How Many Hours Is Too Many For a Disk Drive?

Topic

New Reply

Since disk drives are moving parts, they have a finite lifespan before they will go “poof” leaving you and your data wondering what happened.
We all know the importance of a rigorous back up plan, but how many hours can accrue on a drive collect before it should be considered suspect?
There is no absolute answer of course. I wonder what the drive manufactuers claim for MTBF (mean time between failures).
I have seen several drives with 35,000 hours or better, that seem to run just fine, and pass crystaldiskinfo easily.
For a computer that runs continuously that amounts to only about 4 years of use, which is probably easier on a drive than stop-start every day.
And no doubt de-fragging every week or so is a serious detriment to a drive’s longevity.

Any opinions out there?
rstew

Reply | Quote

Replies

Good points to ponder! Let’s see if others weigh in with their observations

"Take care of thy backups and thy restores shall take care of thee." Ben Franklin, revisted

Reply | Quote

Difficult to give any definite answer to this, all component failures will coincide somewhere along a ‘bathtub’ profile – think ‘U’ shaped, left = new> right = end of life.

Most failures occur within the first days/weeks of use (left side slope > declining failures with time) then leveling out, very low failure rates after say, 2 months until approaching the end of life period (bottom of the bathtub, U shape curve), finally the curve/failures rise again quite steeply.

The ‘normal’ low failure period might be anywhere from 9 months to several (8+) years, dependent upon many factors, cost and intended use being just two of them.

Reply | Quote

I wonder what the drive manufactuers claim for MTBF (mean time between failures).rstew

rstew.

You can usually find the MTBF on the Manufacturer’s website.

Remember what MEAN is:

The mean is the average of all numbers and is sometimes called the arithmetic mean. To calculate mean, add together all of the numbers in a set and then divide the sum by the total count of numbers. For example, in a data center rack, five servers consume 100 watts, 98 watts, 105 watts, 90 watts and 102 watts of power, respectively. The mean power use of that rack is calculated as (100 + 98 + 105 + 90 + 102 W)/5 servers = a calculated mean of 99 W per server. Intelligent power distribution units report the mean power utilization of the rack to systems management software.

Thus, one drive may die after 1 hour while another may last for 10 years of continuous use. The problem is which one did you get? Thus, IMHO a rigorous backup plan and run them till they drop, or at least start showing problems. HTH :cheers:

May the Forces of good computing be with you!

RG

PowerShell & VBA Rule!
Computer Specs

Reply | Quote

In years past, 1980’s to early 1990’s I was updating my HDDs almost every year but because of technology advances in speed and size. Since the mid-late 1990’s I have been adding new drives about every 3 years and retiring the old ones, mostly again for the latest technology. I have only ever had one out of dozens fail and it was in a laptop that saw hard industrial strength use.

For an office environment I would recommend changing every 3 years and for home use or not as important office computers every 5 years. YMMV

Most people just backup once a week or never and “hope”.

Reply | Quote

I have worked with servers that have been running for 7 years and still had the original drives. I’ve also seen drives die after a week. The only time I’ve seen mechanical failure was old desktop drives – you hit them on the side with the blunt end of a screwdriver to start them, then backup.

cheers, Paul

Reply | Quote

Those of us that have worked on IBM Deathstars or a 10/20/30GB series by Fujitsu around the turn of the century are well aware of that

That data is from Backblaze, it’s one of 4 or 5 articles/blogs they’ve published on the topic. Google published one a few years ago, too. These all relate to server workloads, iirc.

For a cross-section (server+home/business/gaming usage) there’s been a periodic compilation based on hardware returns from a French -based etailer over the last 6 years, it’s short term based as in returns under warranty, so up to 12 months in the main sections and 6 months or so in the slightly later data in the conclusion section. Unfortunately, only the early reports were translated into English, though once you’ve studied some of the English versions (take a look at the OCZ SSDs, for example), working with the data from the later French only reports is quite straightforward.

