User-Agent: Microsoft-Entourage/11.2.1.051004
Date: Tue, 01 Nov 2005 15:49:14 +0000
Subject: Re: 8.1beta3 performance
From: Dave Page <dpage@vale-housing.co.uk>
To: Jon Brisbin <jon.brisbin@npcinternational.com>,
	<pgsql-performance@postgresql.org>
Message-ID: <BF8D3EFA.49BF%dpage@vale-housing.co.uk>
Thread-Topic: [PERFORM] 8.1beta3 performance
Thread-Index: AcXe+87xDZdLP0rvEdqyyQARJHpWaA==
In-Reply-To: <20051101085037.1b12b077@fc63r41.npci.com>
Mime-version: 1.0
Content-type: text/plain;
	charset="US-ASCII"
Content-transfer-encoding: 7bit


On 1/11/05 2:50 pm, "Jon Brisbin" <jon.brisbin@npcinternational.com> wrote:

> On Mon, 31 Oct 2005 17:16:46 -0600
> "PostgreSQL" <martin@portant.com> wrote:
> 
>> We're running 8.1beta3 on one server and are having ridiculous
>> performance issues.  This is a 2 cpu Opteron box and both processors
>> are staying at 98 or 99% utilization processing not-that-complex
>> queries.  Prior to the upgrade, our I/O wait time was about 60% and
>> cpu utilization rarely got very high, now I/O wait time is at or near
>> zero.
>> 
>> I'm planning to go back to 8.0 tonight or tomorrow night but thought
>> I'd check the pqsql-performance prophets before I gave it up.
> 
> I have a stock FreeBSD 5.4 box that I put 8.1 on last night. I ran
> pgbench against it and my tps dropped from ~300tps in 8.0.3 to 20tps
> in 8.1. That's right. 20. No changes in any system configuration. No
> data in the new 8.1 database, only the pgbench init'ed stuff. 25
> clients, 100 and 1000 transactions with a scaling factor of 10, which
> gives me 1,000,000 tuples to shoot through.
> 
> I wiped out the 8.1 installation, put 8.0.4 in it's place, and
> pgbenched it again. ~300tps again.
> 
> It's not a problem with system configuration if 8.0 works fine, but 8.1
> has problems, unless there is something that 8.1 needs tweaked that 8.0
> doesn't. In that case, I just need to know what that is and I can tweak
> it.

Hi Jon,

Did you run the bundled version of pgbench against it's own installation?
There we some changes to pgbench for 8.1, and I have to wonder (bearing in
mind I haven't really looked at them) whether they could be affecting things
in any way. Do you get comparable results running the 8.0 pgbench against
both server versions?

Regards, Dave