Announcement

Collapse
No announcement yet.

Linux 3.7 File-System Benchmarks: EXT4, Btrfs, XFS

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

  • Linux 3.7 File-System Benchmarks: EXT4, Btrfs, XFS

    Phoronix: Linux 3.7 File-System Benchmarks: EXT4, Btrfs, XFS

    In this article are benchmarks of the latest Linux 3.7 kernel development code of the EXT4, XFS, and Btrfs file-systems.

    http://www.phoronix.com/vr.php?view=18153

  • #2
    Is the kernel getting slower each release? (except the graphics stack)

    Comment


    • #3
      Thanks for the HDD tests in future.

      Comment


      • #4
        Originally posted by bachinchi View Post
        Is the kernel getting slower each release? (except the graphics stack)
        Huh?
        I did a fast count and its seems that the number of benchmarks won by 3.7 outweigh the number of benchmarks lost. (7:3 in the case of ext4)
        I'd take btrfs out of the equation until its reaches a stable status (mid-development-code tends to be performance regression happy).

        - Gilboa
        DEV: Intel S2600C0, 2xE52658V2, 32GB, 4x2TB + 2x3TB, GTX780, F21/x86_64, Dell U2711.
        SRV: Intel S5520SC, 2xX5680, 36GB, 4x2TB, GTX550, F21/x86_64, Dell U2412..
        BACK: Tyan Tempest i5400XT, 2xE5335, 8GB, 3x1.5TB, 9800GTX, F21/x86-64.
        LAP: ASUS N56VJ, i7-3630QM, 16GB, 1TB, 635M, F21/x86_64.

        Comment


        • #5
          Thanks for the benchmarks. It looks like ext4 continues to perform better than btrfs even on SSDs (which in theory should be btrfs' strong point)...
          BTW I'd like to see a comparison between the last kernel releases (3.7 vs 3.6 vs 3.5...)

          Comment


          • #6
            Is there any news of FTRFS : Fractal Tree FS , by TokuDB ?
            Last i heard was their presentation of FTRFS on some DB convention in june/july. Its supposed to be the thing for next gen DB's.

            Comment


            • #7
              any ideas as of why XFS would perform so much worse in the Threaded I/O Tester v0.3.3 test comparing to the 3.5 kernel?

              Comment


              • #8
                Hard to compare

                I've tried both Ext4 and Btrfs on my laptops Samsung SSD and found that Btrfs with compression and discard works the best for me. Certainly the benchmarks from last year prove the compression with Btrfs vastly improves performance:

                http://www.phoronix.com/scan.php?pag...lzo_2638&num=2

                Comment


                • #9
                  Originally posted by Nextweek View Post
                  Certainly the benchmarks from last year prove the compression with Btrfs vastly improves performance:
                  No, they prove nothing of the sort. All they prove is that if you are writing a stream of zeros to your drive, the performance will be better with compression. With real data that does not compress so easily, the performance will be completely different than the benchmarks.

                  Comment


                  • #10
                    Could you provide reference?

                    Originally posted by jwilliams View Post
                    No, they prove nothing of the sort. All they prove is that if you are writing a stream of zeros to your drive, the performance will be better with compression. With real data that does not compress so easily, the performance will be completely different than the benchmarks.
                    It is considered inappropriate to claim something without a reference, could you link me to where it says that the tests were just streams of zeros?

                    Comment


                    • #11
                      So where in this test do you think Michael hit those performance-losses a SSD can show due to too heavy wear?

                      Comment


                      • #12
                        Originally posted by Nextweek View Post
                        could you link me to where it says that the tests were just streams of zeros?
                        Name a benchmark that you are interested in, and if you lack the basic google skills to look it up, I will help you.

                        Comment


                        • #13
                          Btrfs decides whether or not something is worth compressing and only compresses it if it is worth it.

                          EDIT:
                          Well it's a parameter you can pass, at least. I think this is the default behavior.

                          Comment


                          • #14
                            Originally posted by Nextweek View Post
                            It is considered inappropriate to claim something without a reference, could you link me to where it says that the tests were just streams of zeros?
                            The code for 3 different FS benchmark utilites I inspected myself say so.

                            Basically any benchmark that doesn't write just stream of zeros is "a bit" vocal about it.

                            Comment


                            • #15
                              Originally posted by jwilliams View Post
                              No, they prove nothing of the sort. All they prove is that if you are writing a stream of zeros to your drive, the performance will be better with compression. With real data that does not compress so easily, the performance will be completely different than the benchmarks.
                              Yep, and don't most modern SSD's do some kind of compression at the firmware level anyway?

                              Comment

                              Working...
                              X