Krisztin - stock.adobe.com
HDD specs: Assess MTTF, AFR and UER to get the drives you need
In part one of a two-part series, we look at key hard disk drive specs, including mean time to failure, annual failure rate and unrecoverable error rate, and how they relate to each other
Hard disk drives (HDDs) deliver high storage capacities at low cost, but HDD models can differ considerably.
Capacity and price are not the only criteria that should determine the choice of HDD. Other specifications such as reliability and ability to handle challenging operating conditions must also be taken into account.
HDDs are still the most common form of data storage and often the best-suited to storing increasing volumes of data. HDD technology is still being developed and is significantly cheaper per unit of capacity than flash storage.
This article is the first of two that provide an overview of the most important HDD specifications.
In this piece, we will look at mean time to failure (MTTF), annual failure rate (AFR) and unrecoverable error rate (UER), and how they differ between different classes of drives.
The second article will dig into sustained data rates, Serial ATA (SATA) versus serial-attached SCSI (SAS), block write sizes, on-drive security and how manufacturers increase drive density.
HDDs specified for different applications
But not every HDD is suitable for every application. Drives are specifically tailored to different areas of application and differ in terms of interface, performance and reliability.
PC HDDs, for example, are usually designed for a daily operating time of eight to 16 hours and an annual workload of 55TB (terabytes), so they can’t be used in network-attached storage (NAS) systems or servers. Even though they are cheaper, they would wear out quickly due to continuous use and the probability of failure would increase.
Enterprise HDDs must support 24/7 operation and cope with higher workloads. For NAS and surveillance drives, it is generally 18TB annually, while for enterprise drives, it is 550TB per year.
Manufacturer specifications for MTTF only apply if this value, known as the “workload rate” or “rated workload” in data sheets, is adhered to. If the rated workload is exceeded, the HDD won’t fail immediately, but the probability of failure heightens and MTTF decreases.
MTTF and AFR: Important calculations
MTTF is a statistical value that describes the average operating time until the HDD fails. Depending on the HDD model, it lies between one million and 2.5 million hours. For a single drive, that’s of limited significance. A drive could fail at any time, so regular backups and RAID configurations are required to protect against data loss.
With a larger number of hard drives, however, MTTF helps to estimate how regularly failures could occur. With an MTTF of one million hours and one million drives, a drive failure each hour would be expected, or with 1,000 drives there would be a failure every 1,000 hours.
For HDDs in 24/7 operation, annual failure rate (AFR) can be determined from MTTF, which is more intuitive as a percentage value.
Simplified, this can be calculated as follows: Annual operating time of 8,760 hours divided by MTTF in hours multiplied by 100.
So, an enterprise drive with MTTF of 1.4 million hours therefore has an AFR of 0.625%, and in a datacentre with 100,000 drives, 625 would be expected to fail per year. If the operator opts for enterprise drives with an MTTF of 2.5 million hours, AFR is 0.35% and only 350 hard disks are likely to fail in the same period – 275 fewer. Costly maintenance work is thus significantly reduced.
UER: Picking the right drives for the environment
Less serious than failures are read errors, which are usually compensated for by internal error correction. This is not always successful, which is why UER is a key HDD spec.
Sources of UERs can vary – with dust, electromagnetic radiation or a poorly executed write operation being among the main reasons.
For PC, NAS and surveillance HDDs, UER is 1 in 1,014, with an uncorrectable error occurring on average every 1,014 bits read, or every 12.5TB. In contrast, enterprise HDDs with a UER of 1 in 1,015 experience a read error only every 125TB.
If the annual workload of a PC hard disk (55TB) is distributed evenly, there is an uncorrectable read error about every 2.7 months. If, however, it were confronted with the annual workload of an enterprise hard disk (550TB), there would be an error every 0.27 months, or every eight days.
MTTF and UER as specified are only achieved if HDDs are used within the environmental parameters they’re meant for. PC drives, for example, are usually designed to work at temperatures between 0°C and 60°C, while enterprise drives are designed for 5°C to 55°C operation because they are installed in air-conditioned rooms.
Manufacturers also provide information on sensitivity to shock and vibration. NAS and enterprise HDDs are somewhat less sensitive. Because several of them are incorporated into one device, their rotational vibrations can amplify each other, but storage arrays have vibration sensors and control mechanisms to compensate for this.
Rainer W. Kaese is senior manager for business development in storage products at Toshiba Electronics Europe.
Read more on storage specs
- Storage performance metrics – five key areas to look at: We look at the essentials in storage performance metrics: capacity, throughput and read/write capability, IOPS and latency, and hardware longevity measured by failure rates.
- Cloud storage – key storage specifications: We look at the key specs in cloud storage, including availability – such as five nines – bandwidth, IOPS and latency, capacity and tiering functionality, egress charges and security.
