To calculate MTBF as something usable, divide 8,760 hours by the quoted figure: 200,000 hours becomes an annual failure rate of about 4.38%, which for 1,000 switches means roughly 44 expected failures a year and a spare pool of about 10 units — not a 22.8-year service life.
Rayin (Shenzhen Rayin Technology) is a manufacturer of GPON OLTs and industrial Ethernet switches.
The 22.8-year figure is the one everybody quotes, and it is the one that misleads. MTBF is derived from a failure rate, not measured as a lifetime. It describes how often failures occur across a large population, not how long any individual unit will last.
Most reliability arguments in a procurement meeting eventually arrive at the same question: how to calculate MTBF into something the business can plan against, rather than a figure that merely sounds reassuring on a brochure. The rest of this article is that arithmetic, applied to a real fleet size.
MTBF is the reciprocal of the failure rate: MTBF = 1 / λ. Turning hours into years is an arithmetic convenience, not a statement about service life.
200,000 hours corresponds to 4.38% annual failure rate. 300,000 hours corresponds to 2.92%. That is a halving of failures, not a doubling of life.
The spare parts pool is driven by monthly failure count and repair turnaround, not by the cumulative three-year repair count. Ten spares out of 1,000 units is normal.
Two identical MTBF numbers can differ by a factor of several once you know the standard, the ambient temperature and whether the power supply, fan and optics were included.
These three get used interchangeably in conversation and they mean entirely different things.
MTTF (Mean Time To Failure) applies to non-repairable items — a capacitor, an optical module. It is the average operating time until failure.
MTTR (Mean Time To Repair) is the average time to restore service, and it includes fault localisation, spare replacement and verification. It decides how long an outage actually lasts.
MTBF (Mean Time Between Failures) applies to repairable products and is the average time between two failures. Mathematically it is the reciprocal of the failure rate λ, so MTBF = 1 / λ. That single line is the whole MTBF formula, and everything else follows from it.
| Metric | Applies to | What it answers | Typical use |
|---|---|---|---|
| MTTF | Non-repairable components | How long until this part fails | Component selection, derating decisions |
| MTTR | Repairable systems | How long an outage lasts | Spares strategy, service contracts |
| MTBF | Repairable systems | How often failures occur across a population | Reliability comparison, warranty forecasting |
The weight of that last sentence matters. A switch rated at 200,000 hours MTBF can fail tomorrow or run for a decade. What the number says is that across a large installed base, one failure occurs per 200,000 device-hours on average.
Take a year as 8,760 hours (365 x 24). Three conversions, side by side:
| Conversion | Result | What it is good for |
|---|---|---|
| 200,000 ÷ 8,760 | ≈ 22.8 years | Building intuition only — not a lifetime |
| 8,760 ÷ 200,000 | 4.38% annual failure rate | Engineering decisions |
| 730 ÷ 200,000 | 0.365% monthly failure rate | Spare parts sizing |
Now the comparison that actually changes a purchase decision. 100,000 hours MTBF corresponds to 8.76% annual failure rate. 300,000 hours corresponds to 2.92%. So a data sheet moving from 150,000 to 300,000 hours moves the expected annual failure rate from 5.84% to 2.92% — the failure rate halves. The life has not doubled, because the figure never described life in the first place.
Field conditions: a city deploys 1,000 industrial switches. The vendor quotes 200,000 hours MTBF and the warranty runs three years, which is 3 x 8,760 = 26,280 hours.
Step one, expected failures per unit over three years. 26,280 ÷ 200,000 = 0.1314 failures. Using the exponential distribution properly gives 1 − e^(−0.1314) = 0.1231. At low failure rates the two are close; take 0.13.
Step two, expected returns across the fleet. 1,000 x 0.1314 ≈ 131 units over three years, or about 44 a year.
Step three, the spare pool. The common industry practice is to hold 0.5% to 1% of the installed base: 1,000 x 1% = 10 units. That is not in conflict with 131 cumulative repairs, because spares are a rotation stock for fast replacement. Failed units come back into the pool once repaired, so as long as turnaround is shorter than the failure interval, ten units keep the rotation running.
