[opensuse] Failure Trends in a Large Disk Drive Population [ Was: About Backing Up]
On 2/20/07, David Brodbeck <gull@gull.us> wrote:
Stevens wrote:
Auto manufacturers try to predict how their interiors and their paints will last, too, but until both are subjected to the Texas sun they are only guessing. The North has salt that kills cars; in the South it is the sun. Only when they obtain empirical data can they be sure and that data takes a long time to gather. The same goes for optical media manufacturers. Any longevity rating is a SWAG, at best, which is the reason for my cynical view.
Sure. As anyone who's ever had a couple of hard disks fail can attest, MTBF numbers are mostly fiction.
I don't know about that, it is amazing how often disk drives fail shortly after the warranty expires. Actually, the Google paper on this is very interesting. http://216.239.37.132/papers/disk_failures.pdf To grossly simplify, commodity disk drives fail at a rate of roughly 10% per year starting in the 3rd year, independent of activity, temperature, etc. Greg -- Greg Freemyer The Norcross Group Forensics for the 21st Century -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Tuesday 20 February 2007 13:53, Greg Freemyer wrote:
On 2/20/07, David Brodbeck <gull@gull.us> wrote:
...
Sure. As anyone who's ever had a couple of hard disks fail can attest, MTBF numbers are mostly fiction.
I don't know about that, it is amazing how often disk drives fail shortly after the warranty expires.
Why is that amazing? Does it not make perfect sense to guarantee disks for as long as you say they will probably remain free of error? If the two were out of sync, then we could conclude the manufacturer either is using poor techniques to measure the MTBF or misrepresenting the true value. Now, if they extend the warrantee well beyond the true MTBF, then they'll be paying for a lot of replacements, and it would be foolish or inept of them to do so. If the drives fail long after the warranty period, then they could, without additional expense, extend the warranty further and appear to their potential customers to have great quality products (which their drives do in fact possess). No matter how cynical you are, business still works best when all parties (customers and competing vendors) share accurate and complete information about the products and services being bought and sold.
...
Greg
Randall Schulz -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Tuesday 2007-02-20 at 14:26 -0800, Randall R Schulz wrote:
I don't know about that, it is amazing how often disk drives fail shortly after the warranty expires.
Why is that amazing? Does it not make perfect sense to guarantee disks for as long as you say they will probably remain free of error? If the two were out of sync, then we could conclude the manufacturer either is using poor techniques to measure the MTBF or misrepresenting the true value. Now, if they extend the warrantee well beyond the true MTBF, then they'll be paying for a lot of replacements, and it would be foolish or inept of them to do so. If the drives fail long after the warranty period, then they could, without additional expense, extend the warranty further and appear to their potential customers to have great quality products (which their drives do in fact possess).
Mmm. I don't think that is exact. For instance, cars here used to be guaranteed for a year, but they last ten, or more, with maintenance. Products usually last longer than the warrantied period; but on average, enough of them fail then so that it wouldn't be economical to extend the guaranteed. It is a question of deciding on the point. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.5 (GNU/Linux) Comment: Made with pgp4pine 1.76 iD8DBQFF25hitTMYHG2NR9URAvokAKCMCGkEufTIM0i0YTk5ETGD+wwlLACfWpJS 6/rqFHdf/s0U9Qz/XOhEQGU= =dAjw -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Carlos, On Tuesday 20 February 2007 16:54, Carlos E. R. wrote:
The Tuesday 2007-02-20 at 14:26 -0800, Randall R Schulz wrote:
I don't know about that, it is amazing how often disk drives fail shortly after the warranty expires.
Why is that amazing? Does it not make perfect sense to guarantee disks for as long as you say they will probably remain free of error? If the two were out of sync, then we could conclude the manufacturer either is using poor techniques to measure the MTBF or misrepresenting the true value. Now, if they extend the warrantee well beyond the true MTBF, then they'll be paying for a lot of replacements, and it would be foolish or inept of them to do so. If the drives fail long after the warranty period, then they could, without additional expense, extend the warranty further and appear to their potential customers to have great quality products (which their drives do in fact possess).
Mmm. I don't think that is exact.
It is if the business (the drive manufacturer) is run by rational people. That may be a big "if," but it's true.
For instance, cars here used to be guaranteed for a year, but they last ten, or more, with maintenance. Products usually last longer than the warrantied period; but on average, enough of them fail then so that it wouldn't be economical to extend the guaranteed. It is a question of deciding on the point.
If one's statistics on device failure are good, one can give a maximal warranty consistent with cost containment while simultaneously signaling customers your product's quality. If not, then either the business will not do well or the customers will be unhappy, or both. "Caveat emptor," like Murphy's Law, is readily misunderstood. Properly understood, it's not a statement of ultimate pessimism or cynicism, but rather an observation that a good business tells its customers the truth about its products or services as best it knows that truth and those customers are behooved to pay attention in making their purchasing choices. Then it's government's (and / or the court's) job to hold the business accountable for its claims, if it misrepresents them or fails to stand by its commitments. In the end, it's all statistics. Consider insurance. Those people know how to work statistics to run a cash-cow, guaranteed profit scam^H^H^H^Hbusiness.
-- Cheers, Carlos E. R.
RRS -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Randall R Schulz wrote:
Now, if they extend the warrantee well beyond the true MTBF, then they'll be paying for a lot of replacements, and it would be foolish or inept of them to do so.
Although it's not as if they'd be paying for *new* replacements. When I've had to do a warranty return on a hard disk, I 've never gotten a new replacement; I've gotten someone else's failed drive that they've "refurbished." -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Greg Freemyer wrote:
I don't know about that, it is amazing how often disk drives fail shortly after the warranty expires.
I seem to recall the MTBF figure was usually substantially longer than the warranty. I was going to verify that, but no one seems to quote MTBF anymore for hard disks -- instead they quote error rate, start/stop cycles, and annualized failure rate.
Actually, the Google paper on this is very interesting. http://216.239.37.132/papers/disk_failures.pdf
To grossly simplify, commodity disk drives fail at a rate of roughly 10% per year starting in the 3rd year, independent of activity, temperature, etc.
And yet Seagate claims an annualized failure rate of only 0.34% for their drives. Someone's skewing their figures, and I suspect it isn't Google. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Tuesday 20 February 2007 17:07, David Brodbeck wrote:
Greg Freemyer wrote:
I don't know about that, it is amazing how often disk drives fail shortly after the warranty expires.
I seem to recall the MTBF figure was usually substantially longer than the warranty. I was going to verify that, but no one seems to quote MTBF anymore for hard disks -- instead they quote error rate, start/stop cycles, and annualized failure rate.
Which is more specific information and much more informative than a lumped MTBF number.
...
RRS -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
participants (4)
-
Carlos E. R. -
David Brodbeck -
Greg Freemyer -
Randall R Schulz