Measured under
in-process matching path, single instrument, canary hardware
Synthetic sustained load — headroom beyond any real-world exchange order flow.
- Order-ack latencyp99 9 µs · p50 5 µs
- p50 and p99 shown together — in-process matching path, single instrument, canary hardware.
- Sustained throughput1.1M ops/s
- Single instrument, single thread, 30-minute run — a benchmark, not production traffic.
- Deterministic recovery2.9 s
- WAL replay from genesis, 4 GB log — bit-identical state, every time.
- Ledger conservation0 drift
- Gate-enforced — a property, not a benchmark.
On the throughput number
Synthetic on purpose — that is the point.
The throughput figure is marked honestly: it is a benchmark, not production traffic. No live order flow on any exchange, the largest CEX included, reaches the density sustained on a single instrument. The synthetic nature is not a weakness — it means the measured ceiling sits well above any flow you would actually see. And competitors are not named: their public numbers are marketing claims, not verifiable measurements — the differentiator is the published methodology, not someone else's figures.
Mechanism: synthetic sustained load — headroom beyond any real-world exchange order flow.