Trying to make sense of IoT architecture layers? Think devices, network connectivity, service support (often the cloud) and applications, with security and management running across all of them.
If you are mapping a whole project, our guide to iot development shows where architecture fits among the other decisions.
| Item | Value |
|---|---|
| Layers in the ITU IoT reference model | Four layers: application, service and application support, network, device |
| Capabilities that span those layers | Management and security |
| ITU-T Y.2060 (Overview of the Internet of things) adopted | June 2012 |
| Typical size of proposed generic IoT architectures | Three to five layers |
| Layering in the OSI network model, for comparison | 7-layer |
These figures come from the Digital Watch Observatory summary of the ITU recommendation and the Napier University IoT architecture notes. Standards and summaries can change, so check the source before you rely on them.
Why do IoT systems use layers at all?
In an IoT project, the same outcome can be reached with many different devices and technologies, so every build risks ending up slightly different. A reference architecture gives you a broad-brush framework to follow, so you can focus on how each component works inside instead of reinventing the overall shape, as the Napier University notes explain.
Layers work much like the well-known 7-layer OSI network model. Lower levels sit closer to physical reality, while higher levels relate more to the logical purpose of the system, according to the same Napier notes.
Because IoT is a young and dynamic field, many generic architectures have been proposed, and they typically include three to five layers. Expect to meet models that count layers differently, and read them as different views of the same system. The Napier author picks the ITU model as a good compromise between clarity and detail.
The four layers of the ITU IoT reference model
As of the Digital Watch summary of the June 2012 recommendation, ITU-T Y.2060 explains an IoT reference model built on four layers plus management and security capabilities. The Digital Watch Observatory lists them as application, service and application support, network, and device, and the ITU publication page repeats the same structure.
Read from the top down, the layers are:
- Application layer: holds the IoT applications themselves.
- Service support and application support layer: shared and specialised capabilities that applications draw on.
- Network layer: networking and transport capabilities that carry data and control information.
- Device layer: device capabilities and gateway capabilities.
Data tends to travel upward from the devices, while commands travel back down. Management and security capabilities are associated with every layer rather than sitting in a box of their own.

Many teams shorten this to devices, connectivity, cloud and apps. That is a common shorthand and not ITU wording: the device layer becomes devices, the network layer becomes connectivity, service support is where cloud platforms usually sit, and the application layer becomes apps.
Device layer: direct devices and gateways
The ITU text splits device layer capabilities into two kinds: device capabilities and gateway capabilities. Device capabilities include:
- Direct interaction with the communication network, without using gateway capabilities.
- Indirect interaction, where information passes through gateway capabilities.
- Ad-hoc networking in scenarios that need increased scalability and quick deployment.
- Sleeping and waking-up mechanisms to save energy.
Gateways exist because devices do not all speak the same language. At the device layer, the same ITU page says gateway capabilities support devices connected through wired or wireless technologies such as a CAN bus, ZigBee, Bluetooth or Wi-Fi.
A gateway also converts protocols. The ITU describes two situations that call for it: when devices use different device layer protocols, and when the device layer and network layer use different protocols.
If you are building battery-powered nodes, treat sleep and wake behaviour as an early design choice rather than a late optimisation.
Network layer: how data and commands travel
According to the ITU publication, the network layer consists of two types of capabilities. Networking capabilities cover control functions such as access and transport resource control, mobility management, and authentication, authorization and accounting (AAA).
Transport capabilities focus on connectivity for IoT service and application data. They also carry IoT-related control and management information, so the same path serves both your data and your device administration.
The protocol choices you make live around this layer and the ones above it. Our guide to iot protocols goes deeper into how devices and platforms exchange messages.
Where does the cloud fit? Service support and application layers
The ITU page says the service support and application support layer has two capability groupings. Generic support capabilities are common to different IoT applications, with data processing and data storage given as examples.
Specific support capabilities cater for the requirements of diversified applications and can include various detailed groupings. That lets different IoT applications receive different support functions.
The model does not name a “cloud layer”. In practice, a cloud platform is a common way to supply storage and processing, which is why the cloud is usually drawn at this level.
The application layer sits on top and, in the ITU wording, simply contains the IoT applications. Dashboards, mobile apps and alerting tools are typical examples you might build there.
Management and security across every layer
The ITU reference model pairs its four layers with management capabilities and security capabilities that are associated with those layers. Treat them as questions to ask at each level, not as a separate stage at the end.
The Digital Watch summary also notes that mitigating security threats and protecting privacy and personal data during IoT data transmission are challenges the private sector and technical community keep examining. It lists interconnectivity, heterogeneity and dynamic changes among the fundamental characteristics of the IoT, which is one reason gateways and shared support capabilities matter.
A practical per-layer checklist might ask:
- Device layer: how is each device identified and updated?
- Network layer: who is allowed to connect, and how is access controlled?
- Support layer: where is data stored, and who can process it?
- Application layer: which users can see or command what?
Using the layers to plan your build
The Napier notes state that an architectural design aims to facilitate interoperability, simplify development and ease implementation. A layer diagram helps you reach that by letting you decide on technology one level at a time.
- List what each device must do and whether it talks to the network directly or through a gateway.
- Choose the connectivity and any protocol conversion the gateway must handle.
- Decide which storage and processing capabilities several applications will share.
- Define the application and who uses it.
- Note the management and security needs of every layer.
When you are ready to move from diagram to hardware, our notes on iot prototyping explain how to start small and test the layers in turn.
Frequently asked questions
How many layers does an IoT architecture have?
Show answer
There is no single answer. Published generic models usually land between three to five layers, and the ITU model described above counts four, with management and security alongside.
Is the cloud a separate layer?
Show answer
Not in the ITU model. Cloud services typically provide the data storage and data processing that the service support layer describes.
Does every device need a gateway?
Show answer
No. The ITU text allows devices to reach the network directly or indirectly, and it says a single device does not have to support both routes.
Do you have to follow the ITU model exactly?
Show answer
No. A reference architecture is deliberately generic, and the Napier notes say it can be refined into a specific architecture for a particular application domain. Treat the ITU layers as a starting map for your own design.
