Paper Review : Analysis and Design of an Adaptive Virtual Queue (AVQ) Algorithm for Active Queue Management
Reviewer : Seh Leng Lim
This paper attempts to “market” a particular Active Queue Management (AQM) scheme known as the Adaptive Virtual Queue (AVQ).
The main contribution of the paper is its analysis of the properties of the Adaptive Virtual Queue (AVQ) , and its argument about the superiority in performance of AVQ to other AQM schemes such as RED, REM, PI controller etc. AVQ will benefit efforts to achieve fairness of bandwidth allocation in the Internet.
The key main ideas expounded are :
(a) To provide explicit congestion notification (ECN) marks or drop packets to provide fairness and control queue lengths, the routers have to select packets intelligently in a manner that conveys information about the current state of the network to the users. Algorithms which the routers employ to convey such information are called Active Queue Management (AQM) schemes.
(b) the idea of a virtual-queue AQM scheme, namely the Adaptive Virtual Queue (AVQ) whereby the virtual queue is less than the actual capacity of the link and packets in the real queue are marked/dropped when the virtual buffer overflows
© preference for algorithm with small queue lengths and robustness in the presence of extremely short flows
I think that the paper has a modest contribution (rating of 3) to the study of Active Queue Management algorithms. As can be seen by the number of algorithms that the paper attempts to compare with, there is already a lot of work on Active Queue Management. The authors first demonstrate that their AVQ algorithm stabilizes in the simple test setting of a single 10 Mbps link with further introductions of short flows. Then they use the performance measures of packet losses and link utilization, responsiveness to changing network conditions, and queue length whenever appropriate to benchmark with other AQM algorithms in the presence of only long-lived FTP flows, the dynamic scenario when flows are dropped and then introduced later, the test scenario when short flows are gradually increased, and lastly, making the router drop packets instead of marking them at the onset of congestion. As the test scenarios are pretty simplified, it remains to be seen if AVQ will actually perform better than the other AQM algorithms in the Internet. Also, it is hard to be convinced that due to space constraint, the fairness measures of AVQ cannot be elaborated. Therefore, we do not know if AVQ really supports fairness.
Researchers and builders working with Internet applications may have a better appreciation from this paper of the difficulties involved with network congestion control mechanisms.