Announcement

Collapse
No announcement yet.

Btrfs With Linux 5.10 Brings Some Sizable FSync Performance Improvements

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

  • Btrfs With Linux 5.10 Brings Some Sizable FSync Performance Improvements

    Phoronix: Btrfs With Linux 5.10 Brings Some Sizable FSync Performance Improvements

    The Btrfs file-system is seeing some promising additions with Linux 5.10...

    http://www.phoronix.com/scan.php?pag...s-Fsync-Faster

  • #2
    That and data reservations will use the same ticketing system as for metadata. Future work also includes replacing the locking system with standard locking. That will reduce performance somewhat for some workloads but it should be good in the long run. Glad to see that my favourite filesystem is improving so much. I really hope that we are getting closer to supporting different mount options and/or storage profiles per subvolume. A while ago I suggested being able to assign/tag storage devices to groups and then assigning individual weights for read and/or write for a subvolumes assigned group. That would make a hybrid storage pool (sdd/hdd) much more feasible in my humble opinion. Crossing my fingers for some functionality like that!

    http://www.dirtcellar.net

    Comment


    • #3
      That would be great. Or anybody picking up the hot data relocation patches from IBM. But I don't see any work happening in this area.

      Comment


      • #4
        That's fantastic. Over last few months, I've migrated /home (4TB) and /var (much smaller) to native raid1 on btrfs on my server. I am looking forward to migrate / (root) partition too.

        Comment


        • #5
          My only problem with BTRFS at the moment is that directories that contain more than a few thousand files takes 10-20s to list from cold cache (this on a BTRFS Raid1 system with 110T so could be a case specific problem).

          Comment


          • #6
            Originally posted by F.Ultra View Post
            My only problem with BTRFS at the moment is that directories that contain more than a few thousand files takes 10-20s to list from cold cache (this on a BTRFS Raid1 system with 110T so could be a case specific problem).
            110TB with RAID1? Wouldn't you be better off with other RAID configuration than RAID1?

            Comment


            • #7
              Originally posted by piorunz View Post

              110TB with RAID1? Wouldn't you be better off with other RAID configuration than RAID1?
              Depends how much I/O do you need from the array. RAID5 will get you only 20%-40% of the I/O performance compared to what RAID1 would do.

              Comment


              • #8
                Wouldn't raid 6 be better than 5 for I/O performance?

                Comment


                • #9
                  Originally posted by Snaipersky View Post
                  Wouldn't raid 6 be better than 5 for I/O performance?
                  It's doesn't matter anymore: https://www.zdnet.com/article/why-ra...rking-in-2019/

                  Comment


                  • #10
                    Originally posted by Snaipersky View Post
                    Wouldn't raid 6 be better than 5 for I/O performance?
                    On the same amount of physical disks, RAID6 would perform approximately the same at read and somewhat worse at write (i.e. needs to write parity to one more drive than RAID5).

                    Comment

                    Working...
                    X