Check out the new USENIX Web site. next up previous
Next: Consistency Points and NVRAM Up: WAFL - Write Anywhere Previous: WAFL - Write Anywhere

The Snapshot Facility

Snapshots are a primary benefit of WAFL's write anywhere approach. A snapshot is an on-line, read-only copy of the entire file system. The file server is able to copy the entire file system in just a few seconds. The copy uses no additional disk space until files are changed or deleted due to the use of copy-on-write. Only as blocks in the active file system are modified and written to new locations on disk does the snapshot begin to consume extra space.

Snapshots can be used as an on-line backup capability allowing users to recover their own files. Snapshots can be taken manually, and are also taken on a schedule selected by the file system administrator; a common schedule is hourly snapshots taken every 4 hours throughout the day and kept for 24 hours plus daily snapshots taken every night at midnight and kept for 2 days. With such a frequent snapshot schedule, snapshots provide much more protection from accidental deletion than is provided by daily incremental backups. WAFL allows up to 20 snapshots to be kept at a time. How long snapshots can be kept depends on how quickly the file system changes, but ranges from a few hours to a few weeks. Snapshots also simplify backup to tape. Since snapshots are read-only copies of the entire file system, they allow self-consistent backup from an active system. Instead of taking the system off-line, the system administrator can make a backup to tape of a recently created snapshot.

A WAFL file system can be thought of as a tree of blocks rooted by a data structure that describes the inode file. The inode file in turn contains the inodes that describe the rest of the files in the system including meta-data files such as the free block bitmap. WAFL creates a new snapshot by making a duplicate copy of the root data structure, and updating the block allocation information. Since the root data structure is only 128 bytes, and since only the block allocation information needs to be copied on disk, a new snapshot does not consume significant additional disk space until a user deletes or modifies data in the active file system. WAFL creates a snapshot in just a few seconds.

Due to the presence of snapshots, the question of whether a block is free or in use is more complicated than in file systems without snapshots. Typically, when a file is deleted all of the blocks holding data for that file may be marked as free. In WAFL, however, if the file is in a snapshot, then the blocks must not be reused until all of the snapshots holding the file are deleted. Accordingly, WAFL's free block data structure contains 32 bits per block in the file system instead of the 1 bit per block that is needed without snapshots. The live file system as well as each snapshot is allocated a bit plane in the free block data structure; a block is free only when it is not marked as belonging to either the live file system or any snapshot.


next up previous
Next: Consistency Points and NVRAM Up: WAFL - Write Anywhere Previous: WAFL - Write Anywhere
Logical vs. Physical File System Backup
OSDI '99