EMQX Edge with store-and-forward sits on the factory floor
and buffers all MQTT messages to disk when the upstream connection is lost. When connectivity
returns, buffered messages are automatically forwarded in order. No data loss, no operator action,
no gaps in the time-series.
The full pipeline: EMQX Neuron reads OPC UA tags from
the Enterprise B virtual factory (3 sites, ~3,340 topics) and publishes them as MQTT.
EMQX Edge bridges to
EMQX Cloud, where the rule engine parses payloads and stores
them in EMQX Tables (built-in time-series storage). Grafana
queries the tables to render OEE, production counts, and
process data in real time.
The edge layer (EMQX Neuron + EMQX Edge) runs on-site; the cloud layer (EMQX Cloud + Tables) is a
managed service. Grafana connects to the cloud layer over the WAN.
Additionally, MCP (Model Context Protocol) servers expose both the time-series data
and the broker API to AI agents, enabling natural-language queries against live manufacturing data.
How long to implement
The edge-to-cloud pipeline was configured and running in under a
week, including OPC UA tag mapping, MQTT bridging, rule engine configuration,
time-series storage, and Grafana dashboards. No custom code; everything is configuration on the
EMQX platform.
What it would cost
EMQX Neuron licensing is tag-based, EMQX and EMQX
Cloud
are priced by connections message throughput. For a single-site deployment like this demo,
typical costs start in the low thousands per year. Multi-site scales linearly with one edge node
per site.
