Communication and Networking Technologies

🕸️ 2.02 Network topologies

Any transmission between two devices needs five things:

1. A sender
2. A receiver
3. A transmission medium to carry it
4. The message itself
5. a protocol: the agreed rules both ends follow

Which way can data flow?

Who receives the message?

• Broadcast: one sender, every device on the network receives it.
• Multicast: one sender, a chosen group receives it.
• Unicast: one sender, one receiver.

➖ Point-to-point and bus Figures 2.01–2.02

A point-to-point network is one dedicated link between two end-systems. It can be simplex, half duplex or full duplex. A WAN link between two sites is built this way.

In a bus topology every end-system taps into one shared cable:

• it is a multipoint connection, and a message sent by one is a broadcast that every other end-system receives
• only one end-system can transmit at a time
• it is quite resilient: if one end-system fails, the others carry on (but a break in the main cable stops it)

⭐ Mesh and star Figures 2.03–2.04

In a mesh every end-system has a direct link to every other, so every link can be full duplex and messages never queue for a shared cable. For end-systems it is unrealistic: with n devices you need n(n − 1)/2 cables. It is used where reliability really matters, such as between switches in a WAN and between routers on the Internet.

In a star each end-system has its own link to a central device. Messages go through it, so the central device must not fail. This is the usual layout of a LAN, and the central device is usually the one that connects the LAN to the Internet.

Hybrid networks

A hybrid network joins two or more networks with different topologies, using a special device to connect them. A common example is adding a WiFi network to an existing wired LAN: a wireless access point plugs into the LAN's switch.

🧵 2.03 Transmission media: cables

There are three types of cable: twisted pair, coaxial and fibre-optic. To choose between them, think about:

• Cost
• bandwidth: how much data per second
• attenuation: how quickly the signal weakens with distance, which is worse at high frequencies
• interference: noise picked up from other signals and electrical equipment
• how often repeaters are needed to boost the signal

Table 2.01, rebuilt: comparing the three cable types.

Each type comes in different grades and standards, so these are general trends. Fibre is the usual choice for new installations, but replacing working copper is not always worth the cost.

Twisted pair is the cable of the telephone network and of most high-speed LANs. Twisting each pair of wires together cancels out much of the interference.

Coaxial cable is used for cable TV and has been used for metropolitan area networks (MANs). Fibre-optic cable, usually bundled as in (b), carries data over long distances: between cities, countries and across oceans.

📡 Wireless media Figure 2.07

All three are electromagnetic waves; the only difference between them is their frequency (or wavelength).

Radio 3 kHz–3 GHz · Microwave 3–300 GHz · Infrared 300 GHz–400 THz

Infrared cannot get through walls, so it only works inside one room (think TV remote). For most purposes microwave is the best compromise.

Cable or wireless?

• People call cable guided and wireless unguided, but that is misleading: a microwave beam can be aimed very precisely.
• Wireless frequencies are regulated; you cannot just use any frequency you like.
• Laying a cable across land needs the landowner's permission.
• For global links the choice is between fibre on the sea bed and satellites.
• Wireless suffers more from interference, but usually needs fewer repeaters.
• A mobile phone can only use wireless; at home you can choose either.

🛰️ Satellites Figure 2.08

Satellites carry signals where cables cannot go. Their orbits avoid the two Van Allen belts, zones of charged particles that would damage them.

The big drawback is delay: the signal has a long way to go. GPS and Internet access in remote places still need satellites, but undersea fibre has taken over most long-distance traffic.

🔌 2.04 LAN hardware

Wired LANs

The first LANs used coaxial cable in a bus; now twisted pair and fibre are normal. On a bus, each end-system plugs into a socket on the cable, and each end of the cable needs a terminator so signals are not reflected back. Twisted-pair cables end in RJ-45 connectors. A star needs more, and longer, cables because every end-system has its own run to the centre.

• A long cable weakens the signal (attenuation), so a repeater is fitted to boost it.
• A bus can be split into segments joined by a bridge. The bridge stores the addresses of the end-systems on each side, so it only passes on messages meant for the other segment.
• Every end-system needs a network interface card (NIC). Each NIC has a unique physical address (more in Chapter 17.05).
• In a star the central device is a hub, a switch or a router. Today it is most likely a switch.
• The servers for a LAN are attached like any other end-system.

Wireless LANs

A wireless LAN (WLAN) is usually called WiFi; the standard is IEEE 802.11 and it uses radio frequencies. A wireless access point (WAP) is wired to the LAN, and each device has a wireless network interface card (WNIC).

🔁 2.05 Ethernet

Ethernet is the most common technology for wired LANs. It is standardised as IEEE 802.3, and there have been five generations:

The original, now legacy, Ethernet was a bus, or a star with a hub in the middle. Either way, every message was a broadcast on a shared medium, so two end-systems could transmit at once and their signals would collide.

