Zlatko,

I can appreciate your imaginary situation;  my comments were directed at the
typical scenario for incremental "backup" data, which is easily recovered on
the next night's cycle.  This is another reason there is no single, correct
answer to the question.

Every situation requires one to apply their analytical skills in concert
with TSM capabilities;  for my customers, the large db backups (and
mission-critical, dual-copied archive/redo logs) go straight to tape... as
for archive storage pools, either go straight to tape, or never use the
delete option (I would not advise using the delete option -- there are just
too many ways the process can fail, then you are without recourse!)  This is
a good point for considering RAID-0 (or just simple, non-RAID disk) for
backups;  for mission-critical data like redo logs, consider using RAID-1
(vs. RAID-5) for performance *and* protection.

Your point about how many logical volumes -- using just ONE logical volume
can cause major performance bottlenecks in TSM;  the disk queue for parallel
writes is done on a logical volume basis (hence my reference to "mount
point" wait -- it's really a delay);  it's been a long-time ROT (Rule of
Thumb) to allocate as many logical volumes as one wants parallel sessions...
per Wanda's original comments.  (Unless things have changed, which has not
been indicated in the latest performance info shared by developers.)

Regards,
Don


Don France
Technical Architect -- Tivoli Certified Consultant
San Jose, Ca
(408) 257-3037
mailto:don_france AT att DOT net

Professional Association of Contract Employees
(P.A.C.E. -- www.pacepros.com)