Near Optimal Network Design for Path Pair Availability Guarantees

Guaranteeing high levels of availability in the network in a cost effective manner is of primary importance to network operators and managers. We address the network design problem for path pair availability guarantees, assuming links can be upgraded to have an increased availability. Since the path pair availability constraints are nonlinear and not linearizable in an exact manner, this mathematical problem has been avoided by considering only the working path availability or availability guarantees for the working and backup paths separately in a disaggregated way. In this paper, we present an aggregated model, where only the path pair availabilities must be fulfilled. In this model, we consider a convex relaxation for an approximation of the path pair availability to obtain linear constraints, and describe an iterative approach to tighten the bounds of the solution space, in order to obtain near-optimal solutions. The results show that considering an aggregated model is more cost effective than considering a disaggregated model with explicit values for the availabilities of the working and the backup paths.