From:
Jeroen Vermeulen <jtv(at)xs4all(dot)nl>
To:
Robert Haas <robertmhaas(at)gmail(dot)com>
Cc:
Greg Stark <gsstark(at)mit(dot)edu>, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>,
Bart Samwel <bart(at)samwel(dot)tk>, Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com>,
pgsql-hackers(at)postgresql(dot)org
Subject:
Re: Avoiding bad prepared-statement plans.
Date:
2010-02-26 02:48:05
Message-ID:
4B873665.20208@xs4all.nl (view raw or flat )
Thread:
2010-02-09 12:08:54 from Jeroen Vermeulen <jtv(at)xs4all(dot)nl>
2010-02-09 12:59:15 from Yeb Havinga <yebhavinga(at)gmail(dot)com>
2010-02-09 13:46:55 from Jeroen Vermeulen <jtv(at)xs4all(dot)nl>
2010-02-09 14:53:17 from Mark Mielke <mark(at)mark(dot)mielke(dot)cc>
2010-02-10 00:47:18 from Kris Jurka <books(at)ejurka(dot)com>
2010-02-11 03:12:41 from Bruce Momjian <bruce(at)momjian(dot)us>
2010-02-11 04:07:57 from Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
2010-02-09 22:21:43 from Dimitri Fontaine <dfontaine(at)hi-media(dot)com>
2010-02-09 13:50:32 from Richard Huxton <dev(at)archonet(dot)com>
2010-02-09 14:25:19 from Jeroen Vermeulen <jtv(at)xs4all(dot)nl>
2010-02-09 14:45:57 from Richard Huxton <dev(at)archonet(dot)com>
2010-02-09 14:10:22 from Andres Freund <andres(at)anarazel(dot)de>
2010-02-09 14:28:01 from Jeroen Vermeulen <jtv(at)xs4all(dot)nl>
2010-02-09 16:43:52 from Robert Haas <robertmhaas(at)gmail(dot)com>
2010-02-09 16:59:41 from Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
2010-02-11 03:15:01 from Bruce Momjian <bruce(at)momjian(dot)us>
2010-02-11 06:26:15 from Дмитрий Фефелов <fozzy(at)ac-sw(dot)com>
2010-02-11 12:09:33 from Bart Samwel <bart(at)samwel(dot)tk>
2010-02-11 12:25:28 from Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com>
2010-02-11 12:39:49 from Bart Samwel <bart(at)samwel(dot)tk>
2010-02-11 12:41:12 from Robert Haas <robertmhaas(at)gmail(dot)com>
2010-02-11 12:48:14 from Bart Samwel <bart(at)samwel(dot)tk>
2010-02-11 13:04:46 from Robert Haas <robertmhaas(at)gmail(dot)com>
2010-02-11 16:17:40 from Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
2010-02-11 16:39:11 from Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com>
2010-02-15 19:11:34 from Bruce Momjian <bruce(at)momjian(dot)us>
2010-02-15 20:03:21 from Robert Haas <robertmhaas(at)gmail(dot)com>
2010-02-15 20:05:29 from Bruce Momjian <bruce(at)momjian(dot)us>
2010-02-16 14:28:57 from Greg Stark <gsstark(at)mit(dot)edu>
2010-02-16 14:31:44 from Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com>
2010-02-16 20:17:32 from Bruce Momjian <bruce(at)momjian(dot)us>
2010-02-17 03:12:50 from Greg Stark <gsstark(at)mit(dot)edu>
2010-02-15 19:51:26 from Jeroen Vermeulen <jtv(at)xs4all(dot)nl>
2010-02-16 14:22:00 from Greg Stark <gsstark(at)mit(dot)edu>
2010-02-17 22:52:14 from Jeroen Vermeulen <jtv(at)xs4all(dot)nl>
2010-02-21 12:37:50 from Robert Haas <robertmhaas(at)gmail(dot)com>
2010-02-26 02:48:05 from Jeroen Vermeulen <jtv(at)xs4all(dot)nl>
2010-02-26 03:19:21 from Robert Haas <robertmhaas(at)gmail(dot)com>
2010-02-26 03:40:35 from Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
2010-02-26 04:01:14 from Robert Haas <robertmhaas(at)gmail(dot)com>
2010-02-26 04:25:46 from Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
2010-02-28 14:12:32 from Greg Stark <gsstark(at)mit(dot)edu>
2010-02-26 04:28:14 from Alex Hunsaker <badalex(at)gmail(dot)com>
2010-02-26 04:46:20 from Alex Hunsaker <badalex(at)gmail(dot)com>
2010-02-26 05:11:56 from Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
2010-02-26 06:29:41 from Alex Hunsaker <badalex(at)gmail(dot)com>
2010-02-26 15:07:00 from Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
2010-02-26 15:27:30 from Alex Hunsaker <badalex(at)gmail(dot)com>
2010-02-26 16:51:53 from Robert Haas <robertmhaas(at)gmail(dot)com>
2010-02-26 18:11:44 from Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
2010-02-26 16:50:37 from Robert Haas <robertmhaas(at)gmail(dot)com>
2010-02-26 20:02:40 from Alex Hunsaker <badalex(at)gmail(dot)com>
2010-02-26 08:13:31 from Mark Mielke <mark(at)mark(dot)mielke(dot)cc>
2010-02-26 10:20:17 from Jeroen Vermeulen <jtv(at)xs4all(dot)nl>
2010-02-26 14:49:35 from Mark Mielke <mark(at)mark(dot)mielke(dot)cc>
2010-02-26 16:27:11 from Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
