Orthodromic Routing and Forwarding for Large Satellite Constellations

TL;DR

This work introduces Orthodromic Routing ($OR(r)$) for large satellite constellations with inter-satellite links, addressing scalability by restricting flooding to a radius $r$ and by aggregating addresses on the unit sphere. It blends geographic proximity with localized link-state information, enabling proactive routing with pre-computed SPF trees and parallelizable hardware-forwarding. The key contributions include a concrete addressing/forwarding design, a bounded flooding mechanism, and simulation-backed guidance on how $r$ should scale with link failures to maintain low packet loss. The approach promises line-rate forwarding and robust routing in constellations of thousands of satellites, with practical implications for intercontinental coverage and private networks over the satellite mesh.

Abstract

Low earth orbit satellite constellations with intersatellite links (ISLs) are currently being developed and deployed. The availability of ISLs provides the capability to route across the satellite constellation, rather than using the satellite as a single hop in a bent-pipe configuration. We present a fully distributed solution to routing and forwarding which we call Orthodromic Routing (OR(r) ). OR(r) routing is built on a foundation of both geographic and link state routing to create a hybrid protocol which scales to enormous constellations with excellent failure handling. Our work includes an addressing and forwarding plane for OR(r)which can be implemented in hardware in a highly parallel manner to achieve line rates while only requiring a bounded number of forwarding table entries.

Figures (7)

• Figure 1: An orthodrome is a portion of a great circle at the surface of the earth, however we use the term independently of the radius of the sphere on which we are routing to conveniently express the shortest possible path on that sphere’s surface between two points.
• Figure 2: The OR(r) algorithm simplified. The SPF computations are not done per packet but are pre-computed.
• Figure 3: OR(r) forwarding table comparator tree which can be implemented in hardware or parallel CPU processing in O(log(K)) clock cycles.
• Figure 4: OR(1) single link failure
• Figure 5: OR(2) seeing past a link failure