Network Recovery After Massive Failures

This paper addresses the problem of efficiently restoring sufficient resources in a communications network to support the demand of mission critical services after a large scale disruption. We give a formulation of the problem as an MILP and show that it is NP-hard. We propose a polynomial time heuristic, called Iterative Split and Prune (ISP) that decomposes the original problem recursively into smaller problems, until it determines the set of network components to be restored. We performed extensive simulations by varying the topologies, the demand intensity, the number of critical services, and the disruption model. Compared to several greedy approaches ISP performs better in terms of number of repaired components, and does not result in any demand loss. It performs very close to the optimal when the demand is low with respect to the supply network capacities, thanks to the ability of the algorithm to maximize sharing of repaired resources.