PREPARING FOR A LOW-LATENCY FUTURE INTERNET

Abstract

In this thesis, we proposed environment-aware rate-based congestion control (EvaRate) , a rate-based algorithm that incorporates new buffer estimation and buffer management techniques, which allow EvaRate flows to infer its own and competing flows’ buffer occupancy. With this mechanism, an EvaRate flow is able to determine its operating environment. When in a low-latency (or benevolent) environment, it collaboratively regulates the bottleneck buffer occupancy with other EvaRate flows. When in a buffer-filling (or hostile) environment, it keeps a minimum number of packets in the bottleneck buffer for survival. The results suggest that EvaRate represents a promising path of congestion control: inferring the conditions of the network environment and reacting quickly. We believe that our work will allow the Internet to transition smoothly to a low-latency future without disrupting its current operation, as the proportion of low-latency TCP variant flows continues to increase.