Online analysis of hard drive

[mara-]

Hi,

I just installed Speed Fan and on tab S.M.A.R.T I selected In dept analysis of my main hard disk, WDC WD2000KS-00MJB0 with firmware 02.01C03.

Here are the results:

http://www.hddstatus.com/hdrepshowreport.php?ReportCode=2915893&ReportVerification=BD173F1A

Can somebody explain me those warnings? I know that web page offered explanations, but I don't understand it so it would be great if someone can give me simple explanation of those warnings. Also, I would like to know if there is anything I can to prevent those warnings? And do I need to be worried that hard disk will fail? As I could see values are pretty close to recommended but I would like opinion of someone who understand this better then me.

Cheers

[jaclaz]

NO need to worry, your drive appears in exceptionally good state.

The "analysis" is carried by comparing the values of your hard disk against the data collected by the site over time from users like you that tested their hard disks.

It's a kind of "statistical" data, suffering, as most "pseudo-statistical" data from the main error these kind of data may suffer of, sampling error.

http://en.wikipedia.org/wiki/Sampling_error

Basically, statistical data is more accurate as much you lessen the sampling error (or corrections are introduced in the data).

Example:

your computer shows that has been powered on for less time that the average of samples.

If it's a relatively old model, and you bought it new recently, this can be a valid explanation.

But if a relatively high number of the other samples against which your data is compared were posted by people running thir PC 24/7, it could be another valid explanation.

On the other hand, if a relatively high number of the other samples were coming from, say, part-time office users, that only run their PC 4 to 6 hours per day, and you are running it 24/7 it could be more difficult to find a valid explanation.

The accuracy of such comparisons without the details of the sampling criteria (and corresponding classufication of the data you supplied) and/or number of samples is very, very, very low.

Anyway, "common sense" applies also to senseless data, you have 99 instead of an average of 100 on two of the "warnings", and 92 instead of 100 on the third one.

To give ANY, even minimal, relevance to the warnings, it would mean that the analysis has an accuracy below or around 1% for the first two, which is mathematically and logically impossible, and below or around 8% for the third, which is possible, but very, very, very unlikely.

If you had results like 50/100 or 32/100, while there would still be no concrete reason for being preoccupied, but you might have wanted to carry on some more investigations.

http://www.humorsphere.com/simpsons/homer-simpson-quotes.htm

Oh, people can come up with statistics to prove anything, Kent. 14% of people know that.

S.M.A.R.T. itself is theoretically a predictive algorithm attempting to produce a preventive warning BEFORE disaster strikes, but in "real" world, Murphy's Laws will ALWAYS prevail.

Real world field reports assign a good reliability to SMART predictions only for a limited number of parameters/errors, but overall, a very low one, expecially on single drives.

This one is AFAIK one of the most recent study on a very large sample, and possibly the only one with such a large sample:

http://labs.google.com/papers/disk_failures.pdf

5 Conclusions

In this study we report on the failure characteristics of

consumer-grade disk drives. To our knowledge, the

study is unprecedented in that it uses a much larger

population size than has been previously reported and

presents a comprehensive analysis of the correlation between

failures and several parameters that are believed to

affect disk lifetime.

....

....

One of our key findings has been the lack of a consistent

pattern of higher failure rates for higher temper-ature

drives or for those drives at higher utilization levels.

....

....

Our results confirm the findings of previous smaller

population studies that suggest that some of the SMART

parameters are well-correlated with higher failure probabilities.

....

....

Despite those strong correlations, we find that

failure prediction models based on SMART parameters

alone are likely to be severely limited in their prediction

accuracy, given that a large fraction of our failed drives

have shown no SMART error signals whatsoever. This

result suggests that SMART models are more useful in

predicting trends for large aggregate populations than for

individual components. It also suggests that powerful

predictive models need to make use of signals beyond

those provided by SMART.

drafts.

jaclaz

[mara-]

OK, thank you for such detail explanation. I'm much calmer now.

Cheers

Edited September 6, 2008 by mara-