I have lots of policies that have near identical retention times - the
difference in the policies is often superficial, so there's no reason
they can't stream together. (Ex. windows vs. unix, systems that can't
run until a later time etc).

Network is the biggest bottleneck I know in the backup infrastructure
- a media server with 1Gb inbound can really only stream about 1 1/3
of a tape drive ( or 2 drives with inline copy). Because of this,
having systems fall through to the next storage unit when the first
has enough streams is quite useful.

If my first host got all 8 drives, it'd be driving them at 1/8th speed
and i'd be dead in the water as far as completing my backups in the
window - by going across 3 or 4 (lots of inline copies), I get 4x the
available network bandwidth and am not choking for no good reason.

Storage unit groups help to maximize resource utilization in my
environment and drive down backup times.

-Charlie

On 11/17/05, Dale King <dale AT daleking DOT org> wrote:
> Hi All,
>
> A collegue and I in different parts of our organisation were discussing
> the use of storage unit groups.  My philosophy is that you should have a
> single storage unit per media server - robot pair set to use all drives
> for sanity, with max multiplexing set at the highest you would want to go.
> This is easy to manage and I think is the most efficient way to use your
> drives.
>
> The other theory put forward is that you have multiple storage unit groups
> limiting each individual storage unit to one or two drives depending on
> policy requirements.  My collegue says this has been used to resolve
> problems where multiple policies using different volume pools want to run
> to the same storage unit thereby causing some sort of exclusive lock.
> This to me sounds like a bug because we tested it on our AIX 5.1 MP2
> server and could not get it to fail (second policy running with a
> different volume pool would simply cause a new tape mount and jobs would
> keep running.
>
> My collegue says that Veritas recommended storage unit groups to overcome
> these lock out problems on their Solaris 5.1 media servers.  But I can't
> for the life of me see how it would help.
>
> The reasons for using groups that I know of are:
>         - you have multiple robots and prefer one over the other but are
>           happy to use both
>         - you want to load balance the same robot across two or more media
>           servers for a single policy
>         - you have drive contention and want to set some crazy high
>           multiplexing value on your last few drives so that drives don't
>           fail
>
> Are there any more?  Has anyone been told by Veritas to use groups to
> overcome storage unit availability problems due to multiple
> policies/volume pools?
>
> Your comments appreciated.
>
> Thanks,
> Dale
>
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.1 (GNU/Linux)
>
> iD8DBQFDfQGR/dB3EOgAfiQRAsdbAJ9mHrZWYzs2nuR5XK+TmL5sYe2FSgCfcNm1
> 4yNwjbhhcjXPLeWQWFSkJ6U=
> =ysUc
> -----END PGP SIGNATURE-----
>
>
>