Btrfs Merged Into Mainline Linux Kernel

[phoronix]

Phoronix: Btrfs Merged Into Mainline Linux Kernel

Btrfs, the next-generation Linux file-system conceived by Oracle and designed to compete with some of the features found in Sun's ZFS file-system, has just been merged for the Linux 2.6.29 kernel. Last week we shared that Btrfs was getting ready for the mainline kernel and since then Chris Mason and other kernel hackers have committed several commits to the btrfs-unstable tree. There have been 21 commits to this new open-source file-system in the past four days. This morning Linus Torvalds finally pulled Btrfs into the mainline kernel...

http://www.phoronix.com/vr.php?view=Njk4Mw

[Louise]

For me, 2.6.29 will be the feature wize coolest kernel to date.

2.6.30 will likely be the coolest kernel bug fix wize ;)

[joffe]

So, will it be functional or is it purely for testing still?

[d2kx]

It would be interesting to see SSD benchmarks with the following:

ext3
ext4
btrfs
btrfs (SSD mode)

Does anyone know whether the SSD mode is already implemented and whether it will detect your SSD automatically and enable the optimized mode?

[Naib]

Sweet!!! I am after doing a new server build (present only has a single harddrive) and I /really/ want to try this out

having two HDD in mirrored RAID with a BTRFS should be sweet!!!

[stan]

The benchmarks linked in the Btrfs home page show that Btrfs is phenominally BAD compared to pretty much any other Linux filesystem. For example, look at the Single Disk Mail Server Simulation with one thread (http://btrfs.boxacle.net/repository/single-disk/Initial-compare/Initial-Compare-Single_disk_Mail_server_simulation._num_threads=1. html): Btrfs is eleven times (1100%) less efficient than EXT4. The second worse filesystem after Btrfs is JFS, and even it is 9X more efficient than Btrfs.

Upgrading to Btrfs will mean a sure slowdown for the vast majority of Linux users unless Oracle devs have a miracle up their sleeves.

ZFS, on the other hand, is both faster than EXT4 and more featureful. Can't beat that.

[JeanPaul145]

So, will it be functional or is it purely for testing still?

That's something I've been wondering myself.

[deanjo]

The benchmarks linked in the Btrfs home page show that Btrfs is phenominally BAD compared to pretty much any other Linux filesystem. For example, look at the Single Disk Mail Server Simulation with one thread (http://btrfs.boxacle.net/repository/single-disk/Initial-compare/Initial-Compare-Single_disk_Mail_server_simulation._num_threads=1. html): Btrfs is eleven times (1100%) less efficient than EXT4. The second worse filesystem after Btrfs is JFS, and even it is 9X more efficient than Btrfs.

Upgrading to Btrfs will mean a sure slowdown for the vast majority of Linux users unless Oracle devs have a miracle up their sleeves.

ZFS, on the other hand, is both faster than EXT4 and more featureful. Can't beat that.

You really can't compare a filesystem that still has a lot of debugging code in it. Besides those results are fairly old right now and the last couple months have been about optimization of it in the recent commits.

[Kano]

Well does GRUB(2) support it already? Thats whats needed to use it without boot partition (a very stupid thing to use).

[_txf_]

It would be interesting to see SSD benchmarks with the following:

ext3
ext4
btrfs
btrfs (SSD mode)

Does anyone know whether the SSD mode is already implemented and whether it will detect your SSD automatically and enable the optimized mode?

I imagine this would be done using a script to change the mount option in fstab

[some-guy]

btrfs is still in heavy development, don't expect it to go stable till at least 2.6.34

[deanjo]

btrfs is still in heavy development, don't expect it to go stable till at least 2.6.34

I would wager on a lot sooner then that.

[danjeel]

That's something I've been wondering myself.

"Anyway, give it a good testing. And please do keep in mind that while new
filesystems can be intruiging and exciting, "new" also means "not widely
tested". I'm sure the btrfs people will appreciate people testing, but I
would suggest that you go very very carefully (squashfs is read-only, so
it's presumably less likely to eat your data, but who knows - the
perversity of the universe is endless)." - Linus LKML

[some-guy]

I would wager on a lot sooner then that.
Well, 2.6.34 is in a little more than a year since there are about 4 kernel releases a year

[deanjo]

Well, 2.6.34 is in a little more than a year since there are about 4 kernel releases a year

I'm thinking more like in 6 months. It's a new filesystem, it doesn't have to go through all the worries such as backwards compatibility that a file system such as ext4 has to go through. I have it will hit 1.0 status fairly soon. 2.6.31 or 2.6.32.

[Vighy]

someone who knows these things better said 1 year :D

http://linux-foundation.org/weblogs/lwf/2009/01/11/looking-forward-to-2629/

and he said also:
Btrfs is intended to be the next-generation filesystem which, conceivably, could last us for the next 10-20 years. The filesystem found in 2.6.29 is not, yet, that filesystem, though; Btrfs remains under heavy development.

so... this should be taken as most probable.

[deanjo]

someone who knows these things better said 1 year :D

http://linux-foundation.org/weblogs/lwf/2009/01/11/looking-forward-to-2629/

and he said also:


so... this should be taken as most probable.

Well I recall somewhere in the lkml where Chris was initially aiming for a December 1.0 status but later decided that would be a bit to ambitious as it got closer to Dec. the wiki is still saying 4th Q / 2008. Now with inclusion with the kernel he also expects development to further accelerate.

http://btrfs.wiki.kernel.org/index.php/Development_timeline