Cloud Computing Resource Allocation Using Hybrid Optimization Techniques

Abstract

Efficient cloud resource allocation requires mapping heterogeneous user tasks to virtual machines (VMs) while simultaneously controlling execution time, service cost, load imbalance, and resource under-utilization. Single metaheuristics can provide useful schedules, but they may converge prematurely or spend excessive iterations exploring weak regions of the search space. This methodology paper proposes a Hybrid Genetic Algorithm-Particle Swarm Optimization with Load-aware Repair (HGA-PSO-LR) model for task-to-VM allocation. Genetic search supplies population diversity and global exploration, particle-swarm refinement exploits promising regions, and a final repair phase relocates selected tasks from overloaded VMs when the multi-objective fitness improves. A reproducible proof-of-concept simulation with 80 heterogeneous tasks and eight VMs is used to demonstrate the evaluation procedure. The proposed model is compared with Round Robin, a greedy earliest-completion heuristic, standalone PSO, and standalone GA. The hybrid method obtains the lowest composite objective score (0.451), reduces makespan from 246.54 s under Round Robin to 137.86 s, and increases average VM utilization from 63.54% to 98.64%. These values are specific to the stated fixed-seed simulation and are presented to demonstrate the methodology rather than as universal cloud-performance claims