XFS Developer Takes Shots At Btrfs, EXT4

[phoronix]

Phoronix: XFS Developer Takes Shots At Btrfs, EXT4

Chris Mason of Btrfs fame wasn't the only Linux file-system developer talking to the public last week. While the Btrfs talk was going on in Los Angeles at SCALE 10x, Dave Chinner was down under in Australia at LCA2012 talking about XFS. His talk included some controversial shots at EXT4 and Btrfs...

http://www.phoronix.com/vr.php?view=MTA0NzM

[halo9en]

"EXT4 is not as stable or as well tested as most people think ... EXT4 has become an aggregation of semi-finished projects that don't play well with each other."

FUD, plain and simple. Currently Ext4 is the fastest and most reliable journaled linux fs. The truth is, XFS sucks

[jwilliams]

My mind boggles that ext4 filesystems STILL are not supported with ext-tools for filesystems larger than 16TB.

XFS is the only filesystem I trust to use for >16TB. Other possibilities are JFS, ZFS, and btrfs, but none of those are mature, stable, and actively developed on linux.

Maybe one day btrfs will be stable and not have any flaws (very slow performance in certain situations). But that day is far from today.

[frantaylor]

The truth is, XFS sucks

Got any facts to go with your FUD?

Local experience says that XFS is the best filesystem for big files like mythtv recordings and virtual image files.

Esp. since 2.6.39, XFS is sturdy and fast. I have RAID 0 with battery backup and write caching enabled, yes I am playing with fire but I have good backups. I get 575 MBytes/sec write and 375 Mbytes/sec read performance, which is pretty much hardware limited. I have had numerous system crashes without any data corruption. I can run 8 fully-loaded virtual machines from images on the same XFS filesystem, and they all get good disk io performance. I have tried ext3 and ext4 in this application and the performance is just miserable.

[susikala]

Why do I always get the feeling the articles here have become almost all sensationalist or displaying negative vibes. Is it really necessary to present things in a negative way to get a lot of publicity?

You could formulate it in a much more distant and professional way. "XFS Developer Comments On Other Filesystems", while writing something like "amongst the perceived deficiencies mentioned were ..."

I bet this would make people here react in a much less aggressive way, too. But I guess it's become all about money.

[[Knuckles]]

Honestly, I fully agree about btrfs.
Every time I try btrfs (last try with 3.0 on ubuntu 11.10), I end up with backtraces on my dmesg and strange problems, especially when the filesystem is starting to run low on disk space.

Of course I can try to bug report, and test, and etc, but honestly with so many filesystems available, I just ignore btrfs.

[Ansla]

XFS may be well suited for use cases where files are never/rarely overwritten like a media archive, but keep it off any essential partition like boot/root. Restoring from backups is not that fun when your computer no longer boots. http://ubuntuforums.org/showthread.php?t=1025412

[frantaylor]

XFS may be well suited for use cases where files are never/rarely overwritten like a media archive, but keep it off any essential partition like boot/root. Restoring from backups is not that fun when your computer no longer boots. http://ubuntuforums.org/showthread.php?t=1025412

That's not a bug report, it's a whine about lost files

You can find similar whines about every file system if you google

"never/rarely overwritten"

MORE FUD!!!

virtual machine image files are constantly open and constantly updated while they are open

In FACT, XFS is excellent for files that are constantly being overwritten. It is extent based, so if the file is not fragmented, the application will see performance similar to that of using a raw partition.

[Ansla]

That's not a bug report, it's a whine about lost files

It's not a bug, it's a feature. That's the way XFS is intended to behave in case of a forceful reboot.

You can find similar whines about every file system if you google

Sure, other file systems might corrupt data in case of a forceful reboot, but I'm not aware of any other that will delete your data ON PURPOSE.

"never/rarely overwritten"

MORE FUD!!!

virtual machine image files are constantly open and constantly updated while they are open

In FACT, XFS is excellent for files that are constantly being overwritten. It is extent based, so if the file is not fragmented, the application will see performance similar to that of using a raw partition.

How many times did the host crash/suffer a power-loss or been forcefully rebooted for any other reason?

P.S. In case you didn't know how XFS journaling works (http://en.wikipedia.org/wiki/XFS#Journaling) :

Where recently modified data has not been flushed to disk before a system crash, XFS ensures that any unwritten data blocks are zeroed on reboot, obviating any possible security issues arising from residual data (as far as access through the filesystem interface is concerned, as distinct from accessing the raw device or raw hardware).

[frantaylor]

P.S. In case you didn't know how XFS journaling works (http://en.wikipedia.org/wiki/XFS#Journaling

"Where recently modified data has not been flushed to disk before a system crash, XFS ensures that any unwritten data blocks are zeroed on reboot"

) :

This is what happens when you use Wikipedia as your source!

SEE the FAQ at xfs.org:

Q: Why do I see binary NULLS in some files after recovery when I unplugged the power?

Update: This issue has been addressed with a CVS fix on the 29th March 2007 and merged into mainline on 8th May 2007 for 2.6.22-rc1.