Five years of data show that SSDs are more reliable than HDDs over the long haul

Enlarge / Crucial’s venerable MX500 is one of the SSDs that Backblaze makes use of in its data facilities.


Backup and cloud storage firm Backblaze has printed data evaluating the long-term reliability of solid-state storage drives and conventional spinning laborious drives in its data middle. Based on data collected since the firm started utilizing SSDs as boot drives in late 2018, Backblaze cloud storage evangelist Andy Klein printed a report yesterday exhibiting that the firm’s SSDs are failing at a a lot decrease charge than its HDDs as the drives age.

Backblaze has printed drive failure statistics (and associated commentary) for years now; the laborious drive-focused experiences observe the habits of tens of 1000’s of data storage and boot drives throughout most main producers. The experiences are complete sufficient that we will draw at the very least some conclusions about which firms make the most (and least) reliable drives.

The pattern dimension for this SSD data is far smaller, each in the quantity and selection of drives examined—they’re largely 2.5-inch drives from Crucial, Seagate, and Dell, with little illustration of Western Digital/SanDisk and no data from Samsung drives in any respect. This makes the data much less helpful for evaluating relative reliability between firms, however it will probably nonetheless be helpful for evaluating the total reliability of laborious drives to the reliability of SSDs doing the identical work.

Backblaze's data suggests that HDDs start failing more in five years, while SSDs keep trundling along.

Backblaze’s data suggests that HDDs begin failing more in 5 years, whereas SSDs preserve trundling alongside.

Backblaze makes use of SSDs as boot drives for its servers somewhat than data storage, and its data compares these drives to HDDs that have been additionally getting used as boot drives. The firm says these drives deal with the storage of logs, short-term recordsdata, SMART stats, and different data along with booting—they are not writing terabytes of data each day, however they are not simply sitting there doing nothing as soon as the server has booted, both.

Over their first 4 years of service, SSDs fail at a decrease charge than HDDs total, however the curve appears to be like principally the identical—few failures in 12 months one, a bounce in 12 months two, a small decline in 12 months three, and one other improve in 12 months 4. But when you hit 12 months 5, HDD failure charges begin going upward rapidly—leaping from a 1.83 % failure charge in 12 months 4 to three.55 % in 12 months 5. Backblaze’s SSDs, on the different hand, continued to fail at roughly the identical 1 % charge as they did the 12 months earlier than.

This data—each the reliability hole between them and the reality that HDDs begin to sputter out sooner than SSDs—makes intuitive sense. All else being equal, you’d count on a drive with a bunch of transferring components to have more factors of failure than one with no transferring components. But it is nonetheless fascinating to see that case made with data from 1000’s of drives over a couple of years of use.

Klein speculates that the SSDs “might hit the wall” and begin failing at increased charges as their NAND flash chips put on out. If that have been the case, you’d see the lower-capacity drives start to fail at the next charge than higher-capacity drives since a drive with more NAND has the next write tolerance. You’d additionally seemingly see loads of these drives begin to fail round the identical time since they’re all doing related work. Home customers who are continuously creating, modifying, and transferring round giant multi-gigabyte recordsdata might additionally see their drives put on out sooner than they do in Backblaze’s utilization state of affairs.

For anybody who want to poke at the uncooked data Backblaze makes use of to generate its experiences, the firm makes it accessible for obtain right here.

Leave a Comment