CU-SeeMe connect requests overloaded the network

Steve Womble (womble@austin.ibm.com)
Fri, 6 Jan 1995 10:51:16 -0500


While I was gone on vacation during the holidays, we had a problem with
CU-SeeMe. Since I was gone and the people that were affected are not
network literate, details of what happened are sketchy.

Apparently, CU-SeeMe was up and running with 4 users (3 lurchers - 2
PCs, 1 Mac and 2 senders - PCs) sending/receiving to an AIX reflector
when there was a network glitch which cause CU-SeeMe to loose contact
with at least 2 (possibly all) of the clients connected. When the
network got back to normal CU-SeeMe tried to re-establish connection
to the clients (which possibly killed their local CU-SeeMe program),
but couldn't reconnect. During this time CU-SeeMe repeatably kept sending
out requests to open port 7XXX which caused some of the CU-SeeMe clients
to completely hang and overload our 16MB Token Ring network enough to
bring other systems on the ring almost unusable.

Has anyone seen this happen before? It seems to me that CU-SeeMe should
have a timeout value to stop sending requests after 1-2 minutes of
failures.

Thanks in advance.

Steve
------------------------------------------------------------------------
Steven K. Womble womble@austin.ibm.com