HEFT-Guided Elite PSO for Workload Scheduling in Cloud Computing

Abstract

Efficient workflow scheduling remains a critical challenge in cloud computing due to precedence constraints among dependent tasks and heterogeneous nature of virtual machines. The traditional heuristic scheduling approaches often fail to produce optimum solutions as the workflows scheduling an NP-hard optimization problem. The conventional swarm intelligence algorithms suffer from premature convergence and slow searching problems. So, this paper proposes a HEFT-Guided Elite Particle Swarm Optimization (EPSO) algorithm to improve workflows scheduling problem in heterogeneous cloud environment. The proposed algorithm combines the HEFT policy for VMs initialization, elite class solution preservation to move towards better scheduling solutions and adaptive strategies for self-decision-making during evolution process for exploration and exploitation. The proposed EPSO algorithm is evaluated using four standard workflows datasets namely CyberShake, Montage, Inspiral and Epigenomics. The performance of the proposed EPSO algorithm has been compared with APSO, IPSO, PSOJaya, PSO and GWO algorithm. The EPSO algorithm found better in terms of makespan, cost, degree of imbalance i.e. load distribution among VMs and through put.