Subject: Re: Disk scheduling policy (Re: NEW_BUFQ_STRATEGY)
To: Jason Thorpe <thorpej@wasabisystems.com>
From: Matt Thomas <matt@3am-software.com>
List: tech-perform
Date: 12/01/2003 15:47:44
At 03:16 PM 12/1/2003, Jason Thorpe wrote:
>-----BEGIN PGP SIGNED MESSAGE-----
>Hash: SHA1
>
>
>On Dec 1, 2003, at 3:11 PM, Thor Lancelot Simon wrote:
>
>>Matt Thomas suggested simply starting at whichever end of the other
>>queue is nearer the current head position, which is an interesting
>>idea.  I was pondering zoning the disk, so you'd have multiple "queues"
>>on each side of the "delayed"/"normal" divide, and could try to switch
>>to the one nearest the block you're currently at, but that adds
>>significant complexity.  I'm going to implement a couple of different
>>things and measure them for different workloads.
>
>Presumably this would imply recording the block # of the last request that 
>was issued on the disk?
>
>I can see a danger in this kind of algorithm... if you were to always 
>prefer issuing requests for "near by" blocks, then you could get into a 
>situation where an app that performs many transactions to a localized 
>region could starve other apps whose data is "somewhere else".

My assumption was that you *always* drain the current queue before
switching to other queue.  In that scenario, there is no starvation.


-- 
Matt Thomas                     email: matt@3am-software.com
3am Software Foundry              www: http://3am-software.com/bio/matt/
Cupertino, CA              disclaimer: I avow all knowledge of this message.