How long do hard drives actually live for?

[mouser]

Extremetech has an article today on an issue we've discussed before -- the longevity of hard drives.

This data covers only the first 4 years, but i have to admit it's a bit scary to me:

If you buy a hard drive today, there’s a 90% chance that it will survive for three years. If your drive makes it to the three-year point, you would be wise to back up your data, as there’s a 12% chance per year that your drive will die.

If you don't have a good backup plan in place, start one this week.  I recommend a combination of monthly full drive imaging combined with a constant online backup service for your documents and/or file mirroring.

http://www.extremetech.com/computing/170748-how-long-do-hard-drives-actually-live-for

from slashdot.org

[4wd]

It’s also worth mentioning that Backblaze’s drives are spinning constantly — these failure rates are for drives that are turned on 24/7. Your home computer probably isn’t powered up 24/7, and thus the drives may last longer.

It's nice that they mention it but saying your drives may last longer is a little misleading, IMHO.

Until someone comes up with some data that actually relates to consumer drives and the way they are generally intended to be used then I would have expected the failure rate for them to be slightly higher with the power on/off cycling that a normal consumer computer goes through.

[mouser]

I'm not sure exactly what you're saying -- are you saying that you think consumer drives might be expected to die faster than the backblaze drives, or?

[4wd]

I'm not sure exactly what you're saying -- are you saying that you think consumer drives might be expected to die faster than the backblaze drives, or?

They use consumer drives in a climate controlled environment - their usage scenario is completely different from a normal consumer.

Theirs are constantly spinning 24/7, as they mention - a normal consumer turns off their computer when they're finished, they turn it on when they want to use it.

Power on/off causes components to expand and contract, (resistors, PCBs, etc), as they start at close to ambient temperature and heat up as time goes on.  This movement while extremely small can be enough to stress either the solder joints or the components themselves - something a constantly on device isn't subjected to.

You also have Inrush current^w when a device is turned on - if you look at the manufacturers data for HDDs, Startup current is higher than what's required to keep it spinning - this current, while used predominantly by the motor and its associated components, could possibly affect other components that aren't rated to keep handling such transient surges.

Evening things out though is the comparison of the time BackBlazes HDDs spend spinning, (24/7), and the time a normal consumers HDD is spinning, (eg. 6-18 hours per day, 5-7 days per week, etc).

All I'm saying is that until someone does a comparable study of consumer drives and how they're generally meant to be used then saying they may or may not last longer is a bit pointless.

The only thing you can say about a HDD, (and any storage device), is that it will fail at some point ... and that's what you plan for.

[40hz]

I think 4wd is pointing out there's likely to be a difference in reliability between "consumer" and "enterprise" grade drives intended for workstation and server use. Especially since it's now cheaper for a manufacturer to play the odds and replace failures under warranty than to extensively engineer or test consumer hard drives. That's what "risk management" gets us.

My experience puts useful life of office store type drives between three and five years average.

SATA drives by the same manufacturers intended for server deployment average about seven years before experiencing problems or throwing a warning. They usually have more robust motors and spin at the slower 5300 speeds to keep heat and vibration down.

But a good part if the extended service life might also be attributable to servers usually being kept in climate controlled areas with very clean and tightly regulated power lines. Not something you can assume is the norm for home computers.  


ADDENDUM: whoops! 4wd got in while I was typing!

[40hz]

@ Mouser- IIRC Backblaze uses pretty much nothing but consumer grade hardware in their storage PODs and simply maintains a lot of redundancy and has excess capacity to offset the increased likelihood of hardware failures. More risk management at work.

[4wd]

@ Mouser- IIRC Backblaze uses pretty much nothing but consumer grade hardware in their storage PODs ...

No need to recall, it's mentioned in the article:

It’s worth noting that Backblaze uses normal, consumer-level drives — the kind of drives with 12- or 36-month warranties.

[IainB]

Whilst a disaster contingency strategy of backing-up everything in anticipation/assumption of (say) a 3 or 4-year lifespan for a drive would seem to be prudent and could give you some peace-of-mind, it would have the disadvantage that the strategy would apparently be based on uncertainty - a belief/expectation/anticipation of a failure event at some future guessed-for date. As a strategy that is rather like "working blind" or "in the dark".

The only disaster contingency backup that will be of productive use is likely to be the last backup which chanced to be taken before the disaster (disk failure event) - always assuming that you can use that backup to make a recovery from (and how often does one test for that failure?).

All the other unproductive disaster contingency backing-up and administration of same is going to be time consuming, and the consumption of the necessary unproductive backup resources (e.g., including man-hours, CPU-secs., hardware, and on/off-site space rental costs) are likely to be cumulatively expensive too.

However, if, as well as normal operational backups, you have a tool that is monitoring the state of your hard drives in realtime and which will report faults as soon as deterioration starts to set in, and before failure occurs, then one could arguably be in much better control of risk mitigation.
So, for most PC-users and small client-server operations, I would strongly recommend consideration of something like Hard Disk Sentinel PRO, which is relatively inexpensive.
See here: Hard Disk Sentinel PRO - Mini-Review.

[Edvard]

From the anecdotal-tale-to-the-contrary-department: I still have a handful of less-than-1GB drives that still fire up just fine, though they don't hold much, and I can count on that same hand how many disk drives I've had fail while it was installed and being used.  Every time I've had a disk fail it was because I had taken it out to install a newer, larger one, and later tried to install it in a different machine. 
CD-ROM drives, on the other hand... let's just say I should buy them in bulk. 

[IainB]

@Edvard: Yes, I have never known a hard drive to fail, though it has happened to some people that I know of.
I reckon hard drives are probably over-engineered and likely to last ages - at least with the type of intermittent usage on PCs. I have some pretty old 2½" laptop drives lying around, mostly from 3+ year old/dead laptops, and they all still seem to work fine when I fire them up. (I usually keep them as long-term archives.)
From memory the first and only trouble I have ever had with a hard drive was with the bad blocks on the 5700rpm 7200rpm. drive in my HP ENVY 14 laptop - which was mentioned as being reported on by Hard Drive Sentinel in the review I linked to above. The drive is still working fine with no further trouble, the NTFS "self-healing" of the bad blocks apparently having been catered for by the Win7 operating system.

[techidave]

would it be safe to say that the newer, bigger hard drives with more storage space run hotter than the older 80 gb or less hard drives of yesteryear?  I have taken apart lots of older 30 and 40gb drives and they only have 2 platters in them.  If I am not mistaken, the newer 500 gb drives have 3 or 4 platters in them, thus creating more heat.
and then those enterprised drives that I use in my servers weigh in at twice what the consumer models do.  I am talking 3.5" black label vs blue label WD drives of the same storage size.

[Stoic Joker]

They usually have more robust motors and spin at the slower 5300 speeds to keep heat and vibration down

Where are you finding 5300rpm SATA drives? My WD3200YS RE drives are 7200rpm, and I generally always went for the lower speed drives because I assumed (as you mention) that they'd last longer. Even the Nearlines I used for the storage array of the new server are 7200rpm.

---------------------------------------//----------------------------------------

Also chiming in from the anecdotal department...I've got a couple of consumer level drives in my lab that have been spinning 24/7/365 for close to 10 years. I think they may be zombies...because they should be dead by now...but they just keep going ... Should I shoot them in their little read heads?

[tomos]

Where are you finding 5300rpm SATA drives?

normal SATA drives? Drives with the "Green" description are usually 5300rpm
I have a couple here from Samsung ("EcoGreen") and one from Seagate ("Barracuda Green").

e.g. (no date on this review - hate that...)
http://www.storagereview.com/samsung_spinpoint_f4eg_review_hd204ui

[40hz]

@Stoic- yup, you're right. I forgot when they incorporated vibration compensation and enhanced alignment features, the spindle speeds in data center/enterprise drives went up and are now all either 7200 or better. They're also engineered for much higher continuous heat levels than consumer drives ( ~90°C max) so it's not really an issue for this class of drive any more.

Gotta stop dwelling in the past I do.

