You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Planet router: design how the routing package scales to the whole planet #2657
Production coverage reaches the whole planet within about a year (#2648). planner/routing is a "whole" layer: one routing package over the full coverage, because a route can cross any border. planner/router-build keeps all source OSM objects in memory (planner/router-build/README.md). A planet package therefore needs a very large machine, and every OSM refresh rebuilds all of it.
Decide
One planet package on one large machine, or a partitioned router (for example one package per continent, with overlap at the borders).
If partitioned: the partition rule, how a route that crosses partitions works, and what the iOS offline router and its grid selections (specs/route-package.md) download.
Memory and time per partition, measured on one large region, and the bake machine that this needs.
What the search service and planner/basemap need at planet scale.
The data system offers one hook: a whole layer may declare named partitions, and each partition is staled, baked and published on its own (#2655). This issue decides how the router uses that hook. It does not change the data system.
Activity
timohueser commentedon Nov 4, 2026
This issue is merged into #3000. Workstream 4 (hierarchical search, disk storage, planet RAM and bake memory) answers this question with measurements.
Generated by Claude Code