Resource scheduling for cloud computing workloads using an improved genetic-simulated annealing hybrid algorithm
Abstract
Efficient resource scheduling is essential for reducing service latency and improving resource utilization in cloud data centers. This study develops an improved genetic-simulated annealing hybrid algorithm (GASA) for task-to-virtual-machine allocation under heterogeneous workload conditions. A task-resource scheduling model is formulated by jointly considering makespan, resource utilization, and load balance. The genetic algorithm provides population-level global exploration, while simulated annealing is embedded in the evolutionary process through the Metropolis acceptance rule to preserve search diversity and refine candidate schedules. Adaptive adjustment of genetic operators and elite retention are further used to improve convergence stability. Experiments are implemented in CloudSim using Google Cluster, Alibaba Cluster, and synthetic workloads. The reported results show that the proposed GASA achieves shorter execution time, higher resource utilization, better load balance, and a higher scheduling success rate than standalone GA and SA across the tested workload scales. The study provides a practical hybrid-optimization framework for resource allocation in dynamic cloud environments.