OptiGSM: Greedy-Based Load Balancing With Minimum Switch Migrations in Software-Defined Networks

Software-defined networks (SDNs) are an emerging technology to improve the scalability and network performance of data-intensive applications with high-volume traffic. Load balancing among controllers with several switch migrations is a major obstacle in a multi-controller SDN, degrading the overall system’s performance. To resolve the above issue, this paper proposes OptiGSM, an optimum greedy-based switch migration approach for load balancing among the controllers. OptiGSM starts the switch migration process from an overloaded controller to an underloaded controller if the mean-square deviation of a load of each controller from the average load of all controllers is greater than a pre-defined threshold. A controller is overloaded if its load is greater than the average load of all controllers and vice-versa. We introduce an algorithm for load balancing and prove that it achieves load balancing using the minimum number of switch migrations. Next, we analyze the time and space complexities of the algorithm. Furthermore, we evaluate OptiGSM using simulation and experimental studies. Numerical results indicate that OptiGSM outperforms conventional approaches in terms of the number of switch migrations, response time, throughput, and delay.