A MDP Approach to Fault-Tolerant Routing

摘要

This paper defines a theoretical framework based on Markov Decision Processes (MDP) to deal with fault-tolerant routing algorithms in heterogeneous networks, which are realized through the integration of assorted wired and wireless telecommunication technologies. Such kinds of networks are characterized by fast dynamics of link availabilities, mainly due to the extensive use of wireless technologies. The novelty of this paper is the formulation of the fault-tolerant routing problem as a MDP, which is used to compute the optimal re-routing policy. As in existing fault-tolerant algorithms, when a path becomes unavailable, the traffic flows transmitted over that path are re-routed on another available path; the novelty is that the new selected path is the one that minimizes re-routing occurrences, since it is selected taking into consideration the probability that also the alternative paths can become unavailable in the future. Asa by-product, the optimal path selection for new traffic flows is also obtained. Simulations show the effectiveness of the proposed approach.