Task Scheduling Strategy Based On Evolutionary Algorithms In The Cloud Computing

In the recent years, cloud computing has become a hot topic among the world. Cloud computing mainly provides infrastructure service, platform service and application service via network. The user sends the request to the cloud platform and cloud platform will firstly split the request to many small tasks then send these tasks to the distributed server to compute. At last, the cloud platform collectsthese results that returned by the distributed servers and integrates these results then returns to the user. At the same time, there will be many users request the service from cloud platform, so the efficient task scheduling strategy is essential for cloud computing platform. It is a critical issue in cloud computing that how to. relationalallocation the tasks to ensure the smooth running and make the tasks completing time shortest.Under the normal circumstance, the main goal of the task schedule is reducing the overall execution time and the user expense using the cloud platform and improving the resource utilization to ensure the benefits of cloud computing service provider. It’s necessary to improve the utilization rate of transport resources and guarantee the QoS requirements of users, so the task scheduling in cloud computing is a NP complete problem. As evolutionary algorithms such as PSO, GA algorithm have a good performance in solving the NP problem, so the evolutionary algorithm can be introduced to solve the task scheduling problem.In this paper, we have carried on the following researchesin the context of cloud computing:(1) The main work is doing the relative academic research considering various factors affecting the task schedulingefficiency and establishing the question model that meets the actual cloud computing environment requirements of the task scheduling problem.(2) In this paper, we research the traditional task scheduling algorithm such as PSO, GA and point out the advantages and disadvantages using the theory to prove, then analysis the phenomenon of the local convergence of PSO algorithm and point out the convergence is caused by the decrease of population diversity. At last we propose the PSO-CM algorithm that introduces the crossover and mutation strategy of GA algorithm into PSO algorithm and point out the PSO-CM algorithm can achieve the global convergence through the theoretical proof. (3) In the last section of this paper, we realize PSO-CM algorithm by Matlab and compare the performance between the PSO-CM algorithm and the traditional PSO algorithm on the cloud computing task scheduling. The result shows that:the PSO-CM algorithm can keep the population diversity in the process of task scheduling and can achieve the global convergence.