CHAM: A Multi-Population Hybrid Metaheuristic Algorithm for Optimal Controller Placement in Multi-Controller Software-Defined Networks

Software-defined Networks (SDNs) have become prevalent because of their augmentation in state-of-the-art networking technologies regarding scalability and deployment cost. While SDNs incur in actively reducing initial network design and implementation expenses, they leave significant challenges that may influence the overall performance. The Controller Placement Problem (CPP) is among the primary issues with considerable drawbacks on QoS parameters. However, CPP is known as an NP-Hard problem for which several approaches have been reported that introduce solutions. Due to the extreme volatility of the SDN size in terms of the number of switches and controllers, the CPP is deemed to compel extensive studies for a reliable solution. The current paper introduces a metaheuristic algorithm to achieve a solution for controller replacement. Accordingly, the CPP is formulated as an optimization problem, and an influential chaotic-based multi-population hybrid method (CHAM) is designed to find a solution. The proposed method encompasses two distinct algorithms: artificial ecosystem-based optimization (AEO) and marine predators’ algorithm (MPA). These discrete algorithms are then merged using a multi-population strategy. Next, a local search mechanism is introduced to exploit the existing solutions. As the next step, a chaotic neighborhood search mechanism is also provided in the CHAM to search for promising areas in the subpopulations. Finally, a migration procedure is presented to convey solutions between subpopulations. For evaluation purposes, CHAM was employed in ten real-world SDN networks with different sizes and configurations, and the results are compared with nine existing approaches. The experimental results indicate that the proposed method performs affordably well compared to the existing solutions. For example, the proposed method improves network performance by 25% by creating conditions close to related works.