 2004 01:08:11 -0400 From: Max Baker To: Rod Taylor Cc: Postgresql Performance Subject: Re: Vacuum takes a really long time, vacuum full required Message-ID: <20041024050811.GA12044@warped.org> Mail-Followup-To: Rod Taylor , Postgresql Performance References: <20041019153821.GA21258@warped.org> <1098200416.750.238.camel@home> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1098200416.750.238.camel@home> User-Agent: Mutt/1.4.1i X-AntiAbuse: This header was added to track abuse, please include it with any abuse report X-AntiAbuse: Primary Hostname - roue.portalpotty.net X-AntiAbuse: Original Domain - postgresql.org X-AntiAbuse: Originator/Caller UID/GID - [514 32003] / [47 12] X-AntiAbuse: Sender Address Domain - roue.portalpotty.net X-Source: /usr/bin/mutt X-Source-Args: mutt X-Source-Dir: /home/max X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/392 X-Sequence-Number: 8868 On Tue, Oct 19, 2004 at 11:40:17AM -0400, Rod Taylor wrote: > > Whatever the case, the database still slows down to a halt after a month or > > so, and I have to go in and shut everything down and do a VACUUM FULL by > > hand. One index (of many many) takes 2000 seconds to vacuum. The whole > > process takes a few hours. > > Do a REINDEX on that table instead, and regular vacuum more frequently. > > > $ pg_config --version > > PostgreSQL 7.3.2 > > 7.4.x deals with index growth a little better 7.3 and older did. I did a REINDEX of the database. The results are pretty insane, the db went from 16GB to 381MB. Needless to say things are running a lot faster. I will now take Tom's well-given advice and upgrade to 7.4. But at least now I have something to tell my users who are not able to do a DB upgrade for whatever reason. Thanks for all your help folks! -m Before: # du -h pgsql 135K pgsql/global 128M pgsql/pg_xlog 80M pgsql/pg_clog 3.6M pgsql/base/1 3.6M pgsql/base/16975 1.0K pgsql/base/16976/pgsql_tmp 16G pgsql/base/16976 16G pgsql/base 16G pgsql After Reindex: # du /data/pgsql/ 131K /data/pgsql/global 128M /data/pgsql/pg_xlog 81M /data/pgsql/pg_clog 3.6M /data/pgsql/base/1 3.6M /data/pgsql/base/16975 1.0K /data/pgsql/base/16976/pgsql_tmp 268M /data/pgsql/base/16976 275M /data/pgsql/base 484M /data/pgsql/ After Vacuum: # du /data/pgsql/ 131K /data/pgsql/global 144M /data/pgsql/pg_xlog 81M /data/pgsql/pg_clog 3.6M /data/pgsql/base/1 3.6M /data/pgsql/base/16975 1.0K /data/pgsql/base/16976/pgsql_tmp 149M /data/pgsql/base/16976 156M /data/pgsql/base 381M /data/pgsql/ netdisco=> select relname, relpages from pg_class order by relpages desc; Before: relname | relpages ---------------------------------+---------- idx_node_switch_port_active | 590714 idx_node_switch_port | 574344 idx_node_switch | 482202 idx_node_mac | 106059 idx_node_mac_active | 99842 After: relname | relpages ---------------------------------+---------- node_ip | 13829 node | 9560 device_port | 2124 node_ip_pkey | 1354 idx_node_ip_ip | 1017 idx_node_ip_mac_active | 846 From pgsql-performance-owner@postgresql.org Sun Oct 24 06:46:32 2004 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 382233A3AC3 for ; Sun, 24 Oct 2004 06:46:32 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 51348-03 for ; Sun, 24 Oct 2004 05:46:01 +0000 (GMT) Received: from platonic.cynic.net (platonic.cynic.net [204.80.150.245]) by svr1.postgresql.org (Postfix) with ESMTP id 68AD93A3E68 for ; Sun, 24 Oct 2004 06:46:20 +0100 (BST) Received: from angelic.cynic.net (angelic-platonic.cvpn.cynic.net [198.73.220.226]) by platonic.cynic.net (Postfix) with ESMTP id 6647BC000; Sun, 24 Oct 2004 14:46:24 +0900 (JST) Received: from localhost (localhost [127.0.0.1]) by angelic.cynic.net (Postfix) with ESMTP id 76D518736; Sun, 24 Oct 2004 14:46:16 +0900 (JST) Date: Sun, 24 Oct 2004 14:46:16 +0900 (JST) From: Curt Sampson X-X-Sender: cjs@angelic-vtfw.cvpn.cynic.net To: Tom Lane Cc: Kevin Brown , pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some In-Reply-To: <26467.1098555064@sss.pgh.pa.us> Message-ID: References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> <4859.1097363137@sss.pgh.pa.us> <26467.1098555064@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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/393 X-Sequence-Number: 8869 On Sat, 23 Oct 2004, Tom Lane wrote: > Seems to me the overhead of any such scheme would swamp the savings from > avoiding kernel/userspace copies ... Well, one really can't know without testing, but memory copies are extremely expensive if they go outside of the cache. > the locking issues alone would be painful. I don't see why they would be any more painful than the current locking issues. In fact, I don't see any reason to add more locking than we already use when updating pages. cjs -- Curt Sampson +81 90 7737 2974 http://www.NetBSD.org Make up enjoying your city life...produced by BIC CAMERA From pgsql-performance-owner@postgresql.org Sun Oct 24 15:39:38 2004 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 54CD73A3BF0 for ; Sun, 24 Oct 2004 15:39:38 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49256-04 for ; Sun, 24 Oct 2004 14:39:04 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 82FC43A2BE5 for ; Sun, 24 Oct 2004 15:39:29 +0100 (BST) 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 i9OEdZio008439; Sun, 24 Oct 2004 10:39:35 -0400 (EDT) To: Curt Sampson Cc: Kevin Brown , pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some In-reply-to: References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> <4859.1097363137@sss.pgh.pa.us> <26467.1098555064@sss.pgh.pa.us> Comments: In-reply-to Curt Sampson message dated "Sun, 24 Oct 2004 14:46:16 +0900" Date: Sun, 24 Oct 2004 10:39:35 -0400 Message-ID: <8438.1098628775@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/394 X-Sequence-Number: 8870 Curt Sampson writes: > On Sat, 23 Oct 2004, Tom Lane wrote: >> Seems to me the overhead of any such scheme would swamp the savings from >> avoiding kernel/userspace copies ... > Well, one really can't know without testing, but memory copies are > extremely expensive if they go outside of the cache. Sure, but what about all the copying from write queue to page? >> the locking issues alone would be painful. > I don't see why they would be any more painful than the current locking > issues. Because there are more locks --- the write queue data structure will need to be locked separately from the page. (Even with a separate write queue per page, there will need to be a shared data structure that allows you to allocate and find write queues, and that thing will be a subject of contention. See BufMgrLock, which is not held while actively twiddling the contents of pages, but is a serious cause of contention anyway.) regards, tom lane From pgsql-performance-owner@postgresql.org Sun Oct 24 19:12:05 2004 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 B1CEA3A3F85 for ; Sun, 24 Oct 2004 19:12:00 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 88806-05 for ; Sun, 24 Oct 2004 18:11:30 +0000 (GMT) Received: from mra04.ex.eclipse.net.uk (mra04.ex.eclipse.net.uk [212.104.129.139]) by svr1.postgresql.org (Postfix) with ESMTP id B58123A3F9E for ; Sun, 24 Oct 2004 19:11:49 +0100 (BST) Received: from localhost (localhost.localdomain [127.0.0.1]) by mra04.ex.eclipse.net.uk (Postfix) with ESMTP id 53475133D81 for ; Sun, 24 Oct 2004 19:06:42 +0100 (BST) Received: from mra04.ex.eclipse.net.uk ([127.0.0.1]) by localhost (mra04.ex.eclipse.net.uk [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 20604-01-13 for ; Sun, 24 Oct 2004 19:06:41 +0100 (BST) Received: from dev1 (unknown [82.152.145.238]) by mra04.ex.eclipse.net.uk (Postfix) with ESMTP id F2D16133C61 for ; Sun, 24 Oct 2004 19:06:40 +0100 (BST) From: "Rod Dutton" To: Subject: Queries slow using stored procedures Date: Sun, 24 Oct 2004 19:13:23 +0100 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0000_01C4B9FD.887BB690" X-Mailer: Microsoft Office Outlook, Build 11.0.5510 Thread-Index: AcS1PHuwz718w8JtTi2M+MPIPiou1gEuIX0Q X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1106 Message-Id: <20041024180640.F2D16133C61@mra04.ex.eclipse.net.uk> X-Virus-Scanned: by Eclipse VIRUSshield at eclipse.net.uk X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.3 tagged_above=0.0 required=5.0 tests=HTML_50_60, HTML_FONTCOLOR_BLUE, HTML_MESSAGE X-Spam-Level: X-Archive-Number: 200410/395 X-Sequence-Number: 8871 This is a multi-part message in MIME format. ------=_NextPart_000_0000_01C4B9FD.887BB690 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Hi, Has anybody got any ideas on my recent posting ? (thanks in advance) :- I have a problem where a query inside a function is up to 100 times slower inside a function than as a stand alone query run in psql. The column 'botnumber' is a character(10), is indexed and there are 125000 rows in the table. Help please! This query is fast:- explain analyze SELECT batchserial FROM transbatch WHERE botnumber = '1-7' LIMIT 1; QUERY PLAN ---------------------------------------------------------------------------- ---------------------------------------------------- Limit (cost=0.00..0.42 rows=1 width=4) (actual time=0.73..148.23 rows=1 loops=1) -> Index Scan using ind_tbatchx on transbatch (cost=0.00..18.73 rows=45 width=4) (actual time=0.73..148.22 rows=1 loops=1) Index Cond: (botnumber = '1-7'::bpchar) Total runtime: 148.29 msec (4 rows) This function is slow:- CREATE OR REPLACE FUNCTION sp_test_rod3 ( ) returns integer as ' DECLARE bot char(10); oldbatch INTEGER; BEGIN bot := ''1-7''; SELECT INTO oldbatch batchserial FROM transbatch WHERE botnumber = bot LIMIT 1; IF FOUND THEN RETURN 1; ELSE RETURN 0; END IF; END; ' language plpgsql ; explain analyze SELECT sp_test_rod3(); QUERY PLAN ---------------------------------------------------------------------------- ------------ Result (cost=0.00..0.01 rows=1 width=0) (actual time=1452.39..1452.40 rows=1 loops=1) Total runtime: 1452.42 msec (2 rows) ------=_NextPart_000_0000_01C4B9FD.887BB690 Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable
 
Hi,=  
 
Has anybody got any ideas = on my=20 recent posting ? (thanks in advance) :-<= /DIV>
 
 
I have a = problem=20 where a query inside a function is up to 100 times slower inside a function= than=20 as a stand alone query run in psql.
 
The colum= n=20 'botnumber' is a character(10), is indexed and there are 125000 rows i= n the=20 table.
 
Help=20 please!
 
This quer= y is=20 fast:-
 
explain= =20 analyze  
&nb= sp;=20 SELECT batchserial
  FROM transbatch
  WHERE botnumb= er =3D=20 '1-7'
  LIMIT 1;
           =             &nb= sp;            =             &nb= sp;         =20 QUERY=20 PLAN            = ;            &n= bsp;            = ;            &n= bsp;        =20
-----------------------------------------------------------------------= ---------------------------------------------------------
 Limit&nb= sp;=20 (cost=3D0.00..0.42 rows=3D1 width=3D4) (actual time=3D0.73..148.23 rows=3D1= =20 loops=3D1)
   ->  Index Scan using ind_tbatchx on=20 transbatch  (cost=3D0.00..18.73 rows=3D45 width=3D4) (actual time=3D0.= 73..148.22=20 rows=3D1 loops=3D1)
         Ind= ex Cond:=20 (botnumber =3D '1-7'::bpchar)
 Total runtime: 148.29 msec
(4=20 rows)
 
 
This=20 function is slow:-
 
CREATE=20 OR REPLACE FUNCTION  sp_test_rod3 ( ) returns=20 integer         
as=20 '
DECLARE
  bot char(10);
  oldbatch=20 INTEGER;
BEGIN
 
 =20 bot :=3D ''1-7'';
 
 =20 SELECT INTO oldbatch batchserial
  FROM transbatch
  WHERE= =20 botnumber =3D bot
  LIMIT 1;
 
 =20 IF FOUND THEN
    RETURN 1;
 =20 ELSE
    RETURN 0;
  END=20 IF;
 
END;
'
language plpgsql  ;
 
explain analyze SELECT sp_test_rod3();
           =             &nb= sp;            =   =20 QUERY=20 PLAN            = ;            &n= bsp;            = ; =20
-----------------------------------------------------------------------= -----------------
 Result =20 (cost=3D0.00..0.01 rows=3D1 width=3D0) (actual time=3D1452.39..1452.40 rows= =3D1=20 loops=3D1)
 Total runtime: 1452.42 msec
(2=20 rows)
------=_NextPart_000_0000_01C4B9FD.887BB690-- From pgsql-performance-owner@postgresql.org Sun Oct 24 19:25:55 2004 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 A87C33A3F93 for ; Sun, 24 Oct 2004 19:25:53 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 94664-01 for ; Sun, 24 Oct 2004 18:25:21 +0000 (GMT) Received: from server07.icaen.uiowa.edu (server07.icaen.uiowa.edu [128.255.17.47]) by svr1.postgresql.org (Postfix) with ESMTP id 53D993A3B60 for ; Sun, 24 Oct 2004 19:25:46 +0100 (BST) Received: from server11.icaen.uiowa.edu (server11.icaen.uiowa.edu [128.255.17.51]) by server07.icaen.uiowa.edu (8.12.9/8.12.9) with ESMTP id i9OIPmMq027018; (envelope-from ) Sun, 24 Oct 2004 13:25:48 -0500 (CDT) Received: from [192.168.1.11] (12-217-241-0.client.mchsi.com [12.217.241.0]) by server11.icaen.uiowa.edu (8.12.9/smtp-service-1.6) with ESMTP id i9OIPlBN022250; (envelope-from ) Sun, 24 Oct 2004 13:25:47 -0500 (CDT) Message-ID: <417BF3AD.6010901@johnmeinel.com> Date: Sun, 24 Oct 2004 13:25:49 -0500 From: John Meinel User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Rod Dutton Cc: pgsql-performance@postgresql.org Subject: Re: Queries slow using stored procedures References: <20041024180640.F2D16133C61@mra04.ex.eclipse.net.uk> In-Reply-To: <20041024180640.F2D16133C61@mra04.ex.eclipse.net.uk> X-Enigmail-Version: 0.86.0.0 X-Enigmail-Supports: pgp-inline, pgp-mime Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig14D8A891B7D53A803E849F3A" X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.66n X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.75 on clamav.icaen.uiowa.edu X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/396 X-Sequence-Number: 8872 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig14D8A891B7D53A803E849F3A Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Rod Dutton wrote: > > Hi, > > Has anybody got any ideas on my recent posting ? (thanks in advance) :- > > > I have a problem where a query inside a function is up to 100 times > slower inside a function than as a stand alone query run in psql. > > The column 'botnumber' is a character(10), is indexed and there are > 125000 rows in the table. > [...] I had a similar problem before, where the function version (stored procedure or prepared query) was much slower. I had a bunch of tables all with references to another table. I was querying all of the references to see if anyone from any of the tables was referencing a particular row in the base table. It turned out that one of the child tables was referencing the same row 300,000/500,000 times. So if I happened to pick *that* number, postgres wanted to a sequential scan because of all the potential results. In my testing, I never picked that number, so it was very fast, since it knew it wouldn't get in trouble. In the case of the stored procedure, it didn't know which number I was going to ask for, so it had to plan for the worst, and *always* do a sequential scan. So the question is... In your table, does the column "botnumber" have the same value repeated many, many times, but '1-7' only occurs a few? If you change the function to: CREATE OR REPLACE FUNCTION sp_test_rod3 ( ) returns integer as ' DECLARE bot char(10); oldbatch INTEGER; BEGIN SELECT INTO oldbatch batchserial FROM transbatch WHERE botnumber = ''1-7'' LIMIT 1; IF FOUND THEN RETURN 1; ELSE RETURN 0; END IF; END; ' language plpgsql ; Is it still slow? I don't know if you could get crazy with something like: select 1 where exist(select from transbatch where botnumber = '1-7' limit 1); Just some thoughts about where *I've* found performance to change between functions versus raw SQL. You probably should also mention what version of postgres you are running (and possibly what your hardware is) John =:-> --------------enig14D8A891B7D53A803E849F3A Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFBe/OtJdeBCYSNAAMRAp7GAKCorGfd2EOHXy1WplWCYrQ4B6LeRQCgiXzq QMz9AdKW8Ali61i+xQd0fVk= =UMbP -----END PGP SIGNATURE----- --------------enig14D8A891B7D53A803E849F3A-- From pgsql-performance-owner@postgresql.org Sun Oct 24 20:09:09 2004 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 4FB973A3B60 for ; Sun, 24 Oct 2004 20:09:08 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 01845-07 for ; Sun, 24 Oct 2004 19:08:32 +0000 (GMT) Received: from server07.icaen.uiowa.edu (server07.icaen.uiowa.edu [128.255.17.47]) by svr1.postgresql.org (Postfix) with ESMTP id 0DF983A2BE5 for ; Sun, 24 Oct 2004 20:08:58 +0100 (BST) Received: from server11.icaen.uiowa.edu (server11.icaen.uiowa.edu [128.255.17.51]) by server07.icaen.uiowa.edu (8.12.9/8.12.9) with ESMTP id i9OJ92Mq027718; (envelope-from ) Sun, 24 Oct 2004 14:09:02 -0500 (CDT) Received: from [192.168.1.11] (12-217-241-0.client.mchsi.com [12.217.241.0]) by server11.icaen.uiowa.edu (8.12.9/smtp-service-1.6) with ESMTP id i9OJ91BN028283; (envelope-from ) Sun, 24 Oct 2004 14:09:01 -0500 (CDT) Message-ID: <417BFDCB.3040404@johnmeinel.com> Date: Sun, 24 Oct 2004 14:08:59 -0500 From: John Meinel User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Rod Dutton , pgsql-performance@postgresql.org Subject: Re: Queries slow using stored procedures References: <20041024184053.9819240619B@mra02.ex.eclipse.net.uk> In-Reply-To: <20041024184053.9819240619B@mra02.ex.eclipse.net.uk> X-Enigmail-Version: 0.86.0.0 X-Enigmail-Supports: pgp-inline, pgp-mime Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigC39560A59387A6C5A347CC5D" X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.66n X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.75 on clamav.icaen.uiowa.edu X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/397 X-Sequence-Number: 8873 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigC39560A59387A6C5A347CC5D Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Rod Dutton wrote: > Thank John, > > I am running Postgres 7.3.7 on a Dell PowerEdge 6600 Server with Quad Xeon > 2.7GHz processors with 16GB RAM and 12 x 146GB drives in Raid 10 (OS, WAL, > Data all on separate arrays). > You might want think about upgraded to 7.4, as I know it is better at quite a few things. But I'm not all that experienced (I just had a similar problem). > I did try hard coding botnumber as you suggested and it was FAST. So it > does look like the scenario that you have explained. > There are 2 ways of doing it that I know of. First, you can make you function create a query and execute it. Something like: EXECUTE ''SELECT 1 FROM transbatch WHERE botnumber = '' || quote_literal(botnum) || '' LIMIT 1''; That forces the database to redesign the query each time. The problem you are having is a stored procedure has to prepare the query in advance. > >>does the column "botnumber" have the same value repeated many, many times, > > but '1-7' only occurs a few? > > Yes, that could be the case, the table fluctuates massively from small to > big to small regularly with a real mixture of occurrences of these values > i.e. some values are repeated many times and some occur only a few times. > > I wonder if the answer is to: a) don't use a stored procedure b) up the > statistics gathering for that column ? > I don't believe increasing statistics will help, as prepared statements require one-size-fits-all queries. > I will try your idea: select 1 where exist(select from transbatch where > botnumber = '1-7' limit 1); > > Also, how can I get "EXPLAIN" output from the internals of the stored > procedure as that would help me? > I believe the only way to get explain is to use prepared statements instead of stored procedures. For example: PREPARE my_plan(char(10)) AS SELECT 1 FROM transbatch WHERE botnumber = $1 LIMIT 1; EXPLAIN ANALYZE EXECUTE my_plan('1-7'); > Many thanks, > > Rod > If you have to do the first thing I mentioned, I'm not sure if you are getting much out of your function, so you might prefer to just ask the question directly. What really surprises me is that it doesn't use the index even after the LIMIT clause. But I just did a check on my machine where I had a column with lots of repeated entries, and it didn't use the index. So a question for the true Guru's (like Tom Lane): Why doesn't postgres use an indexed query if you supply a LIMIT? John =:-> --------------enigC39560A59387A6C5A347CC5D Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFBe/3PJdeBCYSNAAMRAvjcAJ4l9eoNP1nU8Brgylbh5wIlB4gW2gCfVbNj uBgQNBZ4m+ZhGFxzDuk2JUk= =T1Dk -----END PGP SIGNATURE----- --------------enigC39560A59387A6C5A347CC5D-- From pgsql-performance-owner@postgresql.org Sun Oct 24 20:11:53 2004 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 29DD83A3B60 for ; Sun, 24 Oct 2004 20:11:52 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 02189-07 for ; Sun, 24 Oct 2004 19:11:16 +0000 (GMT) Received: from server07.icaen.uiowa.edu (server07.icaen.uiowa.edu [128.255.17.47]) by svr1.postgresql.org (Postfix) with ESMTP id 4FDFE3A2BE5 for ; Sun, 24 Oct 2004 20:11:43 +0100 (BST) Received: from server11.icaen.uiowa.edu (server11.icaen.uiowa.edu [128.255.17.51]) by server07.icaen.uiowa.edu (8.12.9/8.12.9) with ESMTP id i9OJBlMq027779; (envelope-from ) Sun, 24 Oct 2004 14:11:47 -0500 (CDT) Received: from [192.168.1.11] (12-217-241-0.client.mchsi.com [12.217.241.0]) by server11.icaen.uiowa.edu (8.12.9/smtp-service-1.6) with ESMTP id i9OJBkBN028623; (envelope-from ) Sun, 24 Oct 2004 14:11:46 -0500 (CDT) Message-ID: <417BFE75.4060006@johnmeinel.com> Date: Sun, 24 Oct 2004 14:11:49 -0500 From: John Meinel User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Rod Dutton , pgsql-performance@postgresql.org Subject: Re: Queries slow using stored procedures References: <20041024184536.2A841406474@mra02.ex.eclipse.net.uk> In-Reply-To: <20041024184536.2A841406474@mra02.ex.eclipse.net.uk> X-Enigmail-Version: 0.86.0.0 X-Enigmail-Supports: pgp-inline, pgp-mime Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig6ACE2FD04CAD41D0F1154357" X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.66n X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.75 on clamav.icaen.uiowa.edu X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/398 X-Sequence-Number: 8874 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig6ACE2FD04CAD41D0F1154357 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Rod Dutton wrote: > I also should add that the sp is only slow when the table is big (probably > obvious!). > > Rod Sure, the problem is it is switching to a sequential search, with a lot of rows, versus doing an indexed search. It's all about trying to figure out how to fix that, especially for any value of botnum. I would have hoped that using LIMIT 1 would have fixed that. John =:-> --------------enig6ACE2FD04CAD41D0F1154357 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFBe/51JdeBCYSNAAMRAj40AKCgWYbCednkxAXUfYYCAaThoxftIACgh7PB x3Wkj4EmZzqicQtMgFgRaAc= =5AfW -----END PGP SIGNATURE----- --------------enig6ACE2FD04CAD41D0F1154357-- From pgsql-performance-owner@postgresql.org Sun Oct 24 20:28:00 2004 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 A13FC3A3F88 for ; Sun, 24 Oct 2004 20:27:59 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 07432-02 for ; Sun, 24 Oct 2004 19:27:23 +0000 (GMT) Received: from server07.icaen.uiowa.edu (server07.icaen.uiowa.edu [128.255.17.47]) by svr1.postgresql.org (Postfix) with ESMTP id 8CF943A3F84 for ; Sun, 24 Oct 2004 20:27:50 +0100 (BST) Received: from server11.icaen.uiowa.edu (server11.icaen.uiowa.edu [128.255.17.51]) by server07.icaen.uiowa.edu (8.12.9/8.12.9) with ESMTP id i9OJRvMq028066 for ; (envelope-from ) Sun, 24 Oct 2004 14:27:57 -0500 (CDT) Received: from [192.168.1.11] (12-217-241-0.client.mchsi.com [12.217.241.0]) by server11.icaen.uiowa.edu (8.12.9/smtp-service-1.6) with ESMTP id i9OJRvBN000968 for ; (envelope-from ) Sun, 24 Oct 2004 14:27:57 -0500 (CDT) Message-ID: <417C023F.1080502@johnmeinel.com> Date: Sun, 24 Oct 2004 14:27:59 -0500 From: John Meinel User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: Sequential Scan with LIMIT X-Enigmail-Version: 0.86.0.0 X-Enigmail-Supports: pgp-inline, pgp-mime Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigF85FC43A9A23B9D3A4C0F238" X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.66n X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.75 on clamav.icaen.uiowa.edu X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/399 X-Sequence-Number: 8875 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigF85FC43A9A23B9D3A4C0F238 Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit I was looking into another problem, and I found something that surprised me. If I'm doing "SELECT * FROM mytable WHERE col = 'myval' LIMIT 1.". Now "col" is indexed, by mytable has 500,000 rows, and 'myval' occurs maybe 100,000 times. Without the LIMIT, this query should definitely do a sequential scan. But with the LIMIT, doesn't it know that it will return at max 1 value, and thus be able to use the index? It seems to be doing the LIMIT too late. The real purpose of this query is to check to see if a value exists in the column, so there might be a better way of doing it. Here is the demo info: # select count(*) from finst_t; 542315 # select count(*) from finst_t where store_id = 539960; 85076 # explain analyze select id from finst_t where store_id = 539960 limit 1; QUERY PLAN ------------------------------------------------------------------------------------------------------------------- Limit (cost=0.00..0.13 rows=1 width=4) (actual time=860.000..860.000 rows=1 loops=1) -> Seq Scan on finst_t (cost=0.00..11884.94 rows=88217 width=4) (actual time=860.000..860.000 rows=1 loops=1) Filter: (store_id = 539960) Total runtime: 860.000 ms Notice that the "actual rows=1", meaning it is aware of the limit as it is going through the table. But for some reason the planner thinks it is going to return 88,217 rows. (This is close to the reality of 85076 if it actually had to find all of the rows). Now, if I do a select on a value that *does* only have 1 value, it works fine: # explain analyze select id from finst_t where store_id = 9605 limit 1; QUERY PLAN ------------------------------------------------------------------------------------------------------------------------ Limit (cost=0.00..3.96 rows=1 width=4) (actual time=0.000..0.000 rows=1 loops=1) -> Index Scan using finst_t_store_id_idx on finst_t (cost=0.00..3.96 rows=1 width=4) (actual time=0.000..0.000 rows=1 loops=1) Index Cond: (store_id = 9605) Total runtime: 0.000 ms And 1 further thing, I *can* force it to do a fast index scan if I disable sequential scanning. # set enable_seqscan to off; # explain analyze select id from finst_t where store_id = 539960 limit 1; QUERY PLAN ------------------------------------------------------------------------------------------------------------------------ Limit (cost=0.00..1.59 rows=1 width=4) (actual time=0.000..0.000 rows=1 loops=1) -> Index Scan using finst_t_store_id_idx on finst_t (cost=0.00..140417.22 rows=88217 width=4) (actual time=0.000..0.000 rows=1 loops=1) Index Cond: (store_id = 539960) Total runtime: 0.000 ms Could being aware of LIMIT be added to the planner? Is there a better way to check for existence? John =:-> PS> I'm using postgres 8.0-beta3 on win32 (the latest installer). --------------enigF85FC43A9A23B9D3A4C0F238 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFBfAI/JdeBCYSNAAMRAi+UAKDFkbJ9JSV6NIJjyii2oPDIMaWj+ACfV4Nv NnPsyuCflirLpwD6pD7YxHs= =UkRy -----END PGP SIGNATURE----- --------------enigF85FC43A9A23B9D3A4C0F238-- From pgsql-performance-owner@postgresql.org Sun Oct 24 21:11:58 2004 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 C31883A3FC3 for ; Sun, 24 Oct 2004 21:11:56 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 11828-10 for ; Sun, 24 Oct 2004 20:11:20 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id EFF0C3A2BE5 for ; Sun, 24 Oct 2004 21:11:47 +0100 (BST) 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 i9OKBrIH017690; Sun, 24 Oct 2004 16:11:53 -0400 (EDT) To: John Meinel Cc: pgsql-performance@postgresql.org Subject: Re: Sequential Scan with LIMIT In-reply-to: <417C023F.1080502@johnmeinel.com> References: <417C023F.1080502@johnmeinel.com> Comments: In-reply-to John Meinel message dated "Sun, 24 Oct 2004 14:27:59 -0500" Date: Sun, 24 Oct 2004 16:11:53 -0400 Message-ID: <17689.1098648713@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/400 X-Sequence-Number: 8876 John Meinel writes: > I was looking into another problem, and I found something that surprised > me. If I'm doing "SELECT * FROM mytable WHERE col = 'myval' LIMIT 1.". > Now "col" is indexed, by mytable has 500,000 rows, and 'myval' occurs > maybe 100,000 times. Without the LIMIT, this query should definitely do > a sequential scan. > But with the LIMIT, doesn't it know that it will return at max 1 value, > and thus be able to use the index? But the LIMIT will cut the cost of the seqscan case too. Given the numbers you posit above, about one row in five will have 'myval', so a seqscan can reasonably expect to hit the first matching row in the first page of the table. This is still cheaper than doing an index scan (which must require reading at least one index page plus at least one table page). The test case you are showing is probably suffering from nonrandom placement of this particular data value; which is something that the statistics we keep are too crude to detect. regards, tom lane From pgsql-performance-owner@postgresql.org Sun Oct 24 22:51:59 2004 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 735523A3BFF for ; Sun, 24 Oct 2004 22:51:58 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 37854-04 for ; Sun, 24 Oct 2004 21:51:21 +0000 (GMT) Received: from server07.icaen.uiowa.edu (server07.icaen.uiowa.edu [128.255.17.47]) by svr1.postgresql.org (Postfix) with ESMTP id 6949B3A2BE5 for ; Sun, 24 Oct 2004 22:51:48 +0100 (BST) Received: from server11.icaen.uiowa.edu (server11.icaen.uiowa.edu [128.255.17.51]) by server07.icaen.uiowa.edu (8.12.9/8.12.9) with ESMTP id i9OLpuMq000340; (envelope-from ) Sun, 24 Oct 2004 16:51:56 -0500 (CDT) Received: from [192.168.1.11] (12-217-241-0.client.mchsi.com [12.217.241.0]) by server11.icaen.uiowa.edu (8.12.9/smtp-service-1.6) with ESMTP id i9OLpPBN019459; (envelope-from ) Sun, 24 Oct 2004 16:51:56 -0500 (CDT) Message-ID: <417C23DC.1090407@johnmeinel.com> Date: Sun, 24 Oct 2004 16:51:24 -0500 From: John Meinel User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: pgsql-performance@postgresql.org Subject: Re: Sequential Scan with LIMIT References: <417C023F.1080502@johnmeinel.com> <17689.1098648713@sss.pgh.pa.us> In-Reply-To: <17689.1098648713@sss.pgh.pa.us> X-Enigmail-Version: 0.86.0.0 X-Enigmail-Supports: pgp-inline, pgp-mime Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigFCA213021A1493416B41EFC0" X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.66n X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.75 on clamav.icaen.uiowa.edu X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/401 X-Sequence-Number: 8877 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigFCA213021A1493416B41EFC0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Tom Lane wrote: > John Meinel writes: > >>I was looking into another problem, and I found something that surprised >>me. If I'm doing "SELECT * FROM mytable WHERE col = 'myval' LIMIT 1.". >>Now "col" is indexed, by mytable has 500,000 rows, and 'myval' occurs >>maybe 100,000 times. Without the LIMIT, this query should definitely do >>a sequential scan. > > >>But with the LIMIT, doesn't it know that it will return at max 1 value, >>and thus be able to use the index? > > > But the LIMIT will cut the cost of the seqscan case too. Given the > numbers you posit above, about one row in five will have 'myval', so a > seqscan can reasonably expect to hit the first matching row in the first > page of the table. This is still cheaper than doing an index scan > (which must require reading at least one index page plus at least one > table page). > > The test case you are showing is probably suffering from nonrandom > placement of this particular data value; which is something that the > statistics we keep are too crude to detect. > > regards, tom lane You are correct about non-random placement. I'm a little surprised it doesn't change with values, then. For instance, # select count(*) from finst_t where store_id = 52; 13967 Still does a sequential scan for the "select id from..." query. The only value it does an index query for is 9605 which only has 1 row. It estimates ~18,000 rows, but that is still < 3% of the total data. This row corresponds to disk location where files can be found. So when a storage location fills up, generally a new one is created. This means that *generally* the numbers will be increasing as you go further in the table (not guaranteed, as there are multiple locations open at any one time). Am I better off in this case just wrapping my query with: set enable_seqscan to off; query set enable_seqscan to on; There is still the possibility that there is a better way to determine existence of a value in a column. I was wondering about something like: SELECT 1 WHERE EXISTS (SELECT id FROM finst_t WHERE store_id=52 LIMIT 1); Though the second part is the same, so it still does the sequential scan. This isn't critical, I was just trying to understand what's going on. Thanks for your help. John =:-> --------------enigFCA213021A1493416B41EFC0 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFBfCPgJdeBCYSNAAMRAgUkAJ9HPwhlnYuLWyDEcuukne0KL7TvsACghgPu 6vL+YolBeCgHTPMTfP/ldm4= =NvTk -----END PGP SIGNATURE----- --------------enigFCA213021A1493416B41EFC0-- From pgsql-performance-owner@postgresql.org Sun Oct 24 22:59:31 2004 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 725ED3A2BE5 for ; Sun, 24 Oct 2004 22:59:29 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 40073-02 for ; Sun, 24 Oct 2004 21:58:41 +0000 (GMT) Received: from mra04.ex.eclipse.net.uk (mra04.ex.eclipse.net.uk [212.104.129.139]) by svr1.postgresql.org (Postfix) with ESMTP id EC6E03A3FEB for ; Sun, 24 Oct 2004 22:59:07 +0100 (BST) Received: from localhost (localhost.localdomain [127.0.0.1]) by mra04.ex.eclipse.net.uk (Postfix) with ESMTP id BBB89133AFE for ; Sun, 24 Oct 2004 22:54:00 +0100 (BST) Received: from mra04.ex.eclipse.net.uk ([127.0.0.1]) by localhost (mra04.ex.eclipse.net.uk [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 31190-01-33 for ; Sun, 24 Oct 2004 22:54:00 +0100 (BST) Received: from dev1 (unknown [82.152.145.238]) by mra04.ex.eclipse.net.uk (Postfix) with ESMTP id ED668133D1E for ; Sun, 24 Oct 2004 22:53:59 +0100 (BST) From: "Rod Dutton" To: Subject: Reindexdb and REINDEX Date: Sun, 24 Oct 2004 23:00:42 +0100 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0012_01C4BA1D.49978060" X-Mailer: Microsoft Office Outlook, Build 11.0.5510 Thread-Index: AcS6EpNUyDmOSHB3T9GPkaaJa1R8cQAAkrug X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1106 Message-Id: <20041024215359.ED668133D1E@mra04.ex.eclipse.net.uk> X-Virus-Scanned: by Eclipse VIRUSshield at eclipse.net.uk X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.5 tagged_above=0.0 required=5.0 tests=HTML_40_50, HTML_MESSAGE X-Spam-Level: X-Archive-Number: 200410/402 X-Sequence-Number: 8878 This is a multi-part message in MIME format. ------=_NextPart_000_0012_01C4BA1D.49978060 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Hi, I have had some performance problems recently on very large tables (10s of millions of rows). A vacuum full did make a large improvement and then dropping & re-creating the indexes also was very beneficial. My performance problem has now been solved. My question is: will using the contrib/reindexdb or REINDEX sql command do essentially the same job as dropping and re-creating the indexes. I.E. do you get a fully compacted and balanced index? If so then I could use contrib/reindexdb or REINDEX instead of drop/recreate. How is concurrency handled by contrib/reindexdb and REINDEX (I know you can create an index on the fly with no obvious lock outs)? Thanks, Rod ------=_NextPart_000_0012_01C4BA1D.49978060 Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable
Hi,
 
I have ha= d some=20 performance problems recently on very large tables (10s of millions of=20 rows).  A vacuum full did make a large improvement and then dropping &= amp;=20 re-creating the indexes also was very beneficial.  My performance prob= lem=20 has now been solved.
 
