Computer Networks Complete Study Notes (AKTU Sem 6)

Unit I · 8 Lectures

Introductory Concepts & Physical Layer

OSI model, TCP/IP, topologies, transmission media, switching techniques

OSI Reference Model ★ VERY HIGH

What is the OSI Model?

The Open Systems Interconnection model is a conceptual framework developed by ISO that divides network communication into 7 distinct layers. Each layer has specific responsibilities and communicates with the layers directly above and below it. The key principle is encapsulation — data is wrapped with headers at each layer on the sending side and unwrapped (de-encapsulated) at each layer on the receiving side.

Application (7)

User interface. Provides services like web browsing, email, file transfer. Protocols: HTTP, FTP, SMTP, DNS.

Presentation (6)

Data translation, encryption/decryption, compression. Ensures data is in readable format (JPEG, MPEG, ASCII).

Session (5)

Manages sessions (dialog control). Establishes, maintains, terminates connections between apps.

Transport (4)

End-to-end delivery, segmentation, error recovery, flow control. Protocols: TCP (reliable), UDP (fast).

Network (3)

Logical addressing (IP), routing, path determination. Devices: Router. Protocols: IP, ICMP, ARP.

Data Link (2)

Physical addressing (MAC), framing, error detection, flow control. Devices: Switch, Bridge.

Physical (1)

Raw bit transmission — electrical/optical/radio signals. Defines cables, connectors, signal voltage. Devices: Hub, Repeater, NIC.

Mnemonic: "Please Do Not Throw Sausage Pizza Away" → Physical, Data Link, Network, Transport, Session, Presentation, Application (bottom to top)

TCP/IP Model vs OSI Model PYQ 2023-24

Feature	OSI Model	TCP/IP Model
Layers	7	4
Developed by	ISO (1984)	DARPA (1970s)
Type	Reference model (theoretical)	Implementation model (practical)
Protocol	Protocol independent	Protocol specific (TCP/IP suite)
Application layer	3 separate layers (5,6,7)	Single Application layer
Usage	Troubleshooting, teaching	Internet (real-world)

Network Topologies PYQ Both Years

MESH TOPOLOGY — PYQ 2024-25 (2 marks)

Links = n(n-1) / 2
For n = 10 nodes: Links = 10 × 9 / 2 = 45 links
I/O ports per node = n - 1 = 9 ports

Switching Techniques

Circuit Switching

Dedicated path established before data. Used in telephone networks. Phases: connection setup → data transfer → teardown. Wasteful if no data to send.

Packet Switching

Data split into packets. Each routed independently. Two types: Datagram (each packet independent) and Virtual Circuit (path pre-established). Used in Internet.

Message Switching

Entire message stored at each node before forwarding (store-and-forward). No dedicated path. Used in email. Slower than packet switching.

Transmission Delay & Latency PYQ 2024-25

TOTAL LATENCY FORMULA — PYQ 2024-25 (7 marks numerical)

Total Latency = Propagation Delay + Transmission Delay + Queuing Delay + Processing Delay

Propagation Delay = Distance / Propagation Speed
Transmission Delay = Message Size / Bandwidth

PYQ Solution: 4 MB frame, 1000 km, 2 Mbps, 5 routers (1μs + 2μs each)
Transmission = (4×8×10⁶) / (2×10⁶) = 16 s
Propagation = 1000×10³ / (2×10⁸) = 0.005 s
Router delays = 5 × (1 + 2) μs = 0.000015 s
Total ≈ 16.005015 s

SIGNAL ATTENUATION — PYQ 2024-25

Attenuation (dB) = 10 · log₁₀(P_out / P_in)
−3 dB ≈ half the power
PYQ: P_in = 2W, attenuation = −3 dB
P_out = 2 × 10^(−3/10) = 2 × 0.501 ≈ 1.002 W ≈ 1 W

Unit II · 8 Lectures

Data Link Layer & MAC

Framing, error detection, flow control, ALOHA, CSMA, bridges, LAN standards

CRC — Cyclic Redundancy Check ★ BOTH YEARS

How CRC Works

CRC is a powerful error detection technique. The sender appends a checksum (remainder) to the data. The receiver performs the same division — if remainder ≠ 0, an error occurred. The "generator polynomial" acts as the divisor. Division uses XOR, not regular subtraction.

Represent data as binary: Convert the polynomial/bit stream to binary. x⁸+x⁶+x³+x²+x+1 → 1 0 1 0 0 1 1 1 1

Append zeros: Append (degree of generator) zeros to the message. Generator x⁴+x²+x+1 → degree 4, so append 4 zeros.

XOR division: Divide augmented message by generator using binary XOR. Bring down bits one by one just like long division.

Get remainder (CRC): The remainder after division is the CRC checksum — append it to the original message.

Receiver check: Receiver divides received codeword by generator. If remainder = 0000 → no error. If ≠ 0 → error detected.

PYQ 2024-25 — CRC EXAMPLE (7 marks)

Message : x⁸+x⁶+x³+x²+x+1 → 1 0 1 0 0 1 1 1 1
Generator : x⁴+x²+x+1 → 1 0 1 1 1
Append 4 zeros → Dividend: 1 0 1 0 0 1 1 1 1 0 0 0 0

Perform XOR long division → Get 4-bit remainder (CRC)
Transmitted codeword = Original message + CRC remainder

Error detection: Invert the 3rd bit, then divide at receiver.
Non-zero remainder → Error detected ✓

Key Point: CRC can detect all single-bit errors, all double-bit errors, all odd number of errors, and all burst errors shorter than the generator polynomial degree.

