Evolution of Networking and Data Communication

A computer network is a collection of two or more autonomous computing devices connected by a communication medium so that they can exchange data and share resources.

Two words in that definition carry marks. Autonomous means no device controls another — each one can work on its own. A keyboard plugged into a CPU is not a network, because the keyboard is not an independent computer. Communication medium means there has to be a real path between them: copper cable, optical fibre, radio or satellite.

Vocabulary you will use for the rest of this unit:

• Node — any device connected to the network: a computer, a printer, a phone, a CCTV camera, a router. Every node needs an address on that network. Nodes that are end devices, and so start or finish a conversation, are also called hosts; intermediate nodes such as routers and switches carry other people’s traffic rather than originating it.
• Server — a node that provides a resource or service: files, printing, a website, a database.
• Client — a node that requests that service. Server and client are roles, not machine types; the same computer can be a server for printing and a client for the internet at the same moment.

Why organisations build networks

• Resource sharing — one printer, one scanner, one internet connection, one licensed copy of software, used by many machines.
• Communication — email, chat, file transfer, video conferencing between people who are not in the same room or the same city.
• Centralised data and backup — student records sit on one server, get backed up once, and everybody reads the same current copy instead of forty different versions living on forty pen drives.
• Reliability — if one machine dies, the data is still on the server and the work continues from another machine.
• Lower cost — many ordinary machines sharing one expensive resource is far cheaper than giving every machine its own.

What a network costs you. Networks are not a free win, and the board does ask for limitations. You pay for cabling, switches and somebody to administer them. Malware or one careless click on a single machine can reach every machine. Private data becomes reachable from outside the room. And if the server or the one internet link fails, everybody stops working at the same instant — a single point of failure that standalone computers never had.

Worked example

A school in Ranchi is setting up a lab of 40 computers. Assume a laser printer costs Rs 12,000, a network switch plus cabling for the room costs Rs 25,000, and one broadband connection costs Rs 1,200 per month.

Counting printing alone, the one-time saving is Rs 4,80,000 − (Rs 12,000 + Rs 25,000) = Rs 4,43,000, and the monthly internet bill falls from Rs 48,000 to Rs 1,200. That is the resource-sharing argument in rupees. The trade-off the school knowingly accepts: the day the switch fails, all 40 machines lose printing and internet together.

The internet was not designed in one go. It arrived in three stages, and each stage solved one specific problem. Learn the stages as problem-and-solution, not as a list of years, and the dates become easy to hold.

Stage 1 — ARPANET (from 1969): can distant computers talk at all?

ARPANET stands for Advanced Research Projects Agency NETwork. In the 1960s computers were rare and hugely expensive, and researchers funded by ARPA (an agency of the US Department of Defense) wanted to use each other’s machines from a distance. ARPANET became operational in 1969, and by December of that year it connected four nodes: UCLA, Stanford Research Institute (SRI), the University of California Santa Barbara, and the University of Utah. The first message attempt, on 29 October 1969, went from UCLA to SRI: the operator began typing LOGIN and the SRI host crashed after the letters LO.

The genuinely new idea in ARPANET was packet switching — chopping a message into small packets that travel independently instead of reserving one telephone-style path for the whole conversation. This is why a damaged link does not kill the whole network: the next packets simply take another route. ARPANET itself was formally decommissioned in 1990, long after its ideas had spread.

Stage 2 — a common protocol (1 January 1983): can different networks talk to each other?

ARPANET originally ran a protocol called NCP (Network Control Protocol), which worked only within ARPANET. Vinton Cerf and Robert Kahn designed TCP/IP (Transmission Control Protocol / Internet Protocol) to let networks built by different people, on different hardware, interconnect. On 1 January 1983 ARPANET switched over to TCP/IP. That switch is the real birthday of internetworking — a network of networks, which is exactly what the word "internet" means.

Stage 3 — NSFNET (from the mid-1980s): can everyone else join?

NSFNET stands for National Science Foundation NETwork. The US National Science Foundation built a backbone linking its supercomputer centres, and regional academic networks plugged into it. Its capacity climbed steadily: the backbone started at 56 kbps, was upgraded to T1 (1.544 Mbps) in 1988, and to T3 (44.736 Mbps) in 1991 — roughly 29 times the T1 rate.

NSFNET had an Acceptable Use Policy that allowed only research and education traffic; running a business over it was forbidden. That single policy line is why the early internet felt academic. When the restriction was relaxed and commercial ISPs (Internet Service Providers) began carrying traffic, the NSFNET backbone was retired in April 1995 and the internet became a public, commercial utility.

