We consider dynamic flow restoration in multi-failure scenarios in OpenFlow-based Software-Defined Networks (SDNs). Flexibility of network configuration in these networks makes it possible to dynamically restore flows in case of multilink failures. To re-route the failed flows, network devices such as switches carry out flow operations, i.e., add new flow-entries to the flow-tables. In disaster scenarios where thousands of flows must be restored in a short time, the time required to perform such operations is significant and must be minimized to maintain a carrier-grade network. Shortest-path based techniques do not take into account the number of flow operations (namely, operation cost) and therefore are inefficient for disaster scenarios. We incorporate the operation cost into the flow restoration problem, and formulate the problem of finding a path 1) with the lowest path cost with capped operation cost, 2) with the lowest possible operation cost, and 3) with the minimum operation cost amongst all the paths with a Dijkstra-like path cost. We propose optimal algorithms with Dijkstra-like complexity for the second and third problems. The simulation results with European Reference Network (ERnet) show that our proposed methods on average can reduce the number of required flow operations up to 15% while the path cost rises less than 3%.
展开▼
