Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> Re: Veritas's implementation/interpretation of RAID 0+1 and RAID 10
In comp.sys.sun.admin Darren Dunham <ddunham_at_redwood.taos.com> wrote:
> In comp.unix.solaris NetComrade <andreyNSPAM_at_bookexchange.net> wrote:
> mirror-stripe is the classic (easy) mirror of stripes.. stripe-mirror
> is the layered stripe of mirrors. It looks like that page is fine
> except for the 1+0 and 0+1 headings. I don't like them anyway because
> they're too easy to get backwards.
Obviously true, given this thread. :-)
>> The way it's supposed to be: >> >> RAID 0+1 >> http://www.acnc.com/04_01_0p1.html
Nope. Consider a 4 disk RAID set; using RAID-5, the system can deal with a single disk failure without losing data, but losing a second disk means losing everything.
> Why would the applications between 1+0 and 0+1 be different?
Define '+' as mirroring, and '*' as striping:
(a+b) * (c+d) is a stripe on mirrored disks, or RAID-1,0.
If disk A fails, the set can also deal with a failure of either disk C or D and stay up. Only if B fails will it lose data.
(a*b) + (c*d) is a mirror consisting of two stripes, or RAID-0,1.
If disk A fails, the set cannot lose either disk C or disk D, but a failure of B won't take down the system. To sum up:
RAID-5: 100% chance of data loss if two disks fail. RAID-1,0: 33% " RAID-0,1: 66% "
Note that this is assuming the RAID-0,1 system does not understand that A ~= C and B ~= D. DiskSuite is supposed to be smart enough so it will use B in place of D in order to keep the filesystem available, thus giving it the same failure probabilities as the 1,0 case.
-Chuck
Chuck Swiger | chuck_at_codefab.com | All your packets are belong to us. -------------+-------------------+----------------------------------- "The human race's favorite method for being in control of the facts is to ignore them." -Celia GreenReceived on Tue Aug 20 2002 - 17:09:11 CDT