industrial//industrial data layer

The layer that turns a plant designed to control machines into an industrial data infrastructure. The full path of a datum: sensor, then PLC, then an edge gateway or IPC, then MQTT or an OPC UA server, and from there fan-out to SCADA, the historian, MES, analytics and the cloud.

The layer that turns a plant designed to control machines into an industrial data infrastructure. The full path of a datum: sensor, then PLC, then an edge gateway or IPC, then MQTT or an OPC UA server, and from there fan-out to SCADA, the historian, MES, analytics and the cloud.

Before: sensor, then PLC, then SCADA. One chain, one consumer, built to control.

Now: sensor, then PLC, then edge, and from the edge in parallel to SCADA, historian, MES, analytics, cloud and AI. One source, many consumers, built to exploit data as well.

The unit of data is the tag with value, timestamp and quality; the other kind of datum is the event. Storage is the historian with deadband compression, alongside a time-series database.

The hard part is not moving bytes but meaning: data contextualization is what turns a register value into "motor M17, batch B562, customer X", and fault semantics is why control protocols cannot be swapped for a REST API. OPC solved interoperability first; OPC UA added the information model.

Zooming out: IIoT connects assets to get more from them, telemetry carries their variables out, fleet data compares many plants at once, and edge AI runs the analysis next to the machine. All of it lives on the boundary described in OT.