Date: Sat, 19 Nov 2005 16:41:47 -0800 (PST) From: Doug White <dwhite@gumbysoft.com> To: Alexander Leidinger <Alexander@Leidinger.net> Cc: delphij@delphij.net, Poul-Henning Kamp <phk@phk.freebsd.dk>, freebsd-current@FreeBSD.org, Robert Watson <rwatson@FreeBSD.org>, Brian Candler <B.Candler@pobox.com> Subject: Re: Logical volume management Message-ID: <20051119163635.E88861@carver.gumbysoft.com> In-Reply-To: <20051119165631.093b90b9@Magellan.Leidinger.net> References: <20051119162854.2656096a@Magellan.Leidinger.net> <13727.1132414598@critter.freebsd.dk> <20051119165631.093b90b9@Magellan.Leidinger.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 19 Nov 2005, Alexander Leidinger wrote:
> You can define more than one resource pool with ZFS, and as I did
> understand it, the pools are distinct. So if you assign disc1 and disc2
> to poolA and disc3 and disc4 to poolB, and you assign just one FS to
> each pool, you still have the same separation like in vinum or
> g{mirror,raid3,stripe}. You just have a more complex distribution
> algorithm.
However the ZMU layer doesn't allow you to stack transforms, so you can't
do RAID10 by stacking a stripe transform on a mirror. Score a point for
GEOM and the oldie-but-goodie SDS. :)
The ZFS code and architecture docs imply that the volume-management
aspects and the filesystem are abstracted from one another and are bundled
only for administrative simplicity. If they wanted to, the ZMU could have
been a separate component and you could have layered UFS on top of it.
Most of the code is available through OpenSolaris if you want to see how
its done.
--
Doug White | FreeBSD: The Power to Serve
dwhite@gumbysoft.com | www.FreeBSD.org
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20051119163635.E88861>