Brenda -

Thanks for the helpful extra info.

The TSM server messages indicate that it was attempting to send data
to the client - corroborated by the SendW condition you reported.
The DB2 log indicates that app trying to get data, and "problem
reading from vendor device", which resolves to the TSM API conduit: a
-50 I/O condition there indicates session broken.  So, both ends were
trying to communicate, but something in between prevented that.

I would look for dsmerror.log/dsierror.log indications of any problem
in the TSM API middleware.  I'd also inspect the AIX Error Log and /
var/adm/messages logs for any indications of errors at the OS level
in that client system.  During the event you might also perform
'lsof' or equivalent on the involved client processes to see their
socket state.

The symptoms seem like something in the network layer, as though the
faucet were shut because something thought that was enough data
already through that single session (service throttling).  You might
try an artificial FTP of an equivalent amount of data (perhaps to /
dev/null, as outlined in NETWORK PERFORMANCE in ADSM QuickFacts) to
see if that gets cut off in much the same way.

   Richard Sims