[ale] ext3 v reiserfs

Greg runman at speedfactory.net
Sun Sep 14 14:36:27 EDT 2003


from
http://lists.debian.org/debian-user/2002/debian-user-200202/msg00824.html

* ext3 enhances further the ext2 filesystem by introducing the notion of
journalling, but at the same time inherits the current limitations of ext2.
* in intermittently-powered machines, ext3 and ReiserFS scale good compared
to XFS, which has the tendency to develop file holes when used on desktop
PCs. [in the university I attend, we have "kiosks" - low end machines with
ext3 filesystems. despite being intermittently powered-on/off by students,
has been quite resistant to filesystem corruption - though I'm praying that
it won't develop hardware errors due to the level of abuse it gets]
* for machines used as proxy servers or used as cache systems, ReiserFS
is quite good as filesystem transactions are committed faster (faster file
delete/update/create rate).
* XFS performs better than ReiserFS and ext3 when used on almost-full
filesystems.
* On small-sized partitions, ReiserFS has a disadvantage of having a
constant
size journal of 32 MB compared to ext3, but this is quickly-offset as the
maximum size is approached, since ext3 journal data adjusts and grows
according
to the data in the disk (since ext3 journal stores meta data and may also
contain file data).

http://oregonstate.edu/~kveton/fs/ seems to be pretty good

http://lists.slug.org.au/archives/slug/2003/09/msg00121.html showed some
test results also.

all of the above came from googling "reiserfs ext3 comparison"

I have had installs on small disks fail when using reiserfs on Suse 8.2 due
to a minimum partition size that is needed for the reiserfs fs ( I thing
around 400 M or so).

The churning could be due to a time variable that ext3 uses to write the
metadata to the disk.  I know that when I set up my RAID array, I had to
tell the box "not" to check the partitions on boot up since I was using ext3
to avoid just that time intensive task at boot-up.  I think that when you
create a filesystem in ext3 you can set a variable that tells the machine
when and if or not to check the filesystem.  I am not sure if this is just
boot-related or not.  I kind of not understand what good logging is if you
don't make use of it, but you have to write the data to disk at some time,
maybe ext3 does this a whole lot more than reiserfs ???   Perhaps your boxes
are just constantly checking/syncing up ???

Danny Cox is the resident RAID guru and I am sure that he has ton's of
in-depth experience with the idiosyncrasies of filesystems.  Perhaps he
could shed some light on this.

I would be interested in the reason you have noticed this.

Greg

> -----Original Message-----
> From: ale-bounces at ale.org [mailto:ale-bounces at ale.org]On Behalf Of Jim
> Popovitch
> Sent: Sunday, September 14, 2003 12:31 PM
> To: Atlanta Linux Enthusiasts
> Subject: RE: [ale] ext3 v reiserfs
>
>
> On Sun, 2003-09-14 at 12:29, Greg wrote:
>
> > Perhaps if you told the list what the problem/task is that you
> are trying to
> > solve/accomplish better answers might come up, as well as other
> pertinient
> > issues ?
>
> Thanks Greg.  There are no problems.  I've just been noticing that ext3
> must do less caching than reiser as i see/hear my harddrive churn more
> with ext3 partitions.  While not scientific, kernel compiles on a reiser
> partition seem to be improved.  I am using 2.4.22 with ext3 and reiser
> both compiled into the kernel.
>
> -Jim P.
>
> _______________________________________________
> Ale mailing list
> Ale at ale.org
> http://www.ale.org/mailman/listinfo/ale
>



More information about the Ale mailing list