Linux 2.6.29 Kernel Released; Hello KMS and Btrfs!

[phoronix]

Phoronix: Linux 2.6.29 Kernel Released; Hello KMS and Btrfs!

With much anticipation, the Linux 2.6.29 kernel was released just moments ago by Linus Torvalds. Besides introducing an interim logo to stand-in for Tux, the Linux 2.6.29 kernel most notably introduces support for kernel mode-setting on Intel hardware...

http://www.phoronix.com/vr.php?view=NzE2NA

[Kano]

Hello ATI:

Make 9-3+ ready for it! I do not want to search for patches!

[chithanh]

Additionally worth mentioning is that 2.6.29 finally introduces support for write barriers in LVM.

[stan]

Btrfs is incredibly slow; I'll wait for IBM to GPL ZFS

All the work going into Btrfs is wasted if IBM takes over Sun and GPLs ZFS. As it stands, Btrfs is incredibly slow, especially for desktop and game-related tasks that deal with small files.

I hope IBM gets to GPL-ing ZFS quickly so as to avoid any further duplication of work.

[sreyan]

All the work going into Btrfs is wasted if IBM takes over Sun and GPLs ZFS. As it stands, Btrfs is incredibly slow, especially for desktop and game-related tasks that deal with small files.

I hope IBM gets to GPL-ing ZFS quickly so as to avoid any further duplication of work.

I don't know if you've ever tried using the latest stable zfs-fuse but it's no speed demon.

GPL-ing ZFS might be the only way to ensure a quick zfs on linux. Of course to be fair there are reasonably fast fuse based filesystems like ntfs3g.

[stan]

I don't know if you've ever tried using the latest stable zfs-fuse but it's no speed demon.

Of course not, zfs-fuse uses a lot of indirections, goes through the userspace/kernel interface, and hasn't received all the love it should in terms of optimizations. A proper in-kernel ZFS that uses the original Sun codebase and is _not_ reimplemented for license incompatibility reasons will be much faster.

[sreyan]

http://www.phoronix.com/scan.php?page=article&item=os_threeway_2008&num=1

Has some interesting file benchmarks comparing zfs speed to ufs and ext3.

While I recognize that ext4 does not have the features and slick user management interface zfs does, it would be interesting to see a new benchmark comparing ext4 to ufs and zfs on a pure speed basis.

[Kano]

Speed is not everything. For long term data storage I would never recommend to use a new introduced filesystem. If you only do benchmarking thats fine

[Wyatt]

Speed is not everything. For long term data storage I would never recommend to use a new introduced filesystem. If you only do benchmarking thats fine

Seriously, QFT.
I've been asking people considering Ext4 for a while: "Do you want a new filesystem that's largely untested or an old filesystem that's proven itself stable over the course of the last decade?"
And I'll continue to do so for the foreseeable future.

That said, I'm all for advancement of the state of the art in filesystems. Bring on ZFS, btrfs, Tux3, SpadFS, and even reiser4; I'm a big fan of choice.

[Hephasteus]

I don't think ext4 is largely untested. It just has be limited to smaller than normal group while it was worked out. Being able to work with very large files much faster than ext3 is the only thing most desktop or even majority of file server users get out it.
UK spending 70 billion dollars a year to spy on it's citizens and all that other stuff all over the world that needs such huge databases is only people it will help. The filesystem isn't the danger it's the anaconda storage rewrite to support it and the new wozniak SSD's and general changes in storage hardware that are likely to occur over the next few years is where the PITA part of it to end users comes in.
If it weren't for all the success of late or rapid movement towards success the fail of this year to linux would be frightening.