On Wed, 27 May 1998 16:05:30 +1000  Simon Burge wrote:

> This compares favorably with Ultrix 4.4 between two 5000/260s (R4400's):
> 
>  mel132:~ 2> ./ttcp -t -s -f m -n 8192 -u 144.136.33.70
>  ttcp-t: buflen=8192, nbuf=8192, align=16384/0, port=5001  udp  -> 144.136.33.70
>  ttcp-t: socket
>  ttcp-t: 67108864 bytes in 11.66 real seconds = 43.93 Mbit/sec +++
>  ttcp-t: 8198 I/O calls, msec/call = 1.46, calls/sec = 703.32
>  ttcp-t: 0.0user 9.5sys 0:11real 81% 30i+56d 17maxrss 0+0pf 0+979csw
> 
> This is between two reasonably busy machines connected via a Cisco
> concentrator.  I'll try again tonight when the machines should be idle,
> and let you know if performance is any better.

On idle machines, I get a bit better performance:

ttcp-t: buflen=8192, nbuf=8192, align=16384/0, port=5001  udp  -> 144.136.33.70
ttcp-t: socket
ttcp-t: 67108864 bytes in 9.19 real seconds = 55.73 Mbit/sec +++
ttcp-t: 8198 I/O calls, msec/call = 1.15, calls/sec = 892.31
ttcp-t: 0.0user 9.1sys 0:09real 99% 35i+56d 18maxrss 0+0pf 0+59csw

Notice almost 100% CPU utilisation.  I guess that 5000/240's probably
couldn't get close to the 55Mbit/sec figure (unless the NetBSD IP stack
is much better than the Ultrix one - I don't know about this).

Simon.