Matheion

MA-NET1 · Module 1

Internet edge — hosts, protocols, and access

Topics: End systems · Protocols · Access networks · Physical media

Overview

Session 1 — The Internet edge and its hosts

Meet Mira, a learner who is building a small phone app that shows the daily outlook from a made-up weather service called Skylark Forecast. Every time her app refreshes, bytes leave her phone, travel through a chain of machines, and come back as a forecast. Before we follow those bytes through the middle of the network, we start where Mira stands: at the edge, with her phone, her home connection, and the cables and radio waves that carry her data away.

By the end you should be able to: describe the Internet as a network of networks; say what a host (end system) is and why agreed-upon protocols matter; tell apart the common ways a home or office reaches an access ISP; and match twisted pair, coaxial cable, fiber, and radio links to the situations they suit.

Scroll Concepts → Examples → Quiz. Plan on roughly 55–65 minutes, including time to try the reasoning steps on paper. No special software is needed; a pen and the willingness to pause and re-read a paragraph are enough.

Tip: if a paragraph feels dense, pause and say it back in your own words before moving on; that small habit saves time later.

Concepts

Core ideas — the Internet edge, hosts, protocols, and the links that reach them

Open a weather app on your phone and a forecast appears in a second or two. It feels like the phone simply "knows". In reality, a small conversation has taken place across a very large machine, and this course is about that machine. We will go top-down: begin with the part you can see and touch (your phone, your home connection), and work gradually toward the hidden middle. Throughout, we will follow a learner named Mira, who is writing a phone app that shows the daily outlook from a made-up weather product called Skylark Forecast. The forecast data lives behind the name api.skylark.example, and the app's pictures and icons are served from edge.northline.example, a nearby delivery point run by a made-up network called Northline Relay. Every name, company, and number here is invented for practice.


What do we actually mean by "the Internet"?

There are two good ways to answer, and it helps to hold both at once.

The first is the nuts-and-bolts view: the Internet is an enormous collection of connected machines and the links between them. Some of those machines are the laptops, phones, game consoles, thermostats, and servers that people care about. Others are the relaying machines in the middle whose only job is to pass data along. The links joining them are made of copper, glass, or radio.

The second is the service view: the Internet is an infrastructure that gives applications a way to talk to each other. Mira's app is an application. The Skylark Forecast server program is an application. The Internet's service to them is, roughly, "give me a bunch of bytes, tell me who they are for, and I will do my best to get them there." Notice the words do my best. We will return to them, because they are important.

A short phrase captures the whole shape: the Internet is a network of networks. No single organisation owns it. Mira's home connection belongs to one company, the long-distance links belong to others, and the server she is talking to sits in yet another organisation's network. They cooperate because they all follow the same agreed rules.


End systems, also called hosts

The machines that sit at the outer rim of the Internet and run applications are called end systems, or hosts. The two words mean the same thing in this course. They are "end" systems because data starts or finishes its journey there; nothing beyond them is expected to forward the data further.

Hosts come in two broad flavours, and it is worth meeting both. A client is a host that usually asks: Mira's phone is a client when it requests a forecast. A server is a host that usually answers: the machine behind api.skylark.example waits for requests and replies to them. These words describe roles, not hardware. A powerful laptop can act as a server for a few friends, and a server can briefly act as a client of another server. What matters is who starts the conversation.

Everything that is not a host, such as the relaying machines in the middle of the network, belongs to the network core, which is the subject of the next session. For this session, picture the Internet as a cloud in the middle with hosts around its edge, and our attention is on the edge: the hosts themselves and the access networks that connect them to the cloud.

Common confusion: A host does not have to be a big computer. A doorbell camera, a smart plug, and a car's entertainment system all send and receive data over the Internet, so all of them count as hosts. Likewise, "server" does not mean "a machine in a cold room"; it means "the side that waits for requests".


Why we need protocols

Imagine Mira walking up to a stranger and saying "forecast". The stranger might hand her a newspaper, answer in another language, or walk away. For two people to cooperate, they need a shared pattern: who speaks first, what form the question takes, what counts as a valid answer, and what happens if no answer comes.