[tomos]

^ should have known that was not about consumer drives...

______________________________

@ Mouser- IIRC Backblaze uses pretty much nothing but consumer grade hardware in their storage PODs and simply maintains a lot of redundancy and has excess capacity to offset the increased likelihood of hardware failures. More risk management at work.

With the harddrive shortage Backblaze just started using everything and anything.
The last time they recommended drives was in February this year - at that stage they were using enterprise I think.
Hang on...

http://blog.backblaze.com/2013/02/20/180tb-of-good-vibrations-storage-pod-3-0/

Here is a list of the drives we have tested and currently use:

   4 TB drives

   Hitachi – HDS5C4040ALE630 (Just starting to use these, but they look good.)

   3 TB drives

   Hitachi- HDS5C3030ALA630 (low power)
   Hitachi – HDS723030ALA640 (high power)
   Western Digital – WDC WD30EFRX‑68AX9N0 (RED)
   Seagate – ST3000DM001‑9YN166 (slightly higher failure rate)

__________

as for now - from their blogpost (via link in OP)

Types Of Hard Drives In The Analysis

Backblaze has standardized on “consumer-grade” hard drives. While hard drive companies say these drives are not designed to work in RAID arrays or the 24×7 workload of a data center environment, Backblaze uses software redundancy to protect data. In a future blog post we will delve into the statistics comparing “consumer” and “enterprise” hard drives.

my emphasis
http://blog.backblaze.com/2013/11/12/how-long-do-disk-drives-last/

[Stoic Joker]

Gotta stop dwelling in the past I do.

Na... It's fun! Besides I think you're just screwing with me to make sure I'm awake... ...Which I wasn't...which I guess is why we just had a conversation about basically mythical 5300rpm drives. Instead of the 5400 rpm variety which are actually manufactured. (Tehehe - Oops!)

[40hz]

Gotta stop dwelling in the past I do.

Na... It's fun! Besides I think you're just screwing with me to make sure I'm awake... ...Which I wasn't...which I guess is why we just had a conversation about basically mythical 5300rpm drives. Instead of the 5400 rpm variety which are actually manufactured. (Tehehe - Oops!)

Guess who's had two hours sleep in the last 24...c'mon guess!

Besides, what's another 100rpm give or take once you clear 5000 anyway.

(Boy do I ever need 6 straight hours of sleep!)

[IainB]

Oops. Sorry. I just noticed that in my comment to @Edvard (above) I had erroneously given the disk rotation speed as:  5700rpm.
Corrected now to what it should have been: 7200rpm.

[tomos]

The harddrive rotation speed plot thickens....

[IainB]

I did a DuckGo search for "normal operating temperature for hard drives" and came up with lots of useful results. For example, this one from Seagate:
(Copied below sans embedded hyperlinks/images, with some of my emphasis.)

What is the normal operating temperature for Seagate disk drives?

Discusses the normal parameters for operating temperatures for Seagate drives.

The drive should never exceed the temperature ranges below. If the drives ever exceed these temperature ranges then the drive is considered "overheated" or is not getting adequate air flow from your current case environment.

With our newer model drives the maximum temperature is now at 60 degrees Celsius.

The operating temperature range for most Seagate hard drives is 5 to 50 degrees Celsius. A normal PC case should provide adequate cooling.

However, if your enclosure is unable to maintain this range, we suggest that you contact your system manufacturer for information on cooling and ventilation hardware that is compatible with your specific configuration.

The answer to this question depends on your case environment. If you have adequate cooling, it is probably not necessary. If you feel that you need additional cooling, use your favorite internet search engine and enter the keywords "drive bay cooling kit".

REFERENCE TO THIRD PARTIES AND THIRD PARTY WEB SITES. Seagate references third parties and third party products as an informational service only, it is not an endorsement or recommendation - implied or otherwise - of any of the listed companies. Seagate makes no warranty - implied or otherwise - regarding the performance or reliability of these companies or products. Each company listed is independent from Seagate and is not under the control of Seagate; therefore, Seagate accepts no responsibility for and disclaims any liability from the actions or products of the listed companies. You should make your own independent evaluation before conducting business with any company. To obtain product specifications and warranty information, please contact the respective vendor directly. There are links in this document that will permit you to connect to third-party web sites over which Seagate has no control. These links are provided for your convenience only and your use of them is at your own risk. Seagate makes no representations whatsoever about the content of any of these web sites. Seagate does not endorse or accept any responsibility for the content, or use, of any such web sites.

