[Iccrg] Forcing flow congestion response without policing flows

Bob Briscoe rbriscoe at jungle.bt.co.uk
Thu Nov 13 16:56:29 GMT 2008


Folks,

This paper should help in the discussions on
- the open issues draft
- Mat Mathis's 'TCP unfriendly' talk next week in Minneapolis

Policing Freedom to Use the Internet Resource Pool
Arnaud Jacquet (BT), Bob Briscoe (BT & UCL) & Toby Moncaster (BT)
To appear: Workshop on Re-Architecting the Internet (ReArch'08) (Dec 2008)
<http://www.cs.ucl.ac.uk/staff/B.Briscoe/pubs.html#polfree>

The paper focuses on the 'e2e argument' point in the open issues 
draft: What's the minimum needed in the network layer to ensure 
fairness can be controlled and congestion collapse can be prevented 
(ie. first order & second order resource allocation)?

The paper analyses the cross-effect of unresponsive flows on 
responsive flows in an environment where only overall congestion 
caused by a user throughout the whole Internet is policed, rather 
than policing individual flows.

Prevention of congestion collapse is perhaps a more important angle 
on what Wes said a couple of months back in his review of the open 
issues draft:
"flow-rate fairness is a special case of cost-fairness"

_____________
The summary sentences of most relevance to this w-g are:

"Not all applications have to respond quickly to congestion, as long 
as the overall response is sufficient. If it isn't, the policer will eventually
force even inelastic flows to terminate."

"...prevalent per flow congestion responsiveness certainly prevents 
congestion collapse. But it is a fallacy that it also controls fair 
sharing of resources, particularly given it ignores sharing over time.

However, our congestion policer shows the reverse approach is 
fruitful. Ensuring fair shares of everyone's overall congestion 
contribution does also ensure per flow congestion 
responsiveness---both voluntarily and ultimately by enforced 
intervention (S.5)."




Bob



____________________________________________________________________________
Bob Briscoe, <bob.briscoe at bt.com>      Networks Research Centre, BT Research
B54/77 Adastral Park,Martlesham Heath,Ipswich,IP5 3RE,UK.    +44 1473 645196 




More information about the Iccrg mailing list