Design Uber

August 28, 2024 6:15 PM PDT

---

The meeting focused on the system design of a ride-sharing application similar to Uber. Key topics included the architecture for handling driver and user locations, non-functional requirements, API design, and data management strategies. The discussion also covered load calculations, request handling, and the use of key-value stores for efficient data retrieval.

Key Discussion Points

Non-Functional Requirements

• Latency: Target latency of 1 second.
• Throughput: Discussion on the number of riders and drivers that the system should handle.

Solutions for Nearby Drivers

1. Latitude and Longitude (la, lo).
2. Geohash.
3. QuadTree.

API Design

• User API:
  • Send location API to request a ride.
• Driver API:
  • Approve ride requests and view information about users.

Data Structures

• Car Object:
  • Attributes: id, latitude, longitude, TTL (Time To Live), state, information.
  • Size estimates:
    • 4 million drivers -> 1 million daily active users (DAU).
    • Estimated data size calculations:
      • 1M * 20B = 20MB
      • 1M * 200B = 200MB

Load Calculations

• Request Handling:
  • Request a car: 10 million requests at 200 requests per second.
  • Location updates from drivers: 1 million drivers updating every 5 seconds.
  • Total estimated requests: 1,000 requests per second with peak loads of 1,500 requests per second.

Data Management

• Storage Options:
  • Trade-offs between key-value stores and traditional databases.
  • Key-Value Store: Redis with snapshot and write-ahead logging.

Geospatial Data Handling

• Geo Operations:
  • Use of Geoadd and Geodel for managing car locations.
  • Key: Geohash; Value: Set of Cars.
  • Operations to move cars between geohashes.

Request Handling Mechanisms

• Communication Methods:
  • Long polling for search services.
  • WebSocket for real-time updates.
  • Short polling and Server-Sent Events (SSE).

Race Condition Management

• Discussion on handling race conditions when multiple riders request the same driver.

Conclusion

The meeting provided a comprehensive overview of the design considerations for a ride-sharing application. Key areas of focus included API design, data management strategies, and ensuring efficient handling of requests and updates. Further exploration of the discussed topics is recommended to refine the design and address potential challenges.