Why nowadays hard disk is not that reliable and durable?

[danielyamseng]

Why nowadays the hard disk is not reliable and durable? I had lots of hard disk and most of it start to have problem within 2 years on average( i.e some 2 yrs, 3 yrs and 5 yrs).

This showed that it has a high defect rate.

How 'bout you guys?

[Wilt]

Do you like to play catch with your drives, or abuse your laptops? In my history going back 25+ years in personal computing, I have only had a single failure...of a USED drive! My daughter had one drive fail in her laptop a couple years ago, too.

(knock on wood)

[FLiPMaRC]

In my 11 years hear at work, I've only seen a handfull of drive failures in hundreds of computers. It's not very common, unless the temperatures are constantly above tolerances, or is constantly being moved as Wilt says :lol:

[ed rader]

Why nowadays the hard disk is not reliable and durable? I had lots of hard disk and most of it start to have problem within 2 years on average( i.e some 2 yrs, 3 yrs and 5 yrs).

This showed that it has a high defect rate.

How 'bout you guys?


i've never had a problem with any drive .... knock wood :D.

ed rader

[krb]

One other issue is data density. This is over simplifying, but if a drive has ten times the capacity then a defect that would have lost one bit of data before will now lose ten bits of data. Basically, failures become more obvious.

[ToyTrains]

Yes drives do fail and it seems the newer larger ones fail more frequently.

Don't store your photo's on just a single drive. You can't have too many backups and the prices of drives are really low (and keep getting lower).

[ToyTrains]

Do you like to play catch with your drives, or abuse your laptops?

Drives these days can even withstand major trauma. After the Challenger shuttle crash, they were able to retrieve most of the data from the hard drives even after the fiery crash to earth. Pretty amazing.

[rklepper]

But then they do not use hard drives that are purchased off of newegg.

:)

Drives these days can even withstand major trauma. After the Challenger shuttle crash, they were able to retrieve most of the data from the hard drives even after the fiery crash to earth. Pretty amazing.

[ToyTrains]

But then they do not use hard drives that are purchased off of newegg.:)
Now that is the truth :(

[Mark1]

I found a statistic on hard drives this spring....cant find it again. But it is like 3% are DOA right out of the box. Then 10% die off every year.

[wickerprints]

With increasing data density comes much higher probabilities of data error, not because there is more data, but because it takes much less to do damage. It's amazing that these things work at all.

Hard drive heads are tiny...orders of magnitude smaller than they were two decades ago. They are therefore that much more sensitive to vibration, contamination, and wear. The platter has to be that much more perfect--the slightest defect in the magnetic layer means data cannot be written to that area.

To compensate, modern drives have more sophisticated controllers that automatically map out bad areas of the drive. In fact, the true "defect rate" is masked by the fact that many drives transparently and automatically hide these data errors from the end user. Only until the drive suffers a more catastrophic error (e.g., head crash), would a problem become evident.

The reality is that hard drive technology has maybe another 5-10 years of useful life. We are moving away from electromechanical data storage, and toward solid-state. We're pushing up against the physical barriers of areal density using magnetic effects. Non-volatile memory is the future, as is more exotic technologies not yet ready for market. When the economic costs of data reliability, density, permanence, and convenience for hard drives becomes higher than that for competing technologies (which is already happening in certain portable devices), then we can finally kiss these spinning platters goodbye.

[danielyamseng]

those failure one is the one I put it on the external Hard disk casing. I connect the laptop thru USB connector. Is this is the reason why I always has problem?

when I run a check the hard disk always has error sector this and that error.

Now the my worst nightmare was the 1 Tera hard disk can't be recognize as a drive.
Any toold tips or software that could recover the date from it?

[Paul Li]

This is scary...Is there any solid method of storage, besides online hard drive?

[KayMomto3]

This is scary...Is there any solid method of storage, besides online hard drive?

That's why I backup on a couple of different media - hard drive, portable hard drive, DVD and online) not necessarily all at once but you get the idea). It's a pain - but have a hard drive fail once when you haven't backed up makes you a "little" paranoid.

