Research On Hybrid Particle Swarm Algorithm And Applied In Cloud Computing Task Scheduling

Cloud computing is an emerging technology; it is the development of distributedcomputing, parallel computing and grid computing. Virtualization technology is one of the key technologies in the cloud computing which uses a mature virtualization technology make all kinds of resources into aresources’pool, and provide on-demand service to users on the Internet.Cloud computing is a user-centric business model, the core of this model is "apply as needed, amount billing", in this model, two of users most concerns are the tasks processing time and the tasks’implementation costs. The cost minimization with due dates in cloud computing workflow is an intractable problem. Taking the characteristics in cloud computing of pay-per-use and resource virtualization into account, in this paper, we present a QoS-based hybrid particle swarm optimization (GHPSO) to schedule applications to cloud resources. The main research work of this paper includes:(l)Research on cloud computing technology and commonly used heuristic algorithms, and establish the mathematical model of cloud workflow task scheduling based on the "pay for use" and resource virtualization features.(2)On consideration of that the standard particle swarm algorithm is only applicable to continuous field problem, in this paper, we propose an improvement strategy. By introducing crossover and mutation of genetic algorithm is embedded into the particle swarm optimization algorithm, so that hybrid particle swarm algorithm proposed in this paper can play a role in the discrete problem.(3)In addition, variability index, changing with the number of iterations, is proposed to ensure that population can have higher global search ability during the early stage of evolution, without the premature phenomenon. In order to improve the search accuracy, hill-climbing algorithm was introduced into the particle swarm algorithm.(4)Finally, the experiments were designed to compare hybrid particle swarm algorithm with standard particle swarm from different environment and point of view to get their features. The simulation results show that thehybrid particle swarm algorithm achieves better performance than standard particle swarm algorithm; optimization capability of hybrid particle swarm algorithm increased about10%. Execution time of hybrid particle swarm algorithm is only about10%of the standard particle swarm.