Research And Implementation Of The Virtual Machine Cluster Deploying And Dynamic Migration

The cloud computing platform is a complex systems, and the virtual machine can implement many requirements, such as the application is deployed regardless of the under platform, the physical resource is isolated for many application, and the virtual machine can be migrated. Aimed to integrate dispersed virtual machine and to satisfy the requirement of large scale applications, it is necessarily to implement the virtual machine cluster management. With the efficient cloud virtual machine cluster management strategy, the system can improve cloud platform maintenance quality, so the strategy is an important part of the cloud platform.However, the cloud platform has huge resources scale, various resource style, complex structure and the wide geographical distribution. At the same time, its application has a flexible client request, and the resource style of application is various, the system load is changed dynamic. All the actors challenge the current virtual machine cluster management strategy of cloud platform. Together with the support of national nucleus high technology research group, this paper focuses on two researches. On the one hand, based on the resource classification of virtual cluster, this paper proposes a new deploying technology of virtual machine cluster. On the other hand, based on the performance predicting algorithm, this paper proposes a new migration strategy of virtual machine.To satisfy the requirement of large scale virtual machine, this paper presents the concept of rapid image transfer strategy and the cluster resource type classification. In deploying process of the virtual cluster, to implement the fast response of client’s cluster deploying request, this paper utilize the image parallel multiple pass method and reduce the transfer time of cluster image. In this paper, the cluster resource style is divided as computing intensive clusters, storage intensive clusters, flow intensive clusters. Based on the cluster resource classification strategy, a new deploying node selection algorithm has been implemented, and the selection algorithm supports the balancing of system performance. The result of the experiment verifies that the image transfer strategy can achieve rapid deployment of virtual machine cluster. Moreover, the cluster deploying algorithm can achieve the comprehensive utilization of system resources.To solve the inefficiencies of traditional dynamic migration algorithm, this paper propose a performance predicting algorithm that is named as OSVD(Optimal Singular Value Decompostion). The predicting algorithm comprehensively analyzes the current system resource usage, and it predict the node resource usage of the next period. The predicting result can effectively improve the efficiency of dynamic migration algorithm, and avoid the system interference of sudden load. Based on the OSVD predicting algorithm, the dynamic migration algorithm adapts the migration trigger strategy, the selection strategy of candidate virtual machine, and the migration target selection strategy. The experimental data show that the OSVD predicting algorithm can make a accurate predictions, and the dynamic migration algorithm can achieve load balancing and handle the sudden load.Finally, the paper introduces the general framework of the JTangPlus cloud computing platform, the crucial process of virtual machine cluster deploying and virtual machine dynamic migration. Then the JTangPlus platform final effect is demonstrated.