How long do hard drives actually live for?

[Stoic Joker]

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)

-40hz (November 13, 2013, 08:59 PM)

It starts a bit slow, but has some good points on why one should not skimp on hardware. So while the enterprise class drives have sticker shock price tags...there apparently are actually valid technical reasons for them.

[40hz]

So while the enterprise class drives have sticker shock price tags...

-Stoic Joker (November 14, 2013, 07:02 AM)

I didn't think they were that much more expensive than standard drives for what you were getting when I quick price checked a few online.

Seagate Constellation 7200 RPM 128MB Cache SATA 6.0 @ $115 / 2TB @ $196 / 4TB @ $358

WD WD2000FYYZ 2TB 7200 RPM 64MB Cache SATA 6.0 @ $199

etc. etc. etc...

I guess it's all about expectations and what you're used to seeing?

[mouser]

Let me summarize some conclusions from this and other threads:

• The chance of a catastrophic hard drive failure in any given year is always non-trivial -- people who think there is a one-in-a-thousand chance of hard drive failure are fooling themselves.  Hard drives still fail pretty frequently.
• Sometimes you get some warning -- which is why it's useful to have a SMART drive monitor tool, and to be aware of sounds, etc.  At the first sign of trouble, clone and replace the hard drive and put the old one on the shelf.
• One key question that is a little closer to being answers is one I raised in my post "Should we preemptively retire old hard drives".  This latest article would seem to suggest that after a few years, the rate of failure seems to start increasing, and it really might be a good idea to preemptively replace drives that are 4-5 years old, rather than to assume that a long-lasting drive is something special that will run forever.
• Because it seems clear that for almost all pc components, very high temperatures are bad, and because this is one area where we actually have a little control over -- it's the one area i have started paying more attention to in last years.  So I run tray-resident temperature monitoring software that shows temperatures of all hard drives as well as the cpu, and alerts when they get too high.
• We do not have good data about whether some drives will just run forever.  And we have no good data about which brands and types live longer.

[Stoic Joker]

So while the enterprise class drives have sticker shock price tags...

-Stoic Joker (November 14, 2013, 07:02 AM)

I didn't think they were that much more expensive than standard drives for what you were getting when I quick price checked a few online.

-40hz (November 14, 2013, 09:04 AM)

For sourcing your own perhaps not, but if you're configuring a server from Dell/HP/etc. the price jump between it's a disk (300GB $120) vs. nearline (300GB $250) vs. enterprise (300GB $380) 10k/15k speeds) will make most people wince visibly. Now prices given are approximate (based on memory), but the jumps are within $20 of what I ran into last time a configured a server from HP ... And are why I go (7200rpm) nearline for storage arrays.

[40hz]

For sourcing your own perhaps not

-Stoic Joker (November 14, 2013, 11:54 AM)

To which I say: DIY or Die!

And our clients love us for it!

[40hz]

One key question that is a little closer to being answers is one I raised in my post "Should we preemptively retire old hard drives".  This latest article would seem to suggest that after a few years, the rate of failure seems to start increasing, and it really might be a good idea to preemptively replace drives that are 4-5 years old, rather than to assume that a long-lasting drive is something special that will run forever.

-mouser (November 14, 2013, 09:19 AM)

I'll agree to proactive replacement as long as it's not based strictly on age. I've actually seen more early catastrophic drive failures (1 to 3 months into their service life) than I've seen unexpected failure with drives that have been in service for three to five (or more) years. Like the saying goes, most electronic failures occur very early or very late in life. Electronics hardly ever fail during their midlife.

For a rule of thumb I'd say proactive replacement is best done whenever you start getting that nagging feeling something is wonky with a drive - or when you migrate to a new server or workstation. NEVER recycle mission critical mechanical parts (i.e. HDs) into a new box. Always purchase new drives, and repurpose properly working old ones into less critical roles.

Just my anyway.

 

[sinum]

One online Storage-Provider, using none Server Disks, post his personal stats before some days. As i remember, about 4% fails within 1st 1,5 Years than the fails go down to 1,5% and max lifetime is about 5-6 Years. This means very high risk after 4-5 Years on heavy usage. Maybe someone get this Article again i don't. 

[mouser]

most electronic failures occur very early or very late in life. Electronics hardly ever fail during their midlife.

yes but wouldn't that support the idea of preemtively decommissing a hard drive after 5 or 6 years, whether its displaying any signs of trouble or not?

needless to say, all of this talk about preemptive decomissioning is predicated on the idea that having a hard drive suddenly die on you is a very bad thing to have happen -- something which is true in most cases.

