Katkat
Back

Design a game leaderboard service

MetaSoftware Engineer · Onsite
MediumSoftware EngineerOnsiteSystem Design

Last updated

This question tests the ability to design a high-throughput, low-latency leaderboard system, focusing on data structures for ranking, storage trade-offs, and scalability under heavy write and read loads.

Problem Statement

Design a global leaderboard service for a massive multiplayer mobile game. The game has 50 million daily active users (DAU) and millions of updates to player scores per second.

Functional Requirements

  1. updateScore(userId, score): Update a player's score.
  2. getTopPlayers(k): Retrieve the top k players globally.
  3. getRank(userId): Retrieve the rank of a specific player.
  4. getNearbyPlayers(userId, k): Retrieve the k players ranked just above and below the given user.

Non-Functional Requirements

  • Scalability: Handle high write throughput (millions of updates/sec).
  • Latency: getTopPlayers should be extremely fast, ideally under 100ms.
  • Consistency: Users expect near-real-time updates, though minor eventual consistency is acceptable.

Deliverables

  • Describe the high-level system architecture.
  • Choose a data store and explain why it suits this use case.
  • Detail the API design.
  • Explain how to handle partitioning/sharding to support the scale.

Hints

Hint 1

Consider using a data structure that keeps elements sorted by score while maintaining logarithmic time complexity for insertions and lookups.

Hint 2

A standard relational database might struggle with frequent updates and range queries on a massive scale; explore in-memory stores that support specialized sorted sets.

Hint 3

Think about how you would shard the data; would you shard by userId or by score range, and what are the trade-offs for the requested operations?

Follow-up questions

  • How would you handle a 'hot key' problem if the top 100 players are being queried millions of times per second?
  • How would you adapt the system to handle 'seasonal' leaderboards (e.g., scores reset every week)?
  • If the leaderboard needs to be accurate to the millisecond across multiple geographic regions, what consistency challenges arise and how would you mitigate them?