|
|
 |
| |
|
|
|
Research on Implement Snapshot of pNFS Distributed File System |
|
|
|
PP: 179S-185S |
|
|
|
Author(s) |
|
|
|
Liu-Chao,
Zhang-Jing Wang,
Liu Zhenjun,
Wang-Chao,
Zhang-Jun Wei,
|
|
|
|
Abstract |
|
|
| Snapshot plays an increasingly important role in the online-backup and data-protection of distribute file systems. As a new generation standard of NFS, pNFS doesn’t provide design scheme on snapshot concretely. And the current snapshot implementations depending on their own file system architecture heavily, can’t be used in pNFS. This paper takes a careful study on pNFS structure , adopts a suitable metadata organized solutions, proposes an method to get a consistent state of file system based on COW and a novel source-reserved read/write process flow, which resolve the problem of interaction between client and server, consistence and efficient of creating snapshots. To our knowledge, we firstly implement snapshot function of pNFS in Linux kernel 2.6 and the tests which shows creating snapshot has little affect on performance of pNFS. Snapshot is quick to create(less than 1 second) and can be created frequently (one snapshot per second) without serious performance degradation(less than 10% degradation to write bandwidth) |
|
|
|
|
|
|
|