My questi= on is: will=20 using the contrib/reindexdb or REINDEX sql command do essentially the same = job=20 as dropping and re-creating the indexes.  I.E. do you get a fully comp= acted=20 and balanced index?  If so then I could use contrib/reindexdb or REIND= EX=20 instead of drop/recreate. 
 
How is co= ncurrency=20 handled by contrib/reindexdb and REINDEX (I know you can create an index on= the=20 fly with no obvious lock outs)?
 
Thanks,
 
Rod
 
------=_NextPart_000_0012_01C4BA1D.49978060-- From pgsql-performance-owner@postgresql.org Mon Oct 25 01:16:21 2004 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 300263A2016 for ; Mon, 25 Oct 2004 01:16:20 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 65952-08 for ; Mon, 25 Oct 2004 00:15:47 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 0FEE23A201A for ; Mon, 25 Oct 2004 01:16:13 +0100 (BST) 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 i9P0GHJA028828; Sun, 24 Oct 2004 20:16:17 -0400 (EDT) To: "Rod Dutton" Cc: pgsql-performance@postgresql.org Subject: Re: Reindexdb and REINDEX In-reply-to: <20041024215359.ED668133D1E@mra04.ex.eclipse.net.uk> References: <20041024215359.ED668133D1E@mra04.ex.eclipse.net.uk> Comments: In-reply-to "Rod Dutton" message dated "Sun, 24 Oct 2004 23:00:42 +0100" Date: Sun, 24 Oct 2004 20:16:17 -0400 Message-ID: <28827.1098663377@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/403 X-Sequence-Number: 8879 "Rod Dutton" writes: > My question is: will using the contrib/reindexdb or REINDEX sql command do > essentially the same job as dropping and re-creating the indexes. I.E. do > you get a fully compacted and balanced index? Yes. > How is concurrency handled by contrib/reindexdb and REINDEX (I know you can > create an index on the fly with no obvious lock outs)? In 8.0 they are almost equivalent, but in earlier releases REINDEX takes an exclusive lock on the index's parent table. The details are: DROP INDEX: takes exclusive lock, but doesn't hold it long. CREATE INDEX: takes ShareLock, which blocks writers but not readers. So when you do it that way, readers can use the table while CREATE INDEX runs, but of course they have no use of the dropped index. Putting the DROP and the CREATE in one transaction isn't a good idea if you want concurrency, because then the exclusive lock persists till transaction end. REINDEX before 8.0: takes exclusive lock for the duration. This of course is a dead loss for concurrency. REINDEX in 8.0: takes ShareLock on the table and exclusive lock on the particular index. This means that writers are blocked, but readers can proceed *as long as they don't try to use the index under reconstruction*. If they try, they block. If you're rebuilding a popular index, you have a choice of making readers do seqscans or having them block till the rebuild completes. One other point is that DROP/CREATE breaks any stored plans that use the index, which can have negative effects on plpgsql functions and PREPAREd statements. REINDEX doesn't break plans. We don't currently have any automated way of rebuilding stored plans, so in the worst case you may have to terminate open backend sessions after a DROP/CREATE. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Oct 25 01:30:55 2004 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 BBE413A4027 for ; Mon, 25 Oct 2004 01:30:53 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 72904-04 for ; Mon, 25 Oct 2004 00:30:23 +0000 (GMT) Received: from platonic.cynic.net (platonic.cynic.net [204.80.150.245]) by svr1.postgresql.org (Postfix) with ESMTP id 456563A4024 for ; Mon, 25 Oct 2004 01:30:50 +0100 (BST) Received: from angelic.cynic.net (angelic-platonic.cvpn.cynic.net [198.73.220.226]) by platonic.cynic.net (Postfix) with ESMTP id EE97DC145; Mon, 25 Oct 2004 09:30:57 +0900 (JST) Received: from localhost (localhost [127.0.0.1]) by angelic.cynic.net (Postfix) with ESMTP id 568D18736; Mon, 25 Oct 2004 09:30:56 +0900 (JST) Date: Mon, 25 Oct 2004 09:30:56 +0900 (JST) From: Curt Sampson X-X-Sender: cjs@angelic-vtfw.cvpn.cynic.net To: Tom Lane Cc: Kevin Brown , pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some In-Reply-To: <8438.1098628775@sss.pgh.pa.us> Message-ID: References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> <4859.1097363137@sss.pgh.pa.us> <26467.1098555064@sss.pgh.pa.us> <8438.1098628775@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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/404 X-Sequence-Number: 8880 On Sun, 24 Oct 2004, Tom Lane wrote: > > Well, one really can't know without testing, but memory copies are > > extremely expensive if they go outside of the cache. > > Sure, but what about all the copying from write queue to page? There's a pretty big difference between few-hundred-bytes-on-write and eight-kilobytes-with-every-read memory copy. As for the queue allocation, again, I have no data to back this up, but I don't think it would be as bad as BufMgrLock. Not every page will have a write queue, and a "hot" page is only going to get one once. (If a page has a write queue, you might as well leave it with the page after flushing it, and get rid of it only when the page leaves memory.) I see the OS issues related to mapping that much memory as a much bigger potential problem. cjs -- Curt Sampson +81 90 7737 2974 http://www.NetBSD.org Make up enjoying your city life...produced by BIC CAMERA From pgsql-performance-owner@postgresql.org Mon Oct 25 02:18:07 2004 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 CA9D33A3AE0 for ; Mon, 25 Oct 2004 02:18:05 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 84742-06 for ; Mon, 25 Oct 2004 01:17:32 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 05E733A2016 for ; Mon, 25 Oct 2004 02:18:00 +0100 (BST) 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 i9P1I7tV029891; Sun, 24 Oct 2004 21:18:07 -0400 (EDT) To: Curt Sampson Cc: Kevin Brown , pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some In-reply-to: References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> <4859.1097363137@sss.pgh.pa.us> <26467.1098555064@sss.pgh.pa.us> <8438.1098628775@sss.pgh.pa.us> Comments: In-reply-to Curt Sampson message dated "Mon, 25 Oct 2004 09:30:56 +0900" Date: Sun, 24 Oct 2004 21:18:07 -0400 Message-ID: <29890.1098667087@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/405 X-Sequence-Number: 8881 Curt Sampson writes: > I see the OS issues related to mapping that much memory as a much bigger > potential problem. I see potential problems everywhere I look ;-) Considering that the available numbers suggest we could win just a few percent (and that's assuming that all this extra mechanism has zero cost), I can't believe that the project is worth spending manpower on. There is a lot of much more attractive fruit hanging at lower levels. The bitmap-indexing stuff that was recently being discussed, for instance, would certainly take less effort than this; it would create no new portability issues; and at least for the queries where it helps, it could offer integer-multiple speedups, not percentage points. My engineering professors taught me that you put large effort where you have a chance at large rewards. Converting PG to mmap doesn't seem to meet that test, even if I believed it would work. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Oct 25 02:33:01 2004 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 5D9A33A2019 for ; Mon, 25 Oct 2004 02:32:59 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 86769-08 for ; Mon, 25 Oct 2004 01:32:21 +0000 (GMT) Received: from platonic.cynic.net (platonic.cynic.net [204.80.150.245]) by svr1.postgresql.org (Postfix) with ESMTP id 6FACC3A3AE0 for ; Mon, 25 Oct 2004 02:32:49 +0100 (BST) Received: from angelic.cynic.net (angelic-platonic.cvpn.cynic.net [198.73.220.226]) by platonic.cynic.net (Postfix) with ESMTP id C838AC145; Mon, 25 Oct 2004 10:32:57 +0900 (JST) Received: from localhost (localhost [127.0.0.1]) by angelic.cynic.net (Postfix) with ESMTP id 05DBA8736; Mon, 25 Oct 2004 10:32:55 +0900 (JST) Date: Mon, 25 Oct 2004 10:32:55 +0900 (JST) From: Curt Sampson X-X-Sender: cjs@angelic-vtfw.cvpn.cynic.net To: Tom Lane Cc: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some In-Reply-To: <29890.1098667087@sss.pgh.pa.us> Message-ID: References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> <4859.1097363137@sss.pgh.pa.us> <26467.1098555064@sss.pgh.pa.us> <8438.1098628775@sss.pgh.pa.us> <29890.1098667087@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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/406 X-Sequence-Number: 8882 On Sun, 24 Oct 2004, Tom Lane wrote: > Considering that the available numbers suggest we could win just a few > percent... I must confess that I was completely unaware of these "numbers." Where do I find them? cjs -- Curt Sampson +81 90 7737 2974 http://www.NetBSD.org Make up enjoying your city life...produced by BIC CAMERA From pgsql-performance-owner@postgresql.org Mon Oct 25 16:58:09 2004 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 BB9E63A405A for ; Mon, 25 Oct 2004 02:49:05 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91342-06 for ; Mon, 25 Oct 2004 01:48:27 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id AA0943A2019 for ; Mon, 25 Oct 2004 02:48:55 +0100 (BST) 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 i9P1n3E8006099; Sun, 24 Oct 2004 21:49:03 -0400 (EDT) To: Curt Sampson Cc: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some In-reply-to: References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> <4859.1097363137@sss.pgh.pa.us> <26467.1098555064@sss.pgh.pa.us> <8438.1098628775@sss.pgh.pa.us> <29890.1098667087@sss.pgh.pa.us> Comments: In-reply-to Curt Sampson message dated "Mon, 25 Oct 2004 10:32:55 +0900" Date: Sun, 24 Oct 2004 21:49:03 -0400 Message-ID: <6097.1098668943@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/416 X-Sequence-Number: 8892 Curt Sampson writes: > On Sun, 24 Oct 2004, Tom Lane wrote: >> Considering that the available numbers suggest we could win just a few >> percent... > I must confess that I was completely unaware of these "numbers." Where > do I find them? The only numbers I've seen that directly bear on the question is the oprofile results that Josh recently put up for the DBT-3 benchmark, which showed the kernel copy-to-userspace and copy-from-userspace subroutines eating a percent or two apiece of the total runtime. I don't have the URL at hand but it was posted just a few days ago. (Now that covers all such copies and not only our datafile reads/writes, but it's probably fair to assume that the datafile I/O is the bulk of it.) This is, of course, only one benchmark ... but lacking any measurements in opposition, I'm inclined to believe it. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Oct 25 16:58:17 2004 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 3C9063A4049 for ; Mon, 25 Oct 2004 02:50:29 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 92335-05 for ; Mon, 25 Oct 2004 01:49:58 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id D082F3A4045 for ; Mon, 25 Oct 2004 02:50:27 +0100 (BST) 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 i9P1oZWd006277; Sun, 24 Oct 2004 21:50:35 -0400 (EDT) To: Curt Sampson Cc: pgsql-performance@postgresql.org Subject: Re: First set of OSDL Shared Mem scalability results, some In-reply-to: <6097.1098668943@sss.pgh.pa.us> References: <200410081443.16109.josh@agliodbs.com> <20041009112048.GA665@filer> <27696.1097334448@sss.pgh.pa.us> <20041009203710.GB665@filer> <4859.1097363137@sss.pgh.pa.us> <26467.1098555064@sss.pgh.pa.us> <8438.1098628775@sss.pgh.pa.us> <29890.1098667087@sss.pgh.pa.us> <6097.1098668943@sss.pgh.pa.us> Comments: In-reply-to Tom Lane message dated "Sun, 24 Oct 2004 21:49:03 -0400" Date: Sun, 24 Oct 2004 21:50:34 -0400 Message-ID: <6276.1098669034@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/418 X-Sequence-Number: 8894 I wrote: > I don't have the URL at hand but it was posted just a few days ago. ... actually, it was the beginning of this here thread ... regards, tom lane From pgsql-performance-owner@postgresql.org Mon Oct 25 08:17:22 2004 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 B755F3A3B56 for ; Mon, 25 Oct 2004 08:17:21 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 71174-08 for ; Mon, 25 Oct 2004 07:17:11 +0000 (GMT) Received: from platonic.cynic.net (platonic.cynic.net [204.80.150.245]) by svr1.postgresql.org (Postfix) with ESMTP id 7A63E3A3B14 for ; Mon, 25 Oct 2004 08:17:11 +0100 (BST) Received: from angelic.cynic.net (angelic-platonic.cvpn.cynic.net [198.73.220.226]) by platonic.cynic.net (Postfix) with ESMTP id 4DD79BFFB; Mon, 25 Oct 2004 16:17:09 +0900 (JST) Received: from localhost (localhost [127.0.0.1]) by angelic.cynic.net (Postfix) with ESMTP id BD8DC8736; Mon, 25 Oct 2004 16:17:07 +0900 (JST) Date: Mon, 25 Oct 2004 16:17:07 +0900 (JST) From: Curt Sampson X-X-Sender: cjs@angelic-vtfw.cvpn.cynic.net To: John Meinel Cc: pgsql-performance@postgresql.org Subject: Re: Sequential Scan with LIMIT In-Reply-To: <417C023F.1080502@johnmeinel.com> Message-ID: References: <417C023F.1080502@johnmeinel.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/407 X-Sequence-Number: 8883 On Sun, 24 Oct 2004, John Meinel wrote: > I was looking into another problem, and I found something that surprised > me. If I'm doing "SELECT * FROM mytable WHERE col = 'myval' LIMIT 1.". > Now "col" is indexed... > The real purpose of this query is to check to see if a value exists in > the column,... When you select all the columns, you're going to force it to go to the table. If you select only the indexed column, it ought to be able to use just the index, and never read the table at all. You could also use more standard and more set-oriented SQL while you're at it: SELECT DISTINCT(col) FROM mytable WHERE col = 'myval' cjs -- Curt Sampson +81 90 7737 2974 http://www.NetBSD.org Make up enjoying your city life...produced by BIC CAMERA From pgsql-performance-owner@postgresql.org Mon Oct 25 08:25:29 2004 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 121093A40D0 for ; Mon, 25 Oct 2004 08:25:28 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 76246-01 for ; Mon, 25 Oct 2004 07:25:17 +0000 (GMT) Received: from sue.samurai.com (sue.samurai.com [205.207.28.74]) by svr1.postgresql.org (Postfix) with ESMTP id BF5A93A40FB for ; Mon, 25 Oct 2004 08:25:17 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by sue.samurai.com (Postfix) with ESMTP id AEF70197F4; Mon, 25 Oct 2004 03:25:17 -0400 (EDT) Received: from sue.samurai.com ([127.0.0.1]) by localhost (sue.samurai.com [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 30709-01-9; Mon, 25 Oct 2004 03:25:14 -0400 (EDT) Received: from [61.88.101.19] (unknown [61.88.101.19]) by sue.samurai.com (Postfix) with ESMTP id 8514B197F1; Mon, 25 Oct 2004 03:25:13 -0400 (EDT) Subject: Re: Sequential Scan with LIMIT From: Neil Conway To: Curt Sampson Cc: John Meinel , pgsql-performance@postgresql.org In-Reply-To: References: <417C023F.1080502@johnmeinel.com> Content-Type: text/plain Message-Id: <1098689094.13261.445.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Mon, 25 Oct 2004 17:24:54 +1000 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/408 X-Sequence-Number: 8884 On Mon, 2004-10-25 at 17:17, Curt Sampson wrote: > When you select all the columns, you're going to force it to go to the > table. If you select only the indexed column, it ought to be able to use > just the index, and never read the table at all. Perhaps in other database systems, but not in PostgreSQL. MVCC information is only stored in the heap, not in indexes: therefore, PostgreSQL cannot determine whether a given index entry refers to a valid tuple. Therefore, it needs to check the heap even if the index contains all the columns referenced by the query. While it would be nice to be able to do index-only scans, there is good reason for this design decision. Check the archives for past discussion about the tradeoffs involved. > You could also use more > standard and more set-oriented SQL while you're at it: > > SELECT DISTINCT(col) FROM mytable WHERE col = 'myval' This is likely to be less efficient though. -Neil From pgsql-performance-owner@postgresql.org Mon Oct 25 14:51:39 2004 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 267553A4052 for ; Mon, 25 Oct 2004 14:51:39 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91568-07 for ; Mon, 25 Oct 2004 13:51:30 +0000 (GMT) Received: from smtp105.rog.mail.re2.yahoo.com (smtp105.rog.mail.re2.yahoo.com [206.190.36.83]) by svr1.postgresql.org (Postfix) with SMTP id BD5633A402C for ; Mon, 25 Oct 2004 14:51:32 +0100 (BST) Received: from unknown (HELO phlogiston.dydns.org) (a.sullivan@rogers.com@65.49.125.184 with login) by smtp105.rog.mail.re2.yahoo.com with SMTP; 25 Oct 2004 13:51:33 -0000 Received: by phlogiston.dydns.org (Postfix, from userid 1000) id 7CAE44058; Mon, 25 Oct 2004 09:51:31 -0400 (EDT) Date: Mon, 25 Oct 2004 09:51:31 -0400 From: Andrew Sullivan To: PgSQL - Performance Subject: Re: Performance Anomalies in 7.4.5 Message-ID: <20041025135131.GA2886@phlogiston.dyndns.org> Mail-Followup-To: PgSQL - Performance References: <25228.1098395599@sss.pgh.pa.us> <20041022194058.GC22395@phlogiston.dyndns.org> <417977EE.8050207@zeut.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <417977EE.8050207@zeut.net> User-Agent: Mutt/1.5.6+20040722i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/409 X-Sequence-Number: 8885 On Fri, Oct 22, 2004 at 05:13:18PM -0400, Matthew T. O'Connor wrote: > Yes that is the long term goal, but the autovac in 8.0 is still all or > nothing. Yes, which is why I couldn't use the current iteration for production: the cost is too high. I think this re-inforces my original point, which is that taking away the ability of DBAs to thump the planner for certain tables -- even indirectly -- under certain pathological conditions is crucial for production work. In the ideal world, the wizards and genius planners and such like would work perfectly, and the DBA would never have to intervene. In practice, there are cases when you need to haul on a knob or two. While this doesn't mean that we should adopt the Oracle approach of having knobs which adjust the sensitivity of the knob that tunes the main knob-tuner, I'm still pretty leery of anything which smacks of completely locking the DBA's knowledge out of the system. A -- Andrew Sullivan | ajs@crankycanuck.ca The plural of anecdote is not data. --Roger Brinner From pgsql-performance-owner@postgresql.org Mon Oct 25 15:33:52 2004 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 710643A40CA for ; Mon, 25 Oct 2004 15:33:51 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 08743-10 for ; Mon, 25 Oct 2004 14:33:42 +0000 (GMT) Received: from server07.icaen.uiowa.edu (server07.icaen.uiowa.edu [128.255.17.47]) by svr1.postgresql.org (Postfix) with ESMTP id C49723A40D0 for ; Mon, 25 Oct 2004 15:33:44 +0100 (BST) Received: from server11.icaen.uiowa.edu (server11.icaen.uiowa.edu [128.255.17.51]) by server07.icaen.uiowa.edu (8.12.9/8.12.9) with ESMTP id i9PEXbMq021211; (envelope-from ) Mon, 25 Oct 2004 09:33:37 -0500 (CDT) Received: from [192.168.1.11] (12-217-241-0.client.mchsi.com [12.217.241.0]) by server11.icaen.uiowa.edu (8.12.9/smtp-service-1.6) with ESMTP id i9PEXaJ9001445 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NOT); (envelope-from ) Mon, 25 Oct 2004 09:33:36 -0500 (CDT) Message-ID: <417D0EC2.5020206@johnmeinel.com> Date: Mon, 25 Oct 2004 09:33:38 -0500 From: John Meinel User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Curt Sampson Cc: pgsql-performance@postgresql.org Subject: Re: Sequential Scan with LIMIT References: <417C023F.1080502@johnmeinel.com> In-Reply-To: X-Enigmail-Version: 0.86.0.0 X-Enigmail-Supports: pgp-inline, pgp-mime Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig1883AEDB77A4CC3971BEA57B" X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.66n X-Virus-Scanned: clamd / ClamAV version 0.75.1, clamav-milter version 0.75 on clamav.icaen.uiowa.edu X-Virus-Status: Clean X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/410 X-Sequence-Number: 8886 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig1883AEDB77A4CC3971BEA57B Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Curt Sampson wrote: > On Sun, 24 Oct 2004, John Meinel wrote: > > >>I was looking into another problem, and I found something that surprised >>me. If I'm doing "SELECT * FROM mytable WHERE col = 'myval' LIMIT 1.". >>Now "col" is indexed... >>The real purpose of this query is to check to see if a value exists in >>the column,... > > > When you select all the columns, you're going to force it to go to the > table. If you select only the indexed column, it ought to be able to use > just the index, and never read the table at all. You could also use more > standard and more set-oriented SQL while you're at it: > > SELECT DISTINCT(col) FROM mytable WHERE col = 'myval' > > cjs Well, what you wrote was actually much slower, as it had to scan the whole table, grab all the rows, and then distinct them in the end. However, this query worked: SELECT DISTINCT(col) FROM mytable WHERE col = 'myval' LIMIT 1; Now, *why* that works differently from: SELECT col FROM mytable WHERE col = 'myval' LIMIT 1; or SELECT DISTINCT(col) FROM mytable WHERE col = 'myval'; I'm not sure. They all return the same information. What's also weird is stuff like: SELECT DISTINCT(NULL) FROM mytable WHERE col = 'myval' LIMIT 1; Also searches the entire table, sorting that NULL == NULL wherever col = 'myval'. Which is as expensive as the non-limited case (I'm guessing that the limit is occurring after the distinct, which is causing the problem. SELECT NULL FROM ... still uses a sequential scan, but it stops after finding the first one.) Actually, in doing a little bit more testing, the working query only works on some of the values. Probably it just increases the expense enough that it switches over. It also has the downside that when it does switch to seq scan, it is much more expensive as it has to do a sort and a unique on all the entries. John =:-> --------------enig1883AEDB77A4CC3971BEA57B Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFBfQ7CJdeBCYSNAAMRAu0QAKDOWBXEP/2tshNMNTtGlr5fEPLuPQCeOURu 2Gsj++N4zoUmHhMJqnKvmFE= =sJHL -----END PGP SIGNATURE----- --------------enig1883AEDB77A4CC3971BEA57B-- From pgsql-hackers-owner@postgresql.org Mon Oct 25 16:38:38 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8EB403A4147 for ; Mon, 25 Oct 2004 16:38:35 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 33084-08 for ; Mon, 25 Oct 2004 15:38:29 +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 8CE643A4142 for ; Mon, 25 Oct 2004 16:38:28 +0100 (BST) Received: from unknown (HELO jupiter.black-lion.info) (janwieck@68.80.245.191 with login) by smtp104.mail.sc5.yahoo.com with SMTP; 25 Oct 2004 15:38:27 -0000 Received: from [172.21.8.18] (ismtp.afilias.com [216.217.55.254]) (authenticated bits=0) by jupiter.black-lion.info (8.12.10/8.12.9) with ESMTP id i9PFc994003184; Mon, 25 Oct 2004 11:38:13 -0400 (EDT) (envelope-from JanWieck@Yahoo.com) Message-ID: <417D1D01.3090401@Yahoo.com> Date: Mon, 25 Oct 2004 11:34:25 -0400 From: Jan Wieck User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.1) Gecko/20040707 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Kenneth Marshall Cc: Simon Riggs , pgsql-patches@postgresql.org, pgsql-hackers@postgresql.org, josh@agliodbs.com Subject: Re: ARC Memory Usage analysis References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> In-Reply-To: <20041022200955.GC4382@it.is.rice.edu> 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=1.6 tagged_above=0.0 required=5.0 tests=RCVD_IN_NJABL_DUL, RCVD_IN_SORBS_DUL X-Spam-Level: * X-Archive-Number: 200410/818 X-Sequence-Number: 60321 On 10/22/2004 4:09 PM, Kenneth Marshall wrote: > On Fri, Oct 22, 2004 at 03:35:49PM -0400, Jan Wieck wrote: >> On 10/22/2004 2:50 PM, Simon Riggs wrote: >> >> >I've been using the ARC debug options to analyse memory usage on the >> >PostgreSQL 8.0 server. This is a precursor to more complex performance >> >analysis work on the OSDL test suite. >> > >> >I've simplified some of the ARC reporting into a single log line, which >> >is enclosed here as a patch on freelist.c. This includes reporting of: >> >- the total memory in use, which wasn't previously reported >> >- the cache hit ratio, which was slightly incorrectly calculated >> >- a useful-ish value for looking at the "B" lists in ARC >> >(This is a patch against cvstip, but I'm not sure whether this has >> >potential for inclusion in 8.0...) >> > >> >The total memory in use is useful because it allows you to tell whether >> >shared_buffers is set too high. If it is set too high, then memory usage >> >will continue to grow slowly up to the max, without any corresponding >> >increase in cache hit ratio. If shared_buffers is too small, then memory >> >usage will climb quickly and linearly to its maximum. >> > >> >The last one I've called "turbulence" in an attempt to ascribe some >> >useful meaning to B1/B2 hits - I've tried a few other measures though >> >without much success. Turbulence is the hit ratio of B1+B2 lists added >> >together. By observation, this is zero when ARC gives smooth operation, >> >and goes above zero otherwise. Typically, turbulence occurs when >> >shared_buffers is too small for the working set of the database/workload >> >combination and ARC repeatedly re-balances the lengths of T1/T2 as a >> >result of "near-misses" on the B1/B2 lists. Turbulence doesn't usually >> >cut in until the cache is fully utilized, so there is usually some delay >> >after startup. >> > >> >We also recently discussed that I would add some further memory analysis >> >features for 8.1, so I've been trying to figure out how. >> > >> >The idea that B1, B2 represent something really useful doesn't seem to >> >have been borne out - though I'm open to persuasion there. >> > >> >I originally envisaged a "shadow list" operating in extension of the >> >main ARC list. This will require some re-coding, since the variables and >> >macros are all hard-coded to a single set of lists. No complaints, just >> >it will take a little longer than we all thought (for me, that is...) >> > >> >My proposal is to alter the code to allow an array of memory linked >> >lists. The actual list would be [0] - other additional lists would be >> >created dynamically as required i.e. not using IFDEFs, since I want this >> >to be controlled by a SIGHUP GUC to allow on-site tuning, not just lab >> >work. This will then allow reporting against the additional lists, so >> >that cache hit ratios can be seen with various other "prototype" >> >shared_buffer settings. >> >> All the existing lists live in shared memory, so that dynamic approach >> suffers from the fact that the memory has to be allocated during ipc_init. >> >> What do you think about my other theory to make C actually 2x effective >> cache size and NOT to keep T1 in shared buffers but to assume T1 lives >> in the OS buffer cache? >> >> >> Jan >> > Jan, > >>From the articles that I have seen on the ARC algorithm, I do not think > that using the effective cache size to set C would be a win. The design > of the ARC process is to allow the cache to optimize its use in response > to the actual workload. It may be the best use of the cache in some cases > to have the entire cache allocated to T1 and similarly for T2. If fact, > the ability to alter the behavior as needed is one of the key advantages. Only the "working set" of the database, that is the pages that are very frequently used, are worth holding in shared memory at all. The rest should be copied in and out of the OS disc buffers. The problem is, with a too small directory ARC cannot guesstimate what might be in the kernel buffers. Nor can it guesstimate what recently was in the kernel buffers and got pushed out from there. That results in a way too small B1 list, and therefore we don't get B1 hits when in fact the data was found in memory. B1 hits is what increases the T1target, and since we are missing them with a too small directory size, our implementation of ARC is propably using a T2 size larger than the working set. That is not optimal. If we would replace the dynamic T1 buffers with a max_backends*2 area of shared buffers, use a C value representing the effective cache size and limit the T1target on the lower bound to effective cache size - shared buffers, then we basically moved the T1 cache into the OS buffers. This all only holds water, if the OS is allowed to swap out shared memory. And that was my initial question, how likely is it to find this to be true these days? Jan -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com # From pgsql-hackers-owner@postgresql.org Mon Oct 25 17:06:45 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D1D2B3A4142 for ; Mon, 25 Oct 2004 17:03:37 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 43775-10 for ; Mon, 25 Oct 2004 16:03:32 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 0F9923A4147 for ; Mon, 25 Oct 2004 17:03:31 +0100 (BST) 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 i9PG3D7t020652; Mon, 25 Oct 2004 12:03:13 -0400 (EDT) To: Jan Wieck Cc: Kenneth Marshall , Simon Riggs , pgsql-patches@postgresql.org, pgsql-hackers@postgresql.org, josh@agliodbs.com Subject: Re: [PATCHES] ARC Memory Usage analysis In-reply-to: <417D1D01.3090401@Yahoo.com> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> Comments: In-reply-to Jan Wieck message dated "Mon, 25 Oct 2004 11:34:25 -0400" Date: Mon, 25 Oct 2004 12:03:12 -0400 Message-ID: <20651.1098720192@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/821 X-Sequence-Number: 60324 Jan Wieck writes: > This all only holds water, if the OS is allowed to swap out shared > memory. And that was my initial question, how likely is it to find this > to be true these days? I think it's more likely that not that the OS will consider shared memory to be potentially swappable. On some platforms there is a shmctl call you can make to lock your shmem in memory, but (a) we don't use it and (b) it may well require privileges we haven't got anyway. This has always been one of the arguments against making shared_buffers really large, of course --- if the buffers aren't all heavily used, and the OS decides to swap them to disk, you are worse off than you would have been with a smaller shared_buffers setting. However, I'm still really nervous about the idea of using effective_cache_size to control the ARC algorithm. That number is usually entirely bogus. Right now it is only a second-order influence on certain planner estimates, and I am afraid to rely on it any more heavily than that. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Oct 25 17:34:05 2004 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 901453A412E for ; Mon, 25 Oct 2004 17:34:04 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 60797-01 for ; Mon, 25 Oct 2004 16:33:57 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 2B7E33A3AEC for ; Mon, 25 Oct 2004 17:33:57 +0100 (BST) 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 6553324; Mon, 25 Oct 2004 09:35:21 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 Date: Mon, 25 Oct 2004 09:33:07 -0700 User-Agent: KMail/1.6.2 Cc: Manfred Spraul , Mark Wong , Tom Lane , neilc@samurai.com References: <417221B5.1050704@colorfullife.com> <20041020150528.B7838@osdl.org> <41774D11.1030906@colorfullife.com> In-Reply-To: <41774D11.1030906@colorfullife.com> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410250933.07550.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/420 X-Sequence-Number: 8896 Manfred, > How complicated are Tom's test scripts? His immediate reply was that I > should retest with Fedora, to rule out any gentoo bugs. We've done some testing on other Linux. Linking in pthreads reduced CSes by < 15%, which was no appreciable impact on real performance. Gavin/Neil's full futex patch was of greater benefit; while it did not reduce CSes very much (25%) somehow the real performance benefit was greater. -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Mon Oct 25 17:45:23 2004 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 A51B13A4176 for ; Mon, 25 Oct 2004 17:45:22 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 61378-09 for ; Mon, 25 Oct 2004 16:45:16 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 9B5643A415D for ; Mon, 25 Oct 2004 17:45:15 +0100 (BST) 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 i9PGj6P7021094; Mon, 25 Oct 2004 12:45:06 -0400 (EDT) To: Manfred Spraul Cc: neilc@samurai.com, markw@osdl.org, pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 In-reply-to: <4176A2C1.3070205@colorfullife.com> References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> <417697A5.1050600@colorfullife.com> <7178.1098292535@sss.pgh.pa.us> <4176A2C1.3070205@colorfullife.com> Comments: In-reply-to Manfred Spraul message dated "Wed, 20 Oct 2004 19:39:13 +0200" Date: Mon, 25 Oct 2004 12:45:06 -0400 Message-ID: <21093.1098722706@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/421 X-Sequence-Number: 8897 Manfred Spraul writes: > But: According to the descriptions the problem is a context switch > storm. I don't see that cache line bouncing can cause a context switch > storm. What causes the context switch storm? As best I can tell, the CS storm arises because the backends get into some sort of lockstep timing that makes it far more likely than you'd expect for backend A to try to enter the bufmgr when backend B is already holding the BufMgrLock. In the profiles we were looking at back in April, it seemed that about 10% of the time was spent inside bufmgr (which is bad enough in itself) but the odds of LWLock collision were much higher than 10%, leading to many context swaps. This is not totally surprising given that they are running identical queries and so are running through loops of the same length, but still it seems like there must be some effect driving their timing to converge instead of diverge away from the point of conflict. What I think (and here is where it's a leap of logic, cause I can't prove it) is that the excessive time spent passing the spinlock cache line back and forth is exactly the factor causing that convergence. Somehow, the delay caused when a processor has to wait to get the cache line contributes to keeping the backend loops in lockstep. It is critical to understand that the CS storm is associated with LWLock contention not spinlock contention: what we saw was a lot of semop()s not a lot of select()s. > If it's the pg_usleep in s_lock, then my patch should help a lot: with > pthread_rwlock locks, this line doesn't exist anymore. The profiles showed that s_lock() is hardly entered at all, and the select() delay is reached even more seldom. So changes in that area will make exactly zero difference. This is the surprising and counterintuitive thing: oprofile clearly shows that very large fractions of the CPU time are being spent at the initial TAS instructions in LWLockAcquire and LWLockRelease, and yet those TASes hardly ever fail, as proven by the fact that oprofile shows s_lock() is seldom entered. So as far as the Postgres code can tell, there isn't any contention worth mentioning for the spinlock. This is indeed the way it was designed to be, but when so much time is going to the TAS instructions, you'd think there'd be more software-visible contention for the spinlock. It could be that I'm all wet and there is no relationship between the cache line thrashing and the seemingly excessive BufMgrLock contention. They are after all occurring at two very different levels of abstraction. But I think there is some correlation that we just don't understand yet. regards, tom lane From pgsql-performance-owner@postgresql.org Mon Oct 25 18:07:12 2004 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 7A4863A4174 for ; Mon, 25 Oct 2004 18:07:09 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 67142-10 for ; Mon, 25 Oct 2004 17:07:07 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 12EC83A4052 for ; Mon, 25 Oct 2004 18:07:07 +0100 (BST) 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 6553462; Mon, 25 Oct 2004 10:08:33 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: different io elevators in linux Date: Mon, 25 Oct 2004 10:09:17 -0700 User-Agent: KMail/1.6.2 Cc: Bjorn Bength References: <1098546845.7754.8.camel@localhost> In-Reply-To: <1098546845.7754.8.camel@localhost> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-15" Content-Transfer-Encoding: 7bit Message-Id: <200410251009.17575.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/422 X-Sequence-Number: 8898 Bjorn, > I haven't read much FAQs but has anyone done some benchmarks with > different io schedulers in linux with postgresql? According to OSDL, using the "deadline" scheduler sometimes results in a roughly 5% boost to performance, and sometimes none, depending on the application. We use it for all testing, though, just in case. --Josh -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-hackers-owner@postgresql.org Mon Oct 25 18:14:46 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 3639F3A4143 for ; Mon, 25 Oct 2004 18:14:45 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 72181-08 for ; Mon, 25 Oct 2004 17:14:42 +0000 (GMT) Received: from stark.xeocode.com (gsstark.mtl.istop.com [66.11.160.162]) by svr1.postgresql.org (Postfix) with ESMTP id 7FCE73A3AEC for ; Mon, 25 Oct 2004 18:14:42 +0100 (BST) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1CM8QZ-0001sm-00; Mon, 25 Oct 2004 13:14:39 -0400 To: pgsql-hackers@postgresql.org Subject: Re: [PATCHES] ARC Memory Usage analysis References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> In-Reply-To: <20651.1098720192@sss.pgh.pa.us> From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 25 Oct 2004 13:14:39 -0400 Message-ID: <87mzyavhxc.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/829 X-Sequence-Number: 60332 Tom Lane writes: > However, I'm still really nervous about the idea of using > effective_cache_size to control the ARC algorithm. That number is > usually entirely bogus. It wouldn't be too hard to have a port-specific function that tries to guess the total amount of memory. That isn't always right but it's at least a better ballpark default than a fixed arbitrary value. However I wonder about another approach entirely. If postgres timed how long reads took it shouldn't find it very hard to distinguish between a cached buffer being copied and an actual i/o operation. It should be able to track the percentage of time that buffers requested are in the kernel's cache and use that directly instead of the estimated cache size. Adding two gettimeofdays to every read call would be annoyingly expensive. But a port-specific function to check the cpu instruction counter could be useful. It doesn't have to be an accurate measurement of time (such as on some multi-processor machines) as long as it's possible to distinguish when a slow disk operation has occurred from when no disk operation has occurred. -- greg From pgsql-performance-owner@postgresql.org Sun Oct 31 05:24:29 2004 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 664D93A3AC7 for ; Mon, 25 Oct 2004 18:31:01 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 79206-05 for ; Mon, 25 Oct 2004 17:30:58 +0000 (GMT) Received: from dbl.q-ag.de (dbl.q-ag.de [213.172.117.3]) by svr1.postgresql.org (Postfix) with ESMTP id BC03D3A415D for ; Mon, 25 Oct 2004 18:30:55 +0100 (BST) Received: from [127.0.0.2] (dbl [127.0.0.1]) by dbl.q-ag.de (8.12.3/8.12.3/Debian-6.6) with ESMTP id i9PHUfSL030207; Mon, 25 Oct 2004 19:30:42 +0200 Message-ID: <417D383E.1010807@colorfullife.com> Date: Mon, 25 Oct 2004 19:30:38 +0200 From: Manfred Spraul User-Agent: Mozilla/5.0 (X11; U; Linux i686; fr-FR; rv:1.7.3) Gecko/20040922 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: neilc@samurai.com, markw@osdl.org, pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> <417697A5.1050600@colorfullife.com> <7178.1098292535@sss.pgh.pa.us> <4176A2C1.3070205@colorfullife.com> <21093.1098722706@sss.pgh.pa.us> In-Reply-To: <21093.1098722706@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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/472 X-Sequence-Number: 8948 Tom Lane wrote: >It could be that I'm all wet and there is no relationship between the >cache line thrashing and the seemingly excessive BufMgrLock contention. > > Is it important? The fix is identical in both cases: per-bucket locks for the hash table and a buffer aging strategy that doesn't need one global lock that must be acquired for every lookup. -- Manfred From pgsql-performance-owner@postgresql.org Mon Oct 25 18:45:38 2004 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 C916D3A4188 for ; Mon, 25 Oct 2004 18:45:37 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 84185-09 for ; Mon, 25 Oct 2004 17:45:35 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 605553A4165 for ; Mon, 25 Oct 2004 18:45:35 +0100 (BST) 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 i9PHjTkY021744; Mon, 25 Oct 2004 13:45:29 -0400 (EDT) To: Manfred Spraul Cc: neilc@samurai.com, markw@osdl.org, pgsql-performance@postgresql.org Subject: Re: futex results with dbt-3 In-reply-to: <417D383E.1010807@colorfullife.com> References: <417221B5.1050704@colorfullife.com> <20151.1098222733@sss.pgh.pa.us> <417697A5.1050600@colorfullife.com> <7178.1098292535@sss.pgh.pa.us> <4176A2C1.3070205@colorfullife.com> <21093.1098722706@sss.pgh.pa.us> <417D383E.1010807@colorfullife.com> Comments: In-reply-to Manfred Spraul message dated "Mon, 25 Oct 2004 19:30:38 +0200" Date: Mon, 25 Oct 2004 13:45:28 -0400 Message-ID: <21743.1098726328@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/423 X-Sequence-Number: 8899 Manfred Spraul writes: > Tom Lane wrote: >> It could be that I'm all wet and there is no relationship between the >> cache line thrashing and the seemingly excessive BufMgrLock contention. >> > Is it important? The fix is identical in both cases: per-bucket locks > for the hash table and a buffer aging strategy that doesn't need one > global lock that must be acquired for every lookup. Reducing BufMgrLock contention is a good idea, but it's not really my idea of a fix for this issue. In the absence of a full understanding, we may be fixing the wrong thing. It's worth remembering that when we first hit this issue, I made some simple changes that approximately halved the number of BufMgrLock acquisitions by joining ReleaseBuffer and ReadBuffer calls into ReleaseAndReadBuffer in all the places in the test case's loop. This made essentially no change in the CS storm behavior :-(. So I do not know how much contention we have to get rid of to get the problem to go away, or even whether this is the right path to take. (I am unconvinced that either of those specific suggestions is The Right Way to break up the bufmgrlock, either, but that's a different thread.) regards, tom lane From pgsql-performance-owner@postgresql.org Mon Oct 25 21:53:32 2004 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 94D753A3C35 for ; Mon, 25 Oct 2004 21:53:30 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 78858-01 for ; Mon, 25 Oct 2004 20:53:23 +0000 (GMT) Received: from vt-pe2550-001.VANTAGE.vantage.com (sol.vantage.com [64.80.203.242]) by svr1.postgresql.org (Postfix) with ESMTP id 43B623A3B16 for ; Mon, 25 Oct 2004 21:53:23 +0100 (BST) 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_01C4BAD4.AAC1D4BF" Subject: can't handle large number of INSERT/UPDATEs Date: Mon, 25 Oct 2004 16:53:23 -0400 Message-ID: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985D5@vt-pe2550-001.vantage.vantage.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: can't handle large number of INSERT/UPDATEs Thread-Index: AcS61KspcGsLLkbcT7etdP20bnwpiQ== From: "Anjan Dave" To: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.1 tagged_above=0.0 required=5.0 tests=HTML_60_70, HTML_MESSAGE X-Spam-Level: X-Archive-Number: 200410/424 X-Sequence-Number: 8900 This is a multi-part message in MIME format. ------_=_NextPart_001_01C4BAD4.AAC1D4BF Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Hi, =20 I am dealing with an app here that uses pg to handle a few thousand concurrent web users. It seems that under heavy load, the INSERT and UPDATE statements to one or two specific tables keep queuing up, to the count of 150+ (one table has about 432K rows, other has about 2.6Million rows), resulting in 'wait's for other queries, and then everything piles up, with the load average shooting up to 10+.=20 =20 We (development) have gone through the queries/explain analyzes and made sure the appropriate indexes exist among other efforts put in. =20 I would like to know if there is anything that can be changed for better from the systems perspective. Here's what I have done and some recent changes from the system side: =20 -Upgraded from 7.4.0 to 7.4.1 sometime ago -Upgraded from RH8 to RHEL 3.0 -The settings from postgresql.conf (carried over, basically) are: shared_buffers =3D 10240 (80MB) max_connections =3D 400 sort_memory =3D 1024 effective_cache_size =3D 262144 (2GB) checkpoint_segments =3D 15 stats_start_collector =3D true stats_command_string =3D true=20 Rest everything is at default =20 In /etc/sysctl.conf (512MB shared mem) kernel.shmall =3D 536870912 kernel.shmmax =3D 536870912 =20 -This is a new Dell 6650 (quad XEON 2.2GHz, 8GB RAM, Internal HW RAID10), RHEL 3.0 (2.4.21-20.ELsmp), PG 7.4.1 -Vaccum Full run everyday -contrib/Reindex run everyday -Disabled HT in BIOS =20 I would greatly appreciate any helpful ideas. =20 Thanks in advance, =20 Anjan ------_=_NextPart_001_01C4BAD4.AAC1D4BF Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable

