[CentOS] Installing CentOS on RAID PATA not SATA - A8V
Bogdan Nicolescu
bo2k2 at yahoo.com
Thu May 19 13:26:40 UTC 2005
Thanks for the detailed information. Starting to make
sense.
Bogdan
--- "Bryan J. Smith" <b.j.smith at ieee.org> wrote:
> On Wed, 2005-05-18 at 20:23 -0700, Bogdan Nicolescu
> wrote:
> > My first attempt at RAID...
>
> I purposely call it Fake/Free RAID (FRAID) for a
> reason.
> It is a _regular_ ATA controller with *0* extra
> hardware.
> There is a "trick" 16-bit BIOS that can setup some
> RAID
> organization.
>
> It is _useless_ once a 32/64-bit OS boots.
>
> > Tried a different distribution with the latest
> kernel,
> > and indeed the PTA drives on the Promise
> controller
> > were recognized but they were recognized as two
> > separate drives while the controller was in RAID
> mode
> > (rather than IDE mode).
>
> Correct. Because the system sees the "raw" ATA
> channels.
> There must be a "trick" 32/64-bit driver to match
> the
> "trick" 16-bit BIOS.
>
> This "trick" 32/64-bit driver contains _all_ of the
> RAID functionality. *0* is in the hardware, because
> the
> hardware has *0* RAID firmware other than the
> ultra-simple
> stripe/mirror access when the 16-bit BIOS is
> booting.
>
> About the same 2-3 companies write these drivers,
> and
> _all_ vendors license that code. Because it is the
> core
> IP of those companies, they will _never_ release it.
> As such, the drivers are _always_ "closed source"
> and
> really only available for Windows. The few
> companies that
> do release Linux drivers release kernel
> version-specific
> ones that are binary-only, no source code.
>
> There is a "clean-room" GPL ATA RAID driver in
> ataraid.c.
> And then interface drivers in hptraid.c, pdcraid.c
> and
> silraid.c, but they _rarely_ work with most "FRAID"
> cards accept for very old ones. And they are
> _slower_
> than the OS' RAID (NT LDM, Linux LVM/MD, etc...).
>
> > Are you saying that even if the drives are
> recognized,
> > I will not be able to implement RAID in hardware
> mode?
>
> There is _no_ "hardware" mode. The system _always_
> sees
> the drives. There must be a "trick" driver to
> "hide" the
> drives and make them seem organized.
>
> A _true_ hardware RAID controller _always_ has
> on-board
> intelligence that directly controls the drives, and
> has
> the RAID firmware on-board. In those cases, the
> system
> talks to that on-board intelligence (e.g.,
> microcontroller
> or ASIC) and _never_ actually sees the drives.
> Because
> all of the RAID intelligence is in the firmware, the
> OS
> driver is just a basic block driver, so they are
> GPL.
>
> Examples of true hardware ATA RAID include Promise
> SuperTrak
> (_not_ the FastTrak), Adaptec 2400A/2800A, some (but
> not
> all) LSI Logic MegaRAID and 3Ware Escalade products.
>
> These cards cost between $125-$500+. You will _not_
> find a _true_ hardware RAID controller on a
> mainboard. ;->
>
> > Any other options installing CenOS 4 on
> unsupported hd
> > controller besides:
> > a. install CentOS 4 while drive in recognized hd
> > controller
> > b. get new kernel
> > c. compile/install new kernel (without pulling my
> hair
> > out)
> > d. move drives to unsupported hd controller
> > e. set controller to IDE mode
> > f. boot using new kernel (while praying)
> > g. do software RAID
>
> Yes, software RAID is typically _faster_ than these
> FRAID cards.
> Why? Because instead of using the logic of the
> FRAID controller,
> you get the logic of the OS' buffering. The OS'
> buffering will
> commit writes and interleave reads far better than a
> FRAID
> controller's "trick" driver logic.
>
>
> --
> Bryan J. Smith
> b.j.smith at ieee.org
>
-----------------------------------------------------------------
>
> Beware of those who define their preference in terms
> of hate of
> another option, and not on the positive merits of
> their selection
>
>
> _______________________________________________
> CentOS mailing list
> CentOS at centos.org
> http://lists.centos.org/mailman/listinfo/centos
>
More information about the CentOS
mailing list