Skip to content
IT-502 · Computer Networks/Quick Revision Short Notes

Computer Networks (IT-502) - Unit 1 Short Notes

UNIT 1: Computer Networks Fundamentals


1. Introduction to Computer Networks

Definition: A computer network is an interconnection of autonomous computers/l devices (hosts) via communication links (wired/wireless) and switches/routers, enabling resource sharing and communication using standardized protocols.

Key Components:

  • Hosts/End Systems: Sources/destinations of data (computers, servers, IoT devices).

  • Routers/Switches: Forward packets between networks (layer 3) or within a LAN (layer 2).

  • Communication Links: Guided (cables, fiber) or unguided (radio waves) media.

  • Protocols: Rules governing data format, timing, and error handling (e.g., TCP, IP, Ethernet).

Types of Networks:

Type Scale Example Key Feature
PAN Personal (≤10m) Bluetooth, Zigbee Connects personal devices
LAN Local (building/campus) Ethernet, Wi-Fi High speed, single admin
WAN Wide (city/country) Internet, MPLS Long-distance, multiple admins
Internetwork Global The Internet Network of networks

Network Topologies:

Topology Description Advantages Disadvantages
Bus Single backbone cable Simple, cheap Single point of failure, collisions
Star All nodes connect to central hub/switch Easy to manage, fault isolation Hub/switch failure brings down network
Ring Nodes form closed loop Deterministic access (token) Single node failure breaks ring
Mesh Every node connected to every other High reliability, redundancy Expensive, complex cabling
Hybrid Combination (e.g., star-bus) Flexible, scalable Complex design

Switching Techniques:

  • Circuit Switching: Dedicated path established before communication (e.g., PSTN). Guaranteed bandwidth, but inefficient for bursty traffic.

  • Packet Switching:

    • Datagram (connectionless): Each packet routed independently (e.g., IP). No setup, packets may take different paths, possible reordering.

    • Virtual Circuit (connection-oriented): Logical path established first (e.g., ATM, Frame Relay). Packets follow same path, in-order delivery.

  • Message Switching: Store-and-forward of entire messages (obsolete, used in early email systems). High delay, requires large buffers.

[!TIP] Exam Focus: Distinguish circuit vs. packet switching clearly. Virtual circuit is connection-oriented at network layer, unlike TCP's transport-layer connection.


2. Network Architecture Models

2.1 OSI Reference Model

7 Layers (Bottom-Up):

Layer Function Protocols/Devices Service Primitives
7. Application User interface, network services HTTP, SMTP, DNS REQUEST, INDICATION, RESPONSE, CONFIRM
6. Presentation Data translation, encryption, compression SSL/TLS, JPEG, MPEG —
5. Session Dialog control, synchronization NetBIOS, RPC —
4. Transport End-to-end reliability, flow control TCP (conn-oriented), UDP (conn-less) —
3. Network Routing, logical addressing, fragmentation IP, ICMP, ARP, routers —
2. Data Link Framing, MAC addressing, error control Ethernet, PPP, switches, bridges —
1. Physical Bit transmission over media RS-232, Ethernet (PHY), hubs —

Critique & Limitations:

  • Theoretical elegance but complex; some layers (e.g., session) are thin in practice.

  • Protocol stack overhead; not all functions are strictly layered (e.g., some protocols span layers).

  • Timing mismatch: OSI designed in late 1970s; TCP/IP gained market dominance first.

  • No clear distinction between some services (e.g., presentation/session often merged into application layer in TCP/IP).

2.2 TCP/IP Reference Model

4 Layers:

Layer Function Protocols
Application Combines OSI's App, Pres, Sess HTTP, SMTP, DNS, SSH
Transport End-to-end communication TCP (reliable), UDP (unreliable)
Internet Routing, logical addressing, fragmentation IP, ICMP, ARP
Network Interface (Link) Framing, physical transmission Ethernet, Wi-Fi, PPP

Advantages:

  • Practical, simplified model; widely implemented.

  • Protocols independent of underlying hardware.

  • Scalable and robust (designed for internetworking).

Disadvantages:

  • Less rigorous layering; some functions (e.g., routing) span layers.

  • No standard session/presentation layers; handled by applications.

  • Originally lacked explicit congestion control (added later in TCP).

2.3 Comparison: OSI vs. TCP/IP

Aspect OSI Model TCP/IP Model
Layers 7 layers 4 layers
Approach Theoretical, general Practical, specific to Internet
Protocols Protocol-independent Protocol-specific (TCP/IP suite)
Session/Presentation Explicit layers Merged into Application layer
Network Layer Connection-oriented (X.25) & connectionless Primarily connectionless (IP)
Data Link Layer LLC & MAC sublayers No clear sublayer distinction
Adoption Never fully implemented De facto standard
Routing Network layer handles routing Internet layer handles routing

[!TIP] Exam Key: OSI is reference model; TCP/IP is implementation model. OSI's network layer can support both VC & datagram; TCP/IP's internet layer is strictly datagram. TCP provides connection-oriented service on top of connectionless IP.


3. Physical Layer

3.1 Transmission Media

Guided Media:

  • Twisted Pair (UTP/STP): 100 m limit, 10 Mbps–10 Gbps, cheap, susceptible to EMI.

  • Coaxial Cable: 500 m, 10–500 Mbps, better shielding, used in cable TV.

  • Fiber Optic: 100+ km, 10 Mbps–100 Gbps+, low attenuation, immune to EMI, expensive.

Unguided Media:

  • Radio Waves: Omnidirectional, penetrate walls, Wi-Fi, Bluetooth.

  • Microwave: Directional, line-of-sight, satellite, terrestrial links.

  • Infrared: Short-range, line-of-sight, remote controls.

  • Satellite: Geostationary (~36,000 km, 270 ms delay) or LEO.

Broadband Access:

  • DSL: Uses existing phone lines, frequency division (voice low, data high), ADSL: up to 24 Mbps down, 1.4 Mbps up.

  • Cable: Shared coaxial, DOCSIS standards, up to 1 Gbps down.

  • FTTH: Fiber to home, PON (Passive Optical Network), highest speeds (10+ Gbps).

3.2 Signal Transmission Theory

Nyquist Formula (Noiseless Channel):

$$ \text{Max Bit Rate} = 2B \log_2 M \text{ bps} $$

  • $B$ = bandwidth (Hz)

  • $M$ = number of discrete signal levels

Example: $$\displaystyle B = 3 $$ kHz, $$\displaystyle M = 4 $$ (2 bits/level) → $$\displaystyle 2 \times 3000 \times \log_2 4 = 12,000 $$ bps.

Shannon's Theorem (Noisy Channel):

$$ C = B \log_2 (1 + \text{SNR}) \text{ bps} $$

  • $C$ = channel capacity (max reliable data rate)

  • $$\displaystyle \text{SNR} = \frac{S}{N} $$ (signal power/noise power, often in dB: $$\displaystyle \text{SNR}_{\text{dB}} = 10 \log_{10}(\text{SNR}) $$)

Example: $$\displaystyle B = 4 $$ kHz, $$\displaystyle \text{SNR}_{\text{dB}} = 30 $$ dB → $$\displaystyle \text{SNR} = 1000 $$ → $$\displaystyle C = 4000 \log_2(1001) \approx 39,863 $$ bps.

[!TIP] Common Pitfall: Nyquist assumes no noise and gives theoretical max for given $M$. Shannon gives practical max for given SNR. Increasing $M$ in Nyquist increases bit rate but requires better SNR (Shannon limit).

3.3 Multiplexing

  • FDM: Each signal modulated to different frequency band (e.g., radio stations, cable TV).

  • TDM: Time slots assigned to channels. Synchronous TDM: fixed slots (wastes slots if channel idle). Statistical TDM: slots allocated dynamically (higher efficiency, needs buffering).

TDM Frame Size (bits) = sum of all channel bit rates × slot time.

3.4 Line Coding

Convert binary data to physical signals.

Scheme 0 → Signal 1 → Signal Features
NRZ-L High Low No clock recovery, DC balance issue
NRZ-I Transition at start of 1 No transition for 0 Better clock sync than NRZ-L
RZ Pulse to mid-level then zero Pulse to high then zero Self-clocking, but more bandwidth
Manchester Low→High transition mid-bit High→Low transition mid-bit Self-clocking, DC balanced, used in Ethernet
Diff. Manchester Transition at start of every bit; mid-bit transition for 0 No mid-bit transition for 1 Better noise immunity

Example: Binary 10110 in Manchester: 10 01 10 01 10 (assuming start high? Actually: 1=high→low, 0=low→high → 10 01 10 01 10? Let's define: convention varies; typically 1=transition high→low, 0=low→high. So 1→10, 0→01, 1→10, 1→10, 0→01 → 10 01 10 10 01).


4. Data Link Layer

4.1 Functions & Services

  • Framing: Encapsulate packet into frame (header + payload + trailer).

  • Error Control: Detection (CRC) & correction (Hamming).

  • Flow Control: Prevent fast sender overwhelming slow receiver (stop-and-wait, sliding window).

  • Access Control: Multiple stations sharing medium (MAC protocols).

  • Addressing: MAC address (48-bit, burned into NIC). Types:

    • Unicast: Single destination (LSB of first byte = 0).

    • Multicast: Group (LSB of first byte = 1, e.g., IPv4 224.0.0.0–239.255.255.255 maps to 01:00:5E:xx:xx:xx).

    • Broadcast: All stations (FF:FF:FF:FF:FF:FF).

4.2 Error Detection & Correction

Parity Check:

  • Single bit: detect odd number of errors.

  • 2D parity: detect & correct single-bit errors.

Checksum:

  • Internet checksum (ones complement sum of 16-bit words), used in IP/TCP/UDP.

Simple, but weak error detection.

CRC (Cyclic Redundancy Check):

  • Generator polynomial $G(x)$ (e.g., $$\displaystyle x^3 + x + 1 = 1011 $$).

  • Append $r$ zeros ($$\displaystyle r = \deg(G) $$) to data $D(x)$.

  • Divide $$\displaystyle D(x) \cdot x^r $$ by $G(x)$ (mod 2), get remainder $R(x)$.

  • Transmit $$\displaystyle D(x) \cdot x^r + R(x) $$.

  • Receiver divides by $G(x)$; remainder 0 → no error (if $G(x)$ chosen well, detects all 1,2-bit errors, odd errors, burst ≤ $r$ bits).

Example: Data 1101011011 ($$\displaystyle D(x)=x^9+x^8+x^6+x^4+x^3+x^1+1 $$), $$\displaystyle G(x)=10011 $$ ($$\displaystyle r=4 $$). Compute CRC → codeword.

Hamming Code (Single-bit error correction):

  • $k$ data bits, $r$ parity bits: $$\displaystyle 2^r \ge k + r + 1 $$.

  • Parity bits at positions $$\displaystyle 2^i $$.

  • Compute parity over specific data bits.

  • Receiver recalculates, syndrome points to error position.

4.3 Framing & Stuffing

  • Byte Stuffing (PPP): Escape byte 0x7D used to insert 0x7E (flag) or 0x7D in data.

  • Bit Stuffing (HDLC): After 5 consecutive 1s, insert a 0 to avoid flag 01111110.

Example: Data 011110111110111101111110111101111110 with bit stuffing → 011110111111011111011111101111101111110? Actually: insert 0 after every five 1s. Original: 01111 011111 011111 0111111 011111 011111 0? Let's chunk: 01111 0 11111 0 11111 0 111111 0 11111 0 11111 0? Need exact: 011110111110111101111110111101111110 → break: 01111 011111 011111 0111111 011111 011111 0? Actually count: positions: after first 5 ones? Better: scan left to right: 01111 (5 ones? 0 then four 1s? Actually: 0 1111 → 4 ones. Then 0 → no. Next 11111 → five ones → stuff 0. So: 01111 0 11111 0 11111 0 111111 0 11111 0 11111 0? Let's do properly: string: 0 1 1 1 1 0 1 1 1 1 1 0 1 1 1 1 1 0 1 1 1 1 1 1 0 1 1 1 1 1 0 1 1 1 1 1 0. Count consecutive 1s: after first 0: 1111 (4) → no stuff. Then 0. Then 11111 (5) → stuff 0 after. So becomes 11111 0. Then 0. Then 11111 (5) → stuff 0. Then 0. Then 111111 (6) → after 5 ones, stuff 0, then one more 1? Actually: 111111 → after first 5: 11111 0 1. Then 0. Then 11111 → stuff 0. Then 0. Then 11111 → stuff 0. Then 0. So final: 0 1111 0 11111 0 0 11111 0 0 11111 0 1 0 11111 0 0 11111 0 0? Messy. But exam expects clear step.

4.4 Flow & Error Control Protocols

Stop-and-Wait:

  • Sender sends 1 frame, waits for ACK before next.

  • Efficiency = $$\displaystyle \frac{1}{1 + 2a} $$ where $$\displaystyle a = \frac{\text{propagation delay}}{\text{frame transmission time}} $$.

  • Utilization low for long propagation delays (e.g., satellite).

Sliding Window:

  • Sender can have up to $W$ unacknowledged frames.

  • Receiver window size $$\displaystyle W_r $$ (often 1 for Go-Back-N, >1 for Selective Repeat).

Go-Back-N (GBN):

  • Sender window $$\displaystyle W_s $$, receiver window $$\displaystyle W_r = 1 $$.

  • If frame $n$ lost/damaged, receiver discards all subsequent frames and sends ACK for $n-1$ (duplicate ACK).

  • Sender times out, retransmits $n$ and all following.

  • Efficiency ≈ $$\displaystyle \frac{W_s}{1 + 2a} $$ (if $$\displaystyle W_s \le 1 + 2a $$).

  • Link Utilization = $$\displaystyle \frac{\text{useful time}}{\text{total time}} $$.

Selective Repeat (SR):

  • Both windows $$\displaystyle W_s = W_r = 2^{m-1} $$ (if sequence number $m$ bits).

  • Receiver buffers out-of-order frames, sends ACK for each correctly received.

  • Sender retransmits only missing frames (on timeout or 3 duplicate ACKs).

  • More efficient but requires larger buffers and more complex.

Hybrid ARQ:

  • Combines FEC (Forward Error Correction) and ARQ.

  • Types: Type I (FEC only, no ARQ), Type II (FEC + incremental redundancy), Type III (FEC + ARQ with combining).

4.5 Multiple Access Protocols

Random Access (Contention):

  • Pure ALOHA: Transmit anytime; throughput $$\displaystyle S = G e^{-2G} $$ (max $$\displaystyle S_{\max} = 0.184 $$ at $$\displaystyle G=0.5 $$).

  • Slotted ALOHA: Time slots; throughput $$\displaystyle S = G e^{-G} $$ (max $$\displaystyle S_{\max} = 0.368 $$ at $$\displaystyle G=1 $$).

  • Binary Exponential Backoff (BEB): After collision, wait random $$\displaystyle k \times T_{slot} $$, $$\displaystyle k \in [0, 2^i-1] $$ after $i$ collisions.

