An Approach to Minimizing Downtime Induced by Taking Live Snapshot of Virtual Cluster
Ruiqi Sun, Jie Yang, Zhiqiang He · 2013
Cloud computing has emerged as a new computing paradigm which is adopted by more and more application services based on cluster computing. Live snapshot technology is usually used to ensure the reliability of data and high availability of the running virtual machines. However, existing virtual machine live snapshot technologies are not enough to minimize the downtime induced by taking live snapshot for an entire virtual cluster consisting of a large number of virtual machines. An ordinary approach is taking live snapshot of all virtual machines belonging to the virtual cluster concurrently. But this approach may induce a long downtime of an entire virtual cluster which consists of a large number of virtual machines distributed in a small number of hosts. To address this challenge, this paper surveys the factors which raise the downtime of virtual machines induced by taking live snapshot, and defines five metrics used to determine the concurrency strategy of taking live snapshot of different virtual machines. Based on these metrics, we propose a novel approach named Selection Concurrency Strategy which aims to minimize the downtime of virtual machines, and our experiments show that the downtime of virtual cluster consisting of a large number of virtual machines distributed in a small number of hosts using our approach is 33% shorter than that under ordinary approach.