Message-ID: <4D902809.4020304@2ndQuadrant.com>
Date: Mon, 28 Mar 2011 02:17:45 -0400
From: Greg Smith <greg@2ndQuadrant.com>
User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US;
	rv:1.9.1.16) Gecko/20110307 Icedove/3.0.11
MIME-Version: 1.0
To: pgsql-performance@postgresql.org
Subject: Re: Xeon twice the performance of opteron
References: <AAC806D7-56CD-4BC3-A449-83771F818A4B@torgo.978.org>
In-Reply-To: <AAC806D7-56CD-4BC3-A449-83771F818A4B@torgo.978.org>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

On 03/17/2011 11:13 AM, Jeff wrote:
> three boxes:
>     A: Intel(R) Xeon(R) CPU           E5345  @ 2.33GHz  (Runs query 
> fastest)
>         4MB cache
>     B: Quad-Core AMD Opteron(tm) Processor 2352 (2.1GHZ) (Main 
> production box, currently, middle speed)
>         512k cache
>     C: Quad-Core AMD Opteron(tm) Processor 2378 (2.4GHZ)
>         512k cache

It's possible that transfer speed between the CPU and memory are very 
different between these systems when running a single-core operation.  
Intel often has an advantage there; I don't have any figures on this 
generation of processors to know for sure though.  If you can get some 
idle time to run my stream-scaling tool from 
https://github.com/gregs1104/stream-scaling that might give you some 
insight.

> Now here's where some odd stuff starts piling up: explain analyze 
> overhead on said queries:
> 20ms on A, 50ms on B and 85ms on C(!!)

I found an example in my book where EXPLAIN ANALYZE took a trivial 
COUNT(*) query from 8ms to 70ms.  It's really not cheap for some sorts 
of things.

> I know we're running an old kernel, I'm tempted to upgrade to see what 
> will happen, but at the same time I'm afraid it'll upgrade to a kernel 
> with a broken [insert major subsystem here] which has happened before.

Running a production server on Fedora Core is a scary operation pretty 
much all the time.  That said, I wouldn't consider 2.6.27 to be an old 
kernel--not when RHEL5 is still using 2.6.18.  The kernel version you 
get for FC10 is probably quite behind on updates, though, relative to a 
kernel.org one that has kept getting bug fixes.

-- 
Greg Smith   2ndQuadrant US    greg@2ndQuadrant.com   Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support  www.2ndQuadrant.us
"PostgreSQL 9.0 High Performance": http://www.2ndQuadrant.com/books