Message-Id: <25929.337892@fatcity.com>
From: Nuno Pinto do Souto <nsouto@optusnet.com.au>
Date: Wed, 16 Jul 2003 07:30:05 +1000
Subject: Re: Re: Re: should you seperate indexes from tables in seperate

> why is it useful to seperate different i/o pattersn? such as
> multi-block reads and single block reads? 

Because (assuming adaptive I/O subsystems like EMC's and 
such) each gets optimized into different priority scales and 
device queues.  And of course if they are to the same device, they 
*may* interfere with each other.

Like I said: highly dependent on your I/O hardware.  Doesn't hurt to 
set it up so you can take advantage of any future expansion/upgrade.  
IOW, separate indexes and tables into individual tablespaces, not 
because of I/O performance considerations but because of the 
flexibility it gives you in adapting to future change.

The core of the OFA paper, really.  But I'll let Cary 
speak for that: it's his baby.

Cheers
Nuno Souto
nsouto@optusnet.com.au