Announcement

Collapse
No announcement yet.

Google To Switch To EXT4, Hires Ted To Code

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

  • #11
    Originally posted by apaige View Post
    It would have carried a bit more weight had you mentionned that in the part I quoted. Also, [citation needed]
    Clarification added.
    Michael Larabel
    http://www.michaellarabel.com/

    Comment


    • #12
      deleted - it was unfair of me to say this.

      @ Phoronix - Keep up the good work, but please do not suggest that the PTS was used when other benchmarking apps used were listed, but not PTS.
      Last edited by bugmenot; 01-15-2010, 09:49 AM.

      Comment


      • #13
        With the Linux 2.6.32 kernel, EXT4 lost much ground while Btrfs gained and before that a single commit severely dampened the FS performance.
        It became so slow on my machine that sometimes my friggin cursor freezes.

        ...the performance of EXT4 is looking to be even worse with the forthcoming Linux 2.6.33 kernel.
        oh boy...

        Sometimes I regret switching from ReiserFS.

        Comment


        • #14
          Originally posted by wooptoo View Post
          Sometimes I regret switching from ReiserFS.
          So shrink it or use another drive with RFS and compare the two or do a benchmark. Also, note the differences in features, and then decide overall which you prefer.

          Let us know. :P

          Comment


          • #15
            Google's Michael Rubin shared that they chose EXT4 after benchmarking it as well as XFS and JFS (possibly with our Phoronix Test Suite carrying out some of the testing, which they have used in other areas).
            I'm surprised they didn't choose ext3 on ubuntu.

            Comment


            • #16
              ext4 is safe, probably not fast.

              That being said, where I used to work they *still* use reiserfs3. JFS & XFS just flat aren't stable, they aren't dependable under hardware failure or conditions of power failure (wall or ups, take your pick). The biggest beef I had in the past with the ext? series is the time to fsck is extremely excessive. reiserfs even with rebuild-tree could be back online dramatically sooner than ext? with 16 drive raid6's.

              I haven't run ext4, I'm assuming the time to fsck hasn't improved much.

              Comment


              • #17
                Originally posted by bnolsen View Post
                I haven't run ext4, I'm assuming the time to fsck hasn't improved much.
                I think that is something that was actually significantly improved.

                Comment


                • #18
                  Originally posted by bnolsen View Post
                  reiserfs even with rebuild-tree could be back online dramatically sooner than ext? with 16 drive raid6's
                  It might be fast, but it also has some pretty serious shortcomings (e.g. merging files from ReiserFS images into the filesystem itself). Reiser himself acknowledged this (his solution: ReiserFS v3 is obsolete, use Reiser4).

                  Comment


                  • #19
                    Originally posted by bnolsen View Post
                    I haven't run ext4, I'm assuming the time to fsck hasn't improved much.
                    Actually, fsck is many orders of magnitude faster on ext4 compared to ext3. It completes in ~3 seconds on my 80GB SSD and ~15 seconds on my 1TB disk (versus several minutes on 1TB/ext3).

                    Comment


                    • #20
                      Originally posted by bnolsen View Post
                      ext4 is safe, probably not fast.
                      ...the performance of EXT4 is looking to be even worse with the forthcoming Linux 2.6.33 kernel.
                      It would be great to see some more meaningful file systems benchmarks like: copying, creating, deleting files and directories and random reads, writes times rather then some apache, sqlite benchmarks etc.
                      Last edited by kraftman; 01-18-2010, 07:56 AM.

                      Comment

                      Working...
                      X