Paper Review:
Charge-Sensitive TCP and Rate Control in the Internet
Reviewer: Jie Zhou
Problem
Because Internet is a public domain and its participants potentially do not act cooperatively, we need to design a pricing mechanism for Internet.
Previous approaches including flat rate charging is found to have the problem of ineffeciency. A desirable system should not only be fair, but also
allocate the available bandwidth in such a way that maximize the total utility of users.
Contribution
The paper investigates the fundamental problem of achieving the system optimal rates in a distributed environment.And it introduces an algorithm
that may be deployed over the Internet without significant modification within the network.
Main Ideas
The system optimum is achieved at an equilibrium, when users update their parameters by solving their own optimization problem.
In the case of a single bottleneck, there exists a unique Nash equilibrium, and the equilibrium approaches the system optimum as the number
of users increases.
Critique
This paper is developed upon Kelly's suggestion that the problem of achieving the system optimum can be decomposed into two subproblems: network and
user problems. And the algorithm in the paper is an extension of an algorithm introduced J.Mo and J. Walrand. Though it is not original, the paper
provides some insight into achieving system optimum in terms of both fairness and total utility. So I give it a rating of 3 (modest contribution).
Lession
Due to the size of the Internet, an algorithm that requires an extensive modification inside the network may not be suitable for deployment.