Comparison
Opcenter integration interfaces: OData, OPC UA or MQTT
Short answer
Opcenter Execution supports OData 4.0 for application data and OPC UA for equipment communication; MQTT is not a native interface and comes only from a documented connector or an integration application built on the Foundation stack. Use OData for records, OPC UA for the line, and publish to a broker only from a component that reads one of them.
On this page
- The interfaces Opcenter Execution actually ships
- Matching the requirement to the interface
- OData 4.0: a protocol on REST conventions
- OPC UA: the equipment side
- MQTT: connector or application, never assumed
- The SQL database and the BIL boundary
- Worked example: syncing three years of records at 5,000 a day
- Questions asked
Opcenter Execution — Foundation, Process and Discrete — exposes application data through OData 4.0 and communicates with industrial equipment through OPC UA. MQTT is not a native interface: it arrives through a documented connector for the installed release, or through an integration application you build on the Foundation stack, which is C# on .NET 4.7.2 at the backend with JavaScript at the frontend. The working answer is therefore: OData 4.0 to move manufacturing records, OPC UA to talk to the line, and a broker feed only from a component that reads one of those two. Two claims that travel with this question are wrong: OData is not a superset of REST, and MQTT's low overhead does not cap payload volume or topic count.
The interfaces Opcenter Execution actually ships #
Support at product-family level does not tell you what one installation exposes. Record four things before choosing an endpoint: the exact Opcenter product, the release, the installed components, and the integration configuration. A label such as EXPR is an identification item to resolve against the installed product documentation, not an edition name to design against.
Two more facts set the boundary. Opcenter's data lives in SQL, and in the deployment described in this question BIL is the tool that connects to another external database — a database path, not a messaging interface. Module-to-module integration is its own configured exercise; configuring Opcenter APS and Execution Foundation links is separate from exposing execution data to outside systems.
Matching the requirement to the interface #
| Requirement | Interface | Exchange model | Deciding check |
|---|---|---|---|
| Query or exchange manufacturing application records | OData 4.0 | Request/response over HTTP | Required entities and operations appear in the deployed service contract |
| Communicate with PLCs and other automation | OPC UA (IEC 62541) | Client/server sessions with subscriptions | The configured interface exposes the required equipment data and operations |
| Publish Opcenter events to a broker | MQTT via documented connector or integration app | Broker-mediated publish/subscribe | Installed documentation names the connector, or your app owns the mapping |
| Move existing database records | Documented database path (SQL, BIL) | Bulk file or database transfer | Supported access method, data ownership and synchronization behavior are defined |
| Link planning to execution | Configured module link (APS to Execution) | Vendor-configured integration | The link is configured and documented for the installed releases |
Read the table as the decision it is: application-data access goes to the OData contract check, equipment communication to the OPC UA configuration check, broker publication to the connector inventory, and bulk history to the database path.
OData 4.0: a protocol on REST conventions #
REST is an architectural style for resource-oriented interfaces. OData 4.0 is a protocol specification, standardized at OASIS, that defines service metadata, querying, filtering and paging conventions on top of HTTP in that style. It is not a superset of REST — that phrase confuses a protocol with an architectural style — and the tagline "the best way to REST" is marketing shorthand, not a specification relationship.
A working HTTP connection proves transport only, not data access. Work the contract:
- Read the documented service address and authentication model; expect a configured endpoint and an authorized integration identity.
- Request the service metadata and confirm the entities, properties, relationships and operations you need are in the model.
- Run one bounded query for a record you know — a production order, a lot — and check the identifier and field values that return.
- Exercise each write or business operation separately: read access does not imply write access, and successful HTTP transport does not make a transaction workflow supported.
- Use the documented filtering, field selection and paging, and follow the service's continuation mechanism; the first response is not the dataset.
OPC UA: the equipment side #
OPC UA, standardized as IEC 62541, is the interface Opcenter holds toward industrial automation. Product-level support still leaves the deployed configuration in charge of what is exposed, so the check is concrete: the configured interface must expose the equipment data and operations the integration needs, and when it does not, the fix belongs in configuration, not in guesswork about addresses.
For Opcenter Execution Process, Siemens publishes an OPC UA Information Model Kit for manufacturing use cases (support entry 109814835, document version 3.0) that includes integration guidance and an example TIA Portal project; start from that kit rather than from a blank server configuration.
MQTT: connector or application, never assumed #
MQTT is a broker-mediated publish/subscribe protocol: producers publish messages to named topics and subscribers receive the matches. Nothing in the Opcenter Execution interface list makes it a built-in feature, so two legitimate paths exist. One is a documented connector for the installed release, and only if the documentation for that release names one. The other is an integration application on the Foundation stack — C# on the .NET 4.7.2 backend, JavaScript on the frontend — that acts as a client to OData or OPC UA and publishes to your broker.
