OpenVPN 2.6.0 low throughput

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • bhandari_nikhil
    Junior Member
    • Jul 2024
    • 3

    OpenVPN 2.6.0 low throughput

    I am using OpenVPN 2.6.0 site to site on custom hardware. I have ixia tester connected on both ends on LAN ports pumping TCP traffic to the other end. The setup is something like this:

    Ixia <= LAN => Machine 1 (Server) <== WAN ==> Machine 2 (Client) <== LAN ==> Ixia

    It is giving me ~800 Mbps bidirectional (so ~400 Mbps) where all links are 1G links. On this same setup, I earlier had OpenVPN 2.4.7 running on both ends and it was giving ~1.2 Gbps bidirectional (so ~600 Mbps). I am using OpenSSL 3.0 with OpenVPN 2.6.0 but even if I use OpenSSL 1.1.1v with OpenVPN 2.6.0, the throughput degradation is same. AES-256-GCM is the chosen cipher in both cases.

    Can someone explain what could be the reason for the degradation when everything else is constant ? I can send the server and client configs, if needed, but they are the same except that in OpenVPN 2.6.0, we are using "compress migrate" option.
  • jonathanlee
    Junior Member
    • Apr 2024
    • 6

    #2
    Do you have a crypt chip if so have you enabled off boarding to your crypt chip?

    Comment

    • bhandari_nikhil
      Junior Member
      • Jul 2024
      • 3

      #3
      No, I am not offloading to any crypt chip. I have not enabled DCO.

      Comment

      • bhandari_nikhil
        Junior Member
        • Jul 2024
        • 3

        #4
        After some trials, found that the following commit is causing the performance degradation:
        This function does most of the state transitions in the TLS state machine. Moving it into its own function removes an intention area and makes tls_process function easier to understand as the loop ...


        Especially, the following change causes the degradation:

        + /* If any of the state changes resulted in the to_link buffer being
        + * set, we are also active */
        + if (to_link->len)
        + {
        + return true;
        + }

        Can someone explain why this change was introduced ?

        Comment

        • Pippin
          Administrator
          • Dec 2023
          • 18

          #5
          Originally posted by bhandari_nikhil
          After some trials, found that the following commit is causing the performance degradation:
          Can someone explain why this change was introduced ?
          Hi, would you be willing to ask this on the mailing list?
          You can subscribe here:


          Please also link to this thread if you do subscribe.

          Thanks.

          Comment

          • Pippin
            Administrator
            • Dec 2023
            • 18

            #6
            Hi again,

            I just saw you already commented on the commit.

            Thanks.
            Last edited by Pippin; 2024-08-07, 12:18 PM.

            Comment

            Working...
            😀
            😂
            🥰
            😘
            🤢
            😎
            😞
            😡
            👍
            👎