[WBEL-users] Intel ICH5R SATA RAID

[Tarjei Knapstad]

On Wed, 2004-10-20 at 18:29, Bob Ramstad wrote:
> I am building a new WBEL3 system which has a motherboard with the
> Intel ICH5R SATA RAID chipset.
> 
> >From what I can tell, this is "fake raid" i.e. RAID1 is accomplished
> through software drivers.
> 
> http://users.pandora.be/TheBlackUnicorn/linux/
> 
> It looks like the appropriate drivers are available in 2.4.26.
> 
> >From my end, it looks like respin1 treats the SATA drives as if they
> were SCSI.  Even if I have them combined in the ICH5R setup as if they
> are RAID1, they show up as two separate SCSI drives.
> 
> I have a few questions, wondering if anyone else has any experience with this...
> 
> 1) Given that the ICH5R RAID1 is software RAID, should I expect
> similar performance using Linux software RAID i.e. /dev/mdX?
> 
> 2) If so, any reason to prefer ICH5R RAID1 or Linux RAID1 over the other?
> 
> 3) If ICH5R RAID1 is preferred, any simple strategies for installing
> WBEL other than waiting for respin2 or installing respin1 to a
> traditional IDE drive first, upgrading the kernel and copying over?
> 
> My inclination is to leave the two drives as separate SATA drives, let
> them be recognized as SCSI drives by respin1, and use Linux software
> RAID.  I do have a slight concern / fear that this is suboptimal, or
> won't work, hence my questions.
> 
> Thanks in advance for any help.
> 

Hi Bob,

I had the exact same problem when I wanted to install FC1 on a partition
on a RAID-0 array I had set up on the ICH5R. I didn't know at the time
that ICH5R was indeed a software RAID solution, but I eventually gave up
and bought a separate drive for linux (my raid array already had lots of
Windows stuff set up).

Anyway, from what I've read on the Fedora lists later on there's not
much of a performance gap (if any) between ICH5R "native" RAID and using
the linux software RAID, which IIRC was the suggested solution from the
Fedora community. You might find it useful to search the various Fedora
mailing list archives.

The best solution is probably just to try it out though. You could
compare drive benchmark results for a linux software raid to for
instance the ones in this review at HardOCP:

http://www.hardocp.com/article.html?art=NDg0LDM=

(see the "IDE/ATA performance" bit)

Best of luck! It would be interesting if you could post your findings if
you decide to compare linux software raid to intels software raid on
windows.

--
Tarjei