Efficient Live Migration of Virtual Machines with a Novel Data Filter

Abstract

Live migration of virtual machines (VM) is useful for resource management of data centers and cloud platforms. The pre-copy algorithm is widely used for memory migration. It is very efficient to deal with the memory migration of read-intensive workloads. But for write-intensive workloads, the pre-copy’s straightforward iteration strategy will become inefficient. In this paper, we propose a novel data filter to improve the pre-copy algorithm in this inefficient situation. In each round of iteration, the data filter forecasts the pages which will be subsequently dirtied, and then filters them from the send list. This prevents the pages from being repeatedly transmitted, thus reducing migration time and bandwidth resource consumption. Meanwhile, the data filter also checks if the previously filtered pages should be re-added to the send list. This ensures that the downtime will not be increased. Experimental results show that the improved algorithm effectively reduces the amount of migrated data, while keeping the downtime at the same level.