A protocol is exactly that shared pattern, written down precisely enough that two machines built by strangers can follow it. A protocol defines three kinds of things. It defines the format of messages: which fields appear, in what order, with what meaning. It defines the order in which messages are exchanged: for example, "the client sends a request, then the server sends a reply". And it defines the actions each side takes when a message is sent, received, or fails to arrive: for example, "if no reply comes in five seconds, ask again".

Think of a polite phone call. You say hello; the other person says hello back; you say who you are; they say who they are; you ask your question. If the line goes quiet, you say "are you still there?" Each of those moves is part of an unwritten human protocol. Computers need theirs written precisely, because a machine will not "figure out what you meant".

Mira's app and the Skylark server speak an application protocol of this kind. Underneath it sit other protocols that deal with reliability, addressing, and the physical transmission of signals. Because everybody follows the same protocols, a phone made by one company can talk to a server built by another. That is the real reason protocols matter: they are the agreements that let the network of networks function as a single thing. The next four sessions are largely a tour of these agreements.


Access networks: how a host reaches the first relay

An access network is whatever connects an end system to the first relaying machine on its way into the rest of the Internet, sometimes called the edge router. The organisation that operates the connection is the access ISP (Internet service provider). For Mira, the access ISP is BayLink Access. Further in, BayLink hands her traffic to bigger networks such as SpineNet, which we will meet next session.

Access networks come in several families. Do not memorise product names; instead, keep a mental map of what physical idea each one uses.

Home access over existing wires. Some homes already have a telephone line or a television cable running into them. An ISP can reuse that existing wire to carry data as well. A phone-line style connection is a dedicated path between the home and the ISP's nearby equipment: no neighbour shares your stretch of wire, though the ISP's own equipment further in is shared. A television-cable style connection, by contrast, is often a shared medium in the neighbourhood: the cable passes several homes, and they divide its capacity among themselves. This is why a shared connection can feel slower in the evening when neighbours are all online. Both of these styles are often asymmetric: the rate for downloading to your home is usually larger than the rate for uploading from it, because most households receive far more than they send.

Home access over fiber. Some ISPs run optical fiber all the way to, or very near, the home. Fiber offers far higher capacity than copper wire over longer distances, so these connections can be fast in both directions. Often one fiber from the ISP's building is split to serve a group of homes, so the capacity near the street is still shared at some level.

Enterprise and campus access. A company, school, or office typically runs its own local network, usually built from Ethernet cabling and wireless access points, and connects that whole local network to an ISP through one or a few high-capacity links. Each desk gets a short cable (or a wireless signal) to a local switch; the switches are joined together, and a single path leads out to the ISP. Local networks like this are called LANs (local area networks).

Wireless access. Instead of a cable to every device, a base station sends radio signals that devices pick up. There are two broad styles. A local wireless network, the familiar home or café Wi-Fi, reaches only a few tens of metres: the access point is wired to the ISP, and the radio hop covers just the last few steps. A wide-area wireless network, such as the cellular network, uses base stations (towers) covering kilometres, so a phone can stay connected while its owner travels. In both styles, the radio link is shared: every device that talks to the same base station divides up the same air.

Whatever the technology, an access network is described by two numbers worth remembering. One is its transmission rate, measured in bits per second (bps), with the familiar prefixes: kbps (thousand), Mbps (million), Gbps (billion). The other is whether the link is dedicated or shared, because sharing explains why the rate you actually experience can be lower than the rate printed on the contract.

Common confusion: Network speeds are quoted in bits per second, but file sizes are quoted in bytes, and one byte is eight bits. A connection of 24 Mbps moves at most about 3 megabytes each second, not 24. When you compare numbers, always write the unit beside them before you do any arithmetic.


A quick word about "bandwidth"

You will hear bandwidth used in two slightly different senses. In everyday talk it means "how fast the link is", the transmission rate in bps. In signal processing it means a range of frequencies a medium can carry. This course uses the everyday sense: bandwidth is the link's rate. If you see a sentence like "this link has a bandwidth of 40 Mbps", read it as "this link can push bits onto the medium at 40 million bits per second".


Physical media: what the bits actually travel through

A bit is just a zero or a one. To carry bits between two machines, something physical has to change: a voltage in a wire, a pulse of light in a glass strand, a wave in the air. The material that carries the change is the physical medium. Media fall into two families.

