Date: Sat, 18 Aug 2007 09:11:14 +0100 From: Matthew Seaman <m.seaman@infracaninophile.co.uk> To: Clayton Milos <clay@milos.co.za> Cc: Vivek Khera <vivek@khera.org>, FreeBSD Stable <freebsd-stable@freebsd.org> Subject: Re: large RAID volume partition strategy Message-ID: <46C6A9A2.70209@infracaninophile.co.uk> In-Reply-To: <000f01c7e122$1d91a390$0301a8c0@claylaptop> References: <31BB09D7-B58A-47AC-8DD1-6BB8141170D8@khera.org> <b41c75520708171510i38594117sc55dfabf06ea302@mail.gmail.com> <000f01c7e122$1d91a390$0301a8c0@claylaptop>
next in thread | previous in thread | raw e-mail | index | archive | help
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256
Clayton Milos wrote:
> If you want awesome performance and reliability the real way to go is
> RAID10 (or more correctly RAID 0+1).
RAID10 and RAID0+1 are very different beasts. RAID10 is the best
choice for a read/write intensive f/s with valuable data, exactly
what you need to support a RDBMS. It is built by pairing up all of
the drives as RAID1 mirrors[*] and then creating a RAID0 stripe
across all of the mirrors. It's the least economical RAID setup,
giving you a usable space which is 50% of the total raw disk space,
but it is the most resilient -- potentially being able to survive
half of the drives failing -- and much the best performing of the
RAID types.
RAID0+1 on the other hand is what you give to someone you don't like
very much. In this case, you divide the disks into two equal sets,
create a RAID0 stripe over each set and then a RAID1 mirror over the
stripes. It has the /delightful/ feature that failure of any one
drive immediately puts half of the available disks out of action: ie
it is *less* resilient than any other RAID setup (other than a RAID0
stripe over all the drives). Space economy-wise it's exactly like
RAID10 and performance characteristics are pretty similar to RAID10,
leading to the obvious conclusion: use RAID10 instead.
Cheers,
Matthew
[*] The correctly paranoid sysadmin will of course ensure that each
of the disks in those pairs hangs off a different bus, comes from a
different manufacturing batch and is preferably connected to a
different controller and with different, independent power supplies.
Or, in extreme cases, that each half of the mirrors are in
completely different datacenters.
- --
Dr Matthew J Seaman MA, D.Phil. 7 Priory Courtyard
Flat 3
PGP: http://www.infracaninophile.co.uk/pgpkey Ramsgate
Kent, CT11 9PW
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.4 (FreeBSD)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
iD8DBQFGxqmh8Mjk52CukIwRCPvyAJ4k/POTK9Moqu80nV9TKHZqLIC5ngCfYEd4
oiV2MAAiFIXcNSTSiCM4D6M=
=GDZN
-----END PGP SIGNATURE-----
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?46C6A9A2.70209>