The full power of TDengine, now free forever for up to 5,000 tags.

Explore More

Industrial Data Integration Architecture: Connecting PLCs, SCADA, and Historians

Juno Qiu

November 9, 2026 /

A plant can have a functioning SCADA system, a trusted historian, and years of equipment data, yet still struggle to give another application the information it needs. The problem is not always a missing connection. It may be that current values and historical records live in different systems, that a new consumer would place too much load on a production server, or that the same measurement has different names across sites.

An industrial data integration architecture defines how authorized systems access and use that data. It also establishes where records remain, where copies are created, and who is responsible for maintaining each path.

For an OT architect, the starting point is a practical choice: should a consumer read selected data from an existing source, should selected data be ingested into another store, or should the design combine both approaches? The right answer can differ by data source and use case within the same plant.

Start with the job and the source of record

Before choosing a gateway or integration platform, identify the decision the data needs to support. A maintenance team reviewing equipment behavior over the last six months has different requirements from an application displaying current operating conditions. The first may need reliable historical access; the second may care more about how quickly changes become available.

For each intended use, establish:

  • The consumer and its task. Who needs the data, and what will they do with it?
  • The required time range and resolution. Does the task need current values, raw historical records, summaries, or a combination?
  • The source of record. Which system owns each measurement or context field, and which team can authorize access?
  • The operating constraints. What source load, network paths, access permissions, and service interruptions are acceptable?

A PLC or controller may provide current process values. SCADA may supply operator-facing information and, depending on its configuration, historical records. An existing historian may hold the longer-term process history used by engineers. These are common roles, not guarantees about every installation. Confirm what each system actually stores and exposes before drawing the data path.

Also check whether the same measurement already appears in more than one system. A downstream application that combines both copies without a source rule may count the same observation twice or compare values with different sampling and time semantics.

Compare the architecture choices

This article compares two approaches to accessing and storing selected industrial data: reading it from an existing source through an authorized interface, and ingesting it into another store. These are not an exhaustive classification of industrial integration architectures. A hybrid design combines the two approaches for different data or consumers, so they are not mutually exclusive.

DecisionSource-side accessSelective ingestion or replicationHybrid: combine the paths
Historical accessThe consumer depends on the source interface and the history it exposes.The destination serves the history actually copied; verify backfill separately from ongoing collection.Keep some history at the source and copy a defined subset; specify which system answers each query.
Source-system loadConsumer queries reach the source or its access service; agree limits with its owner.Collection and backfill load the source; subsequent queries against the copy can reduce direct query demand.Source queries and collection both need limits; the combined load must be assessed.
Network boundaryEach authorized access path must cross the required zones, or terminate at an approved intermediary.Collection needs an approved path to the destination; consumers use the destination’s authorized path.Both routes require review; a shared store does not make remaining source access disappear.
Interruption and recoverySource or network downtime can interrupt access; the source owner and consumer must agree how to handle it.The integration owner must establish buffering, gap detection and replay where supported; the destination owner validates the copy.Assign those responsibilities for the copied subset and separately define what happens when source-side access fails.
Operational fitUseful when the source can serve the required queries and consumers accept its availability and interface constraints.Useful when a defined subset needs independent query access, provided the cost and accuracy of maintaining it are acceptable.Useful when data sets have different access needs and the team can operate both paths.

The choice is not simply between “centralized” and “distributed.” A design can preserve local systems while making selected information available elsewhere. What matters is whether each path meets the requirements of its consumers without compromising the systems that produce the data.

One plant, three ways to serve an application

Consider a plant that wants a maintenance application to review equipment history. A PLC controls a machine, an existing SCADA system presents operating information, and a site historian stores selected measurements. The plant team confirms which system owns the measurements and which authorized interfaces expose them. The example is illustrative; it is not a prescribed topology.

The three diagrams in the following sections focus on how a maintenance application uses data already stored in an existing historian. They do not show how measurements originally reach the historian from PLCs or SCADA. If the application also needs information held only in SCADA, an additional authorized access or integration path must be assessed separately.

