Products · Factory Data Backbone
One namespace for the whole factory
Every machine is read once and published into one named tree on a licensed, clustered MQTT broker. Every screen, report and AI assistant subscribes to the tree instead of the machine. You own the broker, the tree and the data.
Licensed broker. Clustered. Yours to keep.
The problem
Today every screen reads every PLC.
Each new dashboard opens another connection to the same controller. Two screens disagree because they each work out the same thing. Nobody can tell a live number from one that stopped updating at lunch.
The Backbone reads each machine once and publishes the answer.
How it fits
Machines in, one tree, everything subscribes
Read-only adapters take what each controller already speaks and publish it into a clustered broker. From then on nothing on the office side ever talks to a machine; it talks to the tree.
One named tree
- Site
- Packaging
- Line 3
- Filler
- Rate
- State
- Status
- Capper
- Line 4
Site / Packaging / Line 3 / Filler / Rate
Value
118 /min
Quality
Good
Updated
2 s ago
Held for
Late joiners
Every machine, line and value has one permanent address. Add a screen and it subscribes to the address; nobody re-reads the PLC.
Design rules
Seven things that stay true
These are the rules the whole platform is built to, and they are written into the acceptance tests at every stage.
Read once, publish once
Each machine is read by exactly one adapter. Every screen, report and assistant subscribes to the result instead of opening its own connection to the controller.
State decided once
Running, slow, stopped, planned or no data is decided in one place and published for everyone. Two screens can never disagree about the same line.
Live now, history kept
The current value of everything is available the moment you connect. Every change is also stored, so this morning and last year are one query apart.
One-way by design
Data flows from the plant to the broker and never back. No dashboard, browser or report can move a machine, which is also what gets it approved.
Named once, versioned forever
Every asset gets one permanent address. The way data is described is versioned, so adding a screen later never means rebuilding what is already there.
One database
Trends, stops, shifts, products and reasons live in the same database, so OEE is a query rather than a spreadsheet exercise.
Separated zones
The machine network, the data hosts and the office are kept apart. The floor cannot be reached from a browser, and an office problem cannot reach the floor.
The broker
Licensed and clustered, on purpose
Free broker tiers are single-node and not licensed for a platform delivered to you. The Backbone runs on a commercially licensed cluster: three nodes, so losing one changes nothing.
Default
EMQX Enterprise
The default choice. Clustered, encrypted, with the durability and monitoring a plant needs. We size and install it; the licence is in your name.
Alternative
HiveMQ Platform
Where your group already runs HiveMQ, the Backbone runs on it one-for-one. Same tree, same rules, same screens.
Yours
No lock-in to us
The broker, the database and the tree are standard, documented and held by you. If we disappeared tomorrow, another integrator could pick it up.
Independent system integrator. EMQX and HiveMQ are trademarks of their respective owners; no partnership is implied.
What plugs in
In from the plant, out to whoever needs it
Nothing on the floor is replaced or reprogrammed. Whatever a controller already speaks is what the adapter reads.
In
- Siemens S7-1500 / S7-1200 (OPC UA)
- Older S7 families
- Modbus TCP meters & utilities
- Rockwell EtherNet/IP
- WinCC Unified runtimes
- IO-Link masters
- Devices with an HTTP API
Out
- Plant Dashboard Server →
- Grafana trends & alerts
- AI Maintenance Assistant →
- Predictive maintenance →
- ERP & CMMS
- Power BI & Excel
From counters to OEE
Every line's state, decided once
The Backbone decides whether each line is running, slow, stopped or planned, and publishes it for every screen. Stops get a reason from the floor in under thirty seconds. OEE comes out of the database, per product, and says so plainly when a figure still rests on a placeholder.
Morning shift · packaging
- Running
- Slow
- Stopped
- Planned
- No data
Left alone safely
Secured, monitored, backed up
A platform the whole plant leans on has to be boring. The handover gate is a reboot of every host with nobody logged in, and everything coming back on its own.
Encrypted, per-client identity
Every adapter, screen and assistant connects with its own credentials over an encrypted link and can only see what its role allows.
Machines are read-only
The adapters cannot write. There is no path from a browser to a controller, full stop.
Monitored, with five standing alerts
A source going quiet, a host filling up, a broker node dropping, history stalling or the tree shrinking is raised to you and to us before anyone notices a blank tile.
Backed up nightly, restore tested
Database, broker state and configuration are backed up every night. The restore is proven at handover and once a year, not assumed.
Sizing
Three sizes, one design
Start where the plant is. Every tier runs the same software and the same rules, so growing is an install, not a rebuild.
| Essential | High availabilityDefault | Multi-site | |
|---|---|---|---|
| Broker | One licensed broker node | Three-node broker cluster | A cluster per site, bridged to a central one |
| Database | Single database, nightly backup | Primary and live replica | Per site, with a central roll-up |
| Hardware | One small industrial PC or VM | Three small industrial PCs | Per site plus central |
| Survives | A crash or a reboot | Losing a whole host | Losing a host per site, or the centre |
| Fits | 1 to 10 lines, one department | The whole plant | Groups with several plants |
How it goes in
Six stages, each with a gate
Fixed scope per stage. The next stage is not invoiced until the gate of the last one has passed in front of you.
01
Discover
Every controller on site is listed with what it can give and how it can be reached. You sign off the list, the tier and the naming before anything is installed.
02
Install
Hosts, cluster, database, monitoring and backup go in. The gate: reboot everything with nobody logged in and it all comes back, and a restore from backup works.
03
Contract & state
The way data is described is fixed and versioned, and line state goes live for the first lines. The gate: two independent screens show the same state from the same source.
04
Coverage
Lines are connected in the order that matters to you. The gate per line: its data matches the list, its history is filling, and its tile appears without a code change.
05
Screens
The factory overview, department pages, trend and OEE reports, the reason console and alerts. The gate: floor, office and assistant agree on every line.
06
Business context
Orders and work orders arrive against the right line, so performance is measured per product automatically.
Where it sits
Which one do you need?
Three different jobs. Most plants end up with all three, and they do not compete.
Operator screens · WinCC Unified
Runs the machine
On the panel beside the line. Operators start, stop, set and acknowledge from it.
See operator screens →Factory Data Backbone · this page
Connects the factory
Reads every machine once, publishes one tree on a clustered broker, keeps the history and decides each line's state.
Plant Dashboard Server
Shows the factory
Read-only screens in any browser, fed by the Backbone, or reading devices directly on a small site.
See the dashboard server →Tell us how many lines
Send through how many lines and controllers you have, and roughly what they are. We will come back with the tier that fits, what the first stage covers, and what you would see at the end of it.