Maximum read/write operations on a drive

[jondelac]

Hi... What are the average read write cycles to a drive's sector before
it goes bad? For example, in NAND based flash memory, after 100,000
read write cycles, they go bad... What are the general numbers for
popular hard drives?

Jon

 

[Arno Wagner]

Hi... What are the average read write cycles to a drive's sector before
it goes bad? For example, in NAND based flash memory, after 100,000
read write cycles, they go bad... What are the general numbers for
popular hard drives?

Infinity. Even if you wirte the same sector over and over again, it
will take longer to go bad than the drive.

Flash has the problem that electrons need to be forced through an
isolator on writes. That damages the isolator over time. Changing
the magnetic orientation (HDD), does not damage the material. As
there is also no head-surface contact in a HDD, there is no machanic
damage to the sector either.

Arno

 

[mike]

Infinity. Even if you wirte the same sector over and over again, it
will take longer to go bad than the drive.

Interesting perspective on "infinity"
My definition of a bad drive is when you can no longer read/write one or
more sectors???? Head-surface contact is a significant failure mode
for hard drives.

Depending on exactly what you mean by flash "read write cycles"...
reads are much less destructive than writes. Writes may be accomapnied
by a read, but reads rarely include writes.

Butwhaddaiknow?
mike

 

[jondelac]

Infinity. Even if you wirte the same sector over and over again, it
will take longer to go bad than the drive.

Flash has the problem that electrons need to be forced through an
isolator on writes. That damages the isolator over time. Changing
the magnetic orientation (HDD), does not damage the material. As
there is also no head-surface contact in a HDD, there is no machanic
damage to the sector either.

Arno

Thanks... what then causes disks to have a life time? If a disk were
not mobile (that is, not in a laptop etc) and stored at ideal
tempratures, what will cause sectors to go bad?

Jon.

 

[Rod Speed]

What are the average read write cycles
to a drive's sector before it goes bad?

On average the drive gets discarded because
its too small and too slow before it dies.

For example, in NAND based flash memory,
after 100,000 read write cycles, they go bad...

Hard drive technology doesnt have that limitation.

What are the general numbers for popular hard drives?

It doesnt die and gets replaced when its too small or too slow.

 

[Rod Speed]

Arno Wagner wrote

Interesting perspective on "infinity"

Nope, he's right. As long as the drive hasnt died, and most
dont die, they get replaced because they are too small or
too slow, you can read and write the sectors as often as
you like, there is no limitation on the read/write cycles at all.

My definition of a bad drive is when you can no longer read/write one or more sectors????

Different matter entirely. And when the drive dies, that isnt
determined by the number of read and write cycles anyway.

Head-surface contact is a significant failure mode for hard drives.

Nope, that doesnt happen much at all.

 

[Rod Speed]

(e-mail address removed) wrote

Arno Wagner wrote

Thanks... what then causes disks to have a life time?

Mostly now its not cooling them properly.

If a disk were not mobile (that is, not in a laptop etc) and stored
at ideal tempratures, what will cause sectors to go bad?

Nothing, and that is why the number of read and write cycles is unlimited.

Better to say unlimited than infinity.

 

[mike]

On average the drive gets discarded because
its too small and too slow before it dies.




Hard drive technology doesnt have that limitation.




It doesnt die and gets replaced when its too small or too slow.

I must be living in a different universe.
I have rarely retired a drive because it was too small or too
slow. I retire drives when the bearing noise gets so loud I can't
stand it or when a head crash (physical contact between head and media)
has rendered too many sectors unreadable.

I've all but quit buying used hard drives because they ALWAYS
have a bunch of bad sectors.

I was poking fun, but I still have a symantic difficulty with the use of
the word
"infinite" to describe the life of anything...your preferred diety
excepted. Just because you can reverse the magnetic domains a LOT of
times, doesn't mean that sectors don't fail much sooner for a variety of
reasons.

It's gonna be interesting to see what happens when the flash-assisted
hard drives get into wide use and you get the worst of both worlds.
"Extra, Extra, Read all about it!
New FlashKill virus retargets your swap space to flash and destroys
your hard drive within days...film at eleven."
Can't wait to see the warranty disclaimers. I hope they have the
foresight to include a jumper to bypass the flash.
mike

 

[Bob Willard]

I must be living in a different universe.
I have rarely retired a drive because it was too small or too
slow. I retire drives when the bearing noise gets so loud I can't
stand it or when a head crash (physical contact between head and media)
has rendered too many sectors unreadable.

I've all but quit buying used hard drives because they ALWAYS
have a bunch of bad sectors.

I was poking fun, but I still have a symantic difficulty with the use of
the word
"infinite" to describe the life of anything...your preferred diety
excepted. Just because you can reverse the magnetic domains a LOT of
times, doesn't mean that sectors don't fail much sooner for a variety of
reasons.

