| （英） |
End-to-end communication quality on the Internet depends on the congestion control algorithms (CCAs) used in transport protocols. Representative CCAs include CUBIC and BBR. Previous studies have reported that when these different CCAs compete on a bottleneck link, throughput unfairness occurs among flows depending on the buffer size of the bottleneck router. To address this issue, we have previously proposed a method to improve fairness between flows by using machine learning to identify the type of CCA at intermediate devices and utilizing programmable data plane capabilities to establish separate buffers for each CCA to control transmission. However, with this packet control method, while throughput between CCAs is fair when the number of flows per CCA varies, fairness in throughput between flows was not addressed. Therefore, in this study, we propose a method that adds a feature to the FQ-VHT algorithm proposed in previous research to set the number of packet transmissions based on the number of flows, thereby ensuring fair throughput among flows. Furthermore, to verify the effectiveness of the proposed method, we conducts validation using the network emulator Mininet. |