Abstract
Satellite networks have been widely investigated both in the business and academia for many years, with many important routing algorithms reported in the literatures. However, fewer existing routing algorithms focus on the trade-off between the routing survivability and the routing computation and storage overheads. Due to topological dynamics, it is difficult to effectively apply the conventional routing protocols such as RIP or OSPF to Low Earth Orbit (LEO) satellite networks. According to the virtual topology model based on virtual node, this paper propose a new fully distributed routing protocol for LEO satellite networks, called Distributed Traffic Balancing Routing (DTBR). The proposed protocol not only guarantees the routing survivability and provides the ability of traffic balancing, but also result in few additional computation and storage overheads only deriving from the information flooding of failed satellites. Simulation results demonstrate positive conclusions of our methods.
Introduction
Routing in Low Earth Orbit (LEO) satellite networks is notoriously difficult because the network topology is constantly changing as satellites move at high speeds. Traditional ground-based protocols like OSPF or RIP fail because they cannot keep up with these dynamics. Existing satellite-specific solutions usually fall into two camps: snapshot-based routing, which requires massive storage for discrete time intervals, and distributed routing, which often fails when a satellite further down the path malfunctions.
This paper introduces Distributed Traffic Balancing Routing (DTBR), a protocol designed to bridge this gap. By combining a virtual topology model with a geographically-aware cost factor, DTBR achieves high survivability and efficient traffic balancing with significantly lower computational and storage overhead than traditional snapshot-based methods.
Methodology
The DTBR protocol relies on three main pillars to manage the deterministic yet dynamic nature of LEO constellations:
- Virtual Topology (VT) Mapping: Instead of routing based on physical satellites, the system routes based on "logical locations" (grid positions over the Earth). A mathematical mapping function allows any satellite to determine its current logical identity and that of its neighbors at any given time $t$.
- Geographically-Aware ISL Cost: To balance traffic, the authors introduce a cost factor ($\lambda$) for Inter-Satellite Links (ISLs). This factor is higher for links passing over "hot spots"—densely populated regions like the Northern Hemisphere—and lower for "cold" regions. This proactively detours traffic away from potentially congested areas.
- Distributed Snapshot Transitions: When a satellite fails, its neighbors flood this information through the network. Every "safe" satellite then independently updates its local topology graph and re-runs Dijkstra’s algorithm to find the best path.
- Source Waiting: To handle failures where a path to the destination simply doesn't exist, a "source waiting" mechanism is implemented. The source terminal queries the satellite to see if the destination is reachable; if not, it waits until orbital movement creates a viable path.
Simulation and Results
The researchers evaluated DTBR using the ns2 simulator on a Walker star constellation with 288 satellites. Traffic was modeled using a Pareto distribution to simulate bursty internet communication, with intensity levels assigned based on global user density forecasts.
Key performance metrics compared DTBR against Datagram Routing Algorithm (DRA) and Satellite Snapshot Routing (SSR):
- Survivability: With the "source waiting" mechanism, DTBR achieved a 0% packet loss rate even when 30% of the satellites in the network had failed. In contrast, the standard DRA protocol saw packet loss soar to nearly 90%.
- Throughput: DTBR showed significantly higher throughput than SSR as terminal bitrates increased, thanks to the $\lambda$ cost factor successfully distributing traffic across the network.
- Delay: While DTBR (without source waiting) had similar end-to-end delays to SSR (~110-130ms), the "source waiting" version naturally introduced higher delays as the fault rate increased, representing the time spent waiting for a viable orbital window.
Conclusion
DTBR exploits the predictability of satellite orbits to provide a robust, low-overhead routing solution. By using a proactive cost factor based on Earth's population density, it anticipates and avoids congestion. While the "source waiting" method ensures 100% delivery, the resulting latency makes it more suitable for non-real-time data. Future work will explore dynamic hot spot adjustments and integration with multi-layered MEO/GEO networks.
Summary updated