A bicriteria optimization approach for robust OSPF routing

For intra-domain routing, open shortest path first (OSPF) is the most commonly used protocol. In OSPF, traffic is routed using paths that are shortest with respect to a set of link costs. One problem is the optimization of OSPF link costs. The problem amounts to finding a set of link costs such that the total network congestion is minimized. We consider robust OSPF routing that takes into account the impact of link failures. The optimized link costs for the non-failure scenario do not necessarily minimize the congestion when a link failure occurs. An OSPF routing solution is robust if it not only leads to low congestion in the non-failure scenario, but also attempts to minimize the impact of link failures. To address this problem, we present a bicriteria optimization model, which simultaneously considers the congestion of the normal state and that of the failure states. Previously proposed methods for bicriteria optimization problems often involve weighting together the two objective functions. We propose a different approach using an artificial objective function. When embedded into a local search algorithm, the function guides the search towards Pareto-optimal solutions. These solutions can be used for analyzing the trade-off between the congestion of the normal state and that of the failure states. The proposed algorithm can therefore be used as an aid for implementing robust OSPF routing.