CSMA (Carrier Sense Multiple Access):

  • 1-persistent: Sense channel; if idle, transmit; if busy, sense continuously.

  • Non-persistent: Sense; if busy, wait random time, then sense again.

  • p-persistent (for slotted): If idle, transmit with prob $p$, else defer to next slot.

CSMA/CD (Collision Detection, Ethernet):

  • Minimum Frame Size ensures collision detected before transmission ends.

$$\displaystyle \text{Min Frame Size} \ge 2 \times \text{Propagation Delay} \times \text{Bandwidth} $$.

Example: 2 km cable, signal speed $$\displaystyle 2 \times 10^8 $$ m/s → propagation delay $$\displaystyle = \frac{2000}{2 \times 10^8} = 10 \ \mu s $$. For 10 Mbps Ethernet → min frame $$\displaystyle = 2 \times 10 \ \mu s \times 10^7 \ \text{bps} = 200 $$ bits ≈ 25 bytes (Ethernet uses 64 bytes).

CSMA/CA (Collision Avoidance, Wi-Fi):

  • DIFS (Distributed InterFrame Space) for data; SIFS (Short IFS) for ACK.

  • RTS/CTS optional to avoid hidden terminal problem.

  • Virtual Carrier Sense: NAV (Network Allocation Vector) indicates channel reserved.

4.6 LAN Standards

Standard Access Method Topology Frame Format Key Features
IEEE 802.3 (Ethernet) CSMA/CD Bus/Star Preamble, Dest MAC, Src MAC, Type, Data, CRC 10 Mbps–400 Gbps, dominant LAN tech
IEEE 802.4 (Token Bus) Token passing Bus (logical ring) Token, control, data Deterministic, used in manufacturing
IEEE 802.5 (Token Ring) Token passing Ring Start delimiter, access control, data, CRC, end delimiter 4/16 Mbps, IBM legacy, FDDI similar (dual ring, 100 Mbps)
IEEE 802.11 (Wi-Fi) CSMA/CA Star (AP) Frame control, duration, addr1–4, seq, data, CRC Infrastructure/ad-hoc, 802.11n/ac/ax (MIMO, OFDM)

4.7 Data Link Protocols

HDLC (High-Level Data Link Control):

  • Bit-oriented, supports both async & sync.

  • Frame Format: Flag 01111110, Address, Control, Information, FCS, Flag.

  • Modes: Normal Response (NRM), Asynchronous Response (ARM), Asynchronous Balanced (ABM).

  • Control Field: $N(S)$ (seq), $N(R)$ (ack), P/F bit.

PPP (Point-to-Point Protocol):

  • Phases: Link establishment (LCP), authentication (PAP/CHAP), network layer (NCP), link termination.

  • Frame Format: Flag, Address (0xFF), Control (0x03), Protocol (e.g., 0x0021 for IP), Data, FCS, Flag.

  • Byte Stuffing: Flag 0x7E, escape 0x7D.

SLIP (Serial Line Internet Protocol):

  • Obsolete, simple: END (0xC0) frame delimiter, ESC (0xDB) stuffing.

  • No error detection, no multi-protocol support.

Frame Relay:

  • Virtual circuit (PVC/SVC), packet-switched, no error correction (only detection, higher layers handle).

  • Architecture: DTE (user device) ↔ DCE (switch) via access link.

  • Frame Format: Flag, DLCI (virtual circuit ID), control, data, FCS, flag.

  • Congestion: DE (Discard Eligible), FECN/BECN bits.

4.8 Bridges & Switches

Device Layer Function Learning Forwarding
Hub Physical Broadcast all ports No No
Bridge Data Link Connect LAN segments, filter by MAC Store-and-forward, learns MAC→port Flood if unknown, forward if known & different port
Switch Data Link Multi-port bridge, per-port collision domain Same as bridge, aging timer Same, often cut-through possible
Router Network Connects networks, routes by IP No (uses routing table) Forward based on longest prefix match

Transparent Bridging:

  • Bridges learn source MAC→port.

  • Forward if dest MAC known & on different port; flood if unknown or broadcast.

  • Spanning Tree Protocol (STP): Prevents loops by blocking redundant links (root bridge, root ports, designated ports).

Switch Learning:

  • Initially MAC table empty.

  • On frame arrival: record source MAC→ingress port.

  • If dest MAC in table & on different port → forward to that port.

  • If dest MAC unknown/broadcast → flood to all except ingress.

  • Entries aged out if inactive for timeout.

[!TIP] Exam Distinction: Hub = layer 1, repeat signal; Switch = layer 2, uses MAC; Router = layer 3, uses IP. Switches break collision domains, not broadcast domains.


5. Network Layer

5.1 Functions & Design Issues

  • Routing: Determine path from source to destination (routing algorithms).

  • Forwarding: Move packet from input to output link (per-hop decision).

  • Congestion Control: Prevent network overload (traffic-aware routing, admission control, load shedding).

  • Fragmentation/Reassembly: Handle MTU differences (IPv4: at routers; IPv6: only at source).

  • Virtual Circuits vs. Datagram Subnets:

    • VC Subnet: Connection setup, per-hop state (VC number), in-order delivery (e.g., Frame Relay, ATM).

