Announcement

Collapse
No announcement yet.

Out-Of-The-Box 10GbE Network Benchmarks On Nine Linux Distributions Plus FreeBSD 12

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

  • #11
    FreeBSD has always dominated Linux, it is a go to OS for DDoS mitigation hardware and ZFS clusters, Linux's TCP/IP stack cannot compete and breaks down even under miedcore attacks. The benchmarks in this article don't seem to reveal FreeBSD's true superiority, tests with more common ethernet card would be welcomed.

    10GbE is still a dumb and expensive choice for typical home enduser, hard to believe that its been more than 15 years and motherboards are still equipped with 1GbE, for $50 bucks and some hunting on ebay you can get yourself 40GbE InfiniBand setup, stuff that is used in supercomputers, no need to buy stupid overpriced ethernet switches as LinusTechTips would wash down to its herd audience.

    Comment


    • #12
      Originally posted by hax0r View Post
      FreeBSD has always dominated Linux, it is a go to OS for DDoS mitigation hardware and ZFS clusters, Linux's TCP/IP stack cannot compete and breaks down even under miedcore attacks. The benchmarks in this article don't seem to reveal FreeBSD's true superiority, tests with more common ethernet card would be welcomed.
      There is nothing to reveal.

      no need to buy stupid overpriced ethernet switches as LinusTechTips would wash down to its herd audience.
      We get it, you watch LinusTechTips.

      Comment


      • #13
        Originally posted by hax0r View Post
        FreeBSD has always dominated Linux, it is a go to OS for DDoS mitigation hardware and ZFS clusters, Linux's TCP/IP stack cannot compete and breaks down even under miedcore attacks. The benchmarks in this article don't seem to reveal FreeBSD's true superiority, tests with more common ethernet card would be welcomed.
        Well, FreeBSD dominated only in file system (ZFS) so far. However, it seems it's going to change, because they're aiming at Linux' ZFS implementation. You should tell them they're going to make a mistake. When comes to more TCP tests it will be interesting to see how (or if) CPU mitigations affect benchmarks. Keep in mind some Linux distributions have debbuging enabled, so it affects performance in some cases.

        Comment


        • #14
          As always, CentOS's out-of-the box power profile decreases performance significantly. If you set your tuned profile, you will see much better performance:

          Last week I spent entirely too much time trying to track down a performance issue for the AArch64/ARM64 build of CentOS. While we don’t and won’t do performance comparisons or optimizations, this was fully in the realm of “something’s wrong here”. After a bit of digging, this issued turns out to impact just about everyone running CentOS on their servers who isn’t doing custom performance tuning.

          Comment


          • #15
            Originally posted by hax0r View Post
            FreeBSD has always dominated Linux, it is a go to OS for DDoS mitigation hardware and ZFS clusters, Linux's TCP/IP stack cannot compete and breaks down even under miedcore attacks. The benchmarks in this article don't seem to reveal FreeBSD's true superiority, tests with more common ethernet card would be welcomed.
            At first I thought you were trolling. However, I would like to verify this myself. Could you elaborate more about this? Are there any benchmarks you are aware of? It seems Linux out of the box is optimised for slower devices and needs little tuning for 10GbE and faster:

            If you are trying to achieve a maximum performance with 10Gb or 40Gb NICs in RHEL or similar  prepare yourself to a battle. This article is for experienced users, don’t mess up with default k…


            Comment


            • #16
              Huh? Out of the box makes no sense to me. I’m not sure you’re comparing the performance of these OS’s. In fact I’m sure you’re not. Without setting all the net. kernel parameters in /etc/systemctl.conf all the same, you’re just testing some generic kernel params likely optimized for 1GBE or some combo and some Random machine(s). To prevent comparing apples to apple trucks all the net. optimization parameters should standardized for 10gbe, be tested on the machine, optimized on the fastest distribution, and retested on all the distro’s

              Comment

              Working...
              X