A plant integration project can start with a modest request: send equipment data to a historian and make it available to an energy dashboard. The scope changes when another application needs the same measurements, a remote site loses connectivity, or engineers discover that identical tag names represent different operating conditions. Protocol selection then affects how the team acquires, describes, distributes, and maintains the data.

For equipment integration, evaluate OPC UA when the project needs structured access to an exposed industrial information model. Evaluate MQTT when the priority is distributing published messages through a broker. A combined architecture may serve both requirements, and OPC UA PubSub explicitly supports MQTT transport. These are architectural options to verify against the products involved. [1] [3] [8]

Start the decision by identifying the controllers that already expose usable interfaces and the applications that need current values, historical records, or equipment status. Include the functions that must continue during an outage, then use those requirements to define the protocol boundary and acceptance test. A successful pilot should demonstrate that the intended applications receive usable data under agreed operating conditions, including the conditions that interrupt normal communication.

What changes when you choose MQTT or OPC UA?

The comparison involves two scopes. MQTT defines a messaging transport whose payload can follow an application-specific format. OPC UA includes an AddressSpace that represents objects and their relationships through nodes, attributes, and references. That distinction affects how much meaning an integration can obtain from its interface and how much the project must define elsewhere. [12] [3]

MQTT: A publish/subscribe messaging protocol that distributes application messages through a server, commonly called a broker, allowing publishers and subscribers to communicate without direct connections to each other. [1]

Use the following table to establish the comparison boundary. The OPC UA column describes conventional client/server integration; its PubSub model is covered separately below.

Decision areaMQTTOPC UA client/server
Data exchangePublishers and subscribers communicate through a brokerClients access services exposed by a server
Data meaningDefined by the payload convention or an additional specificationRepresented through an AddressSpace and information model
Change reportingPublished messages reach matching subscribersMonitoredItems and Subscriptions report changes and events
Project design workDefine topics, payloads, permissions, and consumer behaviorVerify exposed nodes, metadata, services, and server limits
Local deploymentCan use a broker within the site networkCan connect clients and servers within the site network
First acceptance checkCan each intended subscriber interpret the published data?Does the server expose the required information and services?

Protocol features in the table follow the MQTT and OPC UA specifications; the acceptance checks are engineering recommendations. [3] [4] [12]

When does OPC UA fit the integration task?

OPC UA deserves close evaluation when equipment already exposes a useful information model and the integration depends on its structure. A client can work with nodes and their relationships instead of relying entirely on an external description of each signal. The value of that structure depends on what the equipment supplier has actually modeled and exposed. [3]

OPC UA information model: A structured representation of industrial objects, variables, types, and relationships that allows applications to interpret exposed information within an OPC UA AddressSpace. [3]

For a compressor, inspect whether the interface identifies the asset, its operating state, and the measurements needed for the analysis. Then check the associated data: OPC UA DataValue defines status and timestamp fields alongside the value. During commissioning, verify which timestamps are populated, what their source is, and how unusable values are represented to the receiving application. [5]

OPC UA also supports Subscriptions and MonitoredItems for notifications about data changes and events. A comparison that describes it only as repeated polling leaves out this communication mechanism. Review the sampling interval, reporting configuration, filters, and queue behavior supported by the selected server. [4]

Before selecting an OPC UA connection, browse a representative device and build a small mapping specification. Record the required nodes, expected types, units, and behavior when a source becomes unavailable. Ask the supplier to demonstrate the configuration under the intended client workload. Use the existing OPC UA integration guide for a deeper introduction, while keeping the selection exercise focused on your equipment.

When does MQTT fit the integration task?

MQTT is worth evaluating when a project needs published measurements distributed to several applications or collected from remote devices. Its publish/subscribe approach allows applications to communicate through a broker, while its lightweight design addresses environments where network bandwidth and device resources matter. These characteristics support a design discussion; they do not establish a universal performance advantage over OPC UA. [1]

The payload contract needs particular attention. MQTT does not prescribe an industrial equipment model for the message body. A topic such as plant-a/utilities/compressor-02/power can help organize routing, but the project still needs an agreed interpretation of the measurement. Decide where units, timestamps, asset identity, and validity information belong, then document how applications will handle missing fields or later schema changes. [12]

Plan broker placement around the site's network boundaries and operating requirements. A local broker may suit an installation that must remain independent of an external connection. A remote broker may suit a different distribution requirement. In either case, identify who maintains the service, approves topic permissions, manages credentials, and monitors failures. MQTT does not require a public cloud deployment.

