Backup and Recovery

When we talk about computerized information systems we must remember that hardware fails. Hard disks do crash. And when ... not "if' ... when that happens, data will be lost. So, before the data is lost in a hardware failure, you need to make a "backup copy" of it. Historically, this has been done to magnetic tape. It's cheap, it's readily available, and it's widely used. Tape backups can be done on a daily basis, weekly basis, hourly basis, or any other frequency depending on how critical is the data being backed up.

Another reason to make backup copies is human error. Human operators can make mistakes which lead to loss of data or files.

Types of Backup

• Full Backup (copies all the files)
• Incremental Backup (copies only the files that have changed since the last Full or Incremental Backup)
• Differential Backup (copies only the files that have changed since the last Full Backup)

Most systems keep track of creation and modification dates for the files stored on disk. As part of the Backup process the date and time that the file was saved can also be recorded. Incremental and Differential Backups determine which files to save by comparing the current system date and time with the creation, modification, and last-backup times of the files stored on disk.

Once the backup is stored on tape, the question becomes where to store the tape. If you are trying to protect yourself against loss of data from such events as fire or flood which affect your physical plant, you do not want to store your backup media right next to the source computer so that both can be destroyed in the same disaster. Backup media should be stored in a secure, off-site location. Yet, it must be available in the event it is needed for a restore operation.

Factors to Consider

• Volume of data to be backed up (often huge)
• Where is the data? (local or over a network)
• Tape hardware: capacity and performance
• Time window
• Network bandwidth and the impact of other traffic
• Operator support available/desirable?
  • operators present or lights out?

Operating system utilities (VMS BACKUP, Unix tar) are adequate for backup of conventional files, but may not be suitable for backup of data stored using modern database products. It is very important to save a database when it is in a "consistent" state (e.g., not in the middle of a transaction when multiple updates have not completed). Because of the size and complexity of modern database systems, operating system utilities designed to save at the file level are replaced by transaction-aware database backup utilities specific to the database management system product itself.

Further, in today's business environment the system and the database may be in use well beyond normal business hours. It is no longer a safe assumption that the database will be quiet for long hours overnight so that backup can take place. Finding a time window to do backup or any other kind of system maintenance in today's 24x7 workplace can be a challenge. And, on one hand, while the window is shrinking because of the demand for system availability to interactive users, the size of the window needed is growing because more and more huge amounts of data are being stored!