Exploring the "Internet from space" with Hypatia

ABSTRACT

SpaceX, Amazon, and others plan to put thousands of satellites in low Earth orbit to provide global low-latency broadband Internet. SpaceX’s plans have matured quickly, such that their under-deployment satellite constellation is already the largest in history, and may start offering service in 2020.

SpaceX、Amazon等公司计划在低地球轨道部署数千颗卫星，以提供全球低延迟的宽带互联网服务。SpaceX的计划发展迅速，其目前的卫星星座规模已成为历史上最大的未完全部署星座，并预计将于2020年开始提供服务。

The proposed constellations hold great promise, but also present new challenges for networking. To enable research in this exciting space, we present Hypatia, a framework for simulating and visualizing the network behavior of these constellations by incorporating their unique characteristics, such as high-velocity orbital motion.

这些拟议中的卫星星座具有巨大的潜力，但也带来了网络方面的新挑战。为了促进这一前沿领域的研究，我们提出了Hypatia，一个用于模拟和可视化这些星座网络行为的框架，能够考虑其独特特征，如高速轨道运动。

Using publicly available design details for the upcoming networks to drive our simulator, we characterize the expected behavior of these networks, including latency and link utilization fluctuations over time, and the implications of these variations for congestion control and routing.

通过利用公开的网络设计细节驱动我们的模拟器，我们对这些网络的预期行为进行了表征，包括延迟和链路利用率随时间波动的情况，以及这些波动对拥塞控制和路由选择的影响。

INTRODUCTION

The Internet is potentially taking “one giant leap” into space, with plans afoot for large satellite constellations to blanket the globe with low-latency broadband Internet. Numerous competitors have disclosed efforts along these lines, including SpaceX [70], Amazon [8], and Telesat [74]. With 400+ satellites already in orbit, and an increasing launch cadence, SpaceX’s Starlink constellation is promising limited availability of its Internet service already in 2020 [23]. It is thus unsurprising that these ambitious plans for an “Internet from space” have captured the public imagination [10, 28, 55, 62, 75].

互联网有可能迈出“巨大的一步”进入太空，多个计划正在推进，旨在通过大型卫星星座为全球提供低延迟的宽带互联网服务。包括SpaceX [70]、Amazon [8]和Telesat [74]等在内的众多竞争者已经披露了相关努力。SpaceX的Starlink星座已经在轨道上部署了400多颗卫星，并且随着发射频率的增加，预计其互联网服务将在2020年开始有限提供 [23]。因此，这些雄心勃勃的“太空互联网”计划能够引发公众的广泛关注并不令人惊讶 [10, 28, 55, 62, 75]。

While the use of satellites for Internet connectivity is as old as the Internet itself 1 , the under-construction constellations differ fundamentally from past efforts. The distinctions are rooted in the recent improvements in enabling technologies, as well as the goals, but manifest themselves deeply in the design. Unlike existing satellite networks [35–37], the new ones are targeting not only traditional niches such as shipping, satellite telephony, and limited connectivity for rural areas, but also mass market broadband that not only addresses these global coverage issues, but also competes with current terrestrial networks in many markets.

尽管卫星用于互联网连接的历史与互联网本身几乎同龄[1]，但正在建设中的星座与以往的努力在根本上有所不同。这些区别源于近期启用技术的改进以及目标的变化，且在设计上表现得尤为明显。与现有的卫星网络[35–37]不同，新的卫星星座不仅仅瞄准传统的市场细分领域，如航运、卫星电话和偏远地区的有限连接，还将目标扩展至大众市场的宽带服务，这不仅解决了全球覆盖的问题，还将在许多市场中与当前的地面网络展开竞争。

The first design manifestation of this goal is scale: to provision enough access bandwidth for their larger target user population, the new systems need many more satellites than past ones. Starlink, with its hundreds of satellites, is already the largest ever satellite fleet in space history, but eventually, the largest planned constellations will each comprise thousands of satellites [8, 70]. This has only become possible due to favorable trends in space technology, primarily, satellite miniaturization, and reduced launch costs.

这一目标的首个设计体现是规模：为了为其更大目标用户群体提供足够的接入带宽，新系统需要比以往更多的卫星。Starlink凭借其数百颗卫星，已经成为太空历史上规模最大的卫星星座，但最终，计划中的最大卫星星座将由数千颗卫星组成[8, 70]。这一切的实现仅在于太空技术的有利发展趋势，主要包括卫星小型化和发射成本的降低。

The goal of competing outside traditional niches has another important design consequence: operation in low Earth orbit (LEO), at most 2,000 km above Earth’s surface. This is essential for latencies to be comparable to terrestrial networks instead of the hundreds of milliseconds that geostationary orbits (GEO) incur. LEO operation, in turn, further reinforces the need for large scale: from GEO, each satellite is visible to a large terrestrial area, but bringing satellites closer to the Earth necessarily reduces each satellite’s coverage.

在传统细分市场之外竞争的目标带来了另一个重要的设计影响：运行于低地球轨道（LEO），即距离地球表面最多2,000公里。这对于确保延迟能够与地面网络相媲美至关重要，而不是像静止轨道（GEO）那样带来数百毫秒的延迟。LEO运行反过来又进一步强化了大规模部署的需求：从GEO轨道来看，每颗卫星的覆盖范围涉及较大的地面区域，但将卫星部署得更接近地球，必然会减少每颗卫星的覆盖范围。

Large LEO constellations promise global coverage at low-latency and high-bandwidth. However, realizing the full potential of these networks requires addressing new research challenges posed by their unique dynamics. In such constellations, each satellite orbits the Earth every ∼100 minutes, traveling at ∼27,000 kmph. This high-velocity movement of satellites creates not only high churn in the ground to satellite links, but also fluctuations in the structure of end-end paths as the satellites comprising the paths move.

