Announcement

Collapse
No announcement yet.

Finally, Reiser4 Benchmarks Against EXT4 & Btrfs

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

  • Finally, Reiser4 Benchmarks Against EXT4 & Btrfs

    Phoronix: Finally, Reiser4 Benchmarks Against EXT4 & Btrfs

    There is no shortage of EXT4 benchmarks from comparing this evolutionary file-system's performance on netbooks to how it battles the Btrfs file-system to its performance recession. We have even benchmarked it on USB flash drives and on high-end SSDs. We have also delivered numerous Btrfs benchmarks. In this article though we are finally delivering something that has long been requested and that is Reiser4 file-system benchmarks running directly against EXT4 and Btrfs. We have also thrown in the original ReiserFS file-system for comparison too.

    http://www.phoronix.com/vr.php?view=14638

  • #2
    You mention running these tests on the 60gb Vertex, it's the same one I have so I know a little about it.

    How are you resetting the drive after each test? With wiper.sh or TRIM, or secure erase, or garbage collection?

    Comment


    • #3
      Here is a tabular summary, including the Harmonic Mean, which is the measure to use for rates (http://en.wikipedia.org/wiki/Harmonic_mean)
      Benchmark;Reiser4;EXT4;Btrfs;ReiserFS
      Compile_Bench_Initial_Create;75,41;46,58;53,43;13, 42
      Compile_Bench_Compile;51,24;47,74;68,52;22,96
      Compile_Bench_Read_Compile_Tree;169,61;84,41;93,26 ;69,5
      IOZone_2GB_Write_Performance;85,83;71,46;81,31;59, 24
      IOZone_4GB_Write_Performance;74,71;64,99;73,86;54, 96
      IOZone_8GB_Write_Performance;73,95;66,31;72,64;53, 64

      Median_Performance(MB/s);75,06;65,65;73,25;54,3
      Harmonic_Mean_Performance(MB/s);77,69;60,78;71,74;32,23


      Reiser4___100.00%
      EXT4_______78.23%
      Btrfs______92.34%
      ReiserFS___41.48%
      Sex, lies and nihilistic pleasures.

      Comment


      • #4
        There are many more file-system / disk tests that we normally run in such articles, but these other tests were unable to successfully run on the Reiser4 file-system without crashing
        Geez! reiser4 is known to be broken on Zen-Kernel because has not been ported to Linux 2.6.32 yet.

        What was wrong with the official reiser4 patches?

        Comment


        • #5
          fsck performance

          That was an good article. Thanks.

          I think it might be interesting to see the fsck performance in filesystem benchmarks likes these since it delays booting during the periodic times that it is run. Perhaps copy a several hundred gigabyte partition containing data to multiple partitions of the same size but with different filesystems and measure how long it takes fsck (using the same options that are used when it runs during boot) to complete for each one.

          I know about tune2fs and the ability to turn it off but the reality is that the default state is usually to have a periodic fsck enabled. Another facet of this is that the fsck performance impacts boot performance yet most boot performance tests do not test with an fsck (as far as I have seen). Anyway, just an idea.

          Again, thanks for the benchmark!

          Comment


          • #6
            Originally posted by intgr View Post
            Geez! reiser4 is known to be broken on Zen-Kernel
            Ok my fault, the 2.6.33-rc7-zen1 release claims that "reiser4 should work now". But regardless, I would like to see testing with official upstream patches before declaring it as failing the tests.

            Comment


            • #7
              yeah, that conclusion is alsmost as stupid as telling people to use the mm or zen patches.

              DO NOT USE THE ZEN OR MM PATCHES! THEY ARE BROKEN!

              Use the standard patch like reiser4-for-32 you can find in edwards directory on ftp.kernel.org.

              Michael, your test is pretty much fail. You used known broken stuff and then blame reiser4 for problems.

              I haven't had problems in years using reiser4 as long as I stayed with the patches from Edward. zen on the other hand - don't get me started.

              Oh -and a run with lzo compression enabled would have been nice too.
              Or ext3 with barriers forced on.
              Or a little 'lets push reset' test.

              Comment


              • #8
                What I find most interesting is the performance of ext4... Never the fastest, but consistently and boringly middle of the pack.

                That said, I'm not really that interested in "speed" of my file-systems, the performance metric that matters for me is how the file-system behaves when things crash: i.e. How well it preserves my data! The fastest FS in the world isn't that much use if it looses data, and as disks are slow anyway...

                Comment


                • #9
                  Sounds like a conspiracy to me :P

                  Comment


                  • #10
                    Originally posted by RobbieAB View Post
                    What I find most interesting is the performance of ext4... Never the fastest, but consistently and boringly middle of the pack.

                    That said, I'm not really that interested in "speed" of my file-systems, the performance metric that matters for me is how the file-system behaves when things crash: i.e. How well it preserves my data! The fastest FS in the world isn't that much use if it looses data, and as disks are slow anyway...
                    well, in that case you should use reiser4. Reiser4 treats the safety of your data with priority. Being atomic helps a lot.

                    If you don 't care about your data then ext4 might be a good choice.

                    Comment

                    Working...
                    X