Hi,

 

I am dealing with an app here that uses pg to handle a f= ew thousand concurrent web users. It seems that under heavy load, the INSERT a= nd UPDATE statements to one or two specific tables keep queuing up, to the cou= nt of 150+ (one table has about 432K rows, other has about 2.6Million rows), r= esulting in ‘wait’s for other queries, and then everything piles up, wit= h the load average shooting up to 10+.

 

We (development) have gone through the queries/explain analyzes and made sure the appropriate indexes exist among other efforts put in.

 

I would like to know if there is anything that can be changed for better from the systems perspective. Here’s what I have d= one and some recent changes from the system side:

 

-Upgraded from 7.4.0 to 7.4.1 sometime ago

-Upgraded from RH8 to RHEL 3.0<= /p>

-The settings from postgresql.conf (carried over, basica= lly) are:

         &n= bsp;  shared_buffers =3D 10240 (80MB)

         &n= bsp;  max_connections =3D 400

         &n= bsp;  sort_memory =3D 1024

         &n= bsp;  effective_cache_size =3D 262144 (2GB)

         &n= bsp;  checkpoint_segments =3D 15

stats_start_collector =3D true=

stats_command_string =3D true =

Rest everything is at default<= o:p>

 

In /etc/sysctl.conf (512MB sha= red mem)

kernel.shmall =3D 536870912

kernel.shmmax =3D 536870912

 

-This is a new Dell 6650 (quad XEON 2.2GHz, 8GB RAM, Internal HW RAID10), RHEL 3.0 (2.4.21-20.ELsmp), PG 7.4.1=

-Vaccum Full run everyday

-contrib/Reindex run everyday

-Disabled HT in BIOS

 

I would greatly appreciate any helpful ideas.=

 

Thanks in advance,

 

