7 new

1. Network Devices (Hub, Switch, Router)

Analogy:

• Hub = shouting in a room (everyone hears — bad, insecure)
• Switch = whispering directly to the right person in the room
• Router = a post office directing mail between cities

Key Rule: Different subnets MUST be connected through a Router. A switch alone cannot route between subnets.

Interface

• Interface = the connection point between a host/router and a physical link
• Routers have multiple interfaces (e.g., eth0, eth1, eth2), each with its own IP address
• Hosts typically have 1–2 interfaces (wired eth0, wireless wlan0)
• Analogy: Interface = a door of a building. Each door connects to a different street (network). A router has many doors; your MacBook usually just has one or two.

Network Architectures in Practice

Home Network:

Cable/Fiber Modem → Modem Router (Wi-Fi) → Devices (PC, Laptop, Phone, Printer)

Home Network (ADSL):

Phone Jack → Phone Cable → ADSL Splitter → Modem Router
(Phone can still share the same line)

Office Network:

Internet → Firewall → Router → Switches → PCs, IP Phones, Printers, Servers
                              → WiFi Router → Laptops, Smartphones

The Firewall is placed between the internet and the internal router in office setups — it filters traffic before it reaches the internal network.

2. Network Layer: Two Key Functions

Data Plane (Forwarding)

• Operates per-router in hardware → nanosecond speed
• Looks up the destination IP in the local forwarding table → picks an output port
• Determined by: Arriving packet header field → match in local forwarding table → route to output port

Control Plane (Routing) — Two Approaches

1. Traditional Per-Router Routing

• Individual routing algorithm runs on each router
• Routers exchange info with each other (e.g., link states, distance vectors) to build their own forwarding tables
• Protocols: OSPF (within AS), BGP (between AS)
• Think of each taxi driver independently knowing the city map

2. Software-Defined Networking (SDN)

• A remote controller (server) computes routes centrally and pushes them out
• Each router runs a Local Control Agent (CA) that communicates with the remote controller
• The controller installs a Flow Table (richer than a plain forwarding table — can match on ports, MACs, IPs) into each router
• Think of separating the GPS app (control plane) from the car's engine (data plane)
• More centralized, flexible, and programmable than per-router routing

3. IP Datagram Format

The structure of every IP packet — 32 bits wide per row.

 0                   1                   2                   3
 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|Version| IHL |Type of Service|          Total Length           |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|         Identification        |Flags|      Fragment Offset    |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|  Time to Live |    Protocol   |         Header Checksum       |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|                       Source Address                          |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|                    Destination Address                        |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|                    Options (if any)          |    Padding     |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|                          Data (Payload)                       |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

$\boxed{\text{TCP + IP overhead} = 20 \text{ bytes (TCP header)} + 20 \text{ bytes (IP header)} = 40 \text{ bytes minimum}}$

TTL in real life: When you do traceroute, it works by sending packets with TTL=1, TTL=2, TTL=3... Each router that drops the packet (TTL expired) sends back an ICMP "Time Exceeded" message — that's how you see each hop's IP.

4. Network-Layer Service Model

Internet "Best Effort"

• No guarantees on:
  1. Successful delivery to destination
  2. Timing or order of delivery
  3. Bandwidth available to the flow

Analogy: Internet is like dropping a letter in a public mailbox with no tracking — it usually arrives, but there's no promise of when or in what order.
Why does the Internet still work well? TCP (transport layer) adds reliability on top of best-effort IP.

5. IP Address

Structure

• 32-bit identifier, one per interface (not per host — a router has many)
• Written in dotted-decimal notation: 4 octets (0–255) separated by dots

$\boxed{192.168.32.152 = \underbrace{11000000}_{192}.\underbrace{10101000}_{168}.\underbrace{00100000}_{32}.\underbrace{10011000}_{152}}$

Three Notations for IPv4

Bit Value Reference (each octet = 8 bits)

Quick tip: Octet range is always 0–255. Sum the values wherever the bit is 1.
Example: 11000000 = 128 + 64 = 192 ✓

IP Address Analogy to Physical Address

6. Subnet & Subnet Mask

Why Subnets?

• Total IPv4 space: $2^{32} \approx 4.3$ billion IPs
• One flat range causes: broadcast floods, management nightmare, security issues
• Solution: Split IP range into multiple subnets
• Broadcasts are scoped within a subnet — they do NOT pass through routers

Analogy: Subnetting = dividing a city into neighbourhoods. ARP broadcast only floods your neighbourhood (subnet), not the whole city.

IP Address = Network Portion + Host Portion

$\boxed{\text{IP Address} = \underbrace{\text{Network Portion (n bits)}}_{\text{Prefix}} + \underbrace{\text{Host Portion (32-n bits)}}_{\text{Suffix}}}$

Subnet Mask: A 32-bit number — 1s for the network part, 0s for the host part.

Example: 192.168.10.10/24

• /24 → 24 bits network, 8 bits host → $2^8 = 256$ total IPs

3 Special Addresses in Every Subnet

$\boxed{\text{Usable Hosts} = 2^{\text{host bits}} - 2}$

Subtract 2: one for Network Address + one for Broadcast Address.

Example: 192.168.5.10/24

• Network: 192.168.5.0 | HostMin: 192.168.5.1 | HostMax: 192.168.5.254 | Broadcast: 192.168.5.255
• Usable hosts: $2^8 - 2 = \mathbf{254}$

7. Classful IP Addressing

4 main techniques to split IP range: (1) Classful A/B/C ← this lecture, (2) Public/Private, (3) CIDR, (4) VLSM

IP Address Classes

$\boxed{\text{Total Hosts per Network} = 2^{\text{host bits}} - 2}$

How to identify class from first octet:

• A: Starts with 0_______ → 0–127 → but 1–126 usable (127 = loopback)
• B: Starts with 10______ → 128–191
• C: Starts with 110_____ → 192–223

Analogy: Class A = countries (few, enormous). Class B = states/provinces. Class C = neighbourhoods (many, small). Class D = group chat (multicast). Class E = reserved for lab experiments.

Class-by-Class Detail

Class A (/8): $\text{Host bits} = 24 \Rightarrow 2^{24} - 2 = 16{,}777{,}214 \text{ hosts}$

• Example: 10.2.1.1/8 → Network: 10.0.0.0 | Broadcast: 10.255.255.255 | Mask: 255.0.0.0

Class B (/16): $\text{Prefix: }10xxxxxx \Rightarrow 128 \text{ to } 191 \quad ; \quad 2^{16} - 2 = 65{,}534 \text{ hosts}$

• Example: 172.16.1.5/16 → Network: 172.16.0.0 | Broadcast: 172.16.255.255 | Mask: 255.255.0.0
• Example: 128.2.0.0/16 → HostMin: 128.2.0.1 | HostMax: 128.2.255.254 | Broadcast: 128.2.255.255

Class C (/24): $2^8 - 2 = 254 \text{ hosts}$

• Example: 192.168.1.1/24 → Network: 192.168.1.0 | Broadcast: 192.168.1.255 | Mask: 255.255.255.0

Reserved / Special Addresses

8. Public vs Private IP Addresses

Private IP Ranges (RFC 1918 — memorise these!)

Analogy: Private IPs are like apartment numbers — the same number (e.g., 192.168.1.1) can exist in different buildings. But each building's public IP is globally unique on the internet.

Class B Private Range Detail

Valid private Class B subnets: 172.16.X.X through 172.31.X.X

$\text{Number of private Class B subnets} = 2^{16-12} = 16 \text{ subnets}$

Example: 172.18.0.0/16 → Usable: 172.18.0.1 – 172.18.255.254 | Broadcast: 172.18.255.255

NAT (Network Address Translation)

• NAT converts between Public IP ↔ Private IP
• Your home router does NAT — all devices share one public IP, but have unique private IPs inside
• Example: Your phone (192.168.1.5) → Router NAT → 49.228.234.12 (public) → Internet

9. How to Design a Network (Subnetting Steps)

Design Process

Step 1 — Determine number of required Network IDs:

• One per subnet (each department, area, or segment)
• One per WAN link (router-to-router connection counts as a network too!)
• Ask: How many departments? How many router links?

Step 2 — Determine hosts per subnet:

• One per TCP/IP host (PC, printer, phone, server)
• One per router interface (each interface takes one IP)

Step 3 — Create:

• One subnet mask for the entire network
• A unique subnet ID for each physical segment
• A range of host IDs for each subnet

⚠️ Don't forget the WAN link! The link between two routers is itself a network (usually a /30 — only 2 usable hosts). This is a very common exam mistake.