Flow Control — ARQ Protocols PYQ 2023-24

Protocol	Window Size	On Error	Efficiency
Stop & Wait	1	Retransmit 1 frame	Low — sender idle waiting for ACK
Go-Back-N	Up to 2ⁿ⁻¹	Retransmit from error frame onward	Better — W/(1+2a)
Selective Repeat	Up to 2ⁿ⁻¹	Retransmit only lost frame	Best — most efficient

Piggybacking: When node B receives a frame from A and also has data to send back, B's acknowledgment (ACK) is attached to (piggybacked on) the outgoing data frame. This saves bandwidth by combining ACK + data in one frame instead of sending ACK separately. PYQ 2024-25

ALOHA Protocols ★ BOTH YEARS

PYQ 2024-25 — SLOTTED ALOHA NUMERICAL (7 marks)

Given: 10% of slots are idle → P(idle) = 0.10
P(idle) = e^(-G) → 0.10 = e^(-G) → G = -ln(0.10) ≈ 2.303

(i) Channel load G ≈ 2.303
(ii) Throughput S = G·e^(-G) = 2.303 × 0.10 ≈ 0.23 = 23%

CSMA/CD vs CSMA/CA: CSMA/CD (Collision Detection) used in wired Ethernet — station listens while transmitting and stops on collision. CSMA/CA (Collision Avoidance) used in wireless (802.11 WiFi) — cannot detect collision so avoids it using backoff and RTS/CTS.

Bridge & Spanning Tree PYQ 2023-24

How a Learning Bridge Works

A bridge connects two LAN segments at the Data Link layer. It maintains a forwarding table (MAC address → port mapping). When a frame arrives, it reads the source MAC and records it. Then checks the destination MAC: if found → forward to correct port (filtering); if not found → flood (broadcast to all ports). This self-learning process eliminates the need for manual configuration.

Spanning Tree Algorithm (STA): Used to eliminate loops in networks with multiple bridges. One bridge is elected as Root Bridge (lowest Bridge ID). A shortest-path tree is built from root to all other bridges. Ports not on the tree are blocked. This prevents broadcast storms while maintaining redundancy.

Unit III · 8 Lectures

Network Layer

IP addressing, subnetting, ARP/DHCP/ICMP, routing algorithms, IPv6

IPv4 Header Format PYQ 2023-24

Subnetting & CIDR ★ PYQ 2024-25

Subnetting Concept

Subnetting divides a large IP address block into smaller sub-networks. Borrowing bits from the host part increases the network prefix. Each borrowed bit doubles the number of subnets (halves hosts per subnet). CIDR (Classless Inter-Domain Routing) allows arbitrary prefix lengths like /26 instead of strict class boundaries.

PYQ 2024-25 — 192.168.10.0/24 divided into 4 subnets (7 marks)

Need 4 subnets → borrow 2 bits (2² = 4) → new prefix = /26
Hosts per subnet = 2⁶ - 2 = 62 usable hosts

Subnet 1: 192.168.10.0 – 192.168.10.63 (Network: .0, Broadcast: .63)
Subnet 2: 192.168.10.64 – 192.168.10.127 (Network: .64, Broadcast: .127)
Subnet 3: 192.168.10.128 – 192.168.10.191 (Network: .128, Broadcast: .191)
Subnet 4: 192.168.10.192 – 192.168.10.255 (Network: .192, Broadcast: .255)

Subnet mask for /26 = 255.255.255.192

ARP, RARP, DHCP, ICMP PYQ 2024-25

ARP — Address Resolution Protocol

Resolves IP address → MAC address. Sender broadcasts "Who has IP X?" — Owner replies with its MAC. ARP table cached for efficiency. Used within same subnet only.

RARP — Reverse ARP

Resolves MAC address → IP address. Used by diskless workstations that know their MAC but not IP. Now replaced by DHCP/BOOTP.

DHCP — Dynamic Host Config

Automatically assigns IP, subnet mask, gateway, DNS to hosts. Process: DISCOVER → OFFER → REQUEST → ACK (DORA). Leases IPs for a time period.

ICMP — Internet Control Message

Reports errors and diagnostics. Used by ping (Echo Request/Reply) and traceroute (TTL exceeded). Not for data delivery — only control messages.

Distance Vector Routing ★ BOTH YEARS

Bellman-Ford Algorithm

Each router maintains a routing table (distance vector) listing the best known distance to every other router and the next hop. Periodically, routers share their tables with neighbours. On receiving a neighbour's table, update own table: new_cost = cost_to_neighbour + neighbour's_cost_to_dest. Pick the minimum. Used by RIP protocol.

PYQ 2024-25 — DISTANCE VECTOR NUMERICAL (7 marks)

Router C receives:
From B (delay to C = 6): vector (5, 0, 8, 12, 6, 2) → A,B,C,D,E,F
From D (delay to C = 3): vector (16,12, 6, 0, 9, 10)
From E (delay to C = 5): vector (7, 6, 3, 9, 0, 4)

C's new delay to A = min(6+5, 3+16, 5+7) = min(11,19,12) = 11 via B
C's new delay to B = min(6+0, 3+12, 5+6) = min(6,15,11) = 6 via B
C's new delay to D = min(6+12,3+0,5+9) = min(18,3,14) = 3 via D
C's new delay to E = min(6+6, 3+9, 5+0) = min(12,12,5) = 5 via E
C's new delay to F = min(6+2, 3+10,5+4) = min(8,13,9) = 8 via B

Count-to-Infinity Problem: In DVR, if a link fails, routers keep advertising stale (outdated) routes — the distance count increases to infinity before routers realize the destination is unreachable. Fix: Split Horizon (don't advertise a route back to the router you learned it from) or Poison Reverse (advertise broken routes with infinite metric).

IPv6 Features PYQ 2024-25

Larger Address Space

128-bit addresses (vs 32-bit IPv4). Written as 8 groups of 4 hex digits. Virtually unlimited addresses — 340 undecillion.

Simplified Header

Fixed 40-byte header (vs variable IPv4). No checksum (done by transport layer). No fragmentation by routers — only end hosts. Faster processing.

Extension Headers

Optional features (routing, fragmentation, authentication) moved to extension headers chained after main header. Each has "Next Header" field pointing to next extension.

Other Improvements

Built-in IPSec security, autoconfiguration (SLAAC), better multicast, anycast support, no NAT needed, flow labels for QoS.

Unit IV · 8 Lectures

Transport Layer

TCP, UDP, connection management, congestion control, flow control

TCP Segment Header ★ BOTH YEARS

TCP Connection Management — 3-Way Handshake ★ BOTH YEARS

4-Way Termination (Connection Release): FIN → ACK → FIN → ACK. Either side can initiate. After sending FIN, enters TIME_WAIT state (waits 2×MSL) to ensure all packets are delivered before closing.