[RDKirk]

Actually, I'm always a little amazed at how cavalierly people treat hard drives these days. Back when we first started putting them into microcomputers, they were so delicate that the slightest nudge of the computer would crash the heads. We wouldn't even shut a computer off without first running a small file to park the head securely.

I doubt they're any more fragile than they ever were--people just treat them more roughly, especially laptop drives.

[Hermes]

I have a 500MB hard drive that has twice been dropped onto the concrete pavement from several feet in the air and still works fine. In fact, all the hard drives I've ever had corrupt have been very old.

The newer hard drives have better software for working around bad sectors and have power management to increase their lifespan - they seem much more reliable now than they were 10-15 years ago.

[hollis_f]

I found a statistic on hard drives this spring....cant find it again. But it is like 3% are DOA right out of the box. Then 10% die off every year.

10% sounds an awful lot. If that were true then, with 7 hard drives at home, I stand a 50:50 chance of having one die each year.

At work I had 15 of them. The chances of at all of them surviving a year would be only 1 in 4. The chances of all 15 surviving three years is just 0.9% - and I ain't that lucky.

I'd be surprised if the annual failure rate was as high as 0.1% a year.

[SwitchBlade]

I'd say the OP is either highly unfortunate or has been buying extremely cheap non-brand crap. In the last 15 years I've only had two hard drives die, one was a full height 3.5" 40MB Maxtor (for those who remember those paperweights), the other a 3.5" single bay 120MB Western Digital. Most every other drive I've had has been replaced because it's too small to be useful long before it has a chance to fail. Modern hard drives (especially the smaller laptop/ipod style drives) are far more robust than they used to be and I've had a number of drives survive accidents that I'd not expected them to.

Highlight for me is the hard drive in the PC in my car. That puts up with the usual vibration of driving, bumps from poor road surfaces, g-forces from cornering etc, and it's been running fine with no problems for the last 3 years or so.

[gooble]

Why is this scary or even news? If you're not keeping your photographs on at least two different hard drives you're being negligent. I keep all my stuff on at least four drives at any given time and sometimes more. Expect failures and plan for them. Don't keep all you stuff on one drive and hope nothing goes wrong.

[Josh_30]

The only HDD's that I've ever had fail were in notebook PC's. Never had one fail in a desktop, not even the really really old ones I have. I blame the increased heat (lack of notebook cooling mostly) and general "abuse" of carrying the notebook PC around for the failed drives.

I have 3 hard drives in my PC. 1 is for my OS and programs, and general stuff. 2 of them (photo/video only drives) are mirrored so if one fails, I'm good. I also back those drives up on DVD's occasionally and keep the discs in a fireproof/waterproof safe.

[Lowner]

Never known a single failure.

Of course having said that, all mine will now fail together.

[Maxdave]

Unfortunately, unless we "meet" a lot of drives in our work/business life, few of us have had sufficient experience with hard drives to personally acquire meaningful statistics. I might have a bit more experience than the average home user, but nowhere near enough to be statistically significant.

For what it is worth, however ....

It is natural for us to remember an failure rather than the unnoticed continued and seemingly never ending operation of good drives.

I do a number of repairs (and build a number) of both desktop and notebook systems each year, for the last eight years or so. It "seems" to me, that very few (less than 1 in a 100) drives are DOA (in other words this is an exceptional occurrance).

We recommend drive brands to our clients that feature long standard warranties (on the admittedly unproved theory that if a company is willing to offer five years rather than one year, they must have reasons to do so). Of all the non DOA desktop drives we have originally installed or used in new machines, exactly two have failed within the warranty period, and none of the drives that we have used in notebooks have failed prior to warranty expiry. In use, again it "seems" to me that modern laptop drives are just as duarable as their desktop counterparts.

Based of that, I have (unscientifically) suggested to clients that when you reach the end of the warranty period, your level of "concern" should increase.

