Case Study
Scaling 10,000 Players: Distributed State in Real-Time Tactical Environments
A case study in H3 spatial sharding, worker-per-region consistency, and multi-tier storage architecture
TL;DR
- • Designed a distributed real-time game engine targeting 10,000 concurrent players in a single contiguous battle at 2Hz tick rate
- • H3 hexagonal spatial sharding with worker-per-region consistency model (Redis hot state + etcd coordination + DynamoDB durability)
- • Epoch-based fencing tokens for split-brain resolution; influenced by CockroachDB's range leases and Kleppmann's work
- • Architecture validated through theoretical analysis; load testing planned for Q3 2026
- • Aims to exceed EVE Online's 6,557-player record without time dilation (they used 10% real-time at that scale)
- Role
- Solo architect and engineer
- Timeline
- 2026, ongoing
- Stack
- Go, Redis 7, etcd, DynamoDB, Kubernetes, H3 (Uber's hexagonal spatial index)
- Key challenge
- Real-time multiplayer consistency at massive scale without time dilation
- Status
- Architecture validated through theoretical analysis and peer review. Load testing in progress.
Full case study content coming soon.
In the meantime, explore the architecture through these posts: