[e2e] Interpretation of ECN as a less severe congestion signal

[Michael Welzl]

Ongun,


> I may be misunderstanding your point,
> 
> If we build the model just as you proposed, we have two different reactions
> for Ecn Capable ans Non Ecn capable transmission.
> If the point is There shouldn't be two differnt policies on a router for RED
> queues, thats a good point, however for diffserv there are three different
> policies, But I accept that diffserv is a burden for processing, and
> implementing two policies is also a burden.

Actually, dropping in case of very severe congestion is already
specified; I am sorry for having missed this paragraph in RFC 3168:

   First, ECN-Capable routers will only mark packets (as opposed to
   dropping them) when the packet marking rate is reasonably low. During
   periods where the average queue size exceeds an upper threshold, and
   therefore the potential packet marking rate would be high, our
   recommendation is that routers drop packets rather then set the CE
   codepoint in packet headers.

Now all we need to change is the endpoint behaviour so as to
have behaviour which would be less reactive when an ECN flag is
used, but still fair towards TCP.
:)

Minseok Kwon pointed me to his ICC 2002 paper with Sonia Fahmy:

Minseok Kwon and Sonia Fahmy, "TCP Increase/Decrease Behavior for
Explicit Congestion Notification (ECN)"
http://www.cs.purdue.edu/homes/fahmy/papers/papers.html

This is pretty much what I had in mind.

Cheers,
Michael