Disk-to-Disk to Tape

As mentioned earlier in this chapter, disk-based systems have certain disadvantages compared with tape, especially a lack of mobility. Disk-based backup systems can suffer the same failures as primary disk systems. Without off-site backups, system administrators may find themselves with two inert storage systems instead of a viable backup. The rapid restore qualities of disk-to-disk systems, however, are a critical benefit of this technology. For some system administrators, the only way to bring backup windows within service levels is with disk-based backup. By combining disk and tape backup, they may achieve the best of both worlds.

One way of accomplishing this is to have the disk-to-disk backup system periodically dump the contents of the backup set to tape. This has the advantage of not affecting the primary backup's performance. The backup to disk might occur several times a day and serve as the primary restore point, while the tape backup might only occur occasionally and be designed to maintain an off-site capability. As long as the tape backup can be completed by the time the next disk-based backup is to occur, which may be several hours, there are no problems.

A variation on this design is to stage the backup to disk and then use tape for the actual backup. The disk system acts as a large buffer for the tape device. From the point of view of the system, it is a tape backup that can be completed in minutes instead of hours. The disk-based system can also act as a restore point (in essence, a very large cache) for the last backup performed. Previous backups would need to be loaded from tape.

Finally, the tape unit may be used for archive purposes only. Every so often, the contents of the disks are dumped to tapes, which are then carted off-site. This may happen only when data is to be removed from the system, but it still needs to be maintained.

Physically, there are many ways to design a disk-to-disk-to-tape system. The tape system may be fully integrated into the disk-based system and may not even be available directly to any other server or host. Software in the system will automatically handle the transfer of data to tape. This is a common design for smaller, self-contained units. The tape drives may be plugged into a common backplane with the disks or utilize a SCSI or SATA DAS connection.

It is also possible simply to place the disk-based system on the network that connects the backup units to the servers. Now the backup server can choose when and where to back up data to disk or tape. This loose integration has the advantage of flexibility. Some systems may still be backed up to tape, while more critical systems are backed up to disk-based systems. The disk-based backup system is then backed up to tape. If there is a tape library already available on the network, it would continue to be used rather than be replaced.