Decentralized optimal routing for packets flow on data networks

This paper concerns optimization of data traffic flows on a telecommunication network, modelled using a fluid - dynamic approach. Flows can be controlled adjusting traffic distribution and priority parameters. Two cost functionals are considered, which measure average velocity and average travelling time of packets. First we address general optimal control problems, showing that existence of solutions is related to properties of packet loss probability functions. A direct solution of the general optimal control problem corresponds to a centralized policy and is hard to achieve, thus we focus on a decentralized policy and provide solutions for a single node with two entering and two exiting lines and asymptotic costs. Such solutions permit to simulate the behaviour of decentralized algorithms for complex networks. Local optimization ensures very good results also for large networks. The latter is shown by case study of a test telecommunication network.