Research On Cloud Task Scheduling Algorithm Based On The Multi-objective Optimization

As the most widely used commercial distributed computing technology,cloud computing has a large scale and wide user groups,requiring the server system to schedule and manage a variety of the cloud tasks frequently.Aiming at the scheduling of execution time and cost is a NP-hard multi-objective optimization problem,however,the current task scheduling under the cloud environment is generally the execution time or the execution cost of single objective optimization with constraint conditions,incompletely meeting the complex cloud systems with load balancing.Given above motivations.As a consequence,the research on cloud task scheduling algorithm based on multi-objective optimization plays a significant role in the cloud computing.This dissertation analyses the characteristics of cloud computing,improving the cloud task model,setting up the hybrid cloud task scheduling model,selecting the execution time,execution cost and load balance as the optimization goals,optimizing the scheduling process,the main research work is as follows:1)Aming at the characteristics of diversified needs for the cloud tasks,analysing the basic concepts of cloud computing,system structure and technical characteristics,improving the cloud task model,introducing the concept of multi-objective optimization.2)Aming at the hybrid cloud task scheduling requirements,selecting the execution time,execution cost and the load balancing that cloud service providers focus on as the cloud task scheduling optimization goals,setting up the cloud task scheduling model based on multi-objective optimization to handle the hybrid cloud task.3)Aiming at the dynamic changes of the cloud computing and the characteristics of the cloud task scheduling,improving the ant colony genetic algorithm and put forwarding an improved adaptive genetic ant colony algorithm of the multi-objective cloud task scheduling based on it,the algorithm combines genetic algorithm's good global search ability and the advantages of ant colony algorithm's high accuracy,avoiding the deficiency of the late genetic algorithm's disadvantage of failing to solve local search and the lack of initial pheromone of the ant colony optimization algorithm,having the obvious advantages in what cloud users pay attention to--the execution time and execution cost problem as well as the whole cloud system load balancing level index.4)Aiming at the genetic algorithm and the large cloud task scheduling,improving the memetic algorithm,introducing two local heuristic algorithm: mountain climbing algorithm and tabu search algorithm,making full use of genetic algorithm's better global optimization ability and mountain climbing algorithm and tabu search's local optimization ability,avoiding the inadequacy of the genetic algorithm in the late local and mountain climbing algorithm and tabu search's insufficient global search ability,meanwhile,the improved memetic algorithm based on the tabu search shows a higher execution efficiency and better load balancing in large-scale cloud task scheduling environment in the CloudSim simulation platform.