On 04/10/13 14:28, Daniel Llewellyn wrote:
> I'm confused.
>
> surely raid-10 with two devices is no different to raid-1?
Indeed. See here for the "wonderful" things you can do with md RAID 10
http://en.wikipedia.org/wiki/Non-standard_RAID_levels#Linux_MD_RAID_10
> I don't get how raid-10 is any different, or maybe that should be "how it's
> possible to be any different", to so-called raid-0+1: at the end of the
> day, you need two copies of the data (RAID-1) but when you have more than
> two devices you stripe the raid-1 mirror onto two sets of raid-0 drives.
You can do md RAID-10 with 3 disks. Try that with RAID 0+1 (or 1+0).
You'll get (at least, depending on configuration) 2 copies of any given
chunk, on different disks, but split across all members of the array.
> The article Keith linked seems to suggest that RAID0+1 is a raid-0 on top
> of raid-1, as opposed to raid-1 on top of two raid-0 arrays, which I can't
> conceptualise.
You can do that, of course, but it's not really fault-tolerant.
> raid-1 appears as a single contiguous block, so how do you
> then lay a striping pattern on top which requires at least two devices?
> Maybe the author thinks you have two raid-1 arrays, thereby requiring at
> least 4 drives paired into two pairs and then stripe between the two raid-1
> devices? this is getting complicated!
That's the normal way to do that, sometimes with more than 2 RAID 1 pairs.
Stuart
--
Stuart Sears RHCA etc.
"It's today!" said Piglet.
"My favourite day," said Pooh.
--
Please post to: Hampshire@???
Web Interface:
https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL:
http://www.hantslug.org.uk
--------------------------------------------------------------