[40hz]

yes but wouldn't that support the idea of preemtively decommissing a hard drive after 5 or 6 years, whether its displaying any signs of trouble or not?

-mouser (November 14, 2013, 02:11 PM)

Yes, but we're still talking probabilities and distribution curves.

However, if I do understand what you're saying, then yes...it probably wouldn't hurt to replace most drives after 5 years  - although it probably wouldn't absolutely decrease the likelihood of a 'bad surprise' as much as we'd hope. Drives seldom go belly up without at least a few days of sending you an indication that something is going south. At least from my experience they do.

Like I said earlier - if you find yourself worrying about a drive, it's probably a good time to think about a replacement. Those subconscious trouble signals you pick up are often worth paying attention to once you have some experience under your belt.

Not terribly scientific I'm afraid. But it's an approach that's worked well for me. YMMV.

[xtabber]

If a drive is 5-6 years old, it is a couple of HD generations old, which means it can be replaced by a drive that uses less power, runs cooler and has higher data transfer rates, to say nothing of costing much less for the same or greater capacity.

I've never replaced a drive preemptively because I was worried about it failing, but I do upgrade drives from time to time. The older drives are then used for off line (e.g., backup) storage.

[Stoic Joker]

Like I said earlier - if you find yourself worrying about a drive, it's probably a good time to think about a replacement. Those subconscious trouble signals you pick up are often worth paying attention to once you have some experience under your belt.

-40hz (November 14, 2013, 02:22 PM)

I'll second the voodoo science/trust your instincts (e.g. don't ignore that funny feeling) angle, as it has server me well in the past (and bit me a few time when ignored).

Here's one (along those lines) that happened yesterday. Client has a computer lab with 15 machines, that were all (old) donated boxes because they are a nonprofit outfit. They let another group user their lab for a training session, and after the session, the folks then shutdown the machines to be considerate. The machines were supposed to be left on for the nightly maintenance routines to run ... so that hadn't actually been off for about 6 months.

The next morning - after finally cooling completely overnight - 7 of the 15 machines were completely - hardware failure - dead. 3 HDDs, 2 motherboards, 1 question mark, and we're pretty sure the last one is a zombie because it should have been dead months ago given how flakey it acts.

[40hz]

The next morning - after finally cooling completely overnight - 7 of the 15 machines were completely - hardware failure - dead.

-Stoic Joker (November 15, 2013, 07:33 AM)

Sounds like they built up some serious condensation overnight. Especially if this noon-profit throttles down the AC (or shuts it off) to save on power when nobody's around. Turn an older PC on when that happens and **POP**

Similar to the printer 'service issues' we used to get with some clients during the summer. Invariably after a 3-day weekend. Their landlords would seriously cut back on the building's AC over long weekends, and that extra day of increased humidity would make the paper in the printer bulk up. Tuesday morning our phone would be ringing off the hook with service calls for continuous paper jams. We tried telling them to just remove the paper that was in the tray and load a fresh unopened ream. But NO...they wanted it checked.

You know the type:


I want them to get their asses in here - and I want it NOW!



Picked up a lot of 'routine maintenance' service trips that way. Quick $75 half-hour invoice.

Except in cases where some irate yahoo decided to forcefully rip some jammed sheets out of a printer, and not power off first to disengage the drive mechanism. Those service trips were much more...um...profitable.

At least for us.

[IainB]

... Like I said earlier - if you find yourself worrying about a drive, it's probably a good time to think about a replacement. Those subconscious trouble signals you pick up are often worth paying attention to once you have some experience under your belt.
Not terribly scientific I'm afraid. But it's an approach that's worked well for me. YMMV.

-40hz (November 14, 2013, 02:22 PM)

Well, there is worry based on ignorance and worry based on evidence. I'd go for the latter every time, which is when I find HD Sentinel so useful.

[40hz]

^Hunch, vibe, or utility - whatever works best for you since it's your data.  

FWIW there are bad vibes and then there's bad vibes. If you do this stuff for a living the vibes you get (which are really more the product of semi-subconscious observation and analysis linked to a deal of real world working experience) are significantly different than what the average computer user likely feels when they get them. And athough we jokingly refer to it as getting a vibe or having a hunch, it's actually a lot more than that.  

Oh yes, we almost always try to confirm it with a test utility before we take action on what we're intuiting too.  


Yeah...it was that Adtran I had a
bad feeling about yesterday

[SeraphimLabs]

SSDs have come a very long way though. I've begun deploying SSD based workstations where I work because of their speed, in workstations designed for 3 years of service before being repurposed. The MTBF of the devices I am using is comparable to a conventional drive, and should at least make that first 3 years problem-free, while their actual use is such that if one does die prematurely I can have it running again off a spare drive in a couple of hours. After the three years it will be interesting to see how they age.

