Research On Task Prediction And Scheduling Strategy In Cloud Service

Cloud service is the service provided through the cloud,which refers to the network as the medium to obtain the required services in an on-demand and easy to expand way.Relying on data center,cloud services not only provide users with a platform to deploy applications and store data,but also provide users with convenient cloud computing services.Through the accurate prediction of tasks,the operators can have a macro understanding of the tasks to be run in the data center in the future.At the same time,using a reasonable and efficient task scheduling strategy can effectively improve resource utilization,improve service quality and improve user satisfaction.This dissertation takes cloud service as the starting point,data center as the foothold,follows the idea of from outside to inside,and takes task prediction and task scheduling as the starting point to carry out the forward-looking research and internal operation optimization of data center.Prediction work can promote the operation of data center from passive mode to active mode,and scheduling work can optimize task allocation and improve resource utilization Rate and so on.The detailed research work is as follows:1.Aiming at the high non-linear and non-stationary characteristics of cloud task sequence in task prediction,an integrated prediction method consisting of empirical mode decomposition and time convolution network is proposed.Firstly,the task sequence is decomposed into several subsequences by empirical mode decomposition,which weakens the interaction between subsequences of the task sequence.Then,the time convolution network is used to predict the task subsequence,so as to accurately predict the trend of each task subsequence.Finally,the prediction results of all task subsequences are reconstructed,and obtain the real task prediction values.The experimental results show that the method has high prediction accuracy.2.In view of the diversity of tasks,the complexity of data center operating environment and the heterogeneity of cluster scale,a task scheduling method based on IMCSA is proposed.In order to solve the problems of low precision and premature of crow search algorithm,the original random initialization method is optimized by using the idea of reverse learning based on the standard crow search algorithm to increase the coverage of the understanding space.The crows with better adaptability are more likely to be selected.The idea of cross mutation in genetic algorithm is used to improve the way of updating crow's position,so that the crow with better fitness can be retained.Experimental results show that the performance of the improved algorithm is improved.On this basis,it is combined with task scheduling model,and compared with other scheduling algorithms on CloudSim platform.The experimental results verify the effectiveness of IMCSA algorithm improvement,effectively reduce the completion time of tasks,and make the system load balancing.