Announcement

Collapse
No announcement yet.

Docker Benchmarks: Ubuntu, Clear Linux, CentOS, Debian & Alpine

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

  • Docker Benchmarks: Ubuntu, Clear Linux, CentOS, Debian & Alpine

    Phoronix: Docker Benchmarks: Ubuntu, Clear Linux, CentOS, Debian & Alpine

    The latest target of our Linux benchmarking at Phoronix are running various performance benchmarks under different Docker operating system images. The images used for benchmarking were the latest of Ubuntu, Clear Linux, CentOS, Debian, and Alpine while comparing the benchmark results to running on the bare metal host.

    http://www.phoronix.com/vr.php?view=23804

  • #2
    Hi there, long time reader, really nice benchmark

    What about other docker filesystems? The best atm is overlay, in terms of io.

    Comment


    • #3
      Originally posted by theonlydoo View Post
      Hi there, long time reader, really nice benchmark

      What about other docker filesystems? The best atm is overlay, in terms of io.
      While we're at it, what about OSes that are usually run inside Docker containers, like CoreOS?
      It's great to see Docker benchmarked at all, but what I (and theonlydoo) mean, is standard tests don't paint a very accurate picture.

      Comment


      • #4
        Nice benchmarks. It would also be nice to see how systemd-nspawn competes with docker regarding performance.

        Comment


        • #5
          Thanks for the nice benchmarks.

          Comment


          • #6
            Nice benchmark. However there are a couple of issues.

            The docker parameters used are not specified (would be useful to have the Dockerfile and docker command line parameters available and also the docker configuration on the file system, for instance the file system used to store the docker images). Also usually the best option anyway is to use docker volumes instead of writing direct inside the container image. Another possible issue the cpu scaling governor is indicated to be on powersave, this can be a issue along the benchmarks causing more throttling than using for instance the performance option?

            Comment


            • #7
              Originally posted by defaultUser View Post
              Nice benchmark. However there are a couple of issues.

              The docker parameters used are not specified (would be useful to have the Dockerfile and docker command line parameters available and also the docker configuration on the file system, for instance the file system used to store the docker images). Also usually the best option anyway is to use docker volumes instead of writing direct inside the container image. Another possible issue the cpu scaling governor is indicated to be on powersave, this can be a issue along the benchmarks causing more throttling than using for instance the performance option?
              Michael tends to use the default options of the software that he tests, which is only fair. (if there is a magic "make it go faster" option, that should really be the default after all, and most users won't switch from defaults)

              in this case he mentions he uses the standard ubuntu + latest official docker release (including those standard settings that come with that)

              Comment


              • #8
                clearLinux looks like it is doing pretty well, at least with caffe. If I wanted to adopt those modifications into Kubuntu what would be involved?

                Comment


                • #9
                  Originally posted by arjan_intel View Post

                  Michael tends to use the default options of the software that he tests, which is only fair. (if there is a magic "make it go faster" option, that should really be the default after all, and most users won't switch from defaults)

                  in this case he mentions he uses the standard ubuntu + latest official docker release (including those standard settings that come with that)
                  Even assuming default parameters this still an issue. For instance for RHEL and Fedora the default docker file system is different from the Ubuntu one. However this only cover the default docker installation on the host system. The benchmarking process to be done , probably, required Dockerfiles to specify the installation of software and to define where to run these tests.

                  Comment


                  • #10
                    Originally posted by Goddard View Post
                    clearLinux looks like it is doing pretty well, at least with caffe. If I wanted to adopt those modifications into Kubuntu what would be involved?
                    Clear Linux targets Haswell+ Intel CPUs exclusively, so they can choose instructions that are fast on that hardware and absent elsewhere. Kubuntu is a general-purpose distribution, so it can't do this.

                    Clear Linux also does profile-guided optimization - watching a running program to see where it spends its time. Profile-guided optimization would probably help Kubuntu, but again, different CPUs are faster at different tasks, so which CPU should it optimize for?

                    Comment

                    Working...
                    X