Re: [ns] Reg bandwidth utilization!!

[Lloyd Wood <l.wood@eim.surrey.ac.uk>]

On Tue, 10 Jul 2001, T. Chitti Babu wrote:

> here's a program to calculate RTT for every 0.01 secs.
> u may take the average and ofcourse half the value of that average would
> give u END-TO-END DELAY.

that assumes symmetry and the same path and queuing delays forwards
and backwards. Big assumptions.

tcp.cc:	bind("rtt_", &t_rtt_);

ugh. There's one reason why the tcp rtt and rto computation code is so
difficult for mere mortals to follow.

Using the tcp sender's estimates to get end-to-end delay won't be that
accurate. I wrote an awk script to parse tracefiles to extract path
delay information (and then used that to generate about a hundred
graphs in my thesis and get a degree) - see my ns webpage for that and
for a method of determining actual tcp flow goodput.

L.

<L.Wood@surrey.ac.uk>PGP<http://www.ee.surrey.ac.uk/Personal/L.Wood/>