Research On The Availability Of Virtual Machine Hot Standby Based On Xen

Virtualization technology as a key technology of cloud computing can virtualize a number of virtual machines from single physical machine, in order to run multiple isolated operating systems. The virtualization technology increases the security of system, improves resource utilization. With the development of virtualization technology, the high reliability of the virtual machine has become an important factor to obstruct the development of virtual machine technology. The virtual machine hot standby implements high efficiency, low-cost virtual machines fault tolerance mechanisms, because of the virtual machine itself can transplant across platforms and has nothing to do with the underlying hardware, become an important means for the protection of the high reliability of the virtual machine. At the same time, the suspension of virtual machine and service in the virtual machine hot standby because of data synchronization, resulting in greatly reduced availability of the virtual machine hot standby, therefore, how to improve the availability of the virtual machine hot standby became the problem we concern.This thesis describes the development history and current situation of the virtual machine system, analysis the basic principles of virtual machine system and the implement technologies, lists a few typical examples of the virtual machine system, study the changes of memory, CPU, I/O devices Xen had for the virtualization in-depth. After classification research of virtual machine hot standby in the visible current literature, elaborate on the model of typical virtual machine hot standby Remus, propose the necessity of improving the availability of virtual machine hot standby, and in-depth analysis the availability of various types of virtual machine hot standby currently. This thesis design a virtual machine hot standby based on pre-transfer, propose running a data synchronization when the virtual machine is running, and in order to reduce the repetition rate of pre-transfer and stop-transfer, propose SSNUR algorithm to reduce the burden of the primary xirtual machine through forecast dirty pages and transfer the dirty page will not be used in the near future. This thesis test the downtime of Remus before and after improvement based on the virtual machine memory, load and hot standby intervals, experimental results show that the pre-transfer Remus can effectively reduce the downtime of the original Remus, and increase the availability of original Remus.