Every IT-dependent company has some form of data backup in place today. While a growing number of organizations are implementing creative new technologies by leveraging disk-based backups and remote restore of an entire data set or a regular database refresh in a test environment also help promote a certain sense of preparedness.
Unfortunately, the few that actually try restoring all critical data as part of a DR test often realize that results are far from what was expected. Backups are typically scheduled to prevent multiple large backups from running simultaneously, and full backups are staggered throughout the week. So, tape backup infrastructures are usually designed to meet those requirements. However, once a massive restore operation is initiated, the limitations of the backup environment quickly surface.
During backups Data interleaving: Avoid using data interleaving during backups since it slows down the restore performance. If data from multiple systems must be interleaved on the same media, systems of similar criticality should be grouped together. Ideally, data from critical systems should be stored on as few tapes as possible.
Full and incremental backups: Many backup products provide the ability to create full backups from incremental backups. This feature should be leveraged to reduce the number of incremental restores and limit data dispersion.
Prioritization: Data must be categorized based on application restore priority. Combined with the ability to create tape pools provided by many backup software products, high-priority data can be isolated on easily identifiable tape media. Prioritizing also helps control the number of simultaneous restore processes. Single copies of backup tapes should also always be avoided.
At the recovery site
Number of tape drives: The number of drives in tape subsystem should at least match the number of systems that must be restored simultaneously in the event of a disaster.
System and network performance: Ensure that systems and disk subsystems on which data will be restored have enough I/O capacity to handle the potentially large amount of data associated with a full restore. There must also be sufficient bandwidth when restoring across the LAN. While full backups may have previously been staggered throughout the week to accommodate network and backup window limitations, recovering from a disaster may require multiple simultaneous full restore streams.
Include DR testing in your IT budget: Your operational IT budget should include provisions for at least one large scale test restore per year. Testing enables you to uncover limitations and exposures and address them before it's too late.
For more information:
About the author: Pierre Dorion is a certified business continuity professional for Mainland Information Systems Inc.