Guided media keep the signal inside a solid path. Think of water flowing through a pipe: the pipe tells the water where to go. Unguided media let the signal spread through open space, like a sound that fills a room: anyone in range can pick it up.


Guided medium 1: twisted-pair copper

The cheapest and most common wire is twisted pair: two insulated copper wires twisted around each other into a spiral. The twisting is not decoration. Every wire acts a little like an antenna, picking up electrical noise from nearby machines. Twisting the pair makes the noise affect both wires nearly equally, so the receiver, which looks at the difference between them, can cancel most of the interference.

You will meet two families. UTP (unshielded twisted pair) has no metal wrapping around the pairs; it is light, flexible, and cheap, and it is what most office and home cabling uses. STP (shielded twisted pair) wraps the pairs in a metal foil or braid that blocks more outside noise; it costs more and is stiffer to install, so it appears where interference is heavy, such as near large motors. Twisted-pair cables are also sold in numbered categories, where a higher category is built to a tighter standard and carries higher rates over short runs. For our purposes, the idea is simply: twisted pair is cheap, easy, and good for runs of up to roughly a hundred metres at high rates.

Guided medium 2: coaxial cable

Coaxial cable also uses copper, but arranged differently. A central copper conductor sits inside an insulating layer, which sits inside a cylindrical metal shield. Both conductors share the same centre line (that is where the word co-axial comes from). The shield keeps interference out and lets the cable carry a wide range of frequencies, so several signals can share one cable. This is the reason television-style access networks use it: one cable can carry many channels plus data. Coaxial cable is a shared medium in the neighbourhood, as described earlier.

Guided medium 3: optical fiber

Optical fiber is a very thin strand of glass that carries pulses of light. A light source at one end switches on and off to represent ones and zeros, and the glass guides that light to the other end. Because light is not affected by electrical noise, fiber has very low error rates, can carry extremely high rates, and can run for tens of kilometres before the signal needs boosting. These properties make fiber the backbone of long-distance networks, including the undersea ones. Its drawbacks are cost and delicacy: the equipment at the ends is more expensive, and the strand must be handled carefully.


Unguided media: radio

Radio carries a signal as an electromagnetic wave through the air. There is no cable to lay, which is a huge advantage, and the signal can reach devices that move around. The drawbacks are the flip side of the same fact: the signal weakens with distance, bounces off walls, gets blocked by obstacles, suffers interference from other transmitters, and can be overheard by anyone nearby, so the content often needs to be scrambled (encrypted) for privacy.

It is useful to group radio links into a few families by how far they reach, without worrying about exact frequencies. Short-range radio covers a room or an apartment and is used for local wireless networks and for connecting small gadgets. Wide-area radio covers kilometres from a tower and serves mobile phones. Long-haul radio includes satellite links, where a dish on the ground beams a signal up to a satellite and the satellite sends it down to another place. A satellite in a very high orbit covers a huge area, which is excellent for remote regions, but the signal must travel tens of thousands of kilometres up and down, so each trip adds a noticeable delay. Satellites in much lower orbits shorten that trip but need many of them working together.

Common confusion: "Wireless" rarely means entirely wireless. When Mira uses wifi at home, only the last few metres are radio. Behind the access point there is almost always a cable, usually fiber or copper, carrying her data to BayLink Access.


Choosing a medium: the trade-offs in one place

Every medium is a compromise among four things: cost, capacity, distance (how far before the signal fades), and interference (how easily noise corrupts it). Twisted pair wins on cost and ease, so it fills offices and homes. Coaxial cable offers more capacity and shielding at moderate cost. Fiber wins on capacity and distance, at higher cost. Radio wins on convenience and mobility but loses on reliability and privacy. No medium is best in the abstract; the right choice depends on the job.


Pulling the session together

We started with an almost philosophical question, "what is the Internet?", and arrived at a concrete picture. The Internet is a network of networks, and a useful way to organise the picture is edge and core. At the edge sit hosts, clients and servers running applications that talk to each other by following protocols: shared agreements on message format, order, and actions. Hosts reach the Internet through access networks, whether wired over existing copper or cable, over fiber, through a local network in a building, or wirelessly over a short-range or wide-area radio link, and each is operated by an access ISP. Underneath everything, bits ride on physical media, which are either guided (twisted pair, coaxial cable, fiber) or unguided (radio, including satellite).

