RE: Collocation.  As I understand it,  every user gets their own tape until
the tapes run out and then the least used tape is doubled up with the
next user needing a new tape.

I try to get the best of both worlds.  I have collocation ON and I keep the
number of tapes in the pool restricted.  I add
a scratch tape to the pool *only* if I'm registering a new user with
data to backup that is about 3/4 of a tape.  With our use of 3490s, currently,
that means 3/4 of 1.2 Gigs.

Some of my users are 200MB Mac users.  I register a tape for every 5 of them.
With about 55 *active* users I'm experiencing about 8 mounts on days when
the STG hits the migration HI mark.   (We have about 100 registered.)

I am also diverting new users with large backups (500M) through a different
STG by initially registering them in a 'pre-incremental' domain.

My STG disks are 2.5G for incremental's and 1.5G for'new user-first full backup'
They both migrate to the same tape pool.
This seems to be working very nicely, but I'm not sure it can scale-up.

I agree with Paul. If you could codify what he & I have said into the migration
logic then it would certainly work a lot better.    Paul says:  give me a knob
to turn.   I say:  build some *more* smarts into the routine and give me a knob

The logic might be:

When migrating a user ask the question:  is the OCC of this
user high enough to merit its own tape?    Where merit is defined as 3/4 tape.

.. joe.f.