Design Uber/Lyft Location Index (Pt 2, Deep Dives) | Systems Design Questions 3.0 With Ex-Google SWE

A deep dive into designing Uber's location index system, focusing on geospatial sharding strategies. Covers how to handle 1 million driver location pings per second using Uber's H3 hexagonal grid library for geographic bucketing and partition key selection. Explains dynamic shard splitting for hotspot areas (e.g., dense cities vs. rural regions), caching the shard table locally on servers, handling boundary edge cases for proximity queries, and using PostgreSQL with PostGIS and Citus for distributed geospatial indexing. Also briefly compares H3 to geohash-based range partitioning.