* unfortunately, the English link now redirects to the Fr site, the .Be server (behardware.com) cache might be available on web archive sites, or use your favourite translation method.
http://www.hardware.fr/articles/927-1/taux-retour-composants-11.html

Reply | Quote

I think that if anyone could come up with a definite answer to the OP’s question, they could become rich. I’m sure many companies would pay for software or hardware that could accurately predict the life of a drive.

Reply | Quote

“How Many Hours Is Too Many For a Disk Drive?”

The hour you ran it and it failed.

Reply | Quote

And no doubt de-fragging every week or so is a serious detriment to a drive’s longevity.

Defragging a mechanical drive is never a detriment to longevity. Your drives work harder if you NEVER defrag them.
Not only the above, but filling them to capacity is also a detriment.

For a computer that runs continuously that amounts to only about 4 years of use, which is probably easier on a drive than stop-start every day.

A computer may be up and running 24/7, but that does not mean a drive is continuously running. Mech drives are designed and built to “start/stop” frequently.

What is a WELL KNOWN & PROVEN detriment to a mechanical drive:

1. An adverse and prolonged thermal environment.
2. An adverse and prolonged vibration environment.
3. Power failures, electrical variations (voltage and current variations; poor quality power), and electrical surges.
4. Factory defects. There is no such thing as a technology derived by man that isn’t infallible.
5.

Reply | Quote

5.

Clint
?? The unexplained?

Just because you don't know where you are going doesn't mean any road will get you there.

Reply | Quote

Clint
?? The unexplained?

That’ll be your own reflection in the mirror, CLiNT’s a vampire and couldn’t see it …

Reply | Quote

That’ll be your own reflection in the mirror, CLiNT’s a vampire and couldn’t see it …

Whatever I might have missed. I’m sure there’s something.

But basically, a warranty is what you want with any drive, unless you’ve got data too sensitive
to send back, then you’d have to put the drive under a hammer and eat the cost.

Reply | Quote

my win98se disks are still working

it was the power supply tht died

Reply | Quote

I built a new computer for Win7 when it came out and put in a 250GB C: drive and a 500GB D: drive, haven’t had a problem with it being on 24/7 since. I do clean the computer frequently and keep it on a UPS.

Before you wonder "Am I doing things right," ask "Am I doing the right things?"

Reply | Quote

I’ve an XP machine which has been up continuously (apart from maintenance, holidays etc.) since 2003. The disk only died in 2014. It’s now good as new with a drive I took from a video recorder which had never been used.

Somewhat OT, but that reminds me; the machine above is on its 3rd video card. The way it’s orientated in the case means the fan’s on the bottom rather than the top, and only glued on. I’ve had two fall off, frying the card. The last one took the motherboard with it. The present card doesn’t have a fan, and is more powerful. So it’s got its 2nd MB, 3rd video card, and 2nd HD, all with Windows OEM which still activates nicely!

Reply | Quote

So I went to the Seagate site to see what they say about expected MTBF for their drives.
Amazingly they say its about 1.2 MILLION hours!! That is about 136 years.
From their site:
“AFR and MTBF specifications are based on the following assumptions for business critical storage system environments:
• 8,760 power-on-hours per year.
• 250 average motor start/stop cycles per year.
• Operations at nominal voltages.
• Systems will provide adequate cooling to ensure the case temperatures do not exceed 40°C. Temperatures outside the specifications in Section 2.9 will increase the product AFR and decrease MTBF”

Wonder why so many drives die before their time?
rste

Reply | Quote

Wonder why so many drives die before their time?

RStew,

The number that is not there is the Millions of Drives they manufacture to get those Averages and Means numbers! :cheers:

May the Forces of good computing be with you!

RG

PowerShell & VBA Rule!
Computer Specs

Reply | Quote

The drives I use mention 100,000 hours MTBF/Mean Time Between Failures. With 8760 hours in a year that would still be longer than most will keep a computer, a bit over 11 years.

Before you wonder "Am I doing things right," ask "Am I doing the right things?"

Reply | Quote

The drives I use mention 100,000 hours MTBF/Mean Time Between Failures. With 8760 hours in a year that would still be longer than most will keep a computer, a bit over 11 years.

