“One major feature present in Fedora 9 will be the ext4 implementation. The new filesystem will not be the default for the distribution, but will be available for users and systems administrators to enable. New functionality includes larger capacities and online defragmentation, for better performance and more reliability. To find out more, we talked with Eric Sandeen, Fedora project member and filesystem developer at Red Hat.”
I wonder if other big distributions are perhaps also planning to offer ext4 as an alternative file system, maybe already in the installer, in their upcoming releases? Ubuntu? Debian? Gentoo? openSUSE? Mandriva?
Online defrag… selecting just one directory or a few files to defrag… It sounds too good to be true 😉 But those wizards will do it, for sure. I’m very happy with ext*. Most of my data is living in an ext world.
The btrfs filesystem memtioned in the interview is still in heavy development, although it looks like it will have interesting features, like online fscheck. I would like that for ext4.
http://oss.oracle.com/projects/btrfs/
The btrfs will be the next killer fs for Linux. What reiser4 promised, will be delivered by btrfs.
We only have to be patient 😉
Regards Harry
I respectfully disagree. The btrfs feature set was carefully selected to include features that are actually useful in the real world, without all the pie-in-the-sky crap that Reiser4 promised, and without all the VFS polution that it delivered.
It is a feature set that is ambitious enough to be exciting, but not so overly-ambitious as to be unattainable in a reasonable time frame.
And unlike Hans, Chris seems like a good team player, who is interested in working *with* the greater community. And that bodes well for his exciting new filesystem.
As the ext3 vs reiser4 faked benchmark fiasco, and his general demeanor so clearly demonstrated, Hans could not be trusted, nor could he be relied upon to act with the interests of the community at heart.
I’m and ext3/4 fan. But I look forward to a usable btrfs with great anticipation.
Edited 2008-03-10 20:09 UTC
I didn’t ment it on a functional level, but on a more abstract level.
But at least we agree on the fact that btrfs looks very promising 🙂
btrfs looks promising to me in security aspect:)
http://en.wikipedia.org/wiki/BTR-80
I don’t see any mention in that link of CRC checking the passengers. At the end of the run, how do I know that I have the same passengers I started out with? How do I know that one of them was not maliciously replaced in transit? How do I know that when I go to check the passenger upon arrival, I won’t just find garbage… the remains of a cheetohs bag? Or, worse yet, what if one got nulled out by a particularly nasty bump along the way?
Oh fsck! 😉
True, I think the focus should be on ‘real world’ features that are useful rather than great engineering feats but result in nothing beneficial when the rubber hits the road. With that being said, there still needs to be more work done on responsiveness. I have Fedora 8, for example, and when I try to run something heavy CPU utilisation I find that the responsiveness when browsing and downloading websites slows dramatically.
Small things like that need to be fixed. Yes, I know it isn’t explicitly related to the ext4fs, but at the same time, this has to do with the over all need to refocus by Linux developers on what affects the end users on a regular basis.
Wait a minute, isn’t ReiserFS also a real ‘killer’ FS?
Oh sorry, that should have been ‘alleged killer’ FS.
You know, sys admins are also users… why make the distinction…
You know, sys admins are also users… why make the distinction…
Because users aren’t (all) also sysadmins. It’s like saying “squares are all rectangles, so why make the distinction?” Just because squares are all rectangles doesn’t make them the same thing.
Edited 2008-03-10 23:17 UTC
Call me silly… but after all this time, i’m just SO used to saying “ext3”.. It’s imprinted in my brain.. I don’t know if i can break the habit now with 4
I suspect it won’t happen, but it would be cute if ext4 support could be backported to Fedora 8 as well so that users could experiment with it without having to upgrade to Fedora 9 first. The article suggests it might be possible depending on what kernels F8 gets in the future, but it would be preferable to have a definitive statement on the subject.
