Controlling migration performance of virtual machines according to user's requirements

Abstract

Live Virtual Machine (VM) migration is an important technology for cloud management and guaranteeing a high service quality. However, existing studies are mainly focusing on improving migration performance but without much consideration of user's requirements. Also, the migration process is transparent to cloud managers and users. Once a migration is initialized, it is uncontrollable as desired. In this paper, we quantitatively analyze the migration process of pre-copy and figure out the relationship between migration performances and dynamic influence factors. Then, by taking user's requirements into consideration, a migration control algorithm is proposed through tuning the dynamic factors. The migration convergence problem of pre-copy is solved as well with the performance control algorithm. We base our study on Xen platform, and the experimental results verify the efficiency of our migration control algorithm.