查看文章

The transport performance in Mobile Internet is crucial to user experience and directly affects the revenue of providers for Internet applications. This dissertation focuses on network access and congestion control, two key factors affecting transport performance, and conducts research from three aspects: network access measurement, single-path congestion control, and multipath congestion control. For network access measurement, we conduct a measurement study across four cities to analyze the characteristics of the connectivity based on provider-managed WiFi in mobile scenarios. For single-path congestion control, we propose an enhanced BBRv2 algorithm based on delay information to improve the performance of BBRv2 in mobile scenarios. For multipath congestion control, we propose a multipath congestion control algorithm (CCA) selection mechanism, which leverages machine learning to dynamically select suitable CCAs for subflows to improve the performance of multipath transport.