Particle Swarm Optimization with Enhanced Neighborhood Search Algorithm and Probabilistic Crossover Genetic Algorithm for Task Scheduling in Cloud Computing

Abstract

Considering the dynamic and elastic nature of cloud computing services, service providers must provide efficient task-scheduling solutions to accommodate the increasing demands in cloud services while satisfying Service Level Agreements (SLA) cost-effectively. In this thesis, we present two novel task scheduling algorithms in cloud computing: ENS-PSO and PxGA, to minimize the makespan. ENS-PSO improves Particle Swarm Optimization (PSO) by introducing an effective neighborhood search technique. Also, we introduce static and dynamic methods to select ENS-PSO neighborhood search size. PxGA enhances the Genetic Algorithm (GA) by applying a weighted probabilistic approach to the crossover operation. CloudSim toolkit is utilized to evaluate the algorithms in terms of makespan, computational time, degree of imbalance, and energy consumption. The simulation results prove that ENS-PSO and PxGA outperform other classic and recent algorithms. Moreover, at the expense of higher computational time, ENS-PSO outperforms PxGA on the overall makespan by 3-4%.