Anjan

------_=_NextPart_001_01C4BAD4.AAC1D4BF-- From pgsql-performance-owner@postgresql.org Mon Oct 25 22:20:14 2004 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 19DD03A3BF2 for ; Mon, 25 Oct 2004 22:20:13 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 86237-04 for ; Mon, 25 Oct 2004 21:20:02 +0000 (GMT) Received: from tht.net (vista.tht.net [216.126.88.2]) by svr1.postgresql.org (Postfix) with ESMTP id 192C33A3B16 for ; Mon, 25 Oct 2004 22:20:01 +0100 (BST) Received: from [134.22.69.212] (dyn-69-212.tor.dsl.tht.net [134.22.69.212]) by tht.net (Postfix) with ESMTP id AB24676B17; Mon, 25 Oct 2004 17:19:59 -0400 (EDT) Subject: Re: can't handle large number of INSERT/UPDATEs From: Rod Taylor To: Anjan Dave Cc: Postgresql Performance In-Reply-To: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985D5@vt-pe2550-001.vantage.vantage.com> References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985D5@vt-pe2550-001.vantage.vantage.com> Content-Type: text/plain; charset=iso-8859-7 Message-Id: <1098739153.8557.333.camel@home> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Mon, 25 Oct 2004 17:19:14 -0400 Content-Transfer-Encoding: 8bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/425 X-Sequence-Number: 8901 On Mon, 2004-10-25 at 16:53, Anjan Dave wrote: > Hi, > > > > I am dealing with an app here that uses pg to handle a few thousand > concurrent web users. It seems that under heavy load, the INSERT and > UPDATE statements to one or two specific tables keep queuing up, to > the count of 150+ (one table has about 432K rows, other has about > 2.6Million rows), resulting in �wait�s for other queries, and then This isn't an index issue, it's a locking issue. Sounds like you have a bunch of inserts and updates hitting the same rows over and over again. Eliminate that contention point, and you will have solved your problem. Free free to describe the processes involved, and we can help you do that. From pgsql-hackers-owner@postgresql.org Mon Oct 25 22:31:03 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 5214B3A421A for ; Mon, 25 Oct 2004 22:31:02 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 89626-07 for ; Mon, 25 Oct 2004 21:30:56 +0000 (GMT) Received: from stark.xeocode.com (gsstark.mtl.istop.com [66.11.160.162]) by svr1.postgresql.org (Postfix) with ESMTP id D5DD63A4211 for ; Mon, 25 Oct 2004 22:30:56 +0100 (BST) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1CMCQX-0004qr-00; Mon, 25 Oct 2004 17:30:53 -0400 To: Greg Stark Cc: pgsql-hackers@postgresql.org Subject: Re: [PATCHES] ARC Memory Usage analysis References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <87mzyavhxc.fsf@stark.xeocode.com> In-Reply-To: <87mzyavhxc.fsf@stark.xeocode.com> From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 25 Oct 2004 17:30:52 -0400 Message-ID: <87hdoiv62b.fsf@stark.xeocode.com> Lines: 51 User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.3 MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="=BATF-Ruby-Ridge-Europol-USDOJ-Vince-Foster-FIPS140-Albright-Baranyi-" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/850 X-Sequence-Number: 60353 --=BATF-Ruby-Ridge-Europol-USDOJ-Vince-Foster-FIPS140-Albright-Baranyi- Greg Stark writes: > However I wonder about another approach entirely. If postgres timed how long > reads took it shouldn't find it very hard to distinguish between a cached > buffer being copied and an actual i/o operation. It should be able to track > the percentage of time that buffers requested are in the kernel's cache and > use that directly instead of the estimated cache size. I tested this with a program that times seeking to random locations in a file. It's pretty easy to spot the break point. There are very few fetches that take between 50us and 1700us, probably they come from the drive's onboard cache. The 1700us bound probably would be lower for high end server equipment with 10k RPM drives and RAID arrays. But I doubt it will ever come close to the 100us edge, not without features like cache ram that Postgres would be better off considering to be part of "effective_cache" anyways. So I would suggest using something like 100us as the threshold for determining whether a buffer fetch came from cache. Here are two graphs, one showing a nice curve showing how disk seek times are distributed. It's neat to look at for that alone: --=BATF-Ruby-Ridge-Europol-USDOJ-Vince-Foster-FIPS140-Albright-Baranyi- Content-Type: image/png Content-Disposition: inline; filename=plot1.png Content-Transfer-Encoding: base64 iVBORw0KGgoAAAANSUhEUgAAAoAAAAHgCAMAAAACDyzWAAABKVBMVEX///8A AACgoKD/AAAAwAAAgP/AAP8A7u7AQADu7gAgIMD/wCAAgECggP+AQAD/gP8A wGAAwMAAYIDAYIAAgABA/4AwYICAYABAQEBAgAAAAICAYBCAYGCAYIAAAMAA AP8AYADjsMBAwIBgoMBgwABgwKCAAACAAIBgIIBgYGAgICAgQEAgQIBggCBg gGBggICAgEAggCCAgICgoKCg0ODAICAAgIDAYACAwODAYMDAgADAgGD/QAD/ QECAwP//gGD/gIDAoADAwMDA/8D/AAD/AP//gKDAwKD/YGAA/wD/gAD/oACA 4OCg4OCg/yDAAADAAMCgICCgIP+AIACAICCAQCCAQICAYMCAYP+AgADAwAD/ gED/oED/oGD/oHD/wMD//wD//4D//8BUJrxzAAAKoklEQVR4nO3djY7iNgCF UVDe/53bMuG3JMTGyY3DOdKKroRJ2PlqDxkGn04AAAAAAAA/5Xzxd3uauIHV jIX9Nfj+BtZzmwAFSMK9tfPUze2esFRphOeZCK/3qk1802FOs+m4bQ43tfY+ L8GHfOpfj+vkNPf79ObKE+DuDnfApzcu2HM3m51Lg2FOs+m4rU9z68fkoATI g9tCNvfqdPxxxNtx5UesHLf1Y7KJpwCnohq/6Rrvf34a98Ywf8SKs/xEgN16 mgAnoromd3rIdG4CFCBL3Tq6/tD/8SLHdZZ7E+DsAixASj00dn0Tyutk99Tq tOHPzKEanO4Wj8mGXqa6pwBfXqUsePVhBqTQ/9bayQCXvPgVIGUeGnv8Ydf5 6bu9j689bgRImYfIzk8z4Olx5T1fq/zuyy1AFlnr7Z4CZBEBckgCJEqARAmQ KAESJUCiBEiUAIkSIFECJEqARAmQKAGSMwzDefYN+3UEyGJmQKIESJQAiRIg UQIkSoBECZAoARIlQKIESFRZLA+fxrDCh5Tzg4piOT9/RkjrbRr4QSWx3D+P RoA0UhDLy6d1XefDh5vbHb/9yCR+QlkoD7Pc2/rMgFQoCfDxY6otwTRR8SpY gLRTHOD76y8uw1DHhWiiBEiUAIkSIFECJEqARAmQKAESJUCiBEiUAIkSIFEC JEqARAmQKAESJUCiBEiUAIkSIFECJEqARAmQKAESJUCiBEiUAIkSIFECJEqA RAmQKAGSMwzD+d8/rR9WgCxmBiRKgEQJkCgfUk5U1V5xtmmglYq94gRIOzV7 xb3dpstecZSq2itusj4zIBWK94qzBNOSveKIchmGKBeiiRIgUQIkSoBECZAo ARIlQKIESJQAiRIgUQIkSoBECZAoARIlQKIESJQAiRIgUQIkSoBECZAoARIl QKIESJQAiRIgUQIkxzYNpJkBiRIgUQIkSoBECZAoARIlQKIESFTRRjXXzyif /Kzy4sfk11VsVPP6N9s0UK9in5DzNTcB8rWKJfi+a5y94vhGeSj39ddecTRR H6AlmAaqXoQIkCbK3g84t0mcyzDUcSGaKAESJUCiBEiUAIkSIFECJEqARAmQ KAESJUCiBEiUAIkSIFECJEqARAmQKAESJUCiBEiUAIkSIFECJEqARAmQKAES JUCiBEiUAIkSIFECJEqARAmQKAESJUCiBEhU8UY1PqSclmzTQFTFXnECpJ2a veLebtNlrzhKVe0VN1mfGZAKlmCivAghymUYolyIJkqARAmQKAESJUCiBEiU AIkSIFECJEqARAmQKAESJUCiBEiUAIkSIFECJEqARAmQKAESJUCiBEiUAIkS IFECJEqARAmQKAESJUCiBEiUAMkZhuH875/WDytAFjMDEiVAogRIlACJqtio 5n57er0pfkx+XdU2Dc9/s00D9SoCvO7YJUC+98VOSfaK4zulodzuba84Wil6 EfLwX5Zgvlf0o7i5Vx8CpFLJi5CZ6y8uw1DHhWhyvBuGsMEMSJIACbIEkzQI kDBLMFECJGdcgps/rgBZygxIlACJEiBRAiRKgEQJkByXYUgzAxIlQKIESJYA iRIgUQIkSoBECZAoAZLkMgxRAiRKgOR4MwJpZkCiBEiUAMkSIFECJEqARAmQ qOK94nxIOQ0VvAqe25/BNg3UESBRhdcBpzaJs1ccFUp/Fnzdp8ZecTRRGOBf YZZgGir8HvAkQJoqCtBecTTl/YCkeTcMUQIkxxJMlABJswQTJUByLMFECZA0 SzBRAiRKgGQJkCgBkmQJJkqARAmQHBeiSTMDEiVAogRIlACJEiA5XgUTJUDS LMFECZAoARIlQKIESJQAyXEZhjQzIDlmQMIGMyA5ZkCShvJ9Qk4fPqt8vF/z U+Wgipbga2rjX2zTwHfGGXDp3c8PM6AAaaLsRchtBrRXHC2cz0NRKLc72yuO BgqX4NcALcF8py5Ae8XRzlAQi73iaKt4BlxKgCwhQKIESJQASRoESNIY4AqP LEA+GwRIzjAIkJhbfiUXopcTIHOGh/4EyKaGJycBsqFh+F9/AmQjr/WNvwsi QDbw//quv4skQNb2rr7bzz8EyIretjc8/iamAFnJVHzPvwgsQFawLL7/CJDG puN798YXAdLOTHtTn8AhQNqYjW/6TX8C5Gvz7c1/+pAA+cY37V0IkEqf2lv2 XnsBUqFNfP8RIGU+tlf2W0YCZLHG7V0IkCU+t1f565UC5IPV2rsQINPWbe9C gLyzIL02H2kgQF5s1t6FALlZkl7rfWUEyH8S7V0I8NctS2+N9i4E+LMWlrde excC/EGL01u3vYvyfUJ8SHm/lpe3QXp/qvaKs01DdwrS26y9i4JtGuwT0qOS 8rZN70/NXnFvt+myV9zeFKWXaK84lOt3e/aK27ey8iLp3X2xW6YleG+6Km8k wEPosLyRveL61uOk98SF6E4VlrfD9P4IsDdHKW8kwH4cLL0/AuzBEcsbCXDf jlveSIA7deBJ74kAd+dHyhsJcE9+qbyRAHfi58obCTDt9ya9JwKM+cH19g0B BkjvToCbUt4rAW5Eeu8JcG0l6+1vtXchwPUUpfeD7V0IcA3KW0yAbSmvkAAb MenVEWAD0qsnwG8o72sCrCS9NgRYTHotCbCA9NoT4BLLy5NeIQHOU97KBDih YNIT3xcE+Ep5mxLgXVF62mtDgKff+R3cPfr1AKUX9rsBKm8XfjJA6e3HjwWo vL35nQCVt0u/EaD0dqs4ltnPKq98zFUpb9cqArzd7H+bBu3tXt0M2EWA0utB 1QzYwV5x2utAdSh73ytOev2oD3C3S7D2evLFErzLALXXmfJY9rxXnPq6c6AL 0eLr0WECVF+fDhKg+np1iADV168jBKi+jvUfoPy61n2A8utb7wGqr3OdByi/ 3nUdoPz613OA8juAjgPU3xH0G6D8DqHbAPV3DL0GqL+D6D3ADQ7FmjoNUH9H 0WeA+juMrgNc/TisrssATYDH0XOAax+GDQiQqB4DtAIfSMcBrnwUNiFAonoM 8PRfg+sfgy30GuD6h2ATAiRKgEQJkKgeA/Qq+EB6DPBkBjyOPgM0/R1GlwFa f4+jxwB9D3ggewpw4bjX9yKsfLhW4zo5zV6e3uvDPH1I+bCFr5/Cwb9CB396 L4/yvE3Dpv35Cu1i3K8F2OApHPwrdPCn9/Io973i/o3jDIs1CvBxr7hN5z94 XYL1x8YESFazxRwAAAA6U/ia+O/O46Dnm88HKR1WPe50vdpUNO56tX+r00wc r3jYzBe8yeWU56uCC+5+vg96vvl8kNJh1eNO95/3lJ5mxbP75jS3/GepGjbz BS9MZ/6slt77tGmA49iKcee6AM/XOxUHMfUsPz+7XQc49wVvFmDRRHo9n/EL /HCz5EClw25FFI+rOd79H7bscOfa03w45Ab/LHVPb/oLXprO5HMpyng8nzcn 8/EwFcNO08/900nWjNv8cHX/mrXHO4/rQtGw6VMsTWf2yRTd/+H/pYebj6PK h42DKg5XtQTXH67+O4Wp4escr2LY9CmWpjP7ZIruX3wi58pvkr553hXjag9X fZqV/19WHq9u2NoBnopeS1deOKgcNjFg2RnXHK/2cPXjpoevcLyKYbNfuaJ0 AAAAAACgzj9lVEYweskfcwAAAABJRU5ErkJggg== --=BATF-Ruby-Ridge-Europol-USDOJ-Vince-Foster-FIPS140-Albright-Baranyi- This is the 1000 fastest data points zoomed to the range under 1800us: --=BATF-Ruby-Ridge-Europol-USDOJ-Vince-Foster-FIPS140-Albright-Baranyi- Content-Type: image/png Content-Disposition: inline; filename=plot2.png Content-Transfer-Encoding: base64 iVBORw0KGgoAAAANSUhEUgAAAoAAAAHgCAMAAAACDyzWAAABKVBMVEX///8A AACgoKD/AAAAwAAAgP/AAP8A7u7AQADu7gAgIMD/wCAAgECggP+AQAD/gP8A wGAAwMAAYIDAYIAAgABA/4AwYICAYABAQEBAgAAAAICAYBCAYGCAYIAAAMAA AP8AYADjsMBAwIBgoMBgwABgwKCAAACAAIBgIIBgYGAgICAgQEAgQIBggCBg gGBggICAgEAggCCAgICgoKCg0ODAICAAgIDAYACAwODAYMDAgADAgGD/QAD/ QECAwP//gGD/gIDAoADAwMDA/8D/AAD/AP//gKDAwKD/YGAA/wD/gAD/oACA 4OCg4OCg/yDAAADAAMCgICCgIP+AIACAICCAQCCAQICAYMCAYP+AgADAwAD/ gED/oED/oGD/oHD/wMD//wD//4D//8BUJrxzAAAILklEQVR4nO3di1baSgBA UbL8/3+uiiCoWJMZPTz2XsXcW0sY09OZJNjV3Q4AAAAAAODBLMv+4+v2fAO/ 7lDc6+N8A79uX5sAyRwCXPblnWyOvwBWWh/gx/pOApxY+zCDueRmB/PfJfhm v7JfZzCXCPBPGMwlPx/MYcV+W7fPN2t39gcM5pL7HcxVfWXcBAFy4riefXd1 +rYOfvm89a+48Xm/vzMCZwFeiup4LfB+BvZdgE/fv+KGUf7NzgicTYAXojok tzvJ9LsJUID81LGjw9Xm6b2Owyz3RYDfLsACZK2Txg53Pz5OdmetXva0981L TRju7+yMzIep7izAD1cpP7j6MAOy0qe19mKAP7n4FSDrnDR2+p7Xcna2999r jyMBss5JZMvZDLg7XXmXQ5Vjv+sC5Ed+67s+BciPCJC7JEBSAiQlQFICJCVA UgIkJUBSAiQlQFICJCVAUgIkJUBSAiQlQFICJCVAUgIkJUBSAiQlQFICJCVA UgIkJUBSAiQlQFICJCVAUgIkJUBSAiQlQFICJCVAUgIkJUBSAiQlQFICJCVA UgIkJUBSAiQlQFICJCVAUgIkJUBSAiQlQFICJCVAUgIkJUBSAiQlQFICJCVA UgIkJUBSAiQlQFICJCVAUgIkJUBS65tZnn3abN0Zj251My+9PT/ON1t3xsMT IKmtS/BrfMthc/IpEfJTm3r5XJ8ZkO0swaQESMptGFJuRJMSICkBkhIgKQGS EiApAZISICkBkhIgKQGSEiApAZISICkBkhIgKQGSEiApAZISICkBkhIgKQGS EiApAZISICkBkhIgKQGSEiApAZISICkBkhIgKQGSEiApAZISICkBkhIgKQGS EiApAZISICkBkhIgKQGSEiApAZISICkBkhIgKQGSEiApAZISICkBkhIgKQGS EiApAZISICkBkhIgKQGSEiApAZISICkBkhIgKQGSEiApAZISICkBkhIgKQGS EiApAZISIKn1zSzPPm227oxHt7qZl96eH+ebrTvj4W0KcPexw6074+FtCHA5 xrccNuefgx/Z1Mvn+syAbOcckJQASbkNQ8qNaFICJCVAUgIkJUBSAiQlQFIC JCVAUgIkJUBSAiQlQFICJCVAUgIkJUBSAiQlQFICJCVAUgIkJUBSAiQlQFIC JCVAUgIkJUBGPY08WYCMEiApAdJ52tv8fAEyygxISoCkBMjtEiApATJoaAUW IKMESEqAdEbvQwuQUWZAUgIkJUBumAAZZAYkJUBSAqTjPiA1MyApAVIaWX93 AmTUWH8CZJAA6QxfBAuQQWZAUgIkJUBSAiQlQFICpOM2DLHBd+IEyBgBErIE UxrvT4AMGcxPgIwRICFLMKUJ/QmQ7QRIbfQUUIBsZwakNjoBCpAhAiQlQErD 1yACZDsXIbSG34gTINvNmAC3NbMsL4+Xj4fNwM64UU9ZgC/N7Rs8bAZ2xq2a 0t+WZhYBsjd+CrihmX12rz+Om+OnThdk7t7oFfCWXo5T32mEh88NDYebMz4B rm9mWSzB7E1YgTdfBQuQNEC3YR5edhX8Rzvjus3pT4BsJEBSAiQ24xpEgGwz aQIUINsIkJQAKc35XqydANlqRn07AbKVAEkJkM60U0ABstGcCVCAbCRAUgIk Nac/AbKRAOlMuwgWIBuZAUkJkJQASQmQjosQamZASpPeCBEgW1iCiZkB6cyb AAXIBgKkNmkFFiDbCJCUAEkJkJQA6bgKJjbrPrQA2USApARIZ+IpoABZT4DU Zq3AAmQTAZISIKVp/QmQLQRISoB0Zt6FESCrCZDYtDfiBMh6ZkBiZkA6UydA AbKBGZDUvP4EyAYC5F4IkJQASQmQlABJCZCUAEkJkJQAWc+NaFICJCVAOr4d i5oZkJQASQmQeyFAUgIkJUDWcw5ISoCUJv6tTAGylndCiJkB6cydAAXIeu0M uDz7tNm6M27RxP7WN/PS2/PjfLN1Z9yeegkW4IOrA3x90rL/cdwcf/50QeZe TbsFs62XZT/tnUZ4+MykgXHd0nPAt94swQ8svwjZCfChxQEuhzXYbZhHdA0X IX+yM67VzPvQAmQ1AdKZvAILkNVmToACZDUBkhIgHeeA1MyAdMyA1MyApARI SoDcDwGSEiApAZISICkBkhIgKQGSEiApAZISICkBkhIgKQGSEiApAZISICkB khIgKQGSEiApAZISICkBkhIgKQGSEiApAZISICkBkhIgKQGSEiApAZISICkB khIgKQGSEiApAbLS1H8mRICsJUBSAqQz+x/LFCBrmQFJCZCUALkjAiQlQFIC ZCXngKQESEqAdLwTQs0MSEqApARISoCUZl6B7ATIOq6CSQmQ0vT+BMjPPT0J kMrTiYm7FSAXPF0081UEeLUuB1Cb+VVeW4D1seW/xn+TT80NsD42/L6ZwewE yFozg9mNB7g8e/+f+uDwqwZb+Tqgwacv+8fu9exNgPdsQm1fFTT49PcAd2bA OzPa1s8KGnz6sjtZgxdYazTA5WQGvJk/dtyL8yUY/pgAaY2v4gAAAHCbruKa +HCD/W0w6Zj2L30+kmxAx1e/gqPz5SEZHsx13BV8G8HbYNIxHY7r6UiyAb0P 5tOYgrF8cUjGB3MtAS6HgcQBLrsrCvA4mGs4Or8X4BWswe9fzvK+CcfyYSTd gD78aaiPzqdRjA/m/FsTSp++sm4cuy+OcRjg239cwWCW+UfmOpbgF/2KdxzH tSzBHwOsT0gPH+/xHPAqDvFxLFczoCsazPLVpdCEwVzFbZjDKOq7HpdueUQD MhgAAAAAAHjzD9bGNv6j2lw6AAAAAElFTkSuQmCC --=BATF-Ruby-Ridge-Europol-USDOJ-Vince-Foster-FIPS140-Albright-Baranyi- This is the program I wrote to test this: --=BATF-Ruby-Ridge-Europol-USDOJ-Vince-Foster-FIPS140-Albright-Baranyi- Content-Type: text/x-csrc Content-Disposition: attachment; filename=timing.c #include #include #include #include #include #include #include #include int main(int argc, char *argv[]) { int rep = atoi(argv[1]); int i; char *filename; int fd; struct stat statbuf; off_t filesize; unsigned blocksize; void *blockbuf; filename = argv[2]; fd = open(filename, O_RDONLY); fstat(fd, &statbuf); filesize = statbuf.st_size; blocksize = statbuf.st_blksize; blockbuf = malloc(blocksize); srandom(getpid()^clock()); for (i=0;i /tmp/l $ gnuplot gnuplot> set terminal png gnuplot> set output "/tmp/plot1.png" gnuplot> plot '/tmp/l2' with points pointtype 1 pointsize 1 gnuplot> set output "/tmp/plot2.png" gnuplot> plot [0:2000] [0:1000] '/tmp/l2' with points pointtype 1 pointsize 1 -- greg --=BATF-Ruby-Ridge-Europol-USDOJ-Vince-Foster-FIPS140-Albright-Baranyi--- From pgsql-hackers-owner@postgresql.org Mon Oct 25 22:53:45 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 2390B3A3C35 for ; Mon, 25 Oct 2004 22:53:43 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 95841-08 for ; Mon, 25 Oct 2004 21:53:32 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 2DFF13A422C for ; Mon, 25 Oct 2004 22:53:32 +0100 (BST) 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 i9PLrQWC000791; Mon, 25 Oct 2004 17:53:26 -0400 (EDT) To: Greg Stark Cc: pgsql-hackers@postgresql.org Subject: Re: [PATCHES] ARC Memory Usage analysis In-reply-to: <87hdoiv62b.fsf@stark.xeocode.com> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <87mzyavhxc.fsf@stark.xeocode.com> <87hdoiv62b.fsf@stark.xeocode.com> Comments: In-reply-to Greg Stark message dated "25 Oct 2004 17:30:52 -0400" Date: Mon, 25 Oct 2004 17:53:25 -0400 Message-ID: <790.1098741205@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/853 X-Sequence-Number: 60356 Greg Stark writes: > So I would suggest using something like 100us as the threshold for > determining whether a buffer fetch came from cache. I see no reason to hardwire such a number. On any hardware, the distribution is going to be double-humped, and it will be pretty easy to determine a cutoff after minimal accumulation of data. The real question is whether we can afford a pair of gettimeofday() calls per read(). This isn't a big issue if the read actually results in I/O, but if it doesn't, the percentage overhead could be significant. If we assume that the effective_cache_size value isn't changing very fast, maybe it would be good enough to instrument only every N'th read (I'm imagining N on the order of 100) for this purpose. Or maybe we need only instrument reads that are of blocks that are close to where the ARC algorithm thinks the cache edge is. One small problem is that the time measurement gives you only a lower bound on the time the read() actually took. In a heavily loaded system you might not get the CPU back for long enough to fool you about whether the block came from cache or not. Another issue is what we do with the effective_cache_size value once we have a number we trust. We can't readily change the size of the ARC lists on the fly. regards, tom lane From pgsql-hackers-owner@postgresql.org Wed Oct 27 18:09:25 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id ECE523A424A for ; Mon, 25 Oct 2004 23:48:43 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 12603-04 for ; Mon, 25 Oct 2004 22:48:32 +0000 (GMT) Received: from is.rice.edu (is.rice.edu [128.42.42.24]) by svr1.postgresql.org (Postfix) with ESMTP id E4A493A4230 for ; Mon, 25 Oct 2004 23:48:34 +0100 (BST) Received: from localhost (localhost [127.0.0.1]) by localhost.is.rice.edu (Postfix) with ESMTP id E04AC4191D; Mon, 25 Oct 2004 17:48:33 -0500 (CDT) Received: from is.rice.edu ([127.0.0.1]) by localhost (it.is.rice.edu [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 28089-10; Mon, 25 Oct 2004 17:48:30 -0500 (CDT) Received: by is.rice.edu (Postfix, from userid 18612) id 87D0C418C9; Mon, 25 Oct 2004 17:48:30 -0500 (CDT) Date: Mon, 25 Oct 2004 17:48:30 -0500 From: Kenneth Marshall To: Tom Lane Cc: Greg Stark , pgsql-hackers@postgresql.org Subject: Re: [PATCHES] ARC Memory Usage analysis Message-ID: <20041025224830.GQ12041@it.is.rice.edu> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <87mzyavhxc.fsf@stark.xeocode.com> <87hdoiv62b.fsf@stark.xeocode.com> <790.1098741205@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <790.1098741205@sss.pgh.pa.us> User-Agent: Mutt/1.4.2i X-Virus-Scanned: by amavis-20030314-p2 at is.rice.edu X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/942 X-Sequence-Number: 60445 On Mon, Oct 25, 2004 at 05:53:25PM -0400, Tom Lane wrote: > Greg Stark writes: > > So I would suggest using something like 100us as the threshold for > > determining whether a buffer fetch came from cache. > > I see no reason to hardwire such a number. On any hardware, the > distribution is going to be double-humped, and it will be pretty easy to > determine a cutoff after minimal accumulation of data. The real question > is whether we can afford a pair of gettimeofday() calls per read(). > This isn't a big issue if the read actually results in I/O, but if it > doesn't, the percentage overhead could be significant. > How invasive would reading the "CPU counter" be, if it is available? A read operation should avoid flushing a cache line and we can throw out the obvious outliers since we only need an estimate and not the actual value. --Ken From pgsql-hackers-owner@postgresql.org Tue Oct 26 00:11:54 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 8850A3A3AC2 for ; Tue, 26 Oct 2004 00:11:16 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 17224-05 for ; Mon, 25 Oct 2004 23:11:07 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id E425D3A4235 for ; Tue, 26 Oct 2004 00:11:09 +0100 (BST) 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 i9PNB6lF002047; Mon, 25 Oct 2004 19:11:06 -0400 (EDT) To: Kenneth Marshall Cc: Greg Stark , pgsql-hackers@postgresql.org Subject: Re: [PATCHES] ARC Memory Usage analysis In-reply-to: <20041025224830.GQ12041@it.is.rice.edu> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <87mzyavhxc.fsf@stark.xeocode.com> <87hdoiv62b.fsf@stark.xeocode.com> <790.1098741205@sss.pgh.pa.us> <20041025224830.GQ12041@it.is.rice.edu> Comments: In-reply-to Kenneth Marshall message dated "Mon, 25 Oct 2004 17:48:30 -0500" Date: Mon, 25 Oct 2004 19:11:05 -0400 Message-ID: <2046.1098745865@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/860 X-Sequence-Number: 60363 Kenneth Marshall writes: > How invasive would reading the "CPU counter" be, if it is available? Invasive or not, this is out of the question; too unportable. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Oct 26 03:29:24 2004 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 1A7903A3BBE for ; Tue, 26 Oct 2004 03:29:23 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 67283-03 for ; Tue, 26 Oct 2004 02:29:17 +0000 (GMT) Received: from purple.west.spy.net (mail.west.spy.net [66.149.231.226]) by svr1.postgresql.org (Postfix) with ESMTP id 239B83A3B33 for ; Tue, 26 Oct 2004 03:29:17 +0100 (BST) Received: from [192.168.1.50] (dustinti.west.spy.net [192.168.1.50]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (Client did not present a certificate) by purple.west.spy.net (Postfix) with ESMTP id CAB19127; Mon, 25 Oct 2004 19:29:16 -0700 (PDT) In-Reply-To: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985D5@vt-pe2550-001.vantage.vantage.com> References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985D5@vt-pe2550-001.vantage.vantage.com> Mime-Version: 1.0 (Apple Message framework v619) Content-Type: text/plain; charset=WINDOWS-1252; format=flowed Message-Id: Content-Transfer-Encoding: quoted-printable Cc: From: Dustin Sallings Subject: Re: can't handle large number of INSERT/UPDATEs Date: Mon, 25 Oct 2004 19:29:19 -0700 To: "Anjan Dave" X-Mailer: Apple Mail (2.619) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/426 X-Sequence-Number: 8902 On Oct 25, 2004, at 13:53, Anjan Dave wrote: > I am dealing with an app here that uses pg to handle a few thousand=20 > concurrent web users. It seems that under heavy load, the INSERT and=20 > UPDATE statements to one or two specific tables keep queuing up, to=20 > the count of 150+ (one table has about 432K rows, other has about=20 > 2.6Million rows), resulting in =91wait=92s for other queries, and then=20 > everything piles up, with the load average shooting up to 10+. Depending on your requirements and all that, but I had a similar issue=20 in one of my applications and made the problem disappear entirely by=20 serializing the transactions into a separate thread (actually, a thread=20 pool) responsible for performing these transactions. This reduced the=20 load on both the application server and the DB server. Not a direct answer to your question, but I've found that a lot of=20 times when someone has trouble scaling a database application, much of=20 the performance win can be in trying to be a little smarter about how=20 and when the database is accessed. -- SPY My girlfriend asked me which one I like better. pub 1024/3CAE01D5 1994/11/03 Dustin Sallings | Key fingerprint =3D 87 02 57 08 02 D0 DA D6 C8 0F 3E 65 51 98 D8 BE L_______________________ I hope the answer won't upset her. ____________ From pgsql-hackers-owner@postgresql.org Tue Oct 26 06:18:05 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 65B863A3C7F for ; Tue, 26 Oct 2004 06:18:03 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 09697-01 for ; Tue, 26 Oct 2004 05:17:48 +0000 (GMT) Received: from stark.xeocode.com (gsstark.mtl.istop.com [66.11.160.162]) by svr1.postgresql.org (Postfix) with ESMTP id 36A403A3AC4 for ; Tue, 26 Oct 2004 06:17:48 +0100 (BST) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1CMJiH-0000Yt-00; Tue, 26 Oct 2004 01:17:41 -0400 To: Tom Lane Cc: Greg Stark , pgsql-hackers@postgresql.org Subject: Re: [PATCHES] ARC Memory Usage analysis References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <87mzyavhxc.fsf@stark.xeocode.com> <87hdoiv62b.fsf@stark.xeocode.com> <790.1098741205@sss.pgh.pa.us> In-Reply-To: <790.1098741205@sss.pgh.pa.us> From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 26 Oct 2004 01:17:40 -0400 Message-ID: <878y9uukgb.fsf@stark.xeocode.com> Lines: 49 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/875 X-Sequence-Number: 60378 Tom Lane writes: > I see no reason to hardwire such a number. On any hardware, the > distribution is going to be double-humped, and it will be pretty easy to > determine a cutoff after minimal accumulation of data. Well my stats-fu isn't up to the task. My hunch is that the wide range that the disk reads are spread out over will throw off more sophisticated algorithms. Eliminating hardwired numbers is great, but practically speaking it's not like any hardware is ever going to be able to fetch the data within 100us. If it does it's because it's really a solid state drive or pulling the data from disk cache and therefore really ought to be considered part of effective_cache_size anyways. > The real question is whether we can afford a pair of gettimeofday() calls > per read(). This isn't a big issue if the read actually results in I/O, but > if it doesn't, the percentage overhead could be significant. My thinking was to use gettimeofday by default but allow individual ports to provide a replacement function that uses the cpu TSC counter (via rdtsc) or equivalent. Most processors provide such a feature. If it's not there then we just fall back to gettimeofday. Your idea to sample only 1% of the reads is a fine idea too. My real question is different. Is it worth heading down this alley at all? Or will postgres eventually opt to use O_DIRECT and boost the size of its buffer cache? If it goes the latter route, and I suspect it will one day, then all of this is a waste of effort. I see mmap or O_DIRECT being the only viable long-term stable states. My natural inclination was the former but after the latest thread on the subject I suspect it'll be forever out of reach. That makes O_DIRECT And a Postgres managed cache the only real choice. Having both caches is just a waste of memory and a waste of cpu cycles. > Another issue is what we do with the effective_cache_size value once we > have a number we trust. We can't readily change the size of the ARC > lists on the fly. Huh? I thought effective_cache_size was just used as an factor the cost estimation equation. My general impression was that a higher effective_cache_size effectively lowered your random page cost by making the system think that fewer nonsequential block reads would really incur the cost. Is that wrong? Is it used for anything else? -- greg From pgsql-hackers-owner@postgresql.org Tue Oct 26 06:27:48 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 50E803A3B90 for ; Tue, 26 Oct 2004 06:27:47 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 11241-03 for ; Tue, 26 Oct 2004 05:27:32 +0000 (GMT) Received: from stark.xeocode.com (gsstark.mtl.istop.com [66.11.160.162]) by svr1.postgresql.org (Postfix) with ESMTP id 8C2C83A430C for ; Tue, 26 Oct 2004 06:27:33 +0100 (BST) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1CMJrm-0000d7-00; Tue, 26 Oct 2004 01:27:30 -0400 To: Greg Stark , pgsql-hackers@postgresql.org Subject: Re: [PATCHES] ARC Memory Usage analysis References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <87mzyavhxc.fsf@stark.xeocode.com> <87hdoiv62b.fsf@stark.xeocode.com> <790.1098741205@sss.pgh.pa.us> In-Reply-To: <790.1098741205@sss.pgh.pa.us> From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 26 Oct 2004 01:27:29 -0400 Message-ID: <873c02ujzy.fsf@stark.xeocode.com> Lines: 12 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/876 X-Sequence-Number: 60379 Is something broken with the list software? I'm receiving other emails from the list but I haven't received any of the mails in this thread. I'm only able to follow the thread based on the emails people are cc'ing to me directly. I think I've caught this behaviour in the past as well. Is it a misguided list software feature trying to avoid duplicates or something like that? It makes it really hard to follow threads in MUAs with good filtering since they're fragmented between two mailboxes. -- greg From pgsql-hackers-owner@postgresql.org Tue Oct 26 06:54:16 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CBC0E3A3C10 for ; Tue, 26 Oct 2004 06:54:14 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 15216-08 for ; Tue, 26 Oct 2004 05:54:00 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 713393A3B33 for ; Tue, 26 Oct 2004 06:53:59 +0100 (BST) 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 i9Q5rtRE004889; Tue, 26 Oct 2004 01:53:55 -0400 (EDT) To: Greg Stark Cc: pgsql-hackers@postgresql.org Subject: Re: [PATCHES] ARC Memory Usage analysis In-reply-to: <878y9uukgb.fsf@stark.xeocode.com> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <87mzyavhxc.fsf@stark.xeocode.com> <87hdoiv62b.fsf@stark.xeocode.com> <790.1098741205@sss.pgh.pa.us> <878y9uukgb.fsf@stark.xeocode.com> Comments: In-reply-to Greg Stark message dated "26 Oct 2004 01:17:40 -0400" Date: Tue, 26 Oct 2004 01:53:55 -0400 Message-ID: <4887.1098770035@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/877 X-Sequence-Number: 60380 Greg Stark writes: > Tom Lane writes: >> Another issue is what we do with the effective_cache_size value once we >> have a number we trust. We can't readily change the size of the ARC >> lists on the fly. > Huh? I thought effective_cache_size was just used as an factor the cost > estimation equation. Today, that is true. Jan is speculating about using it as a parameter of the ARC cache management algorithm ... and that worries me. regards, tom lane From pgsql-hackers-owner@postgresql.org Tue Oct 26 07:05:14 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4E85E3A3C10 for ; Tue, 26 Oct 2004 07:05:13 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 19796-04 for ; Tue, 26 Oct 2004 06:05:06 +0000 (GMT) Received: from platonic.cynic.net (platonic.cynic.net [204.80.150.245]) by svr1.postgresql.org (Postfix) with ESMTP id 1986F3A3B28 for ; Tue, 26 Oct 2004 07:05:04 +0100 (BST) Received: from angelic.cynic.net (angelic-platonic.cvpn.cynic.net [198.73.220.226]) by platonic.cynic.net (Postfix) with ESMTP id 2472EC003; Tue, 26 Oct 2004 15:05:02 +0900 (JST) Received: from localhost (localhost [127.0.0.1]) by angelic.cynic.net (Postfix) with ESMTP id 6737B8736; Tue, 26 Oct 2004 15:04:58 +0900 (JST) Date: Tue, 26 Oct 2004 15:04:58 +0900 (JST) From: Curt Sampson X-X-Sender: cjs@angelic-vtfw.cvpn.cynic.net To: Greg Stark Cc: pgsql-hackers@postgresql.org Subject: Re: [PATCHES] ARC Memory Usage analysis In-Reply-To: <878y9uukgb.fsf@stark.xeocode.com> Message-ID: References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <87mzyavhxc.fsf@stark.xeocode.com> <87hdoiv62b.fsf@stark.xeocode.com> <790.1098741205@sss.pgh.pa.us> <878y9uukgb.fsf@stark.xeocode.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/878 X-Sequence-Number: 60381 On Tue, 26 Oct 2004, Greg Stark wrote: > I see mmap or O_DIRECT being the only viable long-term stable states. My > natural inclination was the former but after the latest thread on the subject > I suspect it'll be forever out of reach. That makes O_DIRECT And a Postgres > managed cache the only real choice. Having both caches is just a waste of > memory and a waste of cpu cycles. I don't see why mmap is any more out of reach than O_DIRECT; it's not all that much harder to implement, and mmap (and madvise!) is more widely available. But if using two caches is only costing us 1% in performance, there's not really much point.... cjs -- Curt Sampson +81 90 7737 2974 http://www.NetBSD.org Make up enjoying your city life...produced by BIC CAMERA From pgsql-performance-owner@postgresql.org Tue Oct 26 09:18:31 2004 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 BF00D3A3C24 for ; Tue, 26 Oct 2004 09:18:30 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 52214-10 for ; Tue, 26 Oct 2004 08:18:14 +0000 (GMT) Received: from mra02.ex.eclipse.net.uk (mra02.ex.eclipse.net.uk [212.104.129.89]) by svr1.postgresql.org (Postfix) with ESMTP id 042813A3B28 for ; Tue, 26 Oct 2004 09:18:13 +0100 (BST) Received: from localhost (localhost.localdomain [127.0.0.1]) by mra02.ex.eclipse.net.uk (Postfix) with ESMTP id 740E840639F; Tue, 26 Oct 2004 09:12:03 +0100 (BST) Received: from mra02.ex.eclipse.net.uk ([127.0.0.1]) by localhost (mra02.ex.eclipse.net.uk [127.0.0.1]) (amavisd-new, port 10024) with LMTP id 11507-01-42; Tue, 26 Oct 2004 09:12:01 +0100 (BST) Received: from dev1 (unknown [82.152.145.238]) by mra02.ex.eclipse.net.uk (Postfix) with ESMTP id 6C5494070A9; Tue, 26 Oct 2004 09:11:55 +0100 (BST) From: "Rod Dutton" To: "'Anjan Dave'" Cc: Subject: FW: can't handle large number of INSERT/UPDATEs Date: Tue, 26 Oct 2004 09:19:33 +0100 MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit X-Mailer: Microsoft Office Outlook, Build 11.0.5510 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1106 Thread-Index: AcS62HV8K9yQuVetTcSgq4oyu4L6aAAWftkw Message-Id: <20041026081155.6C5494070A9@mra02.ex.eclipse.net.uk> X-Virus-Scanned: by Eclipse VIRUSshield at eclipse.net.uk X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/427 X-Sequence-Number: 8903 >>Eliminate that contention point, and you will have solved your problem. I agree, If your updates are slow then you will get a queue building up. Make sure that:- 1) all your indexing is optimised. 2) you are doing regular vacuuming (bloated tables will cause a slow down due to swapping). 3) your max_fsm_pages setting is large enough - it needs to be big enough to hold all the transactions between vacuums (+ some spare for good measure). 4) do a full vacuum - do one to start and then do one after you have had 2&3 (above) in place for a while - if the full vacuum handles lots of dead tuples then your max_fsm_pages setting is too low. 5) Also try reindexing or drop/recreate the indexes in question as... "PostgreSQL is unable to reuse B-tree index pages in certain cases. The problem is that if indexed rows are deleted, those index pages can only be reused by rows with similar values. For example, if indexed rows are deleted and newly inserted/updated rows have much higher values, the new rows can't use the index space made available by the deleted rows. Instead, such new rows must be placed on new index pages. In such cases, disk space used by the index will grow indefinitely, even if VACUUM is run frequently. " Are your updates directly executed or do you use stored procs? We had a recent problem with stored procs as they store a "one size fits all" query plan when compiled - this can be less than optimum in some cases. We have a similar sounding app to yours and if tackled correctly then all the above will make a massive difference in performance. Rod -----Original Message----- From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of Rod Taylor Sent: 25 October 2004 22:19 To: Anjan Dave Cc: Postgresql Performance Subject: Re: [PERFORM] can't handle large number of INSERT/UPDATEs On Mon, 2004-10-25 at 16:53, Anjan Dave wrote: > Hi, > > > > I am dealing with an app here that uses pg to handle a few thousand > concurrent web users. It seems that under heavy load, the INSERT and > UPDATE statements to one or two specific tables keep queuing up, to > the count of 150+ (one table has about 432K rows, other has about > 2.6Million rows), resulting in ?wait?s for other queries, and then This isn't an index issue, it's a locking issue. Sounds like you have a bunch of inserts and updates hitting the same rows over and over again. Eliminate that contention point, and you will have solved your problem. Free free to describe the processes involved, and we can help you do that. ---------------------------(end of broadcast)--------------------------- TIP 3: 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-hackers-owner@postgresql.org Tue Oct 26 09:51:48 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 19A443A3B18; Tue, 26 Oct 2004 09:51:47 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 68108-02; Tue, 26 Oct 2004 08:51:38 +0000 (GMT) Received: from cmailg1.svr.pol.co.uk (cmailg1.svr.pol.co.uk [195.92.195.171]) by svr1.postgresql.org (Postfix) with ESMTP id 2CAAE3A3B1B; Tue, 26 Oct 2004 09:51:38 +0100 (BST) Received: from modem-1194.llama.dialup.pol.co.uk ([217.135.180.170] helo=[192.168.0.102]) by cmailg1.svr.pol.co.uk with esmtp (Exim 4.14) id 1CMN3B-0005kz-31; Tue, 26 Oct 2004 09:51:29 +0100 Subject: Re: [PATCHES] ARC Memory Usage analysis From: Simon Riggs To: Tom Lane , Jan Wieck Cc: Kenneth Marshall , pgsql-patches@postgresql.org, pgsql-hackers@postgresql.org, josh@agliodbs.com In-Reply-To: <20651.1098720192@sss.pgh.pa.us> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> Content-Type: text/plain Organization: 2nd Quadrant Message-Id: <1098780555.6807.136.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Date: Tue, 26 Oct 2004 09:49:15 +0100 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/880 X-Sequence-Number: 60383 On Mon, 2004-10-25 at 16:34, Jan Wieck wrote: > The problem is, with a too small directory ARC cannot guesstimate what > might be in the kernel buffers. Nor can it guesstimate what recently was > in the kernel buffers and got pushed out from there. That results in a > way too small B1 list, and therefore we don't get B1 hits when in fact > the data was found in memory. B1 hits is what increases the T1target, > and since we are missing them with a too small directory size, our > implementation of ARC is propably using a T2 size larger than the > working set. That is not optimal. I think I have seen that the T1 list shrinks "too much", but need more tests...with some good test results The effectiveness of ARC relies upon the balance between the often conflicting requirements of "recency" and "frequency". It seems possible, even likely, that pgsql's version of ARC may need some subtle changes to rebalance it - if we are unlikely enough to find cases where it genuinely is out of balance. Many performance tests are required, together with a few ideas on extra parameters to include....hence my support of Jan's ideas. That's also why I called the B1+B2 hit ratio "turbulence" because it relates to how much oscillation is happening between T1 and T2. In physical systems, we expect the oscillations to be damped, but there is no guarantee that we have a nearly critically damped oscillator. (Note that the absence of turbulence doesn't imply that T1+T2 is optimally sized, just that is balanced). [...and all though the discussion has wandered away from my original patch...would anybody like to commit, or decline the patch?] > If we would replace the dynamic T1 buffers with a max_backends*2 area of > shared buffers, use a C value representing the effective cache size and > limit the T1target on the lower bound to effective cache size - shared > buffers, then we basically moved the T1 cache into the OS buffers. Limiting the minimum size of T1len to be 2* maxbackends sounds like an easy way to prevent overbalancing of T2, but I would like to follow up on ways to have T1 naturally stay larger. I'll do a patch with this idea in, for testing. I'll call this "T1 minimum size" so we can discuss it. Any other patches are welcome... It could be that B1 is too small and so we could use a larger value of C to keep track of more blocks. I think what is being suggested is two GUCs: shared_buffers (as is), plus another one, larger, which would allow us to track what is in shared_buffers and what is in OS cache. I have comments on "effective cache size" below.... On Mon, 2004-10-25 at 17:03, Tom Lane wrote: > Jan Wieck writes: > > This all only holds water, if the OS is allowed to swap out shared > > memory. And that was my initial question, how likely is it to find this > > to be true these days? > > I think it's more likely that not that the OS will consider shared > memory to be potentially swappable. On some platforms there is a shmctl > call you can make to lock your shmem in memory, but (a) we don't use it > and (b) it may well require privileges we haven't got anyway. Are you saying we shouldn't, or we don't yet? I simply assumed that we did use that function - surely it must be at least an option? RHEL supports this at least.... It may well be that we don't have those privileges, in which case we turn off the option. Often, we (or I?) will want to install a dedicated server, so we should have all the permissions we need, in which case... > This has always been one of the arguments against making shared_buffers > really large, of course --- if the buffers aren't all heavily used, and > the OS decides to swap them to disk, you are worse off than you would > have been with a smaller shared_buffers setting. Not really, just an argument against making them *too* large. Large *and* utilised is OK, so we need ways of judging optimal sizing. > However, I'm still really nervous about the idea of using > effective_cache_size to control the ARC algorithm. That number is > usually entirely bogus. Right now it is only a second-order influence > on certain planner estimates, and I am afraid to rely on it any more > heavily than that. ...ah yes, effective_cache_size. The manual describes effective_cache_size as if it had something to do with the OS, and some of this discussion has picked up on that. effective_cache_size is used in only two places in the code (both in the planner), as an estimate for calculating the cost of a) nonsequential access and b) index access, mainly as a way of avoiding overestimates of access costs for small tables. There is absolutely no implication in the code that effective_cache_size measures anything in the OS; what it gives is an estimate of the number of blocks that will be available from *somewhere* in memory (i.e. in shared_buffers OR OS cache) for one particular table (the one currently being considered by the planner). Crucially, the "size" referred to is the size of the *estimate*, not the size of the OS cache (nor the size of the OS cache + shared_buffers). So setting effective_cache_size = total memory available or setting effective_cache_size = total memory - shared_buffers are both wildly irrelevant things to do, or any assumption that directly links memory size to that parameter. So talking about "effective_cache_size" as if it were the OS cache isn't the right thing to do. ...It could be that we use a very high % of physical memory as shared_buffers - in which case the effective_cache_size would represent the contents of shared_buffers. Note also that the planner assumes that all tables are equally likely to be in cache. Increasing effective_cache_size in postgresql.conf seems destined to give the wrong answer in planning unless you absolutely understand what it does. I will submit a patch to correct the description in the manual. Further comments: The two estimates appear to use effective_cache_size differently: a) assumes that a table of size effective_cache_size will be 50% in cache b) assumes that effective_cache_size blocks are available, so for a table of size == effective_cache_size, then it will be 100% available IMHO the GUC should be renamed "estimated_cached_blocks", with the old name deprecated to force people to re-read the manual description of what effective_cache_size means and then set accordingly.....all of that in 8.0.... -- Best Regards, Simon Riggs From pgsql-hackers-owner@postgresql.org Tue Oct 26 10:53:25 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7BBAA3A3ADB for ; Tue, 26 Oct 2004 10:53:24 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 86970-04 for ; Tue, 26 Oct 2004 09:53:13 +0000 (GMT) Received: from cmailm4.svr.pol.co.uk (cmailm4.svr.pol.co.uk [195.92.193.211]) by svr1.postgresql.org (Postfix) with ESMTP id 9511D3A4350 for ; Tue, 26 Oct 2004 10:53:15 +0100 (BST) Received: from modem-757.monkey.dialup.pol.co.uk ([217.135.210.245] helo=[192.168.0.102]) by cmailm4.svr.pol.co.uk with esmtp (Exim 4.41) id 1CMO0u-0006Kz-JM; Tue, 26 Oct 2004 10:53:12 +0100 Subject: Re: [PATCHES] ARC Memory Usage analysis From: Simon Riggs To: Tom Lane Cc: Greg Stark , pgsql-hackers@postgresql.org In-Reply-To: <4887.1098770035@sss.pgh.pa.us> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <87mzyavhxc.fsf@stark.xeocode.com> <87hdoiv62b.fsf@stark.xeocode.com> <790.1098741205@sss.pgh.pa.us> <878y9uukgb.fsf@stark.xeocode.com> <4887.1098770035@sss.pgh.pa.us> Content-Type: text/plain Organization: 2nd Quadrant Message-Id: <1098784258.6807.155.