User-Agent: Microsoft-Entourage/11.2.1.051004
Date: Wed, 23 Nov 2005 17:50:57 -0800
Subject: Re: Hardware/OS recommendations for large databases (
From: "Luke Lonergan" <llonergan@greenplum.com>
To: stange@rentec.com
cc: "Postgresql Performance" <pgsql-performance@postgresql.org>
Message-ID: <BFAA5C81.145DB%llonergan@greenplum.com>
Thread-Topic: [PERFORM] Hardware/OS recommendations for large databases
 (
Thread-Index: AcXweV01iY4lbERPReOGq+8vxhxRgAAICXgV
In-Reply-To: <4384E685.4060108@rentec.com>
MIME-Version: 1.0
Content-Type: text/plain;
 charset=us-ascii
Content-Transfer-Encoding: 7bit

Alan,

On 11/23/05 2:00 PM, "Alan Stange" <stange@rentec.com> wrote:

> Luke Lonergan wrote:
>> Why not contribute something - put up proof of your stated 8KB versus
>> 32KB page size improvement.
> 
> I did observe that 32KB block sizes were a significant win "for our
> usage patterns".   It might be a win for any of the following reasons:
> (* big snip *)

Though all of what you relate is interesting, it seems irrelevant to your
earlier statement here:

>> Alan Stange <stange@rentec.com> writes:
>> If your goal is sequential IO, then one must use larger block sizes.
>> No one would use 8KB IO for achieving high sequential IO rates.   Simply
>> put, read() is about the slowest way to get 8KB of data.     Switching
>> to 32KB blocks reduces all the system call overhead by a large margin.
>> Larger blocks would be better still, up to the stripe size of your
>> mirror.   (Of course, you're using a mirror and not raid5 if you care
>> about performance.)

And I am interested in seeing if your statement is correct.  Do you have any
proof of this to share?

- Luke