[e2e] a new paper on Adaptive RED

[Saverio Mascolo]

You may be interested to look at "Easy Red: improving fairness via simple
early discard" available from:

http://www-ictserv.poliba.it/mascolo/papers/Easy_red.pdf

Easy RED has the following main features:

1) It DOES NOT average the queue;
2) min_th0=queue capacity/3
3)max_th=queue capacity
4) pdrop=constant=0.01   when instantaneous queue> queue capacity/3
  otherwise is zero

This setting is very easy and gives remarkable throughput and fairness
incremement respect to RED/Gentle_RED.

Finally, why fairness is considered a secondary metric and no experiments
are devoted to it? Fom  VJ [88] paper

"While algorithms at the transport end points can insure the network
cacapcity is not exceeded, they cannot insure fair sharing of that capacity.
Only in gateway, at the convergece of flows, is there enough information to
conontrol sharing and fair allocation"

it seems evident that fairness improvment is a primary issue .

-Saverio

----- Original Message -----
From: "Sally Floyd" <floyd at aciri.org>
To: <end2end-interest at postel.org>
Sent: Wednesday, August 01, 2001 11:55 PM
Subject: [e2e] a new paper on Adaptive RED


> Ramki Gummadi, Scott Shenker, and I have a new paper out on "Adaptive
> RED: An Algorithm for Increasing the Robustness of RED's Active
> Queue Management", available from:
>
> http://www.aciri.org/floyd/papers/adaptiveRed.ps
> http://www.aciri.org/floyd/papers/adaptiveRed.pdf
>
> The abstract is appended below.  The validation test for Adaptive
> RED in the NS simulator can be run with "./test-all-adaptive-red"
> in the directory tcl/test.
>
> Thanks,
> - Sally
> http://www.aciri.org/floyd/
>
> Abstract:
> The RED active queue management algorithm allows network operators
> to simultaneously achieve high throughput and low average delay.
> However, the resulting average queue length is quite sensitive to
> the level of congestion and to the RED parameter settings, and is
> therefore not predictable in advance.  Delay being a major component
> of the quality of service delivered to their customers, network
> operators would naturally like to have a rough a priori
> estimate of the average delays in their congested routers; to
> achieve such predictable average delays with RED would require
> constant tuning of the parameters to adjust to current traffic
> conditions.
>
> Our goal in this paper is to solve this problem with minimal changes
> to the overall RED algorithm.  To do so, we revisit the Adaptive
> RED proposal of Feng et al. from 1997 [FKSS97, FKSS99].  We make
> several algorithmic modifications to this proposal, while leaving
> the basic idea intact, and then evaluate its performance using
> simulation.  We find that this revised version of Adaptive RED,
> which can be implemented as a simple extension within RED routers,
> removes the sensitivity to parameters that affect RED's performance
> and can reliably achieve a specified target average queue length
> in a wide variety of traffic scenarios.  Based on extensive
> simulations, we believe that Adaptive RED is sufficiently robust
> for deployment in routers.
>