Azure IoT Operations
Azure IoT Operations is a set of edge data services (MQTT broker, Akri connectors such as OPC UA, data flows) on Azure Arc-enabled Kubernetes at a site. It collects and normalizes equipment data locally, works offline for up to 72 hours, and sends results to Microsoft Fabric, Event Hubs or other cloud endpoints.1
Also known as Azure IoT Operations enabled by Azure Arc, AIO
- Intended for
- OT and plant engineers, Industrial IT and edge platform teams, Data engineers building IIoT pipelines, Solution architects for manufacturing and energy sites
- Runs on
- On-premises plants and sites on Arc-enabled Kubernetes, AKS Edge Essentials and AKS on Azure Local (Windows), K3s, RKE2, VMware vSphere Kubernetes Service and Red Hat OpenShift (Linux), K3s on small form factor Azure Local (preview), GitHub Codespaces for evaluation
At a glance
- What is it?
- Azure IoT Operations is Microsoft's unified data plane for the edge. It installs a group of Kubernetes-native services on an Azure Arc-enabled cluster that you run on-premises: an MQTT broker for event-driven messaging, Akri connectors and services that talk to devices and discover assets, and data flows that transform and route messages to the cloud. A web UI called the operations experience lets OT users manage assets and data flows, and assets and devices are registered in Azure Device Registry so they appear as Azure resources. It is part of Microsoft's adaptive cloud approach.1
- What does it do?
- It connects to industrial equipment over open standards such as OPC UA and MQTT (and HTTP/REST through connectors), publishes the data to topics on a local MQTT broker, processes and contextualizes it with data flows, and sends it to cloud endpoints: Azure Event Grid or other MQTT brokers, Azure Event Hubs or Kafka, Azure Data Lake Storage, Microsoft Fabric OneLake and Azure Data Explorer. A schema registry in Azure Device Registry handles message serialization. Built-in secrets management, certificate management and secure settings protect connections, and the runtime keeps operating for up to 72 hours without cloud connectivity.1
- Who needs it?
- Manufacturers, energy companies and other operators that need plant-floor data for anomaly detection, predictive maintenance, energy optimization or overall equipment effectiveness, and want to bridge OT and IT systems without custom point-to-point integrations. It suits organizations that already run or plan to run Kubernetes at sites and want to manage those sites centrally through Azure Arc. Azure IoT Central customers looking at migration paths should note that Microsoft's recommended target for IoT Central is IoT Hub, DPS and Fabric, not IoT Operations.1
- What does it need?2
- An Azure Arc-enabled Kubernetes cluster on a supported distribution: AKS Edge Essentials or AKS on Azure Local (Windows), or K3s, RKE2, VMware vSphere Kubernetes Service or Red Hat OpenShift (Linux)
- At least 16 GB RAM and 4 vCPUs for production (32 GB and 8 vCPUs recommended)
- Azure roles such as Azure IoT Operations Onboarding and Contributor to register resource providers
- An Azure Storage account for the schema registry and Azure Key Vault for secure settings
- Firewall access to Azure Arc, Azure CLI and chosen cloud data endpoints
- What it is not
- IoT Operations is not a device connectivity service for millions of internet-connected devices; that role belongs to Azure IoT Hub and the Device Provisioning Service. It is not a dashboarding tool: visualization happens in Microsoft Fabric or Power BI. It is not a replacement for Azure IoT Central's application experience: Microsoft stopped new IoT Central app creation on 23 September 2026 and recommends IoT Hub, DPS and Fabric as the migration path. It also does not run without Kubernetes: it needs an Arc-enabled cluster on a supported distribution.3
Availability and licensing. In production use; Microsoft lists its supported Kubernetes distributions (AKS Edge Essentials, AKS on Azure Local, K3s, VKS, RKE2, Red Hat OpenShift) at a general availability support level, with K3s on small form factor Azure Local in preview (checked 26 October 2026). The product page refers to pricing by node and asset without listing figures here. Related change: new Azure IoT Central app creation stopped on 23 September 2026 and existing apps run until 20 September 2029, with IoT Hub, DPS and Fabric as the recommended path.24
The problem it solves
Industrial sites run equipment that speaks OPC UA, MQTT and proprietary protocols, often on isolated networks. Getting that data into analytics and AI usually means a patchwork of gateways and custom code at each site, which is hard to secure and to manage across many plants.
