Date: Thu, 26 Jul 2007 08:47:13 +0100 (BST) From: "Mark Powell" <M.S.Powell@salford.ac.uk> To: Doug Rabson <dfr@rabson.org> Cc: freebsd-fs@freebsd.org, Pawel Jakub Dawidek <pjd@freebsd.org> Subject: Re: ZfS & GEOM with many odd drive sizes Message-ID: <20070726083224.O68220@rust.salford.ac.uk> In-Reply-To: <1185434973.3698.18.camel@herring.rabson.org> References: <20070725174715.9F47E5B3B@mail.bitblocks.com> <1185389856.3698.11.camel@herring.rabson.org> <20070726075607.W68220@rust.salford.ac.uk> <1185434973.3698.18.camel@herring.rabson.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 26 Jul 2007, Doug Rabson wrote: > When its reading, RAID-Z only has to read the blocks which contain data > - the parity block is only read if either the vdev is in degraded mode > after a drive failure or one (two for RAID-Z2) of the data block reads > fails. Yes, but that article does not mention reading parity. What it's saying is that every block is striped across multiple drives. The checksum for that block thus applies to data which is on multiple drives. Therefore to checksum a block you have to read all the parts of the block from every drive except one in the RAIDz array: "This makes read performance of a RAID-Z pool be the same as that of a single disk, even if you only needed a small read from block D." > For pools which contain a single RAID-Z or RAID-Z2 group, this is > probably a performance issue. Larger pools containing multiple RAID-Z > groups can spread the load to improve this. This isn't something that's immediately obvious, coming from fixed stripe size raid5. Now it seems that the variable stripe size has a rather serious performance penalty. It seems that if you have 8 drives, it'd be much more prudent to make two RAIDz of 3+1 rather than one of 6+2. Cheers. -- Mark Powell - UNIX System Administrator - The University of Salford Information Services Division, Clifford Whitworth Building, Salford University, Manchester, M5 4WT, UK. Tel: +44 161 295 4837 Fax: +44 161 295 5888 www.pgp.com for PGP key
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20070726083224.O68220>