Subject: Re: Parity support for ccd(4)?
To: Thor Lancelot Simon <email@example.com>
From: Chris G Demetriou <Chris_G_Demetriou@BALVENIE.PDL.CS.CMU.EDU>
Date: 08/18/1995 22:58:24
> Is anyone else working on parity support for the ccd driver?
> RAID-4 style parity, with a dedicated parity disk, seems like it might not
> be too tough. Striped parity, probably somewhat harder.
Not to discourage you, but...
in my opinion, it's not likely to be _too_ easy.
Note that while 'ccd' as-is only has to worry about writing data, and
really doesn't have to do anything if a failure occurs, the whole
_point_ of a RAID is protection against failures, which means:
(1) being able to gracefully handle errors -- including making
sure that all of the blocks are read and written in the
right order, so that the data on the disk is consistent,
(2) having a "checker" program that checks the RAID unit
for consistency, and does all of the right things, on
and maybe even (if you're really going to try to do reliability):
(3) running in 'degraded' mode, where one disk is
disabled/unusable (obviously, only as long as necessary
until the system can be taken down to do reconstruction,
unless you want to do...), and
(4) online reconstruction.
Note that last i heard, the latter is still an open research topic. 8-)
There should be a fair amount of information about doing things like
this (software RAID) beneath the CMU PDL Web Pages, which can be found
Indeed, the PDL has done a complete software RAID framework,
RAIDframe, but it appears that information about it isn't present on
the pages yet.