Re: [SAGE] Long lived, cheap data storage

[Brad Knowles <brad@shub-internet.org>]

On 10/8/07, Sam Johnston wrote:

>  We're talking about archiving here so were any of these things to
>  happen it would just be a case of finding an alternative; the
>  probability of Amazon's data centers and your house going away at the
>  same instant are overwhelmingly small (and indeed if this were ever
>  the case you would presumably have more to worry about!).

Many people have said the same thing about double disk failure in a 
RAID array.  So why do we have RAID-6?  Or why would we need RAID-5 
with hot spares?

Double failures are relatively rare, but they do happen.  This is why 
many aircraft typically have triply redundant systems for those 
critical things that cannot be allowed to fail.  And yet, they still 
do fail, and aircraft do fall out of the sky.

NASA uses five separate computers in the shuttle, from two separate 
manufacturers, and at least the majority of them have to agree before 
they will go ahead and take an action (or allow an action to be 
taken).

How far do you want to push it?  Are you satisfied with a 90% 
solution?  Or do you need a few more nines?  How much are you willing 
to pay for those additional nines?


I'm not saying that I've got the One True Answer here, just that this 
is a case where I wouldn't be quite so quick to jump to the position 
you advocate above.

-- 
Brad Knowles <brad@shub-internet.org>
LinkedIn Profile: <http://tinyurl.com/y8kpxu>