Announcement

Collapse
No announcement yet.

Finally, Reiser4 Benchmarks Against EXT4 & Btrfs

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

  • #16
    As the BKL removal patch for ReiserFS was just merged it would be nice to have a version of the test executed comparing ReiserFS BKL vs. BKL free to see the impact. It would also be nice to note for future reference which is used in the test.

    Comment


    • #17
      SQLite makes a gazillion small changes to a single file. Worst-case scenario for a lot of journaling-schemes.

      Comment


      • #18
        What's the difference ReiserFS and Reiser4?

        I like to try this some day...Raiser=Razer

        Comment


        • #19
          Originally posted by intgr View Post
          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?

          What sort of crack is being smoked here?

          You link goes to "pub/linux/kenrel/people/edward"

          And the email link you posted here is a email _from_Edwards_ talking about how Reiser4 is broken in _2.6.32_ due to changesto the kernel introduced since 2.6.31.

          That post says that Reiser4 support is broken in 2.6.32, not that it has anything to do with anything specifically to Zen. So I have to assume that it's the 2.6.32 reiser4 patch in the link you posted that is broken.

          One guy posting broken patches and saying he will get around to fixing the problems in his spare time is not something gives me any confindence at all in the state of Reiser4.

          Comment


          • #20
            drag, the latest resier4 patch FROM EDWARD is NOT BROKEN.

            The stuff in ZEN on the other hand has been broken for ages and I am afraid since Micheal had problems is still broken.


            So, for you: Edward was able to create an unbroken patch for 2.6.32. A patch that is working flawless here. He also posted a series of patches for -mm on 2.2.2010. If you want to look into mailing list archives.

            Comment


            • #21
              Originally posted by patstew View Post
              I'm very aware that anecdote != data, but reiser4 is the only file system I've had where a bunch of files have irrevocably disappeared, and this was when it was supposed to be reasonably usable, not back when it was expected to be unstable, I'd guess sometime not too long before hans was caught. I'd be pretty wary about trusting it again.
              I had the same happening to me with (in ascending order):

              * reiserfs v3.6 (2004 then NEVER more )
              * JFS several times (it would even corrupt filenames, had problems with file creation dates and make files disappear - I gave it a try several times and each time it happened)
              * XFS (it zeroed several of my files during crashes on /home; when trying to use it on / (root) with dmcrypt [luks] it just literally ZEROED my system partition )
              * ext3 (corrupted my data after crashes)


              * ext4
              * reiserfs
              * reiser4
              ==> safe so far for me with current kernel releases

              Comment


              • #22
                WTF - stupid 10 minutes alteration window:

                Originally posted by patstew View Post
                I'm very aware that anecdote != data, but reiser4 is the only file system I've had where a bunch of files have irrevocably disappeared, and this was when it was supposed to be reasonably usable, not back when it was expected to be unstable, I'd guess sometime not too long before hans was caught. I'd be pretty wary about trusting it again.
                I had the same happening to me with (in ascending order):

                * reiserfs v3.6 (2004 then NEVER more, probably some hash problems - I meanwhile switched to tea hash in favor of r5 )
                * JFS several times (it would even corrupt filenames, had problems with file creation dates, problems with "umlaute" ( ) and make files disappear - I gave it a try several times and each time it happened)
                * XFS (it zeroed several of my files during crashes on /home; when trying to use it on / (root) with dmcrypt [luks] it just literally ZEROED my system partition - I tried it three times in 2 years and each time after a hardlock it happened )
                * ext3 (corrupted my data after crashes)


                * ext4
                * reiserfs
                * reiser4
                ==> safe so far for me with current kernel releases

                Comment


                • #23
                  WTF - stupid 10 minutes alteration window:
                  WTF - 45 seconds between posts (I understand this helps against those crazy spammers BUT COME ON !)

