Announcement

Collapse
No announcement yet.

Btrfs / EXT4 / F2FS / XFS Benchmarks On The Linux 4.12 Kernel

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

  • Btrfs / EXT4 / F2FS / XFS Benchmarks On The Linux 4.12 Kernel

    Phoronix: Btrfs / EXT4 / F2FS / XFS Benchmarks On The Linux 4.12 Kernel

    Given the big changes of the Linux 4.12 kernel and a lot of that being block/file-system-related work, here are some fresh benchmarks of the Btrfs, EXT4, F2FS, and XFS file-systems compared to their performance on past kernel releases when using a solid-state drive.

    http://www.phoronix.com/vr.php?view=24668

  • #2
    Anyone know if PTS contains any tests to measure IO latency?

    Comment


    • #3
      Originally posted by peppercats View Post
      Anyone know if PTS contains any tests to measure IO latency?
      At the moment, none come to mind but there are around a thousand tests/suites and do lose track of some of the ones in there at times.... Any good open-source IO latency tests come to mind?
      Michael Larabel
      http://www.michaellarabel.com/

      Comment


      • #4
        With summer my partitions with ext4 or btrfs broke. I switched to XFS and almost no problem since then. I do not know if there are some benchmark for this (temperature / performance) but it does matter if you live in hot summers zones, and AC is not always connected because it is not healthy.

        Comment


        • #5
          I would like to see Btrfs without COW enabled.

          Comment


          • #6
            Originally posted by Steffo View Post
            I would like to see Btrfs without COW enabled.
            For the sake of stating the obvious: "nodatacow" option disables also checksumming, so it's probably not what most people looking at btrfs will use.

            Although why not. While we are at it, how about testing ext4 with data=journal mount option too? that's its equivalent of COW. So we can compare how good is btrfs COW with ext4's.

            Comment


            • #7
              Apples to apples.

              Comment


              • #8
                Inline snapshotting and checksumming wins over any negligible performance differences

                Comment


                • #9
                  I like how f2fs is performing. It means it's perfectly OK for work from some cheap USB key, CF card etc.

                  Comment


                  • #10
                    Originally posted by starshipeleven View Post
                    For the sake of stating the obvious: "nodatacow" option disables also checksumming, so it's probably not what most people looking at btrfs will use.
                    Also,
                    Code:
                    chattr +C
                    on an empty file disables COW too, but only at the file level, and is one of the proper recommended solutions for huge files that already have their own inner layer of journaling (e.g.: databases, virtual machines's harddrive image files, etc.) or do lots of random writes (e.g.: torrent downloads. Which incidently also have their own checksumming).
                    (the other solution includes using the auto-defragmenter).

                    ZFS impelement a different solution, which would be an "auto-nodatacow" for the files detected with excessive random writes.


                    I would have also appreciated a BTRFS vs. ZFS vs. BCacheFS benchmark, to compare approximately similar things.
                    (Well for BTRFS vs. ZFS. For now BCacheFS has really implemented the same feature set yet - e.g.: snapshoting isn't there yet).

                    Comment

                    Working...
                    X