Introduction

     Computer data on disk can be lost because of software or hardware failures. Typically, a file is accidentally removed or an entire disk just stops working. To minimize the effects of such failures, vital data can be periodically copied and stored elsewhere. If the backups are done correctly and preserved properly, after any failure data can be restored to its state as of the previous backup.

     The electronic and physical distance between the data and its backups can be important. If the original data and its backups share the same electronics, then any failure of the electronics that causes original data to be lost can also cause the backups to be lost. For example, rather than storing a backup of a file in the same directory in the same filesystem on the same disk, it is better to store the copy in another directory, better still in another filesystem, even better on another disk, and better still on another machine. All these may share the same physical room in a building that could be destroyed in a disaster, so the best location for backups is offsite; i.e. miles away.

     In recognition of the importance of disaster recovery following destruction of entire buildings, ITC backups are stored offsite from machines containing the original data. Backup data is written over the network, eliminating the need for transport of media which risks damage to the media. Individual files or entire filesystems may be restored if corrupted or lost, or previous versions of them as they existed at some particular time may be recalled. Again, the network eliminates the need for media transport, making data available quickly for restores.

     Many ITC Unix backups are written to tapes mounted in drives by a robot. Day or night the robot mounts tapes faster and more reliably than a human operator. Some backups are written to RAID arrays to allow for faster backups and restores. Tapes, the robot, the arrays and associated machines are located in an unmanned locked room, environmentally-controlled to prevent degradation of hardware.