Hi everyone!
I am a computer science student currently planning my Bachelor’s final thesis. My project goal is to build a self-hosted routing and route optimization service inspired by the GraphHopper Directions API, integrating GraphHopper (for route calculation and distance matrices) with jsprit (for solving Vehicle Routing Problems).
Before I lock in my project architecture, I would love to get some feedback from the community regarding hardware sizing and integration complexity:
-
AWS Instance Sizing (EU Map): I plan to process the full OpenStreetMap extract for Europe supporting three profiles:
car,truck, andsmall_truck. Would an AWSr6a.2xlargeinstance (8 vCPUs, 64 GB RAM) be sufficient for building and running the graph, or is 64 GB likely to bottleneck with these three profiles? -
Integration Complexity (Java): For anyone who has integrated the two libraries natively: how much development effort is typically involved in piping GraphHopper’s matrix calculations (distances/durations) into
jsprit’sVehicleRoutingTransportCostsmatrix? Are there known performance traps or recommended patterns when translating GraphHopper matrix outputs into jsprit problem definitions?
Any advice, hardware recommendations, or architectural pointers would be hugely appreciated!
Thanks in advance!
Tibe