Step four, put it into the operating budget. 131 repairs at roughly ¥300 each for logistics, labour and testing comes to about ¥39,300 over three years. Moving to a 300,000-hour model gives 1,000 x 26,280 ÷ 300,000 ≈ 88 repairs, about ¥26,400, a difference of roughly ¥13,000. If the better model costs ¥50 more per unit, 1,000 units cost ¥50,000 more up front — an equation that does not balance within two or three years.
The lesson is not that MTBF is useless. It is that a reliability figure only carries meaning once you put it on a specific fleet size and a specific time axis. A comparison of one unit against another shows nothing. A comparison of 1,000 units over three years separates them immediately. Anyone who wants to know how to calculate MTBF for their own site only needs the four steps above and the fleet size.
Variable one is the standard behind the number. MIL-HDBK-217F is the earliest widely used stress-analysis method, built on US military component data from the last century. It is conservative on modern integrated devices, so the resulting figures often come out low — which reads as worse reliability. Telcordia SR-332 is the telecom approach and has been the mainstream since the 1990s, with different modelling of components, temperature and stress. The same hardware put through both methods can produce noticeably different numbers. A third category is field statistics gathered by the vendor from shipped units, which is a different basis again. Any MTBF calculation is only as good as the standard it names, which is why the first question in the table below is the one that matters.
Variable two is ambient temperature. Both MIL-HDBK-217F and SR-332 use temperature acceleration models, and the common approximation is that failure rate roughly doubles per 10°C rise, corresponding to an activation energy around Ea ≈ 0.7 eV. On that approximation, the same device at 40°C has a derated failure rate about 2.8 times that at 25°C — a 15°C gap, or about 1.5 of those 10°C steps.
Variable three is scope. Does the MTBF include the power supply? Does it include pluggable parts like fans and optical modules? Does it include chassis cabling and connectors? Different answers produce very different results. A fanless unit and a fanned unit on the same board will normally score differently, because a fan is a mechanical part and its failure rate is usually higher than that of board-level components.
| What the data sheet says | What to ask | Why it matters |
|---|---|---|
| MTBF 200,000 hours | MIL-HDBK-217F or SR-332? | The same hardware can differ by a factor of several |
| No ambient temperature stated | 25°C or 40°C? | A 15°C gap maps to roughly 2.8x |
| No scope stated | Power supply, optics, fans included? | Including fans usually worsens the figure significantly |
| Only "mean time to failure" | MTBF or MTTF? | MTTF belongs to non-repairable items |
| No confidence level | 60% or 90%? | The same data gives different results at different confidence |
Failure rate over time roughly follows a bathtub shape with three sections.
Infant mortality. The first few hundred hours after shipment carry a higher failure rate, driven by poor solder joints, component defects and assembly stress. This section can be screened out deliberately: burn-in before shipping, powering units for a period to expose the ones with latent defects, at the cost of test bench time and electricity.
Useful life. Failure rate is approximately constant, and this is the section MTBF describes. Its length depends on component process and operating temperature.
Wear-out. Approaching end of life, mechanical parts (fans, relays), electrolytic capacitors and laser diodes in optical modules begin to degrade and failure rate climbs. This is where the rule of thumb about replacing industrial equipment every five years comes from.
The value of burn-in is that it moves first-section failures into the factory. For a buyer the questions are direct: is there burn-in before shipment, for how long, at what temperature. A 24-hour burn-in and a 72-hour burn-in produce visibly different first-year field return rates.
The bathtub curve reliability model is also why a plausible-looking MTBF number can hide a wear-out problem: a constant failure rate assumption says nothing about the section where the fans and capacitors start to go. Bathtub curve reliability thinking separates the three phases for exactly that reason — the number you are quoted covers the middle one only.
Availability is A = MTBF ÷ (MTBF + MTTR). An MTBF calculation on its own is only half the picture, because the number that decides how long a site is actually down is the repair time, not the failure interval.
Take 200,000 hours MTBF and 4 hours MTTR: A = 200,000 ÷ 200,004 ≈ 99.998%, which is about 10.5 minutes of downtime a year.
Now take MTTR of 24 hours, where an engineer has to travel and a spare has to be couriered: A = 200,000 ÷ 200,024 ≈ 99.988%, and annual downtime rises to about 63 minutes.