    • Datagram Subnet: No setup, each packet independent, routing per packet (e.g., IP).

5.2 IP Addressing (IPv4)

Classful Addressing:

Class First Bits Network ID Host ID Range (first octet) Default Mask
A 0 8 bits 24 bits 0–127 255.0.0.0
B 10 16 bits 16 bits 128–191 255.255.0.0
C 110 24 bits 8 bits 192–223 255.255.255.0
D 1110 — — 224–239 Multicast
E 1111 — — 240–255 Reserved

Limitations: Wasted addresses (class B too large for many orgs), no subnetting flexibility, routing table explosion.

CIDR (Classless Inter-Domain Routing):

  • Notation: a.b.c.d/n where $n$ = network prefix length.

  • Example: 192.168.1.0/24 → network mask 255.255.255.0.

  • Allows variable-length subnetting, efficient allocation.

  • Aggregation: Route summarization (e.g., 192.168.0.0/16 covers 192.168.0.0/24 to 192.168.255.0/24).

Subnetting:

  • Borrow bits from host portion to create subnets.

  • Subnet Mask: 32-bit number with contiguous 1s for network+subnet.

  • Example: 200.133.175.0/24 → create 16 subnets (need 4 bits, $$\displaystyle 2^4=16 $$). New mask /28 (255.255.255.240).

    • Subnet 0: 200.133.175.0/28 → hosts 200.133.175.1–200.133.175.14, broadcast 200.133.175.15.

    • Subnet 1: 200.133.175.16/28 → hosts 17–30, broadcast 31.

    • ... Subnet 15: 200.133.175.240/28 → hosts 241–254, broadcast 255.

Supernetting (aggregation): Combine contiguous networks into larger prefix (reverse of subnetting).

IPv6 Overview:

  • 128-bit address (hex colon notation: 2001:0db8:85a3::8a2e:0370:7334).

  • Header: Fixed 40 bytes, no options, removed checksum, flow label.

  • Address Types: Unicast (global, link-local fe80::/10), multicast, anycast.

  • No broadcast; uses multicast.

  • Built-in security (IPsec mandatory), autoconfiguration (SLAAC).

5.3 Routing Algorithms

Distance Vector Routing (DVR):

  • Bellman-Ford Equation: $$\displaystyle D_x(y) = \min_v \{ c(x,v) + D_v(y) \} $$.

  • Each router maintains distance vector (cost to each dest).

  • Periodically exchange entire vector with neighbors.

  • Count-to-Infinity Problem: Slow convergence on link failure; Split Horizon, Poison Reverse mitigate.

  • RIP: Uses hop count (max 15), updates every 30 sec.

Link State Routing (LSR):

  • Each router discovers neighbors, measures cost, builds LSP (Link State Packet).

  • Flood LSPs to all routers → each has complete topology.

  • Dijkstra's Algorithm: Compute shortest path tree from self.

    
    Initialize: tree = {self}, dist[self]=0, others=∞
    
    Repeat: pick node w not in tree with min dist[w]
    
            add w to tree, update dist[v] for neighbors v of w: 
    
            dist[v] = min(dist[v], dist[w] + cost(w,v))
    
    
  • OSPF uses LSR, areas to scale.

Comparison:

Feature DVR LSR
Convergence Slow (count-to-infinity) Fast (flooding)
Overhead Periodic full exchange Flood on change, large initial
Scalability Poor (routing loops) Good (hierarchical areas)
Memory Stores only neighbor vectors Stores full topology map
Example RIP, IGRP OSPF, IS-IS

Shortest Path Routing:

  • Metrics: Hop count, bandwidth, delay, reliability, cost.

  • Dijkstra's (LSR) or Bellman-Ford (DVR) compute paths.

Example: Given graph with nodes A,B,C,D, compute shortest path from A to all.

5.4 Internet Control Protocols

ARP (Address Resolution Protocol):

  • Maps IP address → MAC address on local network.

  • Broadcast ARP request: "Who has IP X? Tell Y (MAC Y)".

  • Target replies with ARP reply (unicast).

  • Cache ARP entries (typically 15 min).

RARP (Reverse ARP):

  • Diskless workstation boot: "Who has MAC X? My IP?" → RARP server replies with IP.

  • Obsolete, replaced by BOOTP/DHCP.

ICMP (Internet Control Message Protocol):

  • Error-reporting & diagnostic messages (carried in IP).

  • Common Types:

    • 0/8: Echo Reply/Request (ping).

    • 3: Destination Unreachable (code: network unreachable, port unreachable, etc.).

    • 11: Time Exceeded (TTL expired, traceroute).

    • 5: Redirect (router suggests better next hop).

  • Query messages: 12/13 Timestamp, 17/18 Address Mask Request/Reply.

5.5 Fragmentation & Reassembly

Need: Different networks have different MTU (Maximum Transmission Unit). IP must handle fragmentation if packet larger than outgoing link's MTU.

IPv4 Header Fields:

  • Identification: Same for all fragments of a packet.

  • Flags: DF (Don't Fragment), MF (More Fragments).

  • Fragment Offset: 13 bits, units of 8 bytes → offset = byte number / 8.

Reassembly:

  • Only at destination (or at router if configured).

  • Uses identification, source IP, dest IP, protocol to identify fragments.

  • Waits for all fragments (timer prevents indefinite wait).

  • Problem: One lost fragment → entire packet lost.

Example: 4000-byte packet (20-byte header, 3980 data), MTU 1500 bytes. First fragment: offset=0, MF=1, data 1480. Second: offset=185 (1480/8), MF=1, data 1480. Third: offset=370, MF=0, data 1020.


6. Transport Layer

6.1 Functions & Services

  • Process-to-Process Communication: Uses port numbers (16-bit) to distinguish apps.

  • Reliability: ACKs, retransmission (TCP).

  • Flow Control: Receiver advertises window size.

  • Congestion Control: Adjust sending rate based on network congestion.

  • Connection-oriented vs. Connectionless:

    • Connection-oriented (TCP): Handshake, stateful, reliable, ordered.

    • Connectionless (UDP): No setup, best-effort, low overhead.

6.2 UDP (User Datagram Protocol)

  • Header (8 bytes):

    • Source Port (16), Dest Port (16)

    • Length (header+data, 16)

    • Checksum (16, optional in IPv4, mandatory in IPv6)

  • Features: No connection setup, no congestion control, no ordering, minimal overhead.

  • Applications: DNS, VoIP, video streaming, DHCP.

6.3 TCP (Transmission Control Protocol)

Header Format (20–60 bytes):

Field Size (bits) Purpose
Source Port 16 Sender's port
Dest Port 16 Receiver's port
Sequence Number 32 Byte number of first data in segment
Acknowledgment Number 32 Next expected byte (if ACK flag set)
Data Offset 4 Header length (32-bit words)
Reserved 3 Must be 0
Flags (URG, ACK, PSH, RST, SYN, FIN) 9 Control bits
Window Size 16 Receiver's buffer space (bytes)
Checksum 16 Error detection (covers pseudo-header)
Urgent Pointer 16 Byte offset to urgent data (if URG set)
Options variable MSS, window scale, timestamps, SACK permitted

Connection Establishment: Three-Way Handshake


Client                          Server

  SYN=1, seq=x  ------------>  

                     SYN=1, seq=y, ACK=x+1  <------------

  ACK=y+1  ------------>

  • Prevents old duplicate connection initiations.

  • SYN consumes one sequence number.

Connection Termination: Four-Way Handshake (graceful)


Finisher (e.g., client)          Other (server)

  FIN=1, seq=u  ------------>  

                     ACK=u+1  <------------

                     FIN=1, seq=v  <------------

  ACK=v+1  ------------>

  • Each direction closed separately (half-close possible).

  • TIME_WAIT state (2×MSL) ensures last ACK received, prevents old duplicates.

Flow Control: Sliding window. Receiver advertises Window Size in header. Sender must not send more than rwnd bytes beyond last ACKed byte.

Congestion Control (RFC 5681, updated by RFC 6582, 8312):

  • Slow Start: cwnd starts at 1–10 MSS, doubles each RTT until ssthresh.

  • Congestion Avoidance: After cwnd >= ssthresh, increase cwnd by ~1 MSS per RTT (additive increase).

  • Fast Retransmit: On 3 duplicate ACKs, retransmit missing segment, set ssthresh = cwnd/2, cwnd = ssthresh + 3 MSS.

  • Fast Recovery: On duplicate ACKs, inflate cwnd (simulate ACK of new data), then on new ACK, set cwnd = ssthresh.

  • Timeout: cwnd set to 1 MSS, ssthresh = cwnd/2, enter slow start.

Load Shedding (Congestion Avoidance at source):

  • Leaky Bucket: Output at constant rate, bursty input smoothed (queue discipline).

  • Token Bucket: Tokens accumulate at rate $r$, up to $b$ tokens. Packet of size $s$ needs $s$ tokens. Allows burst up to $b$ tokens.

6.4 Sliding Window in Transport

  • GBN: Sender window $$\displaystyle W_s $$, receiver window $$\displaystyle W_r=1 $$. Cumulative ACK. Retransmit all after timeout/duplicate ACK.

  • SR: Sender/receiver windows $$\displaystyle W_s=W_r=2^{m-1} $$ (if $m$-bit seq num). Individual ACK, selective retransmission.

  • Piggybacking: ACK carried in data segment going opposite direction (saves bandwidth).

Link Utilization (for GBN):

$$ U = \frac{W_s}{1 + 2a} \quad \text{where} \quad a = \frac{\text{propagation delay}}{\text{frame transmission time}} $$

Max $U \approx 1$ if $$\displaystyle W_s \ge 1 + 2a $$.


7. Application Layer

7.1 Domain Name System (DNS)

  • Hierarchy: Root servers (.) → TLD servers (.com, .org, country-code) → Authoritative servers (domain owner).

  • Resolution:

    • Recursive: Resolver asks server, server returns final answer (may query others).

    • Iterative: Server returns referral to lower server if not authoritative.

  • Caching: Resolvers cache responses (TTL in record).

  • Resource Records:

    • A: IPv4 address.

    • AAAA: IPv6 address.

    • CNAME: Canonical name (alias).

    • MX: Mail exchange (priority).

    • NS: Name server.

    • PTR: Reverse lookup (IP → name).

7.2 Email

Architecture:

  • User Agent (UA): Outlook, Thunderbird (compose, send, receive).

  • Mail Server: Stores mailbox, runs SMTP (send) and POP3/IMAP (receive).

  • SMTP (Simple Mail Transfer Protocol):

    • Push protocol (client → server, server → server).

    • Commands: HELO, MAIL FROM, RCPT TO, DATA, QUIT.

    • ESMTP (Extended SMTP): EHLO, SIZE, STARTTLS.

  • POP3 (Post Office Protocol v3):

    • Download-and-delete (or keep) model.

    • Commands: USER, PASS, LIST, RETR, DELE, QUIT.

  • IMAP (Internet Message Access Protocol):

    • Remote mailbox access, folders, partial fetch.

    • Commands: LOGIN, SELECT, FETCH, STORE, LOGOUT.

  • MIME (Multipurpose Internet Mail Extensions):

    • Encodes non-ASCII: Content-Type (text/plain, image/jpeg), Content-Transfer-Encoding (base64, quoted-printable).

7.3 World Wide Web (WWW)

  • HTTP/HTTPS (Hypertext Transfer Protocol / Secure):

    • Request: Method (GET/POST/HEAD), URL, HTTP/1.1, Headers, optional Body.

    • Response: Status Code (200 OK, 404 Not Found, 500), Headers, Body.

    • Stateless (cookies, sessions for state).

    • HTTPS: HTTP over TLS/SSL (port 443).

  • URL: scheme://host:port/path?query#fragment.

  • Browsers: Render HTML, CSS, JavaScript.

7.4 File Transfer (FTP)

  • Two Connections:

    • Control Connection (port 21): persistent, commands/responses.

    • Data Connection (port 20 or dynamic): ephemeral, actual file transfer.

  • Modes:

    • Active: Server opens data connection to client (client sends PORT).

    • Passive: Client opens data connection to server (server sends PASV).

  • Commands: USER, PASS, LIST, RETR, STOR, QUIT.

7.5 Network Management (SNMP)

  • Components:

    • Manager: Central console (e.g., snmpwalk).

    • Agent: Software on managed device (router, switch).

    • MIB (Management Information Base): Database of managed objects (OID tree).

  • Operations: GET, GETNEXT, SET, TRAP (asynchronous notification).

  • Versions: SNMPv1 (no security), v2c (community string), v3 (user authentication, encryption).

7.6 Other Protocols

  • SSH (Secure Shell): Secure remote login (port 22), replaces Telnet/rlogin.

  • TLS/SSL: Cryptographic protocols for secure communication (HTTPS, SMTPS).


8. Cross-Layer & Advanced Topics

8.1 Network Devices Comparison

Device OSI Layer Function Intelligence Broadcast Domain
Hub Physical Signal regeneration None Single
Switch Data Link MAC learning/forwarding Per-port filtering Single (unless VLANs)
Bridge Data Link Connects two LAN segments MAC table Single
Router Network IP routing, subnet separation Routing table Multiple (per interface)
Gateway Application Protocol conversion (e.g., SMTP↱X.400) High Multiple

8.2 Remote Bridging

  • Concept: Bridge two distant LANs over a WAN link (e.g., two offices connected via leased line).

  • Implementation: Transparent bridging (STP) over WAN; may use Tunneling (e.g., GRE).

  • Challenges:

    • Latency: STP convergence slow over high-delay links.

    • Scalability: STP limits topology; remote bridges may cause loops.

    • Bandwidth: WAN links often slower than LANs → bottleneck.

    • STP Issues: Remote link treated as single hop; failure detection slow.

8.3 Quality of Service (QoS)

  • Need: Different applications have different requirements (delay, jitter, loss).

  • Techniques:

    • Traffic Shaping: Smooth bursty traffic (leaky bucket, token bucket).

    • Priority Queuing: Higher priority queues served first.

    • Resource Reservation: RSVP (ReSerVation Protocol) reserves resources along path.

    • Differentiated Services (DiffServ): PHB (Per-Hop Behavior) based on DS field in IP header (EF, AF classes).

    • Integrated Services (IntServ): Fine-grained per-flow reservation (not scalable).

  • Congestion Control vs. QoS: Congestion control prevents overload; QoS manages traffic during congestion.

8.4 Virtual LANs (VLANs)

  • Definition: Logical segmentation of LAN into broadcast domains, regardless of physical location.

  • Benefits: Security (isolate groups), reduced broadcast traffic, flexible administration.

  • VLAN Tagging (IEEE 802.1Q):

    • Inserts 4-byte VLAN tag between source MAC and EtherType.

    • TPID (Tag Protocol Identifier, 0x8100), TCI (Tag Control Information: PCP, DEI, VID).

    • VID (12 bits) supports 4094 VLANs (1–4094, 0 and 4095 reserved).

  • Trunk Port: Carries multiple VLANs (tagged); Access Port: single VLAN (untagged).

