Traffic engineering in hybrid SDN networks with multiple traffic matrices

Traffic engineering (TE) is an efficient tool for optimizing traffic routing and balancing the flows in networks. Traffic is dynamic, previous TE optimization over a single traffic matrix (TM) have some limitations, because a single TM can have big measurement errors and is insufficient to depict the traffic flucuations. Thus, we consider using multiple TMs to overcome these limitations. With the emergence of Software Defined Networking (SDN), we can route flows more flexibly and better balance the flows over multiple TMs. However, due to the difficulties of full SDN deployment, hybrid SDN networks will be the prevailing architectures in the near future. Therefore, optimizing the routing over multiple TMs in a hybrid SDN network is of great interest.In this paper, we first formulate the problem of TE over multiple TMs and prove its NP-hardness. Next, we propose a heuristic algorithm for optimizing routing over multiple TMs by combining offline weight setting optimization with online splitting ratio optimization. Furthermore, we prove that the routes obtained in our algorithm are loop-free, and we provide an upper and lower bound of our proposed algorithm. Finally, we evaluate our method on measured traffic datasets with three network topologies. The results of extensive experiments demonstrate that the maximum link utilization of a network can be optimized better using our proposed algorithm.