Uber — Mock Interview

You're designing Uber for a live system-design interview. Drivers stream their GPS location continuously; a rider requests a ride and expects the nearest available car within seconds; the system matches the two, tracks the trip live, and prices it. The defining challenge is a live geospatial index: millions of drivers each ping every few seconds, so the "who is where" index is rewritten constantly — yet riders query it for nearby cars and must be matched almost instantly, with no driver ever dispatched to two riders at once.

Scale assumptions:

• ~5M active drivers at peak, each pinging location every ~4s (>1M location writes/s).
• ~100M riders; ~10M rides/day; hundreds of thousands of nearby-car queries/s at peak.
• Targets: dispatch a match in < 5s, nearby query < 200 ms p99, location freshness within seconds, 99.99% marketplace availability.

I'll take you through requirements, estimation, architecture, a geo-index + matching deep dive, and trade-offs. Think out loud and state your assumptions.