Re: MTBF: A bunch of bs

Peter wrote:

>> >> Of course it may help for the purchase decision, but what does it help
>> >> you for the time you are using the drive? (It definitely helps the
>> >> person in charge of 1000 drives, for example for the decision about
>> >> how many drives to have in stock for replacements.)
>> >
>> > Once you have it, it helps you to find out how much it will cost you to
> do a
>> > maintenance for it.
>>
>> How do you calculate the maintenance cost of a single drive? Not the
>> average maintenance cost of one drive that's part of a set of hundred or
>> thousand drives; the actual maintenance cost of a single drive. Let's
>> assume an MTBF of 4M4 h,
>
> That is quite high for a 200$ drive.
>
>> or a failure probability of 1% in 5 y, and a price
>> of $200 for the drive. For simplicity let's not consider working time,
>> and call it replacement cost instead of maintenance cost.
>
> If time spent on diagnosing a problem, getting a new drive, installing it,
> restoring data from a backup; is free, it is hardly a business case. Also
> you did not count how much of business you have lost because of a
> downtime.
>
>> What you will come up with is that you may or may not have to spend
> another
>> $200 in those 5 y. The probability for having to spend another $200 is
>> small, but it's greater than 0. The only thing we know for sure is that
> the
>> replacement cost in those 5 y won't be $2 (1% of $200) -- it will be
> either
>> $0 (probable), $200 (unlikely), $400 (even more unlikely), or a still
>> higher multiple of $200 (still more unlikely).
>
> Based on your assumptions, expected maintenance cost is $2.

So you budget $2 and the drive fails. Now what? What have you gained by
budgeting $2?

>> Now apply the same MTBF to an array of 1000 drives. You'll come up with a
>> high probability that the replacement cost over the course of 5 y will be
>> around $2000 (1% of $200,000), or very likely in the range of, say, $1000
>> to $3000. (To give probabilities for any given replacement cost range you
>> need to assume a failure probability distribution, in addition to the
>> MTBF.)
>>
>>
>> >>> It seems that you have a different understanding of probability.
>> >>
>> >> Different from what other understanding?
>> >
>> > That probability of a failure applies the same to a single as to a huge
>> > number of components.
>>
>> Of course; I never said anything different. But what you can do with the
>> probability, its relevancy for a given situation, is quite different for
>> small and large numbers.
>
> Sure it is. A bad decision for only one drive is not as painfull as a bad
> decision with 1000 drives.

Suppose you buy a drive with an MTBF of x instead of one with an MTBF of 2x.
The probability of either failing during its rated service life is very
small, and the one with the longer MTBF can still fail if you are unlucky.

--
--John
to email, dial "usenet" and validate
(was jclarke at eye bee em dot net)
.

Relevant Pages

  • Re: MTBF: A bunch of bs
    ... it helps you to find out how much it will cost you to ... >> How do you calculate the maintenance cost of a single drive? ... >> thousand drives; the actual maintenance cost of a single drive. ... expected maintenance cost is $2. ...
    (comp.sys.ibm.pc.hardware.storage)
  • Re: Resources required to retrieved wiped disk
    ... even sure if bin Laden's disk would be worth the cost. ... Evidenciary procedure: $1000+ per unit of evidence ... Probability of recovery: varies based on how much contiguous you need, ... we know they actually switched the drives, ...
    (sci.crypt)
  • Re: MTBF: A bunch of bs
    ... >> person in charge of 1000 drives, for example for the decision about how ... How do you calculate the maintenance cost of a single drive? ... call it replacement cost instead of maintenance cost. ... The probability for having to spend another $200 is ...
    (comp.sys.ibm.pc.hardware.storage)
  • Re: MTBF: A bunch of bs
    ... >>> person in charge of 1000 drives, for example for the decision about how ... > How do you calculate the maintenance cost of a single drive? ... > call it replacement cost instead of maintenance cost. ... > high probability that the replacement cost over the course of 5 y will be ...
    (comp.sys.ibm.pc.hardware.storage)
  • Re: Purpose-built RAID 0 GEOM-based store
    ... 500GB 7200rpm drives. ... I'd recommend you compare the cost of downtime in case of failure (times the ... Cost of downtime/minute x minutes to restore from backup x probability ...
    (comp.unix.bsd.freebsd.misc)