Paper review: Charge-Sensitive TCQ and Rate Control in the Internet

Reviewer: Hanlin D. Qian

This paper addresses the issue of achieving optimum distribution of network resources in a distributed environment, based only on information available on end hosts.

The main idea of the paper is that by distributing the burden of optimizing utility functions among individual users, a network optimum can be achieved. In other words, as each user tries to maximize his own utility function, an equilibrium is achieved, and the authors argue that this Nash Equilibrium yields an optimum network state. This allows the process of charging per unit of bandwidth dependent upon the utility of the end-user, rather than on the any particular resource points on the network. This way, efficient allocatin of network resources can be achieved with modifications taking place at the end-user and only minor modifications to the network.

Here are several other ideas from this paper:

  1. Users can specify their own utility or objective functions, noting the unit price they're willing to pay for bandwidth.
  2. Price charged is dependent on the user's objective function.
  3. Users can update their willingness to pay.

I give this paper a rating of 3 for modest contribution. The paper has some good ideas, though much of it is based on past studies. I think its main contribution is a breakthrough because it is a way that efficient allocation of network resources can be achieved in a distributed manner.

However, I believe that the model the authors use is still a bit too simple. Many of the theorems proven assume only a single point of netowrk bottleneck. Another problem is that the authors did not really make clear how much modifications need to go into the network. It is my understanding that a little modification is necessary, but how much is that? In general, the paper presents promising prospects of further research in this area.