Personally, I have had one failure, after the warranty period expired (three year warranty in this case). It occurred without warning giving no indication of impending doom.

Maxdave

[ToyTrains]

Maxdave is correct that most of us do not have a sufficiently large sample to draw valid conclusions.

Also some assumptions we derive are not valid. For example Seagate who had a 5 year warranty has now changed to a 3 year warranty. What does that indicate?

Anyone who wants to get an objective look at hard drive failures should read a Google white paper on the subject (they don't provide results by manufacturer, only in aggregate):
http://labs.google.com/papers/disk_failures.pdf (http://labs.google.com/papers/disk_failures.pdf)

They have a sufficiently large sample of over 100,000 consumer grade hard drives from a mix of manufacturers used in their servers. Their drives are essentially in use 24x7.

The annual failure rate was 1.7% in the first year to over 8.6% for three year old drives.

Interestingly there was no correlation between high temperatures and failures, possibly even the opposite of very low temperatures and failures.

The occurence of certain SMART errors (which can map bad sectors to good sectors) did indicate possible future failures at a high rate but drives with no previous errors still had 1/3 to 1/2 of the failures.

The takeaway should be to make multiple backups of your photo's (just because you never had an error doesn't mean you won't in the near future).

[Lowner]

A question.

I only wind up my external hard drive when I'm using DPP, Photoshop or doing invoices (don't ask!). Am I saving the drive by not working it too hard, or is it possibly shortening it's life?

[SwitchBlade]

ToyTrains, that's a bit different though. A home user won't leave a drive on 24/7, which is better for the drive than the constant spin-up, spin-down, head parking of a home drive.

[SuzyView]

I have lost 3 internal HD, but have been lucky to get the data out. Had to install new drives, therefore, software, update all the stuff, etc. Very painful. So, I back up, but you never know. Maybe it is time to do an off-site safe.

[Tsmith]

So far I've had and use nothing but Western Digital Hard Drives and have yet to have a failure since 1996. I still have a 13 GB that works if it was to be hooked up. Currently have 2 Raptors and a 500/250 GB drives in my main system plus another 250 as external backup.

Its nothing but WD for me.

[ToyTrains]

ToyTrains, that's a bit different though. A home user won't leave a drive on 24/7, which is better for the drive than the constant spin-up, spin-down, head parking of a home drive.
I have not seen any studies about whether it is better to leave a drive on or to turn it on/off (and the frequency of turning them on/off). It probably depends on the frequency. If you turn your drive on once a year, it will probably last a long time (or maybe not). Once per month, once per week, once per day, I am not sure. Our intuition on this stuff is often wrong.

My computers are on 24/7 in a relatively good environment. I have experienced drive failures on both internal drives that are always on and external drives that spin down between use.

[Maxdave]

So far I've had and use nothing but Western Digital Hard Drives and have yet to have a failure since 1996. I still have a 13 GB that works if it was to be hooked up. Currently have 2 Raptors and a 500/250 GB drives in my main system plus another 250 as external backup.

Its nothing but WD for me.

We have had good luck with WD longer warranty drives also. It has been our brand of choice for some time (I know someone will contradict this opinion soon; that's what I mean about data being unreliable/invalid due to insufficient experience).

Maxdave

[nikocanion]

i have toshiba and lacie witch are more than 2 years old and still work fine maxtor is un-recommendable to expensive and high % of bad gear.

[Mystwalker]

Until about a year ago, my home was 100% Seagate drives.
Then I tried Seagate's 1TB and 1.5TB ...
Is this when they changed from 5yr to 3yr warranty?

Out of six drives, two were dead out of box, one started losing data requiring me to reformat (later learned I could've sent back to Seagate to recover), and another started making clunking noises on startup - all within first two weeks. Maybe after 15+ years of no HD problems, electronic gods decided it was just "my turn".