What must you verify about security and data delivery?

Begin with the selected configuration: OPC UA defines security modes including None, Sign, and SignAndEncrypt, while MQTT's security considerations describe TLS protection and authentication mechanisms. For either deployment, document which protections are enabled, how credentials are renewed, and which applications can access each resource. [6] [2]

Separate connection protection from protection of the message payload. With brokered communication, encrypted connections can protect traffic between clients and the broker while leaving the broker able to read the content. OPC UA PubSub defines message security separately from transport security, allowing those boundaries to be evaluated independently. Verify the encoding and key-management support in the chosen implementation. [7]

MQTT quality of service: A set of delivery levels that governs message transfer between MQTT protocol participants, providing at-most-once, at-least-once, or exactly-once behavior within the applicable protocol exchange. [12]

QoS 0 permits message loss; QoS 1 permits duplicates; QoS 2 avoids duplicate delivery within its protocol exchange. None of these levels proves that a historian committed a record or that equipment executed a command. Define application acknowledgments and duplicate handling where the business requirement depends on those outcomes. [12]

Recovery needs its own test. MQTT 5 includes session and message expiry controls, while retained messages provide a latest retained value for a topic rather than a time-series archive. Specify publisher-side buffering, broker persistence, expiry settings, and the receiving application's recovery procedure separately. [2]

Disconnect the remote link during the pilot and continue generating representative data. Restore communication, then compare the source records with the stored results. Check for gaps, duplicates, changed timestamps, and delayed values displayed as current. Repeat the exercise with the historian temporarily unavailable while the broker remains online. These tests examine different failure boundaries and give the team evidence for its recovery claim.

How do Sparkplug and OPC UA PubSub change the comparison?

Sparkplug adds defined conventions around MQTT for SCADA and IIoT applications. Its specification covers the topic namespace, payload, and state-management behavior. Birth and death messages help participating applications understand the availability and state of edge nodes and devices. This makes Sparkplug materially different from a project-specific JSON message published through MQTT. [9]

OPC UA PubSub provides another option. The OPC Foundation specification defines how PubSub messages can use MQTT transport. A project can therefore retain OPC UA PubSub message conventions while using a broker for distribution. The receiving applications must support the applicable mappings and configuration. [8]

When evaluating products, request the exact supported specification and version. Ask whether a proposed MQTT interface accepts arbitrary payloads, supports Sparkplug, or implements OPC UA PubSub over MQTT. Have the supplier demonstrate the receiving application consuming a representative message, including its metadata and state changes. Treat these as separate interoperability checks. General MQTT support provides insufficient evidence for the other two capabilities. [8] [9]

Have something in mind to discuss?

We're here to help you find the answers. Let's talk.

Book a call

How can MQTT and OPC UA work together?

Suppose an energy engineer wants to compare compressor power with operating state and discharge pressure. The historian needs records for later analysis, while a live dashboard needs the latest usable readings. The following worked example uses one OPC UA source and an MQTT gateway to supply both consumers. All node IDs, readings, timestamps, and message fields are illustrative design choices, not a customer deployment or a verified CENTO configuration.

Start with the data exposed by the OPC UA server

In this example, the server exposes power and pressure as Double variables and operating state as a String. The table shows their latest values and associated status at the gateway. All times are UTC on October 22, 2026; each row retains its own SourceTimestamp.

