Paper Review: Analysis and Design of an Adaptive Virtual Queue (AVQ) Algorithm for Active Queue Management
Reviewer: Kenneth Chin
Adaptive Virtual Queue (AVQ) is a paritcular scheme of Active Queue Management (AQM). The objective of AVQ is to maintain small queue size and its robustness in the presence of extremely short flows. This paper is presenting a simple rule to design the parameters of the AVQ algorithm so as to achieve its goal.
AVQ has a virtual queue and a real queue. The virtual queue has a smaller capacity than the real queue, but they have the same buffer size. Upon each packet arrival, a fictitious packet is enqueued in the virtual queue if there is sufficient space in the buffer. If the new packet overflows the virtual buffer and the real packet is either marked or dropped. At each packet arrival, the virtual queue capacity is updated based on the arrival rate and some constants like the capcity of a link (e.g. 10Mbps), the desired utilization at the link (e.g. 0.98) and a smoothing (or damping) factor. The AVQ algorithm is stated in section 2 of the paper.
It now comes to decide the values of the parameters of the AVQ algorithm. The simple rule proposed by the author is based on the congestion-avoidance algorithm of the TCP users where the the utility function is -1/(d^2x). The most fascinating thing is the linearization of the non-linear TCP/AQM model, and this is the most important thing i can take away from this paper. Nevertheless, I still have a question on one of the assumptions made by the author which is assuming each user has a common round-trip propagation delay d.
Based on the analysis, I think this paper has a decent contribution towards active queue management. Therefore, it is a 3rd-grade paper.