Another factor to keep in mind in general regarding TSM server
performance...

Easily overlooked is the impact of a large volumehistory inventory,
where its management in the face of substantial volume transitions can
induce appreciable overhead. This is particularly true as the
volumehistory backup file is completely rewritten in the OS file system
each time a volume transition occurs. See IBM site Technote 1177559,
which discusses this.

    Richard Sims      http://people.bu.edu/rbs