At the same time, my oldest working system has been running very nearly 24/7 for the past 8 years, and is still operating the very same hard drives that it has had the whole time- one of which is actually more than 12 years old and still going strong after being given to me because it was 'dead'.

Sure they don't all last that long, gotta run the statistics. But those that do survive a few years often go for a very long time afterward, it's not like a SSD where it will suddenly lose its ability to accept new data due to accumulated wear.

Also it has been a strange month for the electronics. I've had no less than 4 video hardware related failures in November alone, after going many years without ever losing a video card in active service. A workstation that gets powered off every night also turned up one morning with a dead mainboard as well, I've not had that happen in a while either.

[IainB]

@SeraphimLabs: Have you ever tried restoring/recovering hard drives or SSDs using Spinrite?

[40hz]

@SeraphimLabs: Have you ever tried restoring/recovering hard drives or SSDs using Spinrite?

-IainB (November 26, 2013, 05:08 PM)

Did somebody say Spinrite?

Uh-oh...here comes f0dder! 

[IainB]

I was looking for a review of Spinrite on the DC Forum - an actual "use case" - but couldn't find one.
Sure, some positive comments from some people, and mostly negative comments from @f0dder, but not exactly good "reviewer" reports.
You can't establish by hearsay, opinion poll or argument whether a software tool performs its allotted task properly. You need some hard facts first - so, in the absence of a review, I usually do a "suck-it-and-see".

I searched various forums on the Internet to see who had tried Spinrite out and what problems they had, and decided I should give it a whirl. So I bought a copy.
My notes from this trial:

2013-11-22 1039hrs: After a great deal of trouble, finally created a Spinrite CD:
Could only do it on an imation DVD-R 16X4.7GB 2hr DVD.
Had to format it first.
Then used Easy Burner (Free) to write the Spinrite .iso file to the DVD.

When I booted and ran the Spinrite disc at boot, the first time, nothing happened, then the second time it worked (started to load the DOS), but stopped quickly with this error (this was consistently repeatable):

-InitDisk
Invalid Opcode at FFFE 059D 3006 00E2 96D5 0008 0000 96D5 0000 6244 0B63 6218 96D5

Did a search for: "spinrite -InitDiskInvalid Opcode" and came up with lots of stuff  - e.g., "How to Fix Invalid Opcode Error in SpinRite".
Several people had found that:

...go into the BIOS, to the SATA mode setting, and change it from IDE to AHCI.  When I did this and then booted SpinRite again, I noticed that a bunch of the drive fields (hardware addrs, hardware irq, etc) were now listed as "unknown", but SpinRite was able to scan the disk now.
_____________________________

Essentially, the various comments on this indicated that, if you could get at the drive's BIOS settings and tweak them experimentally, then the Invalid Opcode would not occur, and Spinrite would be able to scan the disk.
However, in my case, though the user is able get into the BIOS (no problem), there is no facility to access/tweak the BIOS settings for the hard drive, so I never got to find out for myself whether Spinrite could do its stuff as documented.
So I shall ask for my money back - not because the product is "no good" (I don't know that), but because it couldn't run on the drive I had specifically wanted to use it on.

That is why I asked:

@SeraphimLabs: Have you ever tried restoring/recovering hard drives or SSDs using Spinrite?
___________________

-IainB (November 26, 2013, 05:08 PM)

I am genuinely interested to know. What the Spinrite author says when describing what his software does actually makes a lot of sense to me - at least, I think I understood what he was talking about - and though hard drive manufacturers have definitely upped their game by building in "self-healing" functionality (auto-copy of data from bad sectors to spare sectors), Spinrite apparently not only uses that technology but also does something that the hard drive manufacturers do not.
As I understand it, Spinrite tries to force the disk surface in the bad sectors to get back to full magnetic strength by persistently reading and writing and may even do a reverse polarity write to the drive's surface. It also persists in reading bad sector data by having the read head come in from multiple different directions to perform the read, until it gets a parity check (a good read).
Whether that forcing makes for a permanently reusable disk surface in that formerly "bad" sector is unknown, but it would presumably have to help.
Of course, if it is the drive armatures/mechanisms that are failing, then, as @f0dder seems to suggest, all that extra reading/writing could "overstress" the drive and make it fail real quick, so taking an image of the drive could be good risk-avoidance.
In my case, because of the SMART errors reported by HDS, I knew it was bad sectors, and the drive's performance was still 100%.

