Book a consultation

HiveMQ

HiveMQ Unified Namespace: Edge, Plant Broker and Enterprise

HiveMQ unified namespace explained: how HiveMQ positions UNS, the edge to plant to enterprise pattern, Sparkplug or JSON topics, and how it fails.

Vladimir Romanov


A HiveMQ unified namespace is three things: an MQTT broker, a way to get plant data into it, and a set of rules for what the topics and payloads look like. HiveMQ sells the first two and publishes guidance on the third. The rules are the part the plant has to own. This page covers how HiveMQ describes the pattern, which of its products sit at which level, when Sparkplug or plain JSON fits, and how a HiveMQ-based namespace goes wrong. For the generic pattern we assume our UNS guide, and for broker selection our UNS broker page. The rest of the HiveMQ cluster covers the products themselves.

How HiveMQ positions the unified namespace

HiveMQ’s blog series on the subject, written by Kudzai Manditereza, presents the UNS as an alternative to the centralised, layer-by-layer ISA-95 data architecture. In Part 1 (22 February 2026), components publish contextualised, changed-only data into a shared real-time hierarchy that HiveMQ calls the single source of truth. The article gives three principles: edge-driven, domain-owned data; an open architecture with MQTT as the standard infrastructure; and federated data governance.

It names the components as an MQTT broker, an IIoT platform that shapes the data, and data persistence, where a historian or SQL database subscribes. In a separate post, “Bringing Clarity to the Unified Namespace Approach” (22 February 2026, by Magnus McCune), HiveMQ states that the term was coined by Walker Reynolds. The sources give no year, so we give none.

Our read: HiveMQ treats the namespace as a current-state exchange, not a database, which matches how we design them. It also places the modelling work outside the broker, in an IIoT platform. That is the gap HiveMQ’s own modelling products aim at, covered on the Pulse page. A historian sits beside the broker; our data historian guide covers where.

A HiveMQ unified namespace: edge, plant broker, enterprise

In the same article, HiveMQ Edge converts point-to-point protocols to MQTT and can run an embedded broker for a production area or line. Devices that already speak MQTT, or connectivity tools such as KEPServerEX, feed a raw MQTT namespace. An IIoT platform shapes that into the UNS, and MQTT bridges connect brokers across levels and sites.

A Siemens Architecture Hub page, labelled a partner solution, describes a variant. PLCs and drives connect to Siemens Industrial Edge. A HiveMQ instance in each factory enforces topic structure, metadata and access control for HMI, SCADA and analytics. Local brokers bridge or tunnel selected topics to a central HiveMQ cluster serving ERP, MES and cloud analytics. The page is marketing and contains no joint-development text. Its claim that local brokers keep running during outages is a design statement, not a measured result.

Edge, plant broker and enterprise levels of a HiveMQ unified namespace Three columns from left to right. At the edge, PLCs, drives and sensors feed HiveMQ Edge, which runs protocol adapters, an embedded broker and ISA-95 topic prefixing. A bridge carries topics to the plant broker, which holds the site namespace and runs Data Hub validation, with SCADA, MES and historian consumers subscribed above it. A second bridge carries selected topics to an enterprise broker cluster, with ERP, lakehouse, Kafka and cloud analytics above it. A dashed band across the bottom shows an optional modelling and governance layer. Each level keeps its own broker; only selected topics move up EDGE (ISA-95 levels 1 to 2) PLANT (levels 2 to 3) ENTERPRISE (level 4) PLCs, drives, sensors S7, Modbus, OPC UA servers Site consumers SCADA, MES, historian Enterprise consumers ERP, lakehouse, Kafka, cloud HiveMQ Edge protocol adapters, embedded broker, ISA-95 topic prefix Plant broker site namespace, Data Hub schema checks, access control Enterprise broker cluster selected topics from every site, one enterprise view bridge bridge Optional modelling and governance layer information models, catalogue, schema rules (HiveMQ Pulse, status unconfirmed) Site screens read the plant broker, not the enterprise one, so a WAN outage does not blank them. Arrows show the direction data moves. Illustrative, not a HiveMQ drawing.
The pattern HiveMQ and Siemens both describe: a broker at each level and bridges that move only selected topics upward. The modelling layer is optional, and its product status is the open question on the Pulse page.

Why the plant keeps its own broker. A central-only broker makes every plant screen depend on the WAN. HiveMQ’s documented multi-site tool is bridging, not a cluster stretched across sites; our multi-site architecture page and the bridging page cover the detail.

Topic design guidance HiveMQ publishes

HiveMQ’s published guidance is general. Part 1 defers hierarchy design to a later part. Its FAQ gives four rules: define context centrally while letting local teams add detail; use standard templates per asset type; fix units and naming conventions; and validate payloads against schemas.

The Sparkplug namespace article (19 February 2026) warns that inconsistent topic structures harm interoperability and scalability, and says standardised topics make fine-grained access control easier. It does not mention ISA-95. HiveMQ Edge has an open-source feature that prefixes topics in the ISA-95 style, per the Edge documentation.

