RAID6? (was Re: Anyone Actually Using Virtual Linux Servers?)

[Derek Atkins]

Jarod Wilson <jarod-ajLrJawYSntWk0Htik3J/w at public.gmane.org> writes:

> Yes, it requires at least 4 disks instead of 3 disks like RAID5, and your 
> capacity is number of drives minus two. With RAID5, in the 3-disk case, you 
> have two data stripes and one parity stripe for every write. With RAID6 in 
> the 4-disk case, you have two data stripes and two parity stripes for every 
> write. With that in place, RAID6 can survive a double-disk failure of any two 
> disks in the array, whereas with RAID10, certain double-disk failures can 
> still be fatal (and of course in RAID5, they're always fatal).

Yeah, I read that much on wikipedia..

> Personally, I moved from RAID5 to RAID6 after getting bit by a second disk 
> going haywire while a RAID5 array was being rebuilt after replacing a disk 
> that had failed. Went RAID6 over RAID10 primarily since it should survive 
> failing any two disks, and in cursory testing, the throughput was still more 
> than adequate (some have complained about the computational intensity of 
> RAID6, but with a modern multi-core cpu, *shrug*).

Fair enough..  For me I'm pondering this for the case of an 8-core
vmware host server.

>> ISTR a problem in that with RAID5 if 
>> you had a partial disk failure (i.e. it just returned bogus data) that
>> you could corrupt your data because Linux S/W RAID didn't do checksum
>> verification on reads -- so it was "safer" to use RAID10.  Is this
>> "fixed" in raid6?
>
> Hrm. I wasn't aware of any such issue, so I haven't a clue if its fixed or 
> not...

IIRC:  http://www.miracleas.com/BAARF/RAID5_versus_RAID10.txt

> Jarod Wilson
> jarod-ajLrJawYSntWk0Htik3J/w at public.gmane.org

-derek

-- 
       Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
       Member, MIT Student Information Processing Board  (SIPB)
       URL: http://web.mit.edu/warlord/    PP-ASEL-IA     N1NWH
       warlord-DPNOqEs/LNQ at public.gmane.org                        PGP key available

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.