Research On Monitoring And Controlling Computer Behavior Through Parallelized Virtual Machine

As the popularity and the emergence of multi-processor servers, traditional virtual machines which can not make full use of the multi-processor resources is not suitable to be used on multi-processor servers. Therefore, developing a parallel virtual machine to simulate a full computer system to investigate its behavior is up to our agenda.This dissertation describes the use of a new parallel virtual machine to simulate,investigate and control a typical computer system. Using Virtual Machine to investigate the behavior of computer systems differs significantly from traditional approaches which could only observe parts of a computer system. This virtual machine parallelized the multi-processor simulation and was able to provide a detailed and factual view into the computer system. The users can get a better observation and understanding of the computer system in a detailed manner using virtual machines.Two key issues were brought about by the parallelization. First of all, when virtual machine is idle, the parallel execution routine would consume much system computing power. This dissertation presents a method which could make the virtual machine convert between serial and parallel simulation modes. When the virtual machine is idle, it will switch automatically into serial simulation mode to reduce overhead. Then it will switch back into parallel simulation mode when the virtual machine is busy. Secondly, the parallelization reduces the authenticity of virtual machine in simulating a real computer system. This dissertation presents a rapid simulating model and an actual simulation model. Moreover, the virtual machine can dynamically switch between these two modes on the requests of users. Users can freely choose the required model. Finally, we provide re-manipulation mechanism of the information out side the virtual machine and show it in graphic format in the client, which help the user to get a better understanding of the computer system more easily.The test result shows that using the"self-adapt parallel/serial conversion"technique, the overload on the real server dropt while introducing little overhead in the virtual machine. Users can get detailed information of the computer behavior in actual simulation and become 2.60 times faster when apply to the fast simulation mode.