Thanks for the detailed information. Starting to make sense.
Bogdan
--- "Bryan J. Smith" b.j.smith@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@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@centos.org http://lists.centos.org/mailman/listinfo/centos
-
Bogdan Nicolescu