Re: RAID 5 vs RAID 10 benchmark

Frank van Bortel <fvanbortel_at_netscape.net> wrote:
> Jonathan Lewis wrote:
> > I thought Hamming codes were a little sophisticated -
> > whereas Raid-5 parity simply does an XOR across
> > the data blocks to generate the parity block. That
> > way ANY one block is equivalent to the XOR
> > across all the other blocks in same stripe, which
> > is why you can reconstruct any failed block if
> > you have all but one of the blocks in the stripe.
> >
>
> Hmmmm, looks like my reply to you never made it...
> It read something like:
> Looks like you are correct about this, but how is
> data reconstructed then? Mind you - a missing disk,
> as well as a data error should be corrected, if it
> just happens to be.

RAID-5 doesn't correct individual byte-size data errors (or at least, the RAID-5ness isn't relevant to doing so.) If the data disk is alive and happily spewing gibberish, then you get gibberish. When reading, the RAID-5 parity information gets consulted only if the "data" disk has gone offline.

Xho

-- 
-------------------- http://NewsReader.Com/ --------------------
Usenet Newsgroup Service                        $9.95/Month 30GB