Initial performance data

[Peter Van Epp]

	I've spent the last little while poking at performance. I have a 
reference IP only tcpdump file that I am playing back via tcpreplay
(initially at 1 meg per second so the sniffer doesn't overrun). As hoped
all three of the sniffer tcpdump and argus on FreeBSD give reproducable 
results (at least as far as file size of output file goes) and tcpdump reports
receiving the same number of packets as the sniffer at 5002, argus reports
seeing 4990 from ra. However speeding up there looks to be a break point (which
I think is FreeBSD/bpf) at between 20 and 30 megs. 20 is fine, 30 loses about
200 packets on tcpdump (the sniffer has gone west long ago, my new wire speed
one isn't here yet). That stays reasonably constant (~200 packets lost) up
to 95 megs (as fast as tcpreplay will go).
	The upshot of all this is I have a sniffer trace (from the 1 meg 
run) which gives me the size of all the packets in the stream and IP/port 
numbers to verify that ra and whats on the wire agree.
	I'll pop 2.0.0j in and see what that does in a little while.

Peter Van Epp / Operations and Technical Support 
Simon Fraser University, Burnaby, B.C. Canada