Newegg allowed me to return and get WD Caviar blacks. No problems - yet. Hope it's another 15+ years :)

I think biggest threat to anything electronic are dust and heat. You hope the hd is sealed so dust is not a problem. But with these drives getting bigger and bigger, I think heat will become more of an issue. The 1TB and 1.5TB we have today are no bigger then the 160/320/500s from 3-5 years ago - I can almost swear the 40 and 80s of 5+ years back were of similar size, but haven't seen touched one in ages. The bigger drives today are also cheaper then smaller drives in past. Maybe drive manufacturers are cutting corners to remain profitable? Maybe these drives are now made in 3rd world countries where QA isn't as stringent as they were here in past?

I always have 4 copies of all my pictures (HD, two external drives, one DVD) - hopefully that's enough. Do not want to be overparanoid.

[awdark]

those failure one is the one I put it on the external Hard disk casing. I connect the laptop thru USB connector. Is this is the reason why I always has problem?

when I run a check the hard disk always has error sector this and that error.

Now the my worst nightmare was the 1 Tera hard disk can't be recognize as a drive.
Any toold tips or software that could recover the date from it?

Are these externals you are using the small ones that plug with USB only or the type that use a external power adapter? The USB only ones sometimes have issue with power because they want to draw more than the USB port offers. Thats not to say the AC adapters are much better because the adapter can go bad too.

Also sometimes those drive enclosures go bad. I know a few people who had the USB>IDE/SATA board go bad and they thought the drive was a loss but I told them to take it apart and plug the drive directly into their computer and their data was fine. The brand doesn't seem to really matter, they just randomly go bad sometimes.

Recover My Files is not free and takes a while but it has saved my butt many times already!

Anyways, I think it has to do with the density of the disks and possibly the use of RoHS components, the lead free solder is more brittle and provide more opportunity for the controller boards to fail. And new technology forces companies to push out new drives before thoroughly testing them like in the case of the Seagate 1.5tb drives which were failing due to firmware issues.

[ToyTrains]

Until about a year ago, my home was 100% Seagate drives.
Then I tried Seagate's 1TB and 1.5TB ...
Is this when they changed from 5yr to 3yr warranty?

Out of six drives, two were dead out of box, one started losing data requiring me to reformat (later learned I could've sent back to Seagate to recover), and another started making clunking noises on startup - all within first two weeks. Maybe after 15+ years of no HD problems, electronic gods decided it was just "my turn".
+1

In the last 6 months probably 50% of the large drives from Seagate have failed, some within a month.

For me the warranty on a dead drive is useless. Once I put my data on the disk I won't chance sending it back for replacement. You never know where your data will wind up!

I have now switched to WD 2TB drives. We'll see how they do.

[SwitchBlade]

But with these drives getting bigger and bigger, I think heat will become more of an issue. The 1TB and 1.5TB we have today are no bigger then the 160/320/500s from 3-5 years ago - I can almost swear the 40 and 80s of 5+ years back were of similar size, but haven't seen touched one in ages. The bigger drives today are also cheaper then smaller drives in past. Maybe drive manufacturers are cutting corners to remain profitable? Maybe these drives are now made in 3rd world countries where QA isn't as stringent as they were here in past?

Not really, it's just that modern manufacturing methods allow for stricter tolerances which increase data density. I remember when the first multi GB drives came out I bought a 5.25" drive as it was about £30 cheaper than the equivalent 3.5" one.

As for drive size I've still got an 80MB drive the same size as my 1TB drive. (as are my old 120s, 140s, 250s, 850s, etc). But yes modern drives do get hotter, I'm guessing a large part to that is that the platters are spinning much faster so that kinetic energy is going to bleed off as heat.

