Network infrastructure and server connections

In most plants, systems are connected point to point: PLC to SCADA, SCADA to historian, historian to MES, MES to ERP. Every new application means a new connection, a new mapping and a new maintenance burden. A Unified Namespace (UNS) turns this structure around: every system publishes its current state into one shared namespace, and anyone who needs data reads it from there.

In short

A UNS is not a product but an architectural rule: “Data is published once, to a known address; consumers do not need to know the publisher.” In practice it is usually built with an MQTT broker and an ISA-95-style topic hierarchy.

The structure of a UNS

Unified Namespace (UNS) hierarchyData is published on a single MQTT broker in an ISA-95-style hierarchy from enterprise down to cell; MES, historian and analytics systems consume from the same namespace.EnterpriseSiteAreaLineCellPublishersPLCs, edge gateways,MES, ERPUNS (MQTT broker)One namespace, a single sourceof truth for current stateConsumersHistorian, analytics,MES, dashboardsenterprise / site / area / line / cell / measurementThe topic name carries the hierarchy; publisher and consumer need not know each other
A UNS is an architectural rule rather than a product: every system publishes, every system reads what it needs from the same place.

Publishers (PLCs, edge gateways, MES, ERP) send their data to the broker under hierarchical topic names. The hierarchy typically runs from enterprise to cell: enterprise / site / area / line / cell / measurement. Consumers (historian, analytics, dashboards, MES) subscribe to the branches they need. The broker is the single source of truth for current state.

What a UNS is not

  • It is not a historian. A UNS carries current state; for history, a historian subscribes to the UNS and records data.
  • It is not just MQTT. MQTT is a common carrier; the real value lies in naming and data-model discipline.
  • It does not replace MES or ERP. They can also be publishers and consumers of the UNS.
  • It is not a package you buy. Products help; the real work is naming and ownership decisions.

Step-by-step setup

  1. Define the naming hierarchy. The ISA-95 levels are a good start. Choose lowercase, space-free, ASCII names; names should be permanent, and only values should change.
  2. Choose the broker topology. A redundant in-plant broker, bridged to a central broker if needed. The broker sits in a DMZ at the IT/OT boundary, and the OT side makes outbound connections only.
  3. Connect the publishers. Legacy devices need an edge gateway: Kepware Server collects the data and publishes it with IoT Gateway; OPC Router can transform and publish source systems. Newer devices can speak MQTT or OPC UA PubSub directly.
  4. Define the data model. Every message should carry value, unit, source timestamp and quality. Sparkplug B provides this structure and the BIRTH/DEATH lifecycle ready-made; alternatively use a documented JSON schema (choosing the data model).
  5. Add consumers. Historian, dashboards, analytics and MES subscribe only to the branches they need.
  6. Set up governance. The owner of each branch, schema versioning, publish-rate rules and topic-based access lists should be written down.

Topic name examples

LevelMeaningExample name
EnterpriseThe organization’s root nameenterprise
SitePhysical locationenterprise/site-1
AreaProduction area or departmententerprise/site-1/filling
LineProduction lineenterprise/site-1/filling/line-2
CellMachine or stationenterprise/site-1/filling/line-2/filler
MeasurementData point…/filler/temperature

Common mistakes

  • A flat “topic soup”: topics with no hierarchy, named by each team on its own, turn the UNS into a mess.
  • Ownerless branches: topics with no clear responsible party go stale over time.
  • Publishing everything every second: publishing on change (deadband) is often enough.
  • Leaving security to the end: TLS, authentication and topic-based access should be designed from the start.
  • Skipping quality and time: a value alone is not enough (data quality and timestamps).

Start with a small pilot

Pick a single line, roughly twenty to fifty measurements and one consumer (for example a dashboard). Test naming and the data model in this pilot, then extend to other lines and systems. For using OPC UA and MQTT together see OPC UA or MQTT?; for terminology see the glossary.

Frequently asked questions

Is Sparkplug B mandatory for a UNS?

No. It is a common choice because it provides a shared payload format and device lifecycle rules, but a UNS can also be built with a documented schema.

Where is historical data kept in a UNS?

A historian application subscribes to the UNS and records data with timestamps. The UNS itself carries current state.

Do I have to use the ISA-95 hierarchy?

It is not required, but it is a good start because it offers a familiar and extensible structure. What matters is that the hierarchy is consistent and permanent.