IoT Operations puts a standard set of data services on a Kubernetes cluster at the site, normalizes data before it leaves, and lets central IT manage all sites from Azure through Arc while data stays on local infrastructure until it is sent to the destination you choose.
How it works
An IoT Operations instance runs on an Azure Arc-enabled Kubernetes cluster at the edge:
- MQTT broker: edge-native broker that components and devices publish to and subscribe from.
- Akri connectors and services: for example the connector for OPC UA and the HTTP/REST connector, which publish device data to MQTT topics and can discover assets automatically.
- Data flows: subscribe to topics, transform and contextualize messages using schemas, and route them to cloud endpoints.
- Operations experience: a web UI for OT users to manage assets and data flows.
Azure Resource Manager resources for the instance live in a supported Azure region, but the runtime and operational data remain on your premises; data goes directly from the site to the storage or analytics region you pick. Deployments use either test settings (for evaluation) or secure settings (secrets and user-assigned managed identity, for production).2
Diagram as a list
Copilot & business apps
- Operations experience web UIOT users manage assets and data flowsConnects to Azure Arc and Azure Device Registry
Data & analytics
- Akri connectors (OPC UA, HTTP/REST)Read device data, discover assets and publish to MQTTConnects to MQTT broker
- MQTT brokerEdge-native publish and subscribe messagingConnects to Data flows
- Data flowsTransform, contextualize and route messages
- Microsoft Fabric OneLake and Power BIReal-time dashboards and analytics
- Event Hubs, Event Grid, Data Lake Storage, Data ExplorerOther cloud destinations for processed data
Security, identity & governance
- Azure Key Vault and managed identitySecrets and identity for secure settings deploymentsConnects to Data flows
Cloud & hybrid infrastructure
- Industrial assets and devices (OPC UA servers, PLCs, sensors)Produce telemetry on the plant networkConnects to Akri connectors (OPC UA, HTTP/REST)
- Azure Arc-enabled Kubernetes clusterRuns all IoT Operations services at the siteConnects to Azure Arc and Azure Device Registry
- Azure Arc and Azure Device RegistryCloud management of instances, assets and schemas
Capabilities
Edge MQTT broker2
An industrial-grade MQTT broker runs on the cluster for event-driven architectures.
Why it matters: Devices, connectors and data flows exchange messages locally without a cloud round trip.
Limits: Many broker settings cannot be changed after deployment, so plan sizing first.
Connectors and asset discovery2
Akri connectors such as OPC UA and HTTP/REST bring device data onto MQTT, and Akri services can discover devices and assets automatically.
Why it matters: Reduces manual configuration for OT teams.
Limits: Automatic discovery needs resource sync enabled on the instance.
Data flows1
Transform and contextualize messages and route them to cloud endpoints.
Why it matters: Data is normalized at the edge before it reaches analytics.
Limits: Destinations are limited to the supported endpoint types.
Cloud destinations1
Event Grid and other MQTT brokers, Event Hubs or Kafka, Azure Data Lake Storage, Fabric OneLake and Azure Data Explorer.
Why it matters: Fits existing analytics stacks, including Fabric dashboards and Power BI.
Limits: Each destination is a separately billed service.
Asset registry and schemas2
Azure Device Registry stores assets and devices as Azure resources with namespaces and a schema registry.
Why it matters: Assets can be managed from the Azure portal or CLI as well as at the edge.
Limits: The schema registry requires a customer-provided Blob Storage container reachable by the service.
Offline operation1
The services keep running for up to 72 hours without a cloud connection.
Why it matters: Short network outages do not stop local data collection.
Limits: Microsoft notes possible degradation during that period.
Central management through Arc2
Instances are deployed and managed from the cloud and can be grouped by Azure Arc sites.
Why it matters: One team can run many plants with consistent configuration.
