[MiNT] RAID0/1

[Mark Duckworth <mduckworth@atari-source.com>]

This is a re-send.  I don't think my original went through.  If this is a
dupe, I apologize....

Hey guys,

I have an interest in RAID support under MiNT.  Uwe Seimet claims this
could easily be done as a feature under MiNT and that this would
probably be better than doing it through hd-driver.  Through mint you
could simply just have it replicate every operation to more than one
partition(drive) based on a configuration.  The question of ideals is,
do you want RAID that is transparent to the OS or do you want more of a
RAID filesystem driver.

To do it properly however requires implementation of a volume management
system and a concept of logical volumes.  RAID 0 would involve multiple
drives and partitions to create a single logical volume which would
actually decrease reliability because a failure across an array of
drives would cause the entire volume to fail.  I however am interested
in RAID1 which is straight mirroring across 2 or more drives.  It used
to be that I had os level drive corruption.  I no longer have those
problems and now I am interested in preventing data loss due to physical
drive failures.  And I happen to have a BUNCH of identically sized ide
and scsi hdd's.  I realize that for writes this would make the atari
twice as slow, but for reads it can be optimized to be twice as fast, or
at least 110% faster depending on whether the bottle neck is seeks and
reads in the ide drives or the falcon ide bus.  I assume that in some
time if a PCI adapter comes or whatnot, bus speed will not be the issue
and RAID1 reads woudl actually cause a speed increase.  RAID 5 is much
more complicated and probably not necessary since most of us only have
access to 2 ide devices.

Several people have recommended hardware raid solutions to me but I'm
not too interested in stuff like that as the cost increases
exponentially the more I want to RAID.  I'm a fan of software raid, I
use it under linux and it's really fast for me ;-).  I'm just looking
for an opinion as to this feature, whether it could be done or not, how
hard.. worth it.. etc.  In my opinion any cool feature is worth it, but
then I wouldn't be doing the work and we have more important things to
do like shared lib and vm support.

Thanks,
-- 
Mark Duckworth <mduckworth@racingchannel.com>