On the Distinction Between Configuration and Topology in Satellite Networks

Satellite telecommunications constellations in low Earth orbit (LEO) have moved from conceptual ideas to large-scale deployment. Systems such as Starlink [1], Eutelsat OneWeb [2], and Amazon Leo (formerly Project Kuiper) [3] exemplify the new paradigm in global telecommunications, where satellites play a central role.

Unlike geostationary (GEO) satellites, LEO systems offer many attractive communication attributes, including shorter propagation delays and reduced link budget requirements for user terminals. However, these advantages come with a significant challenge. While GEO satellites remain relatively fixed in the Earth-rotating reference frame, LEO satellites complete an orbit in roughly 90 to 120 minutes and travel at speeds of about 7.2 to 7.8 km/s, depending on their altitude. For architectures that employ inter-satellite links (ISLs), this rapid motion creates a network whose geometry, coverage, and feasible connectivity evolve continuously. This dynamic nature makes routing traffic through the mesh a complex and constantly changing problem.

This fast-paced, time-varying environment has motivated considerable academic and network protocol research on optimizing performance in ISL networks. Researchers have focused on improving critical metrics such as end-to-end latency (minimizing the time for packets to traverse the network), system throughput (maximizing aggregate traffic under realistic constraints), and jitter and quality-of-service (reducing latency variability and maintaining service requirements for real-time applications such as video calls or gaming).

Defining the Layers

A common assumption in routing and scheduling research is that the physical distribution of satellites in orbit has already been determined and remains fixed throughout the analysis. (This reflects practical realities—launch economics, regulatory constraints, and orbital mechanics do impose significant constraints on geometric design.) However, treating the configuration as a given can make us overlook a more fundamental question about how the physical arrangement of satellites relates to the network properties we can construct on top of it. To examine this relationship, we first need to establish some terminology.

Think of these layers as nested constraints, as illustrated in Figure 1. The constellation configuration determines what links could possibly exist, inducing a time-varying feasibility graph of geometrically possible connections. The network topology then determines which of those feasible links are actually established and maintained at any given time. Finally, network performance (the metrics we ultimately care about, such as latency and throughput) is computed based on this realized topology. ^[*]

Visualizing How Configuration Shapes Topology

To make this relationship concrete, it helps to examine two cases side by side. Let us consider two different Walker-Delta configurations, each following its own specification $T/P/F$ . In this notation, $T$ represents the total number of satellites in the constellation, $P$ is the number of orbital planes, and $F$ controls the relative along-track phasing between planes. By adjusting $P$ and $F$ , we change the three-dimensional spatial distribution of the nodes, which directly affects which links can be simultaneously short, unobstructed by Earth, and feasible given our system constraints.

To generate a network topology from a given configuration, we use a simplified model that captures the essential physics. An edge can exist in $E(t)$ only if two conditions are met: the instantaneous satellite-to-satellite distance falls within a maximum threshold, and the line-of-sight between satellites is not blocked by Earth. In real systems, additional constraints typically come into play, including antenna steering limits, Doppler effects and pointing losses in the link budget, and per-node limits on the number of simultaneous ISLs that can be maintained. However, this simplified model is sufficient to illustrate the key principles.

In the visualizations below, we contrast two Walker-Delta-based configuration-topology pairs. Both configurations use exactly the same number of satellites ( $T=24$ ) and identical ISL feasibility thresholds. The only difference between them is how the satellites are arranged, specifically the choice of orbital planes and phasing parameters.

Figure 2: The configuration uses 12 planes (consequently, two satellites per plane). In this specific case, we observe a partitioning: the network naturally partitions into two disjointed groups positioned on opposite sides of Earth. While satellites within each group maintain full connectivity with their neighbors, no active ISLs bridge the gap between the two groups. This creates two isolated sub-networks that cannot directly exchange traffic.

Figure 3: This configuration presents a dramatically different picture. It uses only two orbital planes, placing 12 satellites in each plane. Here, the network achieves global connectivity. Every satellite maintains continuous links with its neighbors ahead of and behind it in the same plane (the intra-plane links). Additionally, satellites near the nodal crossings (where the two orbital planes intersect) can establish connections with satellites in the other plane (the inter-plane links).

As these illustrations demonstrate, each configuration offers distinct topological properties, and the optimal choice depends on the specific mission objectives and operational constraints at hand. These represent just two hand-picked examples from among the 60 possible Walker-Delta patterns achievable with 24 satellites (including graph isomorphisms). Each of those 60 patterns would exhibit its own unique connectivity characteristics and performance trade-offs.

The Case for Co-Design

Algorithmic innovation in routing and scheduling remains essential and can achieve remarkable efficiency gains within a given constellation design. However, we must recognize that these algorithms operate within boundaries established by the physical configuration. This observation suggests a complementary opportunity that is often overlooked: co-designing the constellation configuration itself to enable more favorable topological properties, which routing and scheduling algorithms can then better exploit.

This approach admittedly assumes that we have geometric design freedom, which is an expensive proposition in practice. Most satellite systems are designed around a complex web of constraints, including launch vehicle capabilities, orbital mechanics, economic realities, and regulatory requirements, which leave little room for geometric experimentation. Given these constraints, it is natural that researchers focus their efforts where they have the most control. But the thought experiment remains instructive. If we could exercise more deliberate control over the spatial distribution of our satellites during the design phase, what network properties become accessible? What new opportunities might this create for routing and scheduling algorithms? These questions are worth pondering and provide food for thought.

The examples above rely on well-known symmetric configurations for ease of illustration, but future designs need not be limited to these patterns. Asymmetric constellation patterns [4] represent another dimension of co-design freedom, allowing designers to tailor the geometry-topology relationship more precisely to meet specific coverage or performance objectives. This expansion enlarges the design space from a finite catalog to a continuous spectrum of possibilities.

For readers interested in exploring these ideas further, I recommend examining recent work on the co-design of configuration and network topology for complex traffic demand patterns using mathematical optimization techniques [5], as well as analyses of how different network topologies affect the capacity for processing data onboard satellites [6].