pg.ddx.io pgsql-performance@postgresql.org mailing list archive
help / color / mirror / Atom feedpostgresql geqo optimization
2+ messages / 2 participants
[nested] [flat]
* postgresql geqo optimization
@ 2006-02-11 01:46 uwcssa <uwcssa@gmail.com>
0 siblings, 1 reply; 2+ messages in thread
From: uwcssa @ 2006-02-11 01:46 UTC (permalink / raw)
To: pgsql-performance
I have a question with regard to GEQO optimizer of Postgresql.
For complex queries with over 12 tables in a join, (12 is the
default value), the Postgresql optimizer by default will not use the dynamic
programming style optimizer. Instead, it uses genetic algorithm to compute a
sub-optimal query plan. The reason is that GEQO takes sub-seconds to find a
query plan while the DP style optimizer will take minutes or even hours to
optimize a complex query with large join degree.
I am wondering if anyone here ever had complex queries that the GEQO fails
to work properly, i.e., finds a terrible query plan as compared to one
found by DP optimizer (by forcing Postgresql always uses DP). This is
important to me since I am trying to see what type of queries will be worth
spending a lot of time doing a thorough DP optimization (if it is going to
be executed again and again).
thanks a lot!
^ permalink raw reply [nested|flat] 2+ messages in thread
* Re: postgresql geqo optimization
@ 2006-02-11 01:52 Steinar H. Gunderson <sgunderson@bigfoot.com>
parent: uwcssa <uwcssa@gmail.com>
0 siblings, 0 replies; 2+ messages in thread
From: Steinar H. Gunderson @ 2006-02-11 01:52 UTC (permalink / raw)
To: uwcssa <uwcssa@gmail.com>; +Cc: pgsql-performance
On Fri, Feb 10, 2006 at 08:46:14PM -0500, uwcssa wrote:
> I am wondering if anyone here ever had complex queries that the GEQO fails
> to work properly, i.e., finds a terrible query plan as compared to one
> found by DP optimizer (by forcing Postgresql always uses DP). This is
> important to me since I am trying to see what type of queries will be worth
> spending a lot of time doing a thorough DP optimization (if it is going to
> be executed again and again).
There have been a few problems earlier on this list which might have been the
geqo's fault; search the list archives for "geqo" or "genetic", and you
should be able to find them quite easily.
/* Steinar */
--
Homepage: http://www.sesse.net/
^ permalink raw reply [nested|flat] 2+ messages in thread
end of thread, other threads:[~2006-02-11 01:52 UTC | newest]
Thread overview: 2+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2006-02-11 01:46 postgresql geqo optimization uwcssa <uwcssa@gmail.com>
2006-02-11 01:52 ` Steinar H. Gunderson <sgunderson@bigfoot.com>
This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox