Overclock.net banner

[ET] 3TB Seagate Hard Drives Have 43% Failure Rate, According to Backblaze

12K views 181 replies 96 participants last post by  parityboy 
#1 ·
Quote:
Cloud storage provider Backblaze, in its 2014 'Hard Drive Annual Failure Rate' graph, reports that the 3TB Seagate Barracuda drives it used for its storage service had a 43.1% failure rate over 2014. Other Seagate models also performed badly, with the 1.5TB Barracuda 7200.11 failing 23.8% of the time.

It must be noted that Backblaze exposed the drives to heavy stress, using them 24/7 to support user data. Seagate's drives, designed purely to serve as home external storage devices, are not expected to handle more than eight hours use a day, nor be used as part of a massive vibrating enclosure, so it is baffling why Backblaze would continue to use Barracuda HHDs for a task they are so ill-suited to.
Source
 
#2 ·
Maxed it out, used it for 3 years in 5 different systems, still works fine, but it is too small, so I removed it and put in a 5tb
 
#4 ·
That's...high. Are these numbers reflected in normal use aswell?
 
#5 ·
Quote:
It must be noted that Backblaze exposed the drives to heavy stress, using them 24/7 to support user data. Seagate's drives, designed purely to serve as home external storage devices
Why is this being reported when they were used by a cloud storage provider (as in enterprise)? Buy enterprise class HDD's next time then?
 
#7 ·
Quote:
Originally Posted by driftingforlife View Post

Didn't see that. My bad
Not problem. But it's even weirder they are using them in an enterprise usage while they know they are not meant for that. Can't be saving money on that with that high failure rate...
 
#10 ·
Quote:
Originally Posted by Stuuut View Post

Well they shouldn't be used for enterprise usage but that doesn't change the fact that the other drives they use worked just fine.
They might have gotten a lucky batch with the 1,5TB drives. Even that doesn't excuse their IT department for using them and then making some stupid "Year failure rates".

I'm going to buy an enterprise HDD, use it for 1 hour a day and then make an decade failure rate, would be as useful as this report.

Too bad I don't work for Seagate's PR dept. "Well, You're one dumb .... aren't you?" I know they can't say this out loud but c'mon...This is really stupid already.
 
#11 ·
Wait, are these seriously actually external drives? That means the numbers are irrelevant for the desktop version.
 
#12 ·
You guys missed the point by a mile. Backblaze does this report every year about how their consumer arrays stand up to the environment they're using them in. They're very aware of the pitfalls of using non-enterprise hardware for cloud storage. If they were truly concerned with going under because of the high failure rate they would change their strategy. Fact is, Seagate's consumer drives are less reliable under this particular workload than their WD and Hitachi counterparts.
 
  • Rep+
Reactions: Robenger
#15 ·
Quote:
Originally Posted by MadRabbit View Post

Why is this being reported when they were used by a cloud storage provider (as in enterprise)? Buy enterprise class HDD's next time then?
I believe I read on one of their prior reviews that they had compared the two at one point and the cost benefit wasn't worth it. It probably was another brand, but they do take these things into consideration. They buy an assortment and keep track of them all. Id have to go look it up, but it was one of those situations where the Enterprise versions were like 50% more money, but only had a 20% less failure rate kinda thing. So, they could spend a lot more money, deal with the warranty stuff, or go with the cheaper, and they realized it was cheaper to go with the cheaper ones.

I dont know how cloud based storage providers actually allocate their space. If they create smaller sub volumes for users, like I only need 10TB of storage, do they span that out on the same server that has 48x3TB ..... hmm but they need to do some raiding of the storage for redundancy... so they raid 6 or whatever the 48 drives into 40 (2 for parity bits, 6 standbys)? In any case I dono.... The point that I was trying to make, is that it could be normal usage. If i bought one for external storage, it might be treated the same way... not constant read writes... if you think about it (IMO), a lot of cloud based storage is backups. Depending on how they raid and all that, who knows how much the drive is actually working.
 
#16 ·
Quote:
Originally Posted by Mrzev View Post

I believe I read on one of their prior reviews that they had compared the two at one point and the cost benefit wasn't worth it. It probably was another brand, but they do take these things into consideration. They buy an assortment and keep track of them all....
I read the same article by them at one point. If I recall correctly, they were comparing 3TB Seagate drives removed from the "Expansion" series of external drives with the Constellation drives. The enterprise drives were approximately 2.5 times the price, so they were fine with the failure rate of the cheap Seagates.
 
#17 ·
Quote:
Originally Posted by kamimaru View Post

I can only concur with the numbers. My company stopped using seagate for our dedicated servers HDD. These disks were falling like flies past the 10K hours powered on.

Enterprises needs are different, more intensive. In a home usage you'll never reach those numbers.
This.

Seagate QC had gone to crap over the years and I refuse to allow them in my shop. I don't even risk them in my personal computers either. By and large, I use WD. Dead drives are largely a thing of the past for me for any drive under 5 years old. Of course for most new systems I put in, I just go with an SSD for general desktop use. Hard drives are now just "cheap storage" for me and most of the time either go as a "scratch drive/junk storage drive" for a system using an SSD as it's boot drive, or in a RAID array where a HDD failure isn't as horrible.
 
