Subject: Re: differential scsi controllers
To: der Mouse <mouse@rodents.montreal.qc.ca>
From: Jason Thorpe <thorpej@nas.nasa.gov>
List: current-users
Date: 07/02/1998 10:51:21
On Thu, 2 Jul 1998 06:59:24 -0400 (EDT) 
 der Mouse <mouse@Rodents.Montreal.QC.CA> wrote:

 > And this from someone who has a machine that can't run normally without
 > its ccd - it's got seven 80MB drives, of which one is root-and-swap,
 > one is /usr, and the other five are ccd-concatenated and hold
 > everything else the machine has.

mmm, ccd(4) :-)

And, for what it's worth, I've been using the ccd driver for a very long
time (originally on HP 7937H HP-IB drives, 4 striped, holding a large
NeXT software archive at the Oregon State University CS Department :-),
and have relied on it heavily ever since doing all of the ccd work in
NetBSD in 1995 (my development source and object trees have always been
located on ccd volumes since I've been at NASA).

We're also using ccd for our HSM system here at the NAS Facility... we
are using MegaDrive RAID arrays as ccd components (each RAID group appears
as a single SCSI disk instance)... the theory here is that each component
is protected from failing, so simple striping across all of the RAID
groups is safe (and makes parity computation cheaper, because there are
fewer individual disks per I/O involved in the computation).

cvs.netbsd.org also uses 2 4-component ccds: one for cvsroot and source
archives, and one for a backups staging area.

The only thing "fragile" about the ccd driver is that it is only striping
(well, and concatenation) ... if you lose a component, you're screwed, but
that's not the fault of the ccd driver, which doesn't even pretend to handle
dying components... it's designed to stripe, and that's it.

Now, as has been noted, we're addressing the ccd's lack of features with
RAIDFrame, which is currently being ported to the NetBSD kernel (and a
lot of progress is being made).  We'll still keep the ccd driver around
for those who want it, but for people looking for a complete RAID solution
(including mirroring and error recovery), RAIDFrame will be their ticket.

Jason R. Thorpe                                       thorpej@nas.nasa.gov
NASA Ames Research Center                            Home: +1 408 866 1912
NAS: M/S 258-5                                       Work: +1 650 604 0935
Moffett Field, CA 94035                             Pager: +1 650 428 6939