Announcement

Collapse
No announcement yet.

Linux 6.4 Device Mapper To See Improved Concurrent I/O Performance

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

  • Linux 6.4 Device Mapper To See Improved Concurrent I/O Performance

    Phoronix: Linux 6.4 Device Mapper To See Improved Concurrent I/O Performance

    Thanks to forthcoming work out of Red Hat, the Linux 6.4 Device Mapper (DM) code is expected to see some optimization work that can significantly benefit concurrent I/O performance... In one case at least acquiring buffers now about 25 times faster...

    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
    I never noticed any perf impact from a DM device but 25x sounds not bad. Are we talking about LVM, Softraid and the like?

    Comment


    • #3
      Originally posted by Anux View Post
      I never noticed any perf impact from a DM device but 25x sounds not bad. Are we talking about LVM, Softraid and the like?
      It is 25x for a specific part of an operation. That does not translate to 25x read speed, write speed or anything else that is a meaningful benchmark.

      Most likely it would be a few percent on any real world benchmark at most. Maybe a bit more on a very directed synthentic benchmark. Only by profiling that specific function would you see a 25x improvement.

      See also Amdahl's law

      Comment


      • #4
        I wish this would help BTRFS...

        Comment


        • #5
          Oh yeah! Linux FTW! Years of hard work from developers and programmers all around the world is beginning to show awesome improvements!

          Comment


          • #6
            Originally posted by guglovich View Post
            I wish this would help BTRFS...
            Yeah dat. But we'll settle for Stratis...

            Comment


            • #7
              Originally posted by Vorpal View Post
              It is 25x for a specific part of an operation. That does not translate to 25x read speed, write speed or anything else that is a meaningful benchmark.
              Yes, most likely it won't result in more speed, but in less CPU load for the same task.

              Comment


              • #8
                Originally posted by Vorpal View Post
                It is 25x for a specific part of an operation. That does not translate to 25x read speed, write speed or anything else that is a meaningful benchmark.
                Ofcoure it says in the article:
                > benefit concurrent I/O performance... In one case at least acquiring buffers now about 25 times faster.
                And since my desktop is mostly sequential ... it might help in my server but my storage devices and network will limit things. That's why I wrote "not noticed perf impact", even raid 6 + luks is limited by network (1Gb/s).

                Comment

                Working...
                X