The latest articles on DEV Community by Gaurav Sharma (@gaurav_sharma_c2ef5dd7646). https://dev.to/gaurav_sharma_c2ef5dd7646 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F2057038%2F363a6933-a83d-4a8a-af41-8fe5e4d9dbf0.jpg https://dev.to/gaurav_sharma_c2ef5dd7646 en Gaurav Sharma Tue, 31 Mar 2026 16:25:27 +0000 https://dev.to/gaurav_sharma_c2ef5dd7646/designing-a-practical-iiot-architecture-1cnn https://dev.to/gaurav_sharma_c2ef5dd7646/designing-a-practical-iiot-architecture-1cnn <h2> Introduction </h2> <p>Industrial IoT (IIoT) architectures often look clean in diagrams but become messy when implemented in real systems.</p> <p>While exploring this space, I tried to structure a more practical reference architecture that balances clarity with real-world concerns like scalability, reliability, and multi-tenancy.</p> <p>This is a work in progress, and I’d love feedback from others building similar systems.</p> <h2> The Problem with Typical IoT Architectures </h2> <p>Most reference architectures:</p> <ul> <li>stay too high-level</li> <li>don’t separate concerns clearly (telemetry vs command)</li> <li>ignore multi-tenancy</li> <li>don’t address edge vs cloud trade-offs</li> </ul> <p>This creates challenges when moving from POC → production.</p> <h2> Proposed Architecture Overview </h2> <p>At a high level, the system is divided into four layers:</p> <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fr71vwgg0lymy86gyx7qp.png" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fr71vwgg0lymy86gyx7qp.png" alt=" " width="800" height="533"></a></p> <h3> 1. Device &amp; Controller Layer </h3> <ul> <li>Sensors, controllers, PLCs</li> <li>Communicate via MQTT or similar lightweight protocols</li> <li>Separate channels for telemetry and command</li> </ul> <h3> 2. Edge Layer </h3> <ul> <li>Local processing and filtering</li> <li>Handles intermittent connectivity</li> <li>Reduces latency for critical operations</li> </ul> <h3> 3. Cloud Ingestion Layer </h3> <ul> <li>MQTT broker → event streaming pipeline</li> <li>Separation of: <ul> <li>telemetry (data ingestion)</li> <li>command (control path)</li> </ul> </li> </ul> <h3> 4. Application Layer </h3> <ul> <li>Multi-tenant SaaS architecture</li> <li>Monitoring dashboards</li> <li>Command and control APIs</li> </ul> <h2> Key Design Considerations </h2> <h3> Telemetry vs Command Separation </h3> <p>Combining both leads to complexity. Keeping them separate improves scalability and clarity.</p> <h3> Edge vs Cloud Responsibilities </h3> <ul> <li>Edge: real-time, low-latency decisions </li> <li>Cloud: aggregation, analytics, orchestration </li> </ul> <h3> Multi-Tenancy </h3> <p>Needs to be built-in from the start (not bolted on later).</p> <h2> Open Questions / Trade-offs </h2> <p>Some areas I’m still exploring:</p> <ul> <li>How to design reliable command execution (ACK/NACK, retries, idempotency)</li> <li>Where to draw the boundary between edge and cloud</li> <li>Best approaches for tenant isolation at scale</li> </ul> <h2> Reference </h2> <p>I’ve put together a more detailed version (with diagrams) here:</p> <p>👉 <a href="https://gauravs19.github.io/iiot-reference-architecture/" rel="noopener noreferrer">https://gauravs19.github.io/iiot-reference-architecture/</a></p> <h2> Looking for Feedback </h2> <p>If you’ve worked on IoT/IIoT systems:</p> <ul> <li>What patterns have worked well for you?</li> <li>Where do architectures typically break?</li> <li>What would you change here?</li> </ul> <p>Open to suggestions and alternative approaches.</p> iot architecture systemdesign cloud