Introduction

With the rapid development of aerospace technology, recently we have witnessed the vigorous deployment of satellite Internet mega-constellations such as SpaceX’s Starlink [1], Amazon’s Kuiper [2] and Telesat [3]. These constellations will deploy thousands of broadband satellites in low earth orbit (LEO), and each satellite can be equipped with high-speed inter-satellite links (ISLs) [4] as well as ground-satellite links (GSLs) [5] to interconnect with other satellites and ground facilities, extending the boundary of today’s Internet and constructing an integrated space and terrestrial network (ISTN).

随着航空航天技术的迅速发展，近年来我们见证了卫星互联网巨型星座的蓬勃发展，例如SpaceX的Starlink、亚马逊的Kuiper和Telesat。这些星座将在低地球轨道（LEO）部署数千颗宽带卫星，每颗卫星可以配备高速星间链路（ISLs）和地面-卫星链路（GSLs），以与其他卫星和地面设施相互连接，扩展当前互联网的边界，并构建一个集成的空间和陆地网络（ISTN）。

Upon the IP-based infrastructure [6], future ISTNs will support heterogeneous access technologies (e.g., Ka-/Ku-/Eband [5] or personal communications services for phone-directto-satellite access [7]) for global Internet services. ISTN users include not only residential customers who leverage fixed satellite terminals deployed on rooftops [1] to access satellites, but also global mobile users in airplanes [8] and maritime cruises [9] which can roam globally over time. In practice, Starlink has recently released its global roaming services to enable Internet services all around the world [10]. How to efficiently manage the mobility of mobile users (e.g., tracking the location of users and coping with handovers) and keep users’ connections active is an important issue for ISTN operators.

在基于IP的基础设施上，未来的ISTN将支持异构的接入技术（例如Ka-/Ku-/E频段或用于手机直接与卫星通信的个人通信服务），以提供全球互联网服务。ISTN的用户不仅包括利用屋顶部署的固定卫星终端访问卫星的住宅客户，还包括在飞机和海上游轮上全球漫游的移动用户。实际上，Starlink最近推出了全球漫游服务，以在全球范围内提供互联网服务。如何高效管理移动用户的移动性（例如跟踪用户位置并处理切换）并保持用户的连接活跃，对于ISTN运营商来说是一个重要的问题。

Network-layer mobility management (MM) has always been an important technique for managing user mobility, especially for heterogeneous mobile networks because it does not rely on or make any assumption about the underlying wireless access technologies. The network community has a long history studying on network-layer MM solutions such as IETF MIP [11], [12] and its variations [13]–[15]. The core idea behind existing solutions is to deploy anchor points at certain fixed nodes in the network. Each mobile node first registers with a certain anchor point and then informs the anchor of its location updates during the roaming process. In addition, each mobile node is identified by its home address allocated by its corresponding anchor, which further allows users to move from one network to another while maintaining a permanent address to keep connections active.

网络层的移动性管理（MM）一直是管理用户移动性的重要技术，尤其适用于异构移动网络，因为它不依赖于或对底层无线接入技术做任何假设。网络社区长期以来一直在研究网络层MM解决方案，如IETF MIP、及其变体。现有解决方案背后的核心思想是 将锚点部署在网络中的某些固定节点上。每个移动节点首先在某个锚点上注册，然后在漫游过程中通知锚点其位置更新 。此外，每个移动节点由其对应锚点分配的主地址标识，这进一步允许用户在移动到另一个网络时保持永久地址以保持连接活跃。

However, as our quantitative analysis shows in §III, directly applying existing MM solutions in emerging ISTNs may suffer from low connection uninterrupted ratio or high user-perceived latency. The root cause relies on a unique characteristic differentiating ISTNs from other traditional mobile networks: not only the end users are mobile, but also the global network infrastructures (i.e., LEO satellite routers) which participate in route calculation are frequently changing their locations in the network. In an ISTN, if we directly follow existing solutions and deploy anchors on the ground (e.g., at ground stations), the combination of space-ground handovers and routing fluctuations can exacerbate service interruptions when users are roaming around the world. Moreover, as both users and satellites are moving, the user-to-anchor path may be lengthened, causing a significant increase in user-perceived latency over time.

