Historian at the Edge and the Enterprise
Line teams need process history right where they work; the enterprise needs one consolidated record across every site. KŌJŌ Stack runs a historian as an edge workload with bounded, recent-window storage for SCADA and line engineers-then a second KŌJŌ Stack instance subscribes to each site's MQTT or Kafka broker and forwards to a centralized historian on-premises or in the cloud. The same platform serves both tiers, so history is trustworthy end to end.
Architecture Highlights
One Historian, Two Tiers
The Problem
The Edge Needs History It Can Actually Keep
A line engineer needs recent trends at the panel, but edge hardware has limited storage. Full-resolution history for every tag simply does not fit, so teams either keep too little or ship everything upstream and lose fast local access.
The Enterprise Needs One Record Across Many Sites
Central analytics and reporting depend on a consistent, long-term history spanning every plant. When each site keeps its own disconnected store in its own format, enterprise-wide questions require stitching together incompatible records after the fact.
Moving Everything Upstream Is Expensive and Noisy
Forwarding raw, full-rate history from every tag saturates constrained site links and floods the central historian with values that never meaningfully changed. Cost and noise grow with every tag added.
What Fails in Traditional Architectures
Without structured, prepared data at the first mile, downstream systems inherit every inconsistency, gap, and limitation of the raw source data.
Local History Falls Off the Edge
Without bounded, recent-window retention tuned for edge hardware, either storage fills and history is lost, or so little is kept that the panel cannot show a meaningful trend. Line engineers lose the fast local context they rely on.
Central History Is Stitched Together After the Fact
When each site keeps its own disconnected store, enterprise reporting must reconcile incompatible records per plant. Cross-site questions become manual reconstruction rather than a single query against one consistent history.
The Central Historian Drowns in Noise
Forwarding raw, full-rate data from every tag inflates cost and buries meaningful changes under values that never moved. Without edge reduction, scaling to more tags and more sites gets steadily more expensive.
How KŌJŌ Stack Helps
Historian as an Edge Workload
Run a time-series historian directly on the line as a KŌJŌ Stack workload, sized for bounded, recent-window storage. SCADA screens and line engineers query local history on-site, without a round trip to the enterprise.
Reduce Volume Before It Leaves the Site
Report-by-exception captures every meaningful state change while dropping noise, and trend compression stores far less while preserving trend shape within a bounded, configurable accuracy envelope. Constrained links carry signal, not raw volume.
Site-to-Enterprise Forwarding via Brokers
A second KŌJŌ Stack instance subscribes to each site's MQTT or Kafka broker and forwards to a centralized historian-on-premises or in the cloud. Durable buffering and ordered replay maintain data continuity across network outages.
Trustworthy Trends on Both Tiers
Averages and statistics stay accurate even when tags report irregularly or only on change, and each consumer retrieves history in the shape it needs-interpolated, stepped, or cyclic. Edge dashboards and enterprise analytics read from the same dependable history.
Why This Requires First-Mile Data Structuring
The pattern separates two historian responsibilities that are usually forced onto one system. At the edge, a KŌJŌ Stack workload runs a historian sized for the line: it acquires data natively from PLCs, sensors, and SCADA, applies report-by-exception and trend compression to keep a bounded, recent window of history on modest storage, and serves that history locally to operators and engineers. The compressed, reduced stream is published to the site's MQTT or Kafka broker. At the enterprise tier, a second KŌJŌ Stack instance subscribes to those brokers across many sites and forwards into a centralized historian-on-premises or in the cloud-for long-term, cross-site history. Because the same platform governs both tiers, averages and statistics stay accurate on irregular and store-on-change data, each consumer retrieves history in the shape it needs, and durable buffering with ordered replay keeps the two tiers consistent even when the link between them drops. The result is fast local history for the people on the line and one trustworthy record for the enterprise, without paying to move raw full-rate data upstream.
Architecture
Expected Outcomes
Local history for the line, consolidated history for the enterprise
Recent-window retention sized for edge hardware
Buffering and ordered replay maintain continuity across outages
Own the First Mile
Owning the first mile ensures historian at the edge and the enterprise data is consistent, contextualized, and usable across the enterprise.