Re: [ns] window probing in ns

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

On Thu, 21 Jun 2001, gmc50 wrote:

> As you suggested, I connected a lossmonitor to the TCP source and saw that no
> bytes are received by the lossmonitor at the end of a 40.0s simulation if I
> set the TCP window_ = 0. If window_>0, bytes are received by the receiver as
> expected.
> 
> Whenever I set window_ = 0, the trace file is empty indicating that no packets
> are transmitted (using TCPSink or lossmonitor at the receiver side). This
> indicates that no probing from the sender is taking place.

That's pretty convincing. no zero window probe there...

L.

> Thanks,
> 
> Gustavo.
> 
> Lloyd Wood wrote:
> 
> > On Wed, 20 Jun 2001, gmc50 wrote:
> >
> > > Since ns-2 does not support dynamic advertised window, I suppose that it
> > > does not support zero window probe. Is this assumption correct?  (that
> > > is, if the advertised window (window_) = 0, does ns-2 support sender
> > > timer freezing and sending probe packets with exponential backoff (RFC
> > > 1122) to probe for a new advertised window?)
> >
> > if you connect a TCP source to a lossmonitor instead of a TCPSink,
> > you'll see the exponential backoff/probe behaviour.
> >
> > L.
> >
> > > If ns-2 does not support such sender action, is there any practical way
> > > to implement this action or has anybody done this before?

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