Backblaze Publishes Q1 2018 Hard Drive Stats

R

rgMekanic

[H]ard|News
Joined
May 13, 2013
Messages
6,943
Since 2013 Backblaze has published the data from its massive army of hard drives, and today they have released the results of Q1 2018. As of March 31, 2018 they had 100,110 drives, and the 26 GB of SMART data lets us see how these bits of spinning rust are holding up.

Looks like there is less reason to be scared of big 8TB+ drives than many have thought. I really need to upgrade my storage.

The failure rates of all of the larger drives (8-, 10- and 12 TB) are very good, 1.2% AFR (Annualized Failure Rate) or less. Many of these drives were deployed in the last year, so there is some volatility in the data, but you can use the Confidence Interval to get a sense of the failure percentage range.
 
Looking good thus far for the Seagate 10TB Enterprize drives.

I like to see that.

I bought 12 Seagate ST10000NM0016 end of last year (these are essentially identical to the ST10000NM0086 drives Backblaze has, but also come with the encryption feature for quick erase purposes)

Thus far they have 106k drive days with no failures. I'm hoping that holds up for me as well. My 12 drives have ~47k hours of uptime between them thus far.

In the lifetime chart, they actually have the lowest failure rate of all drives!

I was hesitant to buy Seagate drives having heard so much about their lack of reliability, but I have a good feeling about these.
 
I'd still be wary of the 10 and 12tb drives for a little while. The article stated that most of them were deployed less than a year ago, so the data may be a little skewed.
 
I just built a DAS TB3 appliance to hook up to a NUC as a home server and filled it with 8TB drives, nice to know they aren't about to go pop any time soon. :)
 
I need a few of those 10's...5 would be good for starters. My NAS has all 3TB WD Reds in it now.
 
I got a 6TB white label from NewEgg last year as a media drive experiment in my X79 rig. So far so good. After reading a bunch of reviews I dropped a WD Red 8TB in our Kodi box and it's been a champ too(but as some reviews stated you really can hear the step motor)-and this sucker gets a lot of use.
 
Showbiz said:
I'd still be wary of the 10 and 12tb drives for a little while. The article stated that most of them were deployed less than a year ago, so the data may be a little skewed.
Click to expand...


Well, the Seagate Enterprise drives come with 5 years of warranty, so I'm not too worried.

It would take a spectacularly unlikely series of simultaneous failures for me to lose data
 
I've had good luck with the 6TB & 8TB 5400 RPM WD red drives (used in my backup servers)

Also with the WD Gold 7200 4TB drives in production.

Couple dozen drives and no failures so far.

Pretty much the only server drives (other than a few SSD's ) I've bought the last couple years.
 
nutzo said:
I've had good luck with the 6TB & 8TB 5400 RPM WD red drives (used in my backup servers)

Also with the WD Gold 7200 4TB drives in production.

Couple dozen drives and no failures so far.

Pretty much the only server drives (other than a few SSD's ) I've bought the last couple years.
Click to expand...


My 12x 4TB Reds did OK for the 3-4 years I had them in my main server.

None of them outright failed, but two of them started getting a very slight amount of read/write errors, and were immediately replaced under warranty.

So, 1/6 disks replaced in almost 4 years. Not great, but not terrible either.
 
It's nice to see that HGST still has a decent showing after being picked up by WD.
 
The real takeaway is that hard drives have gotten incredibly reliable. You really can't go wrong with any of the major players.
 
parawing742 said:
The real takeaway is that hard drives have gotten incredibly reliable.
Click to expand...

Agreed. But even the best hard drives are susceptible to user volatility. That is, the most douchebag volatile user will be the one whose hard drive fails, and it won't matter how good the brand was when your IT director is throwing you under the bus.
 
I'm not a fan of the current HGST 6TB NAS. p/n 726060ALE614

Its very loud and powers down when set to run with 24/7 uptime.
Its not a good drive for a gaming PC just because of the noise! I had to mount it loose on packaging to prevent anything it touches vibrating. The seek is loud as well.
I often get a delay while the drive powers up even when not trying to access that drive as my swapfile is located on it.
Uptime stats for this drive should be disregarded as a huge chunk of that time the platter may not be spinning.
It feels like WD have dropped the quality in numerous ways and to compensate have reduced uptime.
Not impressed.
 
Look at those sample sizes, that chart is perilously close to completely useless.
 
Twisted Kidney said:
Look at those sample sizes, that chart is perilously close to completely useless.
Click to expand...
They have some more comprehensive data available, which includes all of 2017, 2016, 2015, and one for April 2013 through 12/31/2017 that have far more drives for *some* of the models. It is located HERE just scroll down. Most models fare well, a few not so well.
 
parawing742 said:
The real takeaway is that hard drives have gotten incredibly reliable. You really can't go wrong with any of the major players.
Click to expand...

Very few failures on my servers the last couple years, unlike 4-5 years ago when I had several failures of the Dell rebranded Seagate drives, and a few Dell branded WD drives.

I still have a stack of working 250GB, 500GB & 750GB drives I retired the past few years, and now I'm starting to retire the 1TB drives.

Glad they have been reliable enough that I'm now more likely to retire a drive due to the size, than to have them fail.
 
Twisted Kidney said:
Look at those sample sizes, that chart is perilously close to completely useless.
Click to expand...


Not really. Unless your standard deviation is very high, you don't need massive sample sizes for inferential statistics.

Usually less than 30 pieces is more than enough.

Here, we are dealing with attribute data analyzed using the binomial distribution. For that, I'd say you probablt want at least 60 pieces (well, 59 to be precise) to hit 95% confidence at 95% coverage.

So, most of these have large enough sample sizes for analysis. Some of them are on the low side though.
 
You must log in or register to reply here.
Back
Top