Dynamic Service Function Chaining in SDN-enabled networks with middleboxes

Network functions typically need to be visited in a specific order to meet certain objectives, giving rise to the notion of Service Function Chaining. Software-Defined-Networking enables fine-grained traffic routing optimization while satisfying correct traversal of network functions. In this work, we investigate the problem of maximizing throughput in SDN-enabled networks with respect to service chaining specifications under both traditional and new constraints. Besides the algorithm design, we also derive rigorous performance bounds. In the offline traffic routing case, we propose Traffic-Merging-Algorithm and prove that, although the underlying optimization problem is generally NP-hard, our algorithm can efficiently compute the optimal solution in practical settings. In the online traffic routing case, we propose the Primal-Dual-Update-Algorithm, which comes with a system parameter that trades off the algorithm's throughput competitiveness and its meeting of QoS requirements, and prove that our online algorithm achieves optimal tradeoff. We demonstrate that our solutions can be used to address practical problems by conducting simulation-based evaluation over backbone and data center topologies.