Stupid RAID question

Lennart Sorensen lsorense-1wCw9BSqJbv44Nm34jS7GywD8/FfD2ys at public.gmane.org
Tue Oct 4 21:42:33 UTC 2011


On Tue, Oct 04, 2011 at 04:26:28PM -0400, D. Hugh Redelmeier wrote:
> Yeah.  And soon there will only be two manufacturers, so good luck in
> diversity in systems larger than 2 disks.
> 
> If you really have an IDE / PATA interface, I wonder about putting
> /boot on a Compart Flash device and using one of those inexpensive
> gizmos that allows them to be plugged into PATA.  Apparently not all
> of them support DMA which may matter.  /boot isn't used after booting
> so the small capacity wouldn't matter and the physical space and power
> would be small.  Besides, it would be easy to have another in your
> desk as a spare.
> 
> Note that most current consumer hard drives are not suitable for RAID.
> The can take a large amount of time doing internal error recover,
> enough that the RAID system decides the whole drive is dead.  That is
> a practical disaster because rebuilding an array on a 2TB disk takes a
> large portion of a day.  See for example:
>   <http://en.wikipedia.org/wiki/Time-Limited_Error_Recovery>
> This is a scam as far as I'm concerned.  Perpetrated by the last two
> surviving drive manufacturers.

No I don't consider that a scam.  Most consumers don't use raid, so if
the drive can recover from a failing sector by trying multiple reads to
see if it can get lucky, then that makes it a better drive.

The fact raid systems have a time limit on a drive responding makes the
above behaviour undesirable when using it in a raid.

So the two uses have mutually exclusive needs.

To some extent I think WD sells raid edition drives for more money than
non raid drives just to make sure consumers don't buy them thinking
"raid drives must be better or faster so I will use that" and end up
with a much less reliable drive in their single drive system.  People that
need them will know what they need and get the right one.  Now I would
prefer the old method where a software tool or a jumper could change
the behaviour, but unfortunately we don't get that option anymore.

-- 
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists





More information about the Legacy mailing list