Talk:Annualized failure rate

The formulas in this article contain are not correct. The dimensions are definitely incorrect. The rest is highly suspicious (read: almost certainly wrong).

I don't know which underlying probability distribution might have been used to derive these formulas, but as a simple (in)sanity check of the examples, just consider that the given formulas would imply that no drive can ever survive the MTBF! Rather insane!

Without any explanation of how these formulas are derived, this article is not acceptable and should be removed. Pen-pen (talk) 16:15, 25 November 2007 (UTC)

I have to agree that the article needs improvement. As another example, the statement that with an MTBF of 5000 hours and a population of 100 disks, you can't have 175 disks fail in a year is absurd. You can fail all of the disks once and then fail 3/4 of them a second time before the year is out. This would result in an AFR of 1.75 just as the calculations state. Since there are 8760 operating hours and 5000 hours MTBF, it is not at all unlikely that 3/4 of the disks would fail twice. owendelong 19:05, 17 November 2009 (UTC)

It seems that my CAS doesn't seem to get this formula correctly... With the % sign the MTBF would solve to 12000, something that is low, but atleast not 1.2 million (without the % sign)... Another comment is that Seagate states AFR to show how good there HD's are, but at an AFR of 0,73% and thus an MTBF of 1.2 million it just doesn't seem realistic. --145.94.142.235 07:37, 25 February 2007 (UTC)

Don't use a CAS to do two divisions, especially if you don't understand the math. It's correct. Also, MTBF is always calculated using a sample under accelerated life condtitions, and any one of a number of models. Without going into the validity of this, the AFR can be calculated with a different model, or based on actual returns.71.132.128.11 07:30, 22 April 2007 (UTC)

What's the problem here? Why the quibbling with this excellent, brief article? I wanted to know what the term "Annualized Failure Rate" meant in Seagate's specs for a drive (as distinct from MTBF, which was also quoted in the same specs) so I Googled the term and this article came up. It not only told me exactly how Seagate were using the term, which is exactly what I wanted to know, it also referred me directly to the authoritative PDF product manual that, in turn, referenced all of the associated environmental assumptions, etc., that had been made. A minor clarification to the effect that the article references a particular manufacturer in a particular field's usage may be in order (though that's pretty evident from the footnotes); the ADDITION of other usages of the term - if any - would probably enlighten some readers; and a critique of the statistical methodology of the sort suggested above, under an appropriate subheading, might be a useful ADDITION for others. But, in itself and as-is, this article fits all the requirements of an encyclopaedic reference article and the "disputed" tag should certainly be removed (or at least changed to something indicating that the assumptions made by the referenced source - NOT THE ARTICLE, WHICH IS SIMPLY A 100% FACTUAL SUMMARY OF IT - are [possibly] questionable on statistical grounds). —Preceding unsigned comment added by 122.57.251.17 (talk) 08:58, 14 May 2009 (UTC)

I agree, this is a very useful article. I came here to learn what is AFR (in Seagate drive specs) and now I know it. The calculations are correct for me.

Maybe a general formula should be added before examples to make the article look more scientific. (But keep the examples - often they make the formula easier to understand). —Preceding unsigned comment added by 82.211.196.185 (talk) 22:06, 9 June 2009 (UTC)

"I don't know which underlying probability distribution might have been used to derive these formulas, but as a simple (in)sanity check of the examples, just consider that the given formulas would imply that no drive can ever survive the MTBF! Rather insane!" This is not so. MTBF is not the expected useful life. It really applies to large populations of items working well within their useful life. (re)-interpreting the examples, you would not reasonably expect a disk to last for 100's of years. What you might expect is that of a population of 100's of disks all running for a single year, a few would fail. MTBF is really misleading term, however, it is commonly used. MTBF != useful life... —Preceding unsigned comment added by 80.156.46.177 (talk) 17:50, 25 October 2009 (UTC)

From above "... Another comment is that Seagate states AFR to show how good there HD's are, but at an AFR of 0,73% and thus an MTBF of 1.2 million it just doesn't seem realistic"... What Seagate actually do is run 1000's of disks in an accelerated life environment, measure the failure rates, and from this deduce the real life failure rates (MTBF). This method is not perfect, but is relatively "quick". Other disk users/suppliers (say Google) actually measure the failure rates in field (see labs.google.com/papers/disk_failures.pdf). The whole basis of Seagate's failure rate calculation is that the probability of failure of a single disk run for 3 years is identical to say 6 disks running for 1/2 a year. This is where MTBF gets confusing, because in the extreme you say ahh then 1 disk running for 100 years has the same probability of failure as 100 disks running for one year. In Mathematical terms this is correct, in physical terms this is ridiculous. so MTBF should really be applied to large populations of independent components (like disks), or may be thought of a "total concurrent component mean time between failure". thus an Annualised Failure Rate of 1% means MTBF of 100 years. Personally I prefer AFR to MTBF as it is more intuitive and less likely to give rise to the misunderstandings above (which is common).