Date: Thu, 28 Mar 2013 16:20:59 -0500
From: "ktm@rice.edu" <ktm@rice.edu>
To: Kevin Grittner <kgrittn@ymail.com>
Cc: kelphet xiong <kelphet@gmail.com>,
	"pgsql-performance@postgresql.org" <pgsql-performance@postgresql.org>
Subject: Re: Question about postmaster's CPU usage
Message-ID: <20130328212059.GP32580@aart.rice.edu>
References: 
 <CANnt-4YWdFj5R0J6BkrL1UivU5JSn5-E+3PAyAxY1t0zMV_tRw@mail.gmail.com>
	<1364504622.71422.YahooMailNeo@web162902.mail.bf1.yahoo.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
In-Reply-To: <1364504622.71422.YahooMailNeo@web162902.mail.bf1.yahoo.com>
User-Agent: Mutt/1.5.20 (2009-12-10)
Content-Transfer-Encoding: quoted-printable
Precedence: bulk
Sender: pgsql-performance-owner@postgresql.org

On Thu, Mar 28, 2013 at 02:03:42PM -0700, Kevin Grittner wrote:
> kelphet xiong <kelphet@gmail.com> wrote:
>=20
> > When I use postgres and issue a simple sequential scan for a
> > table inventory using query "select * from inventory;", I can see
> > from "top" that postmaster is using 100% CPU, which limits the
> > query execution time. My question is that, why CPU is the
> > bottleneck here and what is postmaster doing? Is there any way to
> > improve the performance? Thanks!
>=20
> > explain analyze select * from inventory;
> >=20
> > Seq Scan on inventory=A0 (cost=3D0.00..180937.00 rows=3D11745000 width=
=3D16) (actual time=3D0.005..1030.403 rows=3D11745000 loops=3D1)
> >=A0 Total runtime: 1750.889 ms
>=20
> So it is reading and returning 11.7 million rows in about 1 second,
> or about 88 nanoseconds (billionths of a second) per row.=A0 You
> can't be waiting for a hard drive for many of those reads, or it
> would take a lot longer, so the bottleneck is the CPU pushing the
> data around in RAM.=A0 I'm not sure why 100% CPU usage would surprise
> you.=A0 Are you wondering why the CPU works on the query straight
> through until it is done, rather than taking a break periodically
> and letting the unfinished work sit there?
>=20
> --
> Kevin Grittner
> EnterpriseDB: http://www.enterprisedb.com
> The Enterprise PostgreSQL Company
>=20

Alternatively, purchase a faster CPU if CPU is the bottleneck as it
is in this case or partition the work into parallel queuries that can
each use a processor.

Regards,
Ken


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