The Internet, and why it is not the Web

The Internet is the global network of interconnected networks, all speaking TCP/IP, with no single owner. The World Wide Web (WWW) is just one service running on it: documents linked by hyperlinks and fetched over HTTP. It was proposed by Tim Berners-Lee at CERN in 1989, had a working browser and server by the end of 1990, and CERN placed the software in the public domain in 1993, which is why nobody pays a licence fee to make a website. Email, video calls and file transfer all use the Internet without using the Web at all. Treat "the Internet and the WWW are the same thing" as a wrong answer.

India’s own timeline

• ERNET (Education and Research Network) began in 1986, connecting Indian academic and research institutions — India’s own NSFNET-style, research-only network.
• VSNL (Videsh Sanchar Nigam Limited) opened public internet access in India on 15 August 1995. Before that date, being online in India essentially meant being in a university.

Milestones at a glance

Worked example

Arrange these events in chronological order and give a one-line reason for each: (a) NSFNET backbone retired, (b) ARPANET’s first four nodes go live, (c) ARPANET switches to TCP/IP, (d) WWW proposed at CERN, (e) public internet access opens in India.

Answer: (b) 1969 — the first packet-switched network. (c) 1 January 1983 — a common protocol makes internetworking possible. (d) 1989 — Berners-Lee’s proposal at CERN, a service built on a network that already existed. (a) April 1995 — the research backbone shuts down as commercial ISPs take over. (e) 15 August 1995 — VSNL opens public access in India, a few months later the same year. Note the logic of the order: the network came first, the common protocol second, and the Web only after both existed. That is why (d) can never come before (c).

Data communication is the exchange of data between two devices through some transmission medium. "Communication" here does not just mean the data left the sender; it means the receiver got it and understood it. That second half is why a protocol is a component and not an optional accessory.

Five components must be present. The board asks for all five, one line each.

Communication media in one pass

• Guided (wired) — the signal is confined inside something. Twisted pair: cheap, easy, used for ordinary LAN cabling, limited distance. Coaxial cable: better shielded against interference, used for cable TV and older LANs. Optical fibre: carries light through glass, so electrical noise cannot corrupt it; highest capacity, longest distance, highest cost.
• Unguided (wireless) — the signal travels through open space. Radio waves: Wi-Fi and FM, pass through walls, omnidirectional. Microwaves: need line of sight between towers. Infrared: very short range and blocked by walls, which is why a TV remote works only inside the room. Satellite: enormous coverage including remote areas, but the long path introduces noticeable delay.

Protocol: the component that is not a physical object

A protocol is an agreed set of rules that governs how data is formatted, how it is grouped, in what order it is sent, how fast, how errors are detected, and what to do when something is lost. Both ends must use the same protocol. Common examples: TCP/IP (the base suite of the internet), HTTP and HTTPS (web pages), FTP (file transfer), SMTP and POP3 (sending and fetching email), PPP (point-to-point links).

Think of two people on a phone call. Having a working line is not enough. They must share a language — Hindi, Tamil, English — and a convention for taking turns. A sender and receiver need exactly those two things, written down precisely. Without an agreed protocol the bits still arrive; they simply do not mean anything, like a fax machine’s screech coming out of a voice call.

One more point that catches students: sender and receiver are roles that swap constantly. When Meera’s phone sends back the blue tick acknowledgement, her phone is the sender and Aarav’s is the receiver.

Worked example

A cold storage unit in Chennai has a temperature sensor that reports every minute to a monitoring server in Bengaluru over a mobile data connection. Identify the five components of data communication.

Now change one thing: the server replies "reading received". The roles reverse — the server becomes the sender and the sensor becomes the receiver — while the medium and the protocol stay exactly the same. That is the cleanest way to show you understand which components are fixed and which are role-based.

This is the pair the board tests hardest in this chapter, and the pair students confuse most. Fix the idea in one line: bandwidth is what the link could do; data transfer rate is what it actually did.

Bandwidth = capacity

Bandwidth is measured two ways, depending on whether you are talking about the medium or about a digital link.