For analysis rather than transport, Insights Hub carries the Opcenter Intelligence API service, which runs analysis and KPI calculations over MES and IoT data inside the supported Insights Hub services.
Now the performance claim to discard. Low per-message overhead is a wire property; it caps neither topic count nor the volume a plant can publish. What actually limits a cloud sync is payload volume, the broker's configured ceilings — MQTT broker and topic limits configuration shows where those settings live — link bandwidth, and the rate at which the destination accepts and indexes writes. Swapping HTTP for MQTT fixes none of those.
The bridge itself is a solved pattern: bridging OPC UA PLC data to MQTT on an edge PC covers the same architecture when the source is the line rather than the MES. Host the integration application on a fanless mini PC at the plant edge, not on the MES server, so broker and network faults never load Opcenter.
The SQL database and the BIL boundary #
Treat three capabilities as separate: database connectivity, application integration, and change synchronization. BIL is identified in this deployment as the tool that connects to another external database; confirm in the installed documentation whether it reads, writes, or exchanges defined records before calling it a cloud sync mechanism.
The recurring failure is equating access to SQL with permission to bypass application logic. The OData interface applies authorization and business rules that raw tables do not, and a direct edit to application-owned tables can corrupt the data integrity the MES depends on. Business operations go through the documented interface; the database path stays inside its supported boundary.
Worked example: syncing three years of records at 5,000 a day #
The complaint behind this question is slow cloud sync of a large history. Size it with four illustrative inputs for a typical plant — replace them with your own and the method and the decision rule hold:
- Record rate: 5,000 records per day from the production journal
- Serialized size: 2 KB per record crossing the integration interface
- History to move once: 3 years, about 1,095 days
- Day length: 24 h, or 86,400 s
Rule in words: the ongoing stream is record rate multiplied by record size; the one-time history load is the daily volume multiplied by the days of history; the average streaming rate is the daily volume divided by the seconds in a day.
Arithmetic, written out:
- Daily delta: 5,000 records/day × 2 KB/record = 10,000 KB/day = 10 MB/day.
- History load: 10 MB/day × 1,095 days = 10,950 MB, about 11 GB.
- Average streaming rate: 10 MB/day ÷ 86,400 s/day ≈ 0.12 KB/s.
Result: the steady stream averages about 0.12 KB/s — trivial for any interface — while the initial load is 11 GB, which is 1,095 days of delta delivered as one job.
Decision: split the two. For this plant, any one-time transfer above about 30 MB — three days of delta — moves through the bulk database path inside its supported boundary, with a restart position persisted only after the destination acknowledges each batch. The ongoing 10 MB/day delta streams through whichever interface the consumers need; at that volume an OData poll and an MQTT publisher both clear it with orders of magnitude to spare. Volume does not choose the protocol here — the consumers do: pull through OData when systems query on demand, push to a broker when several systems subscribe to the same events.
Questions asked
Is OData 4.0 really a superset of REST?
No, and the distinction matters when you write the client. REST is an architectural style; OData 4.0 is an OASIS protocol specification that adds service metadata, query, filtering and paging conventions on top of HTTP in that style. Any HTTP client can issue a simple request, but filters, expansions and continuation tokens follow OData's defined conventions, so build against the service contract, not against generic REST assumptions.
Are there reference architectures for Opcenter Execution in the cloud?
Yes. AWS publishes deployment guidance covering four architecture design options for Opcenter Execution Foundation, which shows where the MES and its integration components can sit relative to plant systems. Treat it as a sanity check for placement, with the MES close to the line and integration and analysis services outside it, not as a substitute for your installed release's documentation.
What does the Opcenter Intelligence API service give me?
It is the documented analysis path inside Insights Hub: it runs analysis and KPI calculations over MES and IoT data within the supported Insights Hub services. Use it when the goal is dashboards and KPIs in the cloud; use OData, OPC UA or your own MQTT publisher when the goal is moving the data itself to a system you control.
Will MQTT break under a large Opcenter dataset?
No. Low per-message overhead is a property of the wire format; neither topic count nor total published volume is limited by the specification. The real ceilings are the broker's configured limits, the link bandwidth and the rate at which the destination accepts and indexes writes. Move the initial history as a bulk job and stream only the daily delta, which in the worked example is 11 GB once, then 10 MB per day.
Where should the MQTT integration application run?
Not on the MES server. Give it its own host at the plant edge, such as a fanless box or mini PC in the cabinet, so a broker outage or network fault never loads Opcenter, and so you can restart or patch the bridge without a production window on the MES. The application acts as a client to OData or OPC UA and as a publisher to the broker.