Data protection & recovery
Annualized failure rate (AFR)
Annualized failure rate (AFR) is the estimated percentage of devices in a population that will fail over one year of operation. A drive model with an AFR of 1 percent is expected to lose about one unit per hundred in service each year. It is the standard measure for comparing the reliability of drives and for planning around their failure.
How is AFR calculated?
From field data, AFR is the number of failures divided by the total time in service, scaled to a year. If 10,000 drives run for a full year and 120 fail, the AFR is 1.2 percent. When drives enter and leave the fleet during the period, service time is counted in drive-days or drive-years so that partial years are handled correctly. Manufacturers also publish an AFR derived from accelerated testing.
How does AFR relate to MTBF?
Mean time between failures (MTBF) expresses the same information as a time. The two convert into each other.
| MTBF (hours) | Approximate AFR |
|---|---|
| 500,000 | 1.75 percent |
| 1,000,000 | 0.88 percent |
| 1,500,000 | 0.58 percent |
| 2,500,000 | 0.35 percent |
Why do observed rates differ from the specification?
Published figures describe a model under stated conditions. Real fleets vary. Failure rates follow a curve over a drive's life, with a small early peak from manufacturing defects, a long stable period and a rise as drives wear out, often after five years or so. Temperature, vibration, workload and firmware all play a part, and individual models and production batches can deviate sharply from the average. Operators of large fleets therefore track AFR per model and per age group from their own data.
How is AFR used in storage design?
AFR is an input to three calculations. Durability: combined with data rebuild time and the protection scheme, it gives the probability that enough drives fail together to lose data, as described under data durability. Operations: a fleet of 2,000 drives at 1.5 percent AFR will see about 30 failures a year, which sets the stock of spares and the frequency of replacement visits. Refresh planning: when an older generation's AFR climbs well above the fleet average, replacing it becomes cheaper than servicing it. Mixed hardware clusters covers comparing generations.
Frequently asked questions
What is a typical AFR for hard drives?
Fleet studies commonly report averages between about 1 and 2 percent, with wide variation between models and ages.
Does AFR apply to SSDs?
Yes. Flash drives have their own failure rates and also a wear limit that depends on how much is written to them.
Does a low AFR mean data is safe?
No. Any drive can fail. AFR tells how often to expect failures, and redundancy is what keeps data safe when they happen.
Related terms
- Data durability: the calculation AFR feeds.
- Data rebuild: what each failure triggers.
- Flash storage vs HDD: two media with different failure behavior.
- RAID: redundancy sized around expected failures.
- Hardware depreciation: the accounting side of aging hardware.