To summarise as a short list:

  • The Internet is a network of networks that gives applications a "best effort" delivery service.
  • A host (end system) runs applications; a client usually asks and a server usually answers.
  • A protocol fixes message format, message order, and actions on events.
  • Access networks vary in medium and in whether the capacity is dedicated or shared; rates are in bits per second.
  • Twisted pair (UTP, STP), coaxial cable, and fiber are guided; radio and satellite are unguided.
  • Choosing a medium means balancing cost, capacity, distance, and interference.

In the next session we leave the edge and step into the middle of the cloud, where relaying machines called routers decide, packet by packet, where Mira's request goes next.

Examples

Worked examples — hosts, protocols, access, and media around Mira's flat

Four short stories follow. In each one, read the Scene, try to answer on your own, and only then read the working. The reasoning matters more than the final label.


Example 1 — Which devices in the flat are hosts?

Scene. Mira lives in a small flat served by BayLink Access. She wants to know how many of her devices count as hosts (end systems) for the purposes of this course. The devices in the flat are:

  1. Her phone, which runs the Skylark Forecast app.
  2. A wall-mounted tablet that shows the same forecast in the hallway.
  3. A small box on the shelf, provided by BayLink, with one cable to the street and a radio antenna that serves the flat.
  4. A smart bulb that checks in every night with its maker's server to see if a software update exists.
  5. A passive wall socket for the Ethernet cable, which is only a hole with copper behind it.

Working, step by step.

Step 1: state the test. A host is a machine that runs applications and is a source or final destination of data. A device that merely passes data along for others is part of the access or core network, not a host.

Step 2: apply it one item at a time.

  • The phone runs an application and starts conversations: it is a host (and, in the forecast conversation, a client).
  • The tablet also runs an application that requests forecasts: a host.
  • The shelf box has a cable to the street and an antenna to the flat. Its job is to carry other devices' data to and from BayLink. That makes it part of the access network. (The box has a management page of its own, so it also acts as a host in a minor way, but its main role is forwarding.)
  • The smart bulb contacts a server and runs a tiny program to check for updates: a host, even though it has no screen.
  • The wall socket carries no application and no processing: it is just cabling, part of the physical medium.

Interpretation. Three devices in the list are clearly hosts: phone, tablet, bulb. The box is access equipment, and the socket is only a medium. The result surprises many beginners, because the bulb is "smaller" than the box and still counts as a host.

Takeaway. Ask "does this device run an application that creates or consumes data?" If yes, it is a host. If it mostly passes data on for others, it belongs to the network around the hosts.


Example 2 — Reading a protocol conversation

Scene. Mira's app asks api.skylark.example for tomorrow's outlook. To get a feel for what a protocol is, Mira writes the conversation as a list of messages. The format below is invented purely for practice and is not a real protocol.

1. Phone  -> Server : HELLO app=mira-forecast v=3
2. Server -> Phone  : HELLO-OK session=77
3. Phone  -> Server : GET outlook city=harbor-town day=tomorrow session=77
4. Server -> Phone  : OUTLOOK 14C light-rain session=77
5. Phone  -> Server : BYE session=77

What to look for. Mira checks the three ingredients that make something a protocol.

Format. Every line has a verb (HELLO, GET, OUTLOOK, BYE) followed by named fields such as city= and session=. Anyone building a client or server can look up the verbs and fields and know what each means.

Order. The conversation cannot begin with line 3. The server only recognises a session after it has seen HELLO and replied with HELLO-OK. The order is part of the rules.

Actions. Suppose line 2 never arrives. The protocol should say what the phone does: perhaps "wait 3 seconds, then repeat line 1, up to three times, then report failure to the user". Our invented protocol so far has not said this, which is a gap. Real protocols fill such gaps explicitly.

Interpretation. Notice what the protocol does not say. It does not say how the bytes travel, what cables are used, or how the machines in the middle find each other. A protocol at one level of the system deals only with its own concerns. This habit of separating concerns is the seed of the layered stack we meet in Session 4.

Takeaway. A protocol is a precise agreement about message format, message order, and actions on events (including the unhappy ones). If any of the three is missing, two machines built by strangers cannot reliably cooperate.


