Designing A Hybrid Heuristic Mechanism To Achieve Optimal Load Balancing In Software Defined Networks

Abstract

The architecture of Software Defined Networks (SDN) is based on separating the control plane from the data plane, this separation helps making the network to be directly programmable and scalable, and integrating new services such as load balancer, firewall and many other services to the network. In this paper, we propose a hybrid heuristics mechanism to improve performance of SDN controller by using a dynamic load balancer based on both Depth First Search Algorithm (to find paths) and Ant Colony Optimization (to choose the optimal path). The proposed load balancer has designed and run dynamically within the control plane in order to take advantage of comprehensive view of the topology provided by the controller, which ensures load balancing in real time, avoiding bottlenecks and network stability at peak times and under conditions of topology changes or network scaling. In addition, we suggest a new mechanism to solve the issue of handling Address Resolution Protocol (ARP) within the controller to ensure that the proposed load balancer works dynamically, discovers the topology changes, and makes the lowest cost routing decision.