Hard drive reliability continues to decline
What's happening
Backblaze currently has about 285,000 hard drives at its disposal, it was six months ago 250 thousand. The company already published results of their operation.
The main thing to look for in such reports is the annualized fail rate (AFR). The company uses this metric to measure the reliability of both HDDs and SSDs. AFR means the probability of a drive failing within a full year of its operation. The relationship is inverse: the higher the AFR, the lower the reliability.
The average AFR for Q2 2024 was 1.71%. In Q1 this year, it was 1.4%. It turns out that HDD reliability has decreased. The drop is not very significant, but it is there. The trend is typical mainly for new models: among HDDs from previous years, the situation is better. There are drives that have been working without problems for more than eight years.
Authors of the report They saythat there is another dependence: the greater the capacity of a certain model, the higher the probability of failure. This is understandable, because the more capacious the disk, the more complex its structure.
The company's statistics do not include models of which Backblaze has no more than 100 units. In addition, it does not take into account drives that have not accumulated 10,000 days of operation in total for the quarter. But even without them, the report contains enough data – the final result included the indicators of 283,065 drives.
More details
The worst performer was the 12 TB HGST model (HUH721212ALN604), whose AFR jumped to 7.17% during the reporting period. However, there are leaders that work without failures: two models from Seagate, the 14 TB ST14000NM000J and the 16 TB ST16000NM002J. Neither of them failed during the entire quarter. However, the company does not have many of these drives at its disposal.
The oldest model that still works in the company's racks is the Seagate 4 TB (ST4000DM000). One of the drives of this model worked for 9 years 11 months and 23 days at the end of the second quarter. All ST4000DM000 HDDs will soon be replaced with more modern HDDs with a higher capacity. This is necessary to save useful space, because the higher the capacity of the hard drives, the higher the data storage density.
The company has published not only tables, but also charts based on hard drive failure statistics. The first is shown below. It shows the AFR for 14 models with an average age of 60 months or less. Why that? That's the typical warranty period for enterprise-class hard drives.
Interestingly, the disks that fell into the first quadrant on the first diagram are described by the company's specialists as working well, with an AFR of less than 1.5%. The second quadrant shows the characteristics of disks that work in normal mode with an AFR above 1.5%. There are no disks in the third quadrant yet because of their “youth”, there is simply little information about them. The fourth shows new HDDs that have worked for some time. There are not many failures among them.
The second chart shows data for models with an average age of over 60 months. Here, the first quadrant shows well-performing models, the second and third show HDDs that are worrisome. The fourth quadrant shows an HDD model that has no problems yet.
Finally, the third chart shows the lifetime failure rate for nine models over 60 months old – for clarity, the count starts at 24 months. Here, the distribution is shown by the first and second quadrants. The most reliable models are those with nearly vertical lines: red, brown and purple.
HDDs with an increased number of failures are shown by blue and gray stripes. Even less stable HDDs are in the third quadrant. And the black line is a completely “tired” Seagate model with a capacity of 4 TB. That same “long-liver” that I mentioned above.
The results of the report can be used, but with one condition
Full details from Backblaze are available at this linkThe company allows the use of the analysis results for its own purposes, but, as before, three conditions must be met:
Please cite Backblaze as the source if you decide to use the report.
Accept sole responsibility for how you use the information in the report. That is, do not blame Backblaze if its information somehow harms someone.
Information is provided free of charge, commercial use is prohibited.
We plan to publish company statistics approximately every six months to keep you informed. Hopefully, the situation will improve.
Perhaps you or your company have your own statistics? If so, share the information in the comments – we will all be interested!
