Game Activity: Design a Backend System to Track and Query Player Activity in Real Time

Question

## Round 1 - System Design

Design a backend system that tracks player activity for a multiplayer game and answers real-time queries.

**Requirements:**
- Record events: player logs in, logs out, completes a match, earns XP.
- Query: is player X currently online? What is player X's XP for today?
- Query: top 100 players by XP in the last 24 hours (leaderboard).
- Scale: 5 million concurrent players, 100K events/second.

## Architecture

**Event Ingestion:**
- Events published to Kafka partitioned by player_id for ordering guarantees.

**Online Status:**
- Redis key `player:{id}:online` with TTL refreshed on each heartbeat (60s). O(1) lookup.

**XP Tracking:**
- Redis sorted set `leaderboard:YYYYMMDD` with player_id as member, XP as score.
- `ZINCRBY` on each XP event. `ZREVRANGE 0 99` for top 100.
- Expire after 48 hours.

**Persistence:**
- Kafka consumer writes events to Cassandra (partition key = player_id, clustering by timestamp).

## Follow-ups

1. How do you handle the case where a player earns XP offline (mobile game, sync on reconnect)?
2. What happens to the leaderboard at midnight when the key rolls over?
3. How do you prevent cheating clients from inflating XP events?
4. How would you expose a REST API for these queries and ensure p99 latency < 20ms?

Game Activity: Design a Backend System to Track and Query Player Activity in Real Time

Question Details

Full Details

Topics

More from Roblox