[KLUG Members] Re: RAID

[Bryan J. Smith]

--=-y+QSBkN2loIFe0ZYMKHR
Content-Type: text/plain
Content-Transfer-Encoding: quoted-printable

On Sat, 2002-12-14 at 22:15, Peter Buxton wrote:
> I've got an Asus mobo with an integrated Promise PDC 20265 IDE RAID
> controller.  Linux supports Promise 'FastTrak' ATARAID, which supports
> RAID 0. I have two matching WD400's.

- "True Hardware" ATA RAID has vendor GPL drivers ...

Promise only supports the "SuperTrak" ATA RAID controllers which are
microcontroller-driven ATA RAID (like SCSI RAID) with GPL drivers.  This
is because all the 3rd party IP is in card's firmware, and used in its
on-board microcontroller.  The same is true of all 3Ware Escalade and
the Adaptec 2400A products.  These are "true hardware" ATA RAID,
designed almost exactly like SCSI RAID controllers (except the 3Ware,
which has a slightly different, non-blocking I/O approach that is
ATA-only).

- "Software-driven" ATA RAID has binary-only vendor driver support ...

Promise only releases "binary-only" drivers for their ATA controllers
with 16-bit BIOS RAID setup-only (i.e. they are still main CPU-driven,
so still "software RAID"), because the driver contains licensed, 3rd
party IP.  Your kernel choice will be limited.  High Point Technologies
(HPT) does the same.

Now the Linux community _has_ developed a "generic" ATA RAID GPL driver
model for use with "Software-driven" ATA RAID controllers.  The idea is
that the "ataraid.c" driver contains all the RAID logic (replacing the
licensed, 3rd party IP), and then a vendor-specific driver is written
for the various products.  Such drivers that use the "ataraid.c" core
exist for both Promise "FastTrak" (pdcraid.c) series and HPT controllers
(hptraid.c) with 16-bit BIOS RAID setup-only.  But I don't recommend
using them -- especially with newer models.

> Now: considering I am a home user, I have two big points of failure: one
> HD could melt, or the mobo could. Now, if a drive goes, no big deal (so
> say I now). But if the mobo goes, will I need another Promise card (or
> Promise-sporting mobo)

Both of the above, "true hardware" and "software-driven" ATA RAID
products use vendor-specific disk organization.

So yes, you'll need to get another product from the same vendor that
supports the same on-disk organization.

> or is RAID RAID no matter what chipset/software supports it,

No.  On-disk organization is usually vendor-specific.  _Sometimes_ you
can read RAID-1 disks that are simply mirrors -- but I haven't tried
myself.

Especially when some ATA RAID controllers offer the ability to "remap
sectors" ala SCSI-like features.

> and dedicated hardware support is simply more efficient?

If you mean "software-driven," _no_.  OS software RAID is
_just_as_fast_.  Why?  Because _both_ use your main CPU.

But if you mean "true hardware," it depends.  On-board microcontroller
or ASICs can seriously reduce the overhead and load on your CPU.=20
Furthermore, it can queue I/O requests and cache read/writes.

Of course, there are other considerations.  If you're interested in
reading more, please see an unfinished draft article (never published) I
wrote for Sys Admin magazine here: =20
  http://smithconcepts.com/files/articles/ata_raid.pdf

> In sum, am I better off, at my level of use, using software MD, which
> can mount my drives on *any* motherboard under most Linuces (such as
> Knoppix)? TIA.

Again, compared to 16-bit BIOS RAID setup-only cards, OS software RAID
is _no_different_ from a "performance" standpoint.  And as you correctly
identified, using OS software RAID means you _can_ move the volumes from
vendor controller to vendor controller, _unlike_ RAID volumes that are
organized by vendor card/controller BIOS/firmware.

Me?  Unless I go with a 3Ware Escalade ATA RAID product, let alone SCSI
RAID, I use software RAID.

-- Bryan

--=20
Bryan J. Smith, E.I. (BSECE)       Contact Info:  http://thebs.org
[ http://thebs.org/files/resume/BryanJonSmith_certifications.pdf ]
------------------------------------------------------------------
  The more government chooses for you, the less freedom you have.


--=-y+QSBkN2loIFe0ZYMKHR
Content-Type: application/pgp-signature; name=signature.asc
Content-Description: This is a digitally signed message part

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (GNU/Linux)

iD8DBQA9+/2pDjEszaVrzmQRAjqXAJ9iar9Nlx+qxuA62DRDT4rqA0QM/QCgqDv0
GyLOBGEergGD6hSNEOfx9Ak=
=G35+
-----END PGP SIGNATURE-----

--=-y+QSBkN2loIFe0ZYMKHR--