DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=ymail.com;
	h=X-YMail-OSG:Received:X-Rocket-MIMEInfo:X-Mailer:References:Message-ID:Date:From:Reply-To:Subject:To:In-Reply-To:MIME-Version:Content-Type:Content-Transfer-Encoding;
	b=nTd4/55IQjAdeEjRAAZFcdWrIq83LoW+Bm/dG5yi32UneFPXpV+ErOn6dx8RLir95IG+eVC0vS/kV05u25p6nLoGYVEX+ZHOX30BA+mNFsEcwqJFiqwMAeLGjTL1ZjSoodT6723pCVNrlUlqWMCk46SRBn3fQuMsTmz0seUf8qQ=;
References: <515D92BF.5000309@axege.com>
	<CAPmjWd100eemBz1RTw=r24zgh8+YO8iq86fThqsALznrPuyWqg@mail.gmail.com>
	<515D9FC5.7030105@axege.com> <18910.1365092713@sss.pgh.pa.us>
	<515DCB34.6040801@axege.com> <6354.1365102532@sss.pgh.pa.us>
	<515ED7BC.40907@axege.com> <4113.1365171448@sss.pgh.pa.us>
	<515FE547.3010301@axege.com>
Message-ID: <1365261991.56886.YahooMailNeo@web162905.mail.bf1.yahoo.com>
Date: Sat, 6 Apr 2013 08:26:31 -0700 (PDT)
From: Kevin Grittner <kgrittn@ymail.com>
Reply-To: Kevin Grittner <kgrittn@ymail.com>
Subject: Re: What happens between end of explain analyze and end of query
	execution ?
To: Franck Routier <franck.routier@axege.com>,
	"pgsql-performance@postgresql.org" <pgsql-performance@postgresql.org>
In-Reply-To: <515FE547.3010301@axege.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: quoted-printable
Precedence: bulk
Sender: pgsql-performance-owner@postgresql.org

Franck Routier <franck.routier@axege.com> wrote:

> Ok, the problem definitely comes from the
> default_statistics_target which is obviously too high on the
> database.

> Now... can someone help me understand what happens ? Where can I
> look (in pg_stats ?) to see the source of the problem ? maybe a
> column with a huge list of different values the palnner has to
> parse ?

This is a fundamental issue in query planning -- how much work do
you want to do to try to come up with the best plan?=A0 Too little,
and the plan can be unacceptably slow; too much and you spend more
extra time on planning than the improvement in the plan (if any)
saves you.=A0 Reading and processing statistics gets more expensive
as you boost the volume.

What I would suggest is taking the default_statistics_target for
the cluster back down to the default, and selectviely boosting the
statistics target for individual columns as you find plans which
benefit.=A0 Don't set it right at the edge of the tipping point, but
don't automatically jump to 5000 every time either.

--
Kevin Grittner
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company


--=20
Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance