Announcement

Collapse
No announcement yet.

Dbus-Broker 26 Released For High Performance D-Bus

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

  • Dbus-Broker 26 Released For High Performance D-Bus

    Phoronix: Dbus-Broker 26 Released For High Performance D-Bus

    With the BUS1 in-kernel IPC not panning out and not seeing any major code work in nearly two years, the user-space based, D-Bus compatible DBus-Broker remains the performant and current option for those looking at something faster and more reliable than D-Bus itself...

    http://www.phoronix.com/scan.php?pag...Dbus-Broker-26

  • #2
    Would anybody have the patience of explaining me what's wrong with the current implementation of D-Bus? I read that this is a better performance implementation, but in what scenario does D-Bus lack performance? As a user I have never noticed any particular issue...
    The Kernel implementation was supposed to offer something that works from very early boot stages, okay, I can maybe see something here, but what would I get with Dbus-broker?

    [EDIT]
    Found this https://dvdhrm.github.io/rethinking-...s-message-bus/
    Last edited by fguerraz; 20 January 2021, 08:55 AM.

    Comment


    • #3
      Originally posted by fguerraz View Post
      Would anybody have the patience of explaining me what's wrong with the current implementation of D-Bus? I read that this is a better performance implementation, but in what scenario does D-Bus lack performance? As a user I have never noticed any particular issue...
      The Kernel implementation was supposed to offer something that works from very early boot stages, okay, I can maybe see something here, but what would I get with Dbus-broker?

      [EDIT]
      Found this https://dvdhrm.github.io/rethinking-...s-message-bus/
      Or a video with Lennart Poettering
      https://www.youtube.com/watch?v=hAHvn5XOzoc

      Comment


      • #4
        It is not replacement for libdbus, it is replacement for dbus-daemon.

        Comment


        • #5
          Originally posted by fguerraz View Post
          ... in what scenario does D-Bus lack performance? As a user I have never noticed any particular issue...
          D-Bus performance is fine if overhead of the IPC is lost in the execution of the function, for example, to send a "STOP" command to your music player or tell GNOME shell to restart. Where it gets problematical is when you need to invoke a function thousands or hundreds of thousands of times a second and the overhead of using the IPC exceeds the amount of time you have to turn around the call. That is why you probably never noticed it, you haven't pushed it hard enough!

          It's the same reason why monolothic kernels like Linux will always beat micro-kernels performance wise, as they don't have the IPC overhead.

          Comment


          • #6
            Originally posted by browseria View Post

            D-Bus performance is fine if overhead of the IPC is lost in the execution of the function, for example, to send a "STOP" command to your music player or tell GNOME shell to restart. Where it gets problematical is when you need to invoke a function thousands or hundreds of thousands of times a second and the overhead of using the IPC exceeds the amount of time you have to turn around the call. That is why you probably never noticed it, you haven't pushed it hard enough!

            It's the same reason why monolothic kernels like Linux will always beat micro-kernels performance wise, as they don't have the IPC overhead.
            It is a cost that only bites you if you are using the API wrong.

            Comment


            • #7
              Originally posted by carewolf View Post
              It is a cost that only bites you if you are using the API wrong.
              only if you define "wrong" as "in time-critical scenarios"

              Comment


              • #8
                Originally posted by carewolf View Post

                It is a cost that only bites you if you are using the API wrong.
                The concept behind efforts like Dbus-Broker and BUS1 is that we shouldn't need a lot of different IPC solutions to meet our needs if it's just inefficient implementation that's forcing more diversity.

                Comment


                • #9
                  [QUOTE=fguerraz;n1233376]Would anybody have the patience of explaining me what's wrong with the current implementation of D-Bus? I read that this is a better performance implementation, but in what scenario does D-Bus lack performance? As a user I have never noticed any particular issue...
                  [EDIT]

                  that's because as a developer, you look at the current dbus performance and you know that it isn't even realistic to use it for semi-real-time use cases, unlike the IPC of macOS, Android and Windows. Thus the applications that would benefit from more performance don't even exist.

                  Comment

                  Working...
                  X