[e2e] Two questions related to AQM design?

Jiangbin Yang jyang at lanterncom.com
Wed Sep 5 09:03:00 PDT 2001


One objective of AQM is to provide fairness among flows 
in case of dropping packets and avoid TCP synchronization.
Therefore, "fairness" should be a performance measure
besides throughput/delay.

Jiangbin Yang


-----Original Message-----
From: Zhang Miao [mailto:zm at csnet1.cs.tsinghua.edu.cn]
Sent: Wednesday, September 05, 2001 4:59 AM
To: end2end-interest at postel.org
Subject: [e2e] Two questions related to AQM design?


Hi,
  I have some questions related to AQM design.

(1) How to evaluate the performance of different AQM schemes?
    What is the metric?
    Some papers use throughput, queue length and loss rate
 when comparing different AQM schemes, while some papers
 draw a picture of queue dynamics. It's difficult to evalute
 when considering multiple metrics together. As I know, most
 papers use the metrics above individually. But as a matter
 of fact, usually we can't get all the best. For example,
 when we want to get a shorter queue, probably we have to sacrifice 
 the throughput. Then we need to calculate the tradeoff. "Power"
 (Power = Throughput / Delay) is a method to consider both 
 throughput and queue length. But I don't know why "Power" is
 seldom used for AQM evaluation. Does someone have any good
 idea?

(2) How dramatically the traffic changes in the Internet?
    One goal of congestion control is to maximize the utilization of
 the link. If the traffic changes very dramatically, it is nearly 
 impossible to keep both high utilizaion and low loss rate. For example,
 80 flows share a link with utilization of 99.9%, now 80 new flows are
 added, some packets have to be dropped before the end system can react.
 If this scenario happens very frequently (say, every 0.5 second), 
 perhaps no method works well.

Thanks a lot!

*****************************************************************
*    Zhang Miao                                                 *
*    Ph.D candidate,Department of Computer Science & Technology * 
*    Tsinghua University,Beijing,China(100084)                  *
*    Tel: (8610)-62785822                                       *
*    Email: zm at csnet1.cs.tsinghua.edu.cn                        *
*    Web: http://netlab.cs.tsinghua.edu.cn/~zm (domestic only ) *
*****************************************************************



More information about the end2end-interest mailing list