[mythtv-users] Should I use LVM on this new raid?

Jay R. Ashworth jra at baylink.com
Thu Sep 27 03:09:56 UTC 2007


On Wed, Sep 26, 2007 at 06:53:25PM -0400, f-myth-users at media.mit.edu wrote:
>     From: "Jay R. Ashworth" <jra at baylink.com>
>     > So that implies to me that people really believe that business will
>     > *pay* a 6:1 or more price imbalance for those drives, for no good
>     > reason.
> 
> Bought any bottled water recently?
> 
> Seriously:  In many places, what comes out of the tap is better than
> what's in the bottle, at about a 9 -thousand- to 1 price ratio.  Not
> to mention all the other badness in the lifecycle costs that you may
> or may not be paying for directly (fuel to truck all those bottles
> around, fuel to make the plastic [energy & the plastic itself], fuel
> to recycle them...).  And yet, people will pay an -enormous- premium
> on ZERO data just because it makes them feel better and they can't be
> bothered to check the facts.  (Yes, occasionally they buy it for the
> portability.  But why, then, don't the vast majority just -save- a
> single bottle and refill it from the tap?  Yet they don't.)

Sure.

> I have -absolutely- no problem believing that many businesses will pay
> "protection money" for a warm fuzzy feeling that has no grounding in
> reality---especially if it -might once have- and nobody's rigorously
> investigated whether it still holds.  And consider---a -business- will
> pay almost any amount for a disk if they value its contents, because
> just having to swap it out will cost them more in personnel time and
> downtime (or degraded-RAID time, or chance-of-two-failures-before-rebuild)
> than the price differential.  -And-, the guy who pays this protection
> money can then say, "Hey, I bought the most expensive disks out there.
> They -must- be good or they couldn't charge that, right?  So the fact
> that the disk failed isn't MY fault!"---and he keeps his job.

Hmmm...  Across all of commercial IT, I'm still not sure I buy that...

> P.S.  Unless your sample size is -very- large, the low failure rates
> of disks in general mean that you don't actually have statistical
> significance to your data---and people are excellent at deducing
> patterns were no actual patterns are present.  That's the whole reason
> that "statistical significance" exists.

Certainly.  And, as I say, I can't attack the methodology of the Google
paper as I haven't yet read it.  And I'm not a professional
statistician.  I do *have* a professional mathematician available
though; perhaps I can get her opinion on their math.

Cheers
-- jra
-- 
Jay R. Ashworth                   Baylink                      jra at baylink.com
Designer                     The Things I Think                       RFC 2100
Ashworth & Associates     http://baylink.pitas.com                     '87 e24
St Petersburg FL USA      http://photo.imageinc.us             +1 727 647 1274


More information about the mythtv-users mailing list