[mythtv-users] new rig ... help with specs

[Chris Stevens]

On Tue, 2009-09-22 at 17:01 +0100, Robin Hill wrote:
> On Tue Sep 22, 2009 at 09:34:04AM -0600, Brian Wood wrote:
> 
> > On Tuesday 22 September 2009 09:23:27 Chris Stevens wrote:
> > > The situation with RAID hard drives is fairly straight-forward: buy the
> > > enterprise versions - eg. Seagate 750 ES. These have MTB of around
> > > 3,000,000 hours which is over 300 years and are only a bit more
> > > expensive.
> > >
> > > But there is just one primary rule for any HD: KEEP THEM COOL. Drives go
> > > very quickly if you let them overheat. You can actually buy any old
> > > rubbish so long as you follow this rule.
> > 
> > I would think so as well, but:
> > 
> > The Google paper on hard drive reliability seemed to contradict this
> > seemingly common sense idea.
> > 
> > They were unable to show much of a correlation between temperature and
> > failure rates.
> > 
> The thing with the Google paper is that it was all done within
> standard data centre operating conditions - when they showed no
> correlation between failure and temperature, this is within a very
> narrow range of temperatures.
> 
> Current thinking is that most data centres are being run at too low a
> temperature - with proper airflow planning you should be able to run
> them at (or slightly above) normal room temperature.  This doesn't,
> however, (as I can personally attest) mean that drives will work quite
> happily at 60 degrees (though modern CPUs appear okay at 90 degrees C),
> just that there's no need to keep everything below 20.
> 
> Cheers,
>     Robin
Yes, when I say keep them cool I do indeed mean under 60 degrees. I have
used smartmontools on RAIDed drives for 10 years now and they start to
log failures when the temp flag shows it's gone over the limit value
(whatever that is). I have been using 6x750 Barracuda ES drives now
(air-cooled to 30 degrees) for 6 years in a RAID zero config (they are
that reliable) and just simply keep a track of the SMART logs (you can
get an e-mail when a flag is exceeded). I only ever use RAID 0 since I
never get surprise failures (except on Maxtors!). I am now using 1TB
Samsung drives (6TB RAID0) and, again, they are fine when looked after.
I tend to keep over 1000 recordings in Myth (about 2.5TB). Been running
Myth about 10 years. The only problem I ever had was that a Reiser
filesystem starts to swamp CPU when there is only 3 percent free so it
caused stuttering - that kept me mystified for a bi!. Sorry for the
ramble.

Chris