ThingsBoard Industrial Monitoring Architecture Guide

ThingsBoard can collect, process, and visualize industrial data, but it should not be described as a drop-in SCADA replacement. A conventional SCADA, PLC, or safety system may own control, deterministic response, interlocks, and certified functions. ThingsBoard is strongest as a supervisory monitoring and integration layer whose behavior depends on its edition, version, deployment, and engineering.

This guide defines that boundary and a practical path from one synthetic device to an operable installation. Product behavior and packaging were checked against official documentation on 2026-07-24; confirm them again for the exact release and subscription you deploy.

Start with the information model

Use telemetry for timestamped measurements whose history matters: temperature, pressure, energy, or machine state transitions. ThingsBoard stores a key, value, and timestamp; if the device omits the timestamp, the server assigns one. Use attributes for current metadata or configuration instead of manufacturing a time series.

Attribute scope matters. Client attributes are reported by a device. Server attributes are platform-managed and unavailable for device modification. Shared attributes are written by the platform and readable by the device, making them suitable for non-safety configuration such as a reporting interval. They are not a guaranteed command channel: delivery varies by protocol, and the device must validate, apply, and acknowledge changes.

Name keys consistently, include engineering units in documentation, and keep thresholds as attributes only when operations can govern their changes. Dashboards and rules should consume the same stable model rather than each renaming raw tags independently.

Prove the flow with one synthetic device

Create only a synthetic device named tb-lab-pump-01. Publish this deterministic JSON telemetry fixture through a secured test connection:

{ "ts": 1753351200000, "values": { "temperature_c": 84.5, "pressure_bar": 4.2, "running": true, "sequence": 17 } }

Set a server-side temperature_limit_c attribute to 80. The fixture is intentionally above that threshold. It contains no plant address, credential, command, or real equipment identity, so it is safe for a repeatable lab check.

Keep rule chains explicit

Route incoming telemetry through a small Rule Chain: validate required keys and types, enrich the message with temperature_limit_c, compare temperature_c with the threshold, save accepted telemetry, and create or clear an alarm. Send malformed or missing-threshold messages to an observable failure path instead of silently accepting them.

Rule Engine messages carry an originator, type, JSON payload, and metadata. Nodes can filter, enrich, transform, and route them. Avoid hiding core semantics in a large script node when standard nodes express the flow. Debug mode is useful during commissioning but produces additional events; disable or constrain it after diagnosis according to the deployed version.

For the fixture, the deterministic result is HIGH_TEMPERATURE because 84.5 > 80. Replaying the same timestamp and sequence should support an idempotency policy upstream or in the processing design; do not assume every transport delivers exactly once.

Treat alarms as an operating lifecycle

An alarm needs more than a red widget. Define severity, creation condition, clear condition, acknowledgement responsibility, owner, and response procedure. Add hysteresis or duration where noisy signals would chatter. Test normal, active, acknowledged, cleared, stale-data, and missing-attribute states.

Dashboards should expose current value, unit, timestamp or freshness, trend, alarm state, and data-quality context. Color should reinforce abnormal states rather than decorate normal operation. A dashboard is an operator view, not evidence of deterministic latency or functional-safety behavior.

Secure MQTT identity and transport

ThingsBoard documents access-token, MQTT Basic, and X.509 device credentials. For production networks, use MQTT over TLS, validate the server certificate and hostname, give each device a unique identity, rotate credentials, and revoke them when equipment leaves service. Mutual TLS with X.509 may be appropriate when certificate lifecycle tooling exists.

Do not expose an unauthenticated or plaintext broker to untrusted networks. Keep secrets outside articles, images, source control, dashboards, and Rule Chain metadata. Firewall transport endpoints, restrict management access, monitor authentication failures, and test certificate expiry and rotation. Supported credential details are release-dependent, so use the documentation for your installed version rather than copying an old client command.

Put the gateway at the protocol boundary

The open-source ThingsBoard IoT Gateway bridges legacy or third-party systems through connectors such as OPC UA, Modbus, BACnet, and MQTT. Place it near the equipment network to map source identifiers into the platform model and buffer according to tested connector behavior.

The gateway does not turn cloud processing into a control loop. Keep PLC interlocks and time-critical control local. Start read-only, allowlist required tags, document scaling and byte order, and add writes or RPC only after a separate hazard and authorization review. Loss of the gateway or WAN must leave the process in its independently engineered state.

Plan retention, capacity, and rollback

Retention is a capacity decision, not “keep everything.” Estimate devices × keys × reporting rate × retention, then load-test representative payloads, rules, alarms, dashboard queries, reconnect storms, and exports. PostgreSQL is the documented default for time series; supported alternative storage and topology choices vary by edition and version. Cloud plans impose subscription-specific retention and API limits.

Community Edition is self-managed and provides core connectivity, processing, alarms, and dashboards. Professional Edition adds licensed capabilities such as advanced RBAC, integrations, reporting, and other enterprise functions. ThingsBoard Cloud is managed and based on Professional Edition, with plan limits. Verify the current comparison before designing around any feature. Horizontal scaling or high-availability architecture is not a guarantee that your deployment is highly available; prove failure behavior against your own recovery objectives.

Back up the database and every external dependency: configuration, secrets through their approved store, certificates, gateway files, and deployment manifests. Version-control export can help with supported entities in editions/releases that provide it, but it is not a database backup. Before an upgrade, record versions, read the exact upgrade path, take a recoverable backup, rehearse restoration in isolation, and define rollback criteria. A backup is credible only after a timed restore test.

Official references

Accessed 2026-07-24:

ThingsBoard can be a useful industrial data layer when its responsibility is narrow and testable: ingest known data, process it visibly, present trustworthy context, and fail without taking control away from the systems designed to own the process.