Just from this description, it sounds like the maximum cwnd is set to the default 20 packets. When ifqlen is less than 20, you might get a drop before the cwnd hits the limit, so then it would decrease. When ifqlen is greater than 20, the cwnd is allowed to grow to 20 and then doesn't grow anymore. Is this the behavior you're seeing?
If so, I believe you should be able to change the max FullTcp cwnd limit (window_) and see the behavior that you expect. -Michele On Jan 26, 2007, at 8:00 PM, [EMAIL PROTECTED] wrote: > Date: Fri, 26 Jan 2007 09:12:47 +0100 > From: " Andr?s Emilio Arcia Moret " <[EMAIL PROTECTED]> > Subject: [ns] FullTcp problems in wireless > To: [email protected] > Message-ID: > <[EMAIL PROTECTED]> > Content-Type: text/plain; charset=ISO-8859-1; format=flowed > > Hi everybody, > > I'm doing some basic testing of FullTcp for a simple ad-hoc scenario > using 802.11 (2 nodes linked by 802.11 link). So I've been googling > on the subject and I've found [apart from the different threads of > discussions] that Michele Weigle has a patch for the DSDV routing > protocol (which correct the original problem): > > http://www.cs.odu.edu/~mweigle/research/netsim/fulltcp-wlan.html > > After applying the patch to the ns-2.30, FullTcp partially works. Only > if I keep the ifqLen less than 20 packets the _FullTcp_ behavior gives > some soundness. However, when I set ifqLen over 20 the CWND increases > up until 20 packets and freezes there till the end of the simulations. > > Any suggestion is most welcomed. > > PS. I've also tried Mathieu Lacage's 80211 model with kind of the > same outcome. > > -- > Andr?s Arcia
