The Unified Namespace (UNS) stack, explained with real tools

A Unified Namespace (UNS) is an architecture pattern where every system in a plant publishes its current state to one shared, structured place, usually an MQTT broker, instead of wiring point-to-point integrations between each pair of systems.

Consumers (MES, historians, dashboards, AI tools) subscribe to the data they need. Adding a new system means connecting it once to the namespace, not to every other system.

Here are the usual layers and tools commonly used in each. Integration details change often, so always confirm with the vendor.

1. Connectivity (PLCs to standard protocols)

Translate PLC and device protocols into OPC UA or MQTT.

Kepware KEPServerEX (PTC): driver-based OPC UA/DA server with a large PLC and device protocol library.
Softing dataFEED: OPC UA and MQTT gateways for Siemens and Rockwell PLCs.
Matrikon OPC (Honeywell): OPC connectivity, including classic-to-UA tunnelling.
Cirrus Link MQTT Modules: MQTT/Sparkplug B modules for Ignition, for report-by-exception data movement.

2. MQTT broker (the namespace itself)

Holds the live, hierarchical topic tree that everything publishes to and subscribes from.

HiveMQ: enterprise MQTT broker, frequently the backbone of a UNS.
EMQX: high-throughput broker with open-source and enterprise editions.
Eclipse Mosquitto: lightweight open-source broker, often the first one in a pilot.

3. Edge and contextualization

Normalize, model and add context (site, area, line, asset) before data reaches the broker or the cloud.

HighByte Intelligence Hub: industrial DataOps that models and contextualizes OT data.
Litmus Edge: vendor-neutral platform for collecting, normalizing and forwarding machine data.
Ignition Edge (Inductive Automation), Siemens Industrial Edge, Node-RED and FlowFuse also commonly sit here.

4. SCADA / HMI

The operators' view and control, often both a publisher to and a consumer of the namespace: Ignition, SIMATIC WinCC and WinCC Unified, WinCC OA, FactoryTalk View, AVEVA System Platform, Proficy iFIX and CIMPLICITY, ICONICS GENESIS64, VTScada.

5. Historian and time-series storage

The namespace holds current state; a historian keeps the history: Canary, AVEVA PI System, Proficy Historian, FactoryTalk Historian, Factry Historian, or general time-series databases like InfluxDB and TimescaleDB.

6. Analytics and industrial AI

Consume contextualized history to find root causes, predict and optimize: Seeq, TrendMiner, dataPARC, Cognite Data Fusion, Siemens Insights Hub, Flow.

Questions to settle before you build

Which MQTT payload convention will you standardize on (Sparkplug B or plain JSON)? Many designs mix Sparkplug for device data with JSON topics for contextualized data.
Where is context (asset hierarchy, units) added: at the edge, or in a hub?
What must stay on-premises, and what can go to the cloud?
Who owns the topic structure and naming rules?

Side-by-side comparisons for each layer

Originally published on PlantWired, a free directory of industrial software with side-by-side comparisons. New comparisons go out in the PlantWired newsletter.