Re: SCSI and EIDE
> Ricardo Kleemann <ricardo@americasnet.com> writes:
> > That's wonderful!
> >
> > Now will Linux implement anything greater than RAID0?
It does. Raid1 is in development. It does partial mirroring using about 1/3rd
of the disk space for 'backup' data.
> > Would you say your performance is significantly increased with striping?
I have an ncr53c815 card with two 312mb XT 1986 scsi-1 SLOW maxtor drives.
They definately perform much better after being raid0'ed. If there would
be a point at which the data read from a drive was faster than the cpu could
crunch the raid0 'virtual' device, it would be slower than no raid at all. I
personally don't see how any setup could be this 'disk fast' and 'cpu slow'.
>From my experience it is significantly faster. It is literally (for me)
doubling the number of read-heads that can be reading or writing data at a
given time.
> > How many drives can be striped?
You can have 4 raid devices. /dev/md0 - /dev/md3. I am not aware of a
limitation on the number of 'block' devices that can be grouped under each
raid device. I would also be surprised if you were not able to increase the
numbers if you needed to. Reminds me of ppp/slip in it's early stages. A
default limit of 4. Then they raised it to 16, and finally to
on-the-fly-creation of devices up to a default limitation of 256, with an
easy #define to change to increase this limitation. I see no reason why the
raid devices will not follow this pattern.
> md just groups a number of physical disk partitions into one logical one,
> /dev/md*.
I'm being picky, but no, it groups block devices. I couldn't get it to work
with a loopback device, so perhaps right now block device = hard drive. But
it shouldn't have to.
One happy raid0 user.
--
Todd Fries .. todd@miango.com
Reply to: