[ns-users] Modeling ns TCP

[rik@scs.leeds.ac.uk]

Despite the fact that I'm not being much of a mathematician, I really
have to model TCP performance in ns (2.1b5) for a paper I'm working on.
What I would like to do is predict the transmission window size at a
given time. 

My results match the ns packet timestamp output up to a certain point
in the simulation but then start to become less accurate. I have read
through the documentation regarding link delays etc. and have the s/b+d
(pkt size/bandwidth + delay) components. However, there appears to be a
queueing or propagation delay which I'm not taking into account at the
sender.

I'd really appreciate it if anyone could point me to the missing
component(s). I currently have 1 RTT in ns as simply:

   transmission delay of PKT = #nodes_out*(s/b+d)
 + transmission delay of ACK = #nodes_return*(s/b+d)

(nodes_out == nodes_return) 

There does, however, seem to be an additional delay somewhere to do
with the propagation of new packets from Tx. Any further information
greatly appreciated.

Rik Wade
--
ATM Group
University of Leeds
UK