A Time-Series Based Precopy Approach for Live Migration of Virtual Machines

Abstract

As an important technique for load balancing in cloud computing, live migration of virtual machines has received increasing attention in recent years. One key issue of live migration of virtual machine is how to quickly and transparently migrate virtual machines to meet Service Level Agreements (SLAs). In this paper, we evaluate traditional precopy approach, and propose an improved time-series based precopy approach for virtual machine migration. With the time-series prediction technique, we identify frequently updated dirty pages (high dirty pages) in the past and future period more precisely, and transmit them in the last round of iteration, in order to reduce unnecessary, repeated transmission of dirty pages. Doing so can significantly reduce the total migration time. Simulation results show that compared with the traditional approach of Xen, our approach can effectively improve the performance of virtual machine migration, including less number of iterations, less down time and migration time, and fewer pages transferred.