Experimental results

We choose the NS-3 simulation platform based on C++ to carry out the experiment. NS-3 is an easy-to-use discrete event network simulator. Our experimental machine uses a CPU of AMD R7-3700X@3.60GHz, 16G of RAM, and runs the operating system Ubuntu 20.04. This simulation experiment mainly includes three kinds of nodes: the Sever node that sends the data, the User node that receives the data and the Router node that forwards the data. On the network routing node, we choose FIFO+DropTail queue management method with a queue length of 100 packets. In the selection of the rate control strategy of the end node, we choose the NewReno scheme optimized based on the classical Reno scheme, the Cubic scheme which is widely used as the default scheme of the Linux system, and the BBR scheme which represents the new control idea, as a comparison scheme to compare the performance with the scheme proposed in this paper (represented as Nodectl). In order to be consistent with the comparison scheme and avoid the interference and influence of other factors, the Nodectl scheme proposed in this paper is the same as the three comparison schemes, and the data sender is selected to control the rate. We carry out simulation experiments from three aspects: start-up speed, throughput capacity and multi-transmission competition. We set the bandwidth of all server nodes and user nodes to 200Mbps, and the delay of sender access link and receiver access link to 5ms. The bandwidth and delay of the bottleneck link vary with the experimental requirements, and the maximum bandwidth of the bottleneck link is less than the bandwidth of the access link at both ends. The purpose of this experiment is to compare the performance of different rate control schemes in the case of limited network resources, so it is necessary to avoid end-node performance restrictions. We test the throughput performance of different schemes under different transmission data sizes. When testing other scenarios, the data size is set to 0, which means there is no limit on the amount of data sent, and the sender will continue to send data until the end of the experiment.In the process of simulation, the NewReno, Cubic and BBR schemes we use are all built-in versions of NS-3. We have implemented the algorithms described in this paper in NS-3. And in order to be consistent with the three comparison schemes, we choose the TCP transport protocol when implementing the Nodectl scheme. In fact, the Nodectl scheme proposed in this paper does not rely on the TCP transport protocol, it can be well applied to a variety of connection-oriented transport protocols driven by the sender or receiver.