Re: Hardware - Mailing list pgsql-general

From Scott Marlowe
Subject Re: Hardware
Date
Msg-id 1170782584.5451.104.camel@state.g2switchworks.com
Whole thread Raw
In response to Re: Hardware  (Lars Heidieker <lars@heidieker.de>)
List pgsql-general
On Tue, 2007-02-06 at 10:33, Lars Heidieker wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
>
> On 6 Feb 2007, at 15:59, Walter Vaughan wrote:
>
> > I need to purchase a new server to put posgresql on that will be
> > acting as the DBMS server for Apache ofBiz soon. While googling
> > around for performance tweaks I saw this at http://revsys.com/
> > writings/postgresql-performance.html
> >
> > <quote>
> > CPUs — The more CPUs the better, however if your database does not
> > use many complex functions your money is best spent on a better
> > disk subsystem. Also, avoid Intel Xeon processors with PostgreSQL
> > as there is a problem with the context switching in these
> > processors that gives sub-par performance. Opterons are generally
> > accepted as being a superior CPU for PostgreSQL databases.
> > </quote>
> >
> > Is this still true in regards to Xeon's? I was looking at a server
> > with Quad Core Xeon 2 5335 @ 2.0GHz.
> >
> > And at http://www.powerpostgresql.com/PerfList
> >
>
> It seems to be outdated. As far as I know it was supposed to warn for
> the HyperThreading things.

It wasn't just hyperthreading.  The older xeons tended to have more CPU
horsepower than cache bandwidth, and this resulted in them going into
swap storms whether hyperthreading was on or not.  It was just easier to
trigger with hyperthreading.

pgsql-general by date:

Previous
From: Richard Huxton
Date:
Subject: Re: Postgres SQL Syntax
Next
From: Tim Tassonis
Date:
Subject: Cursor timeout in postgres