Just always obey the golden rules, I know they are a bit boring these days but there was a reason we started doing it back then. Separate partitions for OS/Software/Data. Better still separate drives. If you run windows the OS will need to be reinstalled a number of times in the computer's life time, just a fact of life post Win95. Always backup data that you *need* to keep, i.e. little point backing up that Simpson's collection because you can re-rip it from DVD easily enough, but that essay you wrote, the novel you were working on, and in this case all those lovely photos should be backed up. Even if it's the same computer, a drive failure is a likely thing to happen, more-so than say lightning striking or your house catching fire or burglary (well dependant on where you live). Lose the drive you have a backup, highly unlikely to lose 2 drives at once unless your PSU throws a right wobbler and puts mains down the molex leads. If you are paranoid about theives/fire than an external drive hidden in the shed or a simple low power backup server is an easy option or dropping an external drive round a relative's place. I tend not to be too trusting with my intellectual property regardless of it's current commercial value.

All that said, you can be as paranoid as you want to be, and tbh despite my preference for Seagate drives due to the longest service I've had from drives, even disks such as the Hitatchi Deathstars aren't unreliable crap any more. Buy a good brand disk and chances are you'll get good use from it.

[danielyamseng]

No wonder I had a high failure rate. all of my HDD is Seagate. Which want is your recommended hard disk? Is it Western Digital?

Which Type I should get? Sata,sata2, etc etc?

There're lots of variation i.e the one with 10000rpm,pata. Not really sure which one has the highest reliability rate.

[pwm2]

I have maybe 100 drives. Haven't seen any DOA. I have lost three disks in 22 years - I would think that I have on average had 15 drives running 24/7 accumulating about 300 years active time.

One was a high-end IBM SCSI drive that went instant dead. One was a brand-new 3.5" disk that slipped off a table when I used it with a USB-to-IDE cable and snagged the cable. The third one was a recent TB-size Seagate that died within 2 months.

No, new drives do not get hotter because of their increased storage capacity. In reality, they get cooler - new drives normally consumes less power because of improvements to the electronics and the reduced weight of the smaller-and-smaller head assemblies.

But each time the drives steps up the rotation speed, there is a bump in power consumption. A 7200rpm 2.5" drive draws more power than a 5400 or 4200rpm drive.

External drives mainly suffers from three problems:
- physical abuse
- unreliable power supplies
- bad cooling

A number of manufactuers of external disks have had big problems with the disks getting overheated. This is greatly affected by how the customer uses a disk. The power consumption increases a lot when the disk is seeking or writing.

The next thing is that many external disks are in the 3.5" format, since 3,5" disks gives more GB/$. But the 3.5" disks are generally more sensitive to shocks than 2.5" disks. Both because of their larger size/bigger mass, and because they are designed for stationary use, while 2.5" and 1.8" disks are mainly designed for mobile use.

A number of years ago, all good manufacturers had long warranty periods. But that was before the price war on disks. Todays disks (almost any kind of disk) are basically for free. Once upon a time, a server-class disk could cost $2000+ or more. Today, the manufactuers can hardly handle a single support call without consuming the little margin there was on the disk, so they cut down on the warranty times to force customers to buy new disks instead of contacting them. The reason isn't that they don't believe in their disks. But a warranty return costs about the same now as it did 10 years ago. But with the sell price of the disk being so much lower, it takes more sold disks now to cover the cost of one single warranty return.

About using disks for backup. Don't store the disks off-line for too long time. When the disk is in use, it will constantly scan the surface, and rewrite sectors that are weak or remap sectors that can't be rewritten without error correction (ECC) needed. When storing the disk for extended times, there will be no scanning of the surface, so weak sectors will not be rewritten. And as with other magnetic media - store at a low temperature.

Next thing: A disk is about the worst possible media for backup. A single drop can instantly change the disk from 100% ok and into a wreck that will require that you pay a data recovery firm a lot of money for restore of the information.

Finally: Redundant disk copies (offline or in RAID configuration) can very quickly die in a fire. Always store one or more copies of the photos and other important documents in a different location.

[pwm2]

No wonder I had a high failure rate. all of my HDD is Seagate. Which want is your recommended hard disk? Is it Western Digital?

Which Type I should get? Sata,sata2, etc etc?

