HGST stays atop the charts in latest Backblaze reliability stats

Cloud backup outfit Backblaze runs its business by buying large quantities of mostly consumer-grade hard drives, running them in close quarters, and planning around the failure of those spinners when they do croak. Along the way, the company releases quarterly reports about the failure rate for the tens of thousands of drives it uses. The company released its latest report just yesterday, and as always, the results are worth a read.

Overall, the company says hard drive failures were down to 1.95% in 2016, compared to 2.47% in 2015 and 6.39% in 2014. HGST drives once again took home the reliability crown with a failure rate of 0.60%. Toshiba drives came in second with a 1.27% failure rate, but the company's 237 deployed drives were a drop in the bucket compared to the total drive pool of over 72,000.

Seagate drives achieved a massive reliability improvement in 2015, with failure rates dropping from 10.68% to 3.48%. That  trend continued in 2016, as 2.65% of Backblaze's Seagates checked out. WD's failure rate also fell, from 6.55% to 3.88%, but that was not enough to keep the company from being at the back of the reliability pack again. We must note that WD drives represent only 2.25% of Backblaze's drive pool, though. The least reliable single model in the company's stable was a Seagate 4TB unit, with a failure rate of 13.57%.

The company includes a special section devoted to 8TB drives, which have so far shown themselves to be fairly reliable in Backblaze service. A particular Seagate 8TB drive accounts for 22% of the company's total drive pool capacity, and shows a lower-than-average failure rate of 1.65%. The company has a few other enterprise 8TB drives from Seagate in its inventory, and future comparisons of enterprise and consumer-targeted disks of the same capacity could be interesting

While Backblaze does use a lot of the sort of drives an enthusiast might use to hold the Steam library in a desktop PC, the operating conditions are quite a bit different. Backblaze packs as many as 60 drives into a 4U server chassis. A desktop PC probably endures more power cycles than a Backblaze server, and the company operates its drives in a narrower temperature window than a computer in a home or office. With all that said, the company's reports are one of the best available resources for comparing hard drive reliability.

The company's report contains detailed information about how the company accounts for its constantly-changing pool of drives and how it calculates failure rates. If reading the report is not enough, the company provides the raw data behind the report on its Hard Drive Test Data page.

Comments closed
    • CuttinHobo
    • 3 years ago

    It makes you wonder how much of the improvement in Seagate and WD’s failure rates is due to these Backblaze reports. It *must* be influencing the powers-that-be to put more emphasis on QC. Speaking for myself, it influenced me to buy HGST for my NAS where I probably would have used WD before. Vote with your wallet, peeps! 🙂

      • Wirko
      • 3 years ago

      2 x WD40EFRX here. By some luck, they’re from different batches and manufacturing dates. Less likely to fail on the same day … or is my hope in vain?

      • Lord.Blue
      • 3 years ago

      Isn’t HGST owned by WD?

        • Leader952
        • 3 years ago

        Again the same question asked over and over again with the question mark on the end that implies that HGST and WD drives are the same.

        Answer they are not the same.

        • CuttinHobo
        • 3 years ago

        Yes, a relatively recent acquisition. They’re no nowhere near integrated.

      • just brew it!
      • 3 years ago

      The Backblaze reports are just the tip of the iceberg. Enterprise storage vendors also deploy massive quantities of mechanical HDDs, and sometimes even use consumer-grade drives (like Backblaze does) for cost-sensitive applications. If you’ve got plenty of redundancy and plan to replace the drives in a few years anyway as larger ones become available, the extra cost of enterprise-grade SAS gear may not be worth it.

      In any deployment where there are 10s of thousands of drives, if there’s a problematic drive model it will become [i<]painfully[/i<] obvious to the vendor of the storage system, who will in turn ensure that it becomes painfully obvious to the supplier of said drives. Enterprise storage vendors just tend not to publicize it when it happens; if you're selling on-prem solutions you don't want to advertise the fact that the drives in your systems are failing! There's also a desire to maintain friendly relations with key suppliers (with the consolidation in the HDD industry there aren't that many alternatives to choose from...), and in some cases there are also NDAs involved.

    • flip-mode
    • 3 years ago

    No Seagate ST3000 drives in the chart because they all died.

      • Prototyped
      • 3 years ago

      The ST4000DX000 appear to be made in the same mould. Though, average age is just over ~3 years. Perhaps we’re reaching the other end of the [url=https://static.googleusercontent.com/media/research.google.com/en//archive/disk_failures.pdf<]bathtub curve[/url<]?

      • colinstu12
      • 3 years ago

      they decommissioned a lot of drives that were in quantities <45. they don’t seem to mix drives in pods. their blog article talks about it.

    • ImSpartacus
    • 3 years ago

    I find it so cool that a private company publishes this kind of info about it’s inner workings. A lot of firms wouldn’t allow this.

    But I think it has helped their name recognition and provided some goodwill on their behalf.

      • UberGerbil
      • 3 years ago

      It’s a huge PR boost for them. They could put out press release after press release about what they do and who among the enthusiast crowd would have ever heard of them? And some enthusiasts do have day jobs, or eventually get them, where they evaluate / recommend offsite backup providers.

      • MOSFET
      • 3 years ago

      It may have even helped overall reliability by spotlighting manufacturers.

      • just brew it!
      • 3 years ago

      As long as their end users aren’t affected when drives fail, it does not hurt them to disclose failure rates; when these articles get linked all over the ‘net, it amounts to free advertising. Also, since they also tend to buy the most cost-effective drives they can find on the open market, they don’t need to worry about souring their relationships with key vendors.

Pin It on Pinterest

Share This