A Genetic Particle swarm optimization based Hybrid Scheduling Algorithm for Cloud Computing Resources

Abstract

As the quick advancement of information technology, cloud computing technology has risen rapidly, but the energy consumption and resource waste generated by data centers are also increasing. Therefore, the study analyzed the hybrid scheduling of cloud computing resources. Firstly, a improved particle swarm optimization algorithm-based resource scheduling model was raised to address the initial placement problem of virtual machines. Secondly, considering that the resource requirements of applications in cloud computing environments are dynamically changing, attention mechanisms and whale optimization algorithms were introduced to optimize the bidirectional long short-term memory network and build a resource demand prediction model. The results showed that when the amount of virtual machines was 200, the energy consumption of the improved particle swarm algorithm was 6.19 kW/h. The completion time of the algorithm always did not exceed 2000ms under different numbers of virtual machines and tasks. When the problem size was 500, the proposed resource demand forecasting model tended to converge at around 100 epochs. The prediction accuracy and recall rate of the proposed resource demand forecasting model were 94.35% and 93.62%, respectively. The experiment outcomes indicate the resource scheduling and resource demand prediction effectiveness of the raised model. The outcomes contribute to improving the service quality and effectiveness of the entire cloud platform, and promoting the development of cloud computing technology.