Re: [HACKERS] nested loops in joins, ambiguous rewrite rules - Mailing list pgsql-hackers

From Bruce Momjian
Subject Re: [HACKERS] nested loops in joins, ambiguous rewrite rules
Date
Msg-id 199901310652.BAA12024@candle.pha.pa.us
Whole thread Raw
In response to Re: [HACKERS] nested loops in joins, ambiguous rewrite rules  (Charles Hornberger <charlie@k4azl.net>)
Responses Re: [HACKERS] nested loops in joins, ambiguous rewrite rules  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
> >We turn on geqo at 8 relations.  Try:  > >    SET GEQO TO 4 >
> >and try the query again.  Let us know.
> 
> Well isn't that something!  Thanks so much for your help!
> 
> I set the GEQO variable to 4 and now the 11.5 minute query
> executes in 6 seconds with this query plan:
> 
> Hash Join  (cost=21.99 size=152 width=124)   width=16)
> 
> 
> Are there any recommendations about what value *ought* to be
> set for GEQO?  It seems to me like 'ON=8' is pretty high -- for
> us, it meant that UNLESS we explicity set that variable for
> every JOIN query of 6-7 tables, the joins were going to bog down
> to a total crawl, while sending memory and CPU consumption
> through the roof (roughly 22MB and 90-95%, respectively, for
> the entire query-processing period).

We have toyed with lowering it to 6.  I think someone said that was too
low, but obviously, some queries need a value of 6 or lower.  I don't
understand when that is the case, though.  Any comments from anyone?

I believe the GEQO README was moved into the documentation in docs.  It
used to be there, but is no longer.


-- Bruce Momjian                        |  http://www.op.net/~candle maillist@candle.pha.pa.us            |  (610)
853-3000+  If your life is a hard drive,     |  830 Blythe Avenue +  Christ can be your backup.        |  Drexel Hill,
Pennsylvania19026
 


pgsql-hackers by date:

Previous
From: Bruce Momjian
Date:
Subject: Re: [HACKERS] Patches
Next
From: Tatsuo Ishii
Date:
Subject: Re: [HACKERS] READ COMMITTED isolevel is implemented ...