Netgate Discussion Forum
    • Categories
    • Recent
    • Tags
    • Popular
    • Users
    • Search
    • Register
    • Login

    Maximized the Speed of Non-Gigabit Internet Connection

    Scheduled Pinned Locked Moved Traffic Shaping
    7 Posts 5 Posters 2.1k Views
    Loading More Posts
    • Oldest to Newest
    • Newest to Oldest
    • Most Votes
    Reply
    • Reply as topic
    Log in to reply
    This topic has been deleted. Only users with topic management privileges can see it.
    • N
      nelioromao
      last edited by

      Tips from an engineer at Ookla

      https://medium.com/speedtest-by-ookla/engineer-maximizes-internet-speed-story-c3ec0e86f37a

      1 Reply Last reply Reply Quote 0
      • GruensFroeschliG
        GruensFroeschli
        last edited by

        Might be worth noting that the linked article describes how to configure QoS rules in pfSense.

        We do what we must, because we can.

        Asking questions the smart way: http://www.catb.org/esr/faqs/smart-questions.html

        1 Reply Last reply Reply Quote 0
        • M
          moscato359
          last edited by

          This is speed as defined by latency and not throughput. This actually reduces throughput.

          1 Reply Last reply Reply Quote 0
          • H
            Harvy66
            last edited by

            @moscato359:

            This is speed as defined by latency and not throughput. This actually reduces throughput.

            Depends. In the simplistic case of one or few flows consuming bandwidth, yes, he shaved off about 0.5Mb/s of his 13Mb/s. In the complex case of many flows, a tail-drop FIFO degrades to about 80% of the link rate because of current TCP implementations. BBR should fix most of that.

            Codel, Red, and other AQMs tend to increasing the average bandwidth while minimizing packetloss and latency, under load. This is even more pronounced with high RTT routes where recovering of many lossed packets can take a while. Using an AQM like Codel reduces loss, and minimizing latency, which not only allows the flow to recover more quickly, but also reduces the aggressiveness of how much the flow slows down.

            The first rule of empirical evidence is to know what you're measuring.

            1 Reply Last reply Reply Quote 0
            • M
              moscato359
              last edited by

              Fair enough.

              I can see that as valid.

              1 Reply Last reply Reply Quote 0
              • M
                micropone
                last edited by

                i got the comcast gigabit he is saying turn OFF traffic shaping?

                with TS removed I got bad bufferfloat F

                with it on set for 950 i get an C

                1 Reply Last reply Reply Quote 0
                • H
                  Harvy66
                  last edited by

                  I would set it to 930. iPerf on my LAN shows about 940Mb/s logical, and about 970Mb/s raw. Toss in some inter-frame overhead and you're at 1Gb/s.

                  1 Reply Last reply Reply Quote 0
                  • First post
                    Last post
                  Copyright 2025 Rubicon Communications LLC (Netgate). All rights reserved.