Announcement

Collapse
No announcement yet.

Btrfs Restoring Support For Swap Files With Linux 4.21

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

  • #71
    Originally posted by edgmnt View Post
    Has anyone benchmarked regular swapfile vs. swap on loopback-mounted file? I figure filesystem compression might offset some of the overhead associated with loopback, that'd be interesting to see.
    Filesystem compression is irrelevant for swap on BTRFS since BTRFS does just reserve an area where the swap file will stay safe (as far as I understood it).
    If any benchmarking is done , it should be good to try with and without zswap enabled as well. ( echo 1 > /sys/module/zswap/parameters/enabled )

    http://www.dirtcellar.net

    Comment


    • #72
      Originally posted by waxhead View Post

      Filesystem compression is irrelevant for swap on BTRFS since BTRFS does just reserve an area where the swap file will stay safe (as far as I understood it).
      Loopback. You could do this even back when BTRFS did not support swapfiles.

      Comment


      • #73
        Originally posted by pal666 View Post
        not having redundancy and checksums was ext4 fault
        Redundancy and checksums do *NOT* make you any safer in case of kernel bugs. In fact btrfs ate my data multiple times because of that.
        ## VGA ##
        AMD: X1950XTX, HD3870, HD5870
        Intel: GMA45, HD3000 (Core i5 2500K)

        Comment


        • #74
          Originally posted by edgmnt View Post

          Loopback. You could do this even back when BTRFS did not support swapfiles.
          Yes, absolutely. It is a bit slow , but then again swapping usually is

          http://www.dirtcellar.net

          Comment


          • #75
            Originally posted by darkbasic View Post

            I'm pretty sure that ZVOL doesn't disable COW in any way, can you please share your sources?
            OK I stand corrected. But that means that there is no way to avoid COW in ZFS even in cases where it's not desirable. So ZFS has actually more problems than I thought.

            Comment


            • #76
              Originally posted by jacob View Post

              OK I stand corrected. But that means that there is no way to avoid COW in ZFS even in cases where it's not desirable. So ZFS has actually more problems than I thought.
              The point is that I never felt the need to disable COW in ZFS, while the opposite is true with btrfs. With ZFS I do use COW for virtual machines with ext4-like performance, just a little bit of tuning is needed: http://www.linuxsystems.it/2018/05/o...t4-benchmarks/
              ## VGA ##
              AMD: X1950XTX, HD3870, HD5870
              Intel: GMA45, HD3000 (Core i5 2500K)

              Comment


              • #77
                Originally posted by darkbasic View Post

                The point is that I never felt the need to disable COW in ZFS, while the opposite is true with btrfs. With ZFS I do use COW for virtual machines with ext4-like performance, just a little bit of tuning is needed: http://www.linuxsystems.it/2018/05/o...t4-benchmarks/
                In my tests depending on the workload the performance of a VM over a ZVOL could get very bad. But it's true that in ZFS random writes have generally a lesser penalty than in Btrfs, presumably due to the block-based nature of ZFS where such a write will basically "just" need to reserve a new block and recalculate its checksum, while in Btrfs it may in some cases involve splitting an extent.

                Comment

                Working...
                X