1 |
Andrew Joyce <joyce@××××××××××××.ca> wrote: |
2 |
> If you consider the case where each drive alone has a a 10% chance |
3 |
> of failing, two drives simultaneously failing should have a 1% |
4 |
> chance of happening. |
5 |
|
6 |
This calculation assumes that the failure of one drive does not give |
7 |
any information about the possible failure of the second drive. That |
8 |
is not the case. If conditions were such that one drive failed, then |
9 |
you should become less confident about your other drives. |
10 |
|
11 |
For example, I have seen an entire room of Sun workstations have their |
12 |
hard drives go bad at once. Sun's explanation was that the bearings |
13 |
in these drives were packed with the wrong lubricant. These |
14 |
improperly lubricated bearings wore out and failed at a more or less |
15 |
constant rate. |
16 |
|
17 |
On another occasion, a friend of mine bought two hard drives, same |
18 |
make and model, with the plan of using the second drive for backup. |
19 |
Both failed at the same time. Two of the same model drive, probably |
20 |
made at the same place and time, and installed into the same |
21 |
computer -- if you think about it, it wasn't a good idea. Whatever |
22 |
would make one drive fail would likely cause the other to fail. |
23 |
|
24 |
(To put that in jargon: you have to multiply by the _conditional_ |
25 |
probability that the second drive will fail, contingent upon the |
26 |
failure of the first drive.) |
27 |
|
28 |
-- |
29 |
Barry.Schwartz@×××.org |
30 |
"I'm a war president!" |
31 |
http://www.winternet.com/~trashman/Bush_or_Pooch.html |
32 |
http://www.winternet.com/~trashman/Music.html |
33 |
|
34 |
-- |
35 |
gentoo-security@g.o mailing list |