A Hybrid NSGA-II for Solving Multi-Objective Controller Placement in SDN

Unlike traditional networks in which both control and data planes are tightly coupled on the same boxes, SDN decouples control and data planes. At the moment, this architecture is facing various challenges such as reliability, resiliency, scalability, and availability that should be considered in its future designs. One of the most important issues to address these challenges is the problem of controller placement, i.e., the deployment of a desired number of controllers within a network so that some requirements, which may be conflicting, are satisfied. Therefore, based on the fact that various types of objectives should be taken into consideration, this problem can be regarded as a multi-objective combinatorial optimization problem (MOCO). Hence, a single optimal placement for these competing objectives could not be achieved and decision makers need to look for an appropriate trade-off among them. An exhaustive evaluation of all possible placements can be performed well for small and medium sized networks. However, considering realistic time and resource constraints, heuristic approaches are needed to adapt and implement for large scale or dynamic networks whose properties change over time. For this purpose, a heuristic algorithm called hybrid NSGA-II is introduced which yields faster computation times and needs much less memory to perform. The results carried out in Matlab 2013b on the internet2 topology showed the efficiency of the proposed method.

Keywords:

Multi-objective Combiatorial Optimization , Pareto frontier , SDN , Controller Placement Problem , Hybrid NSGA-II