There're lots of variation i.e the one with 10000rpm,pata. Not really sure which one has the highest reliability rate.
Seagate disks have traditionally been very good and reliable. But they have recently had a number of issues. They have had disks with firmware errors where the disk has "bricked" itself, incorrectly trying to protect itself from damage. And they have had a couple of external disk enclosures that have either suffered from heat or from too weak power supplies.

The bad thing for you as a customer is that history does not predict the future. IBM/Hitachi has had a very long range of great disks. Then they managed a bad firmware in some models of the TravelStar drives, giving rise to the nicname DeathStar. After correcting the bug, their drives are once more great with good reliability.

When you do buy a disk, you will not know if your specific drive will be good or not. And you will not know if this specific model and hardware revision will be good or not. It will not help what you find by googling for older models from the same manufacturer.

But one thing that is worth noting regarding Seagate, is that they new about the firmware errors in their drives and decided to keep quiet about it for a long time, until their board of directors had decided on what policy to follow and officially putting up a fixed firmware on their web. A number of customers did lose valuable data during this time...

[gjman]

Only one IBM Deskstar (a.k.a DeathStar) has failed on me in over a decade. Off course there was a class action lawsuit and then they sold the Deskstar brand to Hitachi....

Never had a Seagate or WD failure at home.

At work we run 10K RPM Fujitsu SCSI drives all day long...sometimes it takes 72 hours for it run a complete job but they have works flawless for the past 5 years. We are running Serially Attached SCSI (SAS) drives for under a year now in similar high i/o environments and they seem to be working just fine.

[pwm2]

Only one IBM Deskstar (a.k.a DeathStar) has failed on me in over a decade. Off course there was a class action lawsuit and then they sold the Deskstar brand to Hitachi....

Never had a Seagate or WD failure at home.

At work we run 10K RPM Fujitsu SCSI drives all day long...sometimes it takes 72 hours for it run a complete job but they have works flawless for the past 5 years. We are running Serially Attached SCSI (SAS) drives for under a year now in similar high i/o environments and they seem to be working just fine.
Note that the DeskStar is the product name of all consumer-class 3.5" disks from IBM/Hitachi. Within the DeskStar product group, there are regularly released new models based on new designs.

It was the 75GXP series disks that had the firmware problem giving rise to the "DeathStar" moniker. An idle disk didn't regularly move the head to different locations. People who had the affected models and always kept them busy never saw a problem. People who kept the drives spinning without doing any work do often got the problem. Constantly rotating with the heads in the same location made the lubricant on the surface become unevenly distributed which could result in both problems with surface wear and that the head finally splashed into the lubricant.

Anyway, that is past history. It is only a couple of years from now we will know how good the disks are that are currently on the shelves. Any HDD manufactuer may take a wrong step, and it is almost impossible to test everything before releasing a new model or a new firmware. All we can look at right now is value for money, warranty period, availability and warranty/policy track record from previous incidents.

[tim]

I use Seagate disks exclusively, I have for ten years, and I haven't had a failure yet. My data storage needs keep expanding, so a disk only tends to last 2-4 years for me. Having said that I use old disks as swap and such, and they all still work fine.

[Village_Idiot]

I'd say I see about a 50% failure rate...on the cheap Maxtor drives that the government purchased for us since they were the cheapest option...

[Snow001]

A question.

I only wind up my external hard drive when I'm using DPP, Photoshop or doing invoices (don't ask!). Am I saving the drive by not working it too hard, or is it possibly shortening it's life?

I'm in this boat with you. I do not leave my external drives on all the time and only power up when I need to use them. I keep them disconnected from the pc and unplugged from the AC outlet.

I noticed that my older external drives have a power off switch (Maxtor one touch). The new drives from Western digital no longer have this option.

It would be interesting to know if we are really increasing or decreasing the life of the drive by turning it off literally.

[Snow001]

I'm curious if anyone here has any experience in using Blu Ray as a back up method.

Excerpt from the Blu ray FAQ:

" How much data can you fit on a Blu-ray disc?
A single-layer disc can hold 25GB.
A dual-layer disc can hold 50GB.

