Decentralized optimal traffic engineering in the internet

Abstract

Distributed optimal traffic engineering in the presence of multiple paths has been found to be a difficult problem to solve. In this paper, we introduce a new approach in an attempt to tackle this problem. This approach has its basis in nonlinear control theory. More precisely, it relies on the concept of Sliding Modes. We develop a family of control laws, each of them having the property that the steady-state network resource allocation yields the maximum of the given utility function, subject to the network resource constraints. These control laws not only allow each ingress node to independently adjust its traffic sending rate but also provide a scheme for optimal traffic load redistribution among multiple paths. The only nonlocal information needed is binary feedback from each congested node in the path. Moreover, the algorithms presented are applicable to a large class of utility functions, namely, utility functions that can be expressed as the sum of concave functions of the sending rates. We show that the technique can be applied not only to rate adaptive traffic with multiple paths, but also to assured service traffic with multiple paths. Preliminary case studies show that this technique is potentially very useful for optimal traffic engineering in a multiple-class-of-service and multiple-path enabled Internet, e.g., differentiated services enabled multi-protocol label switching networks.