Glossary
Disaster recovery testing
Disaster recovery testing rehearses a recovery plan to show whether systems and data come back within their targets. Tests range from a discussion of the plan around a table to switching production to the alternate site, and each level finds problems the one before it cannot.
Five levels of test
| Test | What happens | What it finds |
|---|---|---|
| Plan review | Owners read the plan against the current environment | Outdated contacts, missing systems |
| Tabletop exercise | The response team talks through a scenario | Unclear authority, missing steps, shared dependencies |
| Component test | One system is restored on its own | Backup defects, wrong credentials, restore speed |
| Parallel test | Systems are recovered at the alternate site while production runs | Capacity gaps, ordering errors, network and DNS issues |
| Full interruption | Production is switched to the alternate site | Everything, at the risk of a real outage |
NIST's guide to test, training and exercise programs, SP 800-84 separates discussion-based tabletop exercises from functional exercises, in which staff perform their duties in a simulated setting. Cost and disruption rise down the table, and so does the number of problems found, which is why tabletops tend to run several times a year and full interruptions rarely.
What every test measures
- Time to restore service, against the recovery time target.
- Age of the recovered data, against the recovery point objective.
- Integrity, showing the data is usable and not merely present.
- Blockers, such as a missing password or a two-hour approval, which become the fix list.
Checks on the backups themselves, described under restore testing, cover the data side of the same question.
Why small tests overstate readiness
A restore of a few terabytes finishes in minutes, and scaling that result up is tempting. It does not scale in a straight line. Restore throughput flattens at the slowest link, and approvals, ticket queues and credentials add time that a small test never meets. A test at realistic size, with the human steps timed alongside the copy, gives the number a plan can rely on.