Mean time between failures

MTBF is the expected time between two consecutive failures of a repairable system, assuming the system is restored to an as-good-as-new condition after each repair. Mathematically, it is the reciprocal of the failure rate (λ) when failures follow an exponential distribution: MTBF = 1/λ. In practice, it is estimated as the total accumulated operating time divided by the number of failures observed during that period.¹ The metric assumes a constant failure rate, which corresponds to the flat portion of the bathtub curve, where early-life and wear-out failures are excluded. This assumption is valid for many electronic components but may not hold for mechanical systems with wear-out modes.²

MTBF is widely used in industries to set maintenance intervals, design redundancy, and predict system availability. For example, in aerospace, MTBF values guide the scheduling of component replacements to prevent in-flight failures. In data centers, MTBF helps design fault-tolerant architectures, as seen in the Google cluster studies that report failure rates of individual servers.³ Standards such as MIL-HDBK-217 and Telcordia SR-332 provide methods for predicting MTBF based on component stress and environmental factors. These predictions are often used in procurement contracts, where manufacturers must demonstrate that their products meet specified MTBF thresholds.

MTBF is often misinterpreted as the expected lifetime of a product, but it is not a guarantee of survival time. For a system with a constant failure rate, the probability of surviving to the MTBF is only about 37%, meaning many units fail before that point.¹ Additionally, MTBF does not account for the severity of failures or the time to repair, which is captured by mean time to repair (MTTR). Availability is calculated as MTBF/(MTBF+MTTR), so a high MTBF does not ensure high availability if repairs are slow.⁴ The metric also assumes that failures are independent and that the system is as good as new after repair, which may not hold in practice.

Beyond the standard definition, MTBF has niche applications and historical quirks. In the early days of computing, the ENIAC had an MTBF of about 2.5 hours, which drove the development of more reliable vacuum tubes.⁵ In the automotive industry, MTBF is used for components like brakes and transmissions, but the failure rate is not constant, so Weibull analysis is often preferred.⁶ Another edge case is the use of MTBF in software reliability, where it is applied to bug occurrence rates, though the underlying assumptions are debated. Additionally, some organizations report MTBF values that are statistically meaningless due to small sample sizes or improper data collection, leading to the phrase "lies, damned lies, and MTBF."²

Failure rate (λ)

The frequency with which a system or component fails, expressed in failures per unit of time.

Mean time to failure (MTTF)

The average time to failure for non-repairable systems, often confused with MTBF.

Mean time to repair (MTTR)

The average time required to repair a failed system and restore it to operation.

Bathtub curve

A graphical representation of failure rate over time, showing early-life, constant, and wear-out phases.

Weibull analysis

A statistical method for modeling failure times that can handle varying failure rates.