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Date: Tue, 26 Oct 2004 10:50:58 +0100 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/881 X-Sequence-Number: 60384 On Tue, 2004-10-26 at 06:53, Tom Lane wrote: > Greg Stark writes: > > Tom Lane writes: > >> Another issue is what we do with the effective_cache_size value once we > >> have a number we trust. We can't readily change the size of the ARC > >> lists on the fly. > > > Huh? I thought effective_cache_size was just used as an factor the cost > > estimation equation. > > Today, that is true. Jan is speculating about using it as a parameter > of the ARC cache management algorithm ... and that worries me. > ISTM that we should be optimizing the use of shared_buffers, not whats outside. Didn't you (Tom) already say that? BTW, very good ideas on how to proceed, but why bother? For me, if the sysadmin didn't give shared_buffers to PostgreSQL, its because the memory is intended for use by something else and so not available at all. At least not dependably. The argument against large shared_buffers because of swapping applies to that assumption also...the OS cache is too volatile to attempt to gauge sensibly. There's an argument for improving performance for people that haven't set their parameters correctly, but thats got to be a secondary consideration anyhow. -- Best Regards, Simon Riggs From pgsql-performance-owner@postgresql.org Tue Oct 26 12:37:43 2004 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 87C823A3B8F for ; Tue, 26 Oct 2004 12:37:41 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 21196-03 for ; Tue, 26 Oct 2004 11:37:24 +0000 (GMT) Received: from obelix.askesis.nl (laudanum.demon.nl [82.161.125.16]) by svr1.postgresql.org (Postfix) with ESMTP id 363353A4398 for ; Tue, 26 Oct 2004 12:37:25 +0100 (BST) content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable X-MimeOLE: Produced By Microsoft Exchange V6.0.6487.1 Subject: Measuring server performance with psql and pgAdmin Date: Tue, 26 Oct 2004 13:37:24 +0200 Message-ID: X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: Measuring server performance with psql and pgAdmin Thread-Index: AcS7UCpCaV53rTi/QIeL+Pxdb9gpCQ== From: "Joost Kraaijeveld" To: "Pgsql-Performance (E-mail)" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/428 X-Sequence-Number: 8904 Hi all, I am (stilll) converting a database from a Clarion Topspeed database to Pos= tgresql 7.4.5 on Debian Linux 2.6.6-1. The program that uses the database u= ses a query like "select * from table" to show the user the contents of a t= able. This query cannot be changed (it is generated by Clarion and the pers= on in charge of the program cannot alter that behaviour). Now I have a big performance problem with reading a large table ( 96713 row= s). The query that is send to the database is "select * from table". "explain" and "explain analyze", using psql on cygwin: munt=3D# explain select * from klt_alg; QUERY PLAN=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20 -----------------------------------------------------------------=20 Seq Scan on klt_alg (cost=3D0.00..10675.13 rows=3D96713 width=3D729)=20 munt=3D# explain analyze select * from klt_alg; QUERY PLAN=20 ---------------------------------------------------------------------------= ---------------------------------------- Seq Scan on klt_alg (cost=3D0.00..10675.13 rows=3D96713 width=3D729) (actu= al time=3D13.172..2553.328 rows=3D96713 loops=3D1) Total runtime: 2889.109 ms (2 rows)=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20=20= =20=20=20=20=20 Running the query (with pgAdmin III): -- Executing query: select * from klt_alg; Total query runtime: 21926 ms. Data retrieval runtime: 72841 ms. 96713 rows retrieved. QUESTIONS: GENERAL: 1. The manual says about "explain analyze" : "The ANALYZE option causes the= statement to be actually executed, not only planned. The total elapsed tim= e expended within each plan node (in milliseconds) and total number of rows= it actually returned are added to the display." Does this time include dat= atransfer or just the time the database needs to collect the data, without = any data transfer? 2. If the time is without data transfer to the client, is there a reliable = way to measure the time needed to run the query and get the data (without t= he overhead of a program that does something with the data)? PGADMIN: 1. What does the "Total query runtime" really mean? (It was my understandin= g that it was the time the database needs to collect the data, without any = data transfer). 2. What does the "Data retrieval runtime" really mean? (Is this including t= he filling of the datagrid/GUI, or just the datatransfer?) TIA Groeten, Joost Kraaijeveld Askesis B.V. Molukkenstraat 14 6524NB Nijmegen tel: 024-3888063 / 06-51855277 fax: 024-3608416 e-mail: J.Kraaijeveld@Askesis.nl web: www.askesis.nl=20 From pgsql-hackers-owner@postgresql.org Tue Oct 26 13:21:01 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 25C463A43D4; Tue, 26 Oct 2004 13:20:58 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 34860-09; Tue, 26 Oct 2004 12:20:45 +0000 (GMT) Received: from cmailg2.svr.pol.co.uk (cmailg2.svr.pol.co.uk [195.92.195.172]) by svr1.postgresql.org (Postfix) with ESMTP id 761533A43BB; Tue, 26 Oct 2004 13:20:47 +0100 (BST) Received: from modem-2767.lynx.dialup.pol.co.uk ([217.135.202.207] helo=[192.168.0.102]) by cmailg2.svr.pol.co.uk with esmtp (Exim 4.14) id 1CMQJb-0003OT-Pz; Tue, 26 Oct 2004 13:20:40 +0100 Subject: Re: [PATCHES] ARC Memory Usage analysis From: Simon Riggs To: Tom Lane , Jan Wieck Cc: Kenneth Marshall , pgsql-patches@postgresql.org, pgsql-hackers@postgresql.org, josh@agliodbs.com In-Reply-To: <1098780555.6807.136.camel@localhost.localdomain> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <1098780555.6807.136.camel@localhost.localdomain> Content-Type: text/plain Organization: 2nd Quadrant Message-Id: <1098793105.6807.193.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Date: Tue, 26 Oct 2004 13:18:25 +0100 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/886 X-Sequence-Number: 60389 On Tue, 2004-10-26 at 09:49, Simon Riggs wrote: > On Mon, 2004-10-25 at 16:34, Jan Wieck wrote: > > The problem is, with a too small directory ARC cannot guesstimate what > > might be in the kernel buffers. Nor can it guesstimate what recently was > > in the kernel buffers and got pushed out from there. That results in a > > way too small B1 list, and therefore we don't get B1 hits when in fact > > the data was found in memory. B1 hits is what increases the T1target, > > and since we are missing them with a too small directory size, our > > implementation of ARC is propably using a T2 size larger than the > > working set. That is not optimal. > > I think I have seen that the T1 list shrinks "too much", but need more > tests...with some good test results > > > If we would replace the dynamic T1 buffers with a max_backends*2 area of > > shared buffers, use a C value representing the effective cache size and > > limit the T1target on the lower bound to effective cache size - shared > > buffers, then we basically moved the T1 cache into the OS buffers. > > Limiting the minimum size of T1len to be 2* maxbackends sounds like an > easy way to prevent overbalancing of T2, but I would like to follow up > on ways to have T1 naturally stay larger. I'll do a patch with this idea > in, for testing. I'll call this "T1 minimum size" so we can discuss it. > Don't know whether you've seen this latest update on the ARC idea: Sorav Bansal and Dharmendra S. Modha, CAR: Clock with Adaptive Replacement, in Proceedings of the USENIX Conference on File and Storage Technologies (FAST), pages 187--200, March 2004. [I picked up the .pdf here http://citeseer.ist.psu.edu/bansal04car.html] In that paper Bansal and Modha introduce an update to ARC called CART which they say is more appropriate for databases. Their idea is to introduce a "temporal locality window" as a way of making sure that blocks called twice within a short period don't fall out of T1, though don't make it into T2 either. Strangely enough the "temporal locality window" is made by increasing the size of T1... in an adpative way, of course. If we were going to put a limit on the minimum size of T1, then this would put a minimal "temporal locality window" in place....rather than the increased complexity they go to in order to make T1 larger. I note test results from both the ARC and CAR papers that show that T2 usually represents most of C, so the observations that T1 is very small is not atypical. That implies that the cost of managing the temporal locality window in CART is usually wasted, even though it does cut in as an overall benefit: The results show that CART is better than ARC over the whole range of cache sizes tested (16MB to 4GB) and workloads (apart from 1 out 22). If we were to implement a minimum size of T1, related as suggested to number of users, then this would provide a reasonable approximation of the temporal locality window. This wouldn't prevent the adaptation of T1 to be higher than this when required. Jan has already optimised ARC for PostgreSQL by the addition of a special lookup on transactionId required to optimise for the double cache lookup of select/update that occurs on a T1 hit. That seems likely to be able to be removed as a result of having a larger T1. I'd suggest limiting T1 to be a value of: shared_buffers <= 1000 T1limit = max_backends *0.75 shared_buffers <= 2000 T1limit = max_backends shared_buffers <= 5000 T1limit = max_backends *1.5 shared_buffers > 5000 T1limit = max_backends *2 I'll try some tests with both - minimum size of T1 - update optimisation removed Thoughts? -- Best Regards, Simon Riggs From pgsql-performance-owner@postgresql.org Tue Oct 26 16:24:27 2004 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 154393A3B5B for ; Tue, 26 Oct 2004 16:24:24 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 02649-10 for ; Tue, 26 Oct 2004 15:24:15 +0000 (GMT) Received: from wproxy.gmail.com (wproxy.gmail.com [64.233.184.207]) by svr1.postgresql.org (Postfix) with ESMTP id C6AD73A441E for ; Tue, 26 Oct 2004 16:24:10 +0100 (BST) Received: by wproxy.gmail.com with SMTP id 64so184259wri for ; Tue, 26 Oct 2004 08:24:08 -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:references; b=a+ihlMJSJkFnekI36iXo6EUGtNjvyfvcBGMrpqe/rQ97eDt4LoPQbrb/8NhrydOU7ezm/j7kl6GMeK4QlzdxKuiTnyVCjuyZkbAQmb3QKwPnHNG+kvGjMBFmrUsfLEsT/GXcjFv/1xK8Xwnyr3WB8/+O71ZbqjKuAgcskcS5o9Y= Received: by 10.38.77.65 with SMTP id z65mr511397rna; Tue, 26 Oct 2004 08:24:08 -0700 (PDT) Received: by 10.38.77.73 with HTTP; Tue, 26 Oct 2004 08:24:08 -0700 (PDT) Message-ID: <38242de904102608246d060f7d@mail.gmail.com> Date: Tue, 26 Oct 2004 09:24:08 -0600 From: Joshua Marsh Reply-To: Joshua Marsh To: pg@fastcrypt.com Subject: Re: Large Database Performance suggestions Cc: pgsql-performance@postgresql.org In-Reply-To: <4178F149.9090408@fastcrypt.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit References: <38242de904102120143c55ba34@mail.gmail.com> <4178F149.9090408@fastcrypt.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/429 X-Sequence-Number: 8905 Thanks for all of your help so far. Here is some of the information you guys were asking for: Test System: 2x AMD Opteron 244 (1.8Ghz) 8GB RAM 7x 72GB SCSI HDD (Raid 5) postrgesql.conf information: #--------------------------------------------------------------------------- # RESOURCE USAGE (except WAL) #--------------------------------------------------------------------------- # - Memory - shared_buffers = 1000 # min 16, at least max_connections*2, 8KB each #sort_mem = 1024 # min 64, size in KB #vacuum_mem = 8192 # min 1024, size in KB sort_mem = 4096000 vacuum_mem = 1024000 # - Free Space Map - #max_fsm_pages = 20000 # min max_fsm_relations*16, 6 bytes each #max_fsm_relations = 1000 # min 100, ~50 bytes each # - Kernel Resource Usage - #max_files_per_process = 1000 # min 25 #preload_libraries = '' #--------------------------------------------------------------------------- # WRITE AHEAD LOG #--------------------------------------------------------------------------- # - Settings - #fsync = true # turns forced synchronization on or off #wal_sync_method = fsync # the default varies across platforms: # fsync, fdatasync, open_sync, or open_datasync #wal_buffers = 8 # min 4, 8KB each # - Checkpoints - #checkpoint_segments = 3 # 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 Everything else are at their defaults. I actually think the WAL options are set to defaults as well, but I don't recall exactly :) As for the queries and table, The data we store is confidential, but it is essentially an account number with a bunch of boolean fields that specify if a person applies to criteria. So a query may look something like: SELECT acctno FROM view_of_data WHERE has_name AND is_active_member AND state = 'OH'; which is explained as something like this: QUERY PLAN ----------------------------------------------------------------- Seq Scan on view_of_data (cost=0.00..25304.26 rows=22054 width=11) Filter: (has_name AND is_active_member AND ((state)::text = 'OH'::text)) (2 rows) Occasionally, because we store data from several sources, we will have requests for data from several sources. We simply intersect the view_of_data table with a sources table that lists what acctno belong to what source. This query would look something like this: SELECT acctno FROM view_of_data WHERE has_name AND is_active_member AND state = 'OH' INTERSECT SELECT acctno FROM sources_data WHERE source = 175; which is explained as follows: QUERY PLAN ------------------------------------------------------------------------------------------- SetOp Intersect (cost=882226.14..885698.20 rows=69441 width=11) -> Sort (cost=882226.14..883962.17 rows=694411 width=11) Sort Key: acctno -> Append (cost=0.00..814849.42 rows=694411 width=11) -> Subquery Scan "*SELECT* 1" (cost=0.00..25524.80 rows=22054 width=11) -> Seq Scan on view_of_data (cost=0.00..25304.26 rows=22054 width=11) Filter: (has_name AND is_active_member AND ((state)::text = 'OH'::text)) -> Subquery Scan "*SELECT* 2" (cost=0.00..789324.62 rows=672357 width=11) -> Seq Scan on sources_data (cost=0.00..782601.05 rows=672357 width=11) Filter: (source = 23) Again, we see our biggest bottlenecks when we get over about 50 million records. The time to execute grows exponentially from that point. Thanks again for all of your help! -Josh On Fri, 22 Oct 2004 07:38:49 -0400, Dave Cramer wrote: > Josh, > > Your hardware setup would be useful too. It's surprising how slow some > big name servers really are. > If you are seriously considering memory sizes over 4G you may want to > look at an opteron. > > Dave > > > > Joshua Marsh wrote: > > >Hello everyone, > > > >I am currently working on a data project that uses PostgreSQL > >extensively to store, manage and maintain the data. We haven't had > >any problems regarding database size until recently. The three major > >tables we use never get bigger than 10 million records. With this > >size, we can do things like storing the indexes or even the tables in > >memory to allow faster access. > > > >Recently, we have found customers who are wanting to use our service > >with data files between 100 million and 300 million records. At that > >size, each of the three major tables will hold between 150 million and > >700 million records. At this size, I can't expect it to run queries > >in 10-15 seconds (what we can do with 10 million records), but would > >prefer to keep them all under a minute. > > > >We did some original testing and with a server with 8GB or RAM and > >found we can do operations on data file up to 50 million fairly well, > >but performance drop dramatically after that. Does anyone have any > >suggestions on a good way to improve performance for these extra large > >tables? Things that have come to mind are Replication and Beowulf > >clusters, but from what I have recently studied, these don't do so wel > >with singular processes. We will have parallel process running, but > >it's more important that the speed of each process be faster than > >several parallel processes at once. > > > >Any help would be greatly appreciated! > > > >Thanks, > > > >Joshua Marsh > > > >P.S. Off-topic, I have a few invitations to gmail. If anyone would > >like one, let me know. > > > >---------------------------(end of broadcast)--------------------------- > >TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org > > > > > > > > > > -- > Dave Cramer > http://www.postgresintl.com > 519 939 0336 > ICQ#14675561 > > From pgsql-hackers-owner@postgresql.org Tue Oct 26 16:30:32 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 4D1483A4448 for ; Tue, 26 Oct 2004 16:30:30 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 07638-03 for ; Tue, 26 Oct 2004 15:30:28 +0000 (GMT) Received: from stark.xeocode.com (gsstark.mtl.istop.com [66.11.160.162]) by svr1.postgresql.org (Postfix) with ESMTP id 57EEF3A442F for ; Tue, 26 Oct 2004 16:30:27 +0100 (BST) Received: from localhost ([127.0.0.1] helo=stark.xeocode.com) by stark.xeocode.com with smtp (Exim 3.36 #1 (Debian)) id 1CMTHE-0003uU-00; Tue, 26 Oct 2004 11:30:24 -0400 To: Curt Sampson Cc: Greg Stark , pgsql-hackers@postgresql.org Subject: Re: [PATCHES] ARC Memory Usage analysis References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <87mzyavhxc.fsf@stark.xeocode.com> <87hdoiv62b.fsf@stark.xeocode.com> <790.1098741205@sss.pgh.pa.us> <878y9uukgb.fsf@stark.xeocode.com> In-Reply-To: From: Greg Stark Organization: The Emacs Conspiracy; member since 1992 Date: 26 Oct 2004 11:30:23 -0400 Message-ID: <87r7nlts34.fsf@stark.xeocode.com> Lines: 39 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/892 X-Sequence-Number: 60395 Curt Sampson writes: > On Tue, 26 Oct 2004, Greg Stark wrote: > > > I see mmap or O_DIRECT being the only viable long-term stable states. My > > natural inclination was the former but after the latest thread on the subject > > I suspect it'll be forever out of reach. That makes O_DIRECT And a Postgres > > managed cache the only real choice. Having both caches is just a waste of > > memory and a waste of cpu cycles. > > I don't see why mmap is any more out of reach than O_DIRECT; it's not > all that much harder to implement, and mmap (and madvise!) is more > widely available. Because there's no way to prevent a write-out from occurring and no way to be notified by mmap before a write-out occurs, and Postgres wants to do its WAL logging at that time if it hasn't already happened. > But if using two caches is only costing us 1% in performance, there's > not really much point.... Well firstly it depends on the work profile. It can probably get much higher than we saw in that profile if your work load is causing more fresh buffers to be fetched. Secondly it also reduces the amount of cache available. If you have 256M of ram with about 200M free, and 40Mb of ram set aside for Postgres's buffer cache then you really only get 160Mb. It's costing you 20% of your cache, and reducing the cache hit rate accordingly. Thirdly the kernel doesn't know as much as Postgres about the load. Postgres could optimize its use of cache based on whether it knows the data is being loaded by a vacuum or sequential scan rather than an index lookup. In practice Postgres has gone with ARC which I suppose a kernel could implement anyways, but afaik neither linux nor BSD choose to do anything like it. -- greg From pgsql-performance-owner@postgresql.org Tue Oct 26 17:39:51 2004 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 862F13A4449 for ; Tue, 26 Oct 2004 17:39:50 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 35967-08 for ; Tue, 26 Oct 2004 16:39:46 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 4352C3A442F for ; Tue, 26 Oct 2004 17:39:47 +0100 (BST) 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 i9QGdiRV010370; Tue, 26 Oct 2004 12:39:44 -0400 (EDT) To: Joshua Marsh Cc: pg@fastcrypt.com, pgsql-performance@postgresql.org Subject: Re: Large Database Performance suggestions In-reply-to: <38242de904102608246d060f7d@mail.gmail.com> References: <38242de904102120143c55ba34@mail.gmail.com> <4178F149.9090408@fastcrypt.com> <38242de904102608246d060f7d@mail.gmail.com> Comments: In-reply-to Joshua Marsh message dated "Tue, 26 Oct 2004 09:24:08 -0600" Date: Tue, 26 Oct 2004 12:39:44 -0400 Message-ID: <10369.1098808784@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/430 X-Sequence-Number: 8906 Joshua Marsh writes: > shared_buffers = 1000 # min 16, at least max_connections*2, 8KB each This is on the small side for an 8G machine. I'd try 10000 or so. > sort_mem = 4096000 Yikes. You do realize you just said that *each sort operation* can use 4G? (Actually, it's probably overflowing internally; I dunno what amount of sort space you are really ending up with but it could be small.) Try something saner, maybe in the 10 to 100MB range. > vacuum_mem = 1024000 This is probably excessive as well. > #max_fsm_pages = 20000 # min max_fsm_relations*16, 6 bytes each > #max_fsm_relations = 1000 # min 100, ~50 bytes each You will need to bump these up a good deal to avoid database bloat. > Occasionally, because we store data from several sources, we will have > requests for data from several sources. We simply intersect the > view_of_data table with a sources table that lists what acctno belong > to what source. This query would look something like this: > SELECT acctno FROM view_of_data WHERE has_name AND is_active_member > AND state = 'OH' INTERSECT SELECT acctno FROM sources_data WHERE > source = 175; IMHO you need to rethink your table layout. There is simply no way that that query is going to be fast. Adding a source column to view_of_data would work much better. If you're not in a position to redo the tables, you might try it as a join: SELECT acctno FROM view_of_data JOIN sources_data USING (acctno) WHERE has_name AND is_active_member AND state = 'OH' AND source = 175; but I'm not really sure if that will be better or not. regards, tom lane From pgsql-performance-owner@postgresql.org Tue Oct 26 18:42:32 2004 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 953C13A3B12 for ; Tue, 26 Oct 2004 18:42:31 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 60110-02 for ; Tue, 26 Oct 2004 17:42:22 +0000 (GMT) Received: from vt-pe2550-001.VANTAGE.vantage.com (sol.vantage.com [64.80.203.242]) by svr1.postgresql.org (Postfix) with ESMTP id A4D023A3BA9 for ; Tue, 26 Oct 2004 18:42:22 +0100 (BST) 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-8859-7" Content-Transfer-Encoding: quoted-printable Subject: Re: can't handle large number of INSERT/UPDATEs Date: Tue, 26 Oct 2004 13:42:21 -0400 Message-ID: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DC@vt-pe2550-001.vantage.vantage.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] can't handle large number of INSERT/UPDATEs Thread-Index: AcS62GKSkO7EAKXVQ/SkOGeKYyw6rwAqfKFQ From: "Anjan Dave" To: "Rod Taylor" Cc: "Postgresql Performance" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/431 X-Sequence-Number: 8907 It probably is locking issue. I got a long list of locks held when we ran s= elect * from pg_locks during a peak time. relation | database | transaction | pid | mode | granted=20 ----------+----------+-------------+-------+------------------+--------- 17239 | 17142 | | 3856 | AccessShareLock | t | | 21196323 | 3875 | ExclusiveLock | t 16390 | 17142 | | 3911 | AccessShareLock | t 16595 | 17142 | | 3782 | AccessShareLock | t 17227 | 17142 | | 3840 | AccessShareLock | t 17227 | 17142 | | 3840 | RowExclusiveLock | t ... ... Vmstat would show a lot of disk IO at the same time. Is this pointing towards a disk IO issue? (to that end, other than a higher= CPU speed, and disabling HT, only thing changed is that it's RAID5 volume = now, instead of a RAID10) -anjan -----Original Message----- From: Rod Taylor [mailto:pg@rbt.ca]=20 Sent: Monday, October 25, 2004 5:19 PM To: Anjan Dave Cc: Postgresql Performance Subject: Re: [PERFORM] can't handle large number of INSERT/UPDATEs On Mon, 2004-10-25 at 16:53, Anjan Dave wrote: > Hi, >=20 >=20=20 >=20 > I am dealing with an app here that uses pg to handle a few thousand > concurrent web users. It seems that under heavy load, the INSERT and > UPDATE statements to one or two specific tables keep queuing up, to > the count of 150+ (one table has about 432K rows, other has about > 2.6Million rows), resulting in =A1wait=A2s for other queries, and then This isn't an index issue, it's a locking issue. Sounds like you have a bunch of inserts and updates hitting the same rows over and over again. Eliminate that contention point, and you will have solved your problem. Free free to describe the processes involved, and we can help you do that. From pgsql-performance-owner@postgresql.org Tue Oct 26 18:49:36 2004 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 B12D53A3B1F for ; Tue, 26 Oct 2004 18:49:35 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 62689-07 for ; Tue, 26 Oct 2004 17:49:25 +0000 (GMT) Received: from tht.net (vista.tht.net [216.126.88.2]) by svr1.postgresql.org (Postfix) with ESMTP id BDEE43A3B12 for ; Tue, 26 Oct 2004 18:49:25 +0100 (BST) Received: from [134.22.69.212] (dyn-69-212.tor.dsl.tht.net [134.22.69.212]) by tht.net (Postfix) with ESMTP id C9A5676A41; Tue, 26 Oct 2004 13:49:25 -0400 (EDT) Subject: Re: can't handle large number of INSERT/UPDATEs From: Rod Taylor To: Anjan Dave Cc: Postgresql Performance In-Reply-To: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DC@vt-pe2550-001.vantage.vantage.com> References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DC@vt-pe2550-001.vantage.vantage.com> Content-Type: text/plain Message-Id: <1098812922.8557.485.camel@home> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Tue, 26 Oct 2004 13:48:42 -0400 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/432 X-Sequence-Number: 8908 On Tue, 2004-10-26 at 13:42, Anjan Dave wrote: > It probably is locking issue. I got a long list of locks held when we ran select * from pg_locks during a peak time. > > relation | database | transaction | pid | mode | granted > ----------+----------+-------------+-------+------------------+--------- > 17239 | 17142 | | 3856 | AccessShareLock | t How many have granted = false? > Vmstat would show a lot of disk IO at the same time. > > Is this pointing towards a disk IO issue? Not necessarily. Is your IO reaching the limit or is it just heavy? From pgsql-performance-owner@postgresql.org Tue Oct 26 18:56:35 2004 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 B06E73A3B40 for ; Tue, 26 Oct 2004 18:56:33 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 63376-10 for ; Tue, 26 Oct 2004 17:56:25 +0000 (GMT) Received: from vt-pe2550-001.VANTAGE.vantage.com (sol.vantage.com [64.80.203.242]) by svr1.postgresql.org (Postfix) with ESMTP id 6B8733A3B2E for ; Tue, 26 Oct 2004 18:56:25 +0100 (BST) 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: can't handle large number of INSERT/UPDATEs Date: Tue, 26 Oct 2004 13:56:25 -0400 Message-ID: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DD@vt-pe2550-001.vantage.vantage.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] can't handle large number of INSERT/UPDATEs Thread-Index: AcS7hCF8Kgl2bmRDTBGmU2oCxry/JwAABxVQ From: "Anjan Dave" To: "Rod Taylor" Cc: "Postgresql Performance" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/433 X-Sequence-Number: 8909 None of the locks are in state false actually. I don't have iostat on that machine, but vmstat shows a lot of writes to the drives, and the runnable processes are more than 1: procs memory swap io system cpu r b swpd free buff cache si so bi bo in cs us sy wa id 1 2 0 3857568 292936 2791876 0 0 0 44460 1264 2997 23 13 22 41 2 2 0 3824668 292936 2791884 0 0 0 25262 1113 4797 28 12 29 31 2 3 0 3784772 292936 2791896 0 0 0 38988 1468 6677 28 12 48 12 2 4 0 3736256 292936 2791904 0 0 0 50970 1530 5217 19 12 49 20 4 2 0 3698056 292936 2791908 0 0 0 43576 1369 7316 20 15 35 30 2 1 0 3667124 292936 2791920 0 0 0 39174 1444 4659 25 16 35 24 6 1 0 3617652 292936 2791928 0 0 0 52430 1347 4681 25 19 20 37 1 3 0 3599992 292936 2790868 0 0 0 40156 1439 4394 20 14 29 37 6 0 0 3797488 292936 2568648 0 0 0 17706 2272 21534 28 23 19 30 0 0 0 3785396 292936 2568736 0 0 0 1156 1237 14057 33 8 0 59 0 0 0 3783568 292936 2568736 0 0 0 704 512 1537 5 2 1 92 1 0 0 3783188 292936 2568752 0 0 0 842 613 1919 6 1 1 92 -anjan -----Original Message----- From: Rod Taylor [mailto:pg@rbt.ca]=20 Sent: Tuesday, October 26, 2004 1:49 PM To: Anjan Dave Cc: Postgresql Performance Subject: RE: [PERFORM] can't handle large number of INSERT/UPDATEs On Tue, 2004-10-26 at 13:42, Anjan Dave wrote: > It probably is locking issue. I got a long list of locks held when we ran select * from pg_locks during a peak time. >=20 > relation | database | transaction | pid | mode | granted > ----------+----------+-------------+-------+------------------+--------- > 17239 | 17142 | | 3856 | AccessShareLock | t How many have granted =3D false? > Vmstat would show a lot of disk IO at the same time. >=20 > Is this pointing towards a disk IO issue? Not necessarily. Is your IO reaching the limit or is it just heavy? From pgsql-performance-owner@postgresql.org Tue Oct 26 19:25:06 2004 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 BEF903A3AC9 for ; Tue, 26 Oct 2004 19:25:04 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 71841-08 for ; Tue, 26 Oct 2004 18:24:55 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 350DD3A3D68 for ; Tue, 26 Oct 2004 19:24:53 +0100 (BST) 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 6558564; Tue, 26 Oct 2004 11:26:17 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: can't handle large number of INSERT/UPDATEs Date: Tue, 26 Oct 2004 11:27:02 -0700 User-Agent: KMail/1.6.2 Cc: "Anjan Dave" , "Rod Taylor" References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DC@vt-pe2550-001.vantage.vantage.com> In-Reply-To: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DC@vt-pe2550-001.vantage.vantage.com> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-7" Content-Transfer-Encoding: 7bit Message-Id: <200410261127.02656.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/434 X-Sequence-Number: 8910 Anjan, > It probably is locking issue. I got a long list of locks held when we ran > select * from pg_locks during a peak time. Do the back-loaded tables have FKs on them? This would be a likely cause of lock contention, and thus serializing inserts/updates to the tables. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Tue Oct 26 19:29:05 2004 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 B38333A448B for ; Tue, 26 Oct 2004 19:29:01 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 72958-09 for ; Tue, 26 Oct 2004 18:28:51 +0000 (GMT) Received: from bayswater1.ymogen.net (host-154-240-27-217.pobox.net.uk [217.27.240.154]) by svr1.postgresql.org (Postfix) with ESMTP id 962C03A3B12 for ; Tue, 26 Oct 2004 19:28:51 +0100 (BST) Received: from [82.68.132.233] (82-68-132-233.dsl.in-addr.zen.co.uk [82.68.132.233]) by bayswater1.ymogen.net (Postfix) with ESMTP id C4026A3637; Tue, 26 Oct 2004 19:28:49 +0100 (BST) Message-ID: <417E9752.5050506@ymogen.net> Date: Tue, 26 Oct 2004 19:28:34 +0100 From: Matt Clark User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Anjan Dave Cc: Rod Taylor , Postgresql Performance Subject: Re: can't handle large number of INSERT/UPDATEs References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DD@vt-pe2550-001.vantage.vantage.com> In-Reply-To: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DD@vt-pe2550-001.vantage.vantage.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/435 X-Sequence-Number: 8911 >I don't have iostat on that machine, but vmstat shows a lot of writes to >the drives, and the runnable processes are more than 1: > > 6 1 0 3617652 292936 2791928 0 0 0 52430 1347 4681 25 >19 20 37 > > Assuming that's the output of 'vmstat 1' and not some other delay, 50MB/second of sustained writes is usually considered 'a lot'. From pgsql-performance-owner@postgresql.org Tue Oct 26 20:18:15 2004 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 C09B23A3D09 for ; Tue, 26 Oct 2004 20:18:12 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 96533-03 for ; Tue, 26 Oct 2004 19:18:10 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 51CF43A3B5B for ; Tue, 26 Oct 2004 20:18:11 +0100 (BST) 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 6558784; Tue, 26 Oct 2004 12:19:36 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Measuring server performance with psql and pgAdmin Date: Tue, 26 Oct 2004 12:20:21 -0700 User-Agent: KMail/1.6.2 Cc: "Joost Kraaijeveld" References: In-Reply-To: MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410261220.21435.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/436 X-Sequence-Number: 8912 Joost, > 1. The manual says about "explain analyze" : "The ANALYZE option causes the > statement to be actually executed, not only planned. The total elapsed time > expended within each plan node (in milliseconds) and total number of rows > it actually returned are added to the display." Does this time include > datatransfer or just the time the database needs to collect the data, > without any data transfer? Correct. It's strictly backend time. > 2. If the time is without data transfer to the > client, is there a reliable way to measure the time needed to run the query > and get the data (without the overhead of a program that does something > with the data)? in PSQL, you can use \timing -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Tue Oct 26 21:19:15 2004 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 CFA023A4479 for ; Tue, 26 Oct 2004 21:19:11 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 20616-05 for ; Tue, 26 Oct 2004 20:19:07 +0000 (GMT) Received: from web50001.mail.yahoo.com (web50001.mail.yahoo.com [206.190.38.16]) by svr1.postgresql.org (Postfix) with SMTP id 560CE3A3BAA for ; Tue, 26 Oct 2004 21:19:07 +0100 (BST) Message-ID: <20041026201906.30736.qmail@web50001.mail.yahoo.com> Received: from [200.93.192.137] by web50001.mail.yahoo.com via HTTP; Tue, 26 Oct 2004 15:19:06 CDT Date: Tue, 26 Oct 2004 15:19:06 -0500 (CDT) From: Jaime Casanova Subject: Re: Sequential Scan with LIMIT To: pgsql-performance@postgresql.org In-Reply-To: <417D0EC2.5020206@johnmeinel.com> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/437 X-Sequence-Number: 8913 --- John Meinel escribi�: > Curt Sampson wrote: > > On Sun, 24 Oct 2004, John Meinel wrote: > > > > > >>I was looking into another problem, and I found > something that surprised > >>me. If I'm doing "SELECT * FROM mytable WHERE col > = 'myval' LIMIT 1.". > >>Now "col" is indexed... > >>The real purpose of this query is to check to see > if a value exists in > >>the column,... > > > > > > When you select all the columns, you're going to > force it to go to the > > table. If you select only the indexed column, it > ought to be able to use > > just the index, and never read the table at all. > You could also use more > > standard and more set-oriented SQL while you're at > it: > > > > SELECT DISTINCT(col) FROM mytable WHERE col = > 'myval' > > > > cjs > > Well, what you wrote was actually much slower, as it > had to scan the > whole table, grab all the rows, and then distinct > them in the end. > > However, this query worked: > > > SELECT DISTINCT(col) FROM mytable WHERE col = > 'myval' LIMIT 1; > > > Now, *why* that works differently from: > > SELECT col FROM mytable WHERE col = 'myval' LIMIT 1; > or > SELECT DISTINCT(col) FROM mytable WHERE col = > 'myval'; > > I'm not sure. They all return the same information. of course, both queries will return the same but that's just because you forced it. LIMIT and DISTINCT are different things so they behave and are plenned different. > > What's also weird is stuff like: > SELECT DISTINCT(NULL) FROM mytable WHERE col = > 'myval' LIMIT 1; why do you want to do such a thing? regards, Jaime Casanova _________________________________________________________ Do You Yahoo!? Informaci�n de Estados Unidos y Am�rica Latina, en Yahoo! Noticias. Vis�tanos en http://noticias.espanol.yahoo.com From pgsql-performance-owner@postgresql.org Tue Oct 26 21:51:00 2004 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 95ABB3A3C59 for ; Tue, 26 Oct 2004 21:50:58 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 36060-03 for ; Tue, 26 Oct 2004 20:50:55 +0000 (GMT) Received: from mail1.catalyst.net.nz (godel.catalyst.net.nz [202.49.159.12]) by svr1.postgresql.org (Postfix) with ESMTP id 98E143A3AFF for ; Tue, 26 Oct 2004 21:50:53 +0100 (BST) Received: from leibniz.catalyst.net.nz ([202.49.159.7] helo=lamb.mcmillan.net.nz) by mail1.catalyst.net.nz with asmtp (TLS-1.0:DHE_RSA_3DES_EDE_CBC_SHA:24) (Exim 4.34) id 1CMYHK-0000aP-VK; Wed, 27 Oct 2004 09:50:51 +1300 Received: from lamb.mcmillan.net.nz (lamb.mcmillan.net.nz [127.0.0.1]) by lamb.mcmillan.net.nz (Postfix) with ESMTP id 97FB7AC21B39; Wed, 27 Oct 2004 09:50:50 +1300 (NZDT) Subject: Re: can't handle large number of INSERT/UPDATEs From: Andrew McMillan To: Anjan Dave Cc: pgsql-performance@postgresql.org In-Reply-To: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985D5@vt-pe2550-001.vantage.vantage.com> References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985D5@vt-pe2550-001.vantage.vantage.com> Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="=-tIJWZPEUWnLHsT0/hc/i" Date: Wed, 27 Oct 2004 09:50:50 +1300 Message-Id: <1098823850.6440.89.camel@lamb.mcmillan.net.nz> Mime-Version: 1.0 X-Mailer: Evolution 2.0.2 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/438 X-Sequence-Number: 8914 --=-tIJWZPEUWnLHsT0/hc/i Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable On Mon, 2004-10-25 at 16:53 -0400, Anjan Dave wrote: > Hi, >=20 >=20=20 >=20 > I am dealing with an app here that uses pg to handle a few thousand > concurrent web users. It seems that under heavy load, the INSERT and > UPDATE statements to one or two specific tables keep queuing up, to > the count of 150+ (one table has about 432K rows, other has about > 2.6Million rows), resulting in =E2=80=98wait=E2=80=99s for other queries,= and then > everything piles up, with the load average shooting up to 10+.