ADSM-L

Re: Using FILE instead of DISK devclass to avoid disk under-utilization

2006-10-26 12:54:03
Subject: Re: Using FILE instead of DISK devclass to avoid disk under-utilization
From: Daniel Clark <dclark AT POBOX DOT COM>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Thu, 26 Oct 2006 12:37:42 -0400
On 10/26/06, Mark Stapleton <mark.s AT evolvingsol DOT com> wrote:
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] On Behalf Of
Bos, Karel
>For normal backup data I like to use JBOD config. No read protection at
>all and maximum usable GB per disk. In order to minimize the number of
>storage pools needed, ITSM 5.3 has the collecation by group option.

There's a certain amount of sense there. What is the purpose of fault
tolerance in a TSM disk storage pool? At the end of a process cycle, all
customer data exists in at least three or four places:
Is having fault tolerance for the disk pool really that necessary? Is
yet a fourth (or fifth) level of redundancy worth it?

I just never found it all that important.

You make an excellent point; I guess my reason for using RAID5 is
based more on "I don't want to deal with any more TSM failure
scenarios than I have to" than a logical reasoning process. I suppose
I should just RTFM and write up a procedure on how to deal with DISK
volume failures. Changing to JBOD from RAID5 may allow me to add the
extra devclass I need without breaking the nightly backup window.

When you use JBOD, do you have one (or more) TSM DISK volume(s) per
physical volume, or do you spread them around, so a single TSM DISK
volume is on multiple physical disks?

From a failure management perspective I would thing the former would
make more sense, but depending on what algorithms TSM uses to disperse
activity amongst DISK devclass volumes, that later may have
significantly better performance characteristics (using multiple
spindles etc).

Thanks,
--
Daniel Joseph Barnhart Clark
http://www.pobox.com/users/dclark