What is MTBF?
MTBF (mean time between failures) is the average time a system runs between one failure and the next. It is calculated as total operating time divided by the number of failures in the period, and higher is better. Together with MTTR it gives availability: MTBF ÷ (MTBF + MTTR).
Why MTBF matters
MTTR tells you how quickly you recover; MTBF tells you how often you have to. A service that recovers in two minutes but fails twice a day is still a poor experience, and the fix is different: fewer bad deploys, more capacity, or removing a fragile dependency.
Watching MTBF over quarters shows whether reliability work is paying off. The metric comes from reliability engineering for repairable equipment; for online services, a failure should mean an incident that affected users.
How to measure MTBF
MTBF = total operating (up) time ÷ number of failures availability = MTBF ÷ (MTBF + MTTR)
Worked example
Over a 30-day month (720 hours) a service had 3 user-facing failures with 1.5 hours of downtime in total.
- Operating time = 720 − 1.5 = 718.5 hours.
- MTBF = 718.5 ÷ 3 = 239.5 hours, roughly 10 days between failures.
- MTTR = 1.5 ÷ 3 = 0.5 hours.
- Availability = 239.5 ÷ (239.5 + 0.5) = 99.79%, the same as 718.5 ÷ 720.
Result: To reach 99.9% with the same MTTR, MTBF would have to be about 499.5 hours, so the service could fail only about once every three weeks.
Common mistakes
- Including downtime in operating time. Use up time only. Dividing the whole period by the number of failures slightly inflates MTBF.
- Very few data points. Three failures make a noisy average. Report the period and the count next to the number.
- Mixing up MTBF and MTTF. MTTF (mean time to failure) is for things that are replaced rather than repaired. Services are repaired, so MTBF is the right term.
- Counting every alert as a failure. Define a failure as a user-facing incident, or MTBF ends up measuring alert noise.
How Vigil handles MTBF
- Vigil records every incident with its start and end time, so failure counts and downtime per monitor come from real data rather than memory.
- The incident timeline shows MTTR, a dashboard tile shows MTTR over 30 days, and the printable SLA report gives availability against your target for any period.
- The SLA calculator converts an availability figure into downtime per month and per year.
MTBF (mean time between failures): frequently asked questions
What is a good MTBF?
There is no general benchmark for software services, because it depends on the system and on how you count failures. Compare a service with its own history and look at the trend.
How is MTBF different from MTTF?
MTTF is used for parts that are replaced after failing, MTBF for systems that are repaired and put back into service. Online services are repaired, so MTBF fits.
Can I get availability from MTBF alone?
No, you also need MTTR. Availability = MTBF ÷ (MTBF + MTTR), so a long MTBF with very long repairs can still mean poor availability.