ITC-LM: A smart iteration-termination criterion based live virtual machine migration

Abstract

Live migration of virtual machines (VMs) plays an important role in grids, clouds and datacenters, and has become the cornerstone of resource management in virtualized systems. The efficiency of live migration depends on the downtime, total migration time and total transferred data. However, while migrating a memory-intensive VM, XEN/KVM often do many useless iterations of memory copy in order to reach expected downtime which can never be reached, leading to a great deal of useless data transferring and insufferable total migration time. It consumes mass of network bandwidth and CPU resource when transferring memory from one to another node. Hence, a critical task is to determine the optimal time to terminate the copy iteration for live migration. In this paper, we propose a smart iteration-termination criterion based live migration which is termed as ITC-LM, to self adaptively control when to terminate iteration. We have implemented ITC-LM into KVM/QEMU. The improvement is significant, especially when migrate a memory-intensive VM. The experimental results show that, our approach can decrease 50.33% of total transferred data on average without impairing migration downtime. © 2013 IFIP International Federation for Information Processing.