Choosing a Data Source
All 31 connectors, what each is for, and how to decide when a system can be reached more than one way.
Infohub ships 31 data source connectors. Most of the time the choice is obvious — you have a Canary historian, so you use the Canary connector. This article covers the cases where it is not obvious, which is more often than you would expect.
The full list#
Process historians
| Connector | Notes |
|---|---|
| Wonderware Historian | AVEVA Historian — the most widely deployed |
| Wonderware Historian 2017 | For the 2017-era client libraries |
| Wonderware Online | The cloud-hosted service |
| Wonderware InTouch | InTouch-specific access |
| Wonderware MES | MES data rather than time series — the richest property set of any connector |
| Canary Historian | Legacy Canary versions |
| Canary Historian V24 | Canary V24 and later |
| Ignition Historian | Inductive Automation Ignition |
| OPC HDA Historian | Any OPC Historical Data Access server |
| OPC UA Historian | OPC UA historical access |
| Siemens Process Historian | |
| Timebase Historian | Flow Software's own time series database |
Live and streaming
| Connector | Notes |
|---|---|
| OPC UA | Live OPC UA values |
| MQTT | MQTT broker subscription |
| Apache Kafka | Kafka topic consumption |
| Telegraf | Telegraf metrics |
Databases
| Connector | Notes |
|---|---|
| Microsoft SQL Database | |
| Oracle Database | |
| PostgreSQL Database | |
| MySQL Database | |
| Snowflake | Cloud data warehouse |
| InfluxDB | Time series database |
| AWS Timestream | AWS time series service |
Cloud platforms and devices
| Connector | Notes |
|---|---|
| Siemens Mindsphere | |
| Netilion | Endress+Hauser Netilion |
| Metering Online | |
| eWon | eWon remote access devices |
| Thin Manager |
Files and services
| Connector | Notes |
|---|---|
| Web Service | Any REST or HTTP API — the general-purpose connector |
| Microsoft Excel | Spreadsheet files |
| Flow Historian Simulator | Built-in simulated data, no external system required |
When a system can be reached more than one way#
This is where real judgement is needed.
Native connector, or its SQL database?#
Most historians have both a dedicated connector and an underlying SQL database you could query directly.
Use the native connector. It understands the product's aggregation semantics, retrieval modes, and quality handling, and it survives schema changes between historian versions. Querying the historian's tables directly means reimplementing that understanding yourself and re-testing it after every historian upgrade.
Go direct to SQL only when you need something genuinely outside the connector's reach — a table of production orders, a custom application's data alongside the tags.
Live, or historical?#
OPC UA reads current values; OPC UA Historian reads stored history.
Historical access is almost always what you want. Infohub calculates over periods, and a live-only source cannot answer "what was the average last Tuesday" — it can only tell you what the value is now. If Infohub is your only historian for a tag, a live source means you begin accumulating history from the day you deploy and have none before it.
Use live sources for values that genuinely have no history worth keeping, or when no historian exists and Infohub is becoming the record.
Which Wonderware connector?#
Four connectors touch the AVEVA/Wonderware family, and they are not interchangeable:
- Wonderware Historian — the standard choice for on-premise historian
- Wonderware Historian 2017 — when the 2017-era client libraries are what is installed
- Wonderware Online — the cloud service, a different product
- Wonderware InTouch / Wonderware MES — InTouch and MES data specifically
The historian connectors depend on client libraries installed on the machine hosting the connection, and version mismatches between the client library and the historian server are a recognised source of connection failures.
Which Canary connector?#
Canary Historian V24 for V24 and later; Canary Historian for earlier versions. The APIs differ enough to warrant separate connectors. If you are unsure which you have, check the server version before configuring, because the wrong connector fails in ways that look like credentials problems.
MQTT: data source or consumer?#
Both exist, and they point in opposite directions. The MQTT data source subscribes — data flows into Infohub. The MQTT Broker consumer publishes — data flows out. If you are trying to get data into Infohub, you want the data source.
Anything with an HTTP API#
The Web Service connector handles any REST or HTTP API, with a rich authentication surface including OAuth 2.0. It is the fallback for systems with no dedicated connector: ERP endpoints, weather services, grid carbon intensity, laboratory systems, cloud platforms.
See Web Service Data Sources and OAuth 2.0 for Web Service Sources.
Deciding by question#
| You have… | Use |
|---|---|
| A process historian with a named connector | That connector |
| A historian with no named connector, but OPC HDA support | OPC HDA Historian |
| An OPC UA server with historical access | OPC UA Historian |
| Only live OPC UA | OPC UA — and accept that history starts now |
| Production or quality data in a relational database | The matching database connector |
| A REST API | Web Service |
| An MQTT broker publishing plant data | MQTT |
| Spreadsheets maintained by a person | Microsoft Excel — but consider manual entry instead |
| No source yet, and you want to learn or demonstrate | Flow Historian Simulator |
Spreadsheets deserve a second thought#
The Excel connector works, and reaching for it is often a sign that something else is the better answer.
A spreadsheet maintained by a person, in a folder, read on a schedule, will eventually be renamed, moved, opened and locked at the wrong moment, or have a column inserted. Each of those breaks retrieval silently.
Where the data originates with a person, manual entry measures with a form are more robust: the values are in the database, they are validated on entry, they are auditable, and no file can go missing. Use the Excel connector when the spreadsheet is a genuine system output you do not control.
Start with the simulator#
The Flow Historian Simulator ships with Infohub and needs no external system. It is the right tool for:
- Learning Infohub, and working through the tutorials
- Building demonstration models
- Testing modelling patterns before real data exists
- Proving the plugin framework works on a new installation
Before you commit#
For any connector new to you, prove it end to end on one measure before modelling against it:
- Create the connection and test it.
- Browse the namespace.
- Create one retrieved measure at the Tag interval.
- Deploy and confirm values arrive.
- Compare a handful of values against the source system directly.
- Confirm it still works as the service account, not just as you.
Step 5 catches unit and scaling surprises. Step 6 catches the integrated-security difference that otherwise appears only after deployment.