Subnetting Examples

Example 1: 4 Subnets from Class C

• Network: 192.168.100.0/24 | Need: 4 subnets (100 users, 60 users, 200 users + WAN)

Example 2: Mixed Class B and C

• Areas: 5000 users, 1000 users, 200 users (+1 WAN link = 4 networks total)

Always count the router-to-router WAN link as a separate network!

10. Network Diagram: Logical vs Physical

Instructor note: This is almost certain on the final exam. The MOST COMMON mistake is not labeling interface names on physical diagrams.

Two Types

Key Difference (EXAM CRITICAL ⚠️)

• Logical diagram: Label the IP addresses on the links/clouds between devices
• Physical diagram: Label the interface name (e.g., eth0) AND the IP address on each interface of each device

Physical Diagram Example:

Host A
  |
 (Network A: 192.168.1.0/24)
  |
Router R1
  eth0: 192.168.1.1       ← label BOTH interface name AND IP
  eth1: 10.0.0.1
  |
 (Network B: 10.0.0.0/30) ← WAN link between routers
  |
Router R2
  eth0: 10.0.0.2
  eth1: 192.168.2.1
  |
 (Network C: 192.168.2.0/24)
  |
Host B

11. How Does a Host Get an IP Address?

Two Methods

Always write IP with CIDR prefix (e.g., 192.168.1.5/24) — without it, you don't know which network it belongs to!

How Does a Network Get Its IP Block?

• Private network: Allocated by Network Admin from RFC 1918 private ranges (free)
• Public IP block: Purchased from ISP (ISP gets ranges from IANA/ICANN)

12. DHCP — Dynamic Host Configuration Protocol

Every home router has DHCP enabled by default!

Purpose

• Host dynamically obtains an IP address when it joins the network
• Supports address reuse — IP is leased, returned when device leaves
• Supports mobile users (join/leave dynamically)
• Ensures no IP conflicts — DHCP tracks who has what
• Can renew lease before it expires

DHCP 4-Step Process (DORA)

Mnemonic: DORA — Discover, Offer, Request, ACK

Why are steps 1 & 2 sometimes skipped? If a client already has a previously allocated IP and wants to reuse it, it can skip straight to Request → ACK [RFC 2131].

Why broadcast (not unicast)? The client has no IP yet — it can't address a unicast packet. The server's reply is also broadcast because the client isn't officially on the network yet.

Packet Details (Exam-Level Detail)

DHCP Discover (Client → Network):

src: 0.0.0.0, port 68  →  dest: 255.255.255.255, port 67
yiaddr: 0.0.0.0  |  transaction ID: 654

DHCP Offer (Server → Network):

src: 223.1.2.5, port 67  →  dest: 255.255.255.255, port 68
yiaddr: 223.1.2.4  |  transaction ID: 654  |  lifetime: 3600 sec

DHCP Request (Client → Network):

src: 0.0.0.0, port 68  →  dest: 255.255.255.255, port 67
yiaddr: 223.1.2.4  |  transaction ID: 655  |  lifetime: 3600 sec

DHCP ACK (Server → Network):

src: 223.1.2.5, port 67  →  dest: 255.255.255.255, port 68
yiaddr: 223.1.2.4  |  transaction ID: 655  |  lifetime: 3600 sec

yiaddr = "your IP address" — the field where the server tells the client its assigned IP.
Port 67 = DHCP server, Port 68 = DHCP client — always.

DHCP Returns More Than Just an IP

When DHCP ACK is received, the client gets:

DHCP Encapsulation Stack

DHCP message
  → encapsulated in UDP
    → encapsulated in IP (dest: 255.255.255.255)
      → encapsulated in Ethernet (dest MAC: FF:FF:FF:FF:FF:FF — broadcast)

DHCP in Practice (Home Network)

ISP Modem (10.10.10.26)
       ↓
  Router (192.168.1.1) ← DHCP server lives here
       ↓
  Switch (192.168.1.2)
   ↙       ↓        ↘
Desktop  Game       Network Printer
(DHCP)  Console     (static: 192.168.1.100)
         (DHCP)
         Mobile Phone (DHCP)
         Laptop (DHCP)

The DHCP server is typically co-located in the router, serving all subnets the router is attached to.

Pro tip: You can reserve a fixed IP for a specific MAC address in the router — the device always gets the same IP via DHCP (useful for printers, cameras).

Quick Reference Summary