[Iccrg] Update to draft-irtf-iccrg-welzl-congestion-control-open-research

[Saverio Mascolo]

dear welzl,

referring to the section '3.2 Challenge 2: Corruption Loss' of the draft, i think that an important point is missing here, which is the basic idea introduced at Mobiocm 2001 in the TCP Westwood paper. 

The key concept of westwood tcp is the following:

- when a packet loss is detected through 3 dupack (or timeout) set the congestion window equal to 
the measured bandwidth at time of congestion. this measurement is obtained by counting and filtering the ack rate. this setting  provides 
a significant goodput improvement in lossy channels, i.e. when packet are lost not due to congestion.

The reason why this setting works very well in lossy channel is that, since losses are  not due to congestion, 
the "blind" by half window reduction 
of New Reno is avoided, i.e. the window is not over-shrinked.


We have run thousands  of tests over real networks using New Reno,  Cubic and Westwood and we have found that all 
basically provide similar performance except in the case of lossy links where tcp westwood provides a neat improvment due to control window shrinking 
based on measured available bandwidth. 

On the other hand it is well-known that using croos-layer signalling is not a working solution due to inter operability, indeed it violates the fundamental design principle of layering.

best,
saverio 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://oakham.cs.ucl.ac.uk/pipermail/iccrg/attachments/20080424/18ae28d9/attachment.html