First of all, no, I am talomg noe ion terms of MTTF because tyhat reduces confusions. Secondly, it is not a gut feeling . That is your department. Its a lifetime as a design engineer talking.
Stiff that wears out has a definite tendencey to wear out after a defined period plus minus a statistical uncertainty.
A product that wears out in three years on average, does not wear out in
30 second nor in 30 years, It twends to wear out in about 3.,
What I am sating is that MTTF is not uniformly distributed in real world systems. In is more or less normally distributed in the sense statisticians use.
By your logic people 30 inches tall and 30 feet tall would be as common as people 6ft tall
For sure an average doesnt vtell you what most values are, but an average with a normal distribution and a standard deviation known, does.
--
New Socialism consists essentially in being seen to have your heart in
the right place whilst your head is in the clouds and your hand is in
someone else's pocket.
Yawn. You are a total asshole and you persist in peddling false logic.
Piss off.
--
New Socialism consists essentially in being seen to have your heart in
the right place whilst your head is in the clouds and your hand is in
someone else's pocket.
Mean Time Between Failures implies a repairable system, i.e. one that may fail, be repaired, fail again, etc. Mean Time To Failure is the term to use on a system that will not be repaired after it fails. Disc drives are normally not often repaired, so MTTF is more appropriate.
You snip my comment that explain my reasoning. Do you expect to be taken seriously?
Ok, that is pretty clear. But it does not indicate that the MTBF will be a lot longer than the end of the bathtub curve. It simply says the two things are different measurements.
When you decide to get rational about discussing this let me know.
+1. I'm one of the people who /does/ understand that MTBF/MTTF apply only during the useful life of the product, which is another figure that has to be quoted in order to make MTBF/MTTF meaningful.
I've seen a disc drive specification of MTTF of a million hours over a lifetime of 5 years. Do your own arithmetic: the quoted MTTF is many times the product's lifetime. The figures are correct.
Beyond the lifetime, the MTTF is normally expected to fall rapidly.
I've see the same from a journalist writing for El Reg. Bearing in mind what he was writing about, he should have known much better.
Yes, it is a mistake commonly made. And when watching the reactions from people here when explaining the situation to them, it is apparently deeply rooted. People don't like to hear that what they have believed to be reading into a large MTBF figure is not really what they always thought it was.
It is often difficult to keep the unit of a specified number away from the "gut feeling" of what a unit usually means. When MTBF is specified as a time (hours), the brain cannot easily turn away from the notion that it must be related to lifetime.
It is similar to the "specific impulse" of a rocket which is specified in seconds. But it is NOT the time the rocket engine will burn during the launch.
But many enterprise drives _are_ 2.5" inch form factor, either for higher revs or just plain storage density. The Seagate Savvios for instance are 2.5" across the range.
It still holds, wear is not the only failure mode and indeed wear can be massively accelerated to the point that the device is still in its early life, e.g. bearings can be misaligned, mechanisms can have swarf caught in them etc. Then there are the other failures
- a bad joint can fail in use or whatever. Infant mortality is essentially things that weren't quite right to begin with but take a little use to manifest themselves - it's a much greater category of failures than the usual age-related wear.
Why do you say higher revs? Maybe I'm not up to date, but the fast spinning drives were always 3.5" because they dissipated a lot more heat and the size helps to get rid of it better. Has that changed?
You seem to be saying that MTTF when used to describe a disk drive is fundamentally different from MTTF when used to describe, say, an elephant. If so, how does the calculation of the average age at death of elephants differ from the calculation of the MTTF of disk drives. If not, why can't I use MTTF for disk drives as an indication of how long my disk drive will last in exactly the same way as I use the average age of death of my elephant calf.
As I pointed out he is using the fundamentally wrong assumption that the distribution of failures is evenly spread over time:
As we know this is completely false in any machine where it tends to have a rash of early failures due to poor manufacturing, then a period of few failures followed by a rising failure rate as age related degradation sets in.
--
New Socialism consists essentially in being seen to have your heart in
the right place whilst your head is in the clouds and your hand is in
someone else's pocket.
NO. I don't make this assumption! I just try to explain you that the MTBF is in no way related to the point in time where the number of failures starts to rise.
That is a common misconception, that is why I try to explain it. But you are not open to the explanation, and neither is Gordon.
So this is a lost case. Go on and believe what you think is right, but I know that it is wrong and that you are being mislead.
you're both wrong. According to those definitions the MTBF is defined as L/n where L is the operational lifetime so far and n is the number of failures during L. IOW this *only* applies to something that gets repaired and put back into service. Since 3.5" and 2.5" drives are not repaired, the MTBF doesn't apply: only the MTTF is applicable and is calculated from MTTF = B10/0.1*n where B10 is the number of operations at which 10% of the devices fail and n is the number of operations.