Illustrative NodeIdValueUnitSourceTimestampStatus
ns=2;s=C02.Power54.2kW10:00:00.000ZGood
ns=2;s=C02.DischargePressure7.1bar10:00:00.020ZGood
ns=2;s=C02.OperatingStateLoadedNot applicable09:59:55.000ZGood

An OPC UA DataValue carries the value, StatusCode, and timestamps. Engineering units are separate variable metadata; the gateway obtains them from the exposed model or a documented mapping. The example maps Good to 0x00000000. Actual servers define their own node IDs and available metadata. [5] [13] [14]

For the power reading, assume the server reports a ServerTimestamp of 10:00:00.040Z. That timestamp describes when the server received or knew the value to be accurate; the SourceTimestamp identifies its origin-side time. The state timestamp is older because that value last changed earlier. Grouping the readings into one message does not make them simultaneous measurements. [5]

Map those values into an explicit MQTT message

The gateway subscribes to the selected OPC UA variables and publishes this application-specific JSON on plant-a/utilities/compressor-02/telemetry. This example chooses QoS 1 and retain=false; those settings are project choices. The payload follows a custom contract, separate from Sparkplug and OPC UA PubSub encodings.

The power object preserves 54.2 kW, its source time, and the original status code. gateway_published_at records a later transport step. After an outage, the receiver can distinguish a delayed measurement from a new reading by retaining both times. In this contract, the gateway keeps event IDs unique across restarts and preserves them on replay; a historian consumer uses the ID to recognize an already processed message. The event ID is an application convention and is separate from an MQTT packet identifier. [2]

Follow the same message through the architecture

One OPC UA source, two MQTT consumers An OPC UA compressor server sends data to a gateway. The gateway maps values, timestamps and status into custom JSON and publishes to an MQTT broker. The broker forwards messages independently to a historian consumer and an energy dashboard.

The gateway owns the node-to-field mapping and any local offline buffer. The broker routes messages to the two subscriptions. The historian consumer validates and commits records; the dashboard interprets the latest readings and their status. These are logical roles that may share a host. A protocol acknowledgment from the broker does not confirm that either consumer completed its work. [12]

The power, state, and pressure fields let the engineer select records associated with loaded operation and compare them with other operating states. The example provides the necessary context; it does not establish an energy baseline or prove avoidable losses from one message. Historical comparisons still need enough operating data for the question being investigated.

Preserve failures as clearly as normal readings

If the OPC UA source reports Bad status, the value is unusable. Under this example's contract, the gateway sends null for that metric and preserves the reported status and available source timestamp. Consumers retain the quality information and avoid presenting the previous good reading as a new valid measurement. If the source timestamp is unavailable, the payload uses null for it. [5]

For a practical acceptance test, interrupt the gateway-to-broker link for 60 seconds while the gateway continues collecting data. If its configured offline buffer covers that interval, replay should preserve the original event IDs and source timestamps. Compare the resulting records with the buffered source set, then repeat with the historian consumer stopped while the broker remains available. Evaluate each consumer's recovery separately; a successful broker delivery alone cannot demonstrate complete history.

Clear next steps you can take with CENTO

Start with one industrial asset group and the applications that need its data. CENTO lists OPC UA and MQTT among its supported acquisition protocols, which provides a basis for an initial integration review. Bring the device interface details, sample values, and required update behavior to that discussion. Confirm the capabilities of the proposed configuration before committing to a wider rollout. [10]

Use the pilot to establish a baseline for data availability and consistency. CENTO's Historian describes acquisition and buffering functions; verify those functions with your sources and expected interruptions. Compare original timestamps, received values, and stored records. Agree on what an acceptable gap means for each measurement and which failures must be reported to the responsible engineering team. [11]

Select one operational question that the dataset must answer, such as how compressor power changes across operating states. For a CENTO project, map the required signals and state information before building the report. Use the OPC UA integration guide to prepare the interface discussion. Treat any proposed reduction in energy use as a hypothesis to investigate with operating data, rather than an outcome established by connectivity.

Approve expansion after the pilot meets its acceptance criteria. In the CENTO integration scope, document supported interfaces, ownership of mappings, access permissions, recovery behavior, and the work required to add another asset. Prioritize changes with a clear operational purpose, and retain the baseline when assessing their results. Repeat the same checks as the installation grows so that the original acceptance criteria remain meaningful.

Explore the demo environment or request a guided walkthrough using your source list and integration requirements.