Announcement

Collapse
No announcement yet.

Bufferbloat Is Still Being Fought In Linux Kernel, Another Big Improvement Queued

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

  • Bufferbloat Is Still Being Fought In Linux Kernel, Another Big Improvement Queued

    Phoronix: Bufferbloat Is Still Being Fought In Linux Kernel, Another Big Improvement Queued

    Bufferbloat is the excess buffering of packets resulting in high latency, jitter, and lower network throughput. There's been efforts to battle bufferbloat within the Linux kernel going back a long time while this week another new patch has surfaced...

    http://www.phoronix.com/scan.php?pag...16-Bufferbloat

  • #2
    In what kind of scenario(s) would a desktop user benefit from this improvement?

    Comment


    • #3
      Originally posted by ArthurBorsboom View Post
      In what kind of scenario(s) would a desktop user benefit from this improvement?
      Transferring data? *runs*

      seriously though - lower power usage will be on the cards when doing 10gbit speeds...

      Comment


      • #4
        Which buffers are we talking about here? Network, memory, I/O, ..?

        Comment


        • #5
          Originally posted by GreatEmerald View Post
          Which buffers are we talking about here? Network, memory, I/O, ..?
          network

          Comment


          • #6
            Originally posted by ArthurBorsboom View Post
            In what kind of scenario(s) would a desktop user benefit from this improvement?
            Reduces CPU load, when someone is flooding you over UDP?, by dropping packets more aggressively.

            Comment


            • #7
              Better protection against DDoS?

              Comment


              • #8
                Originally posted by andrei_me View Post
                Better protection against DDoS?
                IIRC, one of the prime realms of bufferbloat is with TCP connections, which like to have reasonably timely feedback as to how the packets are flowing. Normally buffering is considered good, but when too much buffering happens, the TCP feedback packets get delayed in the buffers. When that happens, TCP starts to interpret those delays as lost packets and requests resends, while the packets are still in-flight - in the buffers. The re-requested packets then contend with the original packets and with the TCP control packets, etc, etc.

                Comment


                • #9
                  Originally posted by ArthurBorsboom View Post
                  In what kind of scenario(s) would a desktop user benefit from this improvement?
                  LAN with a NAS and multiple users wanting to use it at the same time. Lower cpu load on NAS and router usually mean faster LAN transfers.

                  But really this is mostly about the router or the NAS's performance, not your PC's.

                  Comment


                  • #10
                    Originally posted by ArthurBorsboom View Post
                    In what kind of scenario(s) would a desktop user benefit from this improvement?
                    The problem of bufferbloat is not specific for endpoint systems, like servers or desktops, but much more about routers and such. Fixing bufferbloat in edge routers, home routers and all the equipment between the server and the client is where most of the benefits will be evident, as it is a problem that adds congestion, latency and unstability to all kinds of long-lived connections, like streaming, big file transfers or even gaming (funnily enough, the very same problems that were supposed to be fixed by growing the buffers).

                    The tricky part here is to get fixes like this one into the myriad of routers/switches/younameit between the endpoints. Hopefuly most of them are running some kind of linux kernel that will end up being updated (cross your fingers)

                    Comment

                    Working...
                    X