CSMA/CD: carrier sense multiple access with collision detection

1. Before sending, an end-system checks the voltage on the medium (carrier sense).
2. If there is a voltage, someone else is transmitting, so it waits a random time and checks again.
3. If the medium is idle, it starts to transmit.
4. While transmitting, it keeps checking for a collision.
5. If there is no collision, it carries on to the end.
6. If there is a collision, it stops, sends a jamming signal so everyone knows, waits a random time and tries again.

Modern Ethernet is switched. Each end-system has its own full-duplex link to a switch, and the switch buffers frames and sends each one only to its destination. Nothing shares a medium, so there are no collisions and CSMA/CD is no longer needed.

🌍 2.06 The Internet's infrastructure

The Internet is the biggest internetwork in existence. Nobody designed it as a whole; it grew, one connection at a time.

Internet service providers

You connect through an Internet service provider (ISP). ISPs form a hierarchy:

• Access ISPs connect homes and businesses.
• They connect to middle-tier (regional) ISPs…
• …which connect to tier-1 ISPs, the global backbone carriers.
• ISPs at the same level can link at an Internet exchange point (IXP) instead of paying a higher tier to carry their traffic.
• Big content providers run their own networks and connect directly to these levels.

Routers

The backbone is a mesh, and at each node sits a router. When data arrives, the router chooses the best next link towards its destination.

The public switched telephone network (PSTN)

• The old telephone network (sometimes called POTS, plain old telephone service) was built for voice. A modem converted computer data into sound, giving slow dial-up access, shared and only when needed.
• An organisation could instead lease a dedicated line, permanently connected at a guaranteed speed. Leased lines are how WANs (and MANs) were built.
• PSTNs have since upgraded their main lines to fibre and digital technology. They now act as ISPs themselves, offering broadband and WiFi hotspots.

The cell phone network

Mobile phone companies are ISPs too. A phone talks by radio to a cell tower, and the mobile network passes the data on to the Internet.

🕸️ 2.07 Applications that use the Internet

The World Wide Web is not the Internet

People say "the web" and "the Internet" as if they were the same. They are not. The Internet is the internetwork: cables, routers and protocols. The World Wide Web (WWW) is a distributed application that runs on it: a huge collection of websites made of web pages. What makes a web page special is that it contains hyperlinks that jump straight to other pages.

Cloud computing

Cloud computing means providing computing services, usually over the Internet. An organisation can set up a private cloud in one of three ways:

1. it builds and runs the cloud itself, on its own site, on a private network
2. it pays a third party to build and run an on-site system on a private network
3. it pays a third party to build and run a system it reaches over the Internet

The alternative is a public cloud, built, owned and run by a cloud service provider and open to individuals and organisations.

Cloud services are the familiar ones from file servers and application servers, reached through a browser from any device, anywhere. The difference is scale: they run on mainframes or server farms. The service is one of:

• infrastructure provision: storage and processing power you set up yourself
• platform provision: a ready environment to develop and test software
• software provision: finished applications

🎬 Bit streaming Figure 2.09

Media such as video and music can be sent as a stream. A plain byte stream would be far too big, so the file is compressed and sent as a bit stream. There are two kinds:

• On-demand: the file already exists on a server. The user asks for it, the data goes into a buffer, and a media player plays it from there.
• Real-time (live): the event is being recorded as you watch, and the stream is sent out through a content provider's servers.

• The server sends data into the buffer faster than the player removes it.
• The player watches two marks. At the high-water mark it sends control data telling the server to stop (or slow down). At the low-water mark it tells the server to start again.
• The buffer must never fill up (data would be lost) or run empty (the video freezes).
• The user's connection must be fast enough for the bit rate. Servers keep several versions at different compression levels so a slower connection can still play.

🔢 2.08 IP addressing

The Internet runs on the TCP/IP protocol suite (Chapter 17). Every device needs an IP address. The version in general use is IPv4, which uses 32 bits: that allows 2³² addresses. Since 2¹⁰ ≈ 1000, that is roughly 4 × 1000 × 1000 × 1000: about four billion.

With about seven billion people, and roughly half of them online, four billion would just about do if each user needed one address. It is not that simple.

netID and hostID

The original scheme was hierarchical. Some bits give the network (netID) and the rest give the device on it (hostID). Routers only look at the netID to get the data to the right network; the hostID is only examined on arrival. (Strictly, a "host" is a network interface: a router with several interfaces needs several addresses.)

Classes

Addresses were allocated in classes. There were five; these are the first three:

Huge organisations got Class A. The trouble came when LANs of PCs became common: there were too few Class B netIDs, but a Class C network had too few hostIDs.

Dotted decimal notation

The hardware uses the 32-bit binary code. For people, each byte is written as its denary value, 0–255, with dots between: 10000000 00001100 00000010 00011110 is written 128.12.2.30.

