Task Load Characteristics Analysis And Scheduling Algorithm In Cloud Environment

With the growing maturity of cloud computing technology,more and more applications have been deployed to the cloud environment,making it characterized by the variety of tasks and user needs,heterogeneity of resources,and huge amount of data.These features make higher requirements for task scheduling and resource deployment in the cloud environment:not only to meet the needs of users to obtain high quality of service,but also to ensure the maximum benefit of the cloud service provider.However,no matter how to meet the needs of users or ensure maximum benefits,we can not avoid the impact of load changes.Different load and dynamic changes of the same load may cause poor performance of task scheduling and inaccurate deployment of resources.Therefore,through the load analysis,the two-way improvement of task scheduling and resource deployment to complete the bilateral optimization of users and service providers is a problem to be solved in the research of cloud computing.In order to solve the problems mentioned above,this paper first studies the related characteristics of the load in detail.By modeling the Markoff load,the three operators,which can reflect the load characteristics,are extracted,which are persistence,recurrence time and entropy.Among them,persistence reflects the relative stability of load flow;reappearing time reflects the bursting of load flow;entropy reflects the unpredictability of load flow.Secondly,taking the real access of the Wikipedia Wikinews plate as the experimental data set,combining the different load operator values,this paper establishes the function relation of each load characteristic operator with the number of the required virtual machines under different values,and proposes a load aware resource deployment model through the integration of the function relations.To improve the deployment of resources.The experimental results show that the average absolute error value of this model is only 2.6%and the maximum absolute error value is 3.8%compared with the number of resources obtained by simulation experiments in the case of resource utilization rate of 90%and the load rejection rate of 0.The minimum absolute error value is 0%.The results show that the load aware resource deployment model can effectively predict the number of resources needed according to the load characteristics.Because the load aware resource deployment model is optimized for the deployment of cloud resources,the efficiency of cloud task execution is not considered,which belongs to unilateral optimization.Therefore,based on the optimization of cloud resource deployment and genetic algorithm,this paper proposes a bilateral optimization task scheduling strategy and the corresponding system architecture to improve the cloud task scheduling mechanism.The scheduling algorithm is evaluated on the CloudSim simulation platform.The experimental results show that the scheduling algorithm proposed in this paper can effectively improve the efficiency of the task completion.Compared with the FCFS(First Come First Serve)algorithm and the greedy algorithm of the CloudSim platform,the proposed algorithm has a distinct advantage.While meeting the requirements of the task,the number of resources is kept in a proper number and meets the requirements of the resource utilization.