Paper review: Dynamic Behavior of Slowly-Responsive Congestion Control Algorithms

Reviewer: Kevin Hofstra

1. Although some TCP-compatible congestion algorithms appear to be using the same amount of bandwidth, when put in to a highly dynamic environment such as the internet they have been found to not compete equitably especially over time. What aspects of these algorithms are responsible for their remaining TCP-compatible under dynamic conditions?
2. A comparative analysis of several distinct techniques to remain TCP-Compatible over long periods of time in a highly dynamic environment. An overview of all of the possible solutions, where they fit into the spectrum, and there actual effectiveness in a tested dynamic scenario.
3. A. The different TCP-Compatible congestions algorithms (Family):

i. RAP (TCP-equivalent)

ii. AIMD (Multiplicative increase/decrease)

iii. AIMD (Binomial)

iv. TEAR (Binomial)

v. TFRC6 (Binomial)

2. TCP-Compatibleness in a static environment does not directly translate to the dynamic environment over long periods of time.
3. The TCP-Compatible paradigm simply transforms the requirement that all congestion control mechanisms are TCP into all being TCP-compatible. This paradigm is required because of TCP’s widespread acceptance.
4. SlowCC algorithms do not meet TCP fairness because they are too slow to respond to network conditions, and do not take enough available bandwidth.

4. Critique the main contribution
• Significance- 2 The article is more of an evaluation of the current implementations than an actual discovery. There evidence is somewhat conclusive, but should be considered more of a review than a discovery.
• Convincing- 3 Each of the implementations are well explained and they do present some actual network data. It is a difficult task to create a large dynamic environment, but this is necessary for testing how it will perform on the internet.
5. System researchers and builders should recognize how important implementation is before making something a standard. Had some of these congestion control mechanisms been released on the grounds that they were TCP-compatible in a static environment, they would have been found to be lacking in the dynamic environment. They have been deemed safe for deployment, but suffer from a lower level of throughput than normal TCP. The question that needs to be answered is: Can we sacrifice throughput in order to have more stable flows? In the case of streaming media the answer may be yes.