Re: prepared query performs much worse than regular query

From: Rosser Schwarz <rosser(dot)schwarz(at)gmail(dot)com>
To: Richard Yen <dba(at)richyen(dot)com>
Cc: pgsql-performance(at)postgresql(dot)org
Subject: Re: prepared query performs much worse than regular query
Date: 2010-05-22 00:30:23
Message-ID: AANLkTinxV_VP49IhTDOx3UThC7m_dqpUkqyXkNoPGZHB@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

On Fri, May 21, 2010 at 4:53 PM, Richard Yen <dba(at)richyen(dot)com> wrote:
> Any ideas why the query planner chooses a different query plan when using prepared statements?

A prepared plan is the best one the planner can come up with *in
general* for the query in question. If the distribution of the values
you're querying against -- in your case, "owner" and "assignment" --
aren't relatively uniform, that plan is going to be suboptimal, if not
downright pathological, for the more outlying-ly distributed values.

Looking at your prepared plan, it seems that, on average, there are
177 rows for every "assignment", and 184 per "owner". As it turns
out, though, nearly a quarter of your table has an "owner" of -1.
It's not terribly surprising, with a table that big and a distribution
skew of that magnitude, that this query plan, with these arguments,
ends up pretty firmly in the "pathological" category.

rls

--
:wq

In response to

Browse pgsql-performance by date

  From Date Subject
Next Message Matthew Wakeling 2010-05-22 03:26:50 Re: prepared query performs much worse than regular query
Previous Message Richard Yen 2010-05-21 22:53:41 prepared query performs much worse than regular query