然而，如我们的定量分析在§III中所示，在新兴ISTN中直接应用现有的MM解决方案可能会导致连接不中断的比例较低或用户感知延迟较高。其根本原因在于ISTN与其他传统移动网络的不同之处：不仅终端用户是移动的，全球网络基础设施（即LEO卫星路由器）也在频繁改变其在网络中的位置。在ISTN中，如果我们直接遵循现有解决方案并在地面部署锚点（例如在地面站），则空间-地面切换和路由波动的组合可能会加剧服务中断，当用户在全球范围内漫游时。此外，由于用户和卫星都在移动，用户到锚点的路径可能会延长，导致用户感知延迟随时间显著增加。

To facilitate seamless and low-latency satellite Internet service, in this paper we present SKYCASTLE, a novel networklayer global mobility management mechanism for futuristic IP-based ISTNs. At a high level, SKYCASTLE divides all satellites into multiple clusters, and exploits a collection of distributed satellite anchors to manage the mobility of the users and the core network infrastructures. In particular, SKYCASTLE incorporates two techniques for global mobility management.

为了促进无缝和低延迟的卫星互联网服务，本文提出了一种名为SKYCASTLE的新型网络层全局移动性管理机制，适用于未来基于IP的ISTN。从高层次上讲，SKYCASTLE将所有卫星划分为多个集群，并利用一组分布式卫星锚点来管理用户和核心网络基础设施的移动性。特别地，SKYCASTLE结合了两种技术用于全局移动性管理。

First, SKYCASTLE adopts a dynamic-anchor-based MM scheme, together with a convergence-free route mechanism to efficiently track the location of mobile nodes, while providing available forwarding paths when the ISTN topology changes. For users covered by the same cluster and managed by the same satellite anchor, they register with the anchor when they connect to the current cluster, and inform the anchor of their location updates if their locations change. Moreover, the location updates of a ground station are additionally sent to all other anchors in the ISTN to avoid route convergence.

首先，SKYCASTLE采用基于动态锚点的MM方案，结合收敛自由的路由机制，高效地跟踪移动节点的位置，同时在ISTN拓扑变化时提供可用的转发路径。对于由同一集群覆盖并由同一卫星锚点管理的用户，他们在连接到当前集群时在锚点上注册，并在其位置发生变化时通知锚点其位置更新。此外，地面站的位置更新也会发送给ISTN中的所有其他锚点，以避免路由收敛。

Second, SKYCASTLE employs an anchor manager integrating a series of algorithms to judiciously decide how a satellite operator can distribute anchor functionalities from a constellation perspective to reduce the number of interruptions, bound latencies to satisfy various user requirements, and reduce the deployment cost of satellite anchors.

其次，SKYCASTLE采用一个锚点管理器，集成了一系列算法，以从星座的角度出发，合理地决定如何分配卫星操作员的锚点功能，以减少中断次数，满足各种用户延迟要求，并降低卫星锚点的部署成本。

To evaluate the effectiveness of SKYCASTLE, we build an experimental ISTN environment based on a recent simulator [16] for satellite networks, and implement a prototype containing all core functionalities of SKYCASTLE. Through trace-driven evaluations combining real constellation information, mobile user trajectories (e.g., popular flight routes) and network simulation, we demonstrate that for representative global roaming scenarios in ISTNs, SKYCASTLE can: (i) improve the connection uninterrupted time by up to 55.8% and by 34.5% on average; (ii) decrease the user-perceived latency by up to 47.8% and by 21.5% on average, as compared to existing solutions.

为了评估SKYCASTLE的有效性，我们基于最近的卫星网络模拟器建立了一个实验ISTN环境，并实现了包含所有SKYCASTLE核心功能的原型。

通过结合实际星座信息、移动用户轨迹（例如流行的航线）和网络模拟的跟踪驱动评估，我们证明了SKYCASTLE在ISTN的代表性全球漫游场景中可以：（i）提高连接不中断时间最多55.8%，平均提高34.5%；（ii）降低用户感知延迟最多47.8%，平均降低21.5%，与现有解决方案相比。

Contributions of this paper can be concluded as follows.

• We expose and analyze the challenges for global mobility management in emerging ISTNs, where both end users and network infrastructures are inherently mobile (§III).

• We propose SKYCASTLE, a novel global mobility management mechanism exploiting dynamic and distributed satellite anchors to facilitate seamless and low-latency satellite Internet services globally (§IV,§V,§VI).

• We build a SKYCASTLE prototype and conduct trace-driven simulations to show its effectiveness on improving network availability for global roaming in ISTNs (§VII).

本文的贡献可以总结如下：

• 我们揭示并分析了新兴ISTN中全局移动性管理的挑战，在那里终端用户和网络基础设施都具有内在的移动性（§III）
• 我们提出了SKYCASTLE，一种利用动态和分布式卫星锚点的新型全局移动性管理机制，以促进全球无缝和低延迟的卫星互联网服务（§IV、§V、§VI）
• 我们构建了SKYCASTLE原型，并进行了跟踪驱动的模拟，以展示其在ISTN中改善网络可用性和全球漫游方面的有效性（§VII）