A collocated tape pool with a limited number of scratch volumes might be
just the ticket.  What I have done is to create a collocated tape storage
pool with a fixed number of volumes, for instance 50 volumes.  This works
well for sites with tapes of large capacity such as DLT or MAGSTAR.  What
will happen is the first 50 filespaces or the first 50 nodes will each use
a tape volume.  After the first 50 tapes are used up the rest of the data
will be mingled with data from other nodes or filespaces, but collocation
will be maintained.  That is, if two nodes or two filespaces share a tape
those two nodes and filespaces will continue to share that tape until
another tape is needed.  When there is a tape reclamation process run
against these tapes the collocation will be maintained to the extent
possible.  In the long run this will result in the fewest number of tapes
possible perhaps a couple, and not have the down side effect as Andy
mentioned whereby you have a lot of partially filled tapes.

The bottom line in this discussion is related to the restore time required
for the data that you have in your environment.  If you have a very short
restore time requirements then you better consider collocation.  If you do
not have short restore time requirements, perhaps collocation does not buy
you sufficient advantage for the costs incurred.  As Andy mentioned there
is no free lunch in ADSM.

Finally, how often are we really required to restore the entire volume?
  Most restores are for a single file.  You don't necessarily want to
design a solution for something that only happens once in a blue moon.

Kelly
(dictated using IBM ViaVoice)