An advanced parallel hybrid metaheuristic approach for multi-objective optimization in cloud task scheduling

Abstract

Cloud systems are a crucial actor in providing digital services, including real-time data processing and batch job execution. However, these systems also present significant optimization problems due to their complex structures and functions. Two critical metrics that must be minimized are makespan and energy consumption. Achieving optimal performance, profit, and ustainability is possible with a sophisticated multi-objective optimization approach that can reconcile these two important conflicting goals. This paper proposes a parallel hybrid metaheuristic and multi-objective task scheduling approach that finds optimal solutions by considering the balance between makespan and energy. This hybrid approach combines the strengths of Non-dominated Sorting Genetic Algorithm-2 (NSGA-2), and the Strength of Pareto Evolutionary Algorithm 2 (SPEA2). The most important uniqueness of this proposed parallel hybrid method is that it also makes optimum use of system resources, thanks to the master process it uses in the solution process. To test the success of the proposed method, five different cloud system scenarios with different workloads and resources were used. The outcomes are presented for comparison with the results of four classical multi-objective metaheuristic optimization algorithms. The success of the proposed approach is demonstrated.