Solutions · Cloud & warehouse
Feed the data platform without gaps or bespoke pipelines.
Cloud analytics projects usually stall on ingestion: brittle exports, missing periods after every outage, and no agreement on what a field means. Proxus makes the feed a configured target rather than a project.
- Targets
- 18 destination systems, among them Azure Data Lake Storage Gen2, ClickHouse, Kafka, AWS IoT Core and Azure IoT Hub
- Reliability
- Buffered at the edge, forwarded on reconnect
- Contract
- A stable, documented namespace instead of ad-hoc export schemas
A configured target, not an ingestion project.
The cloud or data platform is one more destination: modelled data waits at the gateway through an outage and is written in order when the link returns.
The shift
What actually changes.
How it works today
- An export fails at 02:00 and leaves a hole nobody notices
- Column names are inherited from a PLC address list
- The pipeline was written by someone who has since left
- Volume is whatever the export happened to include
With Proxus
- Buffered delivery replays what the link could not carry
- The namespace is the schema contract, with changes recorded in the audit trail
- Targets are configuration entries your team can read and change
- On-change delivery and filtering at the source size the feed deliberately
How it plays out
A typical sequence, step by step.
An illustrative sequence, not a customer story: written to show the shape of the work rather than to claim a result. Your equipment, sites and systems decide the detail.
-
Design
Agree the contract
The data team reads the namespace model and designs against it, before a single row is shipped.
-
Go-live
The feed starts
Selected signals are delivered to the warehouse or cloud service, with the strategy chosen per target.
-
Outage
The link drops
The gateway queues the feed locally, within its retention, and the historian on your hub keeps serving the plant.
-
Recovery
Backfill without a ticket
Buffered data is forwarded when the link returns, and the target’s status and last success show the feed has caught up.
What Proxus does here
The working parts behind it.
-
Store and forward
Local buffering with replay, so link failures delay data rather than destroy it.
-
A stable contract
Modelled signals with units and context, with changes recorded in the audit trail.
-
Cloud and data platform targets
AWS IoT Core, Azure IoT Hub, Azure Data Lake Storage Gen2, ClickHouse and REST endpoints, initiated from inside your network.
All 18 destination systems -
Streaming targets
Kafka, RabbitMQ, MQTT and Sparkplug B for event-driven consumers.
-
Send only what matters
On-change delivery and filtering at the source to keep bandwidth and storage proportionate.
-
Verifiable delivery
Status, last success, last failure and last error per target, so the data team can see the feed is intact.
Deployment scope
What a first deployment involves.
- You choose
- Which signals leave the site, per target
- You deliver to
- Azure Data Lake Storage Gen2, ClickHouse, Kafka, the AWS and Azure IoT services, and the databases your BI tools read
- You verify
- Delivery status per target, from inside the plant
- You retain
- A copy in your own historian, so plant use never depends on the cloud link
What this enables
- Close the gaps that every outage used to leave in the record
- Give the data team a schema contract rather than an export format
- Keep bandwidth and storage proportionate to the question being asked
- Feed more than one cloud or warehouse from the same collection layer
Proven in production
Feeding data platforms from the plant floor outward.
-
Manufacturing
Digital factory data management
MEXT Digital Factory
Read the case study -
Mining & energy
Mining, fleet and solar on one platform
Kaltun Mining
Ask us about this deployment -
Energy
Grid monitoring at the edge
CK Enerji
Ask us about this deployment -
Manufacturing
Unified Namespace with Oracle ERP
Pilenpak
Ask us about this deployment
Questions
What teams ask before starting.
- How do you avoid gaps in the cloud feed?
- Buffering at the edge with forward-on-reconnect. If the link or the target is unavailable, data queues locally and is delivered when it comes back.
- Does this mean sending everything to the cloud?
- No, and it usually should not. On-change delivery and filtering at the source let you send what the analytics actually needs, which keeps both bandwidth and storage proportionate.
- Who defines the schema the warehouse receives?
- You do, through the namespace model. That model is the contract, and because changes to it are recorded in the audit trail, a change is visible rather than a surprise for the data team.
- Can we feed more than one cloud or warehouse?
- Yes. Each target is configured independently, with its own strategy and format.
Next step
Start with one line, not a programme.
Pick the line that causes the most arguments about numbers. We will scope what it takes to make that one trustworthy, and what the second and tenth would cost in effort.
- 01
Scoping call
Equipment, protocols, sites and the systems the data has to reach.
- 02
Technical walkthrough
A working demo against a scenario that resembles your plant, not a generic tour.
- 03
Pilot plan
A defined first line or site, with the deployment and security questions answered up front.
- 04
Roll-out
The same templates bring the next lines and sites online.
Tell us about your plant
A Proxus engineer replies within one working day.
Rather look around first? Try the live demo