Kernel Log: Ext4 completes development phase as interim step to btrfs

Heise

 

So EXT4 is now out of Dev, I wonder if it is now considered Alpha or Beta quality ?

I'm amazed at how EXT3 has continued to improve, new features, refined code have kept it within reach of more modern file systems (in general usage), EXT4 will only seem like a small step forward as a result.
JFS and Reiser were always in the shadows in EXT3 and EXT4 will for sure kill off JFS (for Linux) and not recomended for product/stable enviroments anymore) at the moment due to part time support. JFS has know issues with fragmentation with time and as the disc fills, which the only developer knows how to fix, but has so far not had time to fix (EXT3 fixed this problem back in 2004 http://lwn.net/Articles/81359/).

Reiser is still in too much flux and I suspect will stay in its experiemental state for a while, especially due to the demand of new features from the VFS (Kernel filesytem API) which seldom appear.
XFS is unique in having a very good niche (large files and large file system performance) and will probably stay around for a while. XFS has SOME horrible code, because it pretty much is XFS for IRIX (Unix) with a translation layer added to be Linux compatible (inodes are different for example) and this has been the cause of some servere bugs cropping up. XFS is now far more power-outage safe though than it used to be, due to the inclusion of write barriers (on by default), though it still appears to lag behind JFS and EXT3.

 

where does zfs stand?

 

See here: Using ZFS though FUSE
The licenses are incompatible for ZFS to be incorporated in the kernel. You would need FUSE.

Maybe you're interested in this quote:

http://en.wikipedia.org/wiki/Btrfs

More on Btrfs.

 

I knew you would be interested
Perhaps early beta?

 

Btrfs and Squashfs merged into Linux kernel

Btrfs and Squashfs merged into Linux kernel

http://www.heise-online.co.uk/news/Btrfs-and-Squashfs-merged-into-Linux-kernel--/112389

 

Re: Btrfs and Squashfs merged into Linux kernel

Better FS is going to be THE next gen filesystem for Linux, already has the backing of Redhat and others and it will be possible to upgrade from EXT4 to Btrfs, which is cool.
It looks like its going to sit somewhere between ZFS and XFS for feature set, which is rather good.

Im currently sticking with EXT3 and will be an early adopter of EXT4 it will be included in Ubuntu 9.04 (http://brainstorm.ubuntu.com/idea/16854/) and EXT4 offers more fragmentation resitance AND quicker fsck and the latest benchmarks are showing it to be almost as good OR better than XFS (which has a more "advanced" features than EXT4 currently).
I also know that EXT4 will have gradual feature improvement, like dynamic inode allocation, but I suspect progress will be slow as focus is shifting towards Btrfs.

I was considering moving to XFS for my media partition, but i've never been comfortable the the size of the code (4x than EXT3), partly because the Linux implementation litterally takes the native Unix code (Unix inode is not exactly the same as a linux inode) and the linux implementation has a chunk of code that acts as a wrapper to translate and adds to lower efficency than ext3 (as an example ) and more suseptible to breakage due to kernal changes. On the flip side it is well used and tested and fixes are quick and quite mature in less and less bugs seen and has the advantage of scaling to high bandwidth AND/OR number of processes/threads better than any other tried and tested FS, which is a situation that EXT3 is not even optimised for because of its limitations.

Anyhow I predict XFS popularity to drop, as EXT4 will be more suitable for more stressful (in terms of bandwidth and fragmentation) and Btrfs on the horison which will should in theory be a better specced FS than XFS.