Backblaze shares Q4, full-year 2016 hard drive failure stats

Shawn Knight

Posts: 15,256   +192
Staff member

Cloud storage provider Backblaze has published Q4 2016 and full-year hard drive failure rate statistics of the 71,939 production hard drives used solely as data drives.

The company maintains another 1,553 HHDs although those are reserved as boot drives and thus, not included in this analysis. Drives that are used for testing purposes and those drive models for which they do not have at least 45 of are also omitted.

Looking at the 2016 full-year data, we see that three drive models – the 8TB HGST HUH728080ALE600, the 8TB Seagate ST800NM0055 and the 4TB Toshiba MD04ABA400V – ended the year with zero failures. It’s worth noting, however, that the drive count on these models is pretty low – 45, 60 and 146, respectively – and that the Seagate drives weren’t deployed until November.

For the year, a total of 1,225 drives kicked the bucket which works out to 3.36 failures a day or around five per workday. The yearly drive failure rate for 2016 was just 1.95 percent, a solid improvement over the 2.47 percent that died in 2015 and far below the 6.39 percent that croaked in 2014.

Backblaze also organized 2016 failure rates by drive size and vendor. Again, the data is skewed a bit due to the number of drives (or lack of) between brands but what we see is that 3TB drives are the most reliable with a failure rate of just 1.40 percent. Looking at manufacturers, HGST came out on top in 2016 with a failure rate of only 0.60 percent with Western Digital having the highest failure rate at 3.88 percent.

Data diehards that want to dig in deeper should check out Backblaze’s blog post on the matter for a more detailed analysis. The company will also be presenting its “Backblaze Hard Drive Stats for 2016” webinar on February 2 should you wish to tune in.

Permalink to story.

 
Wow, HGST and Toshiba are really rocking it. This has been multiple quarters in a row now that they have been consistently more reliable. I actually just switched out two of my WD Black 4 TB (FZEX, latest model) for two 6 TB Toshiba x300 and they have been rock steady. I will add though that the drives seems to have more aggressive parking and I find it having to spin up more often.
 
Seagate's 3TB "Barracuda" HDDs, have the worst star rating on Newegg I've ever seen. And yet, Backblaze's finding seems to contradict that....Er, this quarter

The drives, I would hazard a guess with the best reliability would be single platter 1TB or less units, Seagate brand included. There is much less mass in motion, and they run very, very, cool.

Backblaze purchases vast quantities of consumer level drives, stuffs them into NAS enterprise cabinets, then pumps out this bunch of BS every quarter, in what I believe, is a desperate attempt to keep itself relevant.
 
Seagate's 3TB "Barracuda" HDDs, have the worst star rating on Newegg I've ever seen. And yet, Backblaze's finding seems to contradict that....Er, this quarter

The drives, I would hazard a guess with the best reliability would be single platter 1TB or less units, Seagate brand included. There is much less mass in motion, and they run very, very, cool.

Backblaze purchases vast quantities of consumer level drives, stuffs them into NAS enterprise cabinets, then pumps out this bunch of BS every quarter, in what I believe, is a desperate attempt to keep itself relevant.
Backblaze doesn't show any Seagate brand 3 TB drives. Either you have a typo here or you can't read.
 
Backblaze doesn't show any Seagate brand 3 TB drives. Either you have a typo here or you can't read.
Backblaze has trashed Seagate in every other report they've issued thus far. Am I to believe they didn't include Seagate in this spate of purchasing consumer drives for commercial use?

I do confess to have taken to just skimming their (IMHO), fairly useless reports.

I know Newegg reviews are /iffy with respect to accuracy. Everybody who gets a bad product trashes it mercilessly, and everybody that get the good ones, retires to their computer to enjoy it. But 3TB units, never seem to get the top reviews, regardless of brand. Nevertheless, I have no intention of reading said reviews for the next several hours, to humor you. Sorry man, it's just my way.

BTW, I can't read, I used to work for the Post Office. Although, with the help of "Spell Check" in Firefox, I manage to muddle along nicely, thank you very much.
 
Why would they trash a specific company? From my personal experience I have had the worst luck with Seagate drives, and BB numbers have gone along with my experience of other drives as well. I'm not saying BB is right in everything but they seem to have similar results to my experiences, so I have a degree of trust in their claims. Like any intelligent person though you take it with a grain of salt, because yes, there are haters out there.
 
