Collocation has nothing to do with it. It's the nature of TSM and the VTS
that causes 'problems.' TSM likes to use all the space on a tape. TSM keeps
writing to a tape until it hits end-of-volume (EOV), then asks for another
tape with space on it or a scratch tape.

In a VTS, when you ask for a virtual volume to be mounted, if the data does
not already reside in the VTS cache/disk, then ALL the data on the volume is
recalled from the 'real' 3590 tape into the cache. If there's not room, then
the VTS will remove existing data from the cache based on a LRU algorithm.
If that data hasn't been copied to the 'real' 3590 volumes,that has to
happen first. So, no instead of mounting a tape and positioning to the end
of the data, you have to read the data from tape into the disk cache BEFORE
you can do anything with it. So,as the amount of data on the volume grows,
the amount of time it takes to perform the tape mount increases.

Now when you append data to the virtual volume, the VTS now has to stage
this data back out to the 3590 tapes...all of it. Not just the 'new' data
you appended. So now you're writing back out up to a full 3490 amount of
data to tape. Where the original data existed on tape is now unusable space.
Just like in TSM where expired data on tapes becomes unusable and you need
to reclaim. This happens in the VTS, too. You have regularly scheduled
reclamation tasks, plus thresholds. So, the more you append data to existing
volumes the more frequent you would need to run the reclamation. Just more
overhead in the VTS.

Plus by having to recall the virtual volume into cache before you can append
to it, or even read from it may cause other data withing the VTS to be
removed from cache. This could cause longer mount times for other
applications/jobs streams. You would want to have a larger amount of disk
cache within the VTS for a system used for TSM.

TSM reclamation would drive this system crazy, too. Without collocation,
think of how many input tape mounts it takes to reclaim your offsite storage
pool(s). Each one of those 'mounts' would cause the VTS to read the data
back into cache before it could be read by TSM. Plus the VTS doesn't know
that there is only maybe 10% usable data on that virtual volume. As far as
he's concerned, it's a full tape.

Maybe for a small TSM system you could use a VTS, but TSM is not the right
application for a VTS. IMHO that is.

Bill