On Sun, Jul 3, 2011 at 11:49 AM, Dave <[email protected]> wrote:
> To some extent my experience with hard drives has been random also,
> however I distinguish two types of drive failures.
>
> 1.  A "soft" failure where the drive simply starts having more and more
> bad sectors and has to be replaced.
> 2.  A serious failure "lights out" failure where all data is suddenly
> unavailable.
>
> Without a doubt,  I have had many more WD drives go from being ok one
> day to unavailable the next day than Seagate drives.
>
> I think that Seagate drives tend to fail softly while WDs simply go
> belly up.  "I" (myself and my customers) have had multiple WD drive
> failures like this in the past two years.

Well, I always blamed this on progress in the art of disk diagnostics.
There is a substantial layer of error detection and correction in the
firmware of every drive on the market since mid 90s, called SMART. It
was an IBM invention, but it got included in the ATA spec so that
everybody else had to implement it, too.

SMART looks for errors in the functioning of the drive, and corrects
correctable errors, and replaces sectors that consistently go bad. It
even keeps statistics on the ongoing drive health, so if you are
looking at SMART data, you can sometimes (or maybe even often?)
anticipate the drive going bad, but you'd have to either do it
manually (smartctl or skdump) or automatically (smartd).

If one does NOT pay attention to SMART, the failures appear
suddenly---small errors that would show up without it are masked, and
by the time you start seeing problems in the disk functionality it
probably is too late to get all the data

Google published a paper (
http://labs.google.com/papers/disk_failures.pdf ) looking at their
statistics of disk failures---as you may imagine, they have tons of
experience. They said that SMART doesn't predict all failures---some
disks go bad without any warning whatsoever---but that sector
reallocation events are a strong predictor if impending failure. They
didn't publish manufacturer comparison for partly legal reasons, and
partly because they claimed that there is no significant, consistent
advantage to any particular one.

------------------------------------------------------------------------------
All of the data generated in your IT infrastructure is seriously valuable.
Why? It contains a definitive record of application performance, security 
threats, fraudulent activity, and more. Splunk takes this data and makes 
sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-d2d-c2
_______________________________________________
Emc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-users

Reply via email to