Crow Search Based PSO Algorithm For Load-Aware Workload Scheduling In Cloud Computing

Abstract

Efficient task scheduling and load balancing are critical for improving performance in cloud computing environment containing heterogeneous VMs. This paper proposes an Improved Crow Search-Based Particle Swarm Optimization (ICPSO) algorithm that integrates SMIW inertia weight strategy with a Crow Search-based exploration mechanism to improve exploration-exploitation balance and reduce premature convergence of PSO algorithm. HEFT policy is used for common population initialization, while a combining fitness function considers makespan and degree of imbalance (DI). The proposed algorithm is evaluated using CyberShake, Montage, Inspiral and Sipht workflows each containing 1000 jobs with 50 heterogeneous VMs in CloudSim 3.0. The ICPSO algorithm is compared with IPSO, IJPSO and PSO in terms of makespan, DI, throughput and convergence. Experimental results show that the ICPSO algorithm consistently provides better performance across all workloads. Compared with IJPSO (second best algorithm), ICPSO achieves 11.49-36.70% improvement in makespan and 7.31-41.11% improvement in DI with higher throughput. The convergence results further demonstrate improved search capability of ICPSO and reduced premature convergence. These findings confirm the effectiveness of combining SMIW and Crow Search-based exploration for load-balanced cloud task scheduling.