Research On Resourcescheduling And Load Balancingbased On Cloud Computing

With the rapid development of the Internet and information technologies,the data volume and the corresponding computation requirements have increased quickly.The computing power of traditional computation models cannot meet the needs of users,thus the supply and demand of network resources are imbalanced.Cloud computing enables people to configure,schedule,and utilize resources more flexibly and can solve the problem above.In cloud computing,resource scheduling is a NP-hard problem,because large-scale task scheduling is very complex and uncertain with the increase of users.It is a key to study how to properly allocate and ensure the resource load balance.In this thesis,the task scheduling in cloud computing is divided into two parts,i.e.,task-to-resource scheduling and resource monitoring.Based on the research of user tasks,this thesis classifies user tasks and then establishes the corresponding input models.In the process of task resource scheduling,a genetic algorithm is extended,and a dual adaptive function is added,which takes both efficiency and fairness into account.The performance of the proposed algorithm is verified by experiments.In terms of resource monitoring,the thesis introduces a combination of dynamic migration and dynamic allocation for resources.The dynamic migration is based on the usage of monitored physical resources.When the usage of physical resources crosses the maximum threshold or the minimum threshold,the resources of virtual machines are migrated according to the selection and allocation strategies,which can satisfy the users' service protocol,ensure resource load-balancing and reduce energy consumption.When any parameter of a resource exceeds the pre-defined threshold,the dynamic allocation of resources will be executed.And how much resources to be allocated is determined by the number of remaining tasks and the processing capacity of resources.The allocation strategy proposed in this thesis can ensure the rationality of resource allocation and improve the efficiency of task execution.Finally,R_Migrate_Extend(dynamic migration and allocation of resources)proposed in this thesis is compared with R_NoMigrate,R_Migrate and R_NoMigrate_Extend on the CloudSim simulation platform.Experiments show that R_Migrate_Extend can satisfy the users' service protocol and reduce the energy consumption slightly.The goals of improving the quality of services for users,ensuring resource load-balancing and reducing energy consumption have been reached.