Metaheuristic Techniques for Controller Placement in Software-Defined Networks

Abstract

Software defined networks provides a global network view with centralized management. To maintain the network configuration, multiple controllers are required. The network performance depends on the optimal number of controllers and their placement.Due to the large size and complexity involved, meta-heuristic algorithms are the probable choice that can solve the problems in an acceptable amount of time. This paper addresses the controller placement problem in SDN by using two meta-heuristic techniques. The objective is to find optimal number and location of controllers in the network while minimizing the propagation latency and optimizing cost. A random approach is adopted for initial placement of controllers. The assignment of switches to the controllers is done based on their shortest distance. Then an efficient genetic algorithm based placement solution is proposed to find the optimal location of controllers which minimizes cost. Our proposed genetic algorithm is different from the standard genetic algorithm in terms of generation and replacement for determining the best cost and the optimal location of controllers . The same experiment is done on simulated annealing (SA) and random method. For evaluation purpose, we have used some real topologies. The results of our enhanced GA performs better compared to simulated annealing and random placement approach.