Performance of Metaheuristic Algorithms for the Controller Placement Problem in SDN

Software Defined Networks (SDN) have the potential to improve network resource utilization through its logically centralized architecture. However, the growing interest on its large-scale deployment has raised issues regarding how to better place one or more (physical) SDN controllers under conflicting network performance metrics, which naturally leads to multiobjective optimization problems. This paper presents a performance comparison between two metaheuristic optimization algorithms: the Non-dominated Sorting Genetic Algorithm II (NSGA-II) and the Pareto Simulated Annealing (PSA) by taking into account both latency and bandwidth in the objective function. Both algorithms present good results in terms of accuracy, processing time, and lower utilization of computational resources. In particular, the NSGA-II algorithm showed better convergence properties and a better exploration of the search space.