=20 Hi, We saw a similar problem here that was related to the locking that can happen against referred tables for referential integrity. In our case we had referred tables with very few rows (i.e. < 10) which caused the insert and update on the large tables to be effectively serialised due to the high contention on the referred tables. We changed our app to implement those referential integrity checks differently and performance was hugely boosted. Regards, Andrew. ------------------------------------------------------------------------- Andrew @ Catalyst .Net .NZ Ltd, PO Box 11-053, Manners St, Wellington WEB: http://catalyst.net.nz/ PHYS: Level 2, 150-154 Willis St DDI: +64(4)803-2201 MOB: +64(272)DEBIAN OFFICE: +64(4)499-2267 Chicken Little was right. ------------------------------------------------------------------------- --=-tIJWZPEUWnLHsT0/hc/i Content-Type: application/pgp-signature; name=signature.asc Content-Description: This is a digitally signed message part -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) iD8DBQBBfripjJA0f48GgBIRAqKAAJ0esyvcdqOaBIK+gYL+D7Ql6MOUDwCgvOgO 4Fuwb8gBnPHOHrLZGBFJlMU= =8UWV -----END PGP SIGNATURE----- --=-tIJWZPEUWnLHsT0/hc/i-- From pgsql-performance-owner@postgresql.org Tue Oct 26 22:13:15 2004 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 3CC503A44AB for ; Tue, 26 Oct 2004 22:13:12 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 44291-04 for ; Tue, 26 Oct 2004 21:13:03 +0000 (GMT) Received: from vt-pe2550-001.VANTAGE.vantage.com (sol.vantage.com [64.80.203.242]) by svr1.postgresql.org (Postfix) with ESMTP id 9722C3A4495 for ; Tue, 26 Oct 2004 22:13:04 +0100 (BST) 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: can't handle large number of INSERT/UPDATEs Date: Tue, 26 Oct 2004 17:13:04 -0400 Message-ID: <4BAFBB6B9CC46F41B2AD7D9F4BBAF78501B2751F@vt-pe2550-001.vantage.vantage.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] can't handle large number of INSERT/UPDATEs Thread-Index: AcS7nXxiYanz2kOaQBuW7IrC0zVnhQAAuFnw From: "Anjan Dave" To: "Andrew McMillan" Cc: X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/439 X-Sequence-Number: 8915 Andrew/Josh, Josh also suggested to check for any FK/referential integrity checks, but I am told that we don't have any foreign key constraints. Thanks, anjan -----Original Message----- From: Andrew McMillan [mailto:andrew@catalyst.net.nz]=20 Sent: Tuesday, October 26, 2004 4:51 PM To: Anjan Dave Cc: pgsql-performance@postgresql.org Subject: Re: [PERFORM] can't handle large number of INSERT/UPDATEs On Mon, 2004-10-25 at 16:53 -0400, Anjan Dave wrote: > Hi, >=20 >=20=20 >=20 > I am dealing with an app here that uses pg to handle a few thousand > concurrent web users. It seems that under heavy load, the INSERT and > UPDATE statements to one or two specific tables keep queuing up, to > the count of 150+ (one table has about 432K rows, other has about > 2.6Million rows), resulting in 'wait's for other queries, and then > everything piles up, with the load average shooting up to 10+.=20 Hi, We saw a similar problem here that was related to the locking that can happen against referred tables for referential integrity. In our case we had referred tables with very few rows (i.e. < 10) which caused the insert and update on the large tables to be effectively serialised due to the high contention on the referred tables. We changed our app to implement those referential integrity checks differently and performance was hugely boosted. Regards, Andrew. ------------------------------------------------------------------------ - Andrew @ Catalyst .Net .NZ Ltd, PO Box 11-053, Manners St, Wellington WEB: http://catalyst.net.nz/ PHYS: Level 2, 150-154 Willis St DDI: +64(4)803-2201 MOB: +64(272)DEBIAN OFFICE: +64(4)499-2267 Chicken Little was right. ------------------------------------------------------------------------ - From pgsql-performance-owner@postgresql.org Tue Oct 26 22:15:57 2004 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 DFCE03A3C59 for ; Tue, 26 Oct 2004 22:15:53 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 45725-04 for ; Tue, 26 Oct 2004 21:15:50 +0000 (GMT) Received: from vt-pe2550-001.VANTAGE.vantage.com (sol.vantage.com [64.80.203.242]) by svr1.postgresql.org (Postfix) with ESMTP id 811113A4491 for ; Tue, 26 Oct 2004 22:15:50 +0100 (BST) 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: can't handle large number of INSERT/UPDATEs Date: Tue, 26 Oct 2004 17:15:49 -0400 Message-ID: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DE@vt-pe2550-001.vantage.vantage.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] can't handle large number of INSERT/UPDATEs Thread-Index: AcS7iaSIf5PL3zKfTs2Cep9/+HyB1AAFU4Hg From: "Anjan Dave" To: "Matt Clark" Cc: "Rod Taylor" , "Postgresql Performance" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/440 X-Sequence-Number: 8916 That is 1 or maybe 2 second interval. One thing I am not sure is why 'bi' (disk writes) stays at 0 mostly, it's the 'bo' column that shows high numbers (reads from disk). With so many INSERT/UPDATEs, I would expect it the other way around... -anjan -----Original Message----- From: Matt Clark [mailto:matt@ymogen.net]=20 Sent: Tuesday, October 26, 2004 2:29 PM To: Anjan Dave Cc: Rod Taylor; Postgresql Performance Subject: Re: [PERFORM] can't handle large number of INSERT/UPDATEs >I don't have iostat on that machine, but vmstat shows a lot of writes to >the drives, and the runnable processes are more than 1: > > 6 1 0 3617652 292936 2791928 0 0 0 52430 1347 4681 25 >19 20 37 >=20=20 > Assuming that's the output of 'vmstat 1' and not some other delay,=20 50MB/second of sustained writes is usually considered 'a lot'.=20 From pgsql-performance-owner@postgresql.org Tue Oct 26 22:53:26 2004 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 A208C3A44A8 for ; Tue, 26 Oct 2004 22:53:23 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 59001-04 for ; Tue, 26 Oct 2004 21:53:12 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id C92BE3A44CE for ; Tue, 26 Oct 2004 22:53:14 +0100 (BST) 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 i9QLrE4h013824; Tue, 26 Oct 2004 17:53:14 -0400 (EDT) To: "Anjan Dave" Cc: "Rod Taylor" , "Postgresql Performance" Subject: Re: can't handle large number of INSERT/UPDATEs In-reply-to: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DD@vt-pe2550-001.vantage.vantage.com> References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DD@vt-pe2550-001.vantage.vantage.com> Comments: In-reply-to "Anjan Dave" message dated "Tue, 26 Oct 2004 13:56:25 -0400" Date: Tue, 26 Oct 2004 17:53:14 -0400 Message-ID: <13823.1098827594@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/441 X-Sequence-Number: 8917 "Anjan Dave" writes: > None of the locks are in state false actually. In that case you don't have a locking problem. > I don't have iostat on that machine, but vmstat shows a lot of writes to > the drives, and the runnable processes are more than 1: I get the impression that you are just saturating the write bandwidth of your disk :-( It's fairly likely that this happens during checkpoints. Look to see if the postmaster has a child that shows itself as a checkpointer in "ps" when the saturation is occurring. You might be able to improve matters by altering the checkpoint frequency parameters (though beware that either too small or too large will likely make matters even worse). regards, tom lane From pgsql-performance-owner@postgresql.org Tue Oct 26 23:14:36 2004 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 B90593A44CB for ; Tue, 26 Oct 2004 23:14:19 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 65906-07 for ; Tue, 26 Oct 2004 22:14:09 +0000 (GMT) Received: from vt-pe2550-001.VANTAGE.vantage.com (sol.vantage.com [64.80.203.242]) by svr1.postgresql.org (Postfix) with ESMTP id 568833A3B04 for ; Tue, 26 Oct 2004 23:14:10 +0100 (BST) 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: can't handle large number of INSERT/UPDATEs Date: Tue, 26 Oct 2004 18:14:10 -0400 Message-ID: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DF@vt-pe2550-001.vantage.vantage.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] can't handle large number of INSERT/UPDATEs Thread-Index: AcS7pjPajROyB3T2QBiskTifN/yDSQAAQ+9g From: "Anjan Dave" To: "Tom Lane" Cc: "Rod Taylor" , "Postgresql Performance" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/442 X-Sequence-Number: 8918 It just seems that the more activity there is (that is when there's a lot of disk activity) the checkpoints happen quicker too. Here's a snapshot from the /var/log/messages -=20 Oct 26 17:21:22 vl-pe6650-003 postgres[13978]: [2-1] LOG: recycled transaction log file "0000000B0000007E" Oct 26 17:21:22 vl-pe6650-003 postgres[13978]: [3-1] LOG: recycled transaction log file "0000000B0000007F" ... Oct 26 17:26:25 vl-pe6650-003 postgres[14273]: [2-1] LOG: recycled transaction log file "0000000B00000080" Oct 26 17:26:25 vl-pe6650-003 postgres[14273]: [3-1] LOG: recycled transaction log file "0000000B00000081" Oct 26 17:26:25 vl-pe6650-003 postgres[14273]: [4-1] LOG: recycled transaction log file "0000000B00000082" ... Oct 26 17:31:27 vl-pe6650-003 postgres[14508]: [2-1] LOG: recycled transaction log file "0000000B00000083" Oct 26 17:31:27 vl-pe6650-003 postgres[14508]: [3-1] LOG: recycled transaction log file "0000000B00000084" Oct 26 17:31:27 vl-pe6650-003 postgres[14508]: [4-1] LOG: recycled transaction log file "0000000B00000085" ... I have increased them from default 3 to 15. Haven't altered the frequency though.... Thanks, Anjan=20 -----Original Message----- From: Tom Lane [mailto:tgl@sss.pgh.pa.us]=20 Sent: Tuesday, October 26, 2004 5:53 PM To: Anjan Dave Cc: Rod Taylor; Postgresql Performance Subject: Re: [PERFORM] can't handle large number of INSERT/UPDATEs=20 "Anjan Dave" writes: > None of the locks are in state false actually. In that case you don't have a locking problem. > I don't have iostat on that machine, but vmstat shows a lot of writes to > the drives, and the runnable processes are more than 1: I get the impression that you are just saturating the write bandwidth of your disk :-( It's fairly likely that this happens during checkpoints. Look to see if the postmaster has a child that shows itself as a checkpointer in "ps" when the saturation is occurring. You might be able to improve matters by altering the checkpoint frequency parameters (though beware that either too small or too large will likely make matters even worse). regards, tom lane From pgsql-performance-owner@postgresql.org Tue Oct 26 23:21:50 2004 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 C66593A43E1 for ; Tue, 26 Oct 2004 23:21:45 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 67579-08 for ; Tue, 26 Oct 2004 22:21:41 +0000 (GMT) Received: from ct.radiology.uiowa.edu (ct.radiology.uiowa.edu [129.255.60.186]) by svr1.postgresql.org (Postfix) with ESMTP id DF2CE3A3B4B for ; Tue, 26 Oct 2004 23:21:42 +0100 (BST) Received: from [192.168.1.11] (12-217-241-0.client.mchsi.com [12.217.241.0]) by ct.radiology.uiowa.edu (8.11.6/8.11.6) with ESMTP id i9QMLe322290; Tue, 26 Oct 2004 17:21:40 -0500 Message-ID: <417ECDEF.8030108@johnmeinel.com> Date: Tue, 26 Oct 2004 17:21:35 -0500 From: John Meinel User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Jaime Casanova Cc: pgsql-performance@postgresql.org Subject: Re: Sequential Scan with LIMIT References: <20041026201906.30736.qmail@web50001.mail.yahoo.com> In-Reply-To: <20041026201906.30736.qmail@web50001.mail.yahoo.com> X-Enigmail-Version: 0.86.0.0 X-Enigmail-Supports: pgp-inline, pgp-mime Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enigBA86A89DE219113FA5129D77" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/443 X-Sequence-Number: 8919 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enigBA86A89DE219113FA5129D77 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Jaime Casanova wrote: [...] >> >>I'm not sure. They all return the same information. > > > of course, both queries will return the same but > that's just because you forced it. > > LIMIT and DISTINCT are different things so they behave > and are plenned different. > > > >>What's also weird is stuff like: >>SELECT DISTINCT(NULL) FROM mytable WHERE col = >>'myval' LIMIT 1; > > > why do you want to do such a thing? > > regards, > Jaime Casanova > I was trying to see if selecting a constant would change things. I could have done SELECT DISTINCT(1) or just SELECT 1 FROM ... The idea of the query is that if 'myval' exists in the table, return something different than if 'myval' does not exist. If you are writing a function, you can use: SELECT something... IF FOUND THEN do a ELSE do b END IF; The whole point of this exercise was just to find what the cheapest query is when you want to test for the existence of a value in a column. The only thing I've found for my column is: SET enable_seq_scan TO off; SELECT col FROM mytable WHERE col = 'myval' LIMIT 1; SET enable_seq_scan TO on; My column is not distributed well (larger numbers occur later in the dataset, but may occur many times.) In total there are something like 500,000 rows, the number 555647 occurs 100,000 times, but not until row 300,000 or so. The analyzer looks at the data and says "1/5th of the time it is 555647, so I can just do a sequential scan as the odds are I don't have to look for very long, then I don't have to load the index". It turns out this is very bad, where with an index you just have to do 2 page loads, instead of reading 300,000 rows. Obviously this isn't a general-case solution. But if you have a situation similar to mine, it might be useful. (That's one thing with DB tuning. It seems to be very situation dependent, and it's hard to plan without a real dataset.) John =:-> --------------enigBA86A89DE219113FA5129D77 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFBfs3vJdeBCYSNAAMRAij1AKCzucvN+dDXptIwmK5hsjX9cx2JPwCgsUe1 l5eetHEBY79tw2F3IcY7Qx8= =yeOQ -----END PGP SIGNATURE----- --------------enigBA86A89DE219113FA5129D77-- From pgsql-performance-owner@postgresql.org Tue Oct 26 23:37:47 2004 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 1F9EF3A44EF for ; Tue, 26 Oct 2004 23:37:43 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 75465-01 for ; Tue, 26 Oct 2004 22:37:32 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 229FC3A44D2 for ; Tue, 26 Oct 2004 23:37:34 +0100 (BST) 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 i9QMbX0d014216; Tue, 26 Oct 2004 18:37:33 -0400 (EDT) To: "Anjan Dave" Cc: "Matt Clark" , "Rod Taylor" , "Postgresql Performance" Subject: Re: can't handle large number of INSERT/UPDATEs In-reply-to: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DE@vt-pe2550-001.vantage.vantage.com> References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DE@vt-pe2550-001.vantage.vantage.com> Comments: In-reply-to "Anjan Dave" message dated "Tue, 26 Oct 2004 17:15:49 -0400" Date: Tue, 26 Oct 2004 18:37:32 -0400 Message-ID: <14215.1098830252@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/444 X-Sequence-Number: 8920 "Anjan Dave" writes: > One thing I am not sure is why 'bi' (disk writes) stays at 0 mostly, > it's the 'bo' column that shows high numbers (reads from disk). With so > many INSERT/UPDATEs, I would expect it the other way around... Er ... it *is* the other way around. bi is blocks in (to the CPU), bo is blocks out (from the CPU). regards, tom lane From pgsql-performance-owner@postgresql.org Tue Oct 26 23:46:14 2004 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 129DD3A44F9 for ; Tue, 26 Oct 2004 23:45:41 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 76678-04 for ; Tue, 26 Oct 2004 22:45:37 +0000 (GMT) Received: from vt-pe2550-001.VANTAGE.vantage.com (sol.vantage.com [64.80.203.242]) by svr1.postgresql.org (Postfix) with ESMTP id CEFC23A44E9 for ; Tue, 26 Oct 2004 23:45:37 +0100 (BST) 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: can't handle large number of INSERT/UPDATEs Date: Tue, 26 Oct 2004 18:45:37 -0400 Message-ID: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985E1@vt-pe2550-001.vantage.vantage.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] can't handle large number of INSERT/UPDATEs Thread-Index: AcS7rGXbY+5WynAORYeAsNy4i70odAAAP5XE From: "Anjan Dave" To: "Tom Lane" Cc: "Matt Clark" , "Rod Taylor" , "Postgresql Performance" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/445 X-Sequence-Number: 8921 T2ssIGkgd2FzIHRoaW5raW5nIGZyb20gdGhlIGRpc2sgcGVyc3BlY3RpdmUu IFRoYW5rcyENCg0KCS0tLS0tT3JpZ2luYWwgTWVzc2FnZS0tLS0tIA0KCUZy b206IFRvbSBMYW5lIFttYWlsdG86dGdsQHNzcy5wZ2gucGEudXNdIA0KCVNl bnQ6IFR1ZSAxMC8yNi8yMDA0IDY6MzcgUE0gDQoJVG86IEFuamFuIERhdmUg DQoJQ2M6IE1hdHQgQ2xhcms7IFJvZCBUYXlsb3I7IFBvc3RncmVzcWwgUGVy Zm9ybWFuY2UgDQoJU3ViamVjdDogUmU6IFtQRVJGT1JNXSBjYW4ndCBoYW5k bGUgbGFyZ2UgbnVtYmVyIG9mIElOU0VSVC9VUERBVEVzIA0KCQ0KCQ0KDQoJ IkFuamFuIERhdmUiIDxhZGF2ZUB2YW50YWdlLmNvbT4gd3JpdGVzOiANCgk+ IE9uZSB0aGluZyBJIGFtIG5vdCBzdXJlIGlzIHdoeSAnYmknIChkaXNrIHdy aXRlcykgc3RheXMgYXQgMCBtb3N0bHksIA0KCT4gaXQncyB0aGUgJ2JvJyBj b2x1bW4gdGhhdCBzaG93cyBoaWdoIG51bWJlcnMgKHJlYWRzIGZyb20gZGlz aykuIFdpdGggc28gDQoJPiBtYW55IElOU0VSVC9VUERBVEVzLCBJIHdvdWxk IGV4cGVjdCBpdCB0aGUgb3RoZXIgd2F5IGFyb3VuZC4uLiANCg0KCUVyIC4u LiBpdCAqaXMqIHRoZSBvdGhlciB3YXkgYXJvdW5kLiAgYmkgaXMgYmxvY2tz IGluICh0byB0aGUgQ1BVKSwgDQoJYm8gaXMgYmxvY2tzIG91dCAoZnJvbSB0 aGUgQ1BVKS4gDQoNCgkgICAgICAgICAgICAgICAgICAgICAgICByZWdhcmRz LCB0b20gbGFuZSANCg0K From pgsql-performance-owner@postgresql.org Wed Oct 27 00:00:35 2004 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 80E533A4507 for ; Wed, 27 Oct 2004 00:00:30 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 81736-02 for ; Tue, 26 Oct 2004 23:00:18 +0000 (GMT) Received: from mailman.norchemlab.com (unknown [12.164.217.51]) by svr1.postgresql.org (Postfix) with ESMTP id D465F3A43E1 for ; Wed, 27 Oct 2004 00:00:19 +0100 (BST) Received: from hyperion2.norchemlab.com (fireman.norchemlab.com [12.164.217.50]) by mailman.norchemlab.com (Postfix) with ESMTP id E840C8C8043; Tue, 26 Oct 2004 16:07:16 -0700 (MST) Received: from localhost (localhost.localdomain [127.0.0.1]) by hyperion2.norchemlab.com (8.12.8/8.12.8) with ESMTP id i9QN4ZS5021668; Tue, 26 Oct 2004 16:04:35 -0700 Date: Tue, 26 Oct 2004 16:04:35 -0700 (MST) From: Curtis Zinzilieta To: Tom Lane Cc: Anjan Dave , Matt Clark , Rod Taylor , Postgresql Performance Subject: Re: can't handle large number of INSERT/UPDATEs In-Reply-To: <14215.1098830252@sss.pgh.pa.us> Message-ID: 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/446 X-Sequence-Number: 8922 On Tue, 26 Oct 2004, Tom Lane wrote: > "Anjan Dave" writes: > > One thing I am not sure is why 'bi' (disk writes) stays at 0 mostly, > > it's the 'bo' column that shows high numbers (reads from disk). With so > > many INSERT/UPDATEs, I would expect it the other way around... > > Er ... it *is* the other way around. bi is blocks in (to the CPU), > bo is blocks out (from the CPU). > > regards, tom lane Ummm..... [curtisz@labsoft T2]$ man vmstat FIELD DESCRIPTIONS IO bi: Blocks sent to a block device (blocks/s). bo: Blocks received from a block device (blocks/s). And on my read-heavy 7.4.2 system (running on rh8 at the moment....) (truncated for readability...) [root@labsoft T2]# vmstat 1 procs memory swap io system r b w swpd free buff cache si so bi bo in cs us 0 0 0 127592 56832 365496 2013788 0 1 3 6 4 0 4 2 0 0 127592 56868 365496 2013788 0 0 0 0 363 611 1 1 0 0 127592 57444 365508 2013788 0 0 8 972 1556 3616 11 0 0 1 127592 57408 365512 2013800 0 0 0 448 614 1216 5 0 0 0 127592 56660 365512 2013800 0 0 0 0 666 1150 6 0 3 1 127592 56680 365512 2013816 0 0 16 180 1280 2050 2 0 0 0 127592 56864 365516 2013852 0 0 20 728 2111 4360 11 0 0 0 127592 57952 365544 2013824 0 0 0 552 1153 2002 10 0 0 0 127592 57276 365544 2013824 0 0 0 504 718 1111 5 1 0 0 127592 57244 365544 2013824 0 0 0 436 1495 2366 7 0 0 0 127592 57252 365544 2013824 0 0 0 0 618 1380 5 0 0 0 127592 57276 365556 2014192 0 0 360 1240 2418 5056 14 2 0 0 127592 56664 365564 2014176 0 0 0 156 658 1349 5 1 0 0 127592 55864 365568 2014184 0 0 0 1572 1388 3598 9 2 0 0 127592 56160 365572 2014184 0 0 0 536 4860 6621 13 Which seems appropriate for both the database and the man page.... -Curtis From pgsql-performance-owner@postgresql.org Wed Oct 27 01:09:45 2004 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 D04313A3AFE for ; Wed, 27 Oct 2004 01:09:28 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 96778-08 for ; Wed, 27 Oct 2004 00:09:23 +0000 (GMT) Received: from window.monsterlabs.com (window.monsterlabs.com [216.183.105.176]) by svr1.postgresql.org (Postfix) with SMTP id F22D73A1D9E for ; Wed, 27 Oct 2004 01:09:24 +0100 (BST) Received: (qmail 16708 invoked from network); 27 Oct 2004 00:09:24 -0000 Received: from pcp02680095pcs.nash01.tn.comcast.net (HELO ?192.168.0.114?) (68.52.121.245) by 0 with SMTP; 27 Oct 2004 00:09:24 -0000 In-Reply-To: <1098780555.6807.136.camel@localhost.localdomain> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <1098780555.6807.136.camel@localhost.localdomain> Mime-Version: 1.0 (Apple Message framework v619) Content-Type: text/plain; charset=US-ASCII; format=flowed Message-Id: <74778BDA-27AC-11D9-B369-000D93AE0944@sitening.com> Content-Transfer-Encoding: 7bit Cc: PgSQL - Performance From: Thomas F.O'Connell Subject: Re: [PATCHES] [HACKERS] ARC Memory Usage analysis Date: Tue, 26 Oct 2004 19:09:22 -0500 To: Simon Riggs X-Mailer: Apple Mail (2.619) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/447 X-Sequence-Number: 8923 Simon, As a postgres DBA, I find your comments about how not to use effective_cache_size instructive, but I'm still not sure how I should arrive at a target value for it. On most of the machines on which I admin postgres, I generally set shared_buffers to 10,000 (using what seems to have been the recent conventional wisdom of the lesser of 10,000 or 10% of RAM). I haven't really settled on an optimal value for effective_cache_size, and now I'm again confused as to how I might even benchmark it. Here are the documents on which I've based my knowledge: http://www.varlena.com/varlena/GeneralBits/Tidbits/perf.html#effcache http://www.varlena.com/varlena/GeneralBits/Tidbits/annotated_conf_e.html http://www.ca.postgresql.org/docs/momjian/hw_performance/node8.html From Bruce's document, I gather that effective_cache_size would assume that either shared buffers or unused RAM were valid sources of cached pages for the purposes of assessing plans. As a result, I was intending to inflate the value of effective_cache_size to closer to the amount of unused RAM on some of the machines I admin (once I've verified that they all have a unified buffer cache). Is that correct? -tfo -- Thomas F. O'Connell Co-Founder, Information Architect Sitening, LLC http://www.sitening.com/ 110 30th Avenue North, Suite 6 Nashville, TN 37203-6320 615-260-0005 On Oct 26, 2004, at 3:49 AM, Simon Riggs wrote: > On Mon, 2004-10-25 at 16:34, Jan Wieck wrote: >> The problem is, with a too small directory ARC cannot guesstimate what >> might be in the kernel buffers. Nor can it guesstimate what recently >> was >> in the kernel buffers and got pushed out from there. That results in a >> way too small B1 list, and therefore we don't get B1 hits when in fact >> the data was found in memory. B1 hits is what increases the T1target, >> and since we are missing them with a too small directory size, our >> implementation of ARC is propably using a T2 size larger than the >> working set. That is not optimal. > > I think I have seen that the T1 list shrinks "too much", but need more > tests...with some good test results > > The effectiveness of ARC relies upon the balance between the often > conflicting requirements of "recency" and "frequency". It seems > possible, even likely, that pgsql's version of ARC may need some subtle > changes to rebalance it - if we are unlikely enough to find cases where > it genuinely is out of balance. Many performance tests are required, > together with a few ideas on extra parameters to include....hence my > support of Jan's ideas. > > That's also why I called the B1+B2 hit ratio "turbulence" because it > relates to how much oscillation is happening between T1 and T2. In > physical systems, we expect the oscillations to be damped, but there is > no guarantee that we have a nearly critically damped oscillator. (Note > that the absence of turbulence doesn't imply that T1+T2 is optimally > sized, just that is balanced). > > [...and all though the discussion has wandered away from my original > patch...would anybody like to commit, or decline the patch?] > >> If we would replace the dynamic T1 buffers with a max_backends*2 area >> of >> shared buffers, use a C value representing the effective cache size >> and >> limit the T1target on the lower bound to effective cache size - shared >> buffers, then we basically moved the T1 cache into the OS buffers. > > Limiting the minimum size of T1len to be 2* maxbackends sounds like an > easy way to prevent overbalancing of T2, but I would like to follow up > on ways to have T1 naturally stay larger. I'll do a patch with this > idea > in, for testing. I'll call this "T1 minimum size" so we can discuss it. > > Any other patches are welcome... > > It could be that B1 is too small and so we could use a larger value of > C > to keep track of more blocks. I think what is being suggested is two > GUCs: shared_buffers (as is), plus another one, larger, which would > allow us to track what is in shared_buffers and what is in OS cache. > > I have comments on "effective cache size" below.... > > On Mon, 2004-10-25 at 17:03, Tom Lane wrote: >> Jan Wieck writes: >>> This all only holds water, if the OS is allowed to swap out shared >>> memory. And that was my initial question, how likely is it to find >>> this >>> to be true these days? >> >> I think it's more likely that not that the OS will consider shared >> memory to be potentially swappable. On some platforms there is a >> shmctl >> call you can make to lock your shmem in memory, but (a) we don't use >> it >> and (b) it may well require privileges we haven't got anyway. > > Are you saying we shouldn't, or we don't yet? I simply assumed that we > did use that function - surely it must be at least an option? RHEL > supports this at least.... > > It may well be that we don't have those privileges, in which case we > turn off the option. Often, we (or I?) will want to install a dedicated > server, so we should have all the permissions we need, in which case... > >> This has always been one of the arguments against making >> shared_buffers >> really large, of course --- if the buffers aren't all heavily used, >> and >> the OS decides to swap them to disk, you are worse off than you would >> have been with a smaller shared_buffers setting. > > Not really, just an argument against making them *too* large. Large > *and* utilised is OK, so we need ways of judging optimal sizing. > >> However, I'm still really nervous about the idea of using >> effective_cache_size to control the ARC algorithm. That number is >> usually entirely bogus. Right now it is only a second-order influence >> on certain planner estimates, and I am afraid to rely on it any more >> heavily than that. > > ...ah yes, effective_cache_size. > > The manual describes effective_cache_size as if it had something to do > with the OS, and some of this discussion has picked up on that. > > effective_cache_size is used in only two places in the code (both in > the > planner), as an estimate for calculating the cost of a) nonsequential > access and b) index access, mainly as a way of avoiding overestimates > of > access costs for small tables. > > There is absolutely no implication in the code that > effective_cache_size > measures anything in the OS; what it gives is an estimate of the number > of blocks that will be available from *somewhere* in memory (i.e. in > shared_buffers OR OS cache) for one particular table (the one currently > being considered by the planner). > > Crucially, the "size" referred to is the size of the *estimate*, not > the > size of the OS cache (nor the size of the OS cache + shared_buffers). > So > setting effective_cache_size = total memory available or setting > effective_cache_size = total memory - shared_buffers are both wildly > irrelevant things to do, or any assumption that directly links memory > size to that parameter. So talking about "effective_cache_size" as if > it > were the OS cache isn't the right thing to do. > > ...It could be that we use a very high % of physical memory as > shared_buffers - in which case the effective_cache_size would represent > the contents of shared_buffers. > > Note also that the planner assumes that all tables are equally likely > to > be in cache. Increasing effective_cache_size in postgresql.conf seems > destined to give the wrong answer in planning unless you absolutely > understand what it does. > > I will submit a patch to correct the description in the manual. > > Further comments: > The two estimates appear to use effective_cache_size differently: > a) assumes that a table of size effective_cache_size will be 50% in > cache > b) assumes that effective_cache_size blocks are available, so for a > table of size == effective_cache_size, then it will be 100% available > > IMHO the GUC should be renamed "estimated_cached_blocks", with the old > name deprecated to force people to re-read the manual description of > what effective_cache_size means and then set accordingly.....all of > that > in 8.0.... > > -- > Best Regards, Simon Riggs > > > ---------------------------(end of > broadcast)--------------------------- > TIP 2: you can get off all lists at once with the unregister command > (send "unregister YourEmailAddressHere" to > majordomo@postgresql.org) From pgsql-hackers-owner@postgresql.org Wed Oct 27 01:37:53 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 9AB353A4521; Wed, 27 Oct 2004 01:37:51 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 05354-10; Wed, 27 Oct 2004 00:37:47 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id EA2653A4523; Wed, 27 Oct 2004 01:37:49 +0100 (BST) 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 6560124; Tue, 26 Oct 2004 17:39:16 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-hackers@postgresql.org Subject: Re: [PERFORM] [PATCHES] ARC Memory Usage analysis Date: Tue, 26 Oct 2004 17:39:59 -0700 User-Agent: KMail/1.6.2 Cc: Thomas F.O'Connell , Simon Riggs , PgSQL - Performance References: <1098471059.20926.49.camel@localhost.localdomain> <1098780555.6807.136.camel@localhost.localdomain> <74778BDA-27AC-11D9-B369-000D93AE0944@sitening.com> In-Reply-To: <74778BDA-27AC-11D9-B369-000D93AE0944@sitening.com> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410261739.59814.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/917 X-Sequence-Number: 60420 Thomas, > As a result, I was intending to inflate the value of > effective_cache_size to closer to the amount of unused RAM on some of > the machines I admin (once I've verified that they all have a unified > buffer cache). Is that correct? Currently, yes. Right now, e_c_s is used just to inform the planner and make index vs. table scan and join order decisions. The problem which Simon is bringing up is part of a discussion about doing *more* with the information supplied by e_c_s. He points out that it's not really related to the *real* probability of any particular table being cached. At least, if I'm reading him right. -- --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Wed Oct 27 01:40:53 2004 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 60ED93A4514 for ; Wed, 27 Oct 2004 01:40:51 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 09554-03 for ; Wed, 27 Oct 2004 00:40:39 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id 6B8753A44FD for ; Wed, 27 Oct 2004 01:40:42 +0100 (BST) 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 6560143; Tue, 26 Oct 2004 17:42:09 -0700 From: Josh Berkus Reply-To: josh@agliodbs.com Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: can't handle large number of INSERT/UPDATEs Date: Tue, 26 Oct 2004 17:42:53 -0700 User-Agent: KMail/1.6.2 Cc: "Anjan Dave" , "Tom Lane" , "Rod Taylor" References: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DF@vt-pe2550-001.vantage.vantage.com> In-Reply-To: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985DF@vt-pe2550-001.vantage.vantage.com> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Message-Id: <200410261742.53345.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/449 X-Sequence-Number: 8925 Anjan, > Oct 26 17:26:25 vl-pe6650-003 postgres[14273]: [4-1] LOG: =A0recycled > transaction > log file "0000000B00000082" > ... > Oct 26 17:31:27 vl-pe6650-003 postgres[14508]: [2-1] LOG: =A0recycled > transaction > log file "0000000B00000083" > Oct 26 17:31:27 vl-pe6650-003 postgres[14508]: [3-1] LOG: =A0recycled > transaction > log file "0000000B00000084" > Oct 26 17:31:27 vl-pe6650-003 postgres[14508]: [4-1] LOG: =A0recycled > transaction > log file "0000000B00000085" Looks like you're running out of disk space for pending transactions. Can = you=20 afford more checkpoint_segments? Have you considered checkpoint_siblings? --=20 --Josh Josh Berkus Aglio Database Solutions San Francisco From pgsql-hackers-owner@postgresql.org Wed Oct 27 02:32:30 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 7AAF93A4534 for ; Wed, 27 Oct 2004 02:32:27 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 26646-01 for ; Wed, 27 Oct 2004 01:32:15 +0000 (GMT) Received: from platonic.cynic.net (platonic.cynic.net [204.80.150.245]) by svr1.postgresql.org (Postfix) with ESMTP id 73E753A4533 for ; Wed, 27 Oct 2004 02:32:17 +0100 (BST) Received: from angelic.cynic.net (angelic-platonic.cvpn.cynic.net [198.73.220.226]) by platonic.cynic.net (Postfix) with ESMTP id E7ABAC125; Wed, 27 Oct 2004 10:32:15 +0900 (JST) Received: from localhost (localhost [127.0.0.1]) by angelic.cynic.net (Postfix) with ESMTP id C7FAA8736; Wed, 27 Oct 2004 10:32:13 +0900 (JST) Date: Wed, 27 Oct 2004 10:32:13 +0900 (JST) From: Curt Sampson X-X-Sender: cjs@angelic-vtfw.cvpn.cynic.net To: Greg Stark Cc: pgsql-hackers@postgresql.org Subject: Re: [PATCHES] ARC Memory Usage analysis In-Reply-To: <87r7nlts34.fsf@stark.xeocode.com> Message-ID: References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <87mzyavhxc.fsf@stark.xeocode.com> <87hdoiv62b.fsf@stark.xeocode.com> <790.1098741205@sss.pgh.pa.us> <878y9uukgb.fsf@stark.xeocode.com> <87r7nlts34.fsf@stark.xeocode.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/919 X-Sequence-Number: 60422 On Wed, 26 Oct 2004, Greg Stark wrote: > > I don't see why mmap is any more out of reach than O_DIRECT; it's not > > all that much harder to implement, and mmap (and madvise!) is more > > widely available. > > Because there's no way to prevent a write-out from occurring and no way to be > notified by mmap before a write-out occurs, and Postgres wants to do its WAL > logging at that time if it hasn't already happened. I already described a solution to that problem in a post earlier in this thread (a write queue on the block). I may even have described it on this list a couple of years ago, that being about the time I thought it up. (The mmap idea just won't die, but at least I wasn't the one to bring it up this time. :-)) > Well firstly it depends on the work profile. It can probably get much higher > than we saw in that profile.... True, but 1% was is much, much lower than I'd expected. That tells me that my intuitive idea of the performance model is wrong, which means, for me at least, it's time to shut up or put up some benchmarks. > Secondly it also reduces the amount of cache available. If you have 256M of > ram with about 200M free, and 40Mb of ram set aside for Postgres's buffer > cache then you really only get 160Mb. It's costing you 20% of your cache, and > reducing the cache hit rate accordingly. Yeah, no question about that. > Thirdly the kernel doesn't know as much as Postgres about the load. Postgres > could optimize its use of cache based on whether it knows the data is being > loaded by a vacuum or sequential scan rather than an index lookup. In practice > Postgres has gone with ARC which I suppose a kernel could implement anyways, > but afaik neither linux nor BSD choose to do anything like it. madvise(). cjs -- Curt Sampson +81 90 7737 2974 http://www.NetBSD.org Make up enjoying your city life...produced by BIC CAMERA From pgsql-hackers-owner@postgresql.org Wed Oct 27 03:28:18 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D20BA3A3BF4 for ; Wed, 27 Oct 2004 03:28:16 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 41904-03 for ; Wed, 27 Oct 2004 02:28:11 +0000 (GMT) Received: from mpls-qmqp-01.inet.qwest.net (mpls-qmqp-01.inet.qwest.net [63.231.195.112]) by svr1.postgresql.org (Postfix) with SMTP id 498FD3A3BE5 for ; Wed, 27 Oct 2004 03:28:15 +0100 (BST) Received: (qmail 83001 invoked by uid 0); 27 Oct 2004 02:28:16 -0000 Received: from unknown (63.231.195.9) by mpls-qmqp-01.inet.qwest.net with QMQP; 27 Oct 2004 02:28:16 -0000 Received: from 63-227-127-37.dnvr.qwest.net (HELO ?10.0.0.2?) (63.227.127.37) by mpls-pop-09.inet.qwest.net with SMTP; 27 Oct 2004 02:28:15 -0000 Date: Tue, 26 Oct 2004 20:30:34 -0600 Message-Id: <1098844234.31930.45.camel@localhost.localdomain> From: "Scott Marlowe" To: "Tom Lane" Cc: "Greg Stark" , pgsql-hackers@postgresql.org Subject: Re: [PATCHES] ARC Memory Usage analysis In-Reply-To: <4887.1098770035@sss.pgh.pa.us> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <87mzyavhxc.fsf@stark.xeocode.com> <87hdoiv62b.fsf@stark.xeocode.com> <790.1098741205@sss.pgh.pa.us> <878y9uukgb.fsf@stark.xeocode.com> <4887.1098770035@sss.pgh.pa.us> Content-Type: text/plain Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/921 X-Sequence-Number: 60424 On Mon, 2004-10-25 at 23:53, Tom Lane wrote: > Greg Stark writes: > > Tom Lane writes: > >> Another issue is what we do with the effective_cache_size value once we > >> have a number we trust. We can't readily change the size of the ARC > >> lists on the fly. > > > Huh? I thought effective_cache_size was just used as an factor the cost > > estimation equation. > > Today, that is true. Jan is speculating about using it as a parameter > of the ARC cache management algorithm ... and that worries me. Because it's so often set wrong I take it. But if it's set right, and it makes the the database faster to pay attention to it, then I'd be in favor of it. Or at least having a switch to turn on the ARC buffer's ability to look at it. Or is it some other issue, having to do with the idea of knowing effective cache size cause a positive effect overall on the ARC algorhythm? From pgsql-performance-owner@postgresql.org Wed Oct 27 04:13:33 2004 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 AE2843A450B for ; Wed, 27 Oct 2004 04:12:30 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 54129-09 for ; Wed, 27 Oct 2004 03:12:23 +0000 (GMT) Received: from vt-pe2550-001.VANTAGE.vantage.com (sol.vantage.com [64.80.203.242]) by svr1.postgresql.org (Postfix) with ESMTP id E4F533A3BA3 for ; Wed, 27 Oct 2004 04:12:21 +0100 (BST) 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: can't handle large number of INSERT/UPDATEs Date: Tue, 26 Oct 2004 23:12:20 -0400 Message-ID: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985E5@vt-pe2550-001.vantage.vantage.