8.5 Cryptography Basics

  • Symmetric Encryption: Same key for encrypt/decrypt (AES, DES). Fast, key distribution problem.

  • Asymmetric Encryption: Public/private key pair (RSA, ECC). Slow, solves key distribution.

  • Digital Signature: Hash of message encrypted with sender's private key → authenticity, integrity.

  • Certificates: Bind public key to entity (X.509), issued by CA (Certificate Authority).

  • Role in Security:

    • Confidentiality: Encryption.

    • Integrity: Hash functions (SHA-256), digital signatures.

    • Authentication: Digital signatures, certificates.

    • Non-repudiation: Digital signatures.

8.6 Broadcast & Multicast Routing

  • Broadcast: To all nodes in network (e.g., ARP). Flooding (TTL=1 in IPv4) but wasteful.

  • Multicast: To group of interested nodes (e.g., video streaming).

    • IGMP (Internet Group Management Protocol): Hosts join/leave multicast groups on local network.

    • Multicast Routing Protocols:

      • DVMRP (Distance Vector Multicast Routing Protocol): Reverse path forwarding, prune/graft.

      • PIM (Protocol Independent Multicast): Sparse Mode (RP, shared tree), Dense Mode (flood & prune).

    • Source-Specific Multicast (SSM): Receiver specifies source.

8.7 Network Topology Design

Topology Advantages Disadvantages Use Case
Star Easy to manage, single link failure doesn't affect others Hub/switch failure fatal Ethernet LANs
Bus Simple, cheap Collisions, single failure fatal Legacy Ethernet (10BASE2)
Ring Deterministic access, no collisions Single failure breaks ring Token Ring, FDDI
Mesh High reliability, multiple paths Expensive, complex Backbone networks, Internet
Tree (hierarchical) Scalable, easy to expand Root node failure critical Large LANs, hierarchical networks
Hybrid Flexible, combines benefits Complex design Enterprise networks

Design Considerations:

  • Cost: Cabling, devices.

  • Reliability: Redundancy (mesh, dual-homed star).

  • Scalability: Ease of adding nodes.

  • Performance: Latency, bandwidth, collision domains.

  • Maintainability: Troubleshooting, modularity.

8.8 Internet Standards & RFCs

  • RFC (Request for Comments): IETF documents Internet standards.

    • Categories: Informational, Experimental, Best Current Practice (BCP), Internet Standard (STD).

    • Process: Internet Draft → RFC (after review).

  • TCP/IP Standardization:

    • IETF (Internet Engineering Task Force): Develops standards (working groups).

    • IESG (Internet Engineering Steering Group): Approves standards.

    • IAB (Internet Architecture Board): Oversees architecture.

    • IRTF (Internet Research Task Force): Long-term research.

  • Key RFCs: RFC 791 (IP), RFC 793 (TCP), RFC 768 (UDP), RFC 2460 (IPv6), RFC 1034/1035 (DNS).


[!TIP] Final Exam Strategy:

  1. Definitions first: Always start with clear definition (e.g., "CRC is...").
  1. Formulas boxed: Nyquist, Shannon, throughput calculations.
  1. Diagrams in mind: OSI/TCP/IP layers, TCP header, IPv4 header, Ethernet frame, sliding window.
  1. Compare/contrast: OSI vs TCP/IP, DVR vs LSR, TCP vs UDP, hub/switch/router, circuit/packet switching.
  1. Numerical practice: Subnetting, CRC, Nyquist/Shannon, TDM frame size, min Ethernet frame, ALOHA throughput.
  1. Protocol operations: ARP request/reply, DNS iterative/recursive, TCP handshake, CSMA/CD steps.
  1. Real-world examples: Mention Ethernet (802.3), Wi-Fi (802.11), OSPF, BGP, HTTP/2, QUIC where relevant.

Key Formulas Summary:

  • Nyquist: $$\displaystyle \boxed{R_{\max} = 2B \log_2 M} $$

  • Shannon: $$\displaystyle \boxed{C = B \log_2(1 + \text{SNR})} $$

  • Stop-and-Wait Efficiency: $$\displaystyle \boxed{U = \frac{1}{1 + 2a}} $$, $$\displaystyle a = \frac{\text{prop delay}}{\text{frame tx time}} $$

  • Slotted ALOHA Throughput: $$\displaystyle \boxed{S = G e^{-G}} $$, max $0.368$ at $$\displaystyle G=1 $$

  • Pure ALOHA Throughput: $$\displaystyle \boxed{S = G e^{-2G}} $$, max $0.184$ at $$\displaystyle G=0.5 $$

  • CSMA/CD Minimum Frame: $$\displaystyle \boxed{\text{Min Frame} \ge 2 \times \text{Prop Delay} \times \text{Bandwidth}} $$

  • GBN Utilization: $$\displaystyle \boxed{U = \frac{W_s}{1 + 2a}} $$ (if $$\displaystyle W_s \le 1+2a $$)

Go to where you left off?

Quick Add to Notes

Save questions, your own notes and screenshots into notes filed by unit. It takes a free account.

Create free account

Have an account? Log in