Just a note: I don't want to appear to be just dissing the Win HSM
product; but I -do- want to make sure that nobody confuses it with the
other products TSM has deployed under that name.  It's _different_.


>> On Fri, 20 Jan 2006 09:26:43 +0100, TSM <tsm AT PROFI-AG DOT DE> said:

> another thing to mention: HSM is a product to move "inactive" data
> from the disks, so when there are a lot of changes or retrieves of
> such former hsm migrated files, then this filespace is surely no
> candidate for HSM.


If you'll accept the modification "... no candidate for [ current TSM
Windows ] HSM", then I think we're in complete agreement.


Our largest use of HSM has been for logfiles: Huge directories of
every (say) SMTP or HTTP log from our cluster, dating back years.

Access to the old files is uncommon, but regular, and not time
sensitive.  Once every (say) quarter, we might re-run a "Last 12
months" analysis, or some such.  Once a year we might re run a "Last
four years" analaysis or such.  Worked well for years.  This kind of
use pattern is what I think of when I think "HSM".  You'd be nuts to
try to do this under the windows HSM product.

The largest speculative use we've entertained is related to large
datafiles associated with e.g. physics experiments, with similar usage
patterns: ~10TB of space with relatively rapid access needed, and
hundreds of TB of other data available more slowly from tape, with
access patterns, again, uncommon but regular.

I think these access patterns are right down the middle of mainstream
HSM utilization.


- Allen S. Rout