Re: Questions on query planner, join types, and work_mem - Mailing list pgsql-performance

From Tom Lane
Subject Re: Questions on query planner, join types, and work_mem
Date
Msg-id 7124.1280804145@sss.pgh.pa.us
Whole thread Raw
In response to Re: Questions on query planner, join types, and work_mem  (Peter Hussey <peter@labkey.com>)
List pgsql-performance
Peter Hussey <peter@labkey.com> writes:
> My questions are still
> 1)  Does the planner have any component of cost calculations based on the
> size of work_mem,

Sure.

> and if so why do those calculations  seem to have so
> little effect here?

Since you haven't provided sufficient information to let someone else
reproduce what you're seeing, it's pretty hard to say.  It might have
something to do with the particularly wide join key values you're using,
but that's mere speculation based on the one tidbit you provided.  There
might be some other effect altogether that's making it do the wrong thing.

> 2) Why is the setting of work_mem something left to the admin and/or
> developer?

Because we're not smart enough to find a way to avoid that.

            regards, tom lane

pgsql-performance by date:

Previous
From: Robert Haas
Date:
Subject: Re: Questions on query planner, join types, and work_mem
Next
From: Yeb Havinga
Date:
Subject: Re: Testing Sandforce SSD