OpIATE: Optimization Integrated Adaptive Traffic Engineering

The advent of Multiprotocol Label Switching (MPLS), a connection oriented technology, allows packets in datagram networks to be sent along explicit (not necessarily shortest) paths. This provides scope for the development and more importantly, implementation of traffic engineering schemes that can implicitly or explicitly optimize a network wide performance criterion. We propose a decentralized, optimization integrated adaptive traffic engineering scheme called OpIATE which seeks to adaptively balance the dynamic traffic load experienced in a service provider’s MPLS enabled intra-autonomous system(AS). A key contribution of this paper is to explore the role of non-intrusive feedback parameters, such as ingress-egress flow allocation, to steer sources in a network towards a shared global perspective in terms of network routing. The OpIATE framework depends on the edge nodes in a network having the capability to solve complex flow allocation problems and does not require participation from the interior nodes of a network. Further, no assumptions on scheduling or buffering mechanisms are made. We present an analytical model which establishes the asymptotic convergence of OpIATE flow allocation updates to a globally optimal solution. Finally, we present packet-level simulation results that illustrate the application of OpIATE for different traffic scenarios and show favorable comparison with other adaptive multipath routing schemes like MATE.