[Stoic Joker]

Uh-oh...here comes f0dder!

-40hz (November 26, 2013, 05:32 PM)

Where did he get off to? I haven't seen him in forever.

[SeraphimLabs]

I have used it before on an experimental basis to see what it could do when turned loose on an aging drive, but it was a benchtop drive and not one of my good ones. Didn't really do enough with it to tell you about how it works, other than that it is a very time consuming process.

My last round of hard drive data recovery was done using dd-rescue, which returns a drive image that can be then re-cast back onto a new drive.

[IainB]

^^ Thanks. I got the impression that it could/would be an extremely time-consuming process, since, in bad drive sectors, Spinrite apparently goes down to the level of individual bits of data on the disk, and then resorts to an almost analogue-type approach where the bit has an indistinct magnetic polarity.

[Shades]

If I really need to know if there is a platter problem with a drive, I have found a free tool that tells me the exact location of an error, which type of error and where future problems will arise. It will also point out to you which sections of the drive are faster than others as well.

The tool is called: MHDD  you have to (re-)boot your computer with the included iso, which enables (very) low level access to the drive.  The DOS-like interface will be a turn-off for people that are accustomed to a point-and-click interface, even for keyboard jockeys the interface can be daunting.

It has more tricks up it's sleeve, but I use it mainly to check for (possible) hard disk errors. While this software runs, there is still a rudimentary help system available. But beware, this software comes with a lot of power, so be ready for the responsibility it demands. If in doubt, don't use it as you can do a lot of really irreparable damage.

Depending on the size of the disk a hard disk check can take quite some time, mainly because of the missing speedup functionality of the BIOS and file system being used by the operating system on the hard disk that is being checked.

No link is provided on purpose, only a little bit of <insert preferred search engine here>-fu could get you into a lot of trouble, but that will not be on my head at least.

[IainB]

^^ Thanks @Shades. Never heard of this before. I have downloaded the 2 current versions of MHDD 4.5 and 4.6 and the documentation website page, for reference.
Looks very interesting. Similar to, but probably not quite the same as Spinrite though, methinks.

Yes, MHDD could be drive obliteration time if you're not careful.

[TaoPhoenix]

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.

(see attachment in previous post)

http://www.extremete...es-actually-live-for

from slashdot.org

-mouser (November 12, 2013, 03:47 PM)

Heh my tech skillz are feeble. The best I have done so far is use that Bvckup2 program from that other thread. So my "soft data" is prob in passable shape, except maybe the last few months, but it was an easy prog to run. (Though I should test the "apps" copied over to be sure they are legit files!)

I think on a bigger scale, not counting building a list of progs to install, is thinking about Win 7. Last I knew scuttle was that was due to be "safely supported" for a while ... though with MS's prev rumblings about "not supporting more than three OS's back", and with the advent of 8.1, does that mean they'll drop support as of Win 9? (Will there be a Win 9? Or will  they call it Windows Secure or some other meaningless name?)

I we tried to look pretty far ahead when we custom built my current box with an eye at the future in 2006. First QuadCore Kentsfield series, 2 Terabyte drives, including the fresh D drive I can re-label as C. (Re that article ... if a Hard Drive sits in a "forest" and does absolutely nothing, does it still die?)

So I *think* we looked far enough ahead that my comp can run Win 7, maybe with an extra Ram chip if I feel like splurging. But all that is so far away, and a multi ...week... project I'm not up for yet!

 

P.s. If I have a Fresh OS install maybe Skwire won't hate me anymore!

 

[SeraphimLabs]

Kentsfield Quadcore would put it sometime between 2006 and 2009. Overall a machine that age that was well built and kept clean should survive another 2+ years without too much of a change in the failure rate.

It probably would handle 7 as long as it has a generous amount of RAM, and any performance changes would probably be only slight. The Yorkfield machines I built in 2010 were all outfitted with 4GB of ram and Windows 7 64-bit when they were assembled.

(Re that article ... if a Hard Drive sits in a "forest" and does absolutely nothing, does it still die?)
It still can. If the drive was powered and controlled, Windows will periodically spin it up and give it instructions just to make sure it is still ready to use. Even without data being moved, the drive's mechanical parts are still accumulating some wear and the electronics are still under the effects of aging.

Leaving a drive sitting completely unplugged inside a casing, the only aging phenomena it would see are electronics aging and thermal cycling effects- mechanically it would remain a new drive, and as long as the controller did not fail it would still be a like-new drive when you finally did decide to plug it in and use it.