• Frequency sense (analog). Bandwidth is the range of frequencies the medium can carry: highest frequency minus lowest frequency. The unit is the hertz (Hz), with multiples kHz (10³ Hz), MHz (10⁶ Hz) and GHz (10⁹ Hz). A telephone voice channel passes roughly 300 Hz to 3400 Hz, so its bandwidth is about 3100 Hz — plenty for speech, which is exactly why music sounds thin and flat over a landline.
• Digital sense. Bandwidth is the maximum number of bits the link can carry per second, quoted in bps and its multiples. When your plan says "100 Mbps", that is a bandwidth figure — a ceiling, a promise about the best case.

Data transfer rate = what you actually got

Data transfer rate is the amount of data actually moved per unit of time, and its unit is always bits per second: bps, Kbps, Mbps, Gbps, Tbps. You measure it after the fact, by dividing.

The road analogy is worth remembering because it also explains the gap. Bandwidth is the number of lanes on the Mumbai–Pune expressway. Data transfer rate is how many vehicles actually reached Pune in that hour. The lanes set the ceiling; a toll queue, a breakdown or a slow truck decides the reality. The data transfer rate can never be more than the bandwidth, and in practice it is less.

Why the real rate falls short

• Protocol overhead — every packet carries header bits (addresses, sequence numbers, error checks) that are not part of your file, but still occupy the link.
• Sharing — the ISP’s line into your colony is shared. Between 8 pm and 11 pm everyone is streaming.
• Congestion — packets queue at routers when more traffic arrives than a link can forward.
• Noise, attenuation and distance — a long or weak link corrupts bits, and corrupted data must be re-sent.
• Weak Wi-Fi — the last few metres between your laptop and the router are often the slowest part of the whole journey.
• The far end — a busy server can be slower than your connection.

Units, and the two traps

Trap 1 — bits versus bytes. ISPs advertise in bits (Mbps); your download window shows bytes (MB/s). Divide by 8. A perfect 100 Mbps link delivers at most 100 ÷ 8 = 12.5 MB per second, which is why a "100 Mbps" plan looks like "12 MB/s" on screen and people wrongly complain they were cheated.

Trap 2 — which power to use. In telecommunication, data-rate prefixes are powers of ten: 1 Kbps = 1000 bps. Some textbooks instead use 1 Kbps = 1024 bps. Both appear in answer keys. Write the convention you used in one line at the start of your answer; the method still earns the marks. Everything below uses powers of ten.

Worked example 1 — the exact form the board asks

Meera’s broadband plan is advertised as 100 Mbps. She downloads a 3 GB file and it takes exactly 5 minutes. (a) What data transfer rate did she get? (b) What was the ideal time? (c) Why the difference?

1. Convert the file size to bits: 3 GB = 3 × 10⁹ bytes = 3 × 10⁹ × 8 = 24,000,000,000 bits.
2. Convert time to seconds: 5 minutes = 300 s.
3. Data transfer rate = 24,000,000,000 ÷ 300 = 80,000,000 bps = 80 Mbps.
4. Ideal time at full bandwidth = 24,000,000,000 ÷ 100,000,000 = 240 s = 4 minutes.
5. She achieved 80 ÷ 100 = 80% of the advertised capacity. The missing fifth went into packet headers, a shared line at peak hour and Wi-Fi losses in the last few metres. Note the language: 100 Mbps is bandwidth, 80 Mbps is data transfer rate, and the second can never exceed the first.

Worked example 2 — download time

How long, at best, does a 2 GB movie take on a 50 Mbps link? Bits = 2 × 10⁹ × 8 = 16,000,000,000. Time = 16,000,000,000 ÷ 50,000,000 = 320 s = 5 minutes 20 seconds. "At best" matters: the real time will be longer, because 50 Mbps is the ceiling.

Worked example 3 — how much data a stream eats

An online class runs a video call at 3 Mbps for 40 minutes. Data used = 3,000,000 × (40 × 60) = 7,200,000,000 bits = 7,200,000,000 ÷ 8 = 900,000,000 bytes = 900 MB. That single 40-minute class swallows almost an entire 1 GB daily pack — the practical reason data limits and video quality settings exist.

Part A — the IP address

An IP address (Internet Protocol address) is a unique numeric address assigned to every device on a TCP/IP network so that data can be delivered to the correct machine. It does for data what a postal address does for a letter: without it the packet has nowhere to go.

IPv4 addresses are 32 bits long, written as four decimal numbers called octets, separated by dots. Each octet is 8 bits, so its value runs from 0 to 255. For example 192.168.10.15. The total number of possible IPv4 addresses is 2³² = 4,294,967,296, about 4.3 billion.

