ABSTRACT

End-to-end congestion control algorithms (CCAs) are expected to perform well in any Internet path, including those paths with low-earth orbit (LEO) satellite links. In this paper, we conduct a performance study on various CCAs in an operational LEO satellite network. We find that existing CCAs struggle to deal with the drastic network variations caused by the mobility of LEO satellites, resulting in poor link utilization or high latency. Further, through an in-depth analysis, we identify the fundamental challenge is that existing end-to-end CCAs detect network congestion based on performance changes observed on the sender, but the unique LEO mobility can involve massive non-congestion performance changes which seriously mislead CCA behaviors. Finally, we explore and discuss possible solutions to mitigate the misleading effects of LEO mobility.

端到端拥塞控制算法（CCAs）预计能够在任何互联网路径上表现良好，包括低地球轨道（LEO）卫星链路。本文对在一个实际的LEO卫星网络中不同的拥塞控制算法进行了性能研究。研究发现，现有的拥塞控制算法难以应对由LEO卫星的运动引起的剧烈网络变化，导致链路利用率低或延迟过高。通过深入分析，我们识别出一个根本性挑战：现有的端到端拥塞控制算法基于发送端观察到的性能变化来检测网络拥塞，但LEO卫星的独特移动性可能涉及大量非拥塞性性能变化，严重误导了拥塞控制算法的行为。最后，我们探讨并讨论了减轻LEO移动性误导效应的可能解决方案。

Note

LEO的高速移动性 -> 性能指标改变，但是原因不是拥塞，而是高速移动产生的副作用（“非拥塞网络变化”） -> LEO CC 认为拥塞发生并进行“本毫无必要”的调整 -> 链路利用率低 / 延迟高