[Techtalk] Journalling file systems.

Mary Gardiner linuxchix at puzzling.org
Sat Oct 27 09:45:46 EST 2001

Hi all,

Well, journalling file systems seem to be all the rage at the moment :)
and I have to say that I would be grateful not to have to sit through an
fsck every time a circuit breaker triggers.

At the moment, I'm using ext2fs and vfat (I'm not intending to drop the
vfat, my Windows partition would be unhappy... :) )

I've been exposed to a three way holy war between xfs users, reiserfs
users and ext3fs users, but not a very informative one.

I'm looking for a good rundown on their comparitive strengths.

I need:
 * Something that works with the 2.4.x kernel series, I don't mind
   applying patches (otherwise the options narrow considerably).
 * Something stable, I'm going to run it on my personal box, and would
   be most upset to lose a partition.

I also use Maildirs, so something with good performance on a large
number of small files would be good, although df tells me that I'll run
out of room on /home long before running out of inodes.

I'm being sold the following strengths of each file system:

* Reiser: the aforementioned performance optimisation for large numbers
  of small files, is part of the 2.4.x kernels and so doesn't require a
* ext3: Can convert existing ext2 partitions without moving all the
  data, formatting, and moving back. Can put the journal file on a
  different partition so that the partition can be read as ext2 by, for
  example, a boot floppy.
* xfs: None so far, although my housemates think it is 'cool'.

* Reiser: Makes for an enormous kernel.[1]
* ext3: Appears somewhat unstable, has caused hard locks for users.
* xfs: Again, none.

As you can see, technical details are scarce. Can anyone give me any
pointers to a good rundown on any or all of these file systems (points
to linux-kernel threads etc most appreciated), or share their
experiences with any of these file systems especially as regards data
integrity, system lockups, and failure? And, wrt ext3, is it a good idea
to run the Linus kernel with patches, or the Alan Cox kernels?

I can find plenty of documentation on installing and using them, I'm
more interested in comparisons.


[1] My SO is very big on some kind of inverse kernel power-trip, that
is, compiling the smallest kernel possible, and is hence responsible for
this 'weakness'.

Mary Gardiner
<mary at puzzling.org>

More information about the Techtalk mailing list