Brand wars in the hard disk drive (HDD) realm can be a heated affair. Recently, online backup provider Backblaze stoked the flames by issuing a comparison of HDDs they deploy in their operation, in an attempt to show how they measure HDD reliability. Storage experts are still grumbling over the findings—Backblaze contends that certain HDD brands failed at rates triple that of other brands, and proposed that consumer models outperform enterprise models.
Shortly after the report was published, storage experts took to the Internet penning articles that called Backblaze’s methods dubious, and their findings misleading. We’ll attempt to sort out the arguments for and against this controversial test, and in the process, show how this type of research should be taken with a grain of salt—especially when it pits brand versus brand.
The (In)Famous Backblaze Test
Backblaze uses over 34,000 HDDs in its operation, a combination of Seagate, Hitachi (now part of HGST), and Western Digital devices. Whenever a drive needed replacing, they logged it as a failure. They counted up all the failures by model, and divided the frequency of failures against the number of drives in use. They took into account the average age of each group of specific models, reporting that alongside the annual failure rate. They reported twice; once in January and once in September.
Here’s a graph of the failure rate that sparked the most controversy.
At first glance, the story looks like the Seagate Barracuda LP 1.5 TB and the Seagate Barracuda 7200 1.5 TB are to be avoided. Counter-intuitively, though, Backblaze engineer and report writer Brian Beach notes that his company deploys more Seagate drives than any other brand, citing that Seagate boasts a combination of consistent (but not great) performance and good pricing. Additionally, at the time of the January report, he said that Backblaze was focused on purchasing mostly Seagate Desktop HDD. 15 4 TB, as well as Western Digital Red 3 TB models, for the HDD pods he deployed.
You might think that given the results, Backblaze would have been snapping up as many Hitachi GST Deskstar 7K3000 and Hitachi GST Deskstar 5K3000 as it could get its hands on, but that wasn’t the case—and detractors would call Backblaze on this.
But the number one issue that storage experts had with the report was the proposition that an enterprise model with a zero failure rate takes too long (10 years) to “break even” against cheaper consumer drives with 15 percent failure rate—and that in Backblaze’s findings, consumer drives were slightly more reliable.
The Backblaze Backlash
Storage technology analyst Dan Olds took issue with the small sample size of the enterprise HDDs in Backblaze’s report, which compared 14,719 drive-years of consumer disks versus 368 drive-years of service on enterprise disks. This except is from The Register:
Henry Newman, CEO of storage architecture company Instrumental Inc. and a prominent voice in storage, took Backblaze to school on several issues as well. First, he points out that Backblaze’s 1.5 TB Seagate models were almost eight years old, information that was left out of the study: “It is well known in study after study that disk drives last about 5 years and no other drive is that old,” he writes. “I find it pretty disingenuous to leave out that information.” He adds: “It is no surprise that these old drives are failing. Average [ages] are a pretty useless statistic, and if Seagate drives are so bad then why buy so many new drives? ”
Newman goes on to address the other end of the spectrum, touching on brand new drives, which have, as he calls them, issues with “infant mortality.” He elaborates: “Drive vendors talk about it, RAID vendors talk about it, but there is not a single mention of what, if anything is done in the Backblaze environment or how that figures into any of the calculations.”
The Backblaze study also caught a lot of flak on Spiceworks and other discussion boards. IT professionals deploying storage arrays questioned Backblaze’s infrastructure model. Also, some called into question why anyone would install enterprise model HDDs alongside consumer drives in conditions “not even remotely compliant with manufacturer specs,” according to one popular commenter.
Overall, the advice eschewed by Olds, Newman, and the IT community weighing in on the discussion boards: When looking at any third party research or consumer reviews of HDDs consumers should understand their needs and uses for the gear they’re buying, and deploy it within manufacturer specifications. While it was noble of Backblaze to share its business intelligence, the inherent problems here are that every storage need is unique, and their own environment may not be optimal for running enterprise-grade equipment.
Some key takeaways from all the back-and-forth:
- Enterprise-grade HDDs come with a 3-year warranty. Possibly the best way to fool-proof your investment.
- Doing your own testing is beyond the resources of most SMBs, and third party reliability claims must be taken with a grain of salt.
- HDDs from all vendors can and do fail.
- Statistical historical data does not indicate future results—your results will vary from the mean because that’s how statistics work.
- We would be curious how something like Seagate Constellation enterprise HDDs (for example) would hold up in the Backblaze environment since it is designed for datacenter use.
Undoubtedly, the best price point per-GB storage will remain the most common deciding factor for HDD purchasing. Nonetheless, it is still good practice to keep these perspectives in mind when selecting the best hard disk drive for your operation.
Join the discussion 3 Comments