Research On Virtual Machines' Performance Detection Based On Virtual Machine's Characteristic Scenario

Virtualization, which is an old and important technique in computer field, has regained widespread attention in recent years. Using virtualization method can not only help corporations reduce the costs of the construction of distributed systems greatly, but also optimize various factors such as efficiency, availability, convenience and security. In addition, virtualization is also considered as one of the core technologies of cloud computing. It is able to do efficient integration and utilization to public resources of cloud terminal, and it can be applied in mobile terminal as well, providing management of information security.Under this background, it becomes more and more essential to make improvement and innovation to virtualization method. However, due to the problems of performance anomaly detection brought by its own complex structure and the evaluation difficulties of lack of special and convincing performance measurement standards, advanced research of the virtualization technique is getting trickier. During the study of the optimization of XEN's virtual IO, we thinks that if there is a standard feature running scene to illustrate and differentiate the virtual machine running, it will be a key technique to improve and evaluate the algorithm scheduling.Methods of similar idea include Boost microbenchemark for performance testing of vitual 10, XenMon with precise monitoring and analysis capabilities, and VMmark for cross-platform virtualization testing. But there is not a uniform simulation testing tool for various feature running scenes on virtual machines. Moreover, because of the complexity and variability in virtual machine environment, there are a wide range of complicated-running scenes, which requires that the simulation tools have good dynamics and scalability.Virtual machine's characteristic scenario describes the external phenomena (in host machine or dispatching system) of different operating characteristics due to different applications and system event effects in virtual machines. Based on the classification and analysis of virtual machine running scenes, we developed a virtual machine feature running scene simulator. So far it has various functions including simulation of intensive use of CPU, network transactions of a great amount of data, a large number of network requests. Meanwhile, it provides scalable development framework and user programming interface, supporting further expansion of the simulator functions. This essay first introduces the basic concepts associated with performance detecting of virtual machine and present research situation at home and abroad. After that we describe the core framework of the simulator and its extensible parts for developing. Finally, it shows the design of two scenes, which are intensive use of CPU and massive data transmission on network equipment. By completing detecting and analysis of the tools on Xen, it illustrate the availability and practicality of the simulator.