Doubling MTBF moves availability in the third decimal place. Moving MTTR from 4 hours to 24 hours multiplies downtime by five. That is why on an industrial site a spare pool and a response commitment usually return more than pushing MTBF from 200,000 to 300,000: the first takes MTTR from 24 hours down to 4, the second only improves the expected failure count within a few hours.
To push availability further, the tool is redundancy rather than a better number on the data sheet: dual power supplies, ring protection, dual uplinks. Those make a single failure invisible to traffic, which is equivalent to MTTR of zero and worth far more than the arithmetic on the brochure.
| # | Question | Good answer |
|---|---|---|
| 1 | Which standard is the MTBF based on? | Named down to MIL-HDBK-217F or the Telcordia SR-332 revision |
| 2 | What ambient temperature does it assume? | 25°C or 40°C, or explicitly labelled as measured field conditions |
| 3 | Does it include power supply, fans and optics? | Each item addressed; anything excluded quoted separately |
| 4 | Is there burn-in before shipment, and under what conditions? | Duration, temperature and how far the unit is powered up |
| 5 | What is the actual in-warranty return rate? | Field data is more informative than a theoretical figure |
Question five is the most valuable. A theoretical MTBF can be tuned by methodology; a return rate cannot. A unit quoting 200,000 hours with a 5% actual annual return rate and a unit quoting 150,000 hours with a 1.5% actual return rate are not comparable on paper, and the second one is the easier product to own.
Rayin builds its industrial switches and optical access products around fanless designs, which removes the component with the highest mechanical failure rate. The product line and technical specifications are on the industrial Ethernet switch range page, and the reliability design of the access side — mini GPON OLT and XPON ONU — sits under PON access.
When you ask us for an MTBF figure, ask for the four pieces of context at the same time: standard, temperature, scope and whether burn-in is part of the production flow. A number without them is not a specification, it is a marketing line.
Does 200,000 hours mean the device will fail once every 22 years?
No. The 22.8 years is just hours converted into years to make the magnitude feel concrete. The correct reading is: annual failure rate around 4.38%, so 1,000 units produce about 44 expected failures a year, which averages to one failure per unit every 22.8 years. The two statements are statistically equivalent, but the second does not suggest a service life ceiling.
Why do two units rated at the same 200,000 hours behave so differently in the field?
Three common reasons. The standard and temperature basis differ, and a figure computed at 25°C placed in a 45°C cabinet can double or more in real failure rate. The scope differs, and a figure that includes fans and optics will look worse. And the operating conditions differ — voltage swings, humidity, dust and vibration all accelerate failure, and none of them appear in any MTBF formula. Put your actual cabinet temperature into the quotation request and ask for the number at that temperature.
How many spares should actually be held?
Start at 0.5% to 1% of the installed base, then adjust for repair turnaround. Turnaround is the critical variable: if a failed unit takes 30 days to come back and the monthly failure count is close to the pool size, the pool empties. The rule of thumb is spares ≥ monthly failures x repair turnaround in months x 1.2. With 1,000 units and 44 failures a year, monthly failures are about 3.7; one month of turnaround needs at least 5 spares, three months needs about 14.
Does burn-in screening really reduce field returns?
It reduces first-year returns, specifically the infant mortality portion. Powering units for a period before shipment exposes poor solder joints and early component defects in the factory rather than at the customer site. For a buyer it is worth pinning down duration, temperature and how far the unit is powered up — four hours at room temperature and 48 hours at elevated temperature screen out very different proportions. The cost lands in the unit price, so it is a trade of price against early return rate.
Better MTBF or redundancy — which deserves the budget?
It depends on the cost of a failure. If a failure means the office network is down for a few minutes, spend on MTBF and return rate. If a failure stops a production line or loses CCTV recordings, redundancy pays more: dual supplies, ring protection and dual uplinks keep traffic running through a single failure, which is equivalent to driving MTTR to zero and delivers something no MTBF improvement can match. Most industrial projects end up doing both, in the order redundancy first, numbers second.
How to Read a Hardened Switch Spec Sheet Without Getting Fooled
Managed Industrial Ethernet Switch TCO: The Real 5-Year Cost
About the author: Sara — Sara is a Customer Manager at Rayin with over 10 years of experience in the communications field. In her free time, she enjoys badminton and swimming.
About Rayin: Shenzhen Rayin Technology Co., Ltd. — Company Profile