Why would they trash a specific company? From my personal experience I have had the worst luck with Seagate drives, and BB numbers have gone along with my experience of other drives as well. I'm not saying BB is right in everything but they seem to have similar results to my experiences, so I have a degree of trust in their claims. Like any intelligent person though you take it with a grain of salt, because yes, there are haters out there.
They've already done so. Some of their quarterly reports have singled out Seagate's drive(s) as basically the worst available, due to the highest failure rates. I might add, not undeservedly so. Their last run of 3 TB HDDs was so bad, Newegg seemed to practically have to give them away. The customer reviews were in the basement, and I honestly believed that someone, or everyone on that particular assembly line, needed to be fired. That is, along with the line itself needing to be retooled. The 2TB and 4TB drives, didn't share the same disdain. Disdain they had aplenty, but not nearly on the same grandiose level.

I don't know how you missed it, but (IIRC), all of Backblaze's quarterly reports contained the count in blocks of drives they purchased, the brands involved, and what the failure rate was in each instance. Maybe they've mellowed. Who knows, as I said, I just skimmed it.

They have made themselves sort of infamous, (IMHO), by buying consumer duty drives, and then seeing how many they can break in commercial server usage. While for a time, that was at least entertaining, the novelty has worn off.

These reports, at least to me, are just their way of saying, "hey, we're still here breaking drives, why don't you hire us to do your backups.

Methinks their message is mighty darned counter intuitive.
 
Seagate's 3TB "Barracuda" HDDs, have the worst star rating on Newegg I've ever seen. And yet, Backblaze's finding seems to contradict that....Er, this quarter

The drives, I would hazard a guess with the best reliability would be single platter 1TB or less units, Seagate brand included. There is much less mass in motion, and they run very, very, cool.

Backblaze purchases vast quantities of consumer level drives, stuffs them into NAS enterprise cabinets, then pumps out this bunch of BS every quarter, in what I believe, is a desperate attempt to keep itself relevant.

Blackblaze offers unlimited online backup service at $50. It is pretty common for companies in that industry to use consumer drives to keep prices down. The few that don't charge very high prices.
 
Blackblaze offers unlimited online backup service at $50. It is pretty common for companies in that industry to use consumer drives to keep prices down. The few that don't charge very high prices.
Now see, it would be interesting if they took a block of consumer drives, versus a block of identically branded enterprise drives, compare the failure rates and cost to replace, (in both categories), then show us the bill (s). We could actually learn a lot more that way, than we do now. It might even be worth reading.
 
Now see, it would be interesting if they took a block of consumer drives, versus a block of identically branded enterprise drives, compare the failure rates and cost to replace, (in both categories), then show us the bill (s). We could actually learn a lot more that way, than we do now. It might even be worth reading.

That would be interesting and as a business you'd hope they'd have done that when first deciding what drives to use. Right now we have to assume that HDD manufacturers testing environment and the numbers those generate are reliable.
 
[Now see, it would be interesting if they took a block of consumer drives, versus a block of identically branded enterprise drives, compare the failure rates and cost to replace, (in both categories), then show us the bill (s). We could actually learn a lot more that way, than we do now. It might even be worth reading.

They did that already:
https://www.backblaze.com/blog/enterprise-drive-reliability/

They found a slight decrease in longevity for the enterprise drives as compared to their consumer drives.

Google may have done some work on this too. They also use consumer drives in data centers rather than the more expensive enterprise drives, and they've gone on record asking for even cheaper drives with less emphasis on reliability for data centers, which would be even further from "enterprise" or "server" class drives. With the level of redundancy they have and the massive number of drives in service, the most important factor in reducing their costs is not reliability, but cost of the replacement drives.

Interesting read about that at https://cloudplatform.googleblog.com/2016/02/Google-seeks-new-disks-for-data-centers.html
 
They did that already:
https://www.backblaze.com/blog/enterprise-drive-reliability/

They found a slight decrease in longevity for the enterprise drives as compared to their consumer drives.

Google may have done some work on this too. They also use consumer drives in data centers rather than the more expensive enterprise drives, and they've gone on record asking for even cheaper drives with less emphasis on reliability for data centers, which would be even further from "enterprise" or "server" class drives. With the level of redundancy they have and the massive number of drives in service, the most important factor in reducing their costs is not reliability, but cost of the replacement drives.

Interesting read about that at https://cloudplatform.googleblog.com/2016/02/Google-seeks-new-disks-for-data-centers.html

The only problem with the report you linked is that they specifically state they did not have data on enterprise drivers over 2 years of age, which is really contradictory to their purpose. It's not at all conclusive if they aren't going to do over 2 years and shouldn't have bothered reporting it.

It cannot be denied that consumer drives will be the better option for data centers, simply because of the redundancy they employ. On a consumer level though, I would still like to see if enterprise drives last a significantly longer time. Regular customers do not have the capabilities for massive redundancy.
 
Seagate's 3TB "Barracuda" HDDs, have the worst star rating on Newegg I've ever seen. And yet, Backblaze's finding seems to contradict that....Er, this quarter

The drives, I would hazard a guess with the best reliability would be single platter 1TB or less units, Seagate brand included. There is much less mass in motion, and they run very, very, cool.

