diffusive-routing

Entropy-minimizing network routing — packets are routed along paths of minimum entropic resistance, with the resistance field evolving under a Turing-style reaction-diffusion process driven by network load.

A self-contained synthetic routing testbed: a NetworkX topology (datacenter fat-tree or random), an entropic resistance field ρ_ij = baseline / (1 + Γ_ij · C_ij), a modified-Dijkstra router that follows that field, and a reaction-diffusion update dρ/dt = D∇²ρ - kΓρ + f(load) that lets the field adapt over time. All values reported by the system are computed live from the simulation — there is no fabricated benchmark data.

Installation

pip install diffusive-routing
# with visualisation support (matplotlib)
pip install "diffusive-routing[viz]"

Usage

# Run a simulation cycle on the default datacenter fat-tree topology
diffusive-routing run --topology datacenter-fat-tree --packets 1000

# Compare diffusive routing against a static shortest-path (OSPF-like) baseline
diffusive-routing compare-ospf --topology random --packets 200

# Visualise the entropic resistance field
diffusive-routing visualise --save-path field.png

Or use it as a library:

from diffusive_routing import DiffusiveRouting

system = DiffusiveRouting()
result = system.run_cycle(n_packets=1000)
print(result)
print(system.get_crep_state())   # {"C": ..., "R": ..., "E": ..., "P": ..., "Gamma": ...}
print(system.get_utac_state())   # {"H": ..., "dH_dt": ..., "H_star": ..., ...}

How it works

• Entropic resistance field (entropy_field.py) — per-link resistance derived from the link's CREP state; lower resistance means a more attractive path.
• Reaction-diffusion (reaction_diffusion.py) — the resistance field diffuses across neighbouring links and decays/reinforces based on utilization, so congestion on one link gradually reroutes traffic onto neighbours.
• Router (router.py) — routes each Packet via shortest path under the current resistance weights (a diffusive analogue of Dijkstra).
• S_A / S_V duality (sa_sv_duality.py) — tracks the action entropy of routed paths (S_A, routing cost) against the volume entropy of traffic incident on a destination (S_V), exposing a Lagrangian-style score L_net = -S_A + S_V.
• CREP tensor (crep_network.py) — per-link and network-wide Coherence / Resonance / Emergence / Poetics-diversity state, aggregated into a single Γ via a geometric mean.
• Benchmark (benchmark.py) — compare-ospf runs the same (src, dst) pairs through both diffusive routing and a static shortest-path baseline and reports throughput/hop/resistance deltas.

Role in the GenesisAeon Ecosystem

diffusive-routing is package P30 in the GenesisAeon ecosystem, covering the network science domain. It implements the GenesisAeon Diamond Interface (run_cycle, get_crep_state, get_utac_state, get_phase_events, to_zenodo_record) on top of the UTAC/CREP tensor framework shared across the ecosystem, targeting a routing Γ setpoint of approximately 0.443 (efficiency η = tanh(σ·Γ), σ = 2.2).

Citation

DOI will be assigned automatically on first GitHub Release once Zenodo–GitHub integration is enabled for this repo.

Built with Typer · Rich · NetworkX