Research Of Load Balancing Scheduling For Multiple Data Center

Cloud computing brings not only opportunities for the data center, but alsochallenges. In order to provide services for users efficiently, resources in Cloud datacenter need to be scheduled reasonably. To maintain high performance of physicalserver, the utilization of CPU, memory, storage etc. should be kept below a certain level,and thus the load of physical machines should be balanced. The situation that somemachines are overloaded while some machines are idle should be avoided. Similarly, inmulti-data center environment, to get a higher overall service level, the data centershould try to maintain low resource utilization. Therefore, balancing the load betweendata centers can improve service level.The thesis introduces a Cloud resource sharing model in a multi-data centerenvironment, and implements a Cloud resource sharing system compliance with thismodel. This thesis presents the detailed design and implementation of the system,including the main modules, communication mechanism among modules and savingmethods of system information. Then according to the characteristics of tasks scheduledby the system including considering lifetime and resource sharing, the thesis presents amulti-data center load balancing mechanism used in the model, including both offlineand online situation.Finally, the thesis contrasts the algorithms newly proposed with the classic loadbalancing algorithms, including Round-robin, Random, List Scheduling and LongestProcessing Time First. The indicators taken into consideration are: the number ofrefused requests, the overall imbalancing level, the maximum data center utilization andthe minimum-maximum ratio of comprehensive utilization of the data center. Theexperimental data shows that compared with other algorithms, the proposed algorithmhas obvious advantages in terms of the indexes referred in this thesis.