Effect of unreliable nodes on QoS routing

A number of QoS routing algorithms have been proposed to address the dual objective of selecting feasible paths through the network with enough resources to satisfy a connections' QoS request, while simultaneously utilizing network resources efficiently. However, these routing algorithms and the guarantees they provide do not consider the possibility of node and link failures. The failure of a node or a link along a path can disrupt the continuity of an on-going session and potentially terminate the session. Hence the problem of QoS routing should be extended to incorporate reliability and fault-tolerance requirements. We study the impact of unreliable nodes on QoS routing. We describe a scheme to restore the flows that are disrupted due to node failures to alternate paths. Two prioritized restoration policies, one to maximize the number of disrupted flows that can be restored, and the other to maximize the disrupted demand that can be restored are also presented. We conduct extensive simulations to evaluate the routing performance in the presence of node failures and repairs, as well as the performance of the restoration scheme. Our results indicate: (i) when the availability of the nodes is beyond a certain threshold, the routing as well as the restoration performance is comparable where the number of failures in one is twice the number of failures in the other; (ii) the percentage of the disrupted flows and the percentage of the disrupted demand that can be restored successfully increase with decreasing network load; and (iii) prioritized restoration policies are effective under heavy network load conditions, and their effectiveness decreases with decreasing network load.