Congestion Control — Leaky Bucket & Token Bucket ★ BOTH YEARS

Feature	Leaky Bucket	Token Bucket
Output rate	Constant (fixed rate)	Variable (bursty allowed up to token count)
Burst traffic	Not allowed — excess dropped	Allowed if tokens available
Use case	Strict rate limiting (VoIP)	Flexible rate with burst (most networks)
Token concept	No tokens	Tokens generated at rate r; bucket capacity b
Max burst	None	b (bucket size = max burst size)

Silly Window Syndrome: TCP sends tiny segments, wasting header overhead. Nagle's Algorithm (Sender fix): Buffer small data, send only when window ≥ MSS or previous ACK received. Clark's Algorithm (Receiver fix): Don't send window update until window ≥ MSS or half the buffer is free. PYQ 2024-25

TCP vs UDP

Feature	TCP	UDP
Connection	Connection-oriented (3-way handshake)	Connectionless
Reliability	Reliable — ACK, retransmission	Unreliable — no guarantee
Ordering	In-order delivery guaranteed	No ordering
Speed	Slower (overhead)	Faster (low overhead)
Use case	HTTP, FTP, SMTP, SSH	DNS, DHCP, VoIP, Video streaming, Gaming
Header size	20–60 bytes	8 bytes (fixed)

Unit V · 8 Lectures

Application Layer

DNS, HTTP, FTP, email protocols, SNMP, cryptography, RSA

DNS — Domain Name System PYQ 2024-25

How DNS Works

DNS is the Internet's "phone book" — it translates human-readable domain names (www.google.com) to IP addresses (142.250.82.46). DNS is hierarchical and distributed. The hierarchy: Root → TLD (.com, .in) → Second-level (google) → Subdomain (www). Resolution is recursive (client asks resolver, resolver asks root, then TLD, then authoritative server) or iterative.

RSA Cryptography ★ BOTH YEARS

Public Key Cryptography

Each user has a key pair: Public key (shared with everyone) and Private key (kept secret). What is encrypted with the public key can only be decrypted by the corresponding private key, and vice versa. Eliminates the key distribution problem of symmetric cryptography.

Advantages: No shared secret needed, digital signatures possible, scalable for N users.
Disadvantages: Computationally expensive, vulnerable to man-in-the-middle without PKI, key management required.

RSA ALGORITHM STEPS — PYQ BOTH YEARS

Step 1: Choose two large primes p and q
Step 2: n = p × q (modulus)
Step 3: φ(n) = (p − 1)(q − 1) (Euler's totient)
Step 4: Choose e: 1 < e < φ(n) and gcd(e, φ(n)) = 1
Step 5: Find d: d × e ≡ 1 (mod φ(n)) [using Extended Euclidean Algorithm]
Step 6: Public Key = (e, n), Private Key = (d, n)
Step 7: Encrypt: C = Mᵉ mod n
Step 8: Decrypt: M = Cᵈ mod n

Example: p=3, q=11 → n=33, φ(n)=20, e=7 (gcd(7,20)=1)
d × 7 ≡ 1 (mod 20) → d = 3 (since 3×7=21≡1 mod 20)
Encrypt M=2: C = 2⁷ mod 33 = 128 mod 33 = 29
Decrypt C=29: M = 29³ mod 33 = 24389 mod 33 = 2 ✓

SNMP — Simple Network Management Protocol ★ BOTH YEARS

SNMP Components

Manager (NMS): Central station that monitors network. Polls agents, receives traps, queries MIB.

Agent: Software on managed device (router, switch). Maintains MIB, responds to manager.

MIB (Management Information Base): Hierarchical database of managed objects (variables) on each device.

SNMP PDU Operations

GetRequest: Manager requests value of MIB object from agent.
GetNextRequest: Get next object in MIB tree.
SetRequest: Manager sets value of MIB object.
GetResponse: Agent replies to Get/Set requests.
Trap: Agent sends unsolicited alert to manager (e.g. link down, cold start).

Key Application Protocols PYQ Both Years

Protocol	Port	Transport	Purpose	Key Feature
HTTP	80	TCP	Web pages transfer	Stateless; uses GET, POST, PUT methods
HTTPS	443	TCP	Secure web	HTTP over SSL/TLS encryption
FTP	21 (ctrl), 20 (data)	TCP	File transfer	Two connections: control + data channel
SMTP	25	TCP	Send email	Push protocol; uses MIME for multimedia
POP3	110	TCP	Receive email	Downloads and deletes from server
IMAP4	143	TCP	Receive email	Keeps mail on server; supports folders
DNS	53	UDP (mainly)	Name resolution	Hierarchical distributed database
DHCP	67/68	UDP	IP auto-assignment	DORA process (Discover-Offer-Request-ACK)
Telnet	23	TCP	Remote login	Unencrypted — use SSH instead
SNMP	161/162	UDP	Network management	Manager-Agent model, MIB database

SMTP & MIME: SMTP was originally text-only. MIME (Multipurpose Internet Mail Extensions) extends it to handle images, audio, video, and binary files by encoding them in Base64 and adding Content-Type headers. This is how email attachments work. PYQ 2024-25

IMAP4 vs POP3: POP3 downloads mail and deletes from server — works offline but no sync across devices. IMAP4 keeps mail on server, supports folders, flags, partial download — better for multiple devices. PYQ 2024-25

Data Compression PYQ Both Years

Lossless Compression

Original data recovered perfectly after decompression. Used for text, executables. Algorithms: Huffman coding (variable-length codes based on frequency), LZW (Lempel-Ziv-Welch used in GIF/ZIP), Run-length encoding.

Lossy Compression

Some data discarded — cannot recover original. Used for images, audio, video where slight quality loss is acceptable. Algorithms: JPEG (images), MP3 (audio), MPEG (video). Much higher compression ratios.

