Backup Recommendations

This page contains recommendations for backups of critical data based on the type of risk.

Most data loss causes problems or increases costs, but backup systems can also be costly. It is therefore important to find the appropriate balance for the data being backed up.

  • To minimize the impact of single drive failure, full redundancy of data storage should be implemented on the primary data store (such as RAID-1 or RAID-5).
  • To minimize data loss due to hardware failure, such as failure of SCSI controllers, RAID controllers, connecting leads, drive arrays and motherboards, a regular backup should be performed to a separate physical machine.
  • To minimize the amount of recent data lost in case a restore is necessary, backup frequency should be increased as far as possible, using a technique such as a near live backup, reducing backup lag time. The younger the backups, the less data lost.
  • To minimize data loss due to deletion, a history of backups should be kept. This can be achieved with a rolling backup or a longer backup lag time. The older the backups, the older the data that can be restored. If only a single complete backup is kept, and data is deleted from the primary data store, then as soon as the backup takes place, the data will also be deleted from the backup (so that it remains a mirror of the live/primary data store).
  • To reduce the impact of physical destruction of the primary data store, a backup should be kept in an off-site location. The further away the backup, the less danger of destruction of the primary and the backup at the same time. However, this usually increases backup age or backup lag time. A suitable combination of near live local backups and off-site backups overcomes this problem.
  • To reduce the risk of problems with the backup system, systems and restore plans should be checked and tested on a regular basis. A backup plan is useless if the data cannot be restored in the way that is expected.
metawerx specific

referring pages