It’s a question that pops up on almost every IT message board across the web: are enterprise-class hard drives worth paying more for than desktop hard drives? The cloud-storage company Backblaze compared the rate at which desktop hard drives failed to the rate at which enterprise drives did, and found that—contrary to claims of being more robustly built and more reliable—the enterprise drives failed at a narrowly higher rate than their desktop counterparts.
Enterprise drives are designed to fit into big arrays of hard disks, so they have shorter recovery time limits and are supposed to handle vibrations better and use less power. The first difference that people notice between enterprise and desktop or consumer hard drives, though, is price. Consumer drives can be bought for a hundred bucks or less, while enterprise class can cost twice or even three times as much.
So when Backblaze was starting out, they opted for the less expensive desktop hard drives, set into arrays in Backblaze Storage Pods, running software that could ensure that the end of one of the hard drives wasn’t the end of the data. There are now 45 drives to a pod, in RAID arrays of 15 drives each.
Their business grew to 25,000 consumer hard drives, running in these large arrays for years. They tracked when they needed to replace hard drives, and last month published a report on the life expectancy of hard drives and when, on average, they were giving out.
Over the last two years, Backblaze also used six shelves of enterprise-class drives in commercially purchased Dell Powervault storage systems for administrative and transactional data. Just for an experimental comparison, they put enterprise drives into the same Backblaze Storage Pods that they used desktop drives in. The enterprise drives in both the commercial arrays and the company-specific ones failed at about the same rate—how did that rate compare to regular consumer hard drives?
Backblaze’s co-founder and CEO Gleb Budman explained that the failure rate was the “average number of failures you can expect when you run one disk drive for a year.” Their research on enterprise drives is only on drives that are two years old or less, so it's unclear if enterprise drives hit their stride after two years and are a better long-term investment than consumer drives, which Backblaze found start to break down at an increasing pace after three years.
The comparison between the enterprise and desktop drives has convinced Budman that there’s no advantage of reliability offered by enterprise drives. For him the only reasons someone might consider paying more for an enterprise drive is “if the longer warranty (for some type of accounting reason) or a specific feature such as the ‘Instant Secure Erase’ is important, perhaps.”
I asked him if the longer recovery time limits make arranging an array of the desktop hard drives a problem at all. “We might get better performance in the RAID with shorter timeout values because the RAID software wouldn't have to wait so long to find out one of the drives isn't responding,” Budman offered. “We have not found this to be a big problem, though.”
While it doesn’t make or sell hard drives, it’s not like Backblaze is totally sharing this information altruistically. I’m not disputing any of their conclusions, but it only makes sense that “all hard drives are fallible” is a message that a cloud-storage company might not might mind spreading around.