[Gllug] recommendation for external hard drive

[Peter Childs]

2009/10/22 Hari Sekhon <hpsekhon at googlemail.com>

> Bernard Peek wrote:
> > sean wrote:
> >
> >>  there used to be a web site where an individual was collecting
> >>  failure statistics and providing vendor based graphs, but it was
> >>  threatened with legal action and discontinued the survey 3ish years
> >>  ago. i remember that the best vendor at that time was samsung, with
> >>  hitachi second, and none of the samsung drives i've been buying since
> >>  then have failed. i think maxtor were bottom, but i had more interest
> >>  in what was top, so i could be wrong. it would be useful to have real
> >>  current evidence. google do have that, but won't release it. they
> >>  would probably purchase the majority of their drives from whichever
> >>  vendor is is top of their list, so that might be a clue.
> >>
> >
> > To me it seems self-evident that none of these studies are going to
> > provide the data needed for a sensible comparison. For even the worst
> > drives MTTF is substantially longer than the lifetime of the product. By
> > the time you have collected sufficient evidence to make a judgement
> > about a product it has been superseded and it's too late to take any
> > action on the findings. I remember the IBM "Deathstar" range was off the
> > market before there was much noise about the problem, and the production
> > company was sold off.
> >
> >
> I read a couple years ago that Google ran their infrastructure of
> servers with a single non-raided 80GB Maxtor in there, simply to be as
> cheap as possible because it didn't matter if the system failed, their
> proprietary layer on top deals with the resilience across large numbers
> of nodes.
>
> If that's true, then I'm not sure buying the same drives they buy would
> be beneficial?
>
>
Given the way I've heard Google works, that statement is probably still
true, I suspect they now use something rather larger than 80G disks.

But its widely known they were one of the biggest players behind the new
Journal Free ext4 options.

Google have a policy from what I read, of it it breaks chuck it and replace.
Which is not really an option for most of us....

Peter
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.lug.org.uk/pipermail/gllug/attachments/20091022/5e60244d/attachment.html>
-------------- next part --------------
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug