Guide / Industrial IoT

IIoT platform
architecture.

A practical guide to how industrial IoT systems are structured, from sensor to cloud, including the layers, the protocols, edge computing, device identity, and post-quantum security.

What is IIoT platform architecture?

IIoT platform architecture describes how industrial IoT systems are structured across multiple layers: physical devices and sensors, edge gateways for local processing, a message broker for telemetry transport, a central IoT platform for device management and dashboards, and identity and security layers. Each layer has defined responsibilities, protocols, and scaling characteristics.

Architecture

The seven layers
of an IIoT platform.

Industrial IoT systems are built from distinct functional layers. Understanding each layer’s role helps you choose the right components and avoid common design pitfalls.

01

Device and sensor layer

Physical endpoints: PLCs, RTUs, sensors, meters, actuators, and embedded modules. Devices communicate over field protocols (Modbus, OPC-UA, Wireless M-Bus, BACnet) or directly over IP (MQTT, CoAP, HTTP).

microcontrollersPLCsRTUssensorsModbusOPC-UA
03

Message broker and transport layer

The message broker decouples device producers from platform consumers. MQTT is the dominant protocol for constrained devices; AMQP bridges enterprise services. High-availability brokers with durable queues prevent data loss during outages.

MQTT brokermessage queueAMQPpublish-subscribeQoS
05

Identity and authorization layer

Every device, user, and service must be authenticated and authorized before accessing resources. Runtime policy evaluation prevents the stale-permission windows that occur when permissions are embedded in tokens.

device identityIAMRBACABACJWTauthorization
06

Edge compute and orchestration layer

WebAssembly workloads can be deployed and updated remotely on both cloud nodes and microcontrollers. This layer enables AI inference, data transformation, and protocol bridging at the edge without firmware reflashing.

WebAssemblyedge AIFaaSOTA updatesWasm orchestration
07

Security and cryptography layer

Long-lived IoT devices must remain secure after classical PKI is broken by quantum computers. Hardware-accelerated post-quantum cryptography provides forward-secure device identity that outlives the deployment lifetime.

post-quantum cryptographysecure elementhardware root of trustML-KEMML-DSA

Stack overview

How our stack maps
to each layer.

Each product addresses a specific architectural layer. They can be deployed individually or as an integrated stack.

Applications and dashboards

Magistrala: visual dashboards, alarm management, reports, rules engine

Device management and platform

Magistrala: device provisioning, channel management, multi-tenancy, SDKs

Identity and authorization

Atom: runtime RBAC and ABAC, JWT, OIDC, device credentials, policy engine

Message transport

FluxMQ: MQTT 5.0, AMQP, CoAP, durable queues, 500K+ connections per node

Edge compute and orchestration

Propeller: WebAssembly orchestrator, OCI registry, FaaS at the edge, Zephyr support

Edge gateway (Linux)

A1 Gateway: RISC-V Linux, containers, Wasm, manufactured in the EU

Edge gateway (RTOS / MCU)

A0 Gateway: ESP32-C6 RISC-V, Zephyr, Wi-Fi 6, BLE, 802.15.4, Wireless M-Bus, NB-IoT/LTE-M and sub-GHz modules

Hardware security

Quarc: post-quantum secure element, ML-KEM-768, ML-DSA-65, hardware root of trust

Protocols

Which protocol,
and when.

Protocol selection depends on device constraints, network conditions, data volume, and integration requirements. Most production deployments use several.

TCP / WebSocket 3.1.1 / 5.0

MQTT

Message Queuing Telemetry Transport

Primary IoT telemetry protocol. Lightweight publish-subscribe with QoS 0/1/2, retained messages, and Last Will. MQTT 5.0 adds shared subscriptions, flow control, and user properties.

TCP 0.9.1 / 1.0

AMQP

Advanced Message Queuing Protocol

Enterprise message queuing. Used for integrating IoT data streams with ERP, MES, and cloud services. Supports transactions, routing topologies, and dead-letter queues.

UDP / DTLS RFC 7252

CoAP

Constrained Application Protocol

REST-like protocol optimised for constrained devices with limited bandwidth. Runs over UDP and includes DTLS for security. Common in smart metering and sensor networks.

TCP / WebSocket

OPC-UA

Open Platform Communications Unified Architecture

Industrial automation standard for PLC and SCADA integration. Rich information model, complex address space, and built-in security. Common in manufacturing and process control.

RS-485 / TCP

Modbus

Modbus RTU / TCP

Ubiquitous serial protocol in industrial automation. Runs over RS-485 (RTU) or Ethernet (TCP). No built-in security, so remote access requires gateway-level encryption.

868 MHz RF 868 MHz

Wireless M-Bus

EN 13757

European standard for wireless utility meter reading. Used for smart metering of electricity, gas, heat, and water. Operates at 868 MHz sub-GHz for range and wall penetration.

Design decisions

Edge or cloud
processing?

Where data is processed has significant implications for latency, bandwidth costs, offline resilience, and security.

Process at the edge when

  • Control loops require sub-100ms response times, and cloud round-trips introduce too much latency
  • Devices operate in locations with intermittent connectivity, so local processing must continue offline
  • Data volume is high but only aggregates or anomalies are relevant to the cloud
  • Data sovereignty or regulatory requirements restrict transmission of raw device data off-premises
  • AI inference must run close to the sensor to be actionable, such as vision or predictive maintenance

Process in the cloud when

  • Long-term storage and historical analysis are the primary use case, where cloud storage is cheaper and more scalable
  • Cross-site correlation is needed, aggregating data from hundreds of sites for fleet-level analytics
  • Heavy machine-learning training runs require GPU resources not available at the edge
  • Compliance reporting and audit trails must be tamper-proof and centrally managed
  • Dashboards and alerting are accessed by operators from anywhere

Security

IIoT security
architecture.

Industrial IoT systems have distinct security requirements. Devices are long-lived, operate in physical environments, and often control critical infrastructure.

Device identity

Every device needs a unique cryptographic identity issued at manufacture or provisioning. Device certificates or API keys authenticate the device to the message broker and platform. When a device is decommissioned, its credentials must be revoked and access blocked immediately. Runtime policy evaluation, rather than token-embedded permissions, guarantees this.

Atom, runtime IAM

Transport security

All device-to-broker and broker-to-platform communication should be encrypted: TLS 1.2+ for MQTT and HTTPS, mTLS for inter-service communication, DTLS for CoAP over UDP. For constrained devices where TLS is too heavy, lighter alternatives exist but increase risk; the gateway should terminate TLS on behalf of downstream devices.

Secure boot and OTA updates

Firmware must be signed and the signature verified at boot before execution. OTA updates must be delivered over authenticated, encrypted channels and include rollback protection. A failed update must not brick the device or leave it in an unknown state.

Quarc, hardware-enforced secure boot

Post-quantum preparedness

Industrial devices deployed today may still be in the field when quantum computers can break ECC and RSA. NIST finalised post-quantum cryptography standards in 2024 (FIPS 203, 204, 205). New designs should evaluate ML-KEM for key encapsulation and ML-DSA for signatures, especially for devices with 10+ year deployment lifetimes.

Quarc, PQC secure element

FAQ

Common questions.

What is IIoT platform architecture?

IIoT platform architecture describes how industrial IoT systems are structured: the layers, components, and protocols that connect physical devices to cloud applications. A typical architecture includes device and sensor layers, edge gateways, message brokers, an IoT platform for device management and dashboards, an identity and authorization layer, and optional edge compute and cryptography layers.

What is the difference between IoT and IIoT?

IoT (Internet of Things) is the broad category of connected devices: consumer electronics, smart home devices, wearables. IIoT (Industrial Internet of Things) specifically refers to connected devices in industrial contexts such as manufacturing, energy, utilities, transportation, and infrastructure. IIoT systems typically have higher reliability requirements, longer device lifetimes, and stricter security and compliance needs.

Why is MQTT the dominant protocol for IIoT?

MQTT was designed for constrained devices over unreliable networks. Its publish-subscribe model decouples producers from consumers. Quality of Service levels (0/1/2) provide flexible delivery guarantees. It runs over TCP with minimal overhead, making it suitable for everything from microcontrollers to cloud-scale brokers. MQTT 5.0 adds features like shared subscriptions, topic aliases, and enhanced authentication that enterprise deployments need.

What is an IoT gateway and do I need one?

An IoT gateway is a device that aggregates data from local sensors and field devices and forwards it to a cloud platform. Gateways are necessary when field devices speak non-IP protocols (Modbus, Wireless M-Bus, Bluetooth), when local processing or offline buffering is required, or when you want to reduce cloud bandwidth by aggregating and filtering data at the edge. For devices with native IP and MQTT support, a gateway is optional.

How does edge computing fit into IIoT architecture?

Edge computing moves processing from the cloud to the device or gateway. In IIoT this reduces latency for time-critical control loops, lowers cloud bandwidth costs, and enables operation during cloud connectivity loss. Edge compute layers typically run lightweight runtimes such as containers, WebAssembly, or native binaries on gateway hardware. WebAssembly is increasingly used because it provides sandboxed isolation, sub-10ms startup, and portability from cloud servers down to microcontrollers.

What is a message broker in IIoT architecture?

A message broker is the central routing component for device telemetry. Devices publish messages to topics; platform services subscribe to those topics. The broker decouples producers from consumers, handles delivery guarantees (QoS), and can buffer messages during downstream failures. In enterprise IIoT, brokers also provide durable queues with consumer groups so multiple downstream services can consume the same data stream independently.

How should IoT device identity be managed?

Each device needs a unique cryptographic identity, typically a certificate or API key associated with a device record in the platform. That identity controls which topics the device can publish and subscribe to. Runtime policy evaluation, rather than token-embedded permissions, ensures that when a device is decommissioned or its credentials are revoked, access is blocked immediately without waiting for tokens to expire.

Why does post-quantum cryptography matter for IIoT now?

Industrial devices deployed today often remain in the field for 10 to 20 years. Quantum computers capable of breaking ECC and RSA are projected to become practical within that window. NIST finalised post-quantum cryptography standards in 2024. EU and US regulators project mandatory PQC adoption in critical infrastructure by 2030. Devices that ship now with only classical cryptography will be vulnerable before end-of-life.

Abstract Machines products

Open-source components
for every layer.

Explore all our work

Your next step

Designing an
industrial system?

Tell us about the devices, the data, and the constraints. We will help you choose the right layers and components.

Photography & film

Selected stock imagery illustrates our areas of work. It does not depict Abstract Machines’ team or offices.

Photography under the Unsplash License and Pexels License. Film under the Pexels License.