HUFFMAN CODING — Basic Concept

More frequent symbols → shorter codes
Less frequent symbols → longer codes
Build binary tree bottom-up from frequency table
Left edge = 0, Right edge = 1
Result: No code is a prefix of another (prefix-free code)
Average code length = Σ(frequency × code_length)

Quick Exam Formula Sheet

MESH LINKS

n(n-1)/2

PURE ALOHA

S = G·e^(-2G) max ≈ 18.4%

SLOTTED ALOHA

S = G·e^(-G) max ≈ 36.8%

ATTENUATION

dB = 10·log₁₀(P_out/P_in)

TOTAL LATENCY

Prop + Trans + Queue + Process

RSA ENCRYPT/DECRYPT

C = Mᵉ mod n | M = Cᵈ mod n

SUBNETTING (4 subnets from /24)

/24 → borrow 2 bits → /26

DISTANCE VECTOR

D(x,y) = min{c(x,v) + D(v,y)}

Unit I — Deep Dive

Transmission Media, Encoding & Multiplexing

Guided vs unguided media, signal encoding, TDM/FDM/WDM multiplexing

Transmission Media Classification

Twisted Pair (UTP/STP)

Two insulated copper wires twisted together to reduce EMI. UTP = Unshielded, STP = Shielded. Categories: Cat5e (1 Gbps), Cat6 (10 Gbps up to 55m). Most common for LAN wiring (RJ-45 connector). Cost-effective, easy to install. Susceptible to crosstalk and external interference.

Coaxial Cable

Central copper conductor surrounded by insulation, braided metal shield, outer plastic jacket. Better noise immunity than UTP. Used in cable TV (75Ω) and old 10Base2/10Base5 Ethernet (50Ω). Higher bandwidth than twisted pair. More expensive and harder to install.

Fiber Optic

Transmits light pulses through glass/plastic core via total internal reflection. Single-mode: one light path, long distance (100+ km), laser source. Multi-mode: multiple paths, short distance, LED source. Immune to EMI, no crosstalk, highest bandwidth (100+ Gbps), lightest, but most expensive.

Medium	Max Bandwidth	Max Distance	Attenuation	Cost	Use Case
UTP	1–10 Gbps	100 m	High	Low	LAN, telephone
Coaxial	10 Gbps	500 m	Medium	Medium	Cable TV, old LAN
Single-mode Fiber	100+ Gbps	100+ km	Very Low	High	WAN, backbone, ISP
Multi-mode Fiber	10 Gbps	2 km	Low	Medium-High	Campus LAN, data centers
Radio (WiFi)	9.6 Gbps	~100 m indoors	High (obstacles)	Low	WLAN, mobile
Microwave	Gbps	~50 km (LOS)	Medium	Medium	Backhaul, point-to-point
Satellite	Mbps	Global	High	High	Remote areas, GPS

Signal Encoding Schemes

Encoding	1 represented as	0 represented as	Advantage	Disadvantage
NRZ-L	High voltage	Low voltage	Simple	DC component, sync issues on long same-bit runs
NRZ-I	Transition at bit start	No transition	No baseline wander from 1-strings	Long 0 strings lose sync
Manchester	Low→High mid-bit	High→Low mid-bit	Self-clocking (sync always present)	Requires double bandwidth
Diff. Manchester	No transition at bit start	Transition at bit start	Noise resistant, polarity independent	Complex decoding
AMI (Bipolar)	Alternating +V and −V	Zero voltage	No DC component, error detection if lone zero	Long 0-strings lose sync
4B/5B	Every 4 data bits mapped to 5-bit code (no more than 1 leading 0 or 2 trailing 0s)		Guarantees transitions for sync	20% overhead

Why Manchester Encoding is used in Ethernet: Every bit has a transition in the middle (Low→High for 1, High→Low for 0). This built-in transition allows the receiver to recover the clock signal from the data itself — no separate clock wire needed. Used in 10Base-T Ethernet.

Multiplexing — FDM, TDM, WDM, CDMA

FDM — Frequency Division Multiplexing

Divides total bandwidth into frequency sub-bands. Each user gets a dedicated, non-overlapping band. Guard bands between channels prevent interference. Simultaneous transmission — all channels transmit at same time. Analog. Used in: AM/FM radio, cable TV, ADSL (different upload/download bands), traditional telephone (POTS).

TDM — Time Division Multiplexing

Divides time into slots. Each user gets a slot per frame. All users share full bandwidth but at different moments. Digital. Synchronous TDM: fixed slots assigned even if user has no data (wasteful). Statistical TDM: slots assigned on demand (more efficient). Used in: T1/E1 lines, SONET/SDH, ISDN.

WDM — Wavelength Division Multiplexing

Multiplexing over fiber optic cables using different light wavelengths (colors). DWDM (Dense WDM) carries 160+ channels on a single fiber pair. Each wavelength is essentially a separate FDM channel. Terabit/s capacity. Used in long-distance backbone networks between cities and continents.

CDMA — Code Division Multiple Access

All users transmit simultaneously on the same frequency band. Each user's data is multiplied by a unique orthogonal spreading code (chip sequence). Receiver uses the same code to extract intended signal. Used in: 3G mobile networks (WCDMA), GPS, military communications. Highly resistant to jamming and interference.

Unit II — Deep Dive

Error Detection, LAN Standards & MAC Deep Dive

Parity, checksum, Hamming code, CSMA variants, Ethernet frame, LAN standards

Error Detection — All Methods Compared

Simple Parity Check (VRC)

Adds 1 bit so total 1-bits = even (even parity) or odd (odd parity).

Example: 1010001 → three 1s → add parity bit 1 → 10100011 (even parity).

Detects: All single-bit errors, all odd-number errors.
Misses: Even number of errors (2, 4…).

2D Parity (LRC+VRC)

Organise data as a 2D matrix. Add parity bit per row (VRC) and per column (LRC). Much stronger coverage.

Detects: All 1, 2, and 3-bit errors; most burst errors.
Misses: Specific 4-bit patterns in 2×2 corners.

Internet Checksum

Used by TCP, UDP, IP. Sender: add all 16-bit words (1's complement arithmetic) → complement → attach. Receiver: sum all including checksum → should be all 1s (FFFF).

Advantage: Simple, software-efficient.
Weakness: Can miss some errors CRC would catch.

