Research And Application Of Virtual Machine Optimization Based On Hybrid Virtualization Technology

With the development of computer technology and the popularity of network applications, system virtualization technology is widely used in data centers to consolidate the existing server hardware resources and maximize their utilization by building virtual machine cluster. This could save considerable power consuming in data centers, which is consistent with the concept of green computing. It is critical to minimize the performance overhead incurred by the system level virtualization when deploying the virtual machine cluster. Paravirtualization and hardware-assisted virtualization are the prevailing solutions of system level virtualization techniques, but both exert advantages and disadvantages, respectively, in the performance issue. Therefore, optimizing system virtualization to improve virtual machine performance is one of the current hot topics in both academic and industrial researches of cloud computing infrastructure technology. Based on the analysis and comparison of various categories of system level virtualization techniques, this paper proposes the hybrid virtualization technology, which enhances virtual machine performance by merging the superiority of paravirtualization and hardware-assisted virtualization. The design principle of hybrid virtualization technology is to run paravirtualized operating system in the hardware-assisted virtual machine container. CPU virtualization and memory virtualization adopt the hardware-assisted virtualization solution, while I/O virtualization adopts the optimized paravirtualization solution.The prototype of hybrid virtualization is implemented on the virtualization platform of Xen. Several benchmark tools are utilized for the experiments of performance evaluation to prove the effectiveness of hybrid virtualization, which improves the performance of virtual machines. The experiment results indicate that the hybrid virtual machine manifests nearly 50% and 30% improvement in the micro performance and macro performance, respectively, when running the CPU intensive workloads. Meanwhile, the hybrid virtual machine is more efficient than the hardware-assisted virtual machine when running the I/O intensive workloads. Standing the throughput level, the hybrid virtual machine is capable to save more than 16% CPU utilization compared with that of the hardware-assisted virtual machine.Orienting to the electrical and electronics virtual experiments, the backend simulative computing platform is the processing core of the virtual experiment platform. In order to meet the requirement of tremendously concurrent online manipulation of virtual experiment fronted applications, distributed computing and virtual machine cluster are used to deploy the large scale calculation matrix. Since the hybrid virtualization optimizes the virtual machine and promotes the capability of virtual machine cluster, the backed simulative computing platform can accomplish the tasks more efficiently. The research of this paper is the key technology of the educational virtual experiment platform, and strongly application driven.