ADSM-L

Re: Storage Pool Hierarchy

2004-09-01 03:57:13
Subject: Re: Storage Pool Hierarchy
From: Steven Pemberton <stevep AT IBK.COM DOT AU>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Wed, 1 Sep 2004 17:30:58 +1000
Joni,

How you configure your storage pools is going to depend on the performance and
efficiency that you require, from both on-site and off-site restores.

I like to think that the Policy/Management classes service the "retention"
part of a backup SLA, while the Storage Pool/Volumes used service the
"performance" requirements.

I'd tend to create as few Policy Domains as possible, probably combining the
different Operating Systems. But I would create multiple Policy Domains (to
allow multiple Default Management Classes) to direct the backup data into
specific Storage Pools designed to meet my performance requirements.

The number, and configuration, of the Storage Pools you create depends on the
type of client and their recovery performance needs. Firstly, I'd usually
separate the filesystem  data from the TDP data. I'd probably bundle all the
TDP backups into one Storage Pool, and further separate the filesystem
backups into two or three Storage Pools. The critical (for recovery)
filesystem clients I'd put in their own, smaller pool, while the bulk of the
non-critical nodes into another one or two pools.

And if you need good off-site recovery performance, then you should consider
creating a similar offsite copypool design.

I mentioned this briefly in a talk I gave at the TSM symposium last year:
http://adsm-symposium.oucs.ox.ac.uk/papers/StevenPemberton.pdf

One day, when "group collocation" arrives this may not be necessary, but in
the meantime it's easy to do and performs well.

Regards,
Steven P.

On Tuesday 31 August 2004 01:03, Joni Moyer wrote:
> Hello All!
>
> I know that this is a common sense question concerning storage pools, but I
> was just hoping that there is some type of logical answer.  In the
> beginning I was going to create 5 domains by platform and have a disk pool,
> onsite tape pool and offsite copy pool for each.  Now I am beginning to
> wonder if I don't want to get even more granular than that.  Due to the
> fact that we cannot afford to have our copy pools collocated, but we do
> need the data stored in a less fragmented format.  My question is:  does it
> make sense to create more domains or do I just create multiple storage pool
> hierarchies?  For example, in a UNIX environment should I create 3
> different UNIX domains for Solaris Applications, AIX Notes and AIX
> Applications or do I just create 3 different storage pool hierarchies?
> Also, if I choose this route than how do I specify a default management
> class when I would want 3 different ones?  Maybe I am answering my own
> question here, but any opinions would be appreciated. Thanks!
>
> ********************************
> Joni Moyer
> Highmark
> Storage Systems
> Work:(717)302-6603
> Fax:(717)302-5974
> joni.moyer AT highmark DOT com
> ********************************

--
Steven Pemberton

Mobile: +61/0 418 335 136 | Home: +61/3 9820 5811
Melbourne, Victoria, Australia
http://members.iinet.net.au/~spemberton

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