Announcement

Collapse
No announcement yet.

Btrfs With Linux 6.2 Bringing Performance Improvements, Better RAID 5/6 Reliability

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

  • Btrfs With Linux 6.2 Bringing Performance Improvements, Better RAID 5/6 Reliability

    Phoronix: Btrfs With Linux 6.2 Bringing Performance Improvements, Better RAID 5/6 Reliability

    The Btrfs and EXT4 file-system updates for the Linux 6.2 merge window have been submitted. The Btrfs changes are rather notable with continued performance enhancements as well as making some reliability improvements to its native RAID5/RAID6 modes...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    Michael

    Last sentence: Linux 6.2 is a rather exciting mice of improvements.

    Btrfs With Linux 6.2 Bringing Performance Improvements, Better RAID 5/6 Reliability​

    McbainThat'sTheJoke.gif

    Comment


    • #3
      With the first Release Candidate of Linux Kernel 6.2 I'd like to see some benchmarks .... BTRFS vs ext4 vs XFS ...etc... with kernel 6.0/6.1/6.2-rc1.

      Comment


      • #4
        Originally posted by phoronix View Post
        Phoronix: Btrfs With Linux 6.2 Bringing Performance Improvements, Better RAID 5/6 Reliability

        The Btrfs and EXT4 file-system updates for the Linux 6.2 merge window have been submitted. The Btrfs changes are rather notable with continued performance enhancements as well as making some reliability improvements to its native RAID5/RAID6 modes...

        https://www.phoronix.com/news/Linux-6.2-Btrfs-EXT4
        Given the history and currently accepted problems with RAID5/6... "Better" is the worst possible word here... it really should say "looking for new alpha testers with data to loose"!

        Comment


        • #5
          Finally some native encryption on the horizon... still far away though.

          Comment


          • #6
            As Qu Wenruo points out, this more accurately switches it to do checksum verification during a read-modify-write cycle. It sounds like it's a more constant check to avoid rewriting bad data during partial changes of a stripe (however it got bad - perhaps memory corruption during a read, for example).

            As someone who runs RAID5 btrfs in production (crazy, I know; but the RAID level hasn't been a problem so far, unlike delayed allocation), improvements are welcome. Performance seems good right now, though, especially after recent scrub optimization. In our use case there's limited modification going on, it's more of a media store, so I guess that helps.
            Last edited by GreenReaper; 12 December 2022, 08:35 PM.

            Comment


            • #7
              I'd like Raid6 or Raid6+1 (or the equivalent of Raidz3) and I'm glad Btrfs is getting there. The more in-kernel options we can rely on, the better.

              Comment


              • #8
                Originally posted by GreenReaper View Post
                As Qu Wenruo points out, this more accurately switches it to do checksum verification during a read-modify-write cycle. It sounds like it's a more constant check to avoid rewriting bad data during partial changes of a stripe (however it got bad - perhaps memory corruption during a read, for example).

                As someone who runs RAID5 btrfs in production (crazy, I know; but the RAID level hasn't been a problem so far, unlike delayed allocation), improvements are welcome. Performance seems good right now, though, especially after recent scrub optimization. In our use case there's limited modification going on, it's more of a media store, so I guess that helps.
                your secret is safe with us.

                Comment


                • #9
                  Originally posted by GreenReaper View Post
                  [URL="https://lore.kernel.org/lkml/[email protected]/"]
                  As someone who runs RAID5 btrfs in production (crazy, I know; but the RAID level hasn't been a problem so far, unlike delayed allocation), improvements are welcome...
                  Agreed. I run btrfs RAID5 (data, RAID1 metadata) on my NAS and Jellyfin servers and it's been quite nice in terms of flexibility of adding and moving drives. No linux native filesystem has the unique combination of features that btrfs offers. It's nice to see it continue being developed and improved.

                  I think most people that crap on btrfs have never actually used it, or have never tried to recover a filesystem with it. No it's not perfect. Yes it has shortcomings. But it seems like the common complaint is, "look! the documentation admits RAID5/6 is br0k3n! it's useless! it'll explode your data and launch your cat into orbit!"

                  Comment


                  • #10
                    Originally posted by nranger View Post

                    Agreed. I run btrfs RAID5 (data, RAID1 metadata) on my NAS and Jellyfin servers and it's been quite nice in terms of flexibility of adding and moving drives. No linux native filesystem has the unique combination of features that btrfs offers. It's nice to see it continue being developed and improved.

                    I think most people that crap on btrfs have never actually used it, or have never tried to recover a filesystem with it. No it's not perfect. Yes it has shortcomings. But it seems like the common complaint is, "look! the documentation admits RAID5/6 is br0k3n! it's useless! it'll explode your data and launch your cat into orbit!"
                    My cat never came home after I installed with BTRFS.

                    Comment

                    Working...
                    X