Reliability-aware swarm based multi-objective optimization for controller placement in distributed SDN architecture

• Main Authors: Ibrahim, Abeer A.Z., Hashim, Fazirulhisyam, Sali, Aduwati, Noordin, Nor K., Navaie, Keivan, Fadul, Saber M.E.
• Format: Article
• Published: Elsevier BV 2023
• Online Access: http://psasir.upm.edu.my/id/eprint/110561/; https://linkinghub.elsevier.com/retrieve/pii/S2352864823001700
• Institution: Universiti Putra Malaysia

The deployment of distributed multi-controllers for Software-Defined Networking (SDN) architecture is an emerging solution to improve network scalability and management. However, the network control failure affects the dynamic resource allocation in distributed networks resulting in network disruption and low resilience. Thus, we consider the control plane fault tolerance for cost-effective and accurate controller location models during control plane failures. This fault-tolerance strategy has been applied to distributed SDN control architecture, which allows each switch to migrate to next controller to enhance network performance. In this paper, the Reliable and Dynamic Mapping-based Controller Placement problem (RDMCP) in distributed architecture is framed as an optimization problem to improve the system reliability quality, and availability. By considering the bound constraints, a heuristic state-of-the-art Controller Placement Problem (CPP) algorithm is used to address the optimal assignment and reassignment of switches to nearby controllers other than their regular controllers. The algorithm identifies the optimal controller location, minimum number of controllers, and the expected assignment costs after failure at the lowest effective cost. A metaheuristic Particle Swarm Optimization (PSO) algorithm was combined with RDMCP to form a hybrid approach that improves objective function optimization in terms of reliability and cost-effectiveness. The effectiveness of our hybrid RDMCP-PSO was then evaluated using extensive experiments and compared with other baseline algorithms. The findings demonstrate that the proposed hybrid technique significantly increases the network performance regarding the controller number and load balancing of the standalone heuristic CPP algorithm.