Thingsboard: From Arduino into Scalable Industrial SCADA

There is a myth that Thingsboard is just for “makers” playing with ESP32s. Nothing could be further from the truth.

When configured correctly, Thingsboard is a beast capable of handling millions of messages per second, serving as the operations brain for entire fleets of machines.

The problem? Most use it as a simple raw data viewer.

This guide condenses years of implementing Thingsboard in manufacturing environments (Purdue Model Level 3), transforming it from a “toy” into a Cloud-Native SCADA.

1. Architecture: Don’t Use HSQLDB in Production

First is the database.

• PostgreSQL: For storing entities (devices, users, clients).
• TimescaleDB / Cassandra: For telemetry (time series).
• Redis: For caching.

If you install Thingsboard with the default configuration (HSQLDB), it breaks in a week. Use Docker Compose or Kubernetes to orchestrate these microservices.

2. Rule Chains: The Hidden Brain

This is where beginners fail. They send dirty data from the device and expect to fix it in the Dashboard. Mistake.

The Dashboard should be dumb. Intelligence lives in the Root Rule Chain.

Normalization Pattern

Use Script (Javascript) nodes to standardize payloads before saving them.

// Example: Convert F to C and ensure structure
var newMsg = {};
if (msg.tempF) {
    newMsg.temperature = (msg.tempF - 32) * 5/9;
} else if (msg.tempC) {
    newMsg.temperature = msg.tempC;
}

// Enrich with device metadata (Server Attributes)
newMsg.lineName = metadata.lineName;
newMsg.maintenanceDue = metadata.maintenanceDate < Date.now();

return {msg: newMsg, metadata: metadata, msgType: msgType};

This ensures your database always has clean and enriched data (temperature always in Celsius), regardless of whether the sensor is Chinese, German, or American.

3. Dashboards: High-Performance Design

An Industrial Dashboard is not a color fair. It must answer two questions in less than 3 seconds:

1. Are we producing according to plan?
2. Is something breaking?

Dashboard Hierarchy

• Level 1 (Fleet): Map with states (Green/Red) of all plants.
• Level 2 (Plant/Line): OEE, production count, availability.
• Level 3 (Machine): Raw sensor values, motor currents, temperatures.

Don’t mix levels. A manager doesn’t want to see bearing 3 temperature; they want to see Line 4 OEE.

4. Community vs Professional Edition (PE)

Is it worth paying?

• Use CE (Free) if: You have a strong DevOps team, don’t need automatic PDF reports, and the simple “Tenant” structure is enough.
• Use PE (Paid) if: You need “Whitelabeling” (putting your logo and selling it as yours), native integration with AWS/Azure, or complex Customer Hierarchies. Scheduled PDF report export is often the deciding factor for management.

Conclusion

Thingsboard is the “Swiss Army Knife” of IIoT. But like any sharp tool, it requires discipline. Normalize your data at the edge or in Rule Chains, invest in a robust database, and design with the operator in mind, not the developer.