MTBF (Mean Time Between Failures)
MTBF is the average amount of uptime between incidents — a measure of how often things break.
Mean Time Between Failures measures reliability frequency rather than recovery speed: on average, how long the service runs before the next failure. It is the total operational time divided by the number of failures over that period. Where MTTR asks 'how fast do we recover?', MTBF asks 'how often do we break?' — together they describe two independent dimensions of reliability.
Why it matters: a high MTBF means incidents are rare, which reduces toil, protects your error budget, and keeps on-call sustainable. Trends in MTBF reveal whether your reliability work is actually addressing root causes or merely getting faster at mopping up the same recurring problems. A falling MTBF is an early warning that something systemic is degrading.
A common misconception is that MTBF predicts when the next failure will occur. It is a long-run average, not a countdown — a service with a 30-day MTBF can still fail twice in one week and then run for two months. It also says nothing about how bad each failure is; pairing it with MTTR gives the fuller picture of frequency and impact.
You raise MTBF by removing causes, not symptoms: blameless postmortems, fixing fragile dependencies, adding redundancy, and resisting the urge to simply restart and move on. Each genuine root-cause fix pushes the average time between failures higher.
MTBF = Total operational (up) time ÷ Number of failures
A service that is up for 720 hours in a month and fails 3 times has MTBF = 720 ÷ 3 = 240 hours between failures.
Frequently asked questions
- What is the difference between MTBF and MTTR?
- MTBF measures how often a service fails (frequency); MTTR measures how quickly it recovers (impact per failure). They are independent — you can have rare but long outages, or frequent but brief ones.
- How do I improve MTBF?
- Address root causes rather than restarting after each incident: run blameless postmortems, fix fragile dependencies, add redundancy, and track whether the same failure keeps recurring.
- Does MTBF predict the next outage?
- No. It is a long-run average, not a timer. A service can fail well before or long after its MTBF; the metric describes overall frequency, not a specific date.
Start monitoring in under a minute
Free forever, no credit card. Upgrade when you need finer intervals.