What is DataHub?
DataHub is a place to keep your operational data together with a description of what that data is about.
Without that description, a measurement is just a number in a list. With it, the same measurement knows which pump it came from, which process that pump serves, which site the process runs on, and which business target the site is accountable for.
The one-sentence version
DataHub is a data platform that stores your measurements and your events next to a model of your operation, so that any number can be traced to the physical thing it describes and the business decision it affects.
An analogy: the warehouse and the library
Imagine two buildings.
The first is a warehouse. Millions of items, all in unlabelled boxes, stacked by whoever delivered them. Everything you own is in there. Finding anything specific requires someone who remembers where it was put, and when that person leaves, the knowledge leaves too.
The second is a library. Same volume of material, but every item is catalogued: what it is, what subject it belongs to, which other works it relates to. You don't need to know where anything physically sits. You describe what you want, and the catalogue takes you there.
Most organisations run their operational data like a warehouse. DataHub is the catalogue. The data does not change; what changes is that the data now carries a description of itself, the work called contextualization, and that description is written down somewhere everyone can use rather than living in the heads of a handful of long-serving engineers.
What that looks like in practice
Suppose a plant engineer asks a simple question:
"Why did our energy consumption jump last Tuesday?"
Without a model of the operation, answering this means: find which meters cover the site, in whichever historian holds them; work out which production lines were running, from the MES or a shift log; check whether any maintenance happened, in the CMMS; check whether a set-point changed, in the control system; then reconcile four exports in a spreadsheet. Days of work, and the answer is only as good as the person assembling it.
With a model of the operation, the same question is a traversal. Start at the site, walk down to the lines it contains, to the equipment on those lines, to the meters attached to that equipment; overlay the events raised against any of them in the same window. The answer arrives in minutes, and it is reproducible, the next person asking gets the same answer.
The difference is not the data. Both organisations had the same data. The difference is whether the relationships between things were written down in a form a computer can follow.
What DataHub actually stores
Four things today, and only four. This is deliberate, a small vocabulary is what makes it possible for people who are not data engineers to use the platform. Two more blocks, policies and functions, are on the roadmap.
| Building block | What it is | Example |
|---|---|---|
| Resources | The things in your operation, and how they relate | A pump, a tank, a production line, a site, a KPI |
| Time series | Continuous measurements, arriving over and over | Pressure, temperature, throughput, power draw |
| Events | Things that happened, at a point in time | An alarm fired, a permit closed, a work order completed |
| Data sets | Containers that group related data and control who sees it | "Valve pressure sensors", "Grid asset monitoring" |
Whatever your source systems call things, they arrive in DataHub as one of these four. A work permit being opened, a purchase order being raised, and a sensor alarm are all events. A vibration signal and a financial index are both time series. That reduction is what lets one person query across systems that were never designed to talk to each other.
Kept in step with live data, those four together are what people mean by a digital twin.
Read more about the four building blocks →
What DataHub is not
Being clear about the boundaries saves disappointment later.
- It is not a replacement for your control system. DataHub reads from operational systems; it does not run your plant or hold safety functions.
- It is not a dashboard product. It has charting and exploration built in, but its job is to be the trustworthy layer underneath whatever reporting you already use.
- It is not a data lake. A lake stores files and hopes someone can make sense of them later. DataHub insists that data arrives connected to a model.
- It is not magic. The model has to be built. The platform makes that cheap and keeps it maintainable, but somebody who understands your operation still has to say what the things are. How to build the model →
Who uses it, and for what
| Who | What they do with it |
|---|---|
| Operations and engineering | Investigate incidents by following relationships; monitor live signals; see what a change would affect |
| Subject-matter experts | Curate the model, the description of what the operation is stays with the people who understand it |
| Analysts and data scientists | Query data that already carries its context, and publish results back into the same model |
| Leadership | Get reported figures that can be traced end to end, rather than assembled and asserted |
| Administrators | Run the platform, manage access, control retention and cost |
Why it is open source
An operational data platform sits underneath your reporting, your investigations and eventually your automation. Adopting one is a decade-scale decision, so the honest question is: what happens if the supplier disappears, triples the price, or discontinues the thing you depend on?
Closed platforms answer that with a contract. DataHub answers it structurally: the source is public, you can audit exactly what it does with your data, run it entirely disconnected from the internet, and continue running it yourself if we ever vanish. The licence is the GNU AGPL-3.0, which places no restriction on using DataHub for your own organisation's data.
- The operational data problem: why this keeps happening, and what Industry 4.0 actually requires
- What is a digital twin?: the same idea, under the name your board already uses
- The four building blocks: resources, time series, events and data sets in detail
- What is an ontology?: the model layer, explained without jargon
- Your first hour: build a small working model yourself
- The business case: where the return comes from