#18 ·
I have yet to have a single drive (aside from the horrific Maxxstor I had in 2003) die on me under /NORMAL/ home useage conditions. My system currently has 2 Seagate 1TB drives. Another older system has my friend's old 320GB. After three years one seagate has had some re-allocated sectors but is holding steady. I hope it will last another 2 years. My WD Green 1TB from 2009 is still returning a healthy stat profile (though I never flashed the newer parking firmware). I also still have 4-5 WD IDE drives (and a large array of barely used external IDE backup drives) that all pass extended SMART and other tests. Two IBM servers I received recently had all 6 WD Sata drives (160GB) at near failure. I think they've been running for almost a decade though
smile.gif


On the topic at hand: I guess I would be upset if I had used these things in a 24/7 production environment. For home use, I would never push the drives as much.
 
#19 ·
Why does it matter or not that it is a Backblaze report ?

Cheap Seagates fail significantly more often than other cheap drives (WD or Hitachi) under the same conditions.

When all cheap drives are used in the same manner, it becomes reassuring to know you can beat the crap out of them and they will still work (unlike Seagate).

In Seagate's defense, I've had a 500 GB and 250 GB drive for almost 9 years and they are still going strong.
 
#20 ·
Quote:
Originally Posted by Mrzev View Post

I dont know how cloud based storage providers actually allocate their space. If they create smaller sub volumes for users, like I only need 10TB of storage, do they span that out on the same server that has 48x3TB ..... hmm but they need to do some raiding of the storage for redundancy... so they raid 6 or whatever the 48 drives into 40 (2 for parity bits, 6 standbys)?
More like clusters of servers with a smaller disk space actually : single disk failure is ok with RAID, single server failure also OK with replicas.
Quote:
Originally Posted by Horusrogue View Post

I have yet to have a single drive (aside from the horrific Maxxstor I had in 2003) die on me under /NORMAL/ home useage conditions. My system currently has 2 Seagate 1TB drives. Another older system has my friend's old 320GB. After three years one seagate has had some re-allocated sectors but is holding steady. I hope it will last another 2 years. My WD Green 1TB from 2009 is still returning a healthy stat profile (though I never flashed the newer parking firmware). I also still have 4-5 WD IDE drives (and a large array of barely used external IDE backup drives) that all pass extended SMART and other tests. Two IBM servers I received recently had all 6 WD Sata drives (160GB) at near failure. I think they've been running for almost a decade though
smile.gif


On the topic at hand: I guess I would be upset if I had used these things in a 24/7 production environment. For home use, I would never push the drives as much.
My drives at home are powered off when not in use for a small period that saves energy and lifespan. No wonder they last long. And that's good because a damaged disk can then last longer than expected.
 
#21 ·
Quote:
Originally Posted by MadRabbit View Post

Not problem. But it's even weirder they are using them in an enterprise usage while they know they are not meant for that. Can't be saving money on that with that high failure rate...
The only difference is the firmware and warranty (and cost). Pretty minimal and so the drives can be taken as a close enough approximation as the enterprise drives.
 
#22 ·
Quote:
Originally Posted by MightEMatt View Post

You guys missed the point by a mile. Backblaze does this report every year about how their consumer arrays stand up to the environment they're using them in. They're very aware of the pitfalls of using non-enterprise hardware for cloud storage. If they were truly concerned with going under because of the high failure rate they would change their strategy. Fact is, Seagate's consumer drives are less reliable under this particular workload than their WD and Hitachi counterparts.
This makes no sense what so ever from a company perspective...Why in the world would any company in the world spend money on things they know ARE NOT suitable for them?
 
#23 ·
Quote:
Originally Posted by MadRabbit View Post

This makes no sense what so ever from a company perspective...Why in the world would any company in the world spend money on things they know ARE NOT suitable for them?
As mentioned previously, they already did a test and determined it was more cost efficient to just eat the failures on the bargain drives.
 
#24 ·
So once again, if thats so "cost effective" for them why are they even reporting those failures if they know those dont belong there? This is an totally useless report to start with like I said earlier, since no comapany in their right mind wont do that.
 
#25 ·
Quote:
Originally Posted by MadRabbit View Post

So once again, if thats so "cost effective" for them why are they even reporting those failures if they know those dont belong there? This is an totally useless report to start with like I said earlier.
Data is data, just gives us more statistics to consider. This report goes to show that WD and Hitachi have more rugged budget drives than Seagate by quite a large margain. Is this the same as home use? No. Is it statistically more likely for a cheap Seagate drive to fail before a cheap Hitachi drive? Yes. I don't see why you're so upset that this company is giving us free stress testing statistics.

Edit: Without this data, somebody building a small home storage server might make the mistake of thinking that the consumer Seagates could handle such an environment. Now we know otherwise.
 
#26 ·
Quote:
Originally Posted by MightEMatt View Post

Data is data, just gives us more statistics to consider. This report goes to show that WD and Hitachi have more rugged budget drives than Seagate by quite a large margain. Is this the same as home use? No. Is it statistically more likely for a cheap Seagate drive to fail before a cheap Hitachi drive? Yes. I don't see why you're so upset that this company is giving us free stress testing statistics.
Im not upset. Im just baffled how people read into "statistics" that shouldnt even matter.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top