[net-perf] bwctl testing

[Russ Hobby]

Forwarding from Jeff Boote,

>>The path is the same is each direction.  I'll run a few test from UCLA to 
>>UCI and Abilene in a bit and post them.  A bug in the interaction between 
>>iperf and bwctl loses the results when originating some test from and I 
>>need to track it down and fix it.  In case anybody else has seen it and 
>>has a solution, the temporary file that iperf writes its results to is 
>>being truncated (i.e. the full results don't actually make it into the file).

This typically happens in cases where there is packet loss or a 
particularly bad guess (to large on the send side) for the correct window. 
Basically, bwctl is scheduling a test slot. If the iperf tests takes longer 
than the allocated time slot, then bwctl has to kill iperf. Unfortunately 
there is no way to get the intermediate results from iperf to find out how 
much data is currently buffered up in the send buffer so you could get an 
estimate for how the transfer is going.

My suggestion is to use the -i and -x flags to see the intermediate send 
results. Most likely the session will still not complete on time, but you 
will at least be able to see the build-up of the send buffer if this is the 
problem.

jeff