Brad Lackey writes:

   The Sun Ray 2's have a MUCH different (and faster) x86 processor in
   them.

And sufficiently different that our Ray 2 can be made to function
where our Ray 1 packet loss is 15% or more (as reported by utcapture).

This list has suggested that our problems are possibly duplex mismatch
and also to run "pingplotter" or "mtr".  Networking says the Cisco
switches appear to be fine.  For the Rays in this report, there
is only a single (Cisco) switch stack between server and client.

To use the Ray 2, we moved the server (X4200 running Solaris 10) to a
100 Mb/s line.  There are many posts on this very issue, including
problems noted by Sun with Gb-to-100Mb downsizing, with the blame
sometimes going to switches for dropping UDP packets.  The arrangement
vastly improves packet loss on the Ray 2, but does not help the Ray 1.

To be fair, the Ray 1 functions perfectly if we wire directly to the
server via a $50 Netgear GS108 switch.  We also have very good
performance if we wire through a router (an ordinary Pentium III with
two 3com 905 interfaces), even though this routes over the same wire
where we have packet loss when connected directly.

--Darrel Hankerson
_______________________________________________
SunRay-Users mailing list
[email protected]
http://node1.filibeto.org/mailman/listinfo/sunray-users

Reply via email to