Research On Dynamic Cloud Task Scheduling Algorithm Based On Resource Awareness

With the development and popularization of cloud computing,more and more users feel the convenience of cloud services.How to efficiently schedule cloud tasks has become a major problem that cloud providers need to face when providing services to users.Task scheduling algorithm is the key technology of cloud computing.Aiming at the problem that it is difficult to accurately predict the task execution time in the existing task scheduling methods and can not meet the global priority and task dependency of tasks at the same time,this paper proposes a dynamic cloud task scheduling algorithm TPN-GA based on resource perception,in order to improve the task scheduling efficiency in cloud computing environment.There are several innovations in this thesis:1.A task execution time prediction method based on Elastic Net regression is proposed.The task scheduling data set is constructed by collecting task historical execution data and sensing computing node resources;Outliers in the data set are detected by isolated forest algorithm;The task execution time is predicted by elastic network regression model.Experiments show that the prediction accuracy of this method is higher than that based on BP neural network and other classical regression models.2.A dynamic task scheduling algorithm TP-GA is proposed.Based on the global priority and task dependency of tasks,combined with the prediction results of task execution time,the fitness function,selection operator,crossover operator and mutation operator of genetic algorithm are redesigned to dynamically schedule cloud tasks,improve task scheduling efficiency and shorten the completion time of global jobs.Experiments show that the scheduling efficiency of this method is better than that of traditional genetic algorithm.3.A multi-objective task scheduling algorithm TN-GA is proposed.Based on multi-objective genetic algorithm,fitness function,selection operator,crossover operator and mutation operator are designed.Combined with the prediction results of task execution time,fine scheduling is carried out in the time gap generated by the above dynamic task scheduling process on the premise of meeting the global priority and task dependency,so as to further shorten the global job completion time.Experiments show that this method can further shorten the global job completion time.The TPN-GA algorithm is composed of the TP-GA algorithm and the TN-GA algorithm.Through the analysis of the experimental results,it can be known that the algorithm can meet the needs of dynamic task scheduling in a heterogeneous cloud environment and improve the efficiency of cloud task scheduling.The method proposed in this thesis can mine the value of historical task scheduling data,and improve task scheduling efficiency and profit space for cloud providers.