Tape backup operation must be scheduled by considering the number of concurrent backup and/or restore operations and the number of drives in the ETERNUS tape library. The number of backup and restore operations is not in reference to a single Storage Server but instead refers to all Storage Servers.
If the number of drives is insufficient, some backup and/or restore operations that are activated later must wait for idle drives. When a backup or restore operation that was activated earlier is complete and the relevant drive is released, tape I/O processing for a backup or restore operation that has been made to wait begins. In this case, the backup or restore time is calculated by adding the drive wait time to the time during which tape I/O is actually performed.
The following figure shows the backup processing time when three transaction volumes are backed up concurrently in an environment with two drives.
Figure 5.9 Backup processing time relationships
In light of the above, consider the following two points for scheduling:
Tape I/O performance
When more backup and/or restore operations than the number of drives are performed concurrently, a timeout can be defined for the wait for drives. Also, when a wait for drives occurs, upper limits can be defined so that drive allocation is performed evenly among Storage Servers. Refer to "Chapter 3 Designing the Backup Operation" for more information.
