On Feb 2, 2006, at 2:00 PM, Pugliese, Edward wrote:

...It seems to me that
this is mostly a function of a large number of small files. ...

The classic test is to find or fabricate a large file and back up
just that one object, and contrast throughput with another test
containing the same amount of data but in much smaller files. This is
the Black Box approach, where behavior to varying conditions can be
measured without knowing anything about internals.

IBM provides a variety of TSM tuning guides through the TSM Support
Web page, including the standard TSM Performance Tuning Guide. Don't
overlook other sources, such as the STE seminars, Oxford
presentation, and SHARE presentations. I also collect tips in ADSM
QuickFacts, also largely found in the TSM Wiki. And there are all the
OS and networking things to wade through. There are a lot of factors
to weigh, but it keeps us fully employed.

  Richard Sims