IoT Protocol Comparison: BACnet, Modbus, OPC-UA, MQTT, DNP3
January 10, 2026 · Marcus Chen
Integration work in buildings and industry begins with a confrontation with protocol heritage. The systems on site were not designed to speak to each other; they were designed in different eras, for different network assumptions, by vendors with different priorities. A building management system speaks BACnet because the ASHRAE standard defined how building automation should interoperate; the power meters speak Modbus because it is the de facto language of industrial instrumentation; the utility substation gear speaks DNP3 because the electrical grid has its own reliability requirements; and the new wireless sensors speak MQTT because it is lightweight and internet-native. None of these protocols is “better” in the abstract—each is the right answer to the problem its designers were solving, and integration work is the art of making five right answers coexist.
The Five Protocols in Their Own Words
BACnet (ASHRAE 135) is the building automation standard. It defines a data model—objects for points, schedules, trends, and alarms, each with typed properties—and supports several transports: MS/TP over RS-485 for field-level wiring, IP for building-level networking, and BACnet/SC as a modern encrypted option. Its defining feature is the object model: a temperature sensor, an AHU, or a schedule is a standardized object with standard properties, allowing devices from different manufacturers to interoperate without a custom mapping for each pair. BACnet devices can also discover each other over IP, which is why it became the lingua franca of building automation rather than a curiosity.
Modbus is the opposite philosophy: deliberately minimal. It is a master/slave protocol over serial (RTU) or TCP, built around a simple model of registers and coils. Its genius is its simplicity: any device can implement a Modbus slave with a small amount of code, which is why power meters, drives, flow meters, and a thousand other instruments ship with it as their universal interface. The costs are real: no standardized data model (so “register 40001” means whatever the vendor’s manual says), no discovery, no native security, and no rich semantics.
OPC UA is the industrial interoperability standard that inverts Modbus’s approach. It is a full service-oriented architecture with an information model that can represent types, hierarchy, and relationships—a pump with its associated motor and sensors, not just a pile of registers—plus encryption, authentication, and a subscription model for efficient data delivery, with a newer pub/sub mode for distributed scenarios. Its strength is expressiveness and security; its cost is complexity and the engineering discipline required to use it well.
MQTT is the IoT-native protocol: a lightweight publish/subscribe model over TCP where devices publish to topics and a broker routes messages to subscribers. Its strengths are minimal overhead, support for unreliable networks (three quality-of-service levels), and the natural fit of topics for organizing telemetry. Its trade-offs are that it carries no semantic model—“sensor/room3/temperature” means what you say it means—and that security is broker-dependent: a well-configured broker with TLS and authentication is a solid foundation, but the protocol itself specifies no security.
DNP3 is the utility/SCADA protocol, built for wide-area telemetry in electrical, water, and pipeline systems. Its defining characteristics are robustness on poor links: reliable transport over slow, noisy serial or TCP connections, unsolicited event reporting that avoids wasteful polling, and time-stamped events with quality flags.
A Side-by-Side Reference
| Protocol | Domain | Data model | Transport | Security | Best at |
|---|---|---|---|---|---|
| BACnet | Buildings | Rich, standardized objects | MS/TP, IP, BACnet/SC | Weak by default; BACnet/SC adds it | Interoperability of building systems |
| Modbus | Industrial instruments | Minimal registers/coils | Serial, TCP | None native | Simple device integration |
| OPC UA | Industrial automation | Rich, typed, hierarchical | TCP, pub/sub | Native, strong | Representing real assets with context |
| MQTT | IoT telemetry | None (topic-based) | TCP (plus MQTT-SN) | Broker-dependent | Lightweight telemetry over unreliable links |
| DNP3 | Utilities/SCADA | Points, events, time stamps | Serial, TCP | Secure DNP3 variant | Wide-area, noisy-link reliability |
Why the Choice Matters: The Semantics Problem
A protocol’s choice is a decision about where meaning lives: rich protocols (BACnet, OPC UA) push meaning into the protocol itself—a BACnet sensor arrives with its engineering units, deadband, and alarm limits as typed properties; simple protocols (Modbus, MQTT) push it into the integration layer, where a Modbus register’s meaning lives in the vendor manual and an MQTT reading’s meaning lives in the topic convention you defined. This is the practical difference between systems that integrate in days and systems that integrate in months, and the reason the integration layer—a platform that normalizes all five into one model—always assumes a mix.
Selecting a Protocol for a New Deployment
For a new build rather than an integration, the selection criteria are the deployment’s own:
- What does the device market in this domain speak? Protocol selection is partly a market fact. New building equipment speaks BACnet; new industrial instruments speak Modbus and increasingly OPC UA; new field sensors speak MQTT. Choosing a protocol the market does not speak means custom integration for every device.
- Does the domain demand semantics or simplicity? A system whose value depends on rich context benefits from OPC UA or BACnet’s data models; a simple telemetry stream does not, and paying OPC UA’s complexity for a temperature feed is over-engineering.
- What are the network conditions? MQTT tolerates unreliable connectivity gracefully; DNP3 was built for poor links; a rich polling protocol on a flaky link is a recipe for gaps.
- What security is required? OPC UA and the secure variants (BACnet/SC, DNP3 Secure Authentication) carry security natively; Modbus and plain MQTT require the transport layer to supply it.
The pragmatic default for new telemetry is often MQTT to a secured broker, with the richer protocols reserved for the systems that need their semantics—because no site runs one protocol.
The Integration Reality: Five Protocols, One Model
The honest summary for anyone building an enterprise system is that protocol selection is a fleet decision, not a single choice. A typical facility runs BACnet on the building systems, Modbus on the meters, and MQTT on the wireless field devices, with OPC UA at the process level and possibly DNP3 feeding from a utility interface. The platform’s job is to normalize all of them into a single model—consistent units, naming, and time semantics—so that the analytics above them do not know or care which protocol carried each reading. Integration quality is the product of how well that normalization handles each protocol’s quirks: the BACnet object types, the Modbus register maps, the OPC UA type hierarchy, the MQTT topic taxonomy, and the DNP3 point and event model.
The protocols are not competing technologies to be ranked; they are a legacy the industry inherited, and each is still the best answer to the problem it was built for.