Abstract: A two-phase virtual machine (VM) fault-tolerance allocation method is proposed according to the constantly change of resources request of VMs and the availability levels of applications. In the first stage, the initial allocation plans of VMs are solved according to the change of resources request of VMs. In the second stage, through live migration of VMs technology and hot standby of VMs technology, the fault-tolerance allocation plan of VMs is solved according to the constantly change of the availability levels of applications. Experimental results demonstrate that the proposed method shows up better performance and higher availability compared with the existing methods.