Design real-time game matchmaking service
Overview
This question evaluates a candidate's competence in designing scalable, low-latency distributed systems and matchmaking algorithms, including skills in queue management, skill-based matching, regional sharding, SLO-driven latency targets, consistency models, failure handling, and monitoring.
Roblox
Sep 6, 2025, 12:00 AM
Software Engineer
Onsite
System Design
0
0
Design a Real-Time Game Matchmaking Service
Context
You are designing the matchmaking subsystem for an online multiplayer game that supports multiple regions, game modes (e.g., 1v1, 3v3, 5v5), and both solo and party queues. The service must form balanced matches quickly while respecting latency and regional constraints. It must scale to millions of daily users and handle failures gracefully.
Goals and Requirements
-
Goals
- Latency: Time-to-match (TTM) p50 ≤ 3s, p95 ≤ 10s for casual; p95 ≤ 30s for ranked/high-MMR.
- Fairness/Skill Balance: Minimize team MMR deltas; keep within bands that widen over time in queue.
- Regional Constraints: Prefer local region; allow cross-region fallback with explicit user opt-in or time-based expansion.
- Party Integrity: Parties matched together; preserve role preferences if applicable.
- Backfilling: Fill empty slots when players drop, with minimal disruption.
-
APIs to define
- join_queue
- leave_queue
- heartbeat (client, game server)
- allocate_server / assign_match
- match_status
-
Design aspects to cover
- Data model
- Match formation algorithm (skill bands/ELO, queueing, timeouts, backfilling)
- Scaling and sharding strategy
- Consistency model
- Failure handling and hotspots
- Monitoring/metrics and SLOs
- Capacity estimates
- Trade-offs
Solution
Login required
Show