Installing freeBSD on an Intel RAID5 partition

[Nodje]

   hum
   let's keep practical here please. The question is whether you can use
   SATA RAID as a reasonable HD failure protection system or not.
   Can a Raid1 on two HD, say less than 500Gb, be consider as a good
   protection against HD failure? It still seems to be for me. (I
   consider recovery to be just a bit to bit copy, not sure to rigt here)
   On a Raid5, since there's need for computation in case of HD failure,
   it seems more discutable after the facts that you've exposed. It seems
   that this assumption needs statistics depending on HD size which I'm
   not able to produce. But with reasonably sized disk, it should still
   be ok, isn't it?
   Jeff Mohler wrote:

SATA drives just aint built with the same resiliency as SCSI, hence
the massive difference in cost.


So..as an example, the Hitachi 500G 7K500 drive has a non recoverable
bitrate of 1 in 10^14th.  The 10K300 FCAL (basically scsi) drive is 1
in 10^16th.  Those two zeros mean a _lot_.


I removed a lot of my own math here, knowing that Ive read this
somewhere before..huzzah for google!

[1]http://blogs.netapp.com/dave/2006/03/expect_double_d.html?no_prefetch=1

Im used to working with much larger drives, in very large RGs..so Im
correctable, youre not going to play with the devil TOO much in a home
for small business system, just not enough drives.

But now you can find 1TB drives, and 7 of those in a raid wont be hard
to find pretty soon.

Eventually..you will hit a non recoverable bit error during a
reconstruction, and you wont have parity to go to, to recover it.
Unless youre using a dual parity layout of some type.

Drives are also more common to fail when put into use from being
spares, because theyve never been exercised over a long period of
time..ya never know.

The quality of the firmware that operates consumer SATA isnt near the
level of quality that server drives are either, which can create ghost
errors that dont truly exist, but to the OS are in fact errors which
can shave off a few zeros as well.

On 10/10/07, Nodje [2]<nodje.co at gmail.com> wrote:


 well, you mean on RAID5 then, coz there's probably no math in
reconstructing a RAID1.

 Why would the math on SATA be less reliable than on SCSI???

 Where d'you read that anyway??



 Jeff Mohler wrote:
 Did you know that most "oh my god" RAID failures happen during the
reconstruction of a failed drive?

.Especially on SATA as the non-recoverable-bit-error math is so much
easier to run into.

I think..that on a 500G drive, there are enough bits to read/write
that mathematically you could run into a double-drive failure every
time you have to recover. Although, statistically it wouldnt happen
every time.

No raid solves any backup problem.



 I've been using those Intel RAID with Windows for a couple of years now and
it really helped solve my backup problem.
I think this is simply great, no worries of data loss anymore (at least
coming from hardware failure).

-nodje
_______________________________________________
[3]freebsd-questions at freebsd.org mailing list
[4]http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to
[5]"freebsd-questions-unsubscribe at freebsd.org"

References

   1. http://blogs.netapp.com/dave/2006/03/expect_double_d.html?no_prefetch=1
   2. mailto:nodje.co at gmail.com
   3. mailto:freebsd-questions at freebsd.org
   4. http://lists.freebsd.org/mailman/listinfo/freebsd-questions
   5. mailto:freebsd-questions-unsubscribe at freebsd.org
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 3241 bytes
Desc: S/MIME Cryptographic Signature
Url : http://lists.freebsd.org/pipermail/freebsd-questions/attachments/20071011/8da7e6fc/smime.bin