[net-perf] bwctl testing

[Mark Boolootian]

I'm seeing the same funny performance stuff to UCLA that Mike Iglesias
saw.  I get about 22 mbit/sec sending to UCLA, and 75-85 mbit/sec
receiving.  My system is on a 100 mbit/sec tap at the moment, so I'm
not really concerned about the receive rate.  But the sending rate
is puzzling.

A tcptrace on a capture of a transfer to UCLA reports window issues:

            UCSC                                    UCLA
     max win adv:         1049792 bytes     max win adv:           99328 bytes
     min win adv:         1049792 bytes     min win adv:           17920 bytes

I ran iperf with '-w 1m' and my side claims a 1 mbyte window, though it
is irrelevant as I'm the sender.  On the other hand, UCLA never advertises
anything larger than 100k bytes.

I'm going to keep on digging, but was wondering if this sounded familiar
to anyone.

mb