Journaling file systems used to be an oddity primarily for research purposes, but today it's the default in Linux. Discover the ideas behind journaling file systems, and learn how they provide better integrity in the face of a power failure or system crash. Learn about the various journaling file systems in use today, and peek into the next generation of journaling file systems.
Post a Comment
Member since:
2005-07-06
I find that a pretty daft thing to write. ReiserFS is still used by a great many people because it still does what it was designed to do very well, and there were more people involved with ReiserFS than Hans.
Its usage has diminished mainly because the focus of development was switched from maintenance of Reiser 3 to a new and completely incompatible filesystem in Reiser4 that people weren't just going to reformat and move to overnight. While I still expect lots of useful things to come out of Reiser4, it serves as a bit of a warning to people who think they can just breeze in with a totally new and uber cool filesystem and expect it to be widely adopted.
Additionally, I don't see widespread adoption or support of JFS at all. You usually have to jump through some pretty reasonable hoops to get JFS on most distributions, and XFS holds more confidence for more people in most use cases for such a filesystem.
Edited 2008-06-12 21:52 UTC
Member since:2005-07-06
I find that a pretty daft thing to write. "
I quite agree, it’s not like he loses his copyright because he is imprisoned…
Member since:2006-01-06
Yeah, but if you're the kind of person who refuses to use software from anybody you consider "evil", you're gonna need to add ReiserFS to your list. I mean, if you're gonna lump Microsoft into that category, you have to do the same for Reiser ... right? ;-p
Member since:2005-07-06
Blah, blah, blah, blah, blah, blah blah. Everybody hates Microsoft, there's some conspiracy to hate Microsoft, everybody does everything because they hate Microsoft etc. etc. This complex does the rounds on a lot of MSDN blogs I've seen, and it's just plain sad. Whatever the whys and wherefores of Hans Reiser, the code lives on regardless. Microsoft has a track record of being derogatory about open source software at every turn, so no, many people aren't just going to start liking Microsoft until they throw them a serious bone.
Either deal with that, or get yourself a wigwam outside of Redmond and go and live there.
Edited 2008-06-13 14:31 UTC
Member since:2006-02-05
This is getting REALLY off topic, but it was such a stunning post I couldn't help but comment
Let me get this straight. Your definition of "Evil" is being "derogatory about open source software"? That really blows my mind...
http://www.mono-project.com/Moonlight
They paid for top quality HD codecs (WAY beyond theora, which is the best out there in the oss world), said they wouldn't sue over use of technology they own, and helped out with development.
http://www.microsoft.com/interop/
The big move inside the company to play well with its competitors, which we have already started seeing the results of
http://port25.technet.com/
An effort to help integrate linux into MS based environments
http://www.microsoft.com/opensource/community.mspx
A list of open source businesses that Microsoft has partnered with (MSCP means you get pretty much all microsoft software for free or next to nothing)
http://www.codeplex.com/
Microsofts hosted source code/community site for open source projects
http://www.microsoft.com/interop/osp/default.mspx
Microsofts promise not to sue over use of a whole bunch of protocols and formats they own patents on
Member since:2005-09-14
That doesn't really seem to be the point they were trying to make with that sentence though, all they were saying was that its continued development/enhancement are in a state of flux, since Hans is in fact the principal author/designer, and the fate of Namesys if he does wind up in prison remains to be seen.
Edited 2008-06-13 03:59 UTC
Member since:2006-02-05
It isn't daft, Namesys has pretty much fallen to pieces, and no real work is being done on reiserFS anymore.
Member since:
2006-10-08
Interesting article (allthough it does not go very deep into technology, but introduces a starting point), especially the section "Variations on journaling". It explains how data and metadata is handled for different modes to avoid inconsistencies. Remarkable: The more security, the less performance, as it seems. It has been mentioned that is it not always clear when writes from journal take place, and how metadata is handled in relation to the corresponding data blocks.
This is a difference to performant file systems like the Fast File System (FFS, also known as UFS) used by the BSDs. There, the metadata is written asynchronously in a definite order, so the file system's "on disk status" is always consistent, which may not the case in the mentioned journaling file systems under certain circumstances.
The article mentions that journaling file systems have the advantage that it's not neccessary to wait for the fsck utility at startup when a file system problem is recognized, for example after an unclean system shutdown. BSD's fsck utility is able to run in the background while the system is booting up and running, so there's no need to wait for it to complete. This is not a "journaling FS feature" only.
And I still know XFS from my good old SGI beauties... =^_^=
Member since:
2006-01-19
Does anyone know if Linux / XFS can read Irix disks, or if Linux / JFS can read AIX disks?
Even if the filesystems are on-disk compatible, I imagine there are endianness and incompatible logical volume management issues.
In the days of yore I thought that JFS2 was only in AIX 5 and JFS1 was in AIX 4, Linux, and OS/2, but the Wikipedia article says otherwise.
Member since:2007-06-22
Even if the filesystems are on-disk compatible, I imagine there are endianness and incompatible logical volume management issues.
This may answer your question:
http://oss.sgi.com/projects/xfs/faq.html#useirixxfs
Member since:2006-01-19
Member since:
2007-05-12
They're just incredibly robust and reliable.
It's great to boot my Linux box in the evening and be
*absolutely certain* that it will boot just fine. Sheesh, the number of times that Win XP blue-screened on me in the far-distant past when I used it...
Ok, it's true that OS robustness is due to more than just the file-system, but a good journalling FS sure helps... 
FreeBSD is interesting too, with its "soft-updates" in its file-system. That's ultra-robust as well.
Member since:
2007-07-27
Ive heard that most systems work fine during light load and easy tasks. Even WinXP works fine then. The problem is when you scale up things. Will it still work, under demanding circumstances? Will WinXP work with 99.999% uptime? Will Linux filesystems work with Petabytes of storage, and will it work fast and reliable? And how about fsck petabyte filesystems? Will it take the entire weekend (yes it will)? Etc.
It seems that only ZFS is able to cope with all these issues.
For instance, a new 1TB drive has 20% of its space allocated for error correction. There are so much errors in large drives that it isnt funny. ZFS makes the assumption that the drives, IO cards, etc WILL fail. And ZFS copes with these errors. It is the only one detecting silent corruption (and correcting them). All this issues new drives faces and more, are discussed here:
http://www.acmqueue.org/modules.php?name=Content&pa=showpage&am...
Member since:2006-02-05
There are always trade offs for filesystems. Some work best with large amounts of small files, others work best with constant reads to big files. Some are geared towards the desktop (including stuff like rich metadata, or moving system files to the middle of the platters for quick access), others are geared towards servers (like ZFS, which you just mentioned)
Member since:
2008-03-18
I still think of journalling as an oddity.
While journalling writes the (meta-)data
twice to the disc, softupdates just change
the order of writes, therefore also elimi-
nating the need to run fsck (except to re-
free storage marked as allocated).
That said, our (the BSDs') fsck still sucks.
Another thing is Log-Structured Filesystems
like 4.4BSD LFS. It has a very interesting
concept, but the current implementation lets
people not believe in it and is buggy (while
it worked in NetBSD®, only up to 60% of disc
space can be used cleanly).
Reiserfs looks, to me, more similar to LFS
than to a journalled filesystem. I agree to
keep an eye on it, as it's interesting too.
