We have a Cisco 2924XL inbetween the two NetBSD boxes.  It's what 
we're using for performance measurement.  I'm not sure we'd be able
to capture the traffic at that kind of data rate with stock libpcap/
tcpdump on tiny hardware like this.. and the actual observation could
mess up the results further.  However-- if that is interesting I can
try and get you guys a trace.

I can also provide the traffic generator source if anyone's interested
at looking at it.

Mike

On Fri, Mar 30, 2001 at 03:50:49PM -0800, Jonathan Stone wrote:
> Problems like this are hard to reproduce, which makes them hard to
> diagnose. A timestamped `tcpdump' trace would show the dynamics of the
> acutal TCP behaviour. That would give a lot of insight into what is
> actually happening.
> 
> Exactly which kernel version are you using?  The 1.5 release, or
> something else?  There were some issues with de-tuning checksum
> throughput on certain kernel versions.
> 
> Also, are the two machines connected via a switch, or a shared medium?

-- 
Michael P. Lyle
Chief Technical Officer
Recourse Technologies, Inc.

The contents of this message are confidential.
Copyright 2001 M. Lyle