John -

In a logging-files environment such as you cite, where the data is
largely write-only and historic, a hierarchical storage approach would
be a more reasonable thing, where data over like a week old would
migrate to a cheaper, lower level mass storage area whose entirety
would not have to be regularly scanned for backup.  (It's easy to
incite an Incremental backup on just the data identified by the
migration task.)  Recent data would be held in a higher level area of
much smaller size, whose performance would meet the needs of the
application and be much more reasonable to scan for backup.

We TSM administrators often end up the victims of data architectures
which weren't thought out for all aspects of their management (in our
case, Backup and Restore), and we aren't in a position to re-engineer
the layout.  If the new data is in some way identifiable by name or
timestamp in the directory name, or by identification in some
application logging, it might be possible to focus Incremental backups
on that subset of the file system, rather than Incremental over the
whole thing.  Beyond that, Image or CDP backups may be worth pursuing
further.  Also have the organization consider whether just mirroring
will meet recovery needs: in some implementations, conventional file
backups may not be necessary.

  Richard Sims