Backblaze purchases vast quantities of consumer level drives, stuffs them into NAS enterprise cabinets, then pumps out this bunch of BS every quarter, in what I believe, is a desperate attempt to keep itself relevant.
I do find this drives on consumer level reports really interesting, as stated already trashing Seagate's 3TB drives is not that hard, even though there is no specific data on Seagate's 3TB drives, if you compare HGST Toshiba and WD drives from 3 and 4, even though the 4tb tend to be newer there is a higher failure rate, and it could repeat on the 5tb range too, could it be on density for the "odd" drives?
 
They've already done so. Some of their quarterly reports have singled out Seagate's drive(s) as basically the worst available, due to the highest failure rates. I might add, not undeservedly so. Their last run of 3 TB HDDs was so bad, Newegg seemed to practically have to give them away. The customer reviews were in the basement, and I honestly believed that someone, or everyone on that particular assembly line, needed to be fired. That is, along with the line itself needing to be retooled. The 2TB and 4TB drives, didn't share the same disdain. Disdain they had aplenty, but not nearly on the same grandiose level.

I don't know how you missed it, but (IIRC), all of Backblaze's quarterly reports contained the count in blocks of drives they purchased, the brands involved, and what the failure rate was in each instance. Maybe they've mellowed. Who knows, as I said, I just skimmed it.

They have made themselves sort of infamous, (IMHO), by buying consumer duty drives, and then seeing how many they can break in commercial server usage. While for a time, that was at least entertaining, the novelty has worn off.

These reports, at least to me, are just their way of saying, "hey, we're still here breaking drives, why don't you hire us to do your backups.

Methinks their message is mighty darned counter intuitive.
I see what you are saying... but they do list the drive model numbers, and it in the end it is buyer beware. Ultimately the buyer is responsible for buying the appropriate product for their needs. If anything, it is interesting to see how consumer drives handle the stresses of a enterprise datacenter environment. That is actually quite valuable information, for lots of us.
 
I do find this drives on consumer level reports really interesting, as stated already trashing Seagate's 3TB drives is not that hard, even though there is no specific data on Seagate's 3TB drives, if you compare HGST Toshiba and WD drives from 3 and 4, even though the 4tb tend to be newer there is a higher failure rate, and it could repeat on the 5tb range too, could it be on density for the "odd" drives?
I do have a very much layman's opinion on this topic, but it seems to be bearing itself out in the high failure rates of the humongous capacity HDDs of today.

OK, first I was completely astounded by the massive 160GB capacity of the, (what would now be called "WD Blue"), HDD in my 2005 eMachines! That particular drive is still alive and well after 13 years of daily abuse!

Fast forward until now. Depending on who you talk to, Intel is either sandbagging moving to 10nm process, or is finding it more difficult than they had expected, to do so. (I'm in the latter camp).

With a "hybrid device" such as a HDD, you not only have to shrink electrical pathways, but the tolerances of mechanical devices designed to recover data from the platter. So, we've jumped from perhaps 40GB (!) per platter data capacity, to as much as 4 TB (!!!) per platter, if Seagate produces the 16TB drive they've "threatened" us with..

Some quick math brings us the realization that the points of data on the platter can encompass only 1/100 the area of the 40 GB platter's. It also requires that the tracking movements of the mechanical arm be 1/100 the distance!

So, we're now faced with producing mechanical apparatuses which have machining tolerances approaching the molecular level. It takes machines to build machines, and forgive me for digressing to cliche', but, "there's many a slip twixt the cup and the lip".

As consumers and enterprise alike clamor for higher and higher capacity drives, (IMHO), the problem is going to get worse, before we find out if we can even make it better!

NOTE: The foregoing was an opinion delivered by an old fart and a partial Luddite. Who, for his own purposes in general, requires, and will not seek, drives with capacities exceeding 1TB. My current favorites are the 1TB single platter, and dare I say it, without incurring the wrath and scorn of my fellow members, made by Seagate! They're thin, light, run very cool, and are dirt cheap in actual dollars. Note that I can't possibly have used these for the 13 years of my WD, but all the ones I've purchased have fired right up, been recognized by Windows, and have never made a "mistake", thus far. (y)
 
Last edited:
Blackblaze offers unlimited online backup service at $50. It is pretty common for companies in that industry to use consumer drives to keep prices down. The few that don't charge very high prices.
Now see, it would be interesting if they took a block of consumer drives, versus a block of identically branded enterprise drives, compare the failure rates and cost to replace, (in both categories), then show us the bill (s). We could actually learn a lot more that way, than we do now. It might even be worth reading.

