[e2e] About TCP modeling in an asymmetric link

[Alhussein Abouzeid]

Wei,
Check the following paper and the references therein (I believe Lakshman
et. al., reference included in the paper below, presented a mathematical
analysis of the case you mention, though they don't mention the
"friendliness" stuff).

Balakrishnan, H.; Padmanabhan, V.N., How network asymmetry affects TCP,
IEEE Communications Magazine , Volume: 39 Issue: 4 , April 2001 
Page(s): 60 -67.

Hope this helps,
AA.

--------------------------
Alhussein Abouzeid                              
Assistant Professor, Electrical, Computer and Systems Eng. Dept.     
Rensselaer Polytechnic Institute
http://www.ecse.rpi.edu/Homepages/abouzeid/

On Wed, 6 Mar 2002, Wei Wu wrote:

> Hi, all,
> 
> Recently there are many works on TCP performance modeling and TCP-frieldly 
> congestion control and the performance of TCP with a bottleneck on the data 
> path are well modeled, such as the well-know 1/sqrt(p) law, which the concept
> of "TCP-friendly" is based on. However, I wonder that, when the bottleneck 
> of the TCP connection is on the ACK path, the forward path loss rate p_f and
> the backward path loss rate p_b both existed (moreover, p_f<<p_b), how we 
> can get performance evaluation and whether the traditional TCP-friendly 
> protocols are friendly in this case. How should we consider p_b in the 
> TCP-friendly equation and how to measure it?
> 
> I wonder whether there are any works on this problems. Any hint is helpful.
> Thanks!
> ¡¡¡¡¡¡¡¡¡¡¡¡¡¡
> Best Regards
> Wei Wu
> wuwei at sdp.ee.tsinghua.edu.cn
>