The TCP/IP model divides network communication into four layers: application, transport, internet and link. Each layer handles one job and passes its work to the layer below, so protocols can be designed, replaced and understood independently.
Why is networking split into layers?
A single program that handled everything from choosing a web page to putting voltages on a wire would be enormous, impossible to maintain, and locked to one kind of hardware. Layering breaks the problem into manageable pieces with clear boundaries.
The practical benefits are worth learning as a list, because exam questions ask for them directly:
- Independence. A layer can be changed without rewriting the others. Switching from Wi-Fi to Ethernet changes only the link layer; your browser is unaffected.
- Interoperability. Devices from different manufacturers work together as long as they implement the same layer protocols.
- Simpler design and troubleshooting. A fault can be traced to a layer, narrowing the search enormously.
- Specialisation. Engineers can work on one layer without needing to understand the internals of every other.
The four layers
1. Application layer (top)
This is where the software the user interacts with lives. The application layer decides what is being communicated — a web page request, an email, a file transfer — and formats it according to an agreed protocol.
Protocols found here include HTTP and HTTPS for web traffic, SMTP for sending email, IMAP and POP3 for retrieving it, FTP for file transfer, and DNS for turning domain names into IP addresses.
2. Transport layer
The transport layer establishes an end-to-end connection between two devices and splits the data into packets. It numbers those packets so they can be reassembled in the right order at the far end, and — depending on the protocol — checks that they all arrived.
Two protocols matter here:
- TCP (Transmission Control Protocol) is connection-oriented and reliable. It acknowledges every packet, retransmits anything lost, and delivers data in order. Used for web pages, email and file downloads, where a missing fragment would corrupt the result.
- UDP (User Datagram Protocol) is connectionless and does not guarantee delivery or order. It is much lighter, so it suits live video, voice calls and online games, where a slightly dropped frame matters less than arriving on time.
3. Internet layer (sometimes called the network layer)
The internet layer adds the source and destination IP addresses to each packet and handles routing — deciding which path a packet takes across the network to reach its destination.
Its main protocol is IP (Internet Protocol). Routers operate at this layer: each one reads the destination address, consults its routing table, and forwards the packet one hop closer.
4. Link layer (bottom)
The link layer is the physical connection and the hardware that drives it: the network interface card, the cabling or radio, the switch. It handles MAC addresses, which identify devices on the local network, and converts packets into the electrical, optical or radio signals that actually travel.
Ethernet and Wi-Fi are link-layer technologies.
How the layers work on a real request
When you load a web page:
| Layer | On the sending device | On the receiving device |
|---|---|---|
| Application | The browser builds an HTTP GET request | The web server reads the HTTP request |
| Transport | TCP splits it into numbered packets | TCP reassembles packets in order, requests any missing |
| Internet | IP addresses are added; a route is chosen | IP addresses are checked and stripped |
| Link | Packets become signals on the wire or in the air | Signals are converted back into packets |
Data travels down the layers on the sending device, across the physical network, then up the layers on the receiving device. Each layer on the way down adds its own header information, and the matching layer on the way up reads and removes it. That pairing is what makes layered networking work — each layer talks only to its opposite number.
The TCP/IP model and the OSI model
You may also see the seven-layer OSI model: physical, data link, network, transport, session, presentation and application. It is a more detailed reference framework that predates widespread TCP/IP use.
The two map onto each other loosely — TCP/IP's application layer covers OSI's top three, and its link layer covers OSI's bottom two. GCSE specifications generally examine the four-layer TCP/IP model, so learn that one thoroughly and simply be aware the OSI model exists.
Common exam mistakes
- Putting protocols on the wrong layer. HTTP is application; TCP is transport; IP is internet; Ethernet is link. Learn these four pairings and you will get most of the marks available.
- Saying the transport layer routes packets. Routing is the internet layer's job. The transport layer numbers and reassembles.
- Confusing IP addresses with MAC addresses. IP addresses identify a device on the internet and are used for routing; MAC addresses identify hardware on the local network segment.
- Forgetting the direction of travel. Data goes down the layers to send and up to receive.
- Listing the benefits of layering vaguely. "It makes things easier" earns nothing; "a protocol can be changed without affecting other layers" earns the mark.
Frequently asked questions
What is the difference between TCP and UDP, and when is each used?
TCP is connection-oriented and reliable: it establishes a connection first, numbers every packet, acknowledges receipt and retransmits anything lost, delivering data complete and in order. UDP simply sends packets with no acknowledgement, no retransmission and no guaranteed ordering. TCP suits web pages, email and downloads where every byte must arrive intact. UDP suits live streaming, voice and video calls and online gaming, where waiting for a retransmitted packet would cause a worse experience than dropping it.
Which layer do routers and switches operate at?
Routers operate at the internet layer: they read destination IP addresses and forward packets between different networks. Switches operate at the link layer: they read MAC addresses and forward frames between devices on the same local network. This is a favourite exam distinction, and the reason is worth stating — a router needs to understand IP addressing to choose a path across the internet, whereas a switch only needs to know which port each local device is attached to.
Why does layering make it easier to develop new technology?
Because a new protocol only needs to work correctly with the layers immediately above and below it. When Wi-Fi was introduced it replaced Ethernet at the link layer; every application, every transport protocol and the whole of IP carried on unchanged. Without layering, adding wireless networking would have meant rewriting web browsers. Clear boundaries between layers are what allow one part of the system to be replaced independently.
Do all four layers exist on every device?
On a device running applications — a laptop, a phone, a server — yes, all four are present. Intermediate network equipment implements only the layers it needs: a switch works at the link layer, a router at the link and internet layers. This is efficient, because a router has no reason to understand HTTP in order to forward a packet towards its destination.
For Socratic GCSE Computer Science tutoring — networks, protocols and beyond — visit aitutors.me.