Yev from Backblaze here -> We've done that before to some extent, you can take a look at our article from a while back: https://www.backblaze.com/blog/enterprise-drive-reliability/. The thing is that we don't publish these results as a "buyers guide" for individuals, we simply report on what we see in our environment. We do use consumer-grade (for the most part, although always experimenting with various stuff) drives in our servers, but that's to keep costs down - the entire system is built to allow for drive failure (https://www.backblaze.com/blog/reed-solomon/).

As to Seagate from a previous comment, we had to migrate en-masse from their 3TB variant so you don't see them in our Q4 stats, though if you look at the "all-time" data, it'll appear there. More info on those Seagates can be found in our post here (https://www.backblaze.com/blog/3tb-hard-drive-failure/). That said, we love the new Seagate drives we have in our system, and keep buying them in droves. We think there was just a bad run of drives across the board after the flooding in Thailand, and once those drives were "out of circulation" things seemed to calm down quite a bit. Nowadays we love 'em, and they're a great value.
 
..[ ]...As to Seagate from a previous comment, we had to migrate en-masse from their 3TB variant so you don't see them in our Q4 stats, though if you look at the "all-time" data, it'll appear there. More info on those Seagates can be found in our post here (https://www.backblaze.com/blog/3tb-hard-drive-failure/). That said, we love the new Seagate drives we have in our system, and keep buying them in droves. We think there was just a bad run of drives across the board after the flooding in Thailand, and once those drives were "out of circulation" things seemed to calm down quite a bit. Nowadays we love 'em, and they're a great value.
It's great to hear from you!

(And yes, perhaps somewhat selfishly) Especially when your experiences with, and opinion of, that run of Seagate 3 TB drives, coincides exactly with my appraisal of the situation. I've escaped both of Seagate's major misfortunes. The earlier "high capacity" drive bricking scandal, ( I forget the year), and the 3 TB "bricked at birth, fiasco".

I've had good luck with Seagate's single platter designs, all the way from 320 GB units, to my recent favorite, the 1 TB model. To me at least, they've always seemed to be a touch quieter, faster, and run cooler, than an equivalent capacity WD drive.
 
Last edited:
A long-winded discussion relating to a mechanical, man-made object of which there are various failures. Mankind has yet to build a perfect machine so it is pointless nit-picking of the products because even one or several of the most reliable products will have a failure.
 
A long-winded discussion relating to a mechanical, man-made object of which there are various failures. Mankind has yet to build a perfect machine so it is pointless nit-picking of the products because even one or several of the most reliable products will have a failure.

So because something isn't perfect we shouldn't try at all? The platitudes of reliability matter. There will never exist a perfect machine one, because perfection is an ever escaping concept. Two, machines by their definition are imperfect. Something that is designed to fit a specific roll can for obvious reasons never have a flaw, for the inherit flaw of that idea is that they only fill a specific roll.
 
Your reply is very silly, first, you have basically re-worded what I wrote, so where does it say we shouldn't try. Mankind advances every day to achieve perfection, ad infinitum, hopefully.
 
Your reply is very silly, first, you have basically re-worded what I wrote, so where does it say we shouldn't try. Mankind advances every day to achieve perfection, ad infinitum, hopefully.

"Mankind has yet to build a perfect machine so it is pointless nit-picking of the products because even one or several of the most reliable products will have a failure."

Prey tell, how does one improve without "nit-picking". Perfection is the lack of such nicks and thus your comment of avoiding it is paradoxical to what you said earlier.

So we should somehow make perfect products but not talk about their flaws? Good luck with that logic.

If you think my first comment rewords what you originally said you definitely need to re-read it.

Not that any of this matters because you just came here to piss on everyone's comments. You came to a tech website to state how worthless it is to discuss the particularities of tech. If you don't want to read the "long-winded" discussions, don't.
 
Your reply is very silly, first, you have basically re-worded what I wrote, so where does it say we shouldn't try. Mankind advances every day to achieve perfection, ad infinitum, hopefully.
Yeah, we advance toward self imagined godhood, until the time when we have to piss or procreate, then we're all monkeys again. BTW, where are you on that curve, if you don't mind me asking?
 
Two questions I have always had (I've not read everything in the blogs) is
Are DOA drives included in the failure rate and/or do you breakdown the failure according to number of hours/duty cycle (data/OS) at fail.
If your temperature data is 'POD limited' to 30C. I cannot begin to cool well enough for 30C average, so I wonder what the conclusion was (as opposed to my HGST's 4TB stack running 48-ish, and my manual fan boost when they reach 50C - as opposed to skipping it, since it is Utterly random as to when the single drive in center of stack heats up, I.e., Does >50C temp have adverse affect on longevity in your experience)
dunno if you'll see this, but TIA for reply.

Thanks for the article and follow-up
 
Back