I guess the problem I have is that I have several distros installed on my PC and *none* of them support ext4 at the moment – so it’s either Fedora 9 + ext4 (and I won’t be able to easily read/write the partition in the other OS’es if it did that) or skip using ext4 for a while until I eventually ditch the OS’es that don’t support ext4…
Edited 2008-03-10 22:54 UTC
Doubt it.
Don’t forget that adding ext4 to F8 will also require a long list of package updates: e2fsprogs, system-config-lvm, util-linux, initscripts, parted (and friends), etc.
In general, once FX+1 is released, most of the development resources are put into the current (FX+1) and the next (rawhide) version while the previous version (FX) enters an unofficial maintenance mode. (A rather reasonable decision – people who rather use an older version [F7] – do it because they want a stable/aged release and rather not test new features.)
– Gilboa
Given that Linux has high-end production-quality filesystems like XFS (my personal favorite) and JFS available, why are we still mucking about with trying to bringing ext up to date. Yes, ext4 is a bigger change than ext3 was, but still… We have better filesystems already.
Many people disagree and seem to have had problems with XFS according to comments online. (I can’t find and remember the exact web links just now but a search for “problems with XFS” brings lots of results in Google. Of course, the same probably with other file systems too, however…:P)
New journaling versions of the ext file systems (3 & 4) are meant to be reliable. And rock solid reliability is by far the most important feature a file system can have IMHO.
I don’t know what exactly would be the major benefits of xfs over ext3 or ext4? As far as I can remember to have read XFS failure handling could be improved, the file system cannot be shrunk, some operations can be slow etc.
I’ve seen many people hype the supposedly much better performance of XFS (or ReiserFS, Reiser4 and other file systems) compared to old ext-based file systems. But according to real tests there seems not to be huge performance differences between the journaling file systems in Linux (ext, reiserfs, xfs jfs). In some tests some file system may be slightly faster and in other tests the others may have a small benefit.
Personally I would trade a debatable tiny amount of extra file system speed for greater reliability anytime. Anyway, a lot of the talk concerning file systems seems to be based on subjective feelings mostly.
When push comes to shove, XFS is just not that great. It’s an early 90’s filesystem that carries a seemingly unshakable mystique because Silicon Graphics originally created it. But it’s old, not a good fit for Linux, and of the Linux filesystems is the most likely to trash your data after any sort of unexpected shutdown. It’s pretty fast if you are streaming large data files to or from a very fast raid array. That’s about it. Otherwise, you might as well use a more reliable filesystem.
JFS is pretty good, but has been completely abandoned. My understanding is that it has no maintainer at all these days.
Ext4 is a necessary successor to ext3. And I’m hoping that btrfs will be the next logical step. Let’s try to move forward through the 21st century and not backwards into the 20th.
Before all the insanity around Hans, my money was on ReiserFS as being the next-gen linux filesystem.
Such a shame.
By “insanity” do you mean his inability to work with the kernel devs, or the… err… legal complications which came later?
Reiser4, with its redundant, DIY, vfs layer, was a poorer fit for Linux than was XFS. Add to that, Namesys’ stock in trade was to develop fast, release shoddy, make up the rest with hype, and abandon the code once it was accepted into mainline. And a lot of folks bought into the hype because they wanted to. Especially the Reiser4 hype.
Give me a tried and true, if a bit stodgy, ext3/4 for a while longer. And then, if all goes well, btrfs looks very interesting.
Edited 2008-03-11 01:44 UTC
Yeah, I was referring to the almost surreal story around his wife’s “disappearence”
I guess I was one of the people that bought into the Reiser4 hype. The idea that I liked the most was the extensability through plugins idea. So if you wanted some bit of functionality integrated into the filesystem, you could just code against their API instead of having to learn all the internals and whatnot.
ext3 is really not a bad FS by any means, it just isn’t whiz-bang exciting. This is actually the first time I have heard of BTRFS, but it looks very promising. I worked against ORACLE 9i as a developer for about 3 years, and I will never forgive them for the hell that was learning PL/SQL. But if they manage to pull something like this off and open source it, that will go a long way.