It's gonna be interesting to see what happens when the flash-assisted
hard drives get into wide use and you get the worst of both worlds.
"Extra, Extra, Read all about it!
New FlashKill virus retargets your swap space to flash and destroys
your hard drive within days...film at eleven."
Can't wait to see the warranty disclaimers. I hope they have the
foresight to include a jumper to bypass the flash.
mike

The OP's use of the phrase "read write cycle" makes me wonder if the OP
thinks that reading from a HD is destructive, like reading from a core
stack of yore. Reading from a HD is non-destructive, and there is AFAIK
no known wear-out mechanism associated with reading a sector; in fact,
typical HDs are continuously reading from some track (at least when not
writing or seeking). Writing to a HD is not done with a read-write cycle,
but is done by directly over-writing whatever data is already present; and,
AFAIK, there is no known wear-out mechanism associated with writing a sector
(assuming a decent design and a well-cooled HD).

If you want a spec for the lifetime limit on reading a sector, I'll pick
432,000,000,000. Why that number? A 7200 RPM HD, while idle, is reading
each sector of the track on which it is "parked" 7200 times per minute, or
432,000 times per hour; with the kinda-typical HD MTBF of 1,000,000 hours,
you should expect the HD to fail after ~432,000*1,000,000 sector reads of
any given sector.

{Please, I'm not trying to restart the debate over how meaningless MTBF
specs are; I note only that my "spec" for a lifetime sector read limit is
about as meaningful as the MTBF on which it is based.}

In summary, the magnetic portion of HDs do not wear out, regardless of the
number of reads or writes done to them. Bearings will wear out, and
connectors and electronics will eventually fail, but those events are also
independent of the number of reads or writes.

 

[phreak]

You guys all need to read up a bit on chaos theory and the how/why of
organization devolving into chaos. Or the degradation of magnetic
fields which is a core issue of magnetic storage.

 

[Rod Speed]

Rod Speed wrote

I must be living in a different universe.

Or you havent been doing it for as long as some of us have.

I have rarely retired a drive because it was too small or too slow.

Then you havent been doing it for as long as some of us have.

I retire drives when the bearing noise gets so loud I can't stand it

I've only ever had one of those. And its not even in one of my
systems and the owner just puts up with it. I wouldnt if it was mine.

or when a head crash (physical contact between head and media) has rendered too many sectors
unreadable.

Never ever had one of those, and only one drive failure. I had to
physically kill that drive because of random new bad sectors that
were resumably due to a bad joint or flexible cable break inside
the sealed chamber. I had to kill it because the new bads were
so rare that I didnt want to get involved in endless hassles where
it looked fine when I had RMAed it.

Surprisingly hard to kill too. I first tried grossly over voltaging
pins on the 40 pin connector, nothing did any harm. I then
looked the pinouts of the ics up on the net and zapped ics
themselves. Eventually managed to blow the side right out
of one of the ram chips.

I've all but quit buying used hard drives because they ALWAYS have a bunch of bad sectors.

I dont bother with used hard drives.

I was poking fun, but I still have a symantic difficulty with the use of the word "infinite" to
describe the life of anything...
your preferred diety excepted.

Dont have one of those terminal stupiditys.

Just because you can reverse the magnetic domains a LOT of times, doesn't mean that sectors don't
fail much sooner for a variety of reasons.

They dont fail because of the read write cycles being discussed.

It's gonna be interesting to see what happens when the flash-assisted hard drives get into wide
use and you get the worst of both worlds. "Extra, Extra, Read all about it!
New FlashKill virus retargets your swap space to flash and destroys your hard drive within
days...film at eleven."

It was never seen with drives that can have their microcode changed.

Can't wait to see the warranty disclaimers. I hope they have the foresight to include a jumper to
bypass the flash.

Mindless paranoia.

 

[Rod Speed]

You guys all need to read up a bit on chaos theory
and the how/why of organization devolving into chaos.

Nope, that is completely mindless academic wanking.

Or the degradation of magnetic fields
which is a core issue of magnetic storage.

That isnt a problem seen with read and write cycles.

If anything more reads and writes reduces the chance of it degrading.

 

[Arno Wagner]

Interesting perspective on "infinity"

You can also remove the variable from consideration...

Arno

 

[Arno Wagner]

Thanks... what then causes disks to have a life time? If a disk were
not mobile (that is, not in a laptop etc) and stored at ideal
tempratures, what will cause sectors to go bad?

Mainly surface imperfections if no external damage is done.

However I think most drives die from mishandling.

Arno

 

[Arno Wagner]

I must be living in a different universe.
I have rarely retired a drive because it was too small or too
slow. I retire drives when the bearing noise gets so loud I can't
stand it or when a head crash (physical contact between head and media)
has rendered too many sectors unreadable.