What we would add. Before the first device connects, write down five things: the level names and their order, naming rules, units and timestamp format, the payload schema with a version, and the owner of each branch. Then enforce them. An illustrative branch is enterprise/site/area/line/cell/state. The schema rule is what Data Hub can check at the broker.

Where Sparkplug fits and where plain JSON fits

Sparkplug topics have the form spBv1.0/group_id/message_type/edge_node_id/device_id. They are organised around edge nodes and devices, not around the plant hierarchy, and the payload is Protobuf. What Sparkplug gives you is session state: birth and death messages that tell a host whether data is live. Our Sparkplug page for HiveMQ covers how HiveMQ handles it.

The split we usually draw: Sparkplug where edge nodes and a SCADA host such as Ignition need state, and plain JSON on a plant-hierarchy topic tree for the contextualised data that MES, analytics and cloud consumers read. HiveMQ says its Data Hub Sparkplug module converts Sparkplug data to JSON without writing code, which is the join between the two. Plain JSON is the right choice when the publishers are not Sparkplug devices, for example Edge adapters or a DataOps tool. The Edge documentation we read describes no Sparkplug B publishing adapter, so ask HiveMQ before assuming Edge can emit it.

HiveMQ components in a unified namespace

ComponentRole in the namespaceWhat the sources say, November 2026
HiveMQ EdgeAdapters, embedded broker, ISA-95 topic prefix, bridgesApache 2.0 open-source features; Data Hub on Edge, offline buffering and southbound MQTT are commercial
HiveMQ Enterprise MQTT BrokerPlant and enterprise broker, masterless cluster, retained messagesCommercial; 4.57 is current (3 November 2026), 4.55 the long term support release
HiveMQ Community EditionA single broker for a line or labApache 2.0; listed by Eclipse as Sparkplug compatible
Data HubSchema validation, JavaScript transformation, behaviour policiesHiveMQ’s pricing page lists in-stream data governance in all self-managed plans; free mode is limited to one policy
Sparkplug Aware ExtensionStores and republishes Sparkplug birth certificatesApache 2.0; listed by Eclipse
Pulse (private preview announced February 2025)Information models, catalogue, governanceRelation to the September 2026 Contextualize layer not stated; general availability, licence and price not stated

The Data Hub page and the Edge page go further on the first and fourth rows.

Choosing HiveMQ for a unified namespace

The choice is less about the broker than about who holds the model. These four patterns appear in HiveMQ’s material and in public reference designs.

PatternWho holds the modelFitsWatch
Edge adapters, plant broker, Data HubA written convention plus Data Hub schemasOne site or several with OPC UA, Modbus or S7 sources and a team that will own the conventionMost Edge PLC adapters read only northbound; writes need a commercial southbound mapping
Sparkplug devices and a hostSparkplug group, node and metric namesPlants built around Ignition or similar hostsTopics follow nodes, not the plant hierarchy; Protobuf payloads
A DataOps tool publishing modelled JSONThe DataOps toolMany mixed sources needing shaping before publicationTwo products to support; HiveMQ’s HighByte partner page gives no data-flow detail
HiveMQ modelling productsHiveMQ’s model serverTeams that want vendor-hosted modelling and governanceStatus, licence and price not stated

In our view, HiveMQ fits when the plant wants an open, MQTT 5 broker that can be clustered and bridged, and will write its own namespace convention. When the broker is for one line, HiveMQ Community Edition or another light broker may be enough; see HiveMQ against its alternatives.

Aerial view of a large factory complex with silos, stacks and connected buildings
Every line and building feeds one namespace, so a naming rule that one area ignores becomes a problem for every consumer.

How a HiveMQ unified namespace fails

These are our views from MQTT and UNS design practice, not HiveMQ statements.

  • Topic sprawl. Without a written convention each integrator invents a branch. After two years nobody can say which topic is authoritative. The fix is review at the point of adding a branch, not a cleanup later.
  • No governance at the broker. MQTT accepts any payload on any topic. If Data Hub schema checks are not switched on, a malformed message reaches every subscriber. HiveMQ’s free Data Hub mode is limited to one policy, so check the licence against your plan.
  • Retained-message bloat. Each topic can hold one retained message, and the broker persists them. Decommissioned devices leave retained messages behind unless someone clears them with an empty retained payload. The Sparkplug Aware Extension adds birth certificates as retained messages too.
  • Sparkplug against flat topics. A team that builds an ISA-95 tree and a Sparkplug tree finds two namespaces describing the same asset. Decide which one consumers should trust and map the other into it.
  • Consumers that cannot decode Protobuf. A dashboard tool, a low-code flow or a cloud rule that expects JSON sees binary. Convert once, at a known point, and publish the JSON on a documented topic.
  • Wildcard subscriptions. A consumer that subscribes to # to find data then breaks when volume grows. Give consumers a branch to read.