I find 100,000 hours MTBF way more believable for a drive. I think 1.2 million is utter nonsense.
I also think the 40 C case temperature limit is not well known or understood.

rstew

Reply | Quote

“What me worry? Igor made my Backup!”
“Wazzit noise, that ticking sound? Hooo, that’s bad…uh oh, HD finally quit.”
“Hey, Igor, where’s that backup?”
“…it’s on the 2nd partition of the HD that had the ticking noise, and died…”
“oh poop”

In reality, I had one HD fail, after it took a physcial hit alongside the chassis. Usually, I replaced the computer for upgrading before HDs bit the dust.

"Take care of thy backups and thy restores shall take care of thee." Ben Franklin, revisted

Reply | Quote

one more than the time it takes to back it all up before it fails

there is no absolute time when a HD will stop
so the real answer is 1
back it up daily

defragging may result in less use searching for scattered bits of info to make up for that effort

Since disk drives are moving parts, they have a finite lifespan before they will go “poof” leaving you and your data wondering what happened.
We all know the importance of a rigorous back up plan, but how many hours can accrue on a drive collect before it should be considered suspect?
There is no absolute answer of course. I wonder what the drive manufactuers claim for MTBF (mean time between failures).
I have seen several drives with 35,000 hours or better, that seem to run just fine, and pass crystaldiskinfo easily.
For a computer that runs continuously that amounts to only about 4 years of use, which is probably easier on a drive than stop-start every day.
And no doubt de-fragging every week or so is a serious detriment to a drive’s longevity.

Any opinions out there?
rstew

Reply | Quote

hello everyone;
i believe i could make a contribution to the subject discussed here.
years ago, i used to work for a power supply company where we designed tailor units according to customer specifications, units [PSU] that were to fit inside their systems.
depending on customer’s the spec., we could either PREDICT the MTBF according to calculations or TEST it.
i am adding an attachment which somehow gives an overview of what the term means.
if the system where the unit was to be lodged could tolerate a certain level of failures, for the customer was enough to have a figure to apply to their bank of data, therefore they were asking for PREDICTION, based on calculations.
few decades ago, MIL 217 was the most used method, but as it was fairly “optimistic”, newer calculation methods were used, being Bellcore the most popular lately.
if the PSU was to be lodged in an expensive system (more on that below), then HALT TEST was used (see attachment).
in a VERY simplified way, for the HALT TEST we took 200 units (sometimes 300 or 500) and run them at the most adverse temperature limit according to customer specification for 1000 hours (that is approximately 42 days), non stop.
this was, in fact representative of an equivalent of 200000 unit-hours of test for a single unit.
so 200 thousand hours divided by the number of failures occurring during the test was the MEASURED (MORE REALISTIC) MTBF figure. (more units used if it was foreseen that faillures woukd take longer to occur).
now, let emphasize that these were requirements for a MTBF figure for INDUSTRIAL units.
commercial units (namely the hdd you or i buy at amazon, newegg, ebay or bestbuy, are NOT tested that way, as it easy to understand that the HALT test IS EXPENSIVE (a temperature stabilized chamber + its real estate + cost of 200 units + the cost of UNINTERRUPTIBLE energy for 42 days) (no blackouts tolerated)
what we get “over the counter”, is NOT a unit which MTBF is so representative of is reality.
picture, however a different scenario:
same hdd goes into a server computer controlling an expensive system where reliability IS a concern, where lost of data could impact forlorn profit [lotsa $$$$$$]
for THAT hdd, the server manufactured will require from the hdd manufacturer a HALT tested MTBF figure; and then the server manufacturer will (more likely) use that figure for his CALCULATED prediction of his overall MTBF.
summarizing:
it it the application that defines to what extent the data we’re being given is credible and WHAT it REALLY represents.

Hope it helps.

daniel rozenberg

on a personal note, what i usually do, is to put my machines on standby (it stops the hdd) whenever not using them.

Reply | Quote