大型低地球轨道（LEO）卫星星座有望提供低延迟和高带宽的全球覆盖。然而，要充分实现这些网络的潜力，需要解决其独特动态带来的新研究挑战。在此类星座中，每颗卫星约每100分钟绕地球一圈，飞行速度约为27,000公里每小时。这种高速运动不仅造成地面与卫星链路的高频率变动，还导致端到端路径结构的波动，因为构成这些路径的卫星在轨道上不断变化位置。

At HotNets 2018, three position papers [5, 29, 44] highlighted some of the networking challenges that could potentially arise in LEO networks, e.g., in end-end congestion control [5] and intraconstellation routing [29]. However, progress in precisely fleshing out these challenges and addressing them faces a substantial roadblock: lack of network analysis tools that incorporate the dynamic behavior of LEO networks. This creates a substantial risk that instead of networking research laying out the potential future trajectories for the industry, research will rather lag the industry’s rapid strides. Thus, to help accelerate research on LEO networks, we developed Hypatia 2 , an analysis framework with simulation and visualization modules. Hypatia provides a packet-level LEO network simulator based on ns-3, as well as several types of network visualizations based on Cesium [13], that serve to provide intuition about such networks.

在HotNets 2018会议上，三篇位置论文[5, 29, 44]突出了低地球轨道（LEO）网络中可能出现的一些网络挑战，例如端到端拥塞控制[5]和星座内部路由[29]。然而，在具体阐明这些挑战并加以解决方面，面临着一个重要的障碍：缺乏能够整合LEO网络动态行为的网络分析工具。这带来了一个重大风险，即研究可能无法跟上行业的快速发展，反而会滞后于行业的进展。因此，为了加速LEO网络的研究，我们开发了Hypatia 2，一个包含仿真和可视化模块的分析框架。Hypatia基于ns-3提供了一个数据包级别的LEO网络仿真器，并且基于Cesium [13]提供了几种类型的网络可视化，旨在为此类网络提供直观的理解。

We use Hypatia to analyze the three largest proposed LEO networks: Starlink, Kuiper, and Telesat. Our analysis uses the regulatory information these companies have filed with governing bodies like the Federal Communications Commission (FCC) in the United States, and the International Telecommunication Union (ITU). These filings [47–49, 68, 69, 72] disclose the orbital parameters that describe the structure of the planned constellations. Our simulations of these networks reveal the impact of LEO dynamics on varying path RTTs and packet reordering, as well as fluctuations in available bandwidth along end-end paths. We discuss the implications of these observations for congestion control and routing.

我们使用Hypatia分析了三大拟议中的LEO网络：Starlink、Kuiper和Telesat。我们的分析基于这些公司向监管机构（如美国联邦通信委员会FCC和国际电信联盟ITU）提交的监管信息。这些文件[47–49, 68, 69, 72]披露了描述计划星座结构的轨道参数。我们对这些网络的仿真揭示了LEO动态对路径往返时间（RTT）和数据包乱序的影响，以及沿端到端路径带宽波动的情况。我们讨论了这些观察结果对拥塞控制和路由的影响。

In summary, we make the following contributions:

• We lay out the case for building network analysis tools for upcoming LEO networks. As a first step towards meeting this need, we develop Hypatia, an analysis framework capturing the orbital dynamics of LEO networks.

• We use regulatory filings by the largest three planned LEO networks to evaluate and visualize their networks.

• Using packet-level simulations, we analyze the behavior of individual end-end connections across such networks in terms of their changing latencies and path structure, and show how this impacts congestion control negatively, even in the absence of any competing traffic.

• Further, by simulating traffic constellation-wide, we show that the changes in path structure result in a difficult problem for routing and traffic engineering, as the utilization of paths and links is highly dynamic.

• Hypatia’s visualizations aid intuition about the structure of satellite trajectories and their impact on a constellation’s behavior, and pin-point traffic hotspots in the network and show their evolution over time.

• 我们阐述了为即将到来的低地球轨道（LEO）网络构建网络分析工具的必要性。作为满足这一需求的第一步，我们开发了Hypatia，一个分析框架，用于捕捉LEO网络的轨道动态。
• 我们利用三大计划中的LEO网络的监管文件，对其网络进行评估和可视化。
• 通过包级仿真，我们分析了在这些网络中，单个端到端连接的行为，包括其变化的延迟和路径结构，并展示了这些变化如何在没有任何竞争流量的情况下，对拥塞控制产生负面影响。
• 此外，通过对整个星座的流量进行仿真，我们表明路径结构的变化导致了路由和流量工程中的一个复杂问题，因为路径和链路的利用率高度动态。
• Hypatia的可视化工具有助于直观理解卫星轨迹的结构及其对星座行为的影响，并能够精准定位网络中的流量热点及其随时间演变的情况。

Satellite networking played an important role in laying the foundations of the Internet, and may again provide the impetus for substantial and exciting changes. We hope that Hypatia will serve as an enabler for that work. Hypatia’s source code is available online [40], together with our visualizations [7].

卫星网络在奠定互联网基础方面发挥了重要作用，并可能再次为重大而激动人心的变革提供动力。我们希望Hypatia能够成为这一工作的推动者。Hypatia的源代码已在线发布，同时也包含我们的可视化工具。