#09Real-Time Ride Matching & ETA Prediction
9 / 20
Uber logoUber System
HardDistributed Systems
45–60 min

Real-Time Ride Matching & ETA Prediction

Architect Uber's dispatch and routing system combining geospatial H3 hex indexing, live GPS driver telemetry, graph routing engines, and surge pricing calculators.

Scale & Traffic20M trips/day · 5M active drivers sending GPS every 4s
Monthly Cloud Budget$25,000/month
Latency Target (P99)< 50ms Realtime SLA
Availability SLA99.99% Multi-AZ Active

System Requirements

  • Process 1.25M GPS location pings per second with < 500ms lag
  • Geospatial lookup: find nearest 10 available drivers within 3km in < 15ms
  • ETA graph pathfinding considering historical and live traffic congestion
  • Atomic lock on driver matching to prevent race conditions across riders

Graded Constraints

01Budget limit: $25,000/month
02Persistent WebSocket / TCP connections for driver location streams
03High-throughput message streaming buffer (Kafka)
04In-memory spatial index (Redis Geospatial / memory store)
05Transactional DB for ride state and payment settlement

Ready to design?

Open this challenge in the interactive system design playground and satisfy all architecture constraints.

Explore More

Related Architecture Challenges

All Problems
Real-Time Ride Matching & ETA Prediction — System Design Problem | Mimir Nest