zeleznik@cs.utah.edu (Mike Zeleznik) (06/09/90)
Here's one I thought was really strange...
The following combination of workstation, multiport tranceiver, and
Ethernet MAU do not work well together. However, NO SINGLE unit appears
to be bad; it is only the COMBINATION that has trouble:
Sun Sparcstation
Cabletron MT-800 Multiport Transceiver
HP 30241A MAU Transceiver (SQE always ON)
THE PROBLEM: Network operations on the Sun are very slow. For example, if
the Sun pings other machines, it claims to miss many replies. But a
Sniffer hung on another port of the MT800 sees all requests and replies,
with no collisions. Netstat on the Sun also reports no collisions. The
problem is repeatable on different MT800 ports, with another Sun, and with
another MT800. But, OTHER MACHINES on the MT800 have NO PROBLEMS.
Okay, here's what we saw.
This is the configuration that has the problem:
Sun ---------- MT800 ---------- HP MAU <no good>
Remove the MT800, and the problem goes away:
Sun --------------------------- HP MAU <okay>
Or, replace the MT800 with a DEC or TCL, and the problem goes away:
Sun --- DEC DELNI Multiport --- HP MAU <okay>
Sun --- TCL Multiport --------- HP MAU <okay>
Ah, so you think it is the MT800. Well, you can leave the MT800 in,
but replace the HP MAU, and the problem also goes away:
Sun ---------- MT800 ---------- Cabletron ST-500 MAU <okay>
(this works with SQE either ON or OFF on the ST-500)
Or, you can leave the MT800 and HP MAU *both* in, but stick the
DEC DELNI between them, and the problem also goes away:
Sun ------------ MT800 -- DEC DELNI -- HP MAU <okay>
But, putting the DEC DELNI on the other side DOES NOT fix it:
Sun --- DELNI -- MT800 --------------- HP MAU <no good>
It is like the HP MAU is generating a signal that the MT800 exacerbates
such that the Sun finally does not like it (most of the time). As for the
DELNI, it is like it can somehow clean up the signal enough if put in front
of the MT800, but can not fix it after both the HP MAU and the MT800 are
done with it. I did not have time to look at the actual signals.
Any ideas? In any case, thought someone might appreciate this one.
Thanks,
Mike
Michael Zeleznik Computer Science Dept.
University of Utah
zeleznik@cs.utah.edu Salt Lake City, UT 84112
(01) 581-5617