On 11/22/05 06:51 Scott Ullrich said the following:
> megabits on a 266 megahertz WRAP to 28.5 megabit.   inflight is really
> meant for overly congested links and really hurts performance.    NOTE

robert watson's posts to freebsd-net indicate that in high bandwidth, low 
latency links (such as your average LAN), inflight enabled actually drops 
performance.

on a Celeron 400Mhz with 128MB RAM and intel etherexpress onboard NICs, 
i've been able to get near wire speed throughput (80+Mbps with iperf) on 
the 1.2 and 1.3a1 images, with or without inflight enabled. this doesnt 
indicate much however for the Celeron is beefy enough a processor for what 
we're throwing at it. we'd need to use a processor which maxed out /before/ 
wire speed throughput was reached, if we're to really benchmark the IP 
stack difference between freebsd 4.x and 6.0.

a better benchmark would be doing the same on the soekris net45xx and 48xx 
boxes for the lower speed processor there would max out before the wire 
speed is ever reached.

-- 
Regards,                           /\_/\   "All dogs go to heaven."
dinesh at alphaque dot com                (0 0)    http://www.alphaque.com/
+==========================----oOO--(_)--OOo----==========================+
| for a in past present future; do                                        |
|   for b in clients employers associates relatives neighbours pets; do   |
|   echo "The opinions here in no way reflect the opinions of my $a $b."  |
| done; done                                                              |
+=========================================================================+