An Adapted Nondominated Sorting Genetic Algorithm III (NSGA-III) With Repair-Based Operator for Solving Controller Placement Problem in Software-Defined Wide Area Networks

Optimum controller placement in the presence of several conflicting objectives has received significant attention in the Software-Defined Wide Area Network (SD-WAN) deployment. Multi-objective evolutionary algorithms, like Non-dominated Sorting Genetic
Algorithm II
(NSGA-II) and Multi-objective Particle Swamp Optimization (MOPSO), have proved helpful in solving Controller Placement Problem (CPP) in SD-WAN. However, these algorithms were associated with the challenge of scalability (when there are more than three objectives) for optimization in the SD-WAN. Hence, this study proposed an adapted NSGA-III (A-NSGA-III) to resolve the scalability challenges associated with NSGA-II and MOPSO algorithms in the presence of more than three objectives. This study developed and introduced a repair-based operator into the existing Mechanical Engineering based NSGA-III to propose the A-NSGA-III for optimal controller placement in the SD-WAN. The proposed A-NSGA-III, the NSGA-II and MOPSO algorithms were subjected to evaluation using datasets from Internet2 OS3E WAN topology with six objective functions. The Hypervolume indicator, Percentage Coefficient of Variation (PCV), the percentage difference and the Parallel Coordinate Plots (PCP) confirmed that the proposed A-NSGA-III exhibited high convergence and diversification than the NSGA-II and MOPSO algorithms in the presence of scalability challenge (when the number of objective function exceeded three). The result confirmed that the proposed A-NSGA-III solved the scalability challenges associated with the optimal Controller Placement in the SD-WAN. Hence, A-NSGA-III was recommended over NSGA-II and MOPSO algorithms, subject to the confirmation usage conditions.