nClarity, IoT analytics SaaS platform on Azure
Event-driven telemetry pipeline with zero-loss replay
role: Ran the telemetry pipeline and designed its replay and backfill.
// problem
Field devices stream telemetry continuously. The pipeline has to turn it into alerts for customers and survive outages without losing data.
// approach
- Devices stream through Event Hubs into event-driven Azure Functions and Node.js services.
- Time series land in InfluxDB; a rules engine turns readings into alerts.
- React dashboards deliver readings and alerts to HVAC and refrigeration customers.
- Replay and backfill were designed in from the start.
- Firmware rollouts through Azure Device Update.
// architecture
Field devices
Event Hubs
Ingestion functions
InfluxDB
Rules engine
Alerts
React dashboards
// guardrails and production concerns
- Replay and backfill designed into ingestion, not bolted on after an outage.
- Application Insights alerting.
// outcome
Ingestion for a fleet that reached ~10k devices, was consolidated and is scaling again. After a 16-hour ingestion outage, I proved a zero-loss replay.
// stack
- Azure Functions
- Event Hubs
- IoT Hub
- InfluxDB
- Node.js
- TypeScript
- React