Multi-goal Load Balancing Research In Virtual Environment

The virtualization technology is concerned because of its independence and high resource utilization and easy management in recent years. The common virtualization software Xen that has the advantages of the less resource-intensive, high-performance and open source is often used to build the high-performance and high reliability virtualization environment. The overload of the physical machines will seriously affect the performance of the running virtual machine and even have them go down int the virtualization environment based on the Xen, therefore, how to achieve load balancing of the physical machines in the virtualized environment has become a hot research topic.The paper described the history of the development of virtualization technology, the research background, the research status and the significance, and analysised the basic principle of the virtual machine system and the key technologies of the achievement, focused on the Xen virtualization, live migration technology, Libvirt. This paper proposed a multi-goal load balancing method drawing on ideas and methods of the related literature taking into account three factors of the migration overhead, load balancing of the physical machine’s CPU and memory resources. First, it uses the time series model to predict the overloaded physical machine, second, it uses the minimize migration overhead algorithm to select the virtual machine to be moved for the overloaded physical machine, third, it uses the descending best-fit algorithm to select the target physical machine for the virtual machine to be moved, finally, it uses the live migration command of the Xen to trigger the migration.To validate the performance of the multi-goal load balancing method, it is used to manage the cluster of the real teaching virtualization system based on the Xen. The experimental results show the method can achieve the load balancing of CPU and memory of the physical machines of the cluster with a lower migration overhead in a very short period of time and effectively improve the performance of the cluster.