Enterprise Mining Farm Management System

Bilttek operates one of Europe's largest crypto mining operations. Their existing tooling was a patchwork of vendor dashboards, spreadsheets, and manual SSH checks. Operators had no real-time view of individual miner health and no way to respond automatically when electricity prices spiked.

The business case was clear: if miners could automatically pause during high-price windows and resume when prices dropped, the energy savings would be substantial. But doing that reliably across 10,000+ devices, without missing a beat, required rethinking the entire infrastructure from the ground up.

Every architectural decision was made with one constraint in mind: the system must never lose a command or a telemetry event. That led to a fundamentally event-driven design.

Direct HTTP polling of 10,000 devices was immediately ruled out, the latency, retry complexity, and server load would make it unworkable. Instead, each miner connects via MQTT, a protocol designed for constrained IoT devices with unreliable connections. Commands from the dashboard flow into RabbitMQ, which fans them out to the appropriate MQTT topics and guarantees delivery.

For metrics, PostgreSQL was the right choice for relational data (users, device configs, audit logs, NordPool schedules) but wrong for high-frequency miner telemetry. InfluxDB handled millions of time-series data points per day with efficient retention policies and fast aggregation queries for the real-time dashboard.

Redis served as the distributed state layer, every miner's live status (online/offline, temperature, hashrate, power draw) was cached with short TTLs, giving the UI sub-second reads without putting read pressure on the primary database.