I read this a while back it's about as clear as it's going to get, though has dated very slightly.
Brett
Jeff Waugh wrote:
<quote who="Mary">
On Tue, Jul 29, 2003, Jeff Waugh wrote:
But you're better off choosing ext3, jfs or xfs over reiserfs. :-)
C'mon, back your assertions, it makes world domination easier you know!
:-) Lots of rehashing here, but for the benefit of the list:
Okay, so, reiserfs has no recovery tools. None. If something goes wrong, whammo, you're potentially toast, eggs and bacon. It doesn't use inodes internally, so if you're running an NFS server on top of it, there's a translation layer in between. Slow, and not worth the indirection. It doesn't scale particularly well with SMP. It's a metadata-only journalling filesystem, so you're not protecting the integrity of the data itself, just the description of the data. It has had a number of extents-related issues in the past, writing over files and data that it should not have. Personally, I would not use reiserfs in a production environment, though I do use it for /tmp, for cvs checkouts and for big build trees.
XFS is a long-standing filesystem that has been used on OS/2 and IRIX. It is especially good for high throughput applications, such as media work (which is not surprising given SGI's market). Metadata only journalling, scales incredibly well with multiple CPUs (even under 2.4) and includes POSIX ACLs (even under 2.4), which are kind of cool if you're using recent versions of SAMBA and serving up to Windows PCs. XFS also supports a realtime partition type, which is designed to guarantee very high throughput rates for the most demanding applications (though it will be a while before this is fully supported in Linux).
On the other hand, ext3 is a relatively slow filesystem which is on-disk compatible with ext2, with optional full data journalling (which in some cases actually makes it faster; mail queues are a good example). You can upgrade to ext3 from ext2 without any hassles. There are lots of improvements to ext2/3 all the time, such as Daniel Phillips' htree patch which improves directory indexing performance. Because "basically everyone" uses these filesystems, you can rely on them as the most heavily tested and most likely to be improved filesystems available for Linux.
- Jeff
-- Brett Fenton General Manager NetRegistry Pty Ltd _______________________________________________
http://www.netregistry.com.au/
Tel: +61 2 96996099 | Fax: +61 2 96996088 PO Box 270 Broadway | NSW 2007, Australia
Your Total Internet Business Services Provider Trusted by 10,000s of Oz Businesses Since 1997
This email is from NetRegistry Pty Ltd. The contents of this message are commercial and in confidence to the intended addresseee.
The message may contain copyrighted and/or legally priviledged information. No person or entity other than the intended recipient may read, print or store this message, including any and all attached files.
The intended recipient may not forward this message to any third party without express written permission from NetRegistry Pty Ltd.
-- SLUG - Sydney Linux User's Group - http://slug.org.au/ More Info: http://lists.slug.org.au/listinfo/slug