Limits: Azure Arc site manager is in preview.
Practical use cases
Plant managers learn about equipment anomalies only after a failure.
- Approach
- Connect OPC UA assets to the edge MQTT broker, normalize the data with a data flow, and send it to Fabric for real-time dashboards and alerts.
- Role of Azure IoT Operations
- IoT Operations collects and prepares the data at the site.
- Data, infrastructure and skills
- An Arc-enabled cluster at the plant, OPC UA access, a Fabric workspace.
- Type of benefit
- Faster detection and less downtime
- Caveats
- Requires Kubernetes skills at or for the site.
- First step
- Run the Codespaces quickstart, then connect one real OPC UA server.
Equipment effectiveness metrics are calculated by hand from scattered systems.
- Approach
- Capture availability, performance and quality signals at the edge and route them to analytics.
- Role of Azure IoT Operations
- IoT Operations captures and contextualizes the signals in real time.
- Data, infrastructure and skills
- Asset models and agreed KPI definitions.
- Type of benefit
- Operational insight
- Caveats
- KPI logic still has to be built in the analytics layer.
- First step
- Model one production line's assets in Device Registry.
AI models in the cloud need clean, consistent data from many sites.
- Approach
- Use data flows to apply schemas and context at each site before sending data to OneLake.
- Role of Azure IoT Operations
- IoT Operations acts as the edge data plane feeding the AI data foundation.
- Data, infrastructure and skills
- Schemas in the registry and a destination such as Fabric or Data Lake Storage.
- Type of benefit
- Data quality for AI
- Caveats
- Data residency depends on the destination region you choose.
- First step
- Define a schema for one asset type and apply it in a data flow.
Works with
Requires
- Azure ArcIoT Operations runs on Azure Arc-enabled Kubernetes clusters and is managed through Arc.
Optional integration
- Microsoft FabricData flows can send data to Microsoft Fabric OneLake, and Microsoft suggests Fabric for dashboards.
- Power BIMicrosoft lists Power BI for visualizing sensor data from IoT Operations.
Complementary tools
- Azure LocalAKS on Azure Local is a generally available platform for IoT Operations.
- Azure Kubernetes Service (AKS)AKS Edge Essentials and AKS on Azure Local are supported Kubernetes distributions.
Optional integration for
- Azure ArcAzure IoT Operations runs on Azure Arc-enabled Kubernetes clusters and is managed from the cloud through Arc.
Relationship labels follow Microsoft's documentation. "Alternative approaches" does not mean one is better: each profile says when it fits.
Getting started
Explore in Codespaces
Run the GitHub Codespaces quickstart, which deploys IoT Operations on K3s with an OPC PLC simulator and needs no local setup.
Check: Simulated asset data flows to the configured cloud endpoint.
Prepare a real cluster
Install a supported distribution (for example AKS Edge Essentials or K3s on Ubuntu 24.04) on hardware with at least 16 GB RAM and 4 vCPUs, then Arc-enable it.
Check: The cluster appears as a connected Arc-enabled Kubernetes resource.
Deploy with secure settings
Review deployment planning (many broker settings are fixed after deployment), create Storage and Key Vault, and deploy with secure settings.
Check: The instance shows as healthy and secrets sync from Key Vault.
Connect an asset and a destination
Add an OPC UA asset in the operations experience and a data flow to Fabric OneLake.
Check: Asset data appears in a Fabric real-time dashboard.
Official resources
Deciding whether you need it?
Three free tools apply Microsoft's own published guidance to your situation, and say when Azure IoT Operations or any new technology is not needed.
Sources
Each statement above links to the source it comes from. Labels say who reported it.
- What is Azure IoT Operations? (opens in a new tab)
- Azure IoT Operations deployment overview (opens in a new tab)
- Azure IoT Central is evolving (opens in a new tab)
- Azure IoT Operations product page (opens in a new tab)
- Azure Local product page (opens in a new tab)
- Overview of Azure Arc-enabled Kubernetes (opens in a new tab)
Thank you. Your correction was sent.
The editors check it against the sources. If you left an email address, they may reply about it.