Recently, the AIX 5.1 host for our busiest TSM 5.2.3.0 service,
has been running at 100% CPU for most of the night, some mornings
and even into some afternoons.  We have discovered a few of our
clients (using inappropriate network settings, typically Auto-Dectect
instead of matching the network switch...) have poor performing,
very long running backup sessions that are exhausting the CPU.

TSM provides two server options:
    ThroughputTimeThreshold = MMM [seconds]
    ThroughputDataThreshold = NNN [KB/sec]
to be set causing the TSM service to cancel any client session
that does not perform better than NNN KB/sec after MMM seconds.

Has anyone determined appropriately conservative settings for
these options with TSM client network connections varying from
10Mb,Half-Duplex up to 100Mb,Full-Duplex?

We're aiming to kill the network mis-matched clients and don't want
to suffer with them for too long, but we don't want to kill a client
that might be legitimately slowly progressing through a huge directory.

Please respond if you've already solved this problem satisfactorily
at your site.  Thanks for your help.
--
Jim.Owen AT Yale DOT Edu   (203.432.6693)
Jim.Owen AT Yale DOT Edu   (203.432.6693)