Optimizing workflow scheduling in cloud computing: A mixed-integer programming and adaptive multi-objective particle swarm optimization approach with stochastic energy considerations

Abstract

Cloud computing offers users access to both physical and virtual resources, with an emphasis on workow scheduling to optimize Quality of Service (QoS) parameters like deadlines and budgets. The main challenge in cloud computing is nding optimal scheduling solutions that balance multiple constraints and objectives, especially given the uncertain energy consumption of resources. To address this, the present work introduces a stochastic mixed integer programming (SMIP) model that captures the uncertainty in resource energy consumption and aims to minimize cost, makespan, and energy consumption. Further, the paper proposes an Adaptive Multi-Objective Particle Swarm Optimization (AMOPSO) method, recognizing that workow scheduling is an NP-complete problem best tackled by metaheuristic approaches. AMOPSO iteratively renes solutions to achieve near-optimal solutions. Comparative experiments demonstrate AMOPSO's effectiveness against other models like MOPSO, Hybrid multi-objective Particle Swarm Optimization (HPSO), and Chaotic Squirrel Search Algorithm (CSSA) to solve the workow scheduling problem on an Infrastructure as a Service (IaaS) platform. Numerical results highlight substantial improvements in Generational Distance (GD) and Hypervolume (HV) metrics across different workows (Montage, CyberShake, Epigenomics), with improvements ranging from 20.4% to 85.8% in GD and 7.8% to 260.4% in HV. AMOPSO not only enhances performance but also ensures better convergence and spacing among solutions, establishing it as a more effective approach in the domain of cloud workflow scheduling.