CSE 770 Paper Review

Reviewer: Michela Becchi
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?

Receiver-driven TCP protocols have been proposed as a mean to achieve higher throughput by moving congestion control functionalities to the receiver side. Thus, a central issue which must be addressed when considering these protocols is the action of misbehaving receivers. The paper addresses this topic, categorizing such actions in "denial of service attacks" and "resource stealing". Specifically, the authors analyze the trade-off between the introduction of protection mechanisms against misbehaving receivers and the performance improvements gained through the receiver-driven paradigm itself.

What are the main contributions of the paper and why are they important?

The fact that receiver misbehavior is a significant issue, especially when considering receiver-driven congestion control protocols, is not a new or original consideration. However, the authors make a systematic analysis of the problem. In particular:

they analyze the consequences of manipulating the basic parameters in the TCP throughput formula equation and of the initial congestion window
they point out the limitations of network and edge-router solution for malicious flows detection
they consider long lived as well as short flows, even if the analysis for short flows is brief

Moreover, they propose a solution for malicious receiver detection which aims at adding to the senders as little functionality as possible (Ping and TFRC agent) and making their control decisions independent of receivers feedbacks.

I think that the main contribution of the paper consists in methodologically highlighting how the introduction of means for detecting misbehaving receivers can compromise the gain in performance achieved by receiver-driven protocols.

How significant are these contributions relative to previous work?

I don't know in depth previous work on the topic. In particular, I don't have a sense of how previous work about receiver-driven congestion control protocols and solutions to it really addressed the limitations as done here.

Give detailed comments justifying your view of the paper.

The main issue in the deployment of receiver-driven congestion control protocols is the effect of misbehaving receivers. The authors explore such problem and the limitations of existing solutions. By proposing an own solution, their point out how the introduction (on the sender side) of mechanisms to detect misbehaving receivers can significantly decrease the performance improvements achieved through the use of receiver-driven protocols.

I would rank this as an "average" paper (since I don't feel like putting it on the upper part, bottom half seems more adeguate). In fact, it does not seem to bring breaking news. However, it motivates in a systematic way the limitations of receiver-driven protocols even in presence of mechanisms to avoid the impact of misbehaving receivers. Even if this does not appear to be a great contribution, it may avoid further exploration of these protocols, which seems to be useless.

The reason why I don't feel like putting the paper in the top half is that, while reading it, I have more the impression that it highlights and better defines already considered problems than it really brings new and interesting ideas. For some graphs, I wondered why determined choices of parameters have been made (e.g.: Fig. 4); I think that a thorough explanation of their exploration is missing.