Failure curve of hard drives

  • Thread starter Thread starter Mark M
  • Start date Start date
M

Mark M

I would guess that the failure curve of a hard drive is much like
that for a lot of electronic equipment. It is probably a sort of
wonky "U" shape.

If it fails then the drive is likely to fail quite early on in its
life. After it passes this period then it is more likely to fail
after some years as it gets older.

Does anybody know approximately how long this period is of relatively
high risk of early failure? This must vary from one model of hard
drive to another but, very approximately, after what period of time
would it be?

I may be wrong but I would guess there is a bit of a "knee" in the
shape of the reliability curve where the risk of failure changes.

If I get a new ATA/IDE hard drive I don't like to use it for live
data until it has been running on test for a while. But how long is
sensible for this test period?

At a usage pattern of 10 hours a day for the hard drive, would this
period be three or four days? A week? Three weeks?
 
Mark M said:
I would guess that the failure curve of a hard drive is much like
that for a lot of electronic equipment. It is probably a sort of
wonky "U" shape.
Click to expand...

It's called the "bathtub shape"

Lurch
 
I would guess that the failure curve of a hard drive is much like
that for a lot of electronic equipment. It is probably a sort of
wonky "U" shape.

If it fails then the drive is likely to fail quite early on in its
life. After it passes this period then it is more likely to fail
after some years as it gets older.

Does anybody know approximately how long this period is of relatively
high risk of early failure? This must vary from one model of hard
drive to another but, very approximately, after what period of time
would it be?

I may be wrong but I would guess there is a bit of a "knee" in the
shape of the reliability curve where the risk of failure changes.

If I get a new ATA/IDE hard drive I don't like to use it for live
data until it has been running on test for a while. But how long is
sensible for this test period?

At a usage pattern of 10 hours a day for the hard drive, would this
period be three or four days? A week? Three weeks?
Click to expand...

Mostly what I hear about is situations where the drive is bad right at
the start, or works fine for at least a few months. Unles you hear
strange noises coming from the drive right away it's unlikely you'd
benefit from an extended burn-in of a few days to weeks.

There isn't really much you can do beyond keeping it cool and using a
decent power supply, but if you're that concerned about it then run a
Raid mirror. If wouldn't hurt to check the drive with the
manufacturer's diagnostics right after installing it, and using a disk
scanner to do thorough write/read tests for a few pases, but otherwise
all you can do is replace the drive by the time it's expected lifespan
has elapsed, which used to be 3-4 years, though that might've changed
by now.


Dave
 
kony said:
Mostly what I hear about is situations where the drive is bad right at
the start, or works fine for at least a few months. Unles you hear
strange noises coming from the drive right away it's unlikely you'd
benefit from an extended burn-in of a few days to weeks.

There isn't really much you can do beyond keeping it cool and using a
decent power supply, but if you're that concerned about it then run a
Raid mirror. If wouldn't hurt to check the drive with the
manufacturer's diagnostics right after installing it, and using a disk
scanner to do thorough write/read tests for a few pases, but otherwise
all you can do is replace the drive by the time it's expected lifespan
has elapsed, which used to be 3-4 years, though that might've changed
by now.


Dave
Click to expand...

HDs seem to have MTBFs of ~1 MHrs or so; roughly ten years. Other than
infant mortality (the steep end of the bathtub curve), PC HDs are usually
replaced long before they fail, since few folks keep a PC for a decade.
 
Bob WIllard said:
HDs seem to have MTBFs of ~1 MHrs or so; roughly ten years.
Other than infant mortality (the steep end of the bathtub
curve),
Click to expand...

My question is approximately how long does this steep part last for?
It is clear just from what people have posted here a lot of the
drives which get RMA-ed are relatively new.

PC HDs are usually replaced long before they fail, since
few folks keep a PC for a decade.
Click to expand...

Ah yes, I have some old 80 MB drives which are still going. :-)
 
Bob WIllard said:
HDs seem to have MTBFs of ~1 MHrs or so; roughly ten years.
Click to expand...

Unfortunately that is not what MTBF means. The number in hours is
there for a reason, otherwise they would just publish it as x.y years.

MTBF is about the chance that *your* drive as part of a very large
population will fail.
Problem with chance is that it can happen tomorrow or in over 10 years
 
I would guess that the failure curve of a hard drive is much like
that for a lot of electronic equipment. It is probably a sort of
wonky "U" shape.

If it fails then the drive is likely to fail quite early on in its
life. After it passes this period then it is more likely to fail
after some years as it gets older.

Does anybody know approximately how long this period is of relatively
high risk of early failure? This must vary from one model of hard
drive to another but, very approximately, after what period of time
would it be?

I may be wrong but I would guess there is a bit of a "knee" in the
shape of the reliability curve where the risk of failure changes.

If I get a new ATA/IDE hard drive I don't like to use it for live
data until it has been running on test for a while. But how long is
sensible for this test period?

At a usage pattern of 10 hours a day for the hard drive, would this
period be three or four days? A week? Three weeks?
Click to expand...

The short story is that while that bathtub effect is certainly seen,
it isnt a lot of use for a particular new drive, even if you have the
full stats the manufacturer has for that particular model, with one
they have been producing for a few months etc.

You've basically just got to recognise that the risk is unquantifiable
higher and that you should be fully backed up anyway with the stuff
you'll slash your wrists if you lose, because you might be the sucker
that ends up with the 1 failure in 10K that fangs you on the arse etc.

Fortunately hard drives are now so cheap that you can afford
to have full backups of what is essential, and full image backups
or RAID for convenience on hard drive infant mortality too.

You should also consider running continuous
protection like PowerQuest's V2i Protector.
 
In my view, the statistics don't mean very much. If a hard drive fails
early, it is still under warranty. As for failing late, it will.
 
Back
Top