Protocol Layers and Their Service Models · Computer Networking Interactive

The Internet is organised as five layers, and each one wraps the layer above in its own header before passing it down.

Words you will meet

Layer — one level of the design, offering a service to the level above it.
Service model — what a layer promises to the layer above.
Protocol stack — the five layers taken together.
Message — the application layer’s packet.
Segment — the transport layer’s packet.
Datagram — the network layer’s packet.
Frame — the link layer’s packet.
Header — the bytes a layer puts in front of what it was given.
Payload — what a layer was given, treated as opaque.
Encapsulation — the act of wrapping the payload in a header.

Why this matters

This is the most important page in the chapter. The five layers are the book’s skeleton: chapter 2 is the application layer, chapter 3 the transport layer, chapters 4 and 5 the network layer, chapter 6 the link layer. Everything after this page lives inside one of these boxes.

The four words — message, segment, datagram, frame — are also the vocabulary the rest of the book is written in. Learn them here and the later chapters read smoothly. Skip them and every chapter will be slightly confusing.

Why layer anything?

The Internet has numerous applications and protocols, various kinds of host, packet switches and link media. Given that complexity, can a network architecture be organised at all?

Yes — and the book reaches for an analogy first.

Everyday picture — the book’s airline

Describe the airline system. It has ticketing agents, baggage checkers, gate staff, pilots, aeroplanes and a worldwide system for routing aircraft. One way to describe it is as the series of actions you take: buy a ticket, check your bags, go to the gate, board. The plane takes off and is routed to its destination. After it lands you leave at the gate and claim your bags. If the trip was bad, you complain to a ticket agent.

Now look at that list horizontally instead. There is a ticketing function at each end. There is a baggage function, but only for people who are already ticketed. There is a gate function, but only for people who are already ticketed and baggage-checked. Above the runway there is a routing function.

So the airline has layers.

Where the picture breaks: you experience all the airline’s layers yourself, in order. A packet does not experience anything, and the intermediate machines it passes implement only the lower layers — an aeroplane does not check your ticket again in mid-air.

Each layer, combined with the layers below it, implements some service. At the ticketing layer and below, a person is transferred counter to counter. At the baggage layer and below, a person and their bags are transferred check to claim — but only for someone already ticketed. At the gate layer, gate to gate. At the takeoff/landing layer, runway to runway.

Each layer provides its service in two ways: by doing certain things itself, and by using the services of the layer directly below it. Organising protocols this way is called protocol layering , and what a layer promises upward is its service model .

What layering buys

A layered architecture lets you discuss one well-defined part of a large system. That simplification gives modularity: it becomes far easier to change how a service is implemented.

A layer may change how it works freely. The rest of the system is unaffected, so long as it still offers the same service upward and uses the same services from below.

In plain words

Changing how a service is implemented is very different from changing what the service is.

If the airline decided to board passengers by height, nothing else in the system would need to change. The gate layer still loads and unloads people; it just does it differently. That is the whole value of a service model.

What layering costs

Layering is not universally admired. Some researchers and engineers are vehemently opposed to it, and the book names two real drawbacks.

One layer may duplicate what a lower layer already does. Many protocol stacks provide error recovery both per-link and again end to end.
A layer may need information that exists only in another layer — a timestamp, for example. Reaching for it violates the separation the layers were supposed to give.

The five layers

The protocols of the various layers, taken together, are the protocol stack . The Internet’s has five layers.

The five layers: what each one adds, and who runs it

	Applicationlayer 5	Transportlayer 4	Networklayer 3	Linklayer 2	Physicallayer 1
Its job
Packet is called
Example protocols
Usually implemented in
Which devices run it
How far its header travels

Cells marked ⓘ have a reason behind them — click to read it.

Read the “Packet is called” row. Those four words are the vocabulary this whole book depends on.

Notice that a protocol layer may be implemented in software, in hardware, or in both. Application- and transport-layer protocols are almost always software in the hosts. Because the physical and link layers handle communication over one specific link, they typically live in the network interface card for that link. The network layer is often a mixture.

The airline’s functions were spread across airports and control centres. In the same way, a layer-n protocol is distributed among the hosts, packet switches and other components that make up the network.

The four words

One packet, four names — learn these now

The book uses a different word at each layer, and it is never careless about which. Neither is this site.

Layer	Call it a…	Its header says
application	message	What the application actually wants to send.
transport	segment	Which program on the far host, plus error-detection bits.
network	datagram	The addresses of the two hosts.
link	frame	How to cross this one link — and it changes at every hop.

M is the application's message. Ht, Hn and Hl are the transport, network and link headers. Each layer treats everything handed down to it as one opaque payload and puts its own header in front.

Do not treat these as synonyms

The book uses a different word at each layer and is never careless about which. Neither is this site. Saying “packet” when you mean “frame” will cost you in chapter 6, where the two are genuinely different objects handled by different devices.

Use packet only when the layer is deliberately unspecified.

Encapsulation

Now watch the four names appear and disappear.

At the sending host, an application-layer message is passed down to the transport layer. The transport layer appends its own header — $H_t$ — which the receiving host’s transport layer will read. Message plus $H_t$ is a segment .

We say that the segment encapsulates the message.

The added information might tell the receiving transport layer which application the message belongs to, and might include error-detection bits.

The segment goes down to the network layer, which adds $H_n$ — carrying the source and destination host addresses — creating a datagram . The datagram goes down to the link layer, which adds $H_l$ , creating a frame .

In plain words

At every layer a packet has just two kinds of field: a header and a payload . The payload is normally the entire packet from the layer above.

Each layer refuses to look inside its payload. That refusal is what makes the layers independent.

