4.1.7 - The four-layer TCP-IP model

The stack and its direction

The TCP/IP model is a layered model for exchanging data over networks. Each layer has a particular responsibility and offers a service to the layer above it. This division means an application does not need to control every network interface or work out every step of delivery itself.

From top to bottom, the layers are:

At a sender, data moves down the stack: application -> transport -> internet -> link. At a receiver, it moves up the stack in the reverse order: link -> internet -> transport -> application.

The layers are not four separate routes. They are four stages of processing applied to the same transmission.

Application and transport

The application layer is where network services meet user-facing software. It creates or interprets the application data and applies the rules needed by that kind of service. For example, if a learner sends the text Meeting at 15:30, this layer treats it as the message to be transferred and passes it down. It does not choose a route or produce a radio signal.

The transport layer provides communication between applications on the two end devices. In Pearson's GCSE model, it splits a large item of data into manageable packets for transmission. TCP more precisely calls its pieces segments, but packet is the useful general exam term here.

For a transmission using TCP, the transport layer can:

• add sequence numbers so pieces can be put back into their original order;
• use port numbers to direct data to the correct application service;
• add and check a checksum to detect corruption;
• use acknowledgements to notice missing data and arrange retransmission;
• reassemble correctly received pieces before passing the data to the application layer.

Suppose a message is split into packets numbered 1, 2 and 3. If they arrive as 3, 1, 2, their arrival order does not become the message order. The receiver uses the sequence numbers to restore 1, 2, 3. If packet 2 does not arrive, TCP can identify the gap and have that data retransmitted.

Pearson describes a transmitted packet as having three broad parts:

Exact fields depend on the protocol. The high-value idea is that the payload is the useful data being carried, while headers and footers help the layers deliver and check it.

This distinction lets you explain both a packet arriving out of order and a packet failing to arrive without assigning either problem to IP routing.

Internet and link

The internet layer handles delivery across interconnected networks. It places the source and destination IP addresses in an IP header. The destination IP identifies the final host, while the internet layer and routers use that information to forward the packet towards an appropriate next hop.

IP does not promise that packets will arrive, remain uncorrupted or arrive in order. When reliable delivery is required, TCP at the transport layer supplies the checks, sequencing and retransmission. Saying that the internet layer "makes sure every packet arrives" assigns the job to the wrong layer.

The link layer handles the current connection to a directly connected device, such as the next router. It receives the IP packet from the internet layer and places it in a frame suitable for the local link. Depending on the link technology, it can:

• add source and destination physical addresses, such as MAC addresses;
• add a link header and error-checking footer/trailer around the IP packet;
• use the network interface hardware to turn the frame's bits into suitable electrical, light or radio signals;
• receive those signals at the other end of the link and pass the contained IP packet upward.

Here is the key address distinction:

Imagine a laptop sending to a server beyond its local network. The destination IP names the remote server. The first frame's destination MAC address can name the laptop's router, because that router is the next device on the journey. At the router, the old link frame can be removed and a new one made for the next link.