Note: this came from the PH Linux Newbie forum, moved here ...
On Sat, 1 Sep 2001 at 21:40, Rino Mardo wrote:
> i noticed that JFS is not very popular here.
I'd rank the "popularity" of filesystems in PLUG as follows (although this
is unofficial and just my gist of things):
1. ext2
2. ReiserFS
3. ext3
4. XFS
5. JFS
Why? ext2 is on top for obvious reasons. ReiserFS is second because it's
the next most available "default". ext3 I put before XFS because ... I
don't know why actually and have no basis for that. XFS is there maybe
because I keep talking about it. JFS is probably last because nobody talks
about it enough.
Maybe you (Rino) can help change things by sharing with us your
experiences about JFS?
I just checked out the mongo.pl benchmarks comparing JFS and ReiserFS in
the Namesys website. Interesting note on top of the page:
"Notice that mongo single process does not pass cleanly on JFS, on every�
delete phase there are messages like this: rm: cannot remove directory
'/mnt/testdir1-0-0/d0/d1': Directory not empty Which leads to zero values
on deletes."
This is one reason why I never gave JFS a shot.
The benchmarks aren't very interesting, either. While there are no direct
xfs vs jfs benchmarks in the Namesys website, and no equal kernel to
compile the two, JFS is slower than ReiserFS for slow files, and slower
than XFS for larger files. JFS seems to be a decent "middleground"
filesystem, but otherwise doesn't look too attractive. Maybe you have
other more "enlightening" benchmarks to share, as a JFS user?
> i've tried JFS and one thing that made me stick to it is i don't have
> to downgrade my compiler unlike the requirement of XFS.
This is interesting. I've been using gcc 2.95.{3,4} with XFS and have been
running a quite stable system. Lately I've been doing tests with gcc
3.0.2-pre (since Debian doesn't have 3.0.1 packages, only 3.0.1-pre so I'm
checking the later one out), but life hasn't been so promising with this
combination. I'm not giving up, though.
gcc 2.96 is out of the question, I think. While it _may_ work, it's caused
one hell of a raucus since RedHat first released it with RedHat 7.0.
Besides, with gcc-3.0, why use 2.96?
> my experience with JFS?
Maybe you can share performance issues with us? XFS is particularly slow
on deletes (VERY slow), but since mongo.pl tests of JFS fail on deletes,
there's no point of comparison as far as that particular test suite is
concerned.
> smooth. dito sa amin there are frequent brownouts for no apparent
> reason and i still have to see JFS hiccup.
You mean even data is intact? A "famous" test for both ReiserFS and XFS to
show what metadata journalling is not about is to create a text file using
your favorite text editor, save it, then without running sync, turn off
your machine. On XFS you normally get a file with the size that the text
should have been, filled with binary zeroes. As far as journalling is
concerned this is correct, since metadata points to data that is of
correct size, et al. The data is neither journalled nor synchronous, so
the binary zeroes are theoretically correct. What happens with JFS? I'm
just curious since I haven't had the time to check it out.
Also, it seems JFS doesn't have kernel-specific patches. This is
interesting. How exactly do you go about creating a JFS-enabled kernel?
--> Jijo
---
Federico Sevilla III :: [EMAIL PROTECTED]
Network Administrator :: The Leather Collection, Inc.
GnuPG Key: <http://jijo.leathercollection.ph/jijo.gpg>
_
Philippine Linux Users Group. Web site and archives at http://plug.linux.org.ph
To leave: send "unsubscribe" in the body to [EMAIL PROTECTED]
To subscribe to the Linux Newbies' List: send "subscribe" in the body to
[EMAIL PROTECTED]
