Announcement

Collapse
No announcement yet.

XFS Developer Takes Shots At Btrfs, EXT4

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • XFS Developer Takes Shots At Btrfs, EXT4

    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

  • #2
    "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

    Comment


    • #3
      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.

      Comment


      • #4
        Huh?

        Originally posted by halo9en View Post
        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.

        Comment


        • #5
          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.

          Comment


          • #6
            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.

            Comment


            • #7
              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

              Comment


              • #8
                not a bug

                Originally posted by Ansla View Post
                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.

                Comment


                • #9
                  Originally posted by frantaylor View Post
                  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.

                  Originally posted by frantaylor View Post
                  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.

                  Originally posted by frantaylor View Post
                  "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).
                  Last edited by Ansla; 01-24-2012, 01:13 PM. Reason: Add P.S.

                  Comment


                  • #10
                    wikipedia dain bramage

                    Originally posted by Ansla View Post

                    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.

                    Comment


                    • #11
                      I kinda just agree with susikala. Esp after the fsck.btrfs announce, seems like just drama.
                      XFS is great for some things (that overlap with what btrfs hopefully will be even better at _in the future_). For now I am psyched about XFS improvements -- which is what I thought the talk was about. Why? Because XFS is infinity more reliable right now, and I use it. It would not be news to say that the ext4 has shortcomings and btrfs is still in serious development.
                      In essence what I am saying is that at least the only thing that Mason or Chinner have killed is ext4. ;-P hey where is reiser4 anyway ;-P

                      Comment


                      • #12
                        That's great if it truly is fixed. But I find it strange that even in 2011 some people complain about it. http://www.linuxquestions.org/questi...earing-919115/

                        Comment


                        • #13
                          I think he has a point about EXT4 if you look at where he's coming from

                          Which is big-iron systems. How much testing does EXT4 have of exabyte file systems? Multi-terrabyte files spread across dozens of disks in RAID?

                          Probably quite a bit less than XFS.

                          But it's clearly not true if you are talking about desktop systems.

                          Comment


                          • #14
                            Shrinking?

                            Does XFS support shrinking the filesystem already? Granted, that need doesn't arise often, but is still quite handy when reorganising partition layout..

                            Comment


                            • #15
                              Originally posted by smitty3268 View Post
                              Which is big-iron systems. How much testing does EXT4 have of exabyte file systems? Multi-terrabyte files spread across dozens of disks in RAID?

                              Probably quite a bit less than XFS.

                              But it's clearly not true if you are talking about desktop systems.
                              This. Even though JFS was designed for "huge" FS too, I understand it was only run on such with AIX, and so the linux jfs tools could not cope with >16tb until somewhat recently.

                              Comment

                              Working...
                              X