==> On Fri, 9 Sep 2005 13:57:56 -0400, "Pugliese, Edward" <EPugliese AT NYBOT 
DOT COM> said:


> Out of curiosity, what is the reason for this requirement ?

I can't speak for the original poster, but I can talk about why I'd use it.

> TSM and an appropriate configured library can be pretty fast at restoring
> files even if the last active version was backed up a long time ago.

We've got two corners of our customer space for which this feature would be
welcome.

One corner is the set of animals who are more equal than other animals. Ahem.
If you can always recover your vice president's deleted MP3s without a delay
for a tape mount, that can lead to more happiness for everyone. :)

The other is the no-kidding critical keystone machines in your
infrastructure. Being able to restore them -immediately-, no matter what else
is going on, could feel very nice, at only nominal cost.



But I've got a bet that this won't be forthcoming soon, and a bet about why.

We tend to think of the basic unit of TSM storage as the 'file'.  TSM really
thinks in 'aggregates', groups of files of which it can keep track all at a
time.  That's why there's an option to MOVE DATA, reconstruct=[yes|no].

This detail would dramatically complicate the disk->serial stgpool behavior;
rather than just checking the time on an aggregate, and then moving it or not,
you'd have to check each individual file for active/inactive status.

Then you'd need to either make new aggregates in the diskpool with the
to-be-moved inactive files [brand new operation inside TSM. Scary] or move
each file (or at best set-of-files-from-one-aggregate) as a separate aggregate
in the destination pool.  Database bloat, big time.

I don't know wether reclamation builds -new- aggregates as it goes.  If it
does, then the database bloat would plateau over time; Volumes would be mostly
indistinguaishable after a reclamation.


So it's not that it's impossible, it's that the feature isn't as simple as it
sounds, so its' cost in TSM developer time is disproportionately large.  There
are probably a laundry list of items we'd in fact prefer out of that time.



I'd love to be wrong, though.


- Allen S. Rout