Supporting Differentiated QoS in MPLS Networks

Abstract

This paper proposes a new approach for providing different levels of QoS in IP networks over MPLS. The system’s operator wishes to maximize the throughput of the high priority flows by choosing the less congested paths while taking care of the applications’ requirements. These requirements give rise to an optimization problem that consists of finding resource-constrained shortest paths in a directed graph. The problem can be formally cast in mathematical programming and its solution can follow two directions: (i) a centralized, heuristic approach that aims at approximating the optimal solution in a fast and efficient way; (ii) a distributed approach that relies on Lagrangean relaxation to decompose the optimization problem in small subproblems and thereby divide the computational burden between distributed routers. By means of numerical analysis and simulation, this paper demonstrates the effectiveness of the proposed approaches and shows QoS improvements of the high priority flows.