Introducing Drayroute v1: Sub-millisecond Matrix Calculations for Global Logistics
How we engineered our distributed C++ routing engine with Rust edge bindings to calculate 10,000x10,000 distance matrices in under 15 milliseconds.
Key Article Takeaways
- Custom Contraction Hierarchies (CH) graph solver yielding 100x lower latency than standard Dijkstra variants.
- Zero-copy SIMD parallelization across 64 global edge edge locations.
- Turn-key drop-in HTTP REST and gRPC interfaces for enterprise dispatch systems.
Modern logistics platforms rely heavily on instantaneous matrix calculations. Whether orchestrating drayage truck dispatches from congested ports or optimizing multi-stop urban delivery fleets, calculating turn-by-turn travel times across thousands of origin-destination pairs is the primary bottleneck.
Traditional routing engines often stall when scaling past 1,000x1,000 matrices, taking multiple seconds per payload and consuming gigabytes of memory. With Drayroute v1, we completely re-architected the core solver using low-level C++ with memory-safe Rust bindings deployed directly to regional edge nodes.
By pre-computing shortcut edges via custom Contraction Hierarchies (CH) and leveraging AVX-512 SIMD vector instructions, Drayroute achieves a 10,000x10,000 matrix solution in under 15 milliseconds. This allows dispatchers to simulate thousands of routing scenarios per second in real time.
In addition to raw speed, Drayroute v1 introduces dynamic toll calculation, truck-specific bridge weight enforcement, and live port gate congestion overlays directly integrated into the routing graph.
Ready to test Drayroute in your application?
Get 10,000 free monthly route API credits with zero credit card required.