The collection component that originally supplies the historian may be part of SCADA, the historian, or another approved system. A historian does not necessarily receive its data through SCADA. Each path requires site authorization and a review of control-system load and network zones; the diagrams do not imply unrestricted access to a PLC.

For a separate, vendor-specific example of how an industrial integration hub can connect plant systems and downstream services, see AWS Guidance for Industrial Data Fabric with Ignition. Its components and bidirectional capabilities belong to that specific solution; they do not establish the capabilities of every integration platform.

Read selected data from the source

In a source-side access pattern, an authorized consumer retrieves data through an interface supported by the existing system. For example, a maintenance application might request a limited range of historical measurements from a site historian when an engineer opens an investigation.

Diagram showing a maintenance application accessing an existing historian through an approved interface without creating a separate data copy.
Mobile diagram showing a maintenance application accessing an existing historian through an approved interface without creating a separate data copy.

Source-side access allows an application to use existing historian data through an approved interface.

The historian team sets query and access limits; the application team handles an unavailable source or connection.

This approach can be useful when the source remains the trusted owner of the history and only a small number of consumers need access. It can also avoid maintaining another copy of the selected records.

The design still needs limits. Check whether the interface supports the required historical range, resolution, and data-quality information. Agree on query frequency and expected load with the system owner. If the source or network is unavailable, determine whether the consumer can tolerate the interruption. Access controls and network boundaries still apply even when data stays in its original system.

Ingest a selected subset into another store

In this pattern, an integration component reads or receives approved data and writes it to a separate destination. The destination can then serve a defined set of applications without every consumer querying the original system directly.

Diagram showing selected historian data copied through a collector to a shared store for a maintenance application.
Mobile diagram showing selected historian data copied through a collector to a shared store for a maintenance application.

Selective ingestion copies an approved subset of historian data into a shared store for downstream use.

The historian remains the source of record. The integration team owns collection and gap monitoring, and the destination owner checks that the copy meets agreed history and freshness requirements. Historical backfill and ongoing collection must each be verified.

A separate copy can make sense when several users need the same selected history, when cross-site analysis requires a shared data model, or when the source interface cannot accommodate the intended query workload. But copying values creates a new operating responsibility: someone must know what was copied, what the copy represents, and how discrepancies are detected.

Historical backfill and ongoing ingestion are separate requirements. A connector that receives new values does not necessarily retrieve older records, propagate later source corrections, or reproduce deletions and quality states. Verify each behavior against the specific source, connector, version, and configuration. If the design also needs events, calculations, or asset relationships, check those independently rather than assuming they accompany time-series values.

Combine the two paths

A hybrid design uses source-side access for some information and a separate store for another subset. For instance, a plant might retain detailed process history in its site historian while sending selected measurements to a shared store for comparison across facilities. An authorized application could still retrieve other historical records from the original source when needed.

Diagram showing the same historian serving a maintenance application through approved source-side access and a shared copy of selected data.
Mobile diagram showing the same historian serving a maintenance application through approved source-side access and a shared copy of selected data.

A hybrid architecture combines authorized access to existing historian data with selective ingestion into a shared store.

The team documents which data remains at the source, which subset is copied, and who handles failures on each route.

This is a combination of the first two approaches, not a third, mutually exclusive technology. It can avoid moving data that has no shared use, while giving common applications a consistent place to query selected information.

The trade-off is coordination. The team must tell users which source answers which question, how recently each copy was updated, and what happens if one path is unavailable. If a calculation uses both source-side and copied data, its owner must understand the different update and recovery behavior of each path.

Decide what happens to the existing historian

An existing historian may support production investigations, engineering reports, calculations, or other applications that depend on its history and configuration. A new integration requirement does not by itself mean that the historian should be replaced.

