Btrfs In Linux 3.4 Kernel Has Big Changes

[phoronix]

Phoronix: Btrfs In Linux 3.4 Kernel Has Big Changes

As the latest work queued up for merging in the Linux 3.4 kernel is the Btrfs file-system pull, which Chris Mason describes as "pretty big, picking up patches that have been under development for some time."..

http://www.phoronix.com/vr.php?view=MTA3OTk

[garegin]

wake me up when it's faster and more reliable than ext4. that would take two years. in that time if kristian doesn't get run over by a bus well have wayland working too.

[benmoran]

I started playing around with BTRFS the other day, and it's really cool. The snapshotting works beautifully. For me, the speed is already acceptable. I'm most interested in snapshotting, online resizing, and raid 5/6 support. This would let me replace a combination of software raid, LVM and traditional filesystem, which is not the fastest thing in the world as is.

Looking forward to testing it.


edit: BTRFS doesn't seem to be aimed at Desktop users, so why bother comparing it to Ext4? Ext4 has little to no functionality.

[drag]

edit: BTRFS doesn't seem to be aimed at Desktop users, so why bother comparing it to Ext4? Ext4 has little to no functionality.

Ext4 is a stop-gap file system. It introduced some features necessary to make Ext3 usable on large file systems, such as massive improvements in fsck performance.

Btrfs is the hopeful replacement for the venerable ext* series of file systems and is the sort of thing that is needed for Linux on the desktop. The data protection is much better and it should be rather robust on crappy hardware, which is typical consumer stuff.

The snapshot features has the opportunity to improve the usability of the installing and managing software, for example. Now with btrfs it is possible to roll back a screwed-up upgrade, for example. Also it can be used to enhance in the sandboxing of applications for improved security.

Btrfs, especially with the addition of re-striping features and improved rebalancing code, allows the Linux desktop to do some pretty cool stuff like dynamically expanding AND shrinking of storage. So a desktop user that is running out of storage for their movies could add a big new drive and get it sucked into their btrfs file system in a fairly 'automagic' manner. Later on when they want to remove a slower, smaller drive, they can remove it from the pool with a few simple commands (hopefully there would be integration into stuff like gparted for usability).

With the ability to set data to 'raid1' or 'raid10' and be able to 'scrub' (test all the blocks by reading them in) it will make bitrot a thing of the past. No more silent corruptions in mp3 music databases and all sorts of fun things like that.

So when features like trivially expandable storage, easy drive swapping, snapshotting, robust data protects, and other features get integrated properly into the desktop then it has the potential to improve the usability and capabilities of storage, as it relates to the desktop.

[drag]

Btrfs is the hopeful replacement for the venerable ext* series of file systems and is the sort of thing that is needed for Linux on the desktop. The data protection is much better and it should be rather robust on crappy hardware, which is typical consumer stuff.

Sorry, didn't quite finish the thought.

I meant to say something along the lines of "Btrfs is being touted as the replacement and logical upgrade path from Ext4 for Linux general purpose. So it's being aimed squarely at desktop users".

Desktop users is certainly a target audience for btrfs. So that is why it's worth comparing.

[renkin]

I love how people always bash Michael about the articles all being copy paste from older articles. Or copy paste from developer + some stuff from older articles. Granted, that may be true to some degree, but the same could be said about the forum posts. The threads that follow each Phoronix piece have pretty much the same comments.

Phoronix news bot
......... [ads]

user 1
HURRY UP! I want X feature!

user 2
Meh. Unimpressed. Wake me up when blah blah blah

user 3
Works for me!

** mild debate spans 5 comments
** heated argument spans 5 pages


Anyway, Q, if you're interested, the lead btrfs dev at redhat did some math to basically say that b-trees are NOT made obsolete by fractal trees and that b-trees in many ways is still more efficient than fractals. The thread can be found here.

[Goderic]

wake me up when it's faster and more reliable than ext4. that would take two years. in that time if kristian doesn't get run over by a bus well have wayland working too.

I don't think it's ever going to be faster than ext4 (maybe except for gigantic filesystems). BTRFS was not designed to be faster than ext4, it was designed to have nice features and to be reliable.

[Drago]

I don't think it's ever going to be faster than ext4 (maybe except for gigantic filesystems). BTRFS was not designed to be faster than ext4, it was designed to have nice features and to be reliable.

Slow filesystem can and will be fixed with faster hardware (which we get every year), but not-reliable and not future-full filesystem cannot be fixed with faster hardware.

[siride]

I don't think it's ever going to be faster than ext4 (maybe except for gigantic filesystems). BTRFS was not designed to be faster than ext4, it was designed to have nice features and to be reliable.

And yet there is no fsck and won't be for some time.

[kernelOfTruth]

@Michael:

typo:

reworked mata-data


thanks for the article !