Backblaze publishes quarterly hard drive failure statistics

Shawn Knight

Posts: 15,256   +192
Staff member

Cloud storage specialist Backblaze on Tuesday published its hard drive statistics report for the second quarter, an insightful look at how the thousands of hard drives in its data center fared during the three-month period.

What immediately stands out in Backblaze’s chart are the 228 Seagate ST4000DM000 drives that died during the quarter. But, when you consider that they have 34,693 of these drives in their fleet (50.4 percent of their deployment), the failure number is far more acceptable.

Looking at the annualized failure rate, we see that the Seagate ST400HX000 is actually the highest at 19.81 percent followed by the Toshiba MD04ABA500V at 8.91 percent and the Toshiba DT01ACA300 at 8.68 percent.

It’s worth noting that the annualized failure rate is computed as follows: ((Failures)/(Drive Days/365)) * 100. As Backblaze points out, one should consider the number of “Failures” and “Drive Days” before coming to any judgment about failure rates.

Backblaze’s statistics consist of data drives (not boot drives) and only models which they have 45 or more of deployed.

The cloud storage provider also rolled out 8TB drives in quantity, adding 45 HGST 8TB drives and 2,720 Seagate 8TB drives to the mix during the quarter. Thus far, only three of the Seagate drives have failed; the HGST drives all survived the quarter unscathed.

For those that are curious, the chart below shows Backblaze's annualized hard drive failure rates.

Lead image courtesy stockfoto, Shutterstock

Permalink to story.

 
So when it fails, what happens to all the data you've been saving for years...

a little hint...

image-270x149.jpg
 
As I said with the last hard drive test news, the results and 'testing batches' are so skewed that they can in no way truly represent the overall build quality of any one supplier, especially when you factor in the significant lack of information as to the treatment and conditions under which these drives are under.
 
As I said with the last hard drive test news, the results and 'testing batches' are so skewed that they can in no way truly represent the overall build quality of any one supplier, especially when you factor in the significant lack of information as to the treatment and conditions under which these drives are under.

So I'm assuming because it is a data centre and they need to rely on their drives that conditions must be optimal, treatment must be punishing in the sense of tons of I/O per disk.

The fact that they require tons of hard drives in use doesn't mean those are one batch of disks, they need to replace 2000 a year, without considering the extensions in storage they probably should be making. Also their numbers are closed in to their periods, meaning not old disks.

Definitely enough information to build trends, specially through the years.
 
Those HGST drives are impressively reliable there.

Not surprising. HGST are what used to be Hitachi's HDD business, before Western Digital bought them and began marketing them as HGST. I have a 320GB Hitachi drive that I have been using as my boot drive since 2006, at least until yesterday when I finally switched to a Samsung SSD. It has never so much as hiccupped on me, even after a pretty bad fall.
 
Not surprising. HGST are what used to be Hitachi's HDD business, before Western Digital bought them and began marketing them as HGST. I have a 320GB Hitachi drive that I have been using as my boot drive since 2006, at least until yesterday when I finally switched to a Samsung SSD. It has never so much as hiccupped on me, even after a pretty bad fall.
Welcome to the SSD world!
 
Welcome to the SSD world!
I've had one in my laptop for awhile, because it came with it (albeit, a small one; 128GB). But until recently, when high-end SSDs finally fell below $0.40/GB, I just couldn't justify the premium. Now, I have a 512GB, with 3D NAND that seems to stand the test of time. So now I can fit all my programs on a single drive, and not worry about it slowing to a crawl after a couple years of software updates.
 
So now I can fit all my programs on a single drive, and not worry about it slowing to a crawl after a couple years of software updates.

This is the biggest lie you could've ever heard. The first thing I came up with researching on this topic was this thread that lasted 18 months (Which I'm linking to the final conclusions after all drives died, after 700tb to 2.4PB of writes, more than most consumers will ever write).
 
This is the biggest lie you could've ever heard. The first thing I came up with researching on this topic was this thread that lasted 18 months (Which I'm linking to the final conclusions after all drives died, after 700tb to 2.4PB of writes, more than most consumers will ever write).
Read up on TLC NAND. It ages ridiculously fast, and does in fact slow down - but it is cheaper/GB to produce. MLC NAND, until recently has been too rich for my blood (knowing that I wouldn't go less than 480GB for my next boot drive). 3D NAND finally forced MLC down.

SLC is still stupid expensive, if super durable.
 
Back