Austin Murphy on 16 Oct 2006 19:10:25 -0000


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: [PLUG] RAID cards


On 10/16/06, gabriel rosenkoetter <gr@eclipsed.net> wrote:
On Mon, Oct 16, 2006 at 10:50:30AM -0400, Austin Murphy wrote:
> By standardizing the disk data format, you don't need to have crummy
> BLOBs loaded into your kernel because kernel drivers can be written to
> the standard.  With open kernel drivers and a fast CPU, the fakeraid
> setup could outperform the 3ware setup for less money.
>
> What's wrong with that?

One of us is confused about how the fakeraid cards in question, and
I'm pretty sure it's you. How does having a faster general-purpose
CPU speed up the operations of the embedded circuitry on the card
where the software RAID implementation that's pretending to be
hardware is actually running?

I think you are confused about what is getting off-loaded from the "fakeraid" card. In a "fakeraid" setup, the RAID logic occurs on the system CPU. The "fakeraid" card only has an XOR accelerator to speed up the parity computation. Documentation is sketchy for the most part, but I found some interesting info here: http://tweakers.net/reviews/557/1

My point was that a "fakeraid" setup could *possibly* outperform a
"trueraid" setup for the same reasons that a software RAID setup can:
If you are bottlenecking on CPU, a good main CPU can outperform a
cheap embedded CPU.

In the case of actually using software RAID in the Linux kernel...
the data layout is controlled by the kernel, and it reallyd doesn't
matter how the disk is attached as long as it's the same disk and
you tell the volume manager about it. That is, using a software
volume manager to provide RAID services already resolves this data
layout problem.

These software volume managers are incompatible with each other and incompatible with hardware initiated layouts. Why shouldn't any DDF 1.2 compliant RAID controller (Hardware, fake, or software) be able to read and write any DDF 1.2 compliant RAID set?

This seems like a pretty reasonable goal...

Austin
___________________________________________________________________________
Philadelphia Linux Users Group         --        http://www.phillylinux.org
Announcements - http://lists.phillylinux.org/mailman/listinfo/plug-announce
General Discussion  --   http://lists.phillylinux.org/mailman/listinfo/plug