industrial//industrial data layer//MQTT in the plant

MQTT in the plant is the use of MQTT on the OT/IT boundary to distribute plant data between systems: a PLC or, more often, an edge gateway publishes it to a broker, and SCADA, MES, the historian, analytics and the cloud each subscribe. What travels is telemetry and events, published and subscribed over TCP/IP.

MQTT in the plant is the use of MQTT on the OT/IT boundary to distribute plant data between systems: a PLC or, more often, an edge gateway publishes it to a broker, and SCADA, MES, the historian, analytics and the cloud each subscribe. What travels is telemetry and events, published and subscribed over TCP/IP.

The producer does not need to know its consumers. A new analytics tool is a new subscription, not a new connection to the PLC; that decoupling is what lets a plant add consumers without touching control (publish-subscribe).

MQTT divides the labor with the fieldbuses: Modbus, PROFINET and the like read and write devices, while MQTT distributes information between systems. A sound architecture puts the PLC on a fieldbus, a gateway in the middle and MQTT toward IT and the cloud. MQTT was never meant to move a robot axis (Modbus).

It is often weighed against OPC UA: OPC UA is a server you browse and query and tells you what the data means, MQTT is a stream you subscribe to and gets it to everyone cheaply. Plants that pick one over the other usually end up carrying OPC UA payloads inside MQTT topics.

The topic tree becomes the plant's addressing scheme (factory, line, machine, signal), and with a shared payload convention such as Sparkplug it grows into a unified namespace (data contextualization). Brokers can be on premises, hosted, bridged and redundant; store-and-forward at the gateway covers link outages.