galtsev at kicp.uchicago.edu
Tue Apr 19 18:44:38 UTC 2016
On Tue, April 19, 2016 12:27 pm, Jon Radel wrote:
> On 4/19/16 12:52 PM, Valeri Galtsev wrote:
>> Correct, we didn't hear proof of one or another. I, however, can not
>> of any physical mechanism that can be involved which will lead in case
>> failure of one drive to failure of another. That is why I assume events
>> are (pretty much) independent.
> Personally I think you're thinking about it all wrong. I didn't catch
> anyone claiming that failure of one drive caused the failure of another
> (though I've not slogged through the whole thread). Only that failure
> of drives from the same manufacturing batch have a tendency to have
> similar lifespans, and therefore there is some correlation as to when
> they die. They have the same design flaws. They have the same
> substandard components. Etc., etc. Why do you think that drives
> come with MTBF figures, and that they're not all the same for every
> drive? That's the summary output of somebody's best estimation of the
> failure curve for that particular model.
Jon, the reason for OP to never use "identical" drives in the same RAID
array was that in his opinion there is higher likelyhood of two drives
dies simultaneously (causing data loss) is high.
My opinion is, the above is not correct, I will not repeat my lengthy post
where I explained why.
My opinion that your conclusion that the drives having the same MTBF (or
similar lifespans) are likely to die simultaneously contradicts my
understanding of probability theory (or statistics basics). People life
expectancy in some country is 75 years. Likelyhood that two _given_ people
will die at age 40 is negligible. Even twin brothers are unlikely to die
during the same week or Month. This one is between fundamental assumptions
statistics is based on.
Just my humble opinion.
> Personally I've never had a clear cut example with disk drives, though I
> did own enough 750 mb drives from brand X some years ago to learn that
> the involved mechanism, on average, sucked compared to other brand X
> drives, and they were not to be trusted long term. However, I once saw
> this very clearly in LCD monitors where I had multiples of the same
> model. Turns out that they were caught up in the big "counterfeit"
> capacitor scandal way back when, when due to imperfect industrial
> espionage, a manufacturer only thought they'd stolen the formula for a
> nice electrolyte. Turns out there was one little factor they'd missed,
> the stabilizer that kept the capacitor from dropping dead at pretty much
> the same age as every other capacitor of that model. There most
> certainly was correlation on those monitors failing. Luckily they were
> all still in warranty. :-)
> --Jon Radel
> jon at radel.com
Sr System Administrator
Department of Astronomy and Astrophysics
Kavli Institute for Cosmological Physics
University of Chicago
More information about the freebsd-questions