Payload Formats
UMH-Core recognises three payload formats. Pick the one that matches your sensor / message before you build a bridge or stream-processor.
Time-series / Tags
One numeric/boolean value + timestamp (sensor readings, counters, states)
Data coming from PLCs or sensors
Bridge + tag_processor
TimescaleDB
Relational / JSON
One self-contained business record (order, recipe, batch header)
Data coming from higher-level systems, such as Orders, alarms, set-points, Batch reports, recipes OR time-series data that belongs together, e.g., that has been merged (see further below)
Bridge + nodered_js
PostgreSQL / REST
Binary Blob
File pointer or binary payload (images, PDFs, CNC files)
For everything else
Bridge + nodered_js
S3 Bucket
Why you do not bundle time-series points into one JSON object
The tempting shortcut (UMH Classic)
In UMH Classic you could publish a whole weather snapshot in one go:
Topic:
Payload:
Where that shortcut explodes
No single, unique identifier
You need three dimensions to access a tag: topic, tag name, and JSON path. Every dashboard, rule engine, historian and schema registry must now learn this complex address space.
Real-world timing issues
What happens when the source gives you a new temperature value, but no humidity value? In OPC UA subscriptions, humidity might arrive 100ms later. Do you cache it? How do you know if you're receiving a cached value or if the temperature really stayed constant for hours?
Unintuitive logic for OT users
As an OT person, you simply want to write:
if …weather.temperature > 30 && …isCloudy == true
Not manage local caches or three-dimensional addressing.
Edge-case combinatorics
Missing key? Late arriving field? Partial PLC failure? Each choice (ignore / cache / null) multiplies test cases and bugs.
Merge window guessing
Two PLC variables that change almost at the same time still arrive as separate OPC UA notifications. You must decide how long to wait before you believe you have a "complete" JSON.
Clock skew & racing updates
Reading 100 tags into one JSON can take milliseconds; by the time the last field is read the first may already have changed—yet a single timestamp suggests they belong together.
Topic-vs-payload bikeshedding
Should a nested struct live under weather.* or inside the payload? Different consumers want different splits → endless disagreements.
Result: every integrator ends up writing brittle glue scripts and the Tag-Browser (which once parsed the payload to build a "full-tag-name") turns into a complexity monster.
UMH-Core's guiding rule
"One tag, one message, one topic."
What changes? – You collapse the three dimensionstopic × json-key × payload-format → one dimension topic.
The address of a datapoint is now self-contained and stable; the payload is always the single scalar that belongs to that address.
Benefits:
Intuitive expressions – both OT & IT write rules against plain topics:
if umh.v1.acme._historian.weather.temperature > 30 …Zero merge code – every value is complete at publish time; no caching.
Simplified business discussions – You only need to discuss the "folder structure" (topic hierarchy) with business stakeholders. No need to explain the difference between tag names, topics, JSON paths, and payload formats.
Schema enforcement moves to the topic – data contract can simply specify time-series payload; no further schema needed.
Unlimited fan-out – a client interested only in humidity subscribes once; it is not spammed by temperature updates.
When you do want a bundled JSON
Quality stations or batch reports often require a final "document" after all samples are taken. Create a Stream-Processor, define the merge window & payload, and register an explicit Data Contract (e.g. BatchReport). All assumptions become version-controlled and auditable.
Advanced Examples
CNC Mill with Multiple Sensor Groups
CNC machines often have multiple sensor groups for different aspects of operation. Here's how to organize related sensors using virtual paths:
Topic Structure:
Payloads:
Virtual Path Benefits:
axis.*groups all positioning sensorsmachine_state.*groups operational statusspindle.*groups spindle-related measurementsEach group can be consumed independently for specialized analytics
Packaging Line with Process Data
Packaging equipment often needs to track product flow, quality metrics, and machine health:
Topic Structure:
Analytics Use Cases:
Production Dashboard: Subscribe to
production.*topicsQuality Monitoring: Subscribe to
quality.*topicsPredictive Maintenance: Subscribe to
diagnostics.*topicsOverall Equipment Effectiveness: Combine all virtual paths
Multi-Sensor Aggregation
Some applications require combining multiple sensors into structured payloads:
Bridge Configuration:
Result Topics:
Last updated