We've had a similar remarkable loss of throughput, but between the
corporate IBM mainframe and our HP 3000/996-800.  Last month-end saw a
10-to-15-fold increase in transmission times for a few large files (10-15
hours *each*, vs. 1-1.5 hours each).  Apparently, this occurs only when the
transmission is served by a new 100MB OSA TCP/IP adapter in the mainframe
-- apparently, the originator of the transfer has no bearing on the
problem.  The claim is that this interface was in place during the
month-end close prior to last month-end, and all worked fine; ergo,
something *must* have changed on the HP side.  We've loaded no patches, no
new releases, and made no network configuration changes for several months.
A network analyst in our corp. office contended that he could 'see' packets
arriving at the HP, but nothing being returned; yet, a link trace,
according to the HPRC, shows acknowledgements leaving the system promptly
in response to each packet from the mainframe.

Has anyone else experienced this kind of degradation in a similar
environment?  If so, how did you solve it, if at all?

TIA,

Lee Gunter
============================================================
Supervisor, TRG HP/MPE Systems
The Regence Group
Phone:                   503.375.4498
Fax:                     503.375.4401
E-mail:             [log in to unmask]
============================================================
All opinions expressed are just that -- opinions.  They're
mine and mine alone.