Figure 1.24 — a message crossing a switch and a router

1 / 19

Lin’s laptop · Application — The application has a message to send. At this point it is just data — nothing has been added to it yet.

Scrub through it. Watch two things: the greyed-out layers at the switch and the router, and what happens to the link header each time a hop ends.

Read all 19 steps as text

Lin’s laptop · Application [GET /index.html] — The application has a message to send. At this point it is just data — nothing has been added to it yet.
Lin’s laptop · Transport [Ht | GET /index.html] — The transport layer puts its header Ht in front. Everything below now treats the message as an opaque payload. This is a segment.
Lin’s laptop · Network [Hn | Ht | GET /index.html] — The network layer adds Hn, carrying the addresses of the two hosts. This is a datagram.
Lin’s laptop · Link [Hl | Hn | Ht | GET /index.html] — The link layer adds Hl, which describes how to cross this one link. This is a frame.
Lin’s laptop · Physical [Hl | Hn | Ht | GET /index.html] — The physical layer puts the individual bits of the frame onto the medium.
link-layer switch · Physical [Hl | Hn | Ht | GET /index.html] — The bits arrive and are reassembled into a frame.
link-layer switch · Link [Hn | Ht | GET /index.html] — The switch reads the link header and decides which port to use. It stops here: it does not implement the network layer, so it cannot see the IP addresses inside at all.
link-layer switch · Link [Hl | Hn | Ht | GET /index.html] — A brand-new link header is added for the next link. It may be a completely different link technology — Ethernet here, WiFi next — and the header changes to match.
link-layer switch · Physical [Hl | Hn | Ht | GET /index.html] — The frame goes out onto the next link as bits.
router · Physical [Hl | Hn | Ht | GET /index.html] — The bits arrive and are reassembled into a frame.
router · Link [Hn | Ht | GET /index.html] — The old link header is removed and discarded. It described the link the frame just crossed, and that link is behind us now.
router · Network [Ht | GET /index.html] — The router reads the destination address in Hn and chooses the outgoing link. It stops here: it does not implement the transport layer, so Ht means nothing to it.
router · Link [Hl | Ht | GET /index.html] — A brand-new link header is added for the next link. It may be a completely different link technology — Ethernet here, WiFi next — and the header changes to match.
router · Physical [Hl | Ht | GET /index.html] — The frame goes out onto the next link as bits.
web server · Physical [Hl | Hn | Ht | GET /index.html] — The bits arrive and are reassembled into a frame.
web server · Link [Hn | Ht | GET /index.html] — The old link header is removed and discarded. It described the link the frame just crossed, and that link is behind us now.
web server · Network [Ht | GET /index.html] — The network layer removes Hn and passes the segment up.
web server · Transport [GET /index.html] — The transport layer removes Ht, checks it, and works out which application the message belongs to.
web server · Application [GET /index.html] — The message arrives at the application, byte for byte what was sent.

Not every device has every layer

Routers and link-layer switches are both packet switches, and both organise their software and hardware into layers. But they do not implement all of them — only the bottom ones.

A link-layer switch implements layers 1 and 2.
A router implements layers 1 to 3.
A host implements all five.

So a router can run IP (Internet Protocol) , a layer-3 protocol, and a link-layer switch cannot. A switch does not recognise IP addresses at all; it recognises layer-2 addresses such as Ethernet addresses.

Why hosts do the most work

Hosts implement all five layers. Routers implement three, switches two.

This is consistent with a deliberate design choice: the Internet architecture puts much of its complexity at the edges. Keep it in mind. It explains why TCP (Transmission Control Protocol) can be improved without touching a single router, and it is a theme the book returns to repeatedly.

Everyday picture — the book’s interoffice memo

Alice, in one branch office, sends a memo to Bob in another, through the public postal service.

The memo is the application-layer message. Alice puts it in an interoffice envelope with Bob’s name and department on the front. That interoffice envelope is the segment: it has header information — Bob’s name and department — and it encapsulates the memo.

The sending mailroom puts that envelope inside a second envelope suitable for the postal service, and writes the postal addresses of the two branch offices on it. That outer envelope is the datagram.

The postal service delivers it. The receiving mailroom begins de-encapsulation: it extracts the interoffice envelope and forwards it to Bob. Bob opens it and removes the memo.

Where the picture breaks: the postal service handles one envelope from end to end. A datagram’s outer link-layer envelope is thrown away and rebuilt at every single hop, as the animation above shows. Nothing in the postal system works that way.

Encapsulation can be more complex than this. A large message may be divided into several segments, and each of those might be divided into several datagrams. At the receiving end, a segment must then be reconstructed from its constituent datagrams.

Check yourself

Check yourself — §1.5

0 of 7 answered

1.What is a layer's service model?
2.Name the five layers of the Internet protocol stack, from top to bottom.
3.predictStep the animation to the link-layer switch. Why can it not read the destination IP address?
Look at which layers are drawn as "not implemented" in that column.
4.predictStep the animation through the router. What happens to the link header Hl?
Compare the packet drawn at the router's link layer on the way in and on the way out.
5.Which layers do hosts, routers and link-layer switches implement?
6.A transport-layer packet is called a segment. What does it contain?
7.The book gives two drawbacks of layering. Which is one of them?

What to remember

The Internet stack has five layers — application, transport, network, link, physical — and their packets are called message, segment, datagram, frame, never used loosely.
Encapsulation: every layer treats what it is handed as an opaque payload and puts its own header in front.
Hosts implement all five layers, routers three, switches two. The complexity lives at the edges, and the link header is rebuilt at every hop while the transport header is written once and read once.