RETURN TO SYSTEM DESIGN VAULT (55)BLUEPRINT #40 // Hard
#40HardLocation, Geospatial & Ride-Hailing
Find a Rider for Uber or Uber Eats
ASKED IN INTERVIEWS AT:MetaUberGoogleMicrosoft
EXECUTIVE SUMMARY & PROBLEM STATEMENT
Design the core ride-dispatch matching algorithm that pairs incoming ride/order requests with optimal nearby drivers based on ETA, ratings, and route optimization.
SCALE ESTIMATION & TRAFFIC ENVELOPE
DAILY ACTIVE USERS / TPS15 Million daily completed trips globally
READ / WRITE TRAFFIC RATIOReal-time dispatch optimization
5-YEAR STORAGE CAPACITYIn-memory dispatch graph state
FUNCTIONAL REQUIREMENTS
- [1]Match rider to best available driver in geographical proximity
- [2]Support batch dispatching to optimize overall city wait times
- [3]Handle driver accept/reject timeouts
NON-FUNCTIONAL REQUIREMENTS
- •Dispatch response in < 5 seconds
- •Maximized system efficiency (minimized total city ETA)
- •Fair driver distribution
CORE ARCHITECTURAL COMPONENTS
Dispatch Orchestrator
Batch Aggregator (10s window)
H3 Spatial Candidate Filter
Bipartite Graph Solver
Push Notification Gateway
DATA SCHEMA & PERSISTENCE
RideRequest: id, rider_id, pickup_h3, dest_h3, matched_driver_id, status
DISTRIBUTED TRADE-OFFS & FAILURE MODES
- ⟡Batch matching requests over a 10-second interval yields 15% shorter average customer wait times than instant greedy first-come matching.