Announcement

Collapse
No announcement yet.

There Is Now A WireGuard Benchmark For Testing Linux Networking Performance

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

  • There Is Now A WireGuard Benchmark For Testing Linux Networking Performance

    Phoronix: There Is Now A WireGuard Benchmark For Testing Linux Networking Performance

    With WireGuard added to the Linux 5.6 kernel and it being back-ported to Ubuntu 20.04 LTS and its tools getting packaged up by more Linux distributions, it's finally the year of WireGuard. With its usage set to skyrocket as supported kernels and the WireGuard utilities become available out-of-the-box on more distributions, there is now a WireGuard benchmark for stressing the kernel and its support...

    http://www.phoronix.com/scan.php?pag...uard-Benchmark

  • #2
    Great. So now I'll have one more thing mocking the lousy client my employer makes me use :P

    Comment


    • #3
      This looks extremely useful. I've been struggling to find the right Wireguard MTU.
      • With MTU lower than 1370 I get 500mb download speed and 30 mb upload speed.
      • With MTU Higher than 1370 I get 30mb download speed and 30 mb upload speed.
      It just make no sense to me.

      Comment


      • #4
        This new benchmark won't help you figure out anything at all related to your MTU and has no relation whatsoever. They are not related topics. But, to help with your unrelated issue:

        WireGuard has 60 or 80 bytes overhead. Take your normal MTU and subtract 60 or 80. Then you get your optimal WireGuard MTU.

        For more help with networking stuff, #wireguard in Freenode can help you.

        Comment


        • #5
          Originally posted by Zeioth View Post
          This looks extremely useful. I've been struggling to find the right Wireguard MTU.
          • With MTU lower than 1370 I get 500mb download speed and 30 mb upload speed.
          • With MTU Higher than 1370 I get 30mb download speed and 30 mb upload speed.
          It just make no sense to me.
          I don't seem to have notes on it, but I recall 1370 I think as a fairly common value, even though default/normal MTU goes up to 1500 iirc, was something to do with overheads and possible network hardware or something like that. If you want to do high MTU values, there is jumbo frames with MTU 9000, but you need to specifically toggle that support. I believe that's only useful for large transfers, so while a benchmark might tell you it's faster, you'd get a poorer experience with most network traffic on the internet. Might be due to an increase in latency from the larger packet size.

          Comment


          • #6
            It would fun making a similar teste suite for openvpn, using the sabe tecnics. And then compare them.

            Comment


            • #7
              Originally posted by polarathene View Post

              I don't seem to have notes on it, but I recall 1370 I think as a fairly common value, even though default/normal MTU goes up to 1500 iirc, was something to do with overheads and possible network hardware or something like that. If you want to do high MTU values, there is jumbo frames with MTU 9000, but you need to specifically toggle that support. I believe that's only useful for large transfers, so while a benchmark might tell you it's faster, you'd get a poorer experience with most network traffic on the internet. Might be due to an increase in latency from the larger packet size.
              For some reason the value I was defining in my VPN GUi wasn't the same set in the command "ip link". 1370 seems to actually work fine for both upload and download speed. It's been pretty cool to learn about Jumbo MTU, thanks!

              Comment


              • #8
                Originally posted by Zeioth View Post

                For some reason the value I was defining in my VPN GUi wasn't the same set in the command "ip link". 1370 seems to actually work fine for both upload and download speed. It's been pretty cool to learn about Jumbo MTU, thanks!
                zx2c4 Jason Donenfeld may have more recommendations over WG MTU size.
                Michael Larabel
                http://www.michaellarabel.com/

                Comment


                • #9
                  Originally posted by zx2c4 View Post
                  This new benchmark won't help you figure out anything at all related to your MTU and has no relation whatsoever. They are not related topics. But, to help with your unrelated issue:

                  WireGuard has 60 or 80 bytes overhead. Take your normal MTU and subtract 60 or 80. Then you get your optimal WireGuard MTU.

                  For more help with networking stuff, #wireguard in Freenode can help you.

                  MTU calculation is a pretty big deal performance wise, and it could be easily automatized on the client side by vpn providers.

                  EDIT: Apparently you don't need to calculate your MTU. If you enable IPV6 on your router you get automatically max performance.
                  Last edited by Zeioth; 04-21-2020, 07:02 PM.

                  Comment

                  Working...
                  X