To ensure that the Blu-ray Disc format is easily extendable (future-proof) it also includes support for multi-layer discs, which should allow the storage capacity to be increased to 100GB-200GB (25GB per layer) in the future simply by adding more layers to the discs.

The internal Blu-ray burners are now around $200 and the single layer media is around $45 per 15 pack spindle.
This is of course assuming that they don't introduce a new format in the next 5 years :)

[pwm2]

This is of course assuming that they don't introduce a new format in the next 5 years :)
Blu-ray works quite ok for backup. I normally do buffer backups to DVD-RAM until I have enough information to fill a blu-ray disk.

Newer formats are likely. Support for more than two layers will require that you buy a new drive.

But remember that CD backups are still readable. So is DVD backups. It is quite likely that any new optical standard will try to keep the compatibility with our previous formats - it costs a lot to introduce a new format and we end users don't like our existing media to be unreadable unless the new format is so extremely good that we can accept paying for two drives.

[Wilt]

It would be interesting to know if we are really increasing or decreasing the life of the drive by turning it off literally.

MTBF statistics inherently include power-on time for the mean time failure statistic, so if power is turned off you are suspending time essentially.

[pwm2]

MTBF statistics inherently include power-on time for the mean time failure statistic, so if power is turned off you are suspending time essentially.
But the ageing figures of S.M.A.R.T. supervision also takes into account the number of full power cycles. The $100,000 question is at what power cycle interval these two ageing reasons balances each other.

[Lowner]

"But the ageing figures of S.M.A.R.T. supervision"

Can someone explain the gargon to me please?

[pwm2]

S.M.A.R.T. is a system that is built into most hard disks since a number of years. It measures the number of hours the disk has been running. Number of cold starts. Number of spin ups. Number of seek errors etc. The intention is to have a standard for requesting status information about the disk. Based on the statistics, the S.M.A.R.T. system will also try to compute a heal figure for the disk.

One problem, as shown by the Google report, is that the S.M.A.R.T. system fails to predict a rather large percent of disk failures. But a disk that has started to detect broken sectors have a way higher probability of failing within the following months.

[Wilt]

But the ageing figures of S.M.A.R.T. supervision also takes into account the number of full power cycles. The $100,000 question is at what power cycle interval these two ageing reasons balances each other.

If I only run an external harddrive once a week or once a month, that is not a significant addition to start cycles, compared to a PC which is started up from power-off once a day.

[pwm2]

Correct. But the $100,000 question is still where the break-even is - when it is better to keep it running, or when it is better to turn it off. Is it once-day? Or once-hour?

Another thing - make sure that you regularly keep it on long enough that it has time to do a big self-test. If the drive doesn't get a chance to scan all parts of the surface, it will not be able to find sectors that should be refreshed.

[Wilt]

Correct. But the $100,000 question is still where the break-even is - when it is better to keep it running, or when it is better to turn it off. Is it once-day? Or once-hour?

Another thing - make sure that you regularly keep it on long enough that it has time to do a big self-test. If the drive doesn't get a chance to scan all parts of the surface, it will not be able to find sectors that should be refreshed.

I perhaps oversimplify, but the three alternatives (in the extreme) seem to be

Power up once, leave it running 24/7, and hope to beat the MTBF statistics (e.g. 20000 hours = 2.3 years)
Power up once per day, leaving it running 10 hours/day and hope to beat the MTBF statistics (e.g. 20000 hours = < 5.5 years due to power cycling effects)
Power up once per week, run it for 2 hours and hope to beat the MTBF statistics (e.g. 20000 hours = < 38.5 years due to power cycling)I'd pick #3 for a data backup drive! :)

[ToyTrains]

I perhaps oversimplify, but the three alternatives (in the extreme) seem to be

Power up once, leave it running 24/7, and hope to beat the MTBF statistics (e.g. 20000 hours = 2.3 years)
Power up once per day, leaving it running 10 hours/day and hope to beat the MTBF statistics (e.g. 20000 hours = < 5.5 years due to power cycling effects)
Power up once per week, run it for 2 hours and hope to beat the MTBF statistics (e.g. 20000 hours = < 38.5 years due to power cycling)I'd pick #3 for a data backup drive! :)
I am guessing that we just can't know the answer to this question. We probably are not going to see a long term study as there just is not enough available data.

