A Sliding Mode Approach to Traffic Engineering in Computer Networks

Abstract

In this chapter, we address the problem of optimal Traffic Engineering (TE) in computer networks; i.e., determining an optimal traffic allocation in the presence of both multiple paths and multiple Classes of Service (CoSs). More precisely, we aim at designing data rate adaptation laws that maximize the utilization of the network resources (as measured by a given utility function) subject to link capacity constraints and call service requirements.

This problem, as it is formulated here, is an “easy” problem; i.e., the problem of optimal traffic allocation can be stated as maximizing a concave function subject to linear constraints. Hence, if global information is available, one could use well-known algorithms, such as gradient descent, to determine the optimal traffic allocation.