一种用于存储网络软故障恢复的快照技术的设计和实现

Implementation of Snapshot Used for Failure Recovery in Storage Network

与镜像技术相比,快照具有备份和恢复窗口短、性能损失小、容量利用率高等优点,更适合保护因人为失误等软故障造成的数据损失。提出了一种应用于卷管理的快照技术的设计方案。它可以高效地管理一个数据源的多个存储快照,并提供快照的读写、创建、回收等功能。另外针对快照技术物理容错性差的缺点,还提出了几种快照技术和镜像技术相结合的应用方案,较好地解决了存储网络软、硬故障时数据的保护。

Compared with widely employed split-mirror, snapshot can create copies of data almost instantaneously despite the size of volume, have shorter mean time to recovery, need less physical disk space and suffer less performance loss. So it is more suitable for protecting data in storage network against software and user errors. We propose a new implementation of (volume-based) snapshot, which is efficient especially when a volume owns several snapshots. Difference bitmap and snapshot link table are used to manage snapshots. And an access protocol is also presented for processing I/O on base volume and its snapshots. The implementation of snapshot supports writing and reading on base volume or snapshot volume, establishing and deleting snapshot, and recovering base volume from snapshot. Finally, aim at solving the problem that snapshot cannot protect against hardware failure, three deployment patterns, which combine mirror and snapshot, are introduced. With technologies presented in this paper, data in storage network can be protected against not only soft-failure but also hard-failure efficiently.