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] can't handle large number of INSERT/UPDATEs Thread-Index: AcS7vZgKuirAFby8QOm+KwPcNK4glgACydNC From: "Anjan Dave" To: , Cc: "Tom Lane" , "Rod Taylor" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/450 X-Sequence-Number: 8926 Sm9zaCwNCiANCkkgaGF2ZSBpbmNyZWFzZWQgdGhlbSB0byAzMCwgd2lsbCBz ZWUgaWYgdGhhdCBoZWxwcy4gU3BhY2UgaXMgbm90IGEgY29uY2Vybi4gc2xp Z2h0bHkgbG9uZ2VyIHJlY292ZXJ5IHRpbWUgY291bGQgYmUgZmluZSB0b28u IFdvbmRlciB3aGF0IHBlb3BsZSB1c2UgKGV4YW1wbGVzKSBmb3IgdGhpcyB2 YWx1ZSBmb3IgaGlnaCB2b2x1bWUgZGF0YWJhc2VzIChleGNlcHQgZm9yIGR1 bXAvcmVzdG9yZSkuLi4/DQogDQpJIGRvbid0IGtub3cgd2hhdCBpcyBjaGVj a3BvaW50X3NpYmxpbmcuIEknbGwgcmVhZCBhYm91dCBpdCBpZiB0aGVyZSdz IHNvbWUgaW5mbyBvbiBpdCBzb21ld2hlcmUuDQogDQpUaGFua3MsDQpBbmph bg0KIA0KLS0tLS1PcmlnaW5hbCBNZXNzYWdlLS0tLS0gDQpGcm9tOiBKb3No IEJlcmt1cyBbbWFpbHRvOmpvc2hAYWdsaW9kYnMuY29tXSANClNlbnQ6IFR1 ZSAxMC8yNi8yMDA0IDg6NDIgUE0gDQpUbzogcGdzcWwtcGVyZm9ybWFuY2VA cG9zdGdyZXNxbC5vcmcgDQpDYzogQW5qYW4gRGF2ZTsgVG9tIExhbmU7IFJv ZCBUYXlsb3IgDQpTdWJqZWN0OiBSZTogW1BFUkZPUk1dIGNhbid0IGhhbmRs ZSBsYXJnZSBudW1iZXIgb2YgSU5TRVJUL1VQREFURXMNCg0KDQoNCglBbmph biwgDQoNCgk+IE9jdCAyNiAxNzoyNjoyNSB2bC1wZTY2NTAtMDAzIHBvc3Rn cmVzWzE0MjczXTogWzQtMV0gTE9HOiAgcmVjeWNsZWQgDQoJPiB0cmFuc2Fj dGlvbiANCgk+IGxvZyBmaWxlICIwMDAwMDAwQjAwMDAwMDgyIiANCgk+IC4u LiANCgk+IE9jdCAyNiAxNzozMToyNyB2bC1wZTY2NTAtMDAzIHBvc3RncmVz WzE0NTA4XTogWzItMV0gTE9HOiAgcmVjeWNsZWQgDQoJPiB0cmFuc2FjdGlv biANCgk+IGxvZyBmaWxlICIwMDAwMDAwQjAwMDAwMDgzIiANCgk+IE9jdCAy NiAxNzozMToyNyB2bC1wZTY2NTAtMDAzIHBvc3RncmVzWzE0NTA4XTogWzMt MV0gTE9HOiAgcmVjeWNsZWQgDQoJPiB0cmFuc2FjdGlvbiANCgk+IGxvZyBm aWxlICIwMDAwMDAwQjAwMDAwMDg0IiANCgk+IE9jdCAyNiAxNzozMToyNyB2 bC1wZTY2NTAtMDAzIHBvc3RncmVzWzE0NTA4XTogWzQtMV0gTE9HOiAgcmVj eWNsZWQgDQoJPiB0cmFuc2FjdGlvbiANCgk+IGxvZyBmaWxlICIwMDAwMDAw QjAwMDAwMDg1IiANCg0KCUxvb2tzIGxpa2UgeW91J3JlIHJ1bm5pbmcgb3V0 IG9mIGRpc2sgc3BhY2UgZm9yIHBlbmRpbmcgdHJhbnNhY3Rpb25zLiAgQ2Fu IHlvdSANCglhZmZvcmQgbW9yZSBjaGVja3BvaW50X3NlZ21lbnRzPyAgIEhh dmUgeW91IGNvbnNpZGVyZWQgY2hlY2twb2ludF9zaWJsaW5ncz8gDQoNCgkt LSANCgktLUpvc2ggDQoNCglKb3NoIEJlcmt1cyANCglBZ2xpbyBEYXRhYmFz ZSBTb2x1dGlvbnMgDQoJU2FuIEZyYW5jaXNjbyANCg0K From pgsql-performance-owner@postgresql.org Wed Oct 27 04:21:56 2004 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 A14BD3A3D49 for ; Wed, 27 Oct 2004 04:21:38 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 55937-10 for ; Wed, 27 Oct 2004 03:21:33 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 570293A3BA3 for ; Wed, 27 Oct 2004 04:21:32 +0100 (BST) 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 i9R3LVs4015813; Tue, 26 Oct 2004 23:21:31 -0400 (EDT) To: Curtis Zinzilieta Cc: Anjan Dave , Matt Clark , Rod Taylor , Postgresql Performance Subject: Re: can't handle large number of INSERT/UPDATEs In-reply-to: References: Comments: In-reply-to Curtis Zinzilieta message dated "Tue, 26 Oct 2004 16:04:35 -0700" Date: Tue, 26 Oct 2004 23:21:31 -0400 Message-ID: <15812.1098847291@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/451 X-Sequence-Number: 8927 Curtis Zinzilieta writes: > On Tue, 26 Oct 2004, Tom Lane wrote: >> Er ... it *is* the other way around. bi is blocks in (to the CPU), >> bo is blocks out (from the CPU). > Ummm..... > [curtisz@labsoft T2]$ man vmstat > bi: Blocks sent to a block device (blocks/s). > bo: Blocks received from a block device (blocks/s). You might want to have a word with your OS vendor. My vmstat man page says IO bi: Blocks received from a block device (blocks/s). bo: Blocks sent to a block device (blocks/s). and certainly anyone who's been around a computer more than a week or two knows which direction "in" and "out" are customarily seen from. regards, tom lane From pgsql-performance-owner@postgresql.org Wed Oct 27 05:18:35 2004 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 40D683A458E for ; Wed, 27 Oct 2004 05:18:31 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 74517-07 for ; Wed, 27 Oct 2004 04:18:25 +0000 (GMT) Received: from ct.radiology.uiowa.edu (ct.radiology.uiowa.edu [129.255.60.186]) by svr1.postgresql.org (Postfix) with ESMTP id CA89A3A4583 for ; Wed, 27 Oct 2004 05:18:24 +0100 (BST) Received: from [192.168.1.11] (12-217-241-0.client.mchsi.com [12.217.241.0]) by ct.radiology.uiowa.edu (8.11.6/8.11.6) with ESMTP id i9R4IF324293; Tue, 26 Oct 2004 23:18:15 -0500 Message-ID: <417F2181.4000801@johnmeinel.com> Date: Tue, 26 Oct 2004 23:18:09 -0500 From: John Meinel User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: Curtis Zinzilieta , Anjan Dave , Matt Clark , Rod Taylor , Postgresql Performance Subject: Re: can't handle large number of INSERT/UPDATEs References: <15812.1098847291@sss.pgh.pa.us> In-Reply-To: <15812.1098847291@sss.pgh.pa.us> X-Enigmail-Version: 0.86.0.0 X-Enigmail-Supports: pgp-inline, pgp-mime Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig8A306CDFFF6456D3531A6069" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/452 X-Sequence-Number: 8928 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig8A306CDFFF6456D3531A6069 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Tom Lane wrote: > Curtis Zinzilieta writes: > >>On Tue, 26 Oct 2004, Tom Lane wrote: >> >>>Er ... it *is* the other way around. bi is blocks in (to the CPU), >>>bo is blocks out (from the CPU). > > >>Ummm..... >>[curtisz@labsoft T2]$ man vmstat >> bi: Blocks sent to a block device (blocks/s). >> bo: Blocks received from a block device (blocks/s). > > > You might want to have a word with your OS vendor. My vmstat > man page says > > IO > bi: Blocks received from a block device (blocks/s). > bo: Blocks sent to a block device (blocks/s). > > and certainly anyone who's been around a computer more than a week or > two knows which direction "in" and "out" are customarily seen from. > > regards, tom lane > Interesting. I checked this on several machines. They actually say different things. Redhat 9- bi: Blocks sent to a block device (blocks/s). Latest Cygwin- bi: Blocks sent to a block device (blocks/s). Redhat 7.x- bi: Blocks sent to a block device (blocks/s). Redhat AS3- bi: blocks sent out to a block device (in blocks/s) I would say that I probably agree, things should be relative to the cpu. However, it doesn't seem to be something that was universally agreed upon. Or maybe the man-pages were all wrong, and only got updated recently. John =:-> --------------enig8A306CDFFF6456D3531A6069 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFBfyGBJdeBCYSNAAMRAvI1AJ4i1AzY5iwlsQXVpUy6oUw1+jd/kACguWgx 67UY50j8yjXcHoSo6vdXGrE= =VNsc -----END PGP SIGNATURE----- --------------enig8A306CDFFF6456D3531A6069-- From pgsql-performance-owner@postgresql.org Wed Oct 27 06:03:17 2004 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 46AB43A4593 for ; Wed, 27 Oct 2004 06:03:12 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 84634-09 for ; Wed, 27 Oct 2004 05:03:02 +0000 (GMT) Received: from millenium.mst.co.jp (unknown [210.230.185.241]) by svr1.postgresql.org (Postfix) with ESMTP id 21BC33A458D for ; Wed, 27 Oct 2004 06:03:01 +0100 (BST) Received: from mst1x5r347kymb (lc12114 [192.168.1.114]) by millenium.mst.co.jp (8.11.6p2/3.7W) with SMTP id i9R52Pq30474; Wed, 27 Oct 2004 14:02:25 +0900 Message-ID: <003a01c4bbe2$36ffe700$7201a8c0@mst1x5r347kymb> From: "Iain" To: "Curtis Zinzilieta" , "Tom Lane" Cc: "Anjan Dave" , "Matt Clark" , "Rod Taylor" , "Postgresql Performance" References: <15812.1098847291@sss.pgh.pa.us> Subject: Re: can't handle large number of INSERT/UPDATEs Date: Wed, 27 Oct 2004 14:02:48 +0900 MIME-Version: 1.0 Content-Type: text/plain; format=flowed; charset="iso-2022-jp"; 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/453 X-Sequence-Number: 8929 Turbo linux 7 sems to be agreeing with Curtis, bi: $B%V%m%C%/%G%P%$%9$KAw$i$l$?%V%m%C%/(B (blocks/s)$B!#(B bo: $B%V%m%C%/%G%P%$%9$+$i To: "Curtis Zinzilieta" Cc: "Anjan Dave" ; "Matt Clark" ; "Rod Taylor" ; "Postgresql Performance" Sent: Wednesday, October 27, 2004 12:21 PM Subject: Re: [PERFORM] can't handle large number of INSERT/UPDATEs > Curtis Zinzilieta writes: >> On Tue, 26 Oct 2004, Tom Lane wrote: >>> Er ... it *is* the other way around. bi is blocks in (to the CPU), >>> bo is blocks out (from the CPU). > >> Ummm..... >> [curtisz@labsoft T2]$ man vmstat >> bi: Blocks sent to a block device (blocks/s). >> bo: Blocks received from a block device (blocks/s). > > You might want to have a word with your OS vendor. My vmstat > man page says > > IO > bi: Blocks received from a block device (blocks/s). > bo: Blocks sent to a block device (blocks/s). > > and certainly anyone who's been around a computer more than a week or > two knows which direction "in" and "out" are customarily seen from. > > regards, tom lane > > ---------------------------(end of broadcast)--------------------------- > TIP 4: Don't 'kill -9' the postmaster From pgsql-performance-owner@postgresql.org Wed Oct 27 07:09:58 2004 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 994193A3CED for ; Wed, 27 Oct 2004 07:09:49 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 02209-09 for ; Wed, 27 Oct 2004 06:09:39 +0000 (GMT) Received: from bayswater1.ymogen.net (host-154-240-27-217.pobox.net.uk [217.27.240.154]) by svr1.postgresql.org (Postfix) with ESMTP id 921283A3AE6 for ; Wed, 27 Oct 2004 07:09:39 +0100 (BST) Received: from [82.68.132.233] (82-68-132-233.dsl.in-addr.zen.co.uk [82.68.132.233]) by bayswater1.ymogen.net (Postfix) with ESMTP id CE6C5A387F; Wed, 27 Oct 2004 07:09:37 +0100 (BST) Message-ID: <417F3B91.9080002@ymogen.net> Date: Wed, 27 Oct 2004 07:09:21 +0100 From: Matt Clark User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: John Meinel Cc: Tom Lane , Curtis Zinzilieta , Anjan Dave , Rod Taylor , Postgresql Performance Subject: Re: can't handle large number of INSERT/UPDATEs References: <15812.1098847291@sss.pgh.pa.us> <417F2181.4000801@johnmeinel.com> In-Reply-To: <417F2181.4000801@johnmeinel.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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/454 X-Sequence-Number: 8930 >> and certainly anyone who's been around a computer more than a week or >> two knows which direction "in" and "out" are customarily seen from. >> >> regards, tom lane >> > Apparently not whoever wrote the man page that everyone copied ;-) > Interesting. I checked this on several machines. They actually say > different things. > > Redhat 9- bi: Blocks sent to a block device (blocks/s). > Latest Cygwin- bi: Blocks sent to a block device (blocks/s). > Redhat 7.x- bi: Blocks sent to a block device (blocks/s). > Redhat AS3- bi: blocks sent out to a block device (in blocks/s) > > I would say that I probably agree, things should be relative to the > cpu. However, it doesn't seem to be something that was universally > agreed upon. Or maybe the man-pages were all wrong, and only got > updated recently. > Looks like the man pages are wrong, for RH7.3 at least. It says bi is 'blocks written', but an actual test like 'dd if=/dev/zero of=/tmp/test bs=1024 count=16384' on an otherwise nearly idle RH7.3 box gives: procs memory swap io system cpu r b w swpd free buff cache si so bi bo in cs us sy id 0 0 0 75936 474704 230452 953580 0 0 0 0 106 2527 0 0 99 0 0 0 75936 474704 230452 953580 0 0 0 16512 376 2572 0 2 98 0 0 0 75936 474704 230452 953580 0 0 0 0 105 2537 0 0 100 Which is in line with bo being 'blocks written'. M From pgsql-performance-owner@postgresql.org Sun Oct 31 05:42:58 2004 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 45CC93A44E5 for ; Wed, 27 Oct 2004 02:08:26 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 16559-06 for ; Wed, 27 Oct 2004 01:08:12 +0000 (GMT) Received: from ciar.org (45-97-229-216.sr2.pon.net [216.229.97.45]) by svr1.postgresql.org (Postfix) with SMTP id 201B13A4534 for ; Wed, 27 Oct 2004 02:08:14 +0100 (BST) Received: (qmail 8829 invoked from network); 27 Oct 2004 07:57:49 -0000 Received: from localhost (HELO hardpoint.hardpointi.com) (1004@127.0.0.1) by localhost with SMTP; 27 Oct 2004 07:57:49 -0000 Received: (from ttk2@localhost) by hardpoint.hardpointi.com (8.12.10/8.12.10/Submit) id i9R7vmKE008825 for pgsql-performance@postgresql.org; Wed, 27 Oct 2004 00:57:48 -0700 Date: Wed, 27 Oct 2004 00:57:48 -0700 From: TTK Ciar To: pgsql-performance@postgresql.org Subject: psql large RSS (1.6GB) Message-ID: <20041027075748.GA8355@hardpoint.ciar.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.4.1i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/473 X-Sequence-Number: 8949 Hello! My name is TTK, and I'm a software engineer at the Internet Archive's Data Repository department. We have recently started using postgresql for a couple of projects (we have historically been a MySQL outfit), and today my co-worker noticed psql eating memory like mad when invoked with a simple select statement incorporating a join of two tables. The platform is a heavily modified RedHat 7.3 Linux. We are using version 7.4.5 of postgresql. The invocation was via sh script: #!/bin/bash outfile=$1 if [ -z "$outfile" ]; then outfile=/0/brad/all_arcs.txt fi /usr/lib/postgresql/bin/psql -c 'select ServerDisks.servername,ServerDisks.diskserial,ServerDisks.diskmountpoint,DiskFiles.name,DiskFiles.md5 from DiskFiles,ServerDisks where DiskFiles.diskserial=ServerDisks.diskserial;' -F ' ' -A -t -o $outfile .. and the tables in question are somewhat large (hundreds of GB's of data), though we didn't expect that to be an issue as far as the psql process was concerned. We monitored server load via 'top -i -d 0.5' and watched the output file for data. Over the course of about 200 seconds, psql's RSS climbed to about 1.6 GB, and stayed there, while no data was written to the output file. Eventually 10133194 lines were written to the output file, all at once, about 1.2GB's worth of data. I re-ran the select query using psql in interactive mode, and saw the same results. I re-ran it again, using "explain analyse", and this time psql's RSS did *not* increase significantly. The result is here, if it helps: brad=# explain analyse select ServerDisks.servername,ServerDisks.diskserial,ServerDisks.diskmountpoint,DiskFiles.name,DiskFiles.md5 from DiskFiles,ServerDisks where DiskFiles.diskserial=ServerDisks.diskserial; QUERY PLAN ------------------------------------------------------------------ Hash Join (cost=22.50..65.00 rows=1000 width=274) (actual time=118.584..124653.729 rows=10133349 loops=1) Hash Cond: (("outer".diskserial)::text = ("inner".diskserial)::text) -> Seq Scan on diskfiles (cost=0.00..20.00 rows=1000 width=198) (actual time=7.201..31336.063 rows=10133349 loops=1) -> Hash (cost=20.00..20.00 rows=1000 width=158) (actual time=90.821..90.821 rows=0 loops=1) -> Seq Scan on serverdisks (cost=0.00..20.00 rows=1000 width=158) (actual time=9.985..87.364 rows=2280 loops=1) Total runtime: 130944.586 ms At a guess, it looks like the data set is being buffered in its entirety by psql, before any data is written to the output file, which is surprising. I would have expected it to grab data as it appeared on the socket from postmaster and write it to disk. Is there something we can do to stop psql from buffering results? Does anyone know what's going on here? If the solution is to just write a little client that uses perl DBI to fetch rows one at a time and write them out, that's doable, but it would be nice if psql could be made to "just work" without the monster RSS. I'd appreciate any feedback. If you need any additional info, please let me know and I will provide it. -- TTK ttk2@ciar.org ttk@archive.org From pgsql-performance-owner@postgresql.org Wed Oct 27 12:24:03 2004 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 1085F3A3B3E for ; Wed, 27 Oct 2004 12:23:58 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 01371-04 for ; Wed, 27 Oct 2004 11:23:50 +0000 (GMT) Received: from druid.net (druid.net [216.126.72.98]) by svr1.postgresql.org (Postfix) with ESMTP id 2AB1A3A3B2D for ; Wed, 27 Oct 2004 12:23:52 +0100 (BST) Received: from imp.druid.net (imp [216.126.72.111]) by druid.net (Postfix) with SMTP id D8C041A93; Wed, 27 Oct 2004 07:23:54 -0400 (EDT) Date: Wed, 27 Oct 2004 07:23:46 -0400 From: "D'Arcy J.M. Cain" To: Aaron Mulder Cc: pgsql-performance@postgresql.org Subject: Re: Free PostgreSQL Training, Philadelphia, Oct 30 Message-Id: <20041027072346.6e9cd42a.darcy@druid.net> In-Reply-To: References: X-Mailer: Sylpheed version 0.9.12 (GTK+ 1.2.10; i386--netbsdelf) 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/455 X-Sequence-Number: 8931 On Wed, 13 Oct 2004 12:21:27 -0400 (EDT) Aaron Mulder wrote: > All, > My company (Chariot Solutions) is sponsoring a day of free > PostgreSQL training by Bruce Momjian (one of the core PostgreSQL > developers). The day is split into 2 sessions (plus a Q&A session): Is there anyone else from the Toronto area going down that would like to share the driving? I am planning to drive down Friday morning and drive back Sunday. I'm not looking for expense sharing. I just don't want to drive for eight hours straight. -- D'Arcy J.M. Cain | Democracy is three wolves http://www.druid.net/darcy/ | and a sheep voting on +1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner. From pgsql-performance-owner@postgresql.org Wed Oct 27 18:16:29 2004 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 2D72A3A46C2 for ; Wed, 27 Oct 2004 18:16:25 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 40394-07 for ; Wed, 27 Oct 2004 17:16:22 +0000 (GMT) Received: from web50006.mail.yahoo.com (web50006.mail.yahoo.com [206.190.38.21]) by svr1.postgresql.org (Postfix) with SMTP id 832023A4696 for ; Wed, 27 Oct 2004 18:16:21 +0100 (BST) Message-ID: <20041027171621.60824.qmail@web50006.mail.yahoo.com> Received: from [69.65.137.210] by web50006.mail.yahoo.com via HTTP; Wed, 27 Oct 2004 12:16:21 CDT Date: Wed, 27 Oct 2004 12:16:21 -0500 (CDT) From: Jaime Casanova Subject: Re: Sequential Scan with LIMIT To: pgsql-performance@postgresql.org In-Reply-To: <417ECDEF.8030108@johnmeinel.com> 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/456 X-Sequence-Number: 8932 --- John Meinel escribi�: > Jaime Casanova wrote: > [...] > >> > >>I'm not sure. They all return the same > information. > > > > > > of course, both queries will return the same but > > that's just because you forced it. > > > > LIMIT and DISTINCT are different things so they > behave > > and are plenned different. > > > > > > > >>What's also weird is stuff like: > >>SELECT DISTINCT(NULL) FROM mytable WHERE col = > >>'myval' LIMIT 1; > > > > > > why do you want to do such a thing? > > > > regards, > > Jaime Casanova > > > > I was trying to see if selecting a constant would > change things. > I could have done SELECT DISTINCT(1) or just SELECT > 1 FROM ... > The idea of the query is that if 'myval' exists in > the table, return > something different than if 'myval' does not exist. > If you are writing a > function, you can use: > > SELECT something... > IF FOUND THEN > do a > ELSE > do b > END IF; > > The whole point of this exercise was just to find > what the cheapest > query is when you want to test for the existence of > a value in a column. > The only thing I've found for my column is: > > SET enable_seq_scan TO off; > SELECT col FROM mytable WHERE col = 'myval' LIMIT 1; > SET enable_seq_scan TO on; > > My column is not distributed well (larger numbers > occur later in the > dataset, but may occur many times.) In total there > are something like > 500,000 rows, the number 555647 occurs 100,000 > times, but not until row > 300,000 or so. > > The analyzer looks at the data and says "1/5th of > the time it is 555647, > so I can just do a sequential scan as the odds are I > don't have to look > for very long, then I don't have to load the index". > It turns out this > is very bad, where with an index you just have to do > 2 page loads, > instead of reading 300,000 rows. > > Obviously this isn't a general-case solution. But if > you have a > situation similar to mine, it might be useful. > > (That's one thing with DB tuning. It seems to be > very situation > dependent, and it's hard to plan without a real > dataset.) > > John > =:-> > In http://www.postgresql.org/docs/faqs/FAQ.html under "4.8) My queries are slow or don't make use of the indexes. Why?" says: "However, LIMIT combined with ORDER BY often will use an index because only a small portion of the table is returned. In fact, though MAX() and MIN() don't use indexes, it is possible to retrieve such values using an index with ORDER BY and LIMIT: SELECT col FROM tab ORDER BY col [ DESC ] LIMIT 1;" So, maybe you can try your query as SELECT col FROM mytable WHERE col = 'myval' ORDER BY col LIMIT 1; regards, Jaime Casanova _________________________________________________________ Do You Yahoo!? Informaci�n de Estados Unidos y Am�rica Latina, en Yahoo! Noticias. Vis�tanos en http://noticias.espanol.yahoo.com From pgsql-hackers-owner@postgresql.org Wed Oct 27 22:35:06 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D80153A4760; Wed, 27 Oct 2004 22:35:03 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 14404-03; Wed, 27 Oct 2004 21:34:52 +0000 (GMT) Received: from mail.osdl.org (fw.osdl.org [65.172.181.6]) by svr1.postgresql.org (Postfix) with ESMTP id 4345F3A3C1E; Wed, 27 Oct 2004 22:34:53 +0100 (BST) Received: (from markw@localhost) by mail.osdl.org (8.11.6/8.11.6) id i9RLYMM10488; Wed, 27 Oct 2004 14:34:22 -0700 Date: Wed, 27 Oct 2004 14:34:22 -0700 From: Mark Wong To: Jan Wieck Cc: Kenneth Marshall , Simon Riggs , pgsql-patches@postgresql.org, pgsql-hackers@postgresql.org, josh@agliodbs.com Subject: Re: ARC Memory Usage analysis Message-ID: <20041027143422.A6199@osdl.org> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5i In-Reply-To: <417D1D01.3090401@Yahoo.com>; from JanWieck@Yahoo.com on Mon, Oct 25, 2004 at 11:34:25AM -0400 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/955 X-Sequence-Number: 60458 On Mon, Oct 25, 2004 at 11:34:25AM -0400, Jan Wieck wrote: > On 10/22/2004 4:09 PM, Kenneth Marshall wrote: > > > On Fri, Oct 22, 2004 at 03:35:49PM -0400, Jan Wieck wrote: > >> On 10/22/2004 2:50 PM, Simon Riggs wrote: > >> > >> >I've been using the ARC debug options to analyse memory usage on the > >> >PostgreSQL 8.0 server. This is a precursor to more complex performance > >> >analysis work on the OSDL test suite. > >> > > >> >I've simplified some of the ARC reporting into a single log line, which > >> >is enclosed here as a patch on freelist.c. This includes reporting of: > >> >- the total memory in use, which wasn't previously reported > >> >- the cache hit ratio, which was slightly incorrectly calculated > >> >- a useful-ish value for looking at the "B" lists in ARC > >> >(This is a patch against cvstip, but I'm not sure whether this has > >> >potential for inclusion in 8.0...) > >> > > >> >The total memory in use is useful because it allows you to tell whether > >> >shared_buffers is set too high. If it is set too high, then memory usage > >> >will continue to grow slowly up to the max, without any corresponding > >> >increase in cache hit ratio. If shared_buffers is too small, then memory > >> >usage will climb quickly and linearly to its maximum. > >> > > >> >The last one I've called "turbulence" in an attempt to ascribe some > >> >useful meaning to B1/B2 hits - I've tried a few other measures though > >> >without much success. Turbulence is the hit ratio of B1+B2 lists added > >> >together. By observation, this is zero when ARC gives smooth operation, > >> >and goes above zero otherwise. Typically, turbulence occurs when > >> >shared_buffers is too small for the working set of the database/workload > >> >combination and ARC repeatedly re-balances the lengths of T1/T2 as a > >> >result of "near-misses" on the B1/B2 lists. Turbulence doesn't usually > >> >cut in until the cache is fully utilized, so there is usually some delay > >> >after startup. > >> > > >> >We also recently discussed that I would add some further memory analysis > >> >features for 8.1, so I've been trying to figure out how. > >> > > >> >The idea that B1, B2 represent something really useful doesn't seem to > >> >have been borne out - though I'm open to persuasion there. > >> > > >> >I originally envisaged a "shadow list" operating in extension of the > >> >main ARC list. This will require some re-coding, since the variables and > >> >macros are all hard-coded to a single set of lists. No complaints, just > >> >it will take a little longer than we all thought (for me, that is...) > >> > > >> >My proposal is to alter the code to allow an array of memory linked > >> >lists. The actual list would be [0] - other additional lists would be > >> >created dynamically as required i.e. not using IFDEFs, since I want this > >> >to be controlled by a SIGHUP GUC to allow on-site tuning, not just lab > >> >work. This will then allow reporting against the additional lists, so > >> >that cache hit ratios can be seen with various other "prototype" > >> >shared_buffer settings. > >> > >> All the existing lists live in shared memory, so that dynamic approach > >> suffers from the fact that the memory has to be allocated during ipc_init. > >> > >> What do you think about my other theory to make C actually 2x effective > >> cache size and NOT to keep T1 in shared buffers but to assume T1 lives > >> in the OS buffer cache? > >> > >> > >> Jan > >> > > Jan, > > > >>From the articles that I have seen on the ARC algorithm, I do not think > > that using the effective cache size to set C would be a win. The design > > of the ARC process is to allow the cache to optimize its use in response > > to the actual workload. It may be the best use of the cache in some cases > > to have the entire cache allocated to T1 and similarly for T2. If fact, > > the ability to alter the behavior as needed is one of the key advantages. > > Only the "working set" of the database, that is the pages that are very > frequently used, are worth holding in shared memory at all. The rest > should be copied in and out of the OS disc buffers. > > The problem is, with a too small directory ARC cannot guesstimate what > might be in the kernel buffers. Nor can it guesstimate what recently was > in the kernel buffers and got pushed out from there. That results in a > way too small B1 list, and therefore we don't get B1 hits when in fact > the data was found in memory. B1 hits is what increases the T1target, > and since we are missing them with a too small directory size, our > implementation of ARC is propably using a T2 size larger than the > working set. That is not optimal. > > If we would replace the dynamic T1 buffers with a max_backends*2 area of > shared buffers, use a C value representing the effective cache size and > limit the T1target on the lower bound to effective cache size - shared > buffers, then we basically moved the T1 cache into the OS buffers. > > This all only holds water, if the OS is allowed to swap out shared > memory. And that was my initial question, how likely is it to find this > to be true these days? > > > Jan > I've asked our linux kernel guys some quick questions and they say you can lock mmapped memory and sys v shared memory with mlock and SHM_LOCK, resp. Otherwise the OS will swap out memory as it sees fit, whether or not it's shared. Mark From pgsql-performance-owner@postgresql.org Wed Oct 27 22:41:05 2004 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 128573A474B for ; Wed, 27 Oct 2004 22:40:57 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 14404-07 for ; Wed, 27 Oct 2004 21:40:48 +0000 (GMT) Received: from mail.osdl.org (fw.osdl.org [65.172.181.6]) by svr1.postgresql.org (Postfix) with ESMTP id DA4D13A4758 for ; Wed, 27 Oct 2004 22:40:49 +0100 (BST) Received: (from markw@localhost) by mail.osdl.org (8.11.6/8.11.6) id i9RLem511952; Wed, 27 Oct 2004 14:40:48 -0700 Date: Wed, 27 Oct 2004 14:40:48 -0700 From: Mark Wong To: Josh Berkus Cc: pgsql-performance@postgresql.org, Bjorn Bength Subject: Re: different io elevators in linux Message-ID: <20041027144048.A11433@osdl.org> References: <1098546845.7754.8.camel@localhost> <200410251009.17575.josh@agliodbs.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5i In-Reply-To: <200410251009.17575.josh@agliodbs.com>; from josh@agliodbs.com on Mon, Oct 25, 2004 at 10:09:17AM -0700 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/457 X-Sequence-Number: 8933 On Mon, Oct 25, 2004 at 10:09:17AM -0700, Josh Berkus wrote: > Bjorn, > > > I haven't read much FAQs but has anyone done some benchmarks with > > different io schedulers in linux with postgresql? > > According to OSDL, using the "deadline" scheduler sometimes results in a > roughly 5% boost to performance, and sometimes none, depending on the > application. We use it for all testing, though, just in case. > > --Josh > Yes, we found with an OLTP type workload, the as scheduler performs about 5% worse than the deadline scheduler, where in a DSS type workload there really isn't much difference. The former doing a mix of reading/writing, where the latter is doing mostly reading. Mark From pgsql-hackers-owner@postgresql.org Thu Oct 28 04:19:05 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id CC5AB3A2BBB for ; Thu, 28 Oct 2004 04:18:48 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 07867-03 for ; Thu, 28 Oct 2004 03:18:39 +0000 (GMT) Received: from rwcrmhc12.comcast.net (rwcrmhc12.comcast.net [216.148.227.85]) by svr1.postgresql.org (Postfix) with ESMTP id 6EF163A1D8B for ; Thu, 28 Oct 2004 04:18:38 +0100 (BST) Received: from sysexperts.com (c-24-6-183-218.client.comcast.net[24.6.183.218]) by comcast.net (rwcrmhc12) with ESMTP id <2004102803183601400085mie>; Thu, 28 Oct 2004 03:18:36 +0000 Received: from localhost (localhost [127.0.0.1]) (uid 1000) by filer with local; Wed, 27 Oct 2004 20:20:45 -0700 id 000055D4.4180658D.00006A59 Date: Wed, 27 Oct 2004 20:20:45 -0700 From: Kevin Brown To: pgsql-hackers@postgresql.org Subject: Re: [PATCHES] ARC Memory Usage analysis Message-ID: <20041028032044.GA17583@filer> Mail-Followup-To: Kevin Brown , pgsql-hackers@postgresql.org References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <87mzyavhxc.fsf@stark.xeocode.com> <87hdoiv62b.fsf@stark.xeocode.com> <790.1098741205@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline In-Reply-To: <790.1098741205@sss.pgh.pa.us> Organization: Frobozzco International User-Agent: Mutt/1.5.6+20040907i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/960 X-Sequence-Number: 60463 Tom Lane wrote: > Greg Stark writes: > > So I would suggest using something like 100us as the threshold for > > determining whether a buffer fetch came from cache. > > I see no reason to hardwire such a number. On any hardware, the > distribution is going to be double-humped, and it will be pretty easy to > determine a cutoff after minimal accumulation of data. The real question > is whether we can afford a pair of gettimeofday() calls per read(). > This isn't a big issue if the read actually results in I/O, but if it > doesn't, the percentage overhead could be significant. > > If we assume that the effective_cache_size value isn't changing very > fast, maybe it would be good enough to instrument only every N'th read > (I'm imagining N on the order of 100) for this purpose. Or maybe we > need only instrument reads that are of blocks that are close to where > the ARC algorithm thinks the cache edge is. If it's decided to instrument reads, then perhaps an even better use of it would be to tune random_page_cost. If the storage manager knows the difference between a sequential scan and a random scan, then it should easily be able to measure the actual performance it gets for each and calculate random_page_cost based on the results. While the ARC lists can't be tuned on the fly, random_page_cost can. > One small problem is that the time measurement gives you only a lower > bound on the time the read() actually took. In a heavily loaded system > you might not get the CPU back for long enough to fool you about whether > the block came from cache or not. True, but that's information that you'd want to factor into the performance measurements anyway. The database needs to know how much wall clock time it takes for it to fetch a page under various circumstances from disk via the OS. For determining whether or not the read() hit the disk instead of just OS cache, what would matter is the average difference between the two. That's admittedly a problem if the difference is less than the noise, though, but at the same time that would imply that given the circumstances it really doesn't matter whether or not the page was fetched from disk: the difference is small enough that you could consider them equivalent. You don't need 100% accuracy for this stuff, just statistically significant accuracy. > Another issue is what we do with the effective_cache_size value once > we have a number we trust. We can't readily change the size of the > ARC lists on the fly. Compare it with the current value, and notify the DBA if the values are significantly different? Perhaps write the computed value to a file so the DBA can look at it later? Same with other values that are computed on the fly. In fact, it might make sense to store them in a table that gets periodically updated, and load their values from that table, and then the values in postgresql.conf or the command line would be the default that's used if there's nothing in the table (and if you really want fine-grained control of this process, you could stick a boolean column in the table to indicate whether or not to load the value from the table at startup time). -- Kevin Brown kevin@sysexperts.com From pgsql-performance-owner@postgresql.org Thu Oct 28 09:01:15 2004 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 5A3FC3A4859 for ; Thu, 28 Oct 2004 09:01:13 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 81756-04 for ; Thu, 28 Oct 2004 08:01:09 +0000 (GMT) Received: from mailer.unicite.fr.netcentrex.net (unknown [62.161.167.249]) by svr1.postgresql.org (Postfix) with ESMTP id E47A83A4828 for ; Thu, 28 Oct 2004 09:01:09 +0100 (BST) Received: from akira ([192.168.101.228]) by mailer.unicite.fr.netcentrex.net with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2657.72) id 42CYBH0S; Thu, 28 Oct 2004 10:01:08 +0200 From: "Alban Medici (NetCentrex)" To: "'Tom Lane'" , "'Thomas F.O'Connell'" Cc: "'PgSQL - Performance'" Subject: Re: Performance Anomalies in 7.4.5 Date: Thu, 28 Oct 2004 10:01:02 +0200 Organization: NetCentrex 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: AcS3uchEuW6rOf5GT3yhEH+DktOYswFCiwAw In-Reply-To: <25228.1098395599@sss.pgh.pa.us> X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1441 Message-Id: <20041028080109.E47A83A4828@svr1.postgresql.org> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/458 X-Sequence-Number: 8934 This topic probably available in 8.x will be very usefull for people just using postgresql as a "normal" Database user. -----Original Message----- From: pgsql-performance-owner@postgresql.org [mailto:pgsql-performance-owner@postgresql.org] On Behalf Of Tom Lane Sent: jeudi 21 octobre 2004 23:53 To: Thomas F.O'Connell Cc: PgSQL - Performance Subject: Re: [PERFORM] Performance Anomalies in 7.4.5 "Thomas F.O'Connell" writes: > -> Nested Loop (cost=0.00..0.01 rows=1 width=8) (actual > time=1.771..298305.531 rows=2452 loops=1) > Join Filter: ("inner".id = "outer".id) > -> Seq Scan on userdata u (cost=0.00..0.00 rows=1 width=8) > (actual time=0.026..11.869 rows=2452 loops=1) > -> Seq Scan on userdata_history h (cost=0.00..0.00 rows=1 > width=8) (actual time=0.005..70.519 rows=41631 loops=2452) > Filter: (id = 18181::bigint) > Total runtime: 298321.926 ms > (7 rows) What's killing you here is that the planner thinks these tables are completely empty (notice the zero cost estimates, which implies the table has zero blocks --- the fact that the rows estimate is 1 and not 0 is the result of sanity-check clamping inside costsize.c). This leads it to choose a nestloop, which would be the best plan if there were only a few rows involved, but it degenerates rapidly when there are not. It's easy to fall into this trap when truncating and reloading tables; all you need is an "analyze" while the table is empty. The rule of thumb is to analyze just after you reload the table, not just before. I'm getting more and more convinced that we need to drop the reltuples and relpages entries in pg_class, in favor of checking the physical table size whenever we make a plan. We could derive the tuple count estimate by having ANALYZE store a tuples-per-page estimate in pg_class and then multiply by the current table size; tuples-per-page should be a much more stable figure than total tuple count. One drawback to this is that it would require an additional lseek per table while planning, but that doesn't seem like a huge penalty. Probably the most severe objection to doing things this way is that the selected plan could change unexpectedly as a result of the physical table size changing. Right now the DBA can keep tight rein on actions that might affect plan selection (ie, VACUUM and ANALYZE), but that would go by the board with this. OTOH, we seem to be moving towards autovacuum, which also takes away any guarantees in this department. In any case this is speculation for 8.1; I think it's too late for 8.0. regards, tom lane ---------------------------(end of broadcast)--------------------------- TIP 2: you can get off all lists at once with the unregister command (send "unregister YourEmailAddressHere" to majordomo@postgresql.org) From pgsql-performance-owner@postgresql.org Thu Oct 28 16:27:30 2004 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 4F20E3A4915 for ; Thu, 28 Oct 2004 16:27:22 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 33533-01 for ; Thu, 28 Oct 2004 15:27:20 +0000 (GMT) Received: from ct.radiology.uiowa.edu (ct.radiology.uiowa.edu [129.255.60.186]) by svr1.postgresql.org (Postfix) with ESMTP id 4F4363A490E for ; Thu, 28 Oct 2004 16:27:19 +0100 (BST) Received: from [192.168.1.11] (12-217-241-0.client.mchsi.com [12.217.241.0]) by ct.radiology.uiowa.edu (8.11.6/8.11.6) with ESMTP id i9SFRF303897; Thu, 28 Oct 2004 10:27:16 -0500 Message-ID: <41810FC7.6070501@johnmeinel.com> Date: Thu, 28 Oct 2004 10:27:03 -0500 From: John Meinel User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Jaime Casanova Cc: pgsql-performance@postgresql.org Subject: Re: Sequential Scan with LIMIT References: <20041027171621.60824.qmail@web50006.mail.yahoo.com> In-Reply-To: <20041027171621.60824.qmail@web50006.mail.yahoo.com> X-Enigmail-Version: 0.86.0.0 X-Enigmail-Supports: pgp-inline, pgp-mime Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig17E31C0633DD7D3107A1AF65" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/459 X-Sequence-Number: 8935 This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig17E31C0633DD7D3107A1AF65 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Jaime Casanova wrote: [...] > > In http://www.postgresql.org/docs/faqs/FAQ.html under > "4.8) My queries are slow or don't make use of the > indexes. Why?" says: > > "However, LIMIT combined with ORDER BY often will use > an index because only a small portion of the table is > returned. In fact, though MAX() and MIN() don't use > indexes, it is possible to retrieve such values using > an index with ORDER BY and LIMIT: > SELECT col > FROM tab > ORDER BY col [ DESC ] > LIMIT 1;" > > So, maybe you can try your query as > > SELECT col FROM mytable > WHERE col = 'myval' > ORDER BY col > LIMIT 1; > > regards, > Jaime Casanova Thanks for the heads up. This actually worked. All queries against that table have turned into index scans instead of sequential. John =:-> --------------enig17E31C0633DD7D3107A1AF65 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (Cygwin) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFBgQ/OJdeBCYSNAAMRApqrAJ4p9MtARog1Loz4zx2h+cRZ1iTGhQCfQeSu pbf0rEN1a+c6NUF72t/7xyE= =RcJj -----END PGP SIGNATURE----- --------------enig17E31C0633DD7D3107A1AF65-- From pgsql-performance-owner@postgresql.org Thu Oct 28 16:38:37 2004 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 8FA2C3A3E5E for ; Thu, 28 Oct 2004 16:38:35 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 37681-01 for ; Thu, 28 Oct 2004 15:38:29 +0000 (GMT) Received: from vt-pe2550-001.VANTAGE.vantage.com (sol.vantage.com [64.80.203.242]) by svr1.postgresql.org (Postfix) with ESMTP id 4AF693A48BD for ; Thu, 28 Oct 2004 16:38:26 +0100 (BST) 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_01C4BD04.2A95B409" Subject: Re: Summary: can't handle large number of INSERT/UPDATEs Date: Thu, 28 Oct 2004 11:38:26 -0400 Message-ID: <4BAFBB6B9CC46F41B2AD7D9F4BBAF7850985F8@vt-pe2550-001.vantage.vantage.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PERFORM] Summary: can't handle large number of INSERT/UPDATEs Thread-Index: AcS61KspcGsLLkbcT7etdP20bnwpiQCLj6bw From: "Anjan Dave" To: Cc: "Anjan Dave" X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.2 tagged_above=0.0 required=5.0 tests=HTML_60_70, HTML_FONTCOLOR_UNKNOWN, HTML_MESSAGE X-Spam-Level: X-Archive-Number: 200410/460 X-Sequence-Number: 8936 This is a multi-part message in MIME format. ------_=_NextPart_001_01C4BD04.2A95B409 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable I would like to thank everyone for their timely suggestions. =20 The problem appears to be resolved now. We verified/modified - locking/indexes/vacuum/checkpoints/IO bottleneck/queries, etc. =20 Couple significant changes were the number of checkpoint segments were increased, and we moved over the database to a new SAN RAID10 volume (which was in plan anyway, just did it sooner). =20 =20 Thanks, Anjan =20 _____ =20 From: Anjan Dave=20 Sent: Monday, October 25, 2004 4:53 PM To: pgsql-performance@postgresql.org Subject: [PERFORM] can't handle large number of INSERT/UPDATEs =20 Hi, =20 I am dealing with an app here that uses pg to handle a few thousand concurrent web users. It seems that under heavy load, the INSERT and UPDATE statements to one or two specific tables keep queuing up, to the count of 150+ (one table has about 432K rows, other has about 2.6Million rows), resulting in 'wait's for other queries, and then everything piles up, with the load average shooting up to 10+.=20 =20 We (development) have gone through the queries/explain analyzes and made sure the appropriate indexes exist among other efforts put in. =20 I would like to know if there is anything that can be changed for better from the systems perspective. Here's what I have done and some recent changes from the system side: =20 -Upgraded from 7.4.0 to 7.4.1 sometime ago -Upgraded from RH8 to RHEL 3.0 -The settings from postgresql.conf (carried over, basically) are: shared_buffers =3D 10240 (80MB) max_connections =3D 400 sort_memory =3D 1024 effective_cache_size =3D 262144 (2GB) checkpoint_segments =3D 15 stats_start_collector =3D true stats_command_string =3D true=20 Rest everything is at default =20 In /etc/sysctl.conf (512MB shared mem) kernel.shmall =3D 536870912 kernel.shmmax =3D 536870912 =20 -This is a new Dell 6650 (quad XEON 2.2GHz, 8GB RAM, Internal HW RAID10), RHEL 3.0 (2.4.21-20.ELsmp), PG 7.4.1 -Vaccum Full run everyday -contrib/Reindex run everyday -Disabled HT in BIOS =20 I would greatly appreciate any helpful ideas. =20 Thanks in advance, =20 Anjan ------_=_NextPart_001_01C4BD04.2A95B409 Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable

I would like to thank everyone for = their timely suggestions.

 

The problem appears to be resolved = now. We verified/modified  - locking/indexes/vacuum/checkpoints/IO = bottleneck/queries, etc.

 

Couple significant changes were the = number of checkpoint segments were increased, and we moved over the database to = a new SAN RAID10 volume (which was in plan anyway, just did it = sooner).

 

 

Thanks,
Anjan

 

From: = Anjan Dave
Sent: Monday, October 25, = 2004 4:53 PM
To: pgsql-performance@postgresql.org
Subject: [PERFORM] can't = handle large number of INSERT/UPDATEs

 

Hi,

 

I am dealing with an app here that uses pg to handle = a few thousand concurrent web users. It seems that under heavy load, the = INSERT and UPDATE statements to one or two specific tables keep queuing up, to the = count of 150+ (one table has about 432K rows, other has about 2.6Million = rows), resulting in ‘wait’s for other queries, and then everything = piles up, with the load average shooting up to 10+. =

 

We (development) have gone through the = queries/explain analyzes and made sure the appropriate indexes exist among other efforts = put in.

 

I would like to know if there is anything that can be changed for better from the systems perspective. Here’s what I = have done and some recent changes from the system = side:

 

-Upgraded from 7.4.0 to 7.4.1 sometime = ago

-Upgraded from RH8 to RHEL = 3.0

-The settings from postgresql.conf (carried over, = basically) are:

         =    shared_buffers =3D 10240 (80MB)

         =    max_connections =3D 400

         =    sort_memory =3D 1024

         =    effective_cache_size =3D 262144 (2GB)

         =    checkpoint_segments =3D 15

stats_start_collector =3D = true

stats_command_string =3D = true

Rest everything is at = default

 

In /etc/sysctl.conf (512MB = shared mem)

kernel.shmall =3D = 536870912

kernel.shmmax =3D = 536870912

 

-This is a new Dell 6650 (quad XEON 2.2GHz, 8GB RAM, Internal HW RAID10), RHEL 3.0 (2.4.21-20.ELsmp), PG = 7.4.1

-Vaccum Full run = everyday

-contrib/Reindex run = everyday

-Disabled HT in BIOS

 

I would greatly appreciate any helpful = ideas.

 

Thanks in advance,

 

Anjan

------_=_NextPart_001_01C4BD04.2A95B409-- From pgsql-performance-owner@postgresql.org Thu Oct 28 17:07:45 2004 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 A5F6E3A3C21 for ; Thu, 28 Oct 2004 17:07:42 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 46741-03 for ; Thu, 28 Oct 2004 16:07:32 +0000 (GMT) Received: from davinci.ethosmedia.com (server226.ethosmedia.com [209.128.84.226]) by svr1.postgresql.org (Postfix) with ESMTP id BA3933A2B4A for ; Thu, 28 Oct 2004 17:07:31 +0100 (BST) 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 6567953 for pgsql-performance@postgresql.org; Thu, 28 Oct 2004 09:08:56 -0700 From: Josh Berkus Organization: Aglio Database Solutions To: pgsql-performance@postgresql.org Subject: Re: Performance Anomalies in 7.4.5 Date: Thu, 28 Oct 2004 09:07:00 -0700 User-Agent: KMail/1.6.2 References: <20041028080109.E47A83A4828@svr1.postgresql.org> In-Reply-To: <20041028080109.E47A83A4828@svr1.postgresql.org> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410280907.00445.josh@agliodbs.com> X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/461 X-Sequence-Number: 8937 Tom, > One drawback to this is that it would require an additional lseek per table > while planning, but that doesn't seem like a huge penalty. Hmmm ... would the additional lseek take longer for larger tables, or would it be a fixed cost? -- Josh Berkus Aglio Database Solutions San Francisco From pgsql-performance-owner@postgresql.org Thu Oct 28 17:31:35 2004 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 C41C33A4906 for ; Thu, 28 Oct 2004 17:31:33 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 54390-05 for ; Thu, 28 Oct 2004 16:31:32 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 575F73A4921 for ; Thu, 28 Oct 2004 17:31:31 +0100 (BST) 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 i9SGVTdC002816; Thu, 28 Oct 2004 12:31:29 -0400 (EDT) To: Josh Berkus Cc: pgsql-performance@postgresql.org Subject: Re: Performance Anomalies in 7.4.5 In-reply-to: <200410280907.00445.josh@agliodbs.com> References: <20041028080109.E47A83A4828@svr1.postgresql.org> <200410280907.00445.josh@agliodbs.com> Comments: In-reply-to Josh Berkus message dated "Thu, 28 Oct 2004 09:07:00 -0700" Date: Thu, 28 Oct 2004 12:31:28 -0400 Message-ID: <2815.1098981088@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/462 X-Sequence-Number: 8938 Josh Berkus writes: >> One drawback to this is that it would require an additional lseek per table >> while planning, but that doesn't seem like a huge penalty. > Hmmm ... would the additional lseek take longer for larger tables, or would it > be a fixed cost? Should be pretty much a fixed cost: one kernel call per table. regards, tom lane From pgsql-performance-owner@postgresql.org Thu Oct 28 18:21:42 2004 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 A360D3A4955 for ; Thu, 28 Oct 2004 18:21:39 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 69862-10 for ; Thu, 28 Oct 2004 17:21:37 +0000 (GMT) Received: from tht.net (vista.tht.net [216.126.88.2]) by svr1.postgresql.org (Postfix) with ESMTP id 4CF213A494A for ; Thu, 28 Oct 2004 18:21:37 +0100 (BST) Received: from [134.22.69.212] (dyn-69-212.tor.dsl.tht.net [134.22.69.212]) by tht.net (Postfix) with ESMTP id 9F22F76A35; Thu, 28 Oct 2004 13:21:35 -0400 (EDT) Subject: Re: Performance Anomalies in 7.4.5 From: Rod Taylor To: Tom Lane Cc: Josh Berkus , Postgresql Performance In-Reply-To: <2815.1098981088@sss.pgh.pa.us> References: <20041028080109.E47A83A4828@svr1.postgresql.org> <200410280907.00445.josh@agliodbs.com> <2815.1098981088@sss.pgh.pa.us> Content-Type: text/plain Message-Id: <1098984056.8557.637.camel@home> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 Date: Thu, 28 Oct 2004 13:20:56 -0400 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/463 X-Sequence-Number: 8939 On Thu, 2004-10-28 at 12:31, Tom Lane wrote: > Josh Berkus writes: > >> One drawback to this is that it would require an additional lseek per table > >> while planning, but that doesn't seem like a huge penalty. > > > Hmmm ... would the additional lseek take longer for larger tables, or would it > > be a fixed cost? > > Should be pretty much a fixed cost: one kernel call per table. Is this something that the bgwriter could periodically do and share the data? Possibly in the future it could even force a function or prepared statement recompile if the data has changed significantly? From pgsql-performance-owner@postgresql.org Thu Oct 28 18:50:44 2004 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 ECF3E3A4906 for ; Thu, 28 Oct 2004 18:50:41 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 82355-08 for ; Thu, 28 Oct 2004 17:50:33 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 5A1283A495A for ; Thu, 28 Oct 2004 18:50:33 +0100 (BST) 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 i9SHoTmo003473; Thu, 28 Oct 2004 13:50:29 -0400 (EDT) To: Rod Taylor Cc: Josh Berkus , Postgresql Performance Subject: Re: Performance Anomalies in 7.4.5 In-reply-to: <1098984056.8557.637.camel@home> References: <20041028080109.E47A83A4828@svr1.postgresql.org> <200410280907.00445.josh@agliodbs.com> <2815.1098981088@sss.pgh.pa.us> <1098984056.8557.637.camel@home> Comments: In-reply-to Rod Taylor message dated "Thu, 28 Oct 2004 13:20:56 -0400" Date: Thu, 28 Oct 2004 13:50:29 -0400 Message-ID: <3472.1098985829@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/464 X-Sequence-Number: 8940 Rod Taylor writes: > On Thu, 2004-10-28 at 12:31, Tom Lane wrote: >> Should be pretty much a fixed cost: one kernel call per table. > Is this something that the bgwriter could periodically do and share the > data? I think a kernel call would be cheaper. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 29 00:47:11 2004 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 EC5863A2B41 for ; Fri, 29 Oct 2004 00:47:04 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 91701-04 for ; Thu, 28 Oct 2004 23:46:56 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [66.143.173.58]) by svr1.postgresql.org (Postfix) with ESMTP id 2388D3A2B10 for ; Fri, 29 Oct 2004 00:46:58 +0100 (BST) Received: by flake.decibel.org (Postfix, from userid 1001) id A93A71C900; Thu, 28 Oct 2004 23:46:58 +0000 (GMT) Date: Thu, 28 Oct 2004 18:46:58 -0500 From: "Jim C. Nasby" To: Tom Lane Cc: John Meinel , pgsql-performance@postgresql.org Subject: Re: Sequential Scan with LIMIT Message-ID: <20041028234658.GI55164@decibel.org> References: <417C023F.1080502@johnmeinel.com> <17689.1098648713@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <17689.1098648713@sss.pgh.pa.us> X-Operating-System: FreeBSD 4.10-RELEASE-p3 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/465 X-Sequence-Number: 8941 On Sun, Oct 24, 2004 at 04:11:53PM -0400, Tom Lane wrote: > But the LIMIT will cut the cost of the seqscan case too. Given the > numbers you posit above, about one row in five will have 'myval', so a > seqscan can reasonably expect to hit the first matching row in the first > page of the table. This is still cheaper than doing an index scan > (which must require reading at least one index page plus at least one > table page). > > The test case you are showing is probably suffering from nonrandom > placement of this particular data value; which is something that the > statistics we keep are too crude to detect. Isn't that exactly what pg_stats.correlation is? -- 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 Fri Oct 29 00:49:44 2004 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 E8E233A3BE8 for ; Fri, 29 Oct 2004 00:49:40 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 89496-08 for ; Thu, 28 Oct 2004 23:49:31 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 4C61F3A2B41 for ; Fri, 29 Oct 2004 00:49:31 +0100 (BST) 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 i9SNnTOk009173; Thu, 28 Oct 2004 19:49:29 -0400 (EDT) To: "Jim C. Nasby" Cc: John Meinel , pgsql-performance@postgresql.org Subject: Re: Sequential Scan with LIMIT In-reply-to: <20041028234658.GI55164@decibel.org> References: <417C023F.1080502@johnmeinel.com> <17689.1098648713@sss.pgh.pa.us> <20041028234658.GI55164@decibel.org> Comments: In-reply-to "Jim C. Nasby" message dated "Thu, 28 Oct 2004 18:46:58 -0500" Date: Thu, 28 Oct 2004 19:49:28 -0400 Message-ID: <9172.1099007368@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/466 X-Sequence-Number: 8942 "Jim C. Nasby" writes: > On Sun, Oct 24, 2004 at 04:11:53PM -0400, Tom Lane wrote: >> The test case you are showing is probably suffering from nonrandom >> placement of this particular data value; which is something that the >> statistics we keep are too crude to detect. > Isn't that exactly what pg_stats.correlation is? No. A far-from-zero correlation gives you a clue that on average, *all* the data values are placed nonrandomly ... but it doesn't really tell you much one way or the other about a single data value. regards, tom lane From pgsql-performance-owner@postgresql.org Fri Oct 29 00:54:40 2004 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 080253A49E7 for ; Fri, 29 Oct 2004 00:54:39 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 92614-10 for ; Thu, 28 Oct 2004 23:54:27 +0000 (GMT) Received: from flake.decibel.org (flake.decibel.org [66.143.173.58]) by svr1.postgresql.org (Postfix) with ESMTP id 4F63C3A49F8 for ; Fri, 29 Oct 2004 00:54:30 +0100 (BST) Received: by flake.decibel.org (Postfix, from userid 1001) id BBC521C902; Thu, 28 Oct 2004 23:54:30 +0000 (GMT) Date: Thu, 28 Oct 2004 18:54:30 -0500 From: "Jim C. Nasby" To: Tom Lane Cc: John Meinel , pgsql-performance@postgresql.org Subject: Re: Sequential Scan with LIMIT Message-ID: <20041028235430.GJ55164@decibel.org> References: <417C023F.1080502@johnmeinel.com> <17689.1098648713@sss.pgh.pa.us> <20041028234658.GI55164@decibel.org> <9172.1099007368@sss.pgh.pa.us> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <9172.1099007368@sss.pgh.pa.us> X-Operating-System: FreeBSD 4.10-RELEASE-p3 i386 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.6i X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/467 X-Sequence-Number: 8943 On Thu, Oct 28, 2004 at 07:49:28PM -0400, Tom Lane wrote: > "Jim C. Nasby" writes: > > On Sun, Oct 24, 2004 at 04:11:53PM -0400, Tom Lane wrote: > >> The test case you are showing is probably suffering from nonrandom > >> placement of this particular data value; which is something that the > >> statistics we keep are too crude to detect. > > > Isn't that exactly what pg_stats.correlation is? > > No. A far-from-zero correlation gives you a clue that on average, *all* > the data values are placed nonrandomly ... but it doesn't really tell > you much one way or the other about a single data value. Maybe I'm confused about what the original issue was then... it appeared that you were suggesting PGSQL was doing a seq scan instead of an index scan because it thought it would find it on the first page if the data was randomly distributed. If the correlation is highly non-zero though, shouldn't it 'play it safe' and assume that unless it's picking the min or max value stored in statistics it will be better to do an index scan, since the value it's looking for is probably in the middle of the table somewhere? IE: if the values in the field are between 1 and 5 and the table is clustered on that field then clearly an index scan would be better to find a row with field=3 than a seq scan. -- 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 Fri Oct 29 15:02:25 2004 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 653723A4BCE for ; Fri, 29 Oct 2004 15:02:23 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 39814-05 for ; Fri, 29 Oct 2004 14:02:15 +0000 (GMT) Received: from sccrmhc12.comcast.net (sccrmhc12.comcast.net [204.127.202.56]) by svr1.postgresql.org (Postfix) with ESMTP id 133293A4BCB for ; Fri, 29 Oct 2004 15:02:18 +0100 (BST) Received: from [127.0.0.1] (c-67-160-247-54.client.comcast.net[67.160.247.54]) by comcast.net (sccrmhc12) with ESMTP id <20041029140217012009laume> (Authid: pathat); Fri, 29 Oct 2004 14:02:18 +0000 Message-ID: <41824E05.8000903@comcast.net> Date: Fri, 29 Oct 2004 07:04:53 -0700 From: Patrick Hatcher User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913) X-Accept-Language: en-us, en MIME-Version: 1.0 To: pgsql-performance@postgresql.org Subject: determining max_fsm_pages 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/468 X-Sequence-Number: 8944 Pg: 7.4.5 8G ram 200G RAID5 I have my fsm set as such: max_fsm_pages = 300000 # min max_fsm_relations*16, 6 bytes each max_fsm_relations = 500 # min 100, ~50 bytes each I just did a vacuum full on one table and saw this result: INFO: analyzing "cdm.cdm_fed_agg_purch" INFO: "cdm_fed_agg_purch": 667815 pages, 3000 rows sampled, 52089570 estimated total rows My question is this: I have about 8 databases running on this server. When I do a vacuum full on each of these databases, there is a INFO section that I assume is the total pages used for that database. Should add ALL these individual pages together and pad the total and use this as my new max_fsm_pages? Should I do the same thing with max_fsm_relations? TIA Patrick From pgsql-performance-owner@postgresql.org Fri Oct 29 15:29:59 2004 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 997D83A469A for ; Fri, 29 Oct 2004 15:29:57 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 49078-05 for ; Fri, 29 Oct 2004 14:29:49 +0000 (GMT) Received: from sraigw.sra.co.jp (sraigw.sra.co.jp [202.32.10.2]) by svr1.postgresql.org (Postfix) with ESMTP id 010893A4BD8 for ; Fri, 29 Oct 2004 15:29:51 +0100 (BST) Received: from srascb.sra.co.jp (srascb [133.137.8.65]) by sraigw.sra.co.jp (Postfix) with ESMTP id B83F06278D; Fri, 29 Oct 2004 23:29:50 +0900 (JST) Received: from srascb.sra.co.jp (localhost [127.0.0.1]) by localhost.sra.co.jp (Postfix) with ESMTP id A120710CD06; Fri, 29 Oct 2004 23:29:50 +0900 (JST) Received: from sranhm.sra.co.jp (sranhm [133.137.44.16]) by srascb.sra.co.jp (Postfix) with ESMTP id 8695B10CD04; Fri, 29 Oct 2004 23:29:50 +0900 (JST) Received: from localhost (sraihb-hub.sra.co.jp [133.137.8.6]) by sranhm.sra.co.jp (8.9.3+3.2W/3.7W-srambox) with ESMTP id XAA23107; Fri, 29 Oct 2004 23:29:50 +0900 Date: Fri, 29 Oct 2004 23:31:51 +0900 (JST) Message-Id: <20041029.233151.78703659.t-ishii@sra.co.jp> To: pathat@comcast.net Cc: pgsql-performance@postgresql.org Subject: Re: determining max_fsm_pages From: Tatsuo Ishii In-Reply-To: <41824E05.8000903@comcast.net> References: <41824E05.8000903@comcast.net> X-Mailer: Mew version 2.3 on Emacs 20.7 / Mule 4.1 =?iso-2022-jp?B?KBskQjAqGyhCKQ==?= 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=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/469 X-Sequence-Number: 8945 > Pg: 7.4.5 > 8G ram > 200G RAID5 > > I have my fsm set as such: > max_fsm_pages = 300000 # min max_fsm_relations*16, 6 bytes each > max_fsm_relations = 500 # min 100, ~50 bytes each > > > I just did a vacuum full on one table and saw this result: > INFO: analyzing "cdm.cdm_fed_agg_purch" > INFO: "cdm_fed_agg_purch": 667815 pages, 3000 rows sampled, 52089570 > estimated total rows > > > My question is this: I have about 8 databases running on this server. > When I do a vacuum full on each of these databases, there is a INFO > section that I assume is the total pages used for that database. Should > add ALL these individual pages together and pad the total and use this > as my new max_fsm_pages? Should I do the same thing with max_fsm_relations? I think that's too much and too big FSM affects performance in my opinion. The easiest way to calculate appropreate FSM size is doing vacuumdb -a -v and watching the message. At the very end, you would see something like: INFO: free space map: 13 relations, 1447 pages stored; 1808 total pages needed DETAIL: Allocated FSM size: 100 relations + 1600 pages = 19 kB shared memory. In this case 1808 is the minimum FSM size. Of course this number would change depending on the frequency of VACUUM. Therefore you need some room for the FSM size. -- Tatsuo Ishii From pgsql-performance-owner@postgresql.org Fri Oct 29 15:37:30 2004 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 3DE323A1D9E for ; Fri, 29 Oct 2004 15:37:29 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 48787-09 for ; Fri, 29 Oct 2004 14:37:21 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 25CD43A1D9C for ; Fri, 29 Oct 2004 15:37:24 +0100 (BST) 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 i9TEbOql015930; Fri, 29 Oct 2004 10:37:24 -0400 (EDT) To: Patrick Hatcher Cc: pgsql-performance@postgresql.org Subject: Re: determining max_fsm_pages In-reply-to: <41824E05.8000903@comcast.net> References: <41824E05.8000903@comcast.net> Comments: In-reply-to Patrick Hatcher message dated "Fri, 29 Oct 2004 07:04:53 -0700" Date: Fri, 29 Oct 2004 10:37:23 -0400 Message-ID: <15929.1099060643@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/470 X-Sequence-Number: 8946 Patrick Hatcher writes: > My question is this: I have about 8 databases running on this server. > When I do a vacuum full on each of these databases, there is a INFO > section that I assume is the total pages used for that database. Should > add ALL these individual pages together and pad the total and use this > as my new max_fsm_pages? Should I do the same thing with max_fsm_relations? No, the numbers shown at the end of a vacuum verbose printout reflect the current cluster-wide FSM demand. BTW you do *not* want to use FULL because that's not going to reflect the FSM requirements when you are just running normal vacuums. I would vacuum all your databases (to make sure each one's FSM contents are pretty up-to-date) and then take the numbers shown by the last one as your targets. If you find yourself having to raise max_fsm_relations, it may be necessary to repeat the vacuuming cycle before you can get a decent total for max_fsm_pages. IIRC, the vacuum printout does include in "needed" a count of pages that it would have stored if it'd had room; but this is only tracked for relations that have an FSM relation entry. regards, tom lane From pgsql-hackers-owner@postgresql.org Sat Oct 30 14:39:55 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 25AF23A3E71 for ; Sat, 30 Oct 2004 14:39:54 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 80880-07 for ; Sat, 30 Oct 2004 13:39:48 +0000 (GMT) Received: from smtp109.mail.sc5.yahoo.com (smtp109.mail.sc5.yahoo.com [66.163.170.7]) by svr1.postgresql.org (Postfix) with SMTP id 283043A3E9D for ; Sat, 30 Oct 2004 14:39:51 +0100 (BST) Received: from unknown (HELO jupiter.black-lion.info) (janwieck@68.80.245.191 with login) by smtp109.mail.sc5.yahoo.com with SMTP; 30 Oct 2004 13:39:51 -0000 Received: from [172.21.8.18] (gv90209.gv.psu.edu [146.186.90.209]) (authenticated bits=0) by jupiter.black-lion.info (8.12.10/8.12.9) with ESMTP id i9UDdn94057594; Sat, 30 Oct 2004 09:39:49 -0400 (EDT) (envelope-from JanWieck@Yahoo.com) Message-ID: <418399D7.6090204@Yahoo.com> Date: Sat, 30 Oct 2004 09:40:39 -0400 From: Jan Wieck User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.7.1) Gecko/20040707 X-Accept-Language: en-us, en MIME-Version: 1.0 To: Tom Lane Cc: Greg Stark , pgsql-hackers@postgresql.org Subject: Re: [PATCHES] ARC Memory Usage analysis References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <87mzyavhxc.fsf@stark.xeocode.com> <87hdoiv62b.fsf@stark.xeocode.com> <790.1098741205@sss.pgh.pa.us> <878y9uukgb.fsf@stark.xeocode.com> <4887.1098770035@sss.pgh.pa.us> In-Reply-To: <4887.1098770035@sss.pgh.pa.us> 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=1.6 tagged_above=0.0 required=5.0 tests=RCVD_IN_NJABL_DUL, RCVD_IN_SORBS_DUL X-Spam-Level: * X-Archive-Number: 200410/1001 X-Sequence-Number: 60504 On 10/26/2004 1:53 AM, Tom Lane wrote: > Greg Stark writes: >> Tom Lane writes: >>> Another issue is what we do with the effective_cache_size value once we >>> have a number we trust. We can't readily change the size of the ARC >>> lists on the fly. > >> Huh? I thought effective_cache_size was just used as an factor the cost >> estimation equation. > > Today, that is true. Jan is speculating about using it as a parameter > of the ARC cache management algorithm ... and that worries me. If we need another config option, it's not that we are running out of possible names, is it? Jan -- #======================================================================# # It's easier to get forgiveness for being wrong than for being right. # # Let's break this rule - forgive me. # #================================================== JanWieck@Yahoo.com # From pgsql-hackers-owner@postgresql.org Sat Oct 30 17:53:44 2004 X-Original-To: pgsql-hackers-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id 915A13A3EBC for ; Sat, 30 Oct 2004 17:53:42 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 21298-10 for ; Sat, 30 Oct 2004 16:53:40 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 4EAF83A3EAA for ; Sat, 30 Oct 2004 17:53:40 +0100 (BST) 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 i9UGraPG005986; Sat, 30 Oct 2004 12:53:37 -0400 (EDT) To: Jan Wieck Cc: Greg Stark , pgsql-hackers@postgresql.org Subject: Re: [PATCHES] ARC Memory Usage analysis In-reply-to: <418399D7.6090204@Yahoo.com> References: <1098471059.20926.49.camel@localhost.localdomain> <41796115.2020501@Yahoo.com> <20041022200955.GC4382@it.is.rice.edu> <417D1D01.3090401@Yahoo.com> <20651.1098720192@sss.pgh.pa.us> <87mzyavhxc.fsf@stark.xeocode.com> <87hdoiv62b.fsf@stark.xeocode.com> <790.1098741205@sss.pgh.pa.us> <878y9uukgb.fsf@stark.xeocode.com> <4887.1098770035@sss.pgh.pa.us> <418399D7.6090204@Yahoo.com> Comments: In-reply-to Jan Wieck message dated "Sat, 30 Oct 2004 09:40:39 -0400" Date: Sat, 30 Oct 2004 12:53:36 -0400 Message-ID: <5985.1099155216@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/1002 X-Sequence-Number: 60505 Jan Wieck writes: > On 10/26/2004 1:53 AM, Tom Lane wrote: >> Greg Stark writes: > Tom Lane writes: >>> Another issue is what we do with the effective_cache_size value once we >>> have a number we trust. We can't readily change the size of the ARC >>> lists on the fly. >> > Huh? I thought effective_cache_size was just used as an factor the cost > estimation equation. >> >> Today, that is true. Jan is speculating about using it as a parameter >> of the ARC cache management algorithm ... and that worries me. > If we need another config option, it's not that we are running out of > possible names, is it? No, the point is that the value is not very trustworthy at the moment. regards, tom lane From pgsql-performance-owner@postgresql.org Sat Oct 30 22:13:00 2004 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 4D1A93A3F2C for ; Sat, 30 Oct 2004 22:12:58 +0100 (BST) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 87255-01 for ; Sat, 30 Oct 2004 21:12:47 +0000 (GMT) Received: from web11.manitu.net (web11.manitu.net [217.11.48.111]) by svr1.postgresql.org (Postfix) with ESMTP id 1B23C3A2019 for ; Sat, 30 Oct 2004 22:12:48 +0100 (BST) Received: from [192.168.0.1] (dsl-082-082-202-218.arcor-ip.net [82.82.202.218]) (authenticated) by web11.manitu.net (8.10.2-SOL3/8.10.2) with ESMTP id i9ULCPw23745; Sat, 30 Oct 2004 23:12:25 +0200 Subject: Re: can't handle large number of INSERT/UPDATEs From: Markus Bertheau To: Matt Clark Cc: John Meinel , Tom Lane , Curtis Zinzilieta , Anjan Dave , Rod Taylor , Postgresql Performance In-Reply-To: <417F3B91.9080002@ymogen.net> References: <15812.1098847291@sss.pgh.pa.us> <417F2181.4000801@johnmeinel.com> <417F3B91.9080002@ymogen.net> Content-Type: text/plain Message-Id: <1099170744.2646.3.camel@teetnang> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Date: Sat, 30 Oct 2004 23:12:25 +0200 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/471 X-Sequence-Number: 8947 Turns out the man page of vmstat in procps was changed on Oct 8 2002: http://cvs.sourceforge.net/viewcvs.py/procps/procps/vmstat.8?r1=1.1&r2=1.2 in reaction to a debian bug report: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=157935 -- Markus Bertheau From pgsql-performance-owner@postgresql.org Sun Oct 31 06:03:52 2004 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 1738E3A4045 for ; Sun, 31 Oct 2004 06:03:09 +0000 (GMT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 92972-04 for ; Sun, 31 Oct 2004 06:02:59 +0000 (GMT) Received: from purple.west.spy.net (mail.west.spy.net [66.149.231.226]) by svr1.postgresql.org (Postfix) with ESMTP id B9AD03A4023 for ; Sun, 31 Oct 2004 06:02:56 +0000 (GMT) Received: from [192.168.1.50] (dustinti.west.spy.net [192.168.1.50]) (using TLSv1 with cipher RC4-SHA (128/128 bits)) (Client did not present a certificate) by purple.west.spy.net (Postfix) with ESMTP id 4378A11E; Sat, 30 Oct 2004 23:02:54 -0700 (PDT) In-Reply-To: <20041027075748.GA8355@hardpoint.ciar.org> References: <20041027075748.GA8355@hardpoint.ciar.org> Mime-Version: 1.0 (Apple Message framework v619) Content-Type: text/plain; charset=US-ASCII; format=flowed Message-Id: <8132EBC8-2B02-11D9-A2DE-000A957659CC@spy.net> Content-Transfer-Encoding: 7bit Cc: pgsql-performance@postgresql.org From: Dustin Sallings Subject: Re: psql large RSS (1.6GB) Date: Sat, 30 Oct 2004 23:02:54 -0700 To: TTK Ciar X-Mailer: Apple Mail (2.619) X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200410/474 X-Sequence-Number: 8950 On Oct 27, 2004, at 0:57, TTK Ciar wrote: > At a guess, it looks like the data set is being buffered in its > entirety by psql, before any data is written to the output file, > which is surprising. I would have expected it to grab data as it > appeared on the socket from postmaster and write it to disk. Is > there something we can do to stop psql from buffering results? > Does anyone know what's going on here? Yes, the result set is sent back to the client before it can be used. An easy workaround when dealing with this much data is to use a cursor. Something like this: db# start transaction; START TRANSACTION db# declare logcur cursor for select * from some_table; DECLARE CURSOR db# fetch 5 in logcur; [...] (5 rows) This will do approximately what you expected the select to do in the first, place, but the fetch will decide how many rows to buffer into the client at a time. > If the solution is to just write a little client that uses perl > DBI to fetch rows one at a time and write them out, that's doable, > but it would be nice if psql could be made to "just work" without > the monster RSS. It wouldn't make a difference unless that driver implements the underlying protocol on its own. -- SPY My girlfriend asked me which one I like better. pub 1024/3CAE01D5 1994/11/03 Dustin Sallings | Key fingerprint = 87 02 57 08 02 D0 DA D6 C8 0F 3E 65 51 98 D8 BE L_______________________ I hope the answer won't upset her. ____________ From pgsql-performance-owner@postgresql.org Sun Oct 31 10:27:14 2004 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 35E023A40E6 for ; Sun, 31 Oct 2004 10:27:13 +0000 (GMT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 41598-08 for ; Sun, 31 Oct 2004 10:27:06 +0000 (GMT) Received: from web11.manitu.net (web11.manitu.net [217.11.48.111]) by svr1.postgresql.org (Postfix) with ESMTP id 9BC9F3A40E5 for ; Sun, 31 Oct 2004 10:27:07 +0000 (GMT) Received: from [192.168.0.1] (dsl-082-082-203-101.arcor-ip.net [82.82.203.101]) (authenticated) by web11.manitu.net (8.10.2-SOL3/8.10.2) with ESMTP id i9VAQvw17192; Sun, 31 Oct 2004 11:26:57 +0100 Subject: Re: psql large RSS (1.6GB) From: Markus Bertheau To: TTK Ciar Cc: pgsql-performance@postgresql.org In-Reply-To: <20041027075748.GA8355@hardpoint.ciar.org> References: <20041027075748.GA8355@hardpoint.ciar.org> Content-Type: text/plain; charset=UTF-8 Message-Id: <1099218421.2636.0.camel@teetnang> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Date: Sun, 31 Oct 2004 11:27:01 +0100 Content-Transfer-Encoding: quoted-printable X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=1.1 tagged_above=0.0 required=5.0 tests=RCVD_IN_DSBL X-Spam-Level: * X-Archive-Number: 200410/475 X-Sequence-Number: 8951 =D0=92 =D0=A1=D1=80=D0=B4, 27.10.2004, =D0=B2 09:57, TTK Ciar =D0=BF=D0=B8= =D1=88=D0=B5=D1=82: > brad=3D# explain analyse select ServerDisks.servername,ServerDisks.diskse= rial,ServerDisks.diskmountpoint,DiskFiles.name,DiskFiles.md5 from DiskFiles= ,ServerDisks where DiskFiles.diskserial=3DServerDisks.diskserial; > QUERY PLAN = =20 > ------------------------------------------------------------------ > Hash Join (cost=3D22.50..65.00 rows=3D1000 width=3D274) (actual time=3D= 118.584..124653.729 rows=3D10133349 loops=3D1) > Hash Cond: (("outer".diskserial)::text =3D ("inner".diskserial)::text) > -> Seq Scan on diskfiles (cost=3D0.00..20.00 rows=3D1000 width=3D198= ) (actual time=3D7.201..31336.063 rows=3D10133349 loops=3D1) > -> Hash (cost=3D20.00..20.00 rows=3D1000 width=3D158) (actual time= =3D90.821..90.821 rows=3D0 loops=3D1) > -> Seq Scan on serverdisks (cost=3D0.00..20.00 rows=3D1000 wid= th=3D158) (actual time=3D9.985..87.364 rows=3D2280 loops=3D1) > Total runtime: 130944.586 ms You should run ANALYZE on your database once in a while. --=20 Markus Bertheau From pgsql-performance-owner@postgresql.org Sun Oct 31 18:39:07 2004 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 852EA3A4253 for ; Sun, 31 Oct 2004 18:39:03 +0000 (GMT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 50857-03 for ; Sun, 31 Oct 2004 18:38:57 +0000 (GMT) Received: from themode.com (themode.com [161.58.169.198]) by svr1.postgresql.org (Postfix) with ESMTP id 281C73A4248 for ; Sun, 31 Oct 2004 18:38:57 +0000 (GMT) Received: from localhost (bde@localhost) by themode.com (8.12.11/8.12.10) with ESMTP id i9VIctPR094972; Sun, 31 Oct 2004 13:38:55 -0500 (EST) Date: Sun, 31 Oct 2004 13:38:55 -0500 (EST) From: brew@theMode.com X-X-Sender: mode@themode.com To: Aaron Mulder Cc: pgsql-performance@postgresql.org Subject: Thanks Chariot Solutions 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=0.3 tagged_above=0.0 required=5.0 tests=NO_REAL_NAME X-Spam-Level: X-Archive-Number: 200410/476 X-Sequence-Number: 8952 Many thanks to Chariot Solutions, http://chariotsolutions.com, for hosting Bruce Momjian giving one of his PostgreSQL seminars outside of Philadelphia, PA yesterday. There were about sixty folks there, one person driving from Toronto and another coming from California (!). I found it very enlightening and learned some new things about PostgreSQL, even if I *did* doze off for a few minutes after lunch when all my energy was concentrated in my stomach. But after I got back home I dreamt about PostgreSQL all night long!!! Thanks Bruce and Chariot Solutions. brew ========================================================================== Strange Brew (brew@theMode.com) Check out my Musician's Online Database Exchange (The MODE Pages) http://www.TheMode.com ========================================================================== From pgsql-performance-owner@postgresql.org Sun Oct 31 21:10:52 2004 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 F0D863A42C0 for ; Sun, 31 Oct 2004 21:10:49 +0000 (GMT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 81124-08 for ; Sun, 31 Oct 2004 21:10:31 +0000 (GMT) Received: from trade-india.com (unknown [203.101.29.52]) by svr1.postgresql.org (Postfix) with SMTP id 00D213A4281 for ; Sun, 31 Oct 2004 21:10:30 +0000 (GMT) Received: (qmail 26720 invoked from network); 31 Oct 2004 21:03:10 -0000 Received: from unknown (HELO 192.168.0.66) (192.168.0.66) by 192.168.0.66 with SMTP; 31 Oct 2004 21:03:10 -0000 Received: from 192.168.0.11 (proxying for 203.122.31.2) (SquirrelMail authenticated user mallah) by office1.trade-india.com with HTTP; Mon, 1 Nov 2004 02:33:10 +0530 (IST) Message-ID: <46946.192.168.0.11.1099256590.squirrel@office1.trade-india.com> Date: Mon, 1 Nov 2004 02:33:10 +0530 (IST) Subject: Speeding up Gist Index creations From: mallah@trade-india.com To: pgsql-performance@postgresql.org 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-Spam-Status: No, hits=1.4 tagged_above=0.0 required=5.0 tests=NO_REAL_NAME, PRIORITY_NO_NAME, RCVD_NUMERIC_HELO X-Spam-Level: * X-Archive-Number: 200410/477 X-Sequence-Number: 8953 Hi , Gist indexes take a long time to create as compared to normal indexes is there any way to speed them up ? (for example by modifying sort_mem or something temporarily ) Regds Mallah.