Scaling Real-Time Mobile Game Backends to 50k Concurrent Users with ASP.NET Core & Redis
The Challenge of Mobile Game Launch Spikes
When a mobile game is featured in app stores, traffic does not ramp up linearly — it spikes by 50x in minutes. Every user connection requires token authentication, inventory synchronization, live matchmaking pooling, and state persistence.
If every incoming payload hits a relational database directly, connection pools exhaust within seconds.
```csharp // Example: Thread-safe high-throughput session state caching in Redis public async Task<PlayerSessionState> GetOrRefreshSessionAsync(string playerId, CancellationToken ct) { var cacheKey = $"player:session:{playerId}"; var cached = await _redisDatabase.StringGetAsync(cacheKey); if (cached.HasValue) { // Extend sliding expiration await _redisDatabase.KeyExpireAsync(cacheKey, TimeSpan.FromMinutes(30)); return JsonSerializer.Deserialize<PlayerSessionState>(cached); } // Fallback to cold database storage var state = await _repository.LoadPlayerStateAsync(playerId, ct); await _redisDatabase.StringSetAsync(cacheKey, JsonSerializer.Serialize(state), TimeSpan.FromMinutes(30)); return state; } ```
Architecture Pillars
1. **Separation of Read State vs. Write State:** Read-heavy match stats are stored purely in memory in Redis cluster shards. 2. **WebSocket Keep-Alives & Heartbeat Aggregation:** Preventing TCP socket exhaustion by batching client pings every 5 seconds. 3. **Optimistic Locking for Currency and Inventory:** Atomic Redis scripts (`EVAL`) for currency deduction prevent race conditions without table locks.