An assessment of nature-inspired metaheuristic algorithms for resilient controller placement in software-defined networks

Abstract

Software-defined Networking (SDN) offers flexibility and programmability, making it a desirable option for modern network architecture. SDN provides numerous benefits to network administrators due to its centralized control architecture. This allows network administrators to manage and configure the network from a single location, making it easier to manage and automate network tasks. In a network of multiple controllers, the control plane’s resiliency may impact the overall system’s performance. In a controller failure scenario, switches must be reassigned to other active controllers with adequate capacity. Thus, we define a resilient controller placement (RCP) as an optimization problem. The aim is to design physically distributed and redundant controllers to manage switches with varying resilience levels. The propagation latency may increase due to the reassignment, increasing the network cost. The objective is to determine the number of controllers required, their positions, and the allocation of the network nodes to a particular controller while reducing the average propagation latency and cost in meeting the capacity constraint of the controller. Due to the wide area network (WAN) structure, four nature-inspired metaheuristic algorithms are proposed namely, simulated annealing (RCP-SA), moth-flame optimization algorithm (RCP-MFO), particle swarm optimization (RCP-PSO), and grey wolf optimization algorithm (RCP-GWO). These algorithms are evaluated on three network datasets to determine the optimum controller number and their positions. The experimental results show that RCP-GWO performs better than RCP-SA, RCP-MFO, RCP-PSO, and the Random methods.