HotRestore: a fast restore system for virtual machine cluster
Lei Cui, Jianxin Li, Tianyu Wo, Bo Hu Li, Renyu Yang, Yingjie Cao, Jinpeng Huai · 2014
A common way for virtual machine cluster (VMC) to tolerate failures is to create distributed snapshot and then restore from the snapshot upon failure. However, restor-ing the whole VMC suffers from long restore latency due to large size snapshot files. Besides, different latencies would make the virtual machines not start at the same time. The prior started virtual machine (VM) thus cannot communicate with the VM that is still restoring, leading to the TCP backoff problem. In this paper, we present a novel restore approach called HotRestore, which restores the VMC rapidly with-out compromising performance. Firstly, HotRestore re-stores the single VM through an elastic working set which prefetches the working set in a scalable win-dow size, thereby reducing the restore latency. Sec-ond, HotRestore constructs the communication-induced restore dependency graph, and schedules the restore line to mitigate the TCP backoff problem. Lastly, a restore protocol is proposed to minimize the backoff duration. Additionally, a prototype has been implemented on QE-MU/KVM. The experimental results demonstrate that HotRestore can restore the VMC within a few second-s whilst reducing the TCP backoff duration to merely dozens of milliseconds. 1