A parallel routing algorithm for traffic optimization

Abstract

This chapter applies the general complex network theory to study a parallel routing algorithm called Classified Traffic Routing (CTR) for traffic optimization. The parallel routing algorithm involves two routing tables when forwarding packets. For CTR, performance analysis is performed which focuses on loss, delay, and energy in a scale-free network. Its performance is also compared to those of single routing algorithms. For existing two main kinds of single routing algorithms - shortest path first and congestion avoidance first algorithms, we select one representative algorithm from each kind for comparison. The result shows that good loss or delay performance (but not both) can be obtained for each representative routing algorithm, namely Shortest Path First (SPF) algorithm and Optimal Routing (OR) algorithm. The chapter then discusses a study on energy performance of these two algorithms. The results show that the two algorithms have very different performance on average energy consumption and on distribution of energy consumption among all nodes. This chapter then argues that single routing algorithm could not meet the requirements of different types of traffic while could not balance the energy consumption. In order to provide good loss performance for loss-sensitive traffic and good delay performance for delay-sensitive traffic, and in consideration of energy consumption, forwarding packets with CTR is a good choice. Simulation results show that CTR can give a much more balanced performance on loss, delay, and energy than those of SPF and OR.