Designing a Ride-Sharing System

A ride-sharing platform matches riders with nearby drivers in real-time.

Core Features

• Request a ride
• Match with nearby driver
• Real-time GPS tracking
• Dynamic pricing (surge)
• Payments

Location Service

Drivers send GPS updates every 4 seconds.

Driver App → Location Service → Redis (Geo)

Redis GEO commands:

GEOADD drivers:active lng lat driver_id
GEORADIUS drivers:active -73.9 40.7 5 km

Driver Matching

1. Rider requests ride with pickup coordinates
2. Find all drivers within 5km using GeoSearch
3. Filter: available, correct vehicle type
4. Rank by ETA (maps API)
5. Send request to best driver
6. 10s timeout → try next driver

Real-Time Tracking

• WebSocket connection between rider/driver apps and server
• Driver location pushed to rider every 4s

Driver → Location Service → Kafka → Rider WebSocket

Surge Pricing

Surge = f(demand / supply)
If demand > 2x supply: 1.5x price

Calculate per geographic zone (H3 hexagons) in real-time.

Trip Storage

• Active trips: Redis
• Completed trips: PostgreSQL
• Analytics: Redshift

Maps & ETA

Integrate Google Maps or build routing engine using road graph + Dijkstra's algorithm.

Conclusion

Ride-sharing is a geo-intensive real-time system. Redis GEO, WebSockets, and fast matching algorithms are the core building blocks.