Since 2013, cloud storage firm BackBlaze has published quarterly reports and raw performance data detailing the drives in their data centers. Rather than rely exclusively on enterprise-grade hard drives, BackBlaze uses a mixture of enterprise and consumer hard drives, primarily purchasing whatever can be found cheaply.

For Q3 2017, BackBlaze added 9,599 new drives and decommissioned 6,221 drives, bringing the total to 86,529 hard drives, sourced from 17 different SKUs across all four manufacturers.

Drives are becoming more reliable

According to the report, in Q3 BackBlaze experienced the lowest failure rate of hard drives ever, at 1.84%. The report noted that there are several factors that led to this result, not the least of which is the average age of drives. Only the HGST-produced 4TB drives (specifically, HDS5C4040ALE630) have an average age of more than four years. Of those drives, only four of the 2,490 (0.59%) presently deployed failed in Q3. As the currently deployed drives are relatively new, fewer drives are failing as part of the limitations of their natural lifecycle.

SEE: Power checklist: Managing backups (Tech Pro Research)

BackBlaze has added 1,240 drives with capacities of 10 TB or more this quarter. Of note, all of these are Seagate drives, the ST10000NM0086 and ST12000NM0007, both of which are Helium drives. Typically, large quantities of hard drives are susceptible to premature failures–some drives arrive from the factory dead, or fail in the first few hours of being powered on. This has not occurred with the Seagate Helium drives. While it is possible that quality assurance standards differ for more expensive Helium drives, it is still too early to draw a firm conclusion about the overall reliability of these drives in particular, or Helium drives as a whole.

Deciding between enterprise and consumer drives

In terms of market segmentation, the benefits enterprise drives have over consumer drives is somewhat unclear. While enterprise drives do typically have five-year warranties rather than the industry standard two years warranties for consumer drives, benefits beyond a warranty depend heavily on specific features enabled in drive firmware, and an appropriate use case to utilize those features.

BackBlaze has two similar Seagate 8TB drives deployed–the enterprise ST8000NM0055 and consumer ST8000DM002. Between the two, the enterprise drive had 28 of 14,404 drives (1.04%) fail in Q3, while the consumer drives had 18 of 9,879 (0.72%) fail. For an annualized failure rate, the consumer drive failure is 1.10% versus 1.20% for the enterprise drive.

BackBlaze also found that the enterprise drives consume more power, and can write about 40% more data per day, than consumer drives. However, the enterprise drives come with PowerChoice, which allows the drives to be operated in an energy-saving mode, and the design of BackBlaze’s operations do not account for drive speeds as a bottleneck–faster drives naturally fill up faster, but capacity for accepting data always exists. For this particular use case, the write speeds are not a significant benefit, but this is not applicable across all use cases. With that said, use cases which have a clear benefit–such as database operations–would be run more efficiently from an SSD. Additionally, it is unclear if the PowerChoice setting has an impact on the failure rates of the enterprise drives.

Based on these results, consumer drives can easily be considered competitive with enterprise drives in terms of reliability. If that is your only concern, consumer drives are a safe option.

Drive capacities are increasing rapidly

In Q3, BackBlaze started deploying 10TB and 12TB drives. While the 12TB drives used in the Q3 report are only a trial of 20 drives, the report noted that a 1,200 drive deployment went into production in October. Drive capacities are unlikely to plateau anytime soon as HGST introduced a 14TB drive in October, and Seagate is expected to unveil a 16TB drive next year.

For those interested in studying the data used in these reports, BackBlaze has released the raw data detailing the active hard drives in their data center since Q4 2013. The data can be downloaded in CSV format here.

What’s your view?

Do you have a preferred vendor of hard drives, or one that you take special care to avoid? What measures do you take against individual drive failure? Tell us about your disk deployment strategies and thoughts in the comments.