🧩 Classless inter-domain routing (CIDR) Figure 2.10

CIDR keeps the idea of a netID and a hostID but lets the split point move. An 8-bit suffix says how many bits are the netID. With suffix 21, the netID is 21 bits and 11 bits are left for hostIDs: 2¹¹ = 2048 hosts. No bits are needed to show a class any more, but old Class A, B and C addresses still work with suffixes 8, 16 and 24.

🏢 Sub-netting Figures 2.11–2.12

Sub-netting gives the hostID its own structure. Take an organisation of about 150 workstations, with six department LANs and a head-office LAN.

Under the original scheme each LAN would need its own Class C netID, each pointing at that LAN's gateway (a device that connects networks using different protocols, and acts as, or connects to, a router). That is 7 × 256 = 1792 addresses for 150 machines: 1642 wasted, and nobody else can use them.

With sub-netting the organisation gets one Class C netID, say 194.10.9.0 to 194.10.9.255. Every address points to one router. The router reads the 8-bit hostID: the top 3 bits pick the LAN and the other 5 bits pick the workstation. For example:

• 00001110: workstation 14 on LAN 0, the head office
• 01110000: workstation 16 on LAN 3

Only 106 of the 256 addresses are unused, which leaves room to grow, and six netIDs stay free for other organisations.

🔒 Network address translation (NAT) Figure 2.13

NAT breaks the rule that every IP address must be unique. Large organisations want private networks (intranets) that use the same protocols as the Internet. Isolation used to be the point, for security, but now they want Internet access as well.

The NAT box has one IP address that is visible on the Internet, used for sending and receiving. Inside, addresses must come from three ranges set aside for private networks (you do not need to learn them):

Any number of private networks can use the same addresses at the same time, and neither the Internet nor other private networks know. Software in the NAT box examines every incoming and outgoing transmission and can run a security check before passing incoming data to the right internal address. The loose arrows from the internal router show that the network inside can take any shape.

Static and dynamic addresses

An ISP has a large but limited set of hostIDs, often fewer than its number of customers. Luckily not everyone is online at once. So the ISP normally gives a user a dynamic address: it can change it when it likes, and it goes back into the pool when the user disconnects. A static address never changes, and costs extra.

6️⃣ IPv6

Combining these IPv4 tricks keeps the Internet working, but not for ever. The replacement, IPv6, uses 128-bit addresses: 2¹²⁸ of them, which allows much richer structure. They are written in colon hexadecimal: eight 16-bit groups of four hex digits. Two shortcuts are allowed:

• leading zeros in a group can be dropped: 00A1 → A1
• one run of all-zero groups can be replaced by :: (only once per address)
• an IPv4 address can be written inside IPv6, as in ::192.0.2.46

🔎 2.09 Domain names

Nobody wants to type a dotted decimal address to reach a web page or send an email. The domain name service (DNS, also called the domain name system) was invented in 1983 to give hosts readable names and to look up the IP address for any name.

• The database is spread over a huge number of domain name servers across the Internet, arranged in a hierarchy.
• Powerful root servers sit at the top. They are replicated: several copies of all their data are always kept.
• The name space is split into non-overlapping zones. Each zone has a primary name server holding its database, and secondary servers copy from it.
• There are over 250 top-level domains: generic ones such as .com, .edu and .gov, and country ones such as .uk, .nl and .th.
• A domain name is part of a URL (uniform resource locator), which identifies a web page, or of an email address. A domain is named by the path upwards from it: eng.cisco.com is the eng subdomain of the cisco domain in the com top-level domain.

Name resolution

Looking up a name to find its IP address is name resolution. A query has three possible outcomes:

1. The domain is controlled by the server that was asked: it returns an authoritative, correct address.
2. It is not, but the address is in the server's cache of recent lookups: that is returned, though it might be out of date.
3. The domain is remote: the query goes to a root server, which gives the address of the name server for the top-level domain; that server gives the next one down, and so on until a server can give an authoritative address.

📌 Summary

• Files can be shared using client-server or peer-to-peer networking.
• A LAN is most often a star.
• The main media are copper cable (twisted pair, coaxial), fibre-optic cable and wireless (radio, microwave, infrared).
• Choose a medium by bandwidth, attenuation, interference and the need for repeaters.
• CSMA/CD has been used to detect and deal with collisions on a shared medium.
• The Internet is the largest internetwork; ISPs provide access to it.
• Its infrastructure relies on PSTNs and cell phone companies.
• The World Wide Web is a distributed application that runs on the Internet.
• Addressing today is IPv4; IPv6 is the coming replacement.
• The DNS turns a domain name into an IP address.

🃏 Key-term flashcards

Tap a card to flip it.

✅ Quick quiz

📝 Exam-style questions

Write your answer first, then open the model answer. These are my mark-scheme-style points, not official mark schemes.