The OT/IT Divide: A Historical Accident
Operational Technology (OT) and Information Technology (IT) evolved in complete isolation. OT — the world of PLCs, SCADA systems, industrial sensors, and control networks — was designed around one priority: reliability. Uptime is measured in years. Security models are based on physical isolation. Change management is glacially slow by design.
IT, meanwhile, evolved around agility. Systems are updated frequently, internet connectivity is assumed, and the threat model is completely different.
The result: two worlds that don't speak the same language, use the same protocols, or operate on the same timescales. And between them sits enormous value — data generated on the factory floor that the business layer cannot see, and business intelligence that the operations layer cannot act on.
What IIoT Actually Is
Industrial IoT is not a product or a protocol — it's an architectural approach that creates a secure, bidirectional data bridge between OT systems and IT infrastructure. A well-implemented IIoT architecture allows:
- Real-time visibility into machine performance, utilization, and quality metrics from a web dashboard - Predictive models that identify equipment failure risk days before it happens - ERP systems that automatically adjust production schedules based on machine availability - Energy management systems that optimize consumption based on real-time production load
MQTT and OPC-UA: The Language of IIoT
Two protocols form the backbone of most IIoT implementations.
**MQTT** (Message Queuing Telemetry Transport) is a lightweight publish-subscribe protocol designed for constrained environments and unreliable networks. A temperature sensor publishes to an MQTT broker; a cloud analytics platform subscribes. Simple, efficient, and battle-tested in millions of industrial deployments.
**OPC-UA** (Unified Architecture) is the industrial standard for structured data exchange between PLCs, SCADA systems, and higher-level systems. Where MQTT is lightweight and flexible, OPC-UA is comprehensive — it includes security, authentication, discovery, and a rich data model capable of describing complex industrial assets.
Modern IIoT architectures typically use OPC-UA at the machine level to extract structured data from PLCs, then MQTT to transmit that data efficiently across the plant to an edge computing node, and finally REST APIs or cloud connectors to push aggregated insights to IT systems.
Edge Computing: The Missing Piece
Sending raw sensor data to the cloud for processing creates latency, bandwidth costs, and privacy concerns. Edge computing — processing data at or near the source, on industrial-grade hardware in the plant — solves all three. Edge nodes run local analytics, detect anomalies in real time, and push only relevant events and aggregated metrics to the cloud.
At Abstriq, we deploy edge computing nodes running Node-RED and TimescaleDB alongside every IIoT project. This architecture delivers sub-100ms response times for local alerting while providing cloud dashboards with 1-second data resolution.
Real ROI Numbers From Our Deployments
The business case for IIoT is well established at this point. Across our deployments:
A pharmaceutical manufacturer in Pune reduced unplanned downtime by 38% in the first year by implementing vibration and temperature monitoring on 24 critical pumps and compressors. The predictive maintenance model flagged 7 impending failures that would otherwise have been unplanned shutdowns, each costing 4–8 hours of production time.
An automotive components manufacturer in Manesar increased OEE (Overall Equipment Effectiveness) from 67% to 81% over 18 months by correlating machine data with shift patterns, identifying systematic setup inefficiencies invisible to floor supervisors.
An FMCG packager reduced energy costs by 22% by implementing real-time energy monitoring and automated load shedding during peak tariff windows.
Implementation Roadmap
The classic mistake is trying to boil the ocean — connecting every machine at once. The right approach:
1. **Identify one high-value asset** with clear, measurable outcomes (downtime cost, energy cost, quality defect rate) 2. **Instrument it with low-invasive sensors** — vibration, temperature, current draw 3. **Deploy edge processing and a simple dashboard** — 4–8 weeks 4. **Demonstrate ROI** at the 3-month mark 5. **Scale horizontally** to additional assets using the proven architecture
The first deployment is the hardest. Each subsequent one leverages the same infrastructure and expertise, and typically costs 40–60% less than the first.