The nines, in time
Availability is the share of time a system is up. Each extra "nine" cuts allowed downtime by a factor of ten, and each one is far harder and more expensive to deliver than the last.
| Availability | Downtime per year | Per month |
|---|---|---|
| 99% (two nines) | 87.6 hours (about 3.65 days) | 7.3 hours |
| 99.9% (three nines) | 8.76 hours | 43.8 minutes |
| 99.95% | 4.38 hours | 21.9 minutes |
| 99.99% (four nines) | 52.6 minutes | 4.4 minutes |
| 99.999% (five nines) | 5.26 minutes | 26.3 seconds |
Two details matter when you read an SLA. First, check the measurement window: 99.9% per month is stricter than 99.9% per year, because a single bad day can't be averaged out. Second, check what counts as downtime: many agreements exclude announced maintenance, and a facility outage and an application outage are not the same thing.
Tier figures, with a caveat
| Tier | Defining idea | Commonly cited availability | Downtime per year |
|---|---|---|---|
| Tier I | Basic capacity, single path | 99.671% | about 28.8 hours |
| Tier II | Redundant capacity components | 99.741% | about 22.7 hours |
| Tier III | Concurrently maintainable | 99.982% | about 1.6 hours |
| Tier IV | Fault tolerant | 99.995% | about 26.3 minutes |
These percentages circulate widely, but the Uptime Institute Tier Standard defines Tiers by infrastructure topology, not by an availability number. A Tier IV design operated carelessly can still go down. Treat the figures as a rough sense of scale, not a promise.
What actually drives uptime
Redundant equipment sets the ceiling. People and procedures decide how close a site gets to it. Uptime Institute's outage research has repeatedly found that human error, especially not following procedures, plays a part in a large share of significant outages. The highest-leverage improvements are usually:
- Written, practiced procedures (SOPs, MOPs, and emergency operating procedures) for every switching and maintenance task.
- Change management so risky work is reviewed, scheduled, and has a rollback plan.
- Preventive maintenance that is actually completed, plus testing of generators, UPS systems, and transfers under load.
- Monitoring and alarm management that surfaces real problems instead of burying them in noise.
- Training, so every technician understands the systems and the consequences of each step.
Think you know your Tiers? Ten questions on redundancy, Tiers, and commissioning.
Take the Tiers QuizCourses that protect uptime
Design for availability, then run the site so it delivers.
Tier Infrastructure Fundamentals
Understand and assess TIER I–IV infrastructure without the formal certification track.
Data Center Operations Management
Run daily operations like the teams behind Tier 3/4 facilities — maintenance, incident response, and SLAs.
Incident Response & Troubleshooting
Systematic troubleshooting and root-cause analysis to cut MTTR.
Disaster Recovery & Business Continuity
Design resilient infrastructure and DR plans that survive real failures — not just tabletop exercises.
Preventive Maintenance Planning
Reduce emergency repairs with a structured preventive maintenance program.
Save 42% with the bundle
Operations & Reliability Bundle
Frequently asked questions
How much downtime is 99.99% availability?
99.99 percent availability, often called four nines, allows about 52.6 minutes of downtime per year, roughly 4.4 minutes per month.
How much downtime does a Tier III data center allow?
The figure commonly quoted for Tier III is 99.982 percent availability, about 1.6 hours of downtime per year. That number is an industry rule of thumb, not part of the Uptime Institute Tier Standard, which defines Tiers by topology (Tier III means concurrently maintainable) rather than by an availability percentage.
Does planned maintenance count as downtime?
It depends on the agreement. Many service level agreements exclude announced maintenance windows, while others count any time the service is unavailable. A concurrently maintainable (Tier III) or fault tolerant (Tier IV) facility is designed so that planned maintenance on infrastructure does not require shutting down the IT load.
For planning and education only. Contract terms and design requirements should come from your agreements and your engineering team.