                  Originally posted by patstew View Post
                  I'm very aware that anecdote != data, but reiser4 is the only file system I've had where a bunch of files have irrevocably disappeared, and this was when it was supposed to be reasonably usable, not back when it was expected to be unstable, I'd guess sometime not too long before hans was caught. I'd be pretty wary about trusting it again.
                  I had the same happening to me with (in ascending order):

                  * reiserfs v3.6 (2004 then NEVER more, probably some hash problems - I meanwhile switched to tea hash in favor of r5 )
                  * JFS several times (it would even corrupt filenames, had problems with file creation dates, problems with "umlaute" ( ) and make files disappear - I gave it a try several times and each time it happened)
                  * XFS (it zeroed several of my files during crashes on /home; when trying to use it on / (root) with dmcrypt [luks] it just literally ZEROED my system partition - I tried it three times in 2 years and each time after a hardlock it happened )
                  * ext3 (corrupted my data after crashes)


                  * ext4
                  * reiserfs
                  * reiser4
                  ==> safe so far for me with current kernel releases

                  Comment


                  • #24
                    sorry but this is ridiculous

                    I once posted a correction (which showed an error and got me to post again which said I should wait 45 secs)

                    and the reward is a triple post

                    Comment


                    • #25
                      next run please inclue XFS

                      when you do the next round of such tests, please include XFS. I know that it's now new and sexy, but it's still reliably outperforming many of the filesystems in the test

                      Comment


                      • #26
                        I think the biggest problem here is Michael don't know how to make a kernel for Ubuntu, so he's used Zen from a PPA / http://liquorix.net/

                        Correct me if I'm wrong here

                        Comment


                        • #27
                          Stop with the SSD

                          I said this once before, but I guess Phoronix doesn't care.

                          Please use a standard hard drive. Preferably a laptop drive because those are the majority of hard drives today. Laptops/netops outsell standard desktops now.

                          Btrfs is the only file-system with SSD optimizations that you tested.

                          The other file-systems are designed with rotational media in mind. Reiser4 is especially tweaked to make seeks minimized.

                          With an SSD you are not really testing the algorithms. Seeks are nearly instantaneous. You are testing the CPU/chipset/controller bottleneck.

                          On a traditional HD, the seeks during small and random i/o determine the performance more than anything else.

                          I appreciate the time you put into the testing, but honestly your results are pretty much meaningless for most people. The market for SSD's are still insignificantly small with home users.

                          Comment


                          • #28
                            Originally posted by hechacker1 View Post
                            I said this once before, but I guess Phoronix doesn't care.

                            Please use a standard hard drive. Preferably a laptop drive because those are the majority of hard drives today. Laptops/netops outsell standard desktops now.
                            <SNIP>
                            With an SSD you are not really testing the algorithms. Seeks are nearly instantaneous. You are testing the CPU/chipset/controller bottleneck.
                            <SNIP>
                            I appreciate the time you put into the testing, but honestly your results are pretty much meaningless for most people. The market for SSD's are still insignificantly small with home users.
                            I disagree with you a bit here.

                            it would be nice to see both SSD and rotational tests that are comparable, both so that people considering SSDs can see what difference is made, and also to find out where a particular filesystem may fall behind on a fast disk setup.

                            There was a thread on linux-kernel this past week where EXT4 was showing significantly worse performance than XFS on large file writes, but only on systems with fast disk arrays (basically ext4 maxes out at about 60% the performance of XFS) I don't expect Phoronix to have such high-end disk arrays around to test with, but testing with SSDs is likely to show similar problems.

                            now if all you care about is 'your typical home user', then Phronix tests don't really matter much anyway, your typical home user isn't tweaking things, just installing the default setup.

                            however, there are 'home users' who spend hundreds of dollars for a video card, for people like that, SSDs are very much within their price range.

                            Comment


                            • #29
                              The Larger Problem

                              Which Distribution allows you to boot Reiser4 as the root file system?

                              Butter has a file system converter.

                              Comment


                              • #30
                                Originally posted by dlang View Post
                                now if all you care about is 'your typical home user', then Phronix tests don't really matter much anyway, your typical home user isn't tweaking things, just installing the default setup.
                                That's backwards. Phoronix ONLY tests the default setup. It's us tweakers to which the results don't apply.

                                I wish he would test ext4 on a mechanical laptop drive with barrier=0, noatime vs relatime, and writeback vs ordered. Plus xfs with all the tweaks its fans recommend, and reiser4 with lzo compression enabled.

                                Comment

                                Working...
                                X