2010-02-26 16:54:10 from Robert Haas <robertmhaas(at)gmail(dot)com>
2010-02-26 17:01:59 from Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
2010-02-26 20:11:16 from Yeb Havinga <yebhavinga(at)gmail(dot)com>
2010-02-26 20:29:11 from Mark Mielke <mark(at)mark(dot)mielke(dot)cc>
2010-02-26 20:46:27 from Yeb Havinga <yebhavinga(at)gmail(dot)com>
2010-02-26 22:40:06 from Robert Haas <robertmhaas(at)gmail(dot)com>
2010-02-27 00:03:06 from Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
2010-02-27 00:50:11 from Mark Mielke <mark(at)mark(dot)mielke(dot)cc>
2010-02-28 01:01:43 from Robert Haas <robertmhaas(at)gmail(dot)com>
2010-02-28 04:22:00 from Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
2010-03-01 02:47:38 from Robert Haas <robertmhaas(at)gmail(dot)com>
2010-02-26 18:53:38 from Mark Mielke <mark(at)mark(dot)mielke(dot)cc>
2010-02-26 18:59:22 from Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
2010-02-26 19:22:37 from Mark Mielke <mark(at)mark(dot)mielke(dot)cc>
2010-02-26 19:57:56 from Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
2010-02-26 20:26:05 from Mark Mielke <mark(at)mark(dot)mielke(dot)cc>
2010-02-26 20:47:55 from Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
2010-02-26 19:23:22 from Mark Mielke <mark(at)mark(dot)mielke(dot)cc>
2010-03-01 11:18:01 from Yeb Havinga <yebhavinga(at)gmail(dot)com>
2010-02-28 04:20:38 from Craig Ringer <craig(at)postnewspapers(dot)com(dot)au>
2010-02-28 07:52:55 from Mark Mielke <mark(at)mark(dot)mielke(dot)cc>
2010-03-01 02:51:14 from Robert Haas <robertmhaas(at)gmail(dot)com>
2010-03-02 23:54:28 from Bruce Momjian <bruce(at)momjian(dot)us>
2010-03-03 05:13:46 from Robert Haas <robertmhaas(at)gmail(dot)com>
2010-02-28 08:18:26 from Dimitri Fontaine <dfontaine(at)hi-media(dot)com>
2010-03-02 23:53:56 from Bruce Momjian <bruce(at)momjian(dot)us>
2010-02-18 14:58:25 from "Pierre C" <lists(at)peufeu(dot)com>
2010-02-18 15:09:42 from Dimitri Fontaine <dfontaine(at)hi-media(dot)com>
2010-02-18 16:47:01 from "Pierre C" <lists(at)peufeu(dot)com>
2010-02-18 16:54:31 from Dimitri Fontaine <dfontaine(at)hi-media(dot)com>
2010-02-18 20:19:21 from "Pierre C" <lists(at)peufeu(dot)com>
2010-02-19 02:31:05 from David Christensen <david(at)endpoint(dot)com>
2010-02-19 13:57:51 from Kenneth Marshall <ktm(at)rice(dot)edu>
2010-02-19 17:56:12 from Dimitri Fontaine <dfontaine(at)hi-media(dot)com>
2010-02-11 13:41:02 from Yeb Havinga <yebhavinga(at)gmail(dot)com>
Lists:
pgsql-hackers
Robert Haas wrote:
> On Wed, Feb 17, 2010 at 5:52 PM, Jeroen Vermeulen <jtv(at)xs4all(dot)nl> wrote:
>> I may have cut this out of my original email for brevity... my impression is
>> that the planner's estimate is likely to err on the side of scalability, not
>> best-case response time; and that this is more likely to happen than an
>> optimistic plan going bad at runtime.
>
> Interestingly, most of the mistakes that I have seen are in the
> opposite direction.
I guess there's not much we can do about those, except decide after
running that it's worth optimizing for specific values.
>> Yeb points out a devil in the details though: the cost estimate is unitless.
>> We'd have to have some orders-of-magnitude notion of how the estimates fit
>> into the picture of real performance.
>
> I'm not sure to what extent you can assume that the cost is
> proportional to the execution time. I seem to remember someone
> (Peter?) arguing that they're not related by any fixed ratio, partly
> because things like page costs vs. cpu costs didn't match physical
> reality, and that in fact some attempts to gather better empirically
> better values for things like random_page_cost and seq_page_cost
> actually ended up making the plans worse rather than better. It would
> be nice to see some research in this area...
Getting representative workloads and machine configurations may make
that hard. :/
But all we really want is a check for really obscene costs, as an extra
stopgap so we don't have to wait for the thing to execute before we
decide it's too costly. Surely there must be some line we can draw.
Jeroen
In response to
Responses
pgsql-hackers by date
Next :From: Bruce MomjianDate: 2010-02-26 02:49:10Subject : Re: visibility maps and heap_prunePrevious :From : Bruce MomjianDate : 2010-02-26 02:43:44Subject : Re: Why isn't stats_temp_directory automatically
created?
