ADSM-L

Re: File Device Class and File System Fragmentation with JFS2

2005-12-08 09:47:30
Subject: Re: File Device Class and File System Fragmentation with JFS2
From: "Allen S. Rout" <asr AT UFL DOT EDU>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Thu, 8 Dec 2005 09:47:23 -0500
==> On Wed, 7 Dec 2005 13:38:09 -0600, Andrew Carlson <andyc AT ANDYC.CARENET 
DOT ORG> said:


> We are about to embark on a disk only file device class onsite pool.  I came
> across the recommendation, from Dave Canon's presentation at Oxford
> (http://tsm-symposium.oucs.ox.ac.uk/contributions.html presentation
> Understanding Disk Storage in TSM).  It seems to indicate that you should
> not use scratch volumes - that you should predefine all volumes in the
> storage pool.  I was wondering if anyone has tested this with JFS2, and if
> it is still a problem.  It seems like a waste of time to predefine all those
> volumes, and it would impact what size volume I would be defining.


I would think that the importance of contiguous placement would be very
strongly correlated with the disk tech.  If I go to FILE devclasses on my SSA,
I would think preallocating would be indicated.  If I were deplying on
something more abstracted (shark, netapp, etc) I would hope that the
inefficient placement would be swamped by the cache.

Would your predefined vols be smaller or larger than your scratch vols?  I'm
very interested in your thinking here.


- Allen S. Rout

<Prev in Thread] Current Thread [Next in Thread>