Example 3 — Picking an access technology

Scene. A friend of Mira asks for advice about three places that each need to get online. For each, Mira should choose among: a phone-line style connection, a shared television-cable style connection, a fiber connection, a local wired network plus an ISP link, or a wide-area wireless service.

Place A. A small bakery on a street where optical fiber already runs to the front door. The owner uploads photos of cakes all day and runs a card machine.

Place B. A farmhouse at the end of a track, six kilometres from the nearest telephone exchange, with no fiber or television cable nearby, but with a cell tower visible on the ridge.

Place C. A 40-person design office, where each desk needs a reliable network socket.

Working.

Place A. The bakery uploads a lot, so a symmetric connection matters. Fiber is fast in both directions, and it is already at the door. Choose fiber.

Place B. There is no wired option, and the phone-line style connection loses speed quickly with distance, which a six-kilometre run would make painful. The tower on the ridge suggests wide-area wireless: the signal covers kilometres and needs no new cable. Choose wide-area wireless, with the caveat that the rate will be shared with other users of the tower.

Place C. For 40 desks, give each a short twisted-pair cable (a local wired network, built with Ethernet) to a switch, and connect the whole office to an ISP over one shared high-capacity link, probably fiber. Add a few wireless access points for visitors' phones. Choose a local wired network plus an ISP link.

Interpretation. Notice that the choice depended on three questions: Is a cable already there? How far is it to the ISP's equipment? Is traffic mostly down, mostly up, or both? Nobody asked which technology was "best".

Takeaway. Match the access technology to the building's situation, not to a ranking. And remember that any shared option can feel slower than its printed rate when many neighbours are active at once.


Example 4 — Choosing a medium for each hop

Scene. Northline Relay (Mira's imaginary edge delivery network) is installing three new connections. For each one, decide whether twisted pair, coaxial cable, fiber, or radio is the best fit, and explain.

Link 1. A 15-metre patch between a server and a switch in the same rack room. Cost is a concern; the room is quiet.

Link 2. A 9-kilometre run between two buildings across a field. The run will be above ground and a storm-prone area is known for lightning.

Link 3. A temporary link to a food stall at a street market, for two days, across a road where digging is not allowed.

Working, link by link.

Link 1. Fifteen metres is a very short distance, and twisted pair handles that easily at high rates. The room has little interference, so unshielded twisted pair (UTP) is enough. Cost is lowest. Choose UTP.

Link 2. Nine kilometres is far beyond what twisted pair can cover at a good rate. Fiber can cross that distance without help. A second benefit is relevant here: glass does not conduct electricity, so a lightning strike near the cable will not send a damaging surge into the buildings the way a metal cable could. Choose fiber.

Link 3. The link is temporary, and digging is not allowed, so a cable cannot be laid. A short-range or wide-area radio link solves the problem. The trade-off is a weaker and less predictable signal, plus the need to scramble the content because anyone nearby can receive the signal. Choose radio.

Interpretation. Each decision came from one dominating factor: cost for Link 1, distance and lightning for Link 2, and "no cable allowed" for Link 3. When one factor dominates, choosing is quick. When two factors compete, such as cost versus distance, we weigh them.

Takeaway. There is no universal best medium. Name the constraint that dominates, and the choice usually follows.

Practice check — Internet edge, hosts, and media

Multiple-choice check from the Matheion app lesson. Pick an answer for each question, then submit.

  1. 1.Mira's hallway tablet requests a forecast from `api.skylark.example`. In this exchange, which statement about roles is most accurate?

  2. 2.Which of the following is NOT one of the three things a protocol specifies?

  3. 3.A neighbourhood is served by a television-style coaxial cable that passes forty homes. On a quiet morning Mira's downloads are quick, but in the evening they slow down although her plan has not changed. What is the best explanation?

  4. 4.A plan advertises a downstream rate of 32 Mbps. At the very best, about how many megabytes of a file can arrive each second?

  5. 5.A shielded twisted pair (STP) cable is chosen instead of UTP for a run beside a factory's large motors. Why?

  6. 6.Two office buildings are 7 km apart, and the route crosses an area where lightning strikes are common. Which medium best fits the link?

  7. 7.Mira says: 'At my flat everything is wireless, so my data never touches a cable.' What is the flaw in this statement?