Improved many-objective particle swarm optimization algorithm for scientific workflow scheduling in cloud computing

Abstract

Optimized scientific workflow scheduling can greatly improve the overall performance of cloud computing. As workflow scheduling belongs to NP-complete problem, so, meta-heuristic approaches are more preferred option. Most studies on workflow scheduling in cloud mostly consider at most two or three objectives and there is a lack of effective studies and approaches on problems with more than three objectives remains; because the efficiency of multi-objective evolutionary algorithms (MOEAs) will seriously degrade when the number of objectives is more than three, which are often known as many-objective optimization problems (MaOPs). In this paper, an approach to solve workflow scheduling problem using Improved Many Objective Particle Swarm Optimization algorithm named I_MaOPSO is proposed considering four conflicting objectives namely maximization of reliability and minimization of cost, makespan and energy consumption. Specifically, we use four improvements to enhance the ability of MaOPSO to converge to the non-dominated solutions that apply a proper equilibrium between exploration and exploitation in scheduling process. The experimental results show that the proposed approach can improve up to 71%, 182%, 262% the HyperVolume (HV) criterion compared with the LEAF, MaOPSO, and EMS-C algorithms respectively. I_MaOPSO opens the way to develop a scheduler to deliver results with improved convergence and uniform spacing among the answers in compared with other counterparts and presents results that are more effective closer to non-dominated solutions.