Announcement

Collapse
No announcement yet.

Btrfs With Linux 6.0: Send Protocol v2, ~3x Boost For Direct Read Performance

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

  • Btrfs With Linux 6.0: Send Protocol v2, ~3x Boost For Direct Read Performance

    Phoronix: Btrfs With Linux 6.0: Send Protocol v2, ~3x Boost For Direct Read Performance

    SUSE's David Sterba sent in the Btrfs file-system updates on Monday for the in-development Linux 6.0 kernel...

    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
    Originally posted by phoronix View Post
    Phoronix: Btrfs With Linux 6.0: Send Protocol v2, ~3x Boost For Direct Read Performance
    RAID5/RAID6 fixes
    ...but still not considered stable 👺

    (It's mostly due to Write hole. Would be great if they could leverage the checksum to guess which alternative is correct one, though yes, it would be a lot more complicated than with RAID1 recovery).

    Comment


    • #3
      BTRFS is getting shaped into a killer FS that will blow away all any other FS

      Comment


      • #4
        Originally posted by siyia View Post
        BTRFS is getting shaped into a killer FS that will blow away all any other FS
        Both btrfs and ZFS are very advanced filesystems with amazing feature set and each being slightly better for some particular use cases.
        For a normal desktop / laptop / small RAID1 server, btrfs works really well while offering easy resizing / checksuming / compression / snapshots / send-receive / deduplication, etc. ,
        whereas for large servers, ZFS with RAIDZ and tiered caching is still hard to beat.

        Comment


        • #5
          Surprised me to see the RAID5/6 fixes! Thought it was dropped as it was stained for life.

          I agree with pkese, and use both btrfs (RAID10 SSD at home as in /home!) and ZFS (RAIDZ2 for a small business office NAS, lots of graphics/video assets), but in the end ZFS at home or small/medium biz might not be worth the effort, maybe at a datacenter it is possible to reap the benefits.

          In my case I wanted to try ZFS L2ARC caching with a retail SSD, oh was I so wrong about reliability, ZFS keeps writing to the SSD the whole time (I thought it was a write here and there for some warm data, but no, it almost uses it as RAM), so it killed a new SSD (no NAS branded) in less than a month, and well, we didn't see the benefit on a 5G ethernet network after removing the dead cache drive either, the end users didn't notice it was gone at all. That's the best benchmark, though numbers says otherwise. What I'm not so sure about is how great ARC is, compared to regular buffering/caching in Linux for btrfs, maybe ARC is what is giving all the benefits at that server.

          So, I really want to see tiered cache in btrfs, but now I'm not so sure, and probably will convert the small biz NAS to btrfs in the future, as the OS is Linux (FreeBSD used about 25w extra of power after tuning it a lot, Linux has better power management) and OpenZFS kind of feels like a second class citizen while it is out-of-kernel-tree. Better to leave it for the datacenters. RAID5/6 is a nice space compromise when there are backups, so RAIDZ was something I enjoyed, so still thinking wether to use RAID1 with btrfs in the future at that office.

          tldr; so kids at home, use btrfs and enjoy!

          Comment


          • #6
            Originally posted by DrYak View Post

            ...but still not considered stable 👺

            (It's mostly due to Write hole. Would be great if they could leverage the checksum to guess which alternative is correct one, though yes, it would be a lot more complicated than with RAID1 recovery).
            No, it's not due to write hole. It's absolutely not due to write hole.

            Comment


            • #7
              Finally a reading booster. After ext4 the asynchronous reading was very disappointing, even with 200MB files. If you open a folder with a Deja Dup backup, for example.
              But random writing of small files is a cool thing on HDD with BTRFS.

              Comment


              • #8
                Originally posted by siyia View Post
                BTRFS is getting shaped into a killer FS that will blow away all any other FS
                There's already reiserfs. Can't imagine how you could beat it considering hans's reputation.

                Comment


                • #9
                  Originally posted by DrYak View Post

                  ...but still not considered stable 👺

                  (It's mostly due to Write hole. Would be great if they could leverage the checksum to guess which alternative is correct one, though yes, it would be a lot more complicated than with RAID1 recovery).
                  The raid 5/6 changes do nothing to address the problems, and they're far more extensive than just a simple write hole.

                  problems with btrfs raid: https://lore.kernel.org/linux-btrfs/...ungrycats.org/
                  how to (not reliably) use btrfs raid: https://lore.kernel.org/linux-btrfs/...grycats.org/#t

                  Comment


                  • #10
                    Originally posted by caligula View Post

                    There's already reiserfs. Can't imagine how you could beat it considering hans's reputation.
                    BTRFS is literally the ReiserFS successor.

                    Comment

                    Working...
                    X