HAMMING CODE — SINGLE BIT ERROR CORRECTION

Minimum redundancy bits r: 2ʳ ≥ m + r + 1 (m = data bits)
For m=4: r=3 gives 2³=8 ≥ 4+3+1=8 ✓ → Total codeword = 7 bits

Redundancy bit positions: powers of 2 → positions 1, 2, 4
Data bit positions: 3, 5, 6, 7

Each parity bit covers positions whose binary representation has that bit set:
P1(pos1): covers 1,3,5,7 → P2(pos2): covers 2,3,6,7 → P4(pos4): covers 4,5,6,7

Error detection: Compute syndrome = XOR of positions of all failing parity bits
Syndrome value = exact bit position of error → flip that bit to correct

CSMA/CD vs CSMA/CA — Detailed Comparison

Feature	CSMA/CD (802.3 Ethernet)	CSMA/CA (802.11 WiFi)
Full form	Carrier Sense Multiple Access / Collision Detection	Carrier Sense Multiple Access / Collision Avoidance
Medium	Wired (can measure signal on wire)	Wireless (cannot detect collision — half-duplex, own signal drowns received)
Collision handling	Detect collision while transmitting, send JAM signal, back off	Avoid collision before it happens using backoff + optional RTS/CTS
DIFS/SIFS	Not used	DIFS = wait before transmit; SIFS = short wait before ACK
RTS/CTS	Not used	Optional: Request To Send → Clear To Send (solves hidden node problem)
ACK	No explicit ACK (collision = retransmit)	Every frame must be ACKed; no ACK → retransmit
Efficiency	High when load is low	Lower due to overhead of avoidance mechanisms
Min frame size	64 bytes (ensures collision detected in time)	No similar constraint (no collision detection)

Hidden Node Problem (WiFi): Station A can hear Access Point (AP) but not Station C. C can hear AP but not A. A and C both think channel is idle and transmit simultaneously → collision at AP. Solved by RTS/CTS: A sends RTS to AP; AP broadcasts CTS heard by all stations; C waits during A's transmission.

Ethernet Frame Structure & IEEE 802 Standards

Standard	Technology	Speed	Medium	Use Case
IEEE 802.3	Ethernet (CSMA/CD)	10 Mbps to 400 Gbps	Twisted pair, fiber, coax	Wired LAN — universally used
IEEE 802.3u	Fast Ethernet	100 Mbps	Cat5 UTP or fiber	100Base-TX, office LAN
IEEE 802.3ab	Gigabit Ethernet	1 Gbps	Cat5e/Cat6 UTP	1000Base-T, enterprise LAN
IEEE 802.11b/g/n	Wi-Fi 1/2/4	11/54/600 Mbps	2.4 GHz radio	Home and office WiFi
IEEE 802.11ac	Wi-Fi 5	Up to 3.5 Gbps	5 GHz radio	High-speed WLAN
IEEE 802.15.1	Bluetooth	1–3 Mbps	2.4 GHz, 10m range	WPAN: peripherals, headsets
IEEE 802.15.4	Zigbee	250 Kbps	2.4 GHz, low power	IoT sensors, smart home
IEEE 802.16	WiMAX	Up to 1 Gbps	2–66 GHz radio	WMAN: city-wide broadband
IEEE 802.1D	STP (Spanning Tree)	—	—	Loop prevention in switched LANs
IEEE 802.1Q	VLAN Tagging	—	—	Virtual LANs on shared switches

Spanning Tree Protocol (STP) — Deep Dive

Why Spanning Tree Protocol?

In networks with multiple bridges/switches, there can be multiple paths between nodes. This creates broadcast storms: a broadcast frame loops endlessly causing network meltdown. STP prevents this by building a logical loop-free tree while maintaining physical redundancy.

Root Bridge Election: Bridge with lowest Bridge ID (priority + MAC address) becomes root. All ports on root bridge are designated (forwarding).

Root Port Selection: Each non-root bridge selects the port with least cost to root as its Root Port. Cost based on link speed (1 Gbps = cost 4, 100 Mbps = cost 19, 10 Mbps = cost 100).

Designated Port Selection: For each network segment, the bridge with lowest root path cost has its port as Designated Port (forwarding). The other end becomes Blocked Port.

Blocked Ports: Ports not in Root or Designated role are put in Blocking state — they receive BPDUs but don't forward data frames. Redundancy maintained without loops.

Port States: Blocking → Listening → Learning → Forwarding (and Disabled). Convergence takes ~30-50 seconds in classic STP. RSTP (802.1w) converges in <1 second.

Unit III — Deep Dive

Routing Algorithms, IPv6 & Advanced Network Layer

Dijkstra, RIP, OSPF, BGP, IPv6 headers, VPN, NAT, ICMP in detail

IP Address Classes & Classless Addressing

Class	First Octet Range	Default Mask	Network Bits	Host Bits	Max Hosts	Use
A	1 – 126	255.0.0.0 (/8)	8	24	16,777,214	Very large orgs, ISPs
B	128 – 191	255.255.0.0 (/16)	16	16	65,534	Medium-large orgs
C	192 – 223	255.255.255.0 (/24)	24	8	254	Small orgs, home networks
D	224 – 239	N/A	—	—	—	Multicast groups
E	240 – 255	N/A	—	—	—	Research/reserved

SUBNETTING RULES — KEY FORMULAS

Number of subnets = 2^(borrowed bits)
Hosts per subnet = 2^(remaining host bits) - 2 (subtract network & broadcast)
Subnet mask = all 1s in network portion, all 0s in host portion

CIDR notation: /n means first n bits are network bits
Block size = 256 - last octet of subnet mask

Special addresses:
127.0.0.0/8 = Loopback (localhost)
10.0.0.0/8 = Private Class A
172.16.0.0/12 = Private Class B
192.168.0.0/16 = Private Class C
0.0.0.0 = This host (unspecified)
255.255.255.255 = Limited broadcast

VLSM (Variable Length Subnet Masking): Unlike fixed subnetting, VLSM allows different subnets to have different prefix lengths within the same IP block. This allows efficient allocation — give large subnets to big departments and small subnets to point-to-point WAN links (which only need 2 hosts, /30). Requires classless routing protocols (OSPF, RIPv2, BGP) that carry prefix length info in routing updates.