Every IP address splits into a network part (which network the device is on) and a host part (which device on that network) — the same logic as a PIN code followed by a house number.

Two distinctions worth knowing:

• Static vs dynamic. A static IP is configured once and stays fixed — used for servers and network printers, so people can always find them. A dynamic IP is handed out afresh each time a device connects. That is why your phone’s IP address at school is not the one it had at home.
• Public vs private. Public addresses are unique across the whole internet. Private addresses (ranges beginning 10., 172.16. to 172.31., and 192.168.) are reused inside homes, schools and offices and are not routed on the internet. The address 127.0.0.1 is special: it always means "this machine itself", the loopback address.

Worked example — valid or not?

IPv6, in exactly as much detail as the syllabus needs. Roughly 4.3 billion addresses turned out to be too few for the phones, laptops, cameras and appliances now connected, so IPv6 uses 128 bits, written as eight groups of four hexadecimal digits separated by colons:

2001:0db8:0000:0000:0000:ff00:0042:8329

which can be written more compactly as 2001:db8::ff00:42:8329. That gives 2¹²⁸ = 340,282,366,920,938,463,463,374,607,431,768,211,456 possible addresses. For the exam the useful contrast is short: IPv4 is 32-bit dotted decimal, IPv6 is 128-bit hexadecimal with colons, and IPv6 exists because IPv4 addresses ran out.

Part B — switching techniques

You cannot run a separate wire from every device to every other device; for even a thousand devices that would be 1000 × 999 ÷ 2 = 499,500 links. So devices share links, and at every junction something has to decide where the data goes next. That decision is switching.

Circuit switching

A complete physical path is found and reserved end to end before any data moves, and it stays reserved for the entire conversation — even during the seconds when nobody is speaking. It works in three phases:

1. Circuit establishment (setup) — a path is located and reserved from sender to receiver.
2. Data transfer — everything travels along that one path, arriving in the order it was sent.
3. Circuit disconnection (teardown) — the path is released for other users.

The classic example is the traditional landline telephone network. Advantages: once connected you have guaranteed capacity, a constant delay, in-order delivery, and no address attached to every small piece of data. Disadvantages: you wait through the setup phase before anything can be sent; the reserved capacity is wasted during silence; when all circuits are taken new callers are simply refused; and if one link on the reserved path breaks, the entire call drops and must be set up again.

Packet switching

The message is broken into small pieces called packets. Each packet carries a header plus a slice of the data, and the header holds the source IP address, the destination IP address and a sequence number. Nothing is reserved. Each router reads the destination address and forwards the packet on whichever link is best at that moment (store-and-forward), so different packets of the same message may travel by different routes and arrive out of order or slightly late. The destination uses the sequence numbers to reassemble the message correctly and asks for any missing packet again.

The classic example is the Internet itself. Advantages: links are shared efficiently, since a gap in your traffic is instantly used by someone else’s; there is no setup delay; and if a link fails, later packets simply take another route. Disadvantages: delay varies from packet to packet (jitter), packets can be lost or reordered, every packet carries header overhead, and the receiver has to do reassembly work.

Why a phone call and a video stream fail differently. A landline call is circuit-switched: either you get a reserved path and the quality holds steady for the whole call, or you hear "all lines are busy" and get no call at all. A video stream is packet-switched: it is never refused, but on a busy evening its packets queue behind everyone else’s, so it buffers, freezes and drops to a blurry resolution. Circuit switching rations by refusing new users; packet switching rations by degrading every user a little. That single sentence answers most scenario questions on this topic.

Worked example

A 5 MB file is sent across a packet-switched network with 1000 bytes of data in each packet.

1. How many packets? 5 MB = 5,000,000 bytes; 5,000,000 ÷ 1000 = 5000 packets (plus a header on each one, which is why the bits on the wire exceed 5 MB).
2. Packet 3721 arrives after packet 3900. Is the file corrupted? No. Packets travel independently and may take different routes, so out-of-order arrival is normal. The receiver holds them and reassembles them using the sequence numbers in the headers.
3. Packet 4010 never arrives. What happens? Only that one packet is requested again — not the whole 5 MB. Compare this with circuit switching, where a broken link costs you the entire conversation.
4. Which technique suits a landline voice call, and which suits downloading a PDF? The voice call suits circuit switching: it is continuous, at a steady rate, and sensitive to varying delay, so a reserved path is worth its cost. The PDF download suits packet switching: it is bursty and can tolerate small delays, so reserving a path that would sit idle between bursts would waste capacity.