Administering volume snapshots 313

Cascaded snapshots

to read data from an older snapshot that does not exist in that snapshot, it is obtained by searching recursively up the hierarchy of more recent snapshots.

A snapshot cascade is most likely to be used for regular online backup of a volume where space-optimized snapshots are written to disk but not to tape.

A snapshot cascade improves write performance over the alternative of several independent snapshots, and also requires less disk space if the snapshots are space-optimized. Only the latest snapshot needs to be updated when the original volume is updated. If and when required, the older snapshots can obtain the changed data from the most recent snapshot.

The following points determine whether it is appropriate for an application to use a snapshot cascade:

■Deletion of a snapshot in the cascade takes time to copy the snapshot’s data to the next snapshot in the cascade.

■The reliability of a snapshot in the cascade depends on all the newer snapshots in the chain. Thus the oldest snapshot in the cascade is the most vulnerable.

■Reading from a snapshot in the cascade may require data to be fetched from one or more other snapshots in the cascade.

For these reasons, it is recommended that you do not attempt to use a snapshot cascade with applications that need to remove or split snapshots from the cascade. In such cases, it may be more appropriate to create a snapshot of a snapshot as described in the following section.

See “Adding a snapshot to a cascaded snapshot hierarchy” on page 337 for an example of the use of the infrontof attribute.

Note: Only unsynchronized full-sized or space-optimized instant snapshots are usually cascaded. It is of little utility to create cascaded snapshots if the infrontof snapshot volume is fully synchronized (as, for example, with break- off type snapshots).

Creating a snapshot of a snapshot

For some applications, it may be desirable to create a snapshot of an existing snapshot as illustrated in Figure 9-5.