Live Migration Performance Modelling for Virtual Machines with Resizable Memory

The quality of services is a major concern to the users on cloud computing platforms. Migrating non-stop services across hosts on a cloud is important for many critical applications. Live migration is a useful mechanism for virtual machines to minimize service downtime. Thus, the costs of iterative pre-copy and downtime experienced by a real time application should be as short as possible. In this paper, a general model for live migration is presented. An effective strategy for optimizing the service downtime under this model is suggested. The performance of live migration is evaluated for virtual machines with resizable memory. Our results show that the service downtime can be significantly reduced by dynamically resizing memory size according to the working sets of running tasks