See this as well:
formatting link
which equates the MTTF to the averaged bathtub curve for a sample of devices.
I suppose the bottom line is that when an MTTF is calculated rather than being measured its just advertising bullshit.
--
martin@ | Martin Gregorie
gregorie. | Essex, UK
org |
By that definition MTTF=MTBF for a device that doesnt get repaired.
manufacturers-post-this-how-should-you-i
Well it may or may bot be: For example I have a friend who des as a hoby te solution of enormous matruces - well beyond et capability of RAM so he sets these programs going and leaves them to run for months, with the spinning rust being accessessed constantly.
His MTTF averages between a year and 2 years.
When I was operating servers on the internet a decade or two back, the sort of MTTF we got was 3-5 years. These servers were not under used, but were not constantly accessing data.
On my personal cloud here, I get similar sorts of figures for disks in desktops in daily use, and in servers in daily use.
What is reasonably certain, is that failure within 1 yr. warranty is very rare, and disks lasting more than 5 years are very rare, with 3-5 years being the area where MOST of the drives I use go.
However knowing what I know, YMMV is the order of the day. How the disk is mounted, its usage pattern, number of seeks, temperature humidity and dust all play a huge part. Also number of reboots from cold is pertinent. In some ways a 24x7 server disk although its actually in constant use, tends to last longer as its operating typically in a stable environment, than a desktop disk being booted every day.
And finally, by the time you actually have the life data on any given disk design, chances are its obsolete anyway ;-)
So its lifetime characteristics are rather academic.
It reminds me of my early exposure to undersea cable repeaters. They used germanium transistors. Because, at that time, silicon transistors, although everybody thought they would be better, simply hadn't been around long enough to have real world MTTF figures available,.
Or my question to an oncologist 'what is the long term survival rate of this chemotherapy' .. 'well we have only been using it fifteen years, and so far its better than the rest, but beyond that, we don't know' :-)
Pragmatically, the solution is to have you data on at least two disks and regularly synced, so that if one goes, you simply buy another and replicate it from the other.
All drives are relative crap, but most will last nearly as long as the computer they are connected to.
I run 5-10 years old motherboards and RAM with 3-5 years old disk, in general, as I don't have exacting requirements or money to burn. I expect the disks to fail one day. I just hope they dont all fail at the same time!
>
--
New Socialism consists essentially in being seen to have your heart in
the right place whilst your head is in the clouds and your hand is in
someone else's pocket.
A three way mirror or FEC with two drive redundancy is best because that way when a drive does fail you still have redundancy while you're replacing it (which can be a while if you have to wait on an RMA replacement) and copying data to it (which in a lightly loaded system may be the busiest time the drives have had).
--
Steve O'Hara-Smith | Directable Mirror Arrays
C:>WIN | A better way to focus the sun
The computer obeys and wins. | licences available see
You lose and Bill collects. | http://www.sohara.org/
You may have found a formula that is sometimes used to quickly arrive at a MTxF, but it would be applicable to both MTTF and MTBF. The above cannot be correct because the resulting dimension (unit) is not time. So there must be another factor (length of each operation).
No, it is important to note that although it is a "mean time", it is not to be understood as an elapsed time before failure of the device. It is "mean" because it expresses the failure rate for a large number of devices, not because it averages the elapsed time before a unit fails.
It is, if you are really concerned about uptime, but in fact for my purposes I tend to leave it at two: If it takes me a day to rebuild, its not that hard a task...
--
New Socialism consists essentially in being seen to have your heart in
the right place whilst your head is in the clouds and your hand is in
someone else's pocket.
What you see there is that the typical lifetime of a disk is 3-5 years, and then most of them fail. Under heavy use, for cheap disks, it is shorter. Good disks that are specified for continuous use still achieve
3+ years.
Your disks probably still have an MTBF of 20 years, but that only expresses the number of failures within some specfied time. I.e. it tells you how many disks will fail in the second year. Not how long they will work on average.
I may be goimng senile, but I cannot actually see any difrence.
We nave n devices they fail after X(n) months,. The average of X(0--n) is the MTTF, it is the average life of the units.
What the 'failure rate for a large number of devices' is, if not that, is beyond me..
--
New Socialism consists essentially in being seen to have your heart in
the right place whilst your head is in the clouds and your hand is in
someone else's pocket.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.