What we do here

We design MQTT and Unified Namespace architectures and integrate brokers with PLCs, SCADA, historians and enterprise systems as part of IT/OT projects. We are independent of HiveMQ: we have no commercial relationship with the company and do not resell its products. In a HiveMQ-based namespace our work is:

  • Writing the topic convention, payload schemas and branch ownership before devices connect.
  • Deciding with the team where Sparkplug stays and where JSON takes over.
  • Placing brokers by plant level and network zone, with bridges and failure behaviour specified.
  • Specifying what a consumer is allowed to subscribe to and how it is granted.
  • Testing failure cases on a bench: broker restart, WAN loss, a device that stops publishing.

Our IT/OT integration service covers this work, and an OT network assessment comes first where the zones are unsettled.

Limits and open questions

  • UNS is a pattern, not a standard. HiveMQ’s definition is one among several; write your own.
  • Some HiveMQ details are unconfirmed. We found no HiveMQ statement on placement by ISA-95 or Purdue level, and no document saying whether a cluster may span a WAN.
  • The Eclipse Sparkplug list shows no version or date for HiveMQ’s rows. See the Sparkplug page.
  • Pulse status. General availability, licensing and price are not stated on pages we read.
  • Vendor material. Reference designs here are HiveMQ or partner marketing; none quantifies results for a UNS.

Deeper

  • HiveMQ Sparkplug: the Eclipse listing, the Sparkplug Aware Extension, the Data Hub module and the STATE permission pitfall.
  • HiveMQ Pulse and the 2026 suite: what is confirmed about Pulse, and what to ask before relying on a preview product.
  • Sibling: HiveMQ Edge, where plant data enters the namespace.

Frequently asked questions

Does HiveMQ provide a unified namespace out of the box?

No. HiveMQ provides the pieces: an MQTT broker, HiveMQ Edge to bring plant protocols onto MQTT, and Data Hub to validate and transform messages. The namespace itself, meaning the topic hierarchy, naming, units and payload schemas, is a design the plant writes and enforces. HiveMQ publishes guidance on it and offers commercial modelling products, but none replaces the decision.

Should a HiveMQ unified namespace use Sparkplug or plain JSON?

Often both. Sparkplug suits edge nodes and SCADA host applications such as Ignition, because it carries birth and death state. Plain JSON on an ISA-95 style topic tree suits the contextualised data that MES, analytics and cloud consumers read. A common pattern is Sparkplug inside the plant and a converted JSON namespace above it, which Data Hub's Sparkplug module is meant to support.

How does HiveMQ describe topic design for a UNS?

HiveMQ's published guidance is high level. It advises defining context centrally while letting local teams add detail, using standard templates per asset type, fixing units and naming conventions, and validating payloads against schemas. Its Sparkplug namespace article warns that inconsistent topic structures harm interoperability and scalability. A full hierarchy is left to the plant.

Where does HiveMQ Edge sit in a unified namespace?

At the bottom, near the machines. HiveMQ's articles describe Edge converting point-to-point protocols to MQTT and running an embedded broker for a production area or line. Its ISA-95 topic prefixing is an open-source feature. A plant broker or cluster above it holds the site namespace, and bridges carry selected topics to the enterprise level.

Can a HiveMQ broker store the history of a unified namespace?

It is not meant to. HiveMQ's own UNS article says the namespace should not serve as an ad-hoc query API, and that historians and SQL databases subscribe to it for persistence. The broker holds current state, one retained message per topic, and moves events. Put a historian beside it for trends and audit.

References and further reading

  1. Foundations of Unified Namespace Architecture for IIoT · HiveMQPart 1 of HiveMQ's UNS series, 22 February 2026: principles, components, reference architecture and the topic design FAQ.
  2. Understanding the MQTT Topic Namespace for IIoT · HiveMQThe Sparkplug namespace article, 19 February 2026. Its examples contain errors, so check them against the specification.
  3. HiveMQ Edge documentation · HiveMQWhich Edge features are open source and which need a licence key.
  4. HiveMQ, Architecture Hub · SiemensA partner-solution reference design: local HiveMQ per factory bridged to a central cluster. A marketing page.
  5. Bringing Clarity to the Unified Namespace Approach · HiveMQ
  6. HiveMQ Data Hub documentation · HiveMQData policies, schemas, transformations and behaviour policies.
  7. Sparkplug Compatible Software · Eclipse Foundation

← HiveMQ

Have a project that needs technical ownership?

Send us the situation. If it is not something we should take on, we will say so.

The Plant-Side Brief

The weekly brief for the people who own a plant’s control systems.

One email every Wednesday for engineering and operations leaders responsible for controls, SCADA, MES and OT networks. Skim the top in two minutes; read the Deep Dive when you have fifteen.

  • One modernization problem, worked through to cost, risk and a schedule you can defend. The Deep Dive.
  • End-of-support dates before they reach your plant Windows, servers, controllers and HMI software. Lifecycle Watch.
  • A checklist your engineers can run that week not a framework to admire. Field Kit.
  • Every number sourced, no vendor agenda we are paid by plants, never by vendors.