Date: Tue, 23 Aug 2005 19:24:24 -0400
From: Michael Stone <mstone+postgres@mathom.us>
Subject: Re: Read/Write block sizes
In-reply-to: <60k6icpapm.fsf@dba2.int.libertyrms.com>
To: Chris Browne <cbbrowne@acm.org>
Cc: pgsql-performance@postgresql.org
Mail-Followup-To: Chris Browne <cbbrowne@acm.org>,
	pgsql-performance@postgresql.org
Message-id: <20050823232424.GM8667@mathom.us>
MIME-version: 1.0
Content-type: text/plain; charset=us-ascii; format=flowed
Content-disposition: inline
References: <d4b67d7c2b.d7c2bd4b67@bur-mail1.east.sun.com>
	<60k6icpapm.fsf@dba2.int.libertyrms.com>
User-Agent: Mutt/1.5.10i

On Tue, Aug 23, 2005 at 06:09:09PM -0400, Chris Browne wrote:
>What we have been finding, as RAID controllers get smarter, is that it
>is getting increasingly futile to try to attach knobs to 'disk stuff;'
>it is *way* more effective to add a few more spindles to an array than
>it is to fiddle with which disks are to be allocated to what database
>'objects.'

That statement doesn't say anything about trying to maximize performance
to or from a disk array. Yes, controllers are getting smarter--but they
aren't omnicient. IME an I/O bound sequential table scan doesn't get
data moving off the disk nearly as fast as say, a dd with a big ibs.
Why? There's obviously a lot of factors at work, but one of those
factors is that the raid controller can optimize "grab this meg" a lot
more than it can optimize "grab this 8k". 

Mike Stone