Proportional QoS in Differentiated Services Networks: Capacity Management, Equilibrium Analysis and Elastic Demands

Abstract

The Differentiated Services (Diffserv) architecture is a scalable solution for providing Quality of Service (QoS) over packet switched networks. By its very definition, Diffserv is not intended to provide strict performance guarantees to its subscribers. We purpose in this paper a particular form of relative performance guarantees. Specifically, the network manager’s goal is to maintain pre-defined ratios between common congestion measures over the different service classes. We assume that each service class is advertised with a constant price. Thus, in order to induce its goal, the manager dynamically allocates available capacity between the service classes. This scheme is studied within a network flow model, with self-optimizing users, where each user can choose the amount of flow to ship on each service class according to its service utility and QoS requirements. We pose the entire problem as a non-cooperative game. Concentrating on a simplified single-link model with multiple service classes, we establish the existence and uniqueness of the Nash equilibrium where the relative performance goal is obtained. Accordingly, we show how to compute and sustain the required capacity assignment. The extension to a general network topology is briefly outlined.