Performance Analysis of TCP Congestion Control with and without PEPs over Satellite Links

ABSTRACT Satellite communication presents unique challenges for the Transmission Control Protocol (TCP), including high latency, significant propagation delays, and frequent packet losses that are not necessarily indicative of network congestion. These factors often lead to suboptimal performance for traditional TCP algorithms, which are primarily designed for terrestrial environments. This study evaluates the impact of Performance Enhancing Proxies (PEPs) on a commercial satellite network, focusing on four TCP congestion control algorithms: Cubic, BBR, Hybla, and PCC. In the absence of a PEP, all four algorithms achieve similar steady-state throughputs of approximately 70 Mb/s; however, notable differences emerge in their start-up throughput behaviors. For instance, the use of a PEP dramatically improves the start-up throughput of TCP Cubic commonly the default congestion control algorithm for many Internet servers yielding a performance increase of nearly threefold. Overall, the inclusion of a PEP standardizes performance across different congestion control schemes, highlighting its effectiveness in mitigating the limitations of TCP over satellite links.