Orbital Mesh Routing Engine
Orbital Mesh Routing Engine
OMRE is a complete, production-engineered routing kernel for orbital mesh networks — the layer that decides, on every tick of a constellation's master clock, which satellite forwards each packet to which next hop. It is engineered to the standards a flight-software qualification programme demands: zero heap allocation after construction, bounded worst-case execution time, bit-identical replay, and a full audit-evidence bundle that a buyer's acceptance team can reproduce on their own hardware in under an hour. This listing is for the outright transfer of the entire IP — every line of source, every test, every document, every audit artefact — to a buyer.
What the asset does
Given a registered set of orbital nodes (satellites) and terrestrial nodes (ground stations), OMRE computes the optimal mesh-network traversal paths between every pair, on every tick, with three hard guarantees:
-
Zero heap allocation. No
malloc,new,mmap, orsbrkis ever called after construction. Verified by an interposedoperator newthat aborts on call — the demo binary runs to completion under this trap. -
Bounded worst-case execution time. Every public entry point completes in a number of cycles that depends only on the compile-time bounds in
omre/config.hpp, never on the runtime input. Empirical p99 and max are published; a CI regression gate refuses to ship a build whose worst case has drifted. -
Bit-identical replay. Two runs of the same binary on the same input produce byte-identical output. An opt-in fixed-point arithmetic path extends this guarantee across compilers and host architectures.
A forwarding decision — the operation that runs once per packet, billions of times per orbit — is a single indexed memory load into a precomputed next-hop matrix. Median latency on x86-64: ~10 nanoseconds.
Headline numbers
Measured on a single core of a modern x86-64 host at N = 256 nodes:
| Operation | Median | p99 | Max |
|---|---|---|---|
advance_to_tick — Flat Floyd-Warshall recompute |
37.7 M cycles (~12.6 ms) | 126.7 M cyc | 138.6 M cyc |
advance_to_tick — Hierarchical (cluster-overlay) |
656 K cycles (~219 µs) | 1.2 M cyc | 1.3 M cyc |
query_next_hop — O(1) forwarding decision |
30 cycles (~10 ns) | 34 cyc | (cache-cold outliers) |
query_path — full path reconstruction |
32–48 cycles | 378 cyc | (cache-cold outliers) |
Hierarchical routing delivers a measured 57–65× speedup over the flat path on the bundled constellation, in exchange for a documented and bounded sub-optimality envelope. Engine static footprint (per instance, N = 256): ~660 KiB. Fits in the SRAM of any modern radiation-hardened flight computer.