There were good things about Reiser4. Hans did a great job, on his site, of explaining plugins, dancing trees, files as directories, and the rest.
Where he failed was in being able to convince the kernel devs that the code quality was good enough, that he had thought out all the low level, kernel crashing, consequences of his innovations, and that reiser4 was so wonderful that it deserved its own special VFS.
And I know that it is fashionable to view OSS as a meritocracy. It’s even good to do so, I think. But personalities do matter. And Hans’ skills at working with others, frankly, sucked.
I would rather have less hype, quality code, and a maintainer who plays nicely… a few years in the future, than Hans Reiser and the hype that is reiser4, now.
I tired reading all that “linux VFS” hype. VFS is pushed by RedHat ext* developers, and be just part of ext2. Hans make reiser4 not only for Linux and make good reason do not force reiser4 to use ancient ext2 interface aka “Linux VFS”. Kernel developers sometime says reasonable critycs but for example force all reiser4 to rewrite comment style, error handling (nikita-xxxx) etc. is just crap, they very well know it hurt reiser4 on other application (first reiser ans reiser4 is for some DARPA project and have all that uber-secure features).
So Linux devs just force Namesys to make 2 tree one for linux one for others. For me (and not only me reading some independent comments), Linus just personallu do not like Hans, period.
I call bullshit on that. Yes, VFS is used by ext2… and ext3, and by ext4, and XFS, and JFS, btrfs, 9p, coda, freevxfs… and by all the other fileystems which Linux supports. What made reiser4 so special that it deserved its own VFS, sequestered from the rest? If Hans had worked at getting the features he wanted integrated into the standard VFS that others have been using and improving over the years, then this would not be considered a problem. But that wouldn’t do for Namesys… because in that case other filesystems would have access to the facilities, too, and Namesys did not want that kind of competition. That is part of what I mean when I say that Han and Namesys did not have a “work with” attitude. Having Hans in charge of Linux premier filesystem would be like Xorg putting David Dawes in the position of BDFL.
I notice that you tried to paint this as a case of RedHat self interest. Based upon no evidence, of course. Because the fact of the matter it that Resier4 was always about Namesys’ self-interest. Or why don’t I just come out and say it? Hans’ self interest. Because Namesys has always been a synonym for Hans.
At any rate, this is all water under the bridge, because Reiser4 is dead. It has just taken some people longer than others to realize it.
But in the long run, I can’t help but feel that it is for the best.
“At any rate, this is all water under the bridge, because Reiser4 is dead. It has just taken some people longer than others to realize it. ”
It might be dead but just because Hans Reiser wasn;t a teamplayer doesn’t mean that the ideas of Reiser4 are crap & ex4 and btrfs will be the best thing since sliced bread .
Yes you didn’t say or even suggest it – but I get the feeling that reiser4 as a whole is labeled “arrogant crap” because the teamleader was less than lkml capable .
They still put years of work into it .
It might be an uninformed feeling – but I still feel from posts I read on lkml that ext4 is simply stuff glued onto the old but reliable base of ext2/3 – and that it does very much get prefered treatment because important filesystem developers are gatekeepers & can block things they don’t like .
ext3 – btrfs – xfs benchmarks I saw showed xfs scaling very well – behaving predicably & performing a lot better than ext3 & often equal to btrfs .
btrfs has cool good features but if it outperforms current offerings .. hmm dont think so
XFS deletes large files faster than the other filesystems. For desktop use, this is trivial; for a dedicated server like MythTV where the average file is 1.8Gb or larger and files are being deleted constantly while other disk-intensive activities are going on, it is not trivial.
It’s irritating that the alternate install versions of Ubuntu can correctly create/format an XFS partition but the regular liveCDs have not been able to for at least three major versions.
If I remember correctly IBM and EMC wanted a file system that would work better with very large file systems and have increased reliability. RedHat worked with these vendors to extend ext4. Better defragging, large file systems, more robust. I’m ready to upgrade right now when Fedora 9 final comes out.