Ever notice how a light bulb usually burns out when you turn it on? Also similar to a car (well it doesn't exactly burn out-at least most of the time).

[Wilt]

I am guessing that we just can't know the answer to this question. We probably are not going to see a long term study as there just is not enough available data.

Ever notice how a light bulb usually burns out when you turn it on? Also similar to a car (well it doesn't exactly burn out-at least most of the time).

Yeah I have noticed. And the original Edison bulb is still lit in some fire department, IIRC. But disk drives are so mechanical in nature...bearings wear out just from rotating incessantly. Bulb filaments are stressed from the electricity surge. So are disk drives, but one power up a week vs. on continuously for a month...as you say, there only is endless speculation about which is more 'wearing'.

[pwm2]

I am guessing that we just can't know the answer to this question. We probably are not going to see a long term study as there just is not enough available data.

Ever notice how a light bulb usually burns out when you turn it on? Also similar to a car (well it doesn't exactly burn out-at least most of the time).
My disks will normally run 3-5 years 24/7 before getting replaced. The last 10 years, I have probably had an average of 20 disks running 24/7, with no failure with the exception of a Seagate that died almost directly I got it, and a drive I dropped to the floor.

I'm not so convinced that the expected MTBF is just 2.3 years for 24x7 use. And materials will age even when a device isn't used, so keeping a HDD off doesn't magically stop the ageing.

The only drive that did die a "normal" death did die when I tried to turn on a computer after maintainance, after maybe 2-3 years 24/7 use. I think the status codes indicated a problem with the electronics.

Light bulbs are a bit special. When cold, the current inrush is extremely high because the cold filament has very little resistance. The cold resistance is more than 10 times less than the hot resistance which means a 100W lamp can draw maybe 1.5kW when you turn it on.

But all electronics have a similar property. The electronics gets their supply voltage stabilized by capacitors. And these capacitors are basically small rechargeable batteries with a very fast recharge time. So whenever our electronics is turned on, there will be a very big inrush current while all capacitors recharges. With age, this may kill the electronics, or the power supply feeding the power. If the electronics dies, it may kill the power supply. And if the power supply dies, it may kill the electronics.

Similar thing for mechanics. The load on the motor in the HDD is very high during spin-up time. And while spinning up/down, the platters will not rotate fast enough to make sure that the head can't touch the surface.

Ultimately, quite a high percentage of everything electrical we have tends to die during the startup moment, or in combination with power spikes on the supply voltage or on any inputs.

[pwm2]

Yeah I have noticed. And the original Edison bulb is still lit in some fire department, IIRC. But disk drives are so mechanical in nature...bearings wear out just from rotating incessantly. Bulb filaments are stressed from the electricity surge. So are disk drives, but one power up a week vs. on continuously for a month...as you say, there only is endless speculation about which is more 'wearing'.
Another factor here is how hot we run our drives. The Google study didn't manage to find any correlation between heat and early failures, but they seemed to have their drives run in good environments. Only a very small number of drives in the 45..50°C range where the failures increased.

Some 3.5" drives in external enclosures can become way hotter than that if run with a high load, such as running BitTorrent transfers. If the Google paper had included drives running at up to 60°C (most HDD are specified for use up to 55°C-60°C), I expect the failure figures to have increased very significantly. The problem with the USB enclosures is that the USB bridge chip removes our access to the S.M.A.R.T. supervision, so we can't check how hot the drive is running.

[ToyTrains]

And the original Edison bulb is still lit in some fire department, IIRC.

Fun stuff-Longest Burning Light Bulbs:
http://www.roadsideamerica.com/set/lightbulbs.html (http://www.roadsideamerica.com/set/lightbulbs.html)

I learned something new today :D