Backblaze data suggests young SSD’s are less likely to fail than HDD’s

A recent Backblaze blog post is one of the best comparisons of SSD and HDD reliability currently available and I highly recommend having a read. Unfortunately, many outlets have presented misguided clickbait headings but the facts are simple, the Annualized Failure Rate of HDD’s in the lifetime controlled comparison (average age ~14.2 months) was 31.4% worse for hard drives (SSD = 1.05% AFR, HDD = 1.38% which equates to a difference of 1.314x between rates = +31.4% increased AFR for HDD’s). Despite this proportionally large difference the difference in values falls within the +/- 0.5% confidence interval window and as such we still require more data to be sure but at this point, it suggests that young SSD’s in Backblazes environment were less likely to fail than HDD’s.

For older drives, if we compare Annualized Failure Rates between (04/2013 – 06/2021) by aligning Backblaze datasets for 2014 (HDD) and 2018 (SSD) (as this is when each respective boot drive data set starts) we can get a rough guideline of the year by year failure rates. You can see that the failure rate of HDD’s jumps to a high point rapidly after one year, then drops off but not enough to offset the initially high rate. SSD’s consistently were at least 74% more reliable and even in the SSD’s least reliable year SSD’s are still more than HDD’s most reliable year in this comparison. The one thing we do not yet know from Backblaze data is whether or not SSD’s will follow a similar trend, however, I suspect that if SSD’s do not hit the point where NAND wear becomes a problem that they’ll be about 30%-80% more reliable than HDD’s.

So there you have it between the drives in Backblaze’s environment an SSD that is younger than 4 years old is on average between 30% and 80% less likely to fail than a HDD, for many users (myself included) the performance benefit + the reliability benefit makes the purchase of an SSD as a boot drive a well and truly justified expense. Despite all this, the #1 thing we should note is that all of this is still based on limited data and as Andy Klein wisely notes in his article that factors such as “cost, speed required, electricity, form factor requirements, and so on” are likely more important to focus on than failure rates.

Thank you Backblaze, for sharing their storage reliability data and for their insights, also thanks to Andy Klein for writing such a great article on the Backblaze blog (which this article is based upon).