Dynamic Behavior of Slowly-Responsive Congestion Control Algorithms
Reviewer: Jie Zhou
Different classes of Internet applications demand different congestion control algorithms. To co-exist with the dominant TCP, these algorithms should
be governed by the TCP-compatibility requirement. However, this requirement is based on static notion of throughput under a steady-state loss rate. How
about its effeciency in practice, where most conditions are dynamic.
This paper investigates the behavior of slowly-responsive, TCP compatible congestion control algorithms under more realistic dynamic network conditions,
addressing the fundamental question of whether these algorithms are safe to deploy in the public Internet.
Slowly-responsive TCP-compatible algorithms may not always get their equitable share when network conditions change dynamically.
Incorporating the principle of packet conservation is crucial in dynamic settings to ensure safety.
I give the paper a rate 3 (modest contribution), because it renders more realistic analysis of slowly-responsive congestion control algorithms in
Internet, but this is not an original problem.
The authors test SlowCC mechanism on four aspects: 1) response to a suden increase in congestion; 2) long-term bandwidth fairness; 3) transient fairness;
4) link utilization. I think it basically covers most we want to know about SlowCC's performance in dynamic setting. Besides, the experiments have been
made in a single-bottleneck "dumbbell" topology with RED queue management at the bottleneck. Though simple, the model is appropriate.
Internet is a environment made up of dynamic conditions. It is dangerous to only study the static behavior of certain algorithm or application.