Greetings, all.

We've got a major system here at UF we're about to do some upgrading on.  The
folks who are closest to the upgrade are asking some questions about recovery
times in the Worst of Cases.  I've had to tell them I can't really predict how
long the restore would take, in practical terms: we've never had to do a
restore quite this big.  Maybe someone out there has.

So here's our configuration:

TSM server housing 2x 3590e drives, SCSI attached.

-J cartridges in use in the storagepools in question.
Tapes filespace-colocated for the node in question.
Tapes filespace-colocated for the node in question.

Filespaces of concern are basically mail stores, with a single file per
message. 320GB total space, 18 million files as of last incremental. Most
files are ~5k, but the average is about 17, because of those nummy email
attachments beloved by all.

Network attach between the TSM server and the new node would be via SP
switch.

Underlying disk attach would be RAID-5s of drawers of 9G SSA disks, 15 disks
and one hot spare per drawer.

Anybody done anything in this vicinity?  Databases really are a different
problem, unfortunately.

It seems unlikely that we'll have to deal with this problem, the upgrade seems
quite solid.  But it'd be nice to give an estimate more precise than "Probably
more than a day.  Probably less than a week".


Allen S. Rout