Efficient real-time scalable make-before-break network re-routing

Dynamic traffic in optical networks leads to capacity fragmentation, which significantly reduces network performance. Consequently, non-disruptive re-routing has attracted a lot of attention from network providers as it dramatically improves the amount of traffic that can be granted with a good quality of service. The recent development of flexible and software-defined networks is urgently calling for faster real-time reconfiguration algorithms at all network layers, while being highly efficient in terms of optimized reconfiguration of connections. We propose a very efficient non-disruptive bandwidth defragmentation heuristic (called mbb_df_h) with a make-before-break (MBB) strategy at the user layer. In order to assess its accuracy, we also develop a near- optimal two-phase defragmentation process. First, we formulate an exact optimization model for the decomposition scheme in order to compute the best re-routing, i.e., with minimum bandwidth requirements. Second, we investigate whether the resulting re-routing can be reached with a MBB policy; if not, we exhibit the best possible MBB re-routing. We conduct extensive numerical experiments on several data sets. We first run experiments on small to medium data sets so that the solution efficiency of the mbb_df_h heuristic can be compared with the two-phase optimization models. Next, we focus on the real-time scalability of the mbb_df_h heuristic on large realistic data sets.