Announcement

Collapse
No announcement yet.

FreeBSD 12.0 Performance Against Windows & Linux On An Intel Xeon Server

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

  • FreeBSD 12.0 Performance Against Windows & Linux On An Intel Xeon Server

    Phoronix: FreeBSD 12.0 Performance Against Windows & Linux On An Intel Xeon Server

    Last week I posted benchmarks of Windows Server 2019 against various Linux distributions using a Tyan dual socket Intel Xeon server. In this article are some complementary results when adding in the performance of FreeBSD 11.2 against the new FreeBSD 12.0 stable release for this leading BSD operating system. As some fun benchmarks to end out 2018, here are the results of FreeBSD 11.2/12.0 (including an additional run when using GCC rather than Clang) up against Windows Server and several enterprise-ready Linux distributions.

    http://www.phoronix.com/vr.php?view=27358

  • #2
    Sometimes I have the sensation the PTS is more optimized for Linux rather that FreeBsd, however is a matter of fact that FreeBsd should be work better on the TCP/IP stack and everything related with the internet and server streaming, however I haven't seen many of these test and perhaps PTS is not able to perform this kind of measurement. The PST clearly shows that linux distros performs better than bsd globally, we need just to see if is true that bsd is better in balancing internet...

    Comment


    • #3
      Originally posted by Danielsan View Post
      Sometimes I have the sensation the PTS is more optimized for Linux rather that FreeBsd, however is a matter of fact that FreeBsd should be work better on the TCP/IP stack and everything related with the internet and server streaming, however I haven't seen many of these test and perhaps PTS is not able to perform this kind of measurement. The PST clearly shows that linux distros performs better than bsd globally, we need just to see if is true that bsd is better in balancing internet...
      I will have some 10G network tests soon. But, no, it's not a matter of PTS itself being optimized for Linux or BSD or not but rather the upstream software PTS is benchmarking.... A lot of these open-source projects aren't tested or tuned by their developers for BSDs. PTS is simply testing the code with not really interfering or modifying it. Just as many open-source projects aren't yet tuned/tested on POWER9.

      IX Systems is making use of PTS as part of their system workflow now though so they have been contributing some improvements for getting more benchmarks working on BSD, etc.
      Michael Larabel
      http://www.michaellarabel.com/

      Comment


      • #4
        Originally posted by Michael View Post

        I will have some 10G network tests soon. But, no, it's not a matter of PTS itself being optimized for Linux or BSD or not but rather the upstream software PTS is benchmarking.... A lot of these open-source projects aren't tested or tuned by their developers for BSDs. PTS is simply testing the code with not really interfering or modifying it. Just as many open-source projects aren't yet tuned/tested on POWER9.

        IX Systems is making use of PTS as part of their system workflow now though so they have been contributing some improvements for getting more benchmarks working on BSD, etc.
        I got it! Thanks for the clarification!
        Now I am really curious to see the new incoming 10G networks test!

        Comment


        • #5
          I'd really like to see FreeBSD with clang 7.0.1 as 6.0.1 is a whole release behind stable. Maybe the differences in performance can be partially explained by this?

          Comment


          • #6
            Looks like the one consistent answer is "Don't use Windows Server unless it's required by the applications you want to run".

            Comment


            • #7
              Originally posted by ssokolow View Post
              Looks like the one consistent answer is "Don't use Windows Server unless it's required by the applications you want to run".
              Or upper management. Executives understand corporate contracts, they don't understand technology. Finance and technology speak different languages.

              Largely been the case since the beginning of commodity computing. Microsoft Windows has improved its stability, security, and performance over the years, but its performance still lags behind the competition almost without regard to what the competition is (Legacy UNIX, BSD, Linux, Oracle, Mainframes) in the back room/data centers. Microsoft tried to push Windows into HPC a while back but from the statistics on what the largest/fastest supercomputers run, I'd say they've been pretty unsuccessful.

              Comment


              • #8
                Originally posted by Danielsan View Post
                Sometimes I have the sensation the PTS is more optimized for Linux rather that FreeBsd, however is a matter of fact that FreeBsd should be work better on the TCP/IP stack and everything related with the internet and server streaming, however I haven't seen many of these test and perhaps PTS is not able to perform this kind of measurement.
                No, bsd is not better at all. It performs worse on high end TCP hardware. This is the reason why the biggest providers use Linux. Google, Youtube, Netflix, Facebook and so on. However, when comes to PTS benchmarks they're sometimes meaningless, because they're made by incompetent person who has no clue about real benchmarking. He's not even able to properly setup systems for specific benchmark like Apache and I'm feeling he will mess up the TCP benchmark as well.

                Comment


                • #9
                  Beside the fact that you are a typical troll is quite known that Netflix for example uses FreeBSD: https://www.freebsdfoundation.org/testimonial/netflix/#!

                  Comment


                  • #10
                    Originally posted by OpenSourceAnarchist View Post
                    I'd really like to see FreeBSD with clang 7.0.1 as 6.0.1 is a whole release behind stable. Maybe the differences in performance can be partially explained by this?
                    13-CURRENT has LLVM7 in base.

                    If you are up to recompiling world and kernel, you could recompile FreeBSD 12-STABLE itself using LLVM7 from ports. Tried it for shits and giggles. Doable. When you exclude unneeded components in /etc/src.conf you'll get away with much less mandatory fixes to the source files.

                    Ports, better to use 6.01 from the base. A lot of the ports would still fail with v7. If you go that way, it'd be nice if you shared your fixes with upstream.
                    Last edited by aht0; 01-02-2019, 08:07 PM.

                    Comment

                    Working...
                    X