Charge-sensitive TCP and rate control in the Internet

Abstract

We investigate the fundamental problem of achieving the system optimal rates in a distributed environment, which maximize the total user utility, using only the information available at the end hosts. This is done by decomposing the system problem into two subproblems-network and user problems-and introducing an incentive-compatible pricing scheme, while maintaining proportional fairness. We demonstrate that when users update their parameters by solving their own optimization problem, at an equilibrium the system optimum is achieved. Furthermore, this algorithm does not require any explicit feedback from the network and can be deployed over the Internet with modifications only on the end hosts. In the second part of the paper we model as a noncooperative game the case where the choice of each user's action has nonnegligible effect on the price per unit flow at the resources and investigate the Nash equilibria of the game. We show, in the simple case of a single bottleneck, that there exists a unique Nash equilibrium of the game. Further, as the number of users increases, the unique Nash equilibrium approaches the system optimum.