I've all but quit buying used hard drives because they ALWAYS
have a bunch of bad sectors.

I was poking fun, but I still have a symantic difficulty with the use of
the word
"infinite" to describe the life of anything...your preferred diety
excepted. Just because you can reverse the magnetic domains a LOT of
times, doesn't mean that sectors don't fail much sooner for a variety of
reasons.

"Infinity" was a figure of speech. I basically meant "not relevant",
since whatever the actual number might be, some other limit ensures
it cannot be reached. This is not so uncommon usage in reliability
analysis.

It's gonna be interesting to see what happens when the flash-assisted
hard drives get into wide use and you get the worst of both worlds.

I guess we will just get HDDs that start out quite fast and then get
slower over time as the flash shrinks. And very happy vendors, since
you need to replace these disks far more often, because they get
slow. A big scam IMO.

"Extra, Extra, Read all about it!
New FlashKill virus retargets your swap space to flash and destroys
your hard drive within days...film at eleven."
Can't wait to see the warranty disclaimers. I hope they have the
foresight to include a jumper to bypass the flash.

As far as I remember, the flash will just have bad sector management
like the disk and will shrink over time or disable itself when some
amount of defects has been reached.

Arno

 

[jondelac]

(e-mail address removed) wrote


Mostly now its not cooling them properly.


Nothing, and that is why the number of read and write cycles is unlimited.

Better to say unlimited than infinity.

I understand now...

How then do they manage this issue in the new Hybrid Drive Samsung and
MS are making (http://en.wikipedia.org/wiki/Hybrid_drive)? Seems like
they use the NAND flash area for the more frequent read/writes and send
it to disk only a few times in larger buffers... Wouldnt that eat away
from the 100,000 cycles much faster?

Jon

 

[Rod Speed]

(e-mail address removed) wrote

Rod Speed wrote

I understand now...

How then do they manage this issue in the new Hybrid Drive Samsung
and MS are making (http://en.wikipedia.org/wiki/Hybrid_drive)?

Its a rather specialised drive intended to minimise drive
spinning time and so power consumption in laptops.

Seems like they use the NAND flash area for the more frequent read/writes

Not really, more to minimise the time the drive spins for.

and send it to disk only a few times in larger buffers...

Most drives dont actually write all that much.

Wouldnt that eat away from the 100,000 cycles much faster?

Sure, but its intended to maximise the time on batterys with a laptop.

 

[jondelac]

(e-mail address removed) wrote



Its a rather specialised drive intended to minimise drive
spinning time and so power consumption in laptops.


Not really, more to minimise the time the drive spins for.


Most drives dont actually write all that much.


Sure, but its intended to maximise the time on batterys with a laptop.

Ok.. then am I accurate in interpreting that somehow 100,000 read/write
cycles of data are acceptible lifetimes for use cases in laptops? I
know there are some specifics I may be glossing over, but generally, if
the unit is used 100,000 times, its done... so that must somehow
eclipse the general life time of laptops (I guess in my experience a
laptop has lasted atmost about 3 years before something went bad, maybe
monitor or something else)

Jon

 

[Arno Wagner]

Ok.. then am I accurate in interpreting that somehow 100,000 read/write
cycles of data are acceptible lifetimes for use cases in laptops? I
know there are some specifics I may be glossing over, but generally, if
the unit is used 100,000 times, its done... so that must somehow
eclipse the general life time of laptops (I guess in my experience a
laptop has lasted atmost about 3 years before something went bad, maybe
monitor or something else)

Or alternatively the manufacturer can charge more for the disk and
after some time it starts to get slow and hopefully the customers are
bying evem more of these ''self-degrading'' disks. Quite a scam that
is in preparation here IMO. Especially as results, e.g. from embedded
Linux, show clearly that you can reduce power consumption very well
otherwise.

One example is write gathering, were you defer writes for as long as
it takes to get a decent amount, or the user requests a shutdown or
the battery goes low. The flash is not needed at all.

Another is that boot-up speed is usually not determined by
loading from disk, but by hardware detection instead.

There is more. These things are completely unecessary. But as
many other bogus ''innovations'' they will find their market.

Arno

 

[Rod Speed]

Ok.. then am I accurate in interpreting that somehow 100,000 read/write
cycles of data are acceptible lifetimes for use cases in laptops?

Its not clear how many cycles it will have yet.

I know there are some specifics I may be glossing over, but
generally, if the unit is used 100,000 times, its done... so that
must somehow eclipse the general life time of laptops (I guess
in my experience a laptop has lasted atmost about 3 years
before something went bad, maybe monitor or something else)

Its obviously possible to socket the flash area so it can be
changed and you dont need to discard the entire laptop.

However, I doubt there will be much of a market for it myself.