This rather makes sense, and in the case of the 7200rpm 500Gb 2½" hard drive in my HP ENVY 14 laptop, it corresponds with snapshots of the daily temps, as below. These are reports from Hard Disk Sentinel PRO:

Daily average temps.:
[attachthumb=#]

Daily max temps.:
[attachthumb=#]

[40hz]

Intel has published a good paper on the differences between enterprise and desktop disk drives. 10-pages of good reading for any who might be interested. (Copy attached. It's small.)

[attachurl=#]

[IainB]

@40hz: Thanks for that. It's all an education for me.
Because I have some problems (identified by HDS PRO) with the HP ENVY 14 laptop hard disk (see here) I was looking at the Performance and Health weighting factors that are applied. The HDS website has a good description here - Health calculation | Hard Disc Sentinel, and differentiates between server disk drives and desktop disk drives.

...The hard disk has 100% condition initially. All critical health-rated S.M.A.R.T. parameters (if they are available, it depends on the manufacturer) decrease this value. These attributes have a pre defined weight and a maximum limit value (the latter defines the maximum degradation in health value for the attribute). The overall health percent value is calculated by multiplying the remaining percent values (100 – degradation %).

Currently, Hard Disk Sentinel has two different such methods. By using the default method, the weights and limits (see below) are lighter. If the more strict, recommended for servers option is used the values are much more strict, the problems may reduce the health much drastically. ...

So, presumably a conventional approach to answering the question "How long do hard drives actually live for?" would be to differentiate between the two types (server disk drives and desktop disk drives) in some similar manner, and analyse and assess the statistical life expectancy and performance correspondingly.

[xtabber]

Intel has published a good paper on the differences between enterprise and desktop disk drives. 10-pages of good reading for any who might be interested. (Copy attached. It's small.)
(see attachment in previous post)

This paper is some 5 years old, which is an eon or two in high tech.  If your focus is on performance, you are more likely to be using SSDs than HDDs today.

If the main thing you care about is reliability, there really is little difference, which is why most of the big cloud companies don't bother with enterprise drives any more.  Google led the way after noting that failure rates were pretty much the same for both types.  Let's just say they had a VERY large sample to test and compare.

Hard drive failure rates are reported as predicted statistical distributions - MTBF or AFR - look those up if you are having trouble sleeping at night and counting sheep isn't working - which means that you only have a projected probability of a drive failing at any time.  That probability goes up as the drive ages, but it remains just a probability.  

Your 4 year-old drive will have a much higher probability of failure during the next year than a 1 year-old drive, but either of them could fail tomorrow, which means that you should be just as well prepared for failure in a brand new drive than in an older one.

[40hz]

So, presumably a conventional approach to answering the question "How long do hard drives actually live for?" would be to differentiate between the two types (server disk drives and desktop disk drives) in some similar manner, and analyse and assess the statistical life expectancy and performance correspondingly.

Pretty much. It all comes down to MTBF and probabilities, so there isn't a fixed answer. Just degrees of confidence with varying odds.

Most modern drives are pretty reliable regardless.

[40hz]

This paper is some 5 years old, which is an eon or two in high tech.  If your focus is on performance, you are more likely to be using SSDs than HDDs today.

I was primarily addressing enterprise or datacenter use. In that environment reliability always takes precedence over performance in all but a few very special situations.

SSDs are generally not considered suitable for full scale deployment in a datacenter settings at this point in time.

Also, although the Intel paper may be a few years old, the engineering considerations and operational concerns it addresses are still just as valid today as they were when it was first published. Product specs may change. But vibration, heat, 'wear & tear' and entropy are still unavoidable concerns no matter what.