Redundancy means providing a backup for something a system needs. If one component fails, another can take over. In a data center, this helps keep services available when equipment breaks or maintenance is needed. section 13.2.1
Backups can be layered rather than relying on one spare component:
- Power: independent feeds, dual power supplies, uninterruptible power supplies (UPS), and generators provide alternatives during an outage. section 13.2.1 section 13.9
- Connectivity and computing: redundant Internet access and extra servers provide ways to keep services running. section 13.2.1
- Location: a geographically separate data center or a pre-configured alternate site provides somewhere to shift operations if the primary site cannot be used. section 13.2.1 section 7.1
How much redundancy is enough depends on how much downtime an organization can tolerate. Higher availability targets require more complex, costly arrangements. A disaster recovery plan should also specify how to activate redundant systems or shift operations; having backups alone does not make recovery automatic. section 13.2.1 section 7.1
|
Availability
|
Nickname
|
Maximum downtime per year
|
|
99%
|
“Two nines”
|
About 3.5
days
|
|
99.9%
|
“Three
nines”
|
About 9
hours
|
|
99.99%
|
“Four nines”
|
About 52 minutes
|
|
99.999%
|
“Five
nines”
|
About 5
minutes
|
Compare availability targets with their maximum yearly downtime before deciding whether additional redundancy is worth its cost. section 13.2.1
When backups are not enough
A 2016 power-surge outage at Delta grounded more than 2,000 flights. Its recovery plan included failover to backup servers, but recovery took nearly six hours. The incident exposed gaps in redundancy and shows why recovery procedures need regular testing. section 7.4












