Mesh Networking in Low Earth Orbit

Traditional networking assumes stable connections and fixed topology. In Low Earth Orbit, nothing is stable—nodes move at 7.66 km/s, links form and break every few minutes, and latency changes continuously. Here's how we built a mesh network that thrives in this chaos.

Traditional vs. Mesh Networking

Terrestrial networks use hub-and-spoke architectures: traffic flows through central routers, switches, and data centers. This works because infrastructure is stationary and links are persistent.

In orbit, this model fails immediately. A satellite in LEO completes one orbit every ~92 minutes, passing over different ground stations and adjacent satellites at different ranges and angles. The network topology is never the same twice.

Traditional Network:
  Client → Router → Core Switch → Server
  Topology: Fixed
  Latency: Predictable
  Links: Persistent

Orbital Mesh Network:
  Node → [dynamic ISL] → Node → [dynamic ISL] → Node
  Topology: Changes every ~5 minutes
  Latency: Variable (40-500ms)
  Links: Formed and broken continuously

Inter-Satellite Links (ISL)

The backbone of our mesh is optical inter-satellite links. Each SolarNode carries four laser communication terminals positioned at cardinal points, enabling simultaneous links to up to four neighboring nodes.

Link Specifications

• Data rate: 10 Gbps per link (bidirectional)
• Wavelength: 1550 nm (eye-safe, telecom-grade)
• Maximum range: 5,000 km (limited by orbital geometry)
• Acquisition time: <2 seconds (coarse pointing + fine tracking)
• Beam divergence: 15 microradians

Laser ISLs have a critical advantage over radio: they don't require spectrum allocation. Radio frequencies in space are congested and regulated, but photons in free space are unlimited. We can add links without regulatory approval.

Link Budget

Transmit power:        +10 dBm (10 mW)
Telescope gain:        +105 dB
Free-space loss (2000km): -290 dB
Receive telescope:     +105 dB
Detector sensitivity:  -45 dBm
──────────────────────────────
Link margin:           +15 dB (comfortable)

Dynamic Routing in Orbit

Our routing protocol, OLSR-Orbital, extends the Optimized Link State Routing protocol with orbital mechanics awareness:

Key Innovations

1. Predictive topology — Since orbital mechanics are deterministic, we can compute the network topology for the next several orbits in advance. Routes are pre-calculated and cached.
2. Contact graph routing — Borrowed from DTN (Delay-Tolerant Networking) research, we model each inter-satellite contact as a time-bounded edge in a contact graph. The routing algorithm finds the time-optimal path through the graph.
3. Multi-path forwarding — Critical traffic is sent along multiple paths simultaneously. The receiver deduplicates. This provides resilience against unexpected link drops.

// Simplified route computation
function computeRoute(src, dst, time) {
  const contacts = predictContacts(time, time + ORBIT_PERIOD);
  const graph = buildContactGraph(contacts);

  // Dijkstra with time-varying edge weights
  return timeAwareDijkstra(graph, src, dst, time);
}

// Routes are recomputed every 30 seconds
// and pre-cached for the next 3 orbital periods

Latency Optimization

End-to-end latency in our mesh depends on the number of hops and the distance of each hop. Light travels at ~300,000 km/s, so a 2,000 km ISL adds ~6.7 ms one-way. A typical path might traverse 2–4 hops:

Path: Ground → Node A → Node B → Node C → Ground

Segment              Distance    Latency
Ground → Node A      408 km      1.4 ms
Node A → Node B      1,800 km    6.0 ms
Node B → Node C      2,200 km    7.3 ms
Node C → Ground      408 km      1.4 ms
Processing (×3 hops) -           1.5 ms
──────────────────────────────────────────
Total one-way:                   17.6 ms
Round-trip:                      35.2 ms

For comparison, a transatlantic fiber optic cable between London and New York adds ~35 ms one-way due to the refractive index of glass slowing light to ~200,000 km/s. Our orbital mesh can actually beat fiber for intercontinental traffic.

Fault Tolerance

The mesh must continue operating when nodes fail or links drop unexpectedly. Our fault tolerance strategy has three tiers:

Tier 1: Link Redundancy

Each node maintains connections to 2–4 neighbors. Losing one link triggers automatic rerouting in <100 ms. No traffic is lost because multi-path forwarding means the alternate path was already carrying copies.

Tier 2: Node Failure

If a node goes silent, neighbors detect the failure within 3 heartbeat intervals (9 seconds). The contact graph is updated and new routes propagated. Workloads hosted on the failed node are re-scheduled on surviving nodes.

Tier 3: Partition Recovery

If the constellation splits into disconnected groups (e.g., multiple node failures), each partition operates autonomously. When the partition heals (nodes come back online or orbit brings partitions within ISL range), state is reconciled using CRDTs (Conflict-free Replicated Data Types).

Network Topology Visualization

Our 6-node initial constellation forms a Walker Delta pattern at 408 km altitude, 53° inclination. At any moment, the topology looks roughly like this:

         SN-03
        /     \
  SN-01 ───── SN-05
   |  \     /  |
   |   SN-02   |
   |  /     \  |
  SN-04 ───── SN-06

ISL links shown (active at T=0)
Links reconfigure every ~5 minutes
Average path length: 2.1 hops
Maximum path length: 3 hops

As the constellation scales to 24 and then 48 nodes, the mesh becomes denser and more resilient. With 48 nodes, every location on Earth is within 1 hop of at least 2 nodes, and the average path length drops below 1.5 hops.

Building a network where the topology never repeats requires embracing change as the default state. Orbital mesh networking isn't about fighting physics—it's about making physics work for you.

References:

• RFC 3626: Optimized Link State Routing Protocol (OLSR)
• RFC 9171: Bundle Protocol Version 7 (DTN)
• Bhattacherjee & Singla, “Network Topology Design at 27,000 km/hour” (CoNEXT 2019)
• SolarNode Constellation Design Document v2.1