A detailed Chinese engineering blog post dissects the architecture behind a 'Nearby People' feature designed for apps with tens of millions of daily active users. The core of the solution relies on Redis Geo commands (GEOADD, GEORADIUS) combined with GeoHash encoding to achieve millisecond-level spatial queries. To handle extreme scale, the author proposes a two-level sharding strategy: first by city, then by a grid system within each city, ensuring that no single Redis instance becomes a bottleneck. The article also addresses practical challenges such as dynamic radius adjustment based on user density, anti-crawler measures, and data consistency across shards. This is not a beginner tutorial but a production-grade reference for senior engineers and architects dealing with LBS at scale. The trade-offs between precision, latency, and storage are clearly explained, making it a valuable resource for anyone building or optimizing location-based features.
This article provides a comprehensive architecture for a 'Nearby People' feature targeting millions of daily active users, using Redis Geo (GeoHash + ZSET) for spatial indexing and city-grid sharding for scalability.