Start by separating three decisions:

  1. Retain the historian as a source. Keep existing workflows in place and provide authorized access to the information another system needs.
  2. Copy selected data while retaining the source. Define the subset, time range, update behavior, and validation process for the new destination.
  3. Evaluate a migration or replacement. Inventory dependent applications, history, asset models, calculations, events, permissions, and operating procedures before deciding what must move or be rebuilt.

These choices can be phased. A team may first make a limited dataset available to a new application and later decide whether a wider change is justified. A successful transfer of values alone does not establish that the original historian’s dependent workflows have been reproduced.

Apply the constraints that change the design

Once the team has identified plausible paths, compare them against the conditions that will affect daily operation.

Time, history, and data quality

A record’s source timestamp may describe when a measurement was taken; a receiving system may also record when it arrived. During an interruption or backlog, those times can differ. Agree which time the consumer needs and how delayed records will be handled.

Check the history available from each source, including its sampling or compression behavior. Determine whether quality information is accessible and how the target represents it. If a source value is uncertain or missing, a downstream trend should not silently present it as a trustworthy measurement. For an example of how quality information can affect historical interpretation, see OPC UA data quality codes for historical analysis.

Where records are copied, agree how to identify duplicates and how to handle late arrivals, corrections, or deletions. The permitted behavior will depend on the source interface and destination; it should be tested rather than assumed.

Asset identity and context

A pressure reading is useful only if the reader can determine what it measures and which equipment produced it. Across systems, the same asset may have different identifiers, names, or units. An integration design needs a mapping that the relevant asset owner can review and maintain.

Preserve a reference to the original source alongside any standardized name. Define who approves changes when equipment is renamed, replaced, or moved. Units, measurement meaning, and operating context may matter as much as the value itself. Industrial data contextualization provides further background on associating measurements with assets and process context.

Availability, security, and ownership

For each path, identify where data can be held during an interruption and who owns recovery. A buffer can protect only the part of the path it covers; it cannot recover measurements that were never collected or persisted. Establish what happens when capacity is exhausted and how operators will detect an incomplete recovery.

Review the intended network path, credentials, and permissions with the OT and security teams. Limit access to what the consumer needs and check how the design behaves during maintenance or failure. NIST’s Guide to Operational Technology Security provides broader guidance for balancing security with OT performance, reliability, and safety requirements.

Ownership should extend beyond the first connection. Assign responsibility for monitoring integration tasks, reviewing failed writes, maintaining mappings, managing credentials, and investigating discrepancies. Without that ownership, a design that works during commissioning can become difficult to trust after the first equipment or configuration change.

Validate a bounded pilot before scaling

Return to the illustrative plant above. Suppose the maintenance application needs both selected equipment history from the site historian and operating information held in the existing SCADA system. The OT team must confirm that each source exposes the required information through an approved interface, then decide which data to access at its source and whether to use a shared copy for a selected subset. This is an illustrative scenario, not a customer deployment.

Choose a small equipment scope and agree on acceptance criteria with the source owners before connecting anything. Then compare the proposed paths against the actual task:

  1. Check the source and access path. Confirm the approved interface, permissions, expected query or collection load, and the time range each source can provide.
  2. Trace known records. Select a few records with known timestamps, values, units, and asset identities. Compare what the consumer receives with the source records.
  3. Test the required history. If using a copy, check the intended backfill and new data separately. Test corrections or deletions only if the use case requires them and the source and integration support them.
  4. Exercise a realistic interruption. With the site owner’s approval, test a defined failure boundary and confirm what data is delayed, recovered, duplicated, or unavailable. Record any gap that cannot be recovered.
  5. Review the operating result. Measure observed latency and source load against the team’s agreed limits. Assign owners for monitoring, mapping changes, access reviews, and recovery tests.

A green connection indicator is a useful starting point, but the acceptance decision should rest on evidence that the selected data is usable for the intended task. If the team cannot explain where a record came from, what its timestamp means, or how an interruption affects it, expanding the architecture to more sources will not resolve those questions.

Related Articles