With a devclass of file, a storage group must have reclamation run
regularly, in some cases this can be costly. I'm told that devclass disk
will end up fragmenting the storage. This puzzles me. Are TSM file
placement algorithms really that bad for disk? It would seem to me that
devclass file just puts the file placement work off onto the underlying OS.

The reason I ask is that we have a setup of two TSM servers that all our
production servers back up to. Then the two TSM servers back up to TSM
servers half a state away. The way off campus servers are two instances
on the same physical server. All four servers have their storage groups
set up as devclass file. Bandwidth is turning out to be a problem.
Backups from the local to remote servers take more than a day and then
reclamation of the copypools from the local to remote servers never gets
a chance to finish. Result is that the remote servers are full. Wouldn't
setting up the remote storage groups as devclass disk at least eliminate
reclamation? Would performance eventually suffer because of disk
fragmentaion?

Thanks in advance for any insight any of you may offer.