Guaranteed Performance Routing of Unpredictable Traffic With Fast Path Restoration

Two-phase routing, where traffic is first distributed to intermediate nodes before being routed to the final destination, has been recently proposed for handling widely fluctuating traffic without the need to adapt network routing to changing traffic. Preconfiguring the network in a traffic-independent manner using two-phase routing simplifies network operation considerably. In this paper, we extend this routing scheme by providing resiliency against link failures through fast path restoration along disjoint end-to-end backup paths. We view this as important progress toward adding carrier-class reliability to the robustness of the scheme so as to facilitate its future deployment in Internet service provider (ISP) networks. On the theoretical side, the main contribution of the paper is the development of linear-programming-based and fast combinatorial algorithms for two-phase routing with fast path restoration so as to minimize the maximum utilization of any link in the network, or equivalently, maximize the throughput. The algorithms developed are fully polynomial time approximation schemes (FPTAS)-for any given epsiv > 0, an FPTAS guarantees a solution that is within a (1+epsiv)-factor of the optimum and runs in time polynomial in the input size and [ 1/(\relax epsiv)]. To the best of our knowledge, this is the first work in the literature that considers making the scheme resilient to link failures through preprovisioned fast restoration mechanisms. We evaluate the performance of fast path restoration (in terms of throughput) and compare it to that of unprotected routing. For our experiments, we use actual ISP network topologies collected for the Rocketfuel project and three research network topologies.