Reviewer: Sailesh Kumar
Date: 10-13-2005
How would you rate this paper, relative to others we have read? bottom 50%
How would you rate your knowledge of the topic of this paper? familiar, but not expert
What problem or issue does the paper address? Why is it important?
This paper seems like a survey paper, where authors survey existing sender driven and receiver driven congestion control protocols. The authors tries to present an array of arguments in order to make a strong case for receiver driven congestion control protocol and then slams it with the obvious deficiencies. Congestion control is certainly an important research area, however, in my opinion, it is an area, where zillions of papers have been published and have gone nowhere. I always wonder whether doing further research and analysis in this area makes any sense at all. However, in any case, the issues that this paper addresses in clearly important considering the recent security threats and outrages. Fair sharing of bandwidth among competing flows is also an important issue.
What are the main contributions of the paper and why are they important?
As I have already mentioned, this paper surveys several receiver driven, sender driven, and network assisted congestion control protocols. The emphasis has been given on the fairness and security. The real contribution is that authors have clearly performed experiments where they have demonstrated the deficiencies of several recently proposed protocols and schemes, like RED-PD, D-WARD and protocols, like TFRC and several other flavors of TCP, like TCP-ELN, etc. Most importantly authors consider the short time scale misbehaviors and also analyze possible solutions. They consider various scenarios of un-congested and congested network and illustrate the results where a misbehaving user tries to reduce its response time. Several possible tweaks that a malicious user can do, have been considered like increasing the initial window size parameter, and turning off the exponential back-off timers. The authors then analyze two potential solutions.
How significant are these contributions relative to previous work?
Frankly speaking, this paper doesn't have any direct contribution, in terms of any novel proposal or ideas. However, authors do a good job in pointing out the deficiencies of the existing methods and schemes and also they perform a good analysis of the potential solutions and their limitations.
Give detailed comments justifying your view of the paper.
This paper makes an argument for the receiver driven transport protocols and argues that this radically new protocol design achieves improves performance and efficiency. However, the vulnerabilities associated with such an approach are many and therefore such an approach requires in depth analysis before any deployment. In fact authors argues that present solutions are vulnerable to many threats like DOS, bandwidth stealing, unfair sharing of resources, etc. A malicious client in these contexts can receive massive unfair share and most importantly, this can be achieved with little effort - just by changing the parameters of the underlying congestion control equations. Various network assisted and end-to-end congestion control schemes have also been analyzed and authors concludes that network assisted schemes in general are not very effective. In the context of end-to-end protocols, authors presents various scenarios, where a malicious user degrades the response time and bandwidth of the background flows, both for the long and short term flows.