IoT systems are not built from a single blueprint. The architecture of an IoT deployment depends heavily on the requirements, scale, latency constraints, and processing needs of the specific use case. This article covers the most widely recognized IoT architecture models and how they relate to each other.
Why There Is No Single IoT Architecture
Unlike traditional software systems, IoT spans a wide spectrum — from a simple temperature sensor reporting to a cloud dashboard, to a complex industrial network processing thousands of events per second at the edge. Because of this diversity, no single architecture is agreed upon universally.
What does exist are reference models that provide common vocabulary and structure for thinking about IoT systems.
Basic Three-Layer Architecture
The simplest and most widely referenced IoT architecture has three layers:
+---------------------------+
| Application Layer | Application-specific services for users
+---------------------------+
| Network Layer | Connects smart things, network devices, servers
+---------------------------+
| Perception Layer | Sensors and environment information
+---------------------------+
- Perception Layer — the physical layer; sensors and actuators that collect data from the real world
- Network Layer — handles connectivity and transmission between devices and servers
- Application Layer — delivers services specific to the use case (smart home, smart city, healthcare, etc.)
This model is clean and intuitive, but it lacks detail about data processing and business logic.
Five-Layer Architecture
To address that gap, the three-layer model is often expanded to five layers:
+-----------------------------+
| Business Layer | Manages the whole IoT system, business models, user privacy
+-----------------------------+
| Application Layer | Delivers application-specific services to the user
+-----------------------------+
| Processing Layer | Stores, analyzes, and processes data
+-----------------------------+
| Transport Layer | Transfers sensor data between layers through the network
+-----------------------------+
| Perception Layer | Sensor and environment information
+-----------------------------+
The additions are:
- Transport Layer — moves data between the perception layer and the processing layer across different networks
- Processing Layer — handles storage, analytics, and data processing (often implemented as middleware or cloud services)
- Business Layer — manages the overall IoT system including applications, business and profit models, and user privacy
Cloud-Centric Architecture
When large-scale centralized processing is required, the cloud-centric model is used. Data processing happens primarily in the cloud rather than at the edge.
+-----------------------------+
| Application Layer |
+-----------------------------+
| Cloud Layer | Centralized processing and storage
+-----------------------------+
| Network Layer |
+-----------------------------+
| Perception Layer | Smart things (sensors, devices)
+-----------------------------+
Cloud-centric architecture keeps the cloud at the center, applications above it, and the network of smart things below it.
Advantages:
- Virtually unlimited processing and storage capacity
- Centralized management and updates
- Easy integration with other cloud services
Disadvantages:
- Latency — data must travel to and from the cloud
- Bandwidth consumption — all raw data is transmitted
- Single point of failure if connectivity is lost
Fog Computing Architecture
Fog computing addresses the limitations of cloud-centric architectures by introducing intermediate processing layers at the edge of the network, before data reaches the cloud.
+-----------------------------+
| Application Layer |
+-----------------------------+
| Cloud Layer |
+-----------------------------+
| Transport Layer |
+-----------------------------+
| Security Layer | Encryption/decryption, data integrity, privacy
| Storage Layer | Temporary storage, data replication, distribution
| Pre-processing Layer | Filtering, processing, analytics of sensor data
| Monitoring Layer | Monitors power, resources, responses, and services
+-----------------------------+
| Physical Layer | Sensors and actuators
+-----------------------------+
Fog architecture inserts four sublayers between the physical and transport layers:
- Monitoring Layer — monitors power, resources, responses, and services across the network
- Pre-processing Layer — performs filtering, processing, and analytics of sensor data before transmission
- Storage Layer — provides temporary storage with data replication and distribution
- Security Layer — performs encryption/decryption and ensures data integrity and privacy
Monitoring and pre-processing are done on the **edge of the network** before sending data to the cloud — reducing latency, bandwidth, and cloud costs.
IoT Taxonomy — Common Layers Across Architectures
Regardless of the specific architecture model used, most IoT systems include the following functional layers:
| Layer | Responsibility |
|---|---|
| Perception | Collects data using various types of sensors |
| Preprocessing | Filters and summarizes data before it enters the network |
| Communication | Different entities communicate over the network using various protocols and standards |
| Middleware | Creates abstraction by hiding hardware details from applications |
| Applications | Various applications that improve user experience and increase efficiency |
Choosing the Right Architecture
The right IoT architecture depends on:
- Latency requirements — real-time control needs edge processing; dashboards can tolerate cloud latency
- Bandwidth constraints — fog/edge models reduce data transmitted to the cloud
- Scale — cloud-centric models handle massive scale more easily
- Security posture — sensitive data may need to stay on-premises or encrypted at the edge
- Cost — fog reduces cloud egress costs; cloud reduces on-premises infrastructure investment
Latency-critical systems --> Fog / Edge architecture
Large-scale analytics --> Cloud-centric architecture
Balanced requirements --> Hybrid (Fog + Cloud)
Understanding these trade-offs is the foundation for designing effective IoT ecosystems.