ARP, RARP, DHCP, ICMP — Detailed Working

ARP — Address Resolution Protocol

Problem: You know destination IP but need MAC address for local delivery.

Process:
1. Check ARP cache — if found, use it
2. Broadcast ARP Request: "Who has IP X? Tell me (my MAC)"
3. Owner replies with ARP Reply (unicast) containing its MAC
4. Sender caches IP→MAC mapping

ARP Spoofing: Attacker sends fake ARP replies to associate their MAC with victim's IP → man-in-the-middle attack.

DHCP — Dynamic Host Configuration

DORA Process:
1. DISCOVER — Client broadcasts "Any DHCP server?" (UDP, src 0.0.0.0, dst 255.255.255.255)
2. OFFER — Server offers: IP, mask, gateway, DNS, lease time
3. REQUEST — Client accepts offer (still broadcast, other servers notified)
4. ACK — Server confirms assignment

Client uses IP for lease duration; sends RENEW at 50% of lease time. Releases IP on clean shutdown.

ICMP — Internet Control Message Protocol

Purpose: Error reporting and network diagnostics. Carried inside IP packets (Protocol = 1).

Key Message Types:
Type 0/8 — Echo Reply/Request (ping)
Type 3 — Destination Unreachable (port/host/net)
Type 5 — Redirect (use better route)
Type 11 — Time Exceeded (TTL=0, used by traceroute)
Type 12 — Parameter Problem

Traceroute works by sending packets with TTL=1,2,3… Each router that decrements TTL to 0 sends ICMP Type 11 back, revealing the path.

NAT — Network Address Translation

Purpose: Allows multiple private IP addresses to share one public IP. Conserves IPv4 address space.

How it works: Router maintains NAT table mapping (private IP, private port) ↔ (public IP, public port). Outgoing: replaces private src IP/port. Incoming: looks up table, replaces public dst with private.

Problem: Breaks end-to-end connectivity; complicates P2P applications. Not needed with IPv6.

Routing Protocols — RIP, OSPF, BGP Deep Dive PYQ 2023-24

Feature	RIP v2	OSPF	BGP-4
Type	IGP (intradomain)	IGP (intradomain)	EGP (interdomain)
Algorithm	Distance Vector (Bellman-Ford)	Link State (Dijkstra SPF)	Path Vector
Metric	Hop count (max 15; 16 = ∞)	Cost = 10⁸/bandwidth	Path attributes (AS-PATH, LOCAL-PREF, MED)
Convergence	Slow (minutes)	Fast (seconds)	Slow (policy-driven)
Transport	UDP port 520	Directly over IP (Protocol 89)	TCP port 179
Update trigger	Every 30 seconds (full table)	On topology change (LSA flooding)	On change (incremental updates)
Scale	Small networks (<15 hops)	Large enterprise (areas/hierarchy)	Internet-scale (entire global routing)
Features	Simple, easy config	Areas, DR/BDR, VLSM, auth	Policy-based, multi-homing, confederations
Loop prevention	Split horizon, poison reverse	SPF guarantees loop-free tree	AS-PATH attribute (own AS = discard)

OSPF Areas: Large networks are divided into areas. Area 0 (backbone area) connects all other areas. Routers within an area have identical LSDB. Area Border Routers (ABRs) connect areas. Autonomous System Boundary Routers (ASBRs) connect OSPF to external routing protocols. This hierarchical design reduces flooding scope and speeds convergence.

IPv6 — Complete Coverage PYQ 2024-25

IPv6 Extension Headers

Optional features moved to chained extension headers:
Hop-by-Hop Options: Options for every router on path
Routing Header: Source routing (specify intermediate nodes)
Fragment Header: Only source can fragment (routers cannot)
Auth Header (AH): Data integrity, authentication
ESP Header: Encryption + auth (IPSec)
Destination Options: Info for final destination only

Each header has a Next Header field pointing to the next one in chain.

IPv6 Address Types & Notation

Notation: 8 groups of 4 hex digits separated by colons
Example: 2001:0db8:85a3:0000:0000:8a2e:0370:7334
Compressed: 2001:db8:85a3::8a2e:370:7334 (:: = consecutive zeros)

Address types:
Unicast — single interface (global, link-local, loopback)
Multicast — group of interfaces (FF00::/8)
Anycast — nearest of a group (same address, multiple nodes)
No broadcast in IPv6 — replaced by multicast

SLAAC: Stateless Address Autoconfiguration — device generates own IPv6 address using prefix + MAC-derived Interface ID (EUI-64). No DHCP needed.

VPN — Virtual Private Network PYQ 2024-25

VPN Architecture & Tunneling

A VPN creates an encrypted tunnel through a public network (Internet). The private IP packet is encapsulated inside a public packet — this is tunneling. The VPN gateway at the receiving end decapsulates and decrypts. From the application's perspective, the remote device appears to be on the local private network.

VPN Type	Description	Protocol	Use Case
Site-to-Site	Connects two entire LANs. Gateways handle encryption transparently.	IPSec (Tunnel mode)	Branch office connectivity
Remote Access	Individual user connects to corporate network from home.	SSL/TLS, IPSec, OpenVPN	Work from home, travel
MPLS VPN	Service provider creates virtual circuits using labels.	MPLS + BGP	Enterprise WAN, ISP services

IPSec Modes: Transport mode — encrypts only the payload (IP header unchanged); used for host-to-host. Tunnel mode — encrypts entire IP packet and adds new IP header; used for VPN gateways (site-to-site). IPSec uses AH (Authentication Header) for integrity without encryption, and ESP (Encapsulating Security Payload) for encryption + authentication.

Unit IV — Deep Dive

TCP Congestion Control, QoS & Transport Layer Complete

Slow start, AIMD, QoS mechanisms, UDP header, port numbers

UDP Header & Comparison with TCP

Feature	TCP	UDP
Connection	Connection-oriented (3-way handshake required)	Connectionless (no setup)
Reliability	Guaranteed delivery via ACK + retransmission	Best-effort, no guarantee
Ordering	Sequence numbers ensure in-order delivery	Packets may arrive out of order
Error control	Checksum + ACK + retransmit	Checksum only (optional)
Flow control	Sliding window (receiver advertises buffer)	None
Congestion control	Slow start, AIMD, fast retransmit	None (can overwhelm network)
Header size	20–60 bytes (variable due to options)	8 bytes (fixed)
Speed	Slower (overhead, wait for ACKs)	Very fast (fire and forget)
Suitable for	HTTP, HTTPS, FTP, SMTP, SSH — anything requiring reliability	DNS, DHCP, VoIP, video streaming, online gaming, SNMP

TCP Congestion Control — Complete AIMD Algorithm

AIMD — Additive Increase Multiplicative Decrease

TCP's congestion control follows AIMD: increase window additively when no congestion (add 1 MSS per RTT in Congestion Avoidance), decrease window multiplicatively when congestion detected (halve cwnd on 3 dup ACKs; reset to 1 on timeout). This creates the characteristic "sawtooth" pattern of TCP throughput over time.

Phase	Trigger	cwnd behavior	Growth
Slow Start	Connection start or timeout recovery	Doubles every RTT (cwnd += 1 per ACK)	Exponential
Congestion Avoidance	cwnd ≥ ssthresh	Increases by 1 MSS per RTT (cwnd += MSS²/cwnd per ACK)	Linear (additive)
Timeout event	Timer expires (severe congestion)	ssthresh = cwnd/2; cwnd = 1 MSS; restart Slow Start	Drastic reset
3 Dup ACKs (Fast Retransmit)	3 same ACKs received (mild congestion)	ssthresh = cwnd/2; cwnd = ssthresh (TCP Reno: Fast Recovery)	Multiplicative decrease
Fast Recovery (TCP Reno)	After 3 dup ACKs	Skip Slow Start; enter Congestion Avoidance at new ssthresh	Linear from ssthresh

TCP THROUGHPUT APPROXIMATION (Mathis Formula)

Throughput ≈ (1.22 × MSS) / (RTT × √Loss_Rate)

Where:
MSS = Maximum Segment Size (typically 1460 bytes for Ethernet)
RTT = Round Trip Time
Loss_Rate = probability of packet loss

This shows throughput degrades with higher RTT and higher loss rates.

Port Numbers — Well-Known, Registered, Dynamic

Range	Category	Examples
0 – 1023	Well-Known Ports (assigned by IANA, require root/admin)	20/21 FTP, 22 SSH, 23 Telnet, 25 SMTP, 53 DNS, 67/68 DHCP, 80 HTTP, 110 POP3, 143 IMAP, 161/162 SNMP, 443 HTTPS
1024 – 49151	Registered Ports (registered with IANA, apps can request)	1433 MS SQL Server, 3306 MySQL, 3389 RDP, 5432 PostgreSQL, 8080 HTTP alternate, 27017 MongoDB
49152 – 65535	Dynamic/Ephemeral Ports (assigned temporarily to clients)	Automatically assigned by OS to client-side of connections; released after connection closes

Connection Identification: A TCP connection is uniquely identified by a 4-tuple: (Source IP, Source Port, Destination IP, Destination Port). This allows one server to handle thousands of simultaneous connections on the same destination port (e.g. port 80), since each client has a unique source port.

Quality of Service — Mechanisms Deep Dive PYQ 2023-24

Scheduling Algorithms

FIFO (First In First Out): Simple queue, no priority. All flows treated equally. Prone to head-of-line blocking — one slow flow blocks all behind it.

Priority Queuing: Multiple queues with priorities. High-priority queue always served first. Low-priority may starve.

WFQ (Weighted Fair Queuing): Each flow gets a proportional share of bandwidth based on its weight. Prevents starvation. Used in routers for QoS. Fair but complex.

Traffic Control Mechanisms

Traffic Shaping: Delays packets to conform to a rate (smooth bursty traffic). Buffer needed. Leaky bucket is the classic shaper.

Traffic Policing: Drops or marks packets that exceed rate (no buffering). Immediate enforcement. Token bucket is a policer.

RED (Random Early Detection): Router drops packets probabilistically as queue fills — before it's completely full. Signals senders to slow down. Avoids global synchronization (all TCP flows slowing together).

RSVP (Resource Reservation Protocol): Reserves bandwidth along the path for a specific flow. Signaling protocol — routers must support it.

Unit V — Deep Dive

Application Layer Protocols — Complete Coverage

HTTP methods & versions, FTP modes, email security, SNMP MIB, full RSA, data compression algorithms

HTTP — Complete Coverage PYQ Both Years

HTTP/1.0 vs HTTP/1.1 vs HTTP/2

HTTP/1.0: Non-persistent — new TCP connection per object. Very inefficient for pages with many images/CSS/JS files.

HTTP/1.1: Persistent connections (keep-alive) + pipelining (send multiple requests without waiting for each response). Much more efficient. Most widely supported.

HTTP/2: Binary framing, multiplexing (many streams on 1 TCP conn, no head-of-line blocking), server push, header compression (HPACK). Major performance improvement.

HTTP/3: Uses QUIC (UDP-based), built-in TLS 1.3, faster connection setup, no TCP head-of-line blocking.

HTTP Request/Response Format

Request:
GET /page.html HTTP/1.1
Host: www.example.com
Accept: text/html
Connection: keep-alive

Response:
HTTP/1.1 200 OK
Content-Type: text/html
Content-Length: 1234
(blank line)
(body: HTML content)

Cookies: HTTP is stateless — cookies stored client-side allow state persistence. Session cookies expire on close; persistent cookies have expiry date.

HTTP Method	Purpose	Idempotent?	Has Body?	Safe?
GET	Retrieve resource	Yes	No	Yes (no side effects)
POST	Submit data (form, create resource)	No	Yes	No
PUT	Create or replace resource entirely	Yes	Yes	No
PATCH	Partially update resource	No	Yes	No
DELETE	Remove resource	Yes	No	No
HEAD	Same as GET but no body returned (check headers)	Yes	No	Yes
OPTIONS	Query allowed methods for a resource	Yes	No	Yes

FTP — Detailed Architecture PYQ Both Years

SNMP — Complete Architecture ★ BOTH YEARS

SNMP PDU	Direction	Purpose
GetRequest	Manager → Agent	Request value of one or more MIB variables
GetNextRequest	Manager → Agent	Get next OID in MIB tree (used for table traversal)
GetBulkRequest	Manager → Agent	Retrieve large blocks of MIB data efficiently (SNMPv2c+)
SetRequest	Manager → Agent	Set value of a MIB variable (configure device)
GetResponse	Agent → Manager	Reply to Get/Set requests with values or error status
Trap	Agent → Manager	Unsolicited notification of important event (cold start, link down, auth failure)
InformRequest	Agent → Manager	Like Trap but requires acknowledgment (SNMPv2c+)

SNMP Versions: SNMPv1 — basic, weak community string security (plaintext). SNMPv2c — adds GetBulk, 64-bit counters, better error handling, still community string auth. SNMPv3 — adds strong security: authentication (MD5/SHA), privacy/encryption (DES/AES), access control. Used in all modern enterprise network management.

Data Compression — Detailed Algorithms PYQ Both Years

Huffman Coding (Lossless)

Idea: Frequent characters get shorter codes; rare characters get longer codes. Result is a variable-length prefix-free code.

Algorithm:
1. Count frequency of each character
2. Build min-heap (priority queue) of all symbols by frequency
3. Repeatedly extract 2 nodes with lowest frequency; combine into parent (sum of frequencies); re-insert
4. Repeat until single root remains
5. Assign 0 to left branches, 1 to right (or vice versa)
6. Path from root to leaf = code for that symbol

Example: If 'a'=5, 'b'=2, 'c'=1: 'a'→0, 'b'→10, 'c'→11

Run-Length Encoding (Lossless)

Idea: Replace runs of identical characters with (count, character) pair.

Example:
Original: AAABBBBBCCDDDDD (15 chars)
Encoded: 3A5B2C5D (8 chars) → 47% compression!

Very effective for images with large uniform areas (FAX, BMP, TIFF). Used in PCX format and as preprocessing step for JPEG/PNG.

Ineffective for already-varied data (may increase size).

LZW — Lempel-Ziv-Welch (Lossless)

Dictionary-based compression. Builds a codebook dynamically from the input data. Repeated patterns are assigned shorter codes.

Used in: GIF images, ZIP files, PDF, Unix compress command.

No need to transmit the dictionary — receiver rebuilds it from the compressed stream using the same algorithm. Effective for text files and binary data with repeating patterns.

JPEG / MPEG (Lossy)

JPEG (images): Divide into 8×8 blocks → DCT (Discrete Cosine Transform) to convert to frequency domain → quantize (discard high-frequency details) → Huffman encode. Human eye less sensitive to high-frequency detail. Quality setting controls how much is discarded.

MPEG (video): Exploits temporal redundancy between frames. I-frames (full), P-frames (predicted from previous), B-frames (bidirectional). Typically 10–50x compression vs raw video.

Cryptography — Symmetric Algorithms Deep Dive

Algorithm	Key Size	Block Size	Status	Notes
DES	56 bits	64 bits	Broken (1999)	Only 2⁵⁶ keys — brute-forced in <24h today
3DES (Triple DES)	112 or 168 bits	64 bits	Deprecated	Apply DES 3 times: encrypt-decrypt-encrypt. Slow.
AES	128, 192, or 256 bits	128 bits	Current standard	Rijndael algorithm. Fast in hardware & software. Used in WPA2, TLS, disk encryption.
RC4	Variable (40–2048 bits)	Stream cipher	Deprecated	Was used in WEP WiFi (broken) and SSL/TLS (deprecated).
ChaCha20	256 bits	Stream cipher	Modern/active	Used in TLS 1.3, WireGuard VPN. Fast on mobile CPUs without AES hardware acceleration.

COMPLETE RSA WITH EXTENDED EUCLIDEAN — EXAM LEVEL

Step 1: p = 5, q = 11 → n = 55, φ(n) = (5-1)(11-1) = 40
Step 2: Choose e = 3 (gcd(3,40)=1 ✓)
Step 3: Find d: 3d ≡ 1 (mod 40)
  Extended Euclidean: 40 = 13×3 + 1 → 1 = 40 - 13×3
  d = -13 mod 40 = 27 (check: 3×27=81=2×40+1 ✓)
Step 4: Public key=(3,55), Private key=(27,55)

Encrypt M=2: C = 2³ mod 55 = 8
Decrypt C=8: M = 8²⁷ mod 55
  8² = 64 mod 55 = 9
  8⁴ = 9² = 81 mod 55 = 26
  8⁸ = 26² = 676 mod 55 = 16
  8¹⁶ = 16² = 256 mod 55 = 36
  27 = 16+8+2+1 → 8²⁷ = 36×16×9×8 mod 55
  = 576×72 mod 55 = 26×17 mod 55 = 442 mod 55 = 2 ✓

In exam: use small numbers and show each step of modular exponentiation

Complete Exam Preparation Roadmap

Based on BCS603 PYQ 2023-24 & 2024-25 pattern analysis — prioritised for maximum marks in 3 hours

Section A — 2 Marks Each (All Compulsory — 14 marks)

Mesh topology: Links = n(n-1)/2 — for 10 nodes = 45

dB attenuation: 10·log₁₀(Pout/Pin), −3dB ≈ half power

Piggybacking: ACK attached to outgoing data frame to save bandwidth

Contention MAC: suitable for low load, bursty, decentralized networks

OSI model: draw all 7 layers with PDU names and example protocols

Congestion policies: slow start, congestion avoidance, fast retransmit, leaky bucket, token bucket, RED, RSVP, ECN

RPC stub: client stub marshals params; server stub unmarshals and calls function

Token bucket: tokens generated at rate r; packet can send if token available; bursts allowed up to bucket size b

Fixed framing: fixed size blocks (ATM 53 bytes); Variable framing: length field or flag bytes delimit frames

Bridge: learns MAC→port by observing source MACs; forwards/filters frames

Section B — 7 Marks Each (Choose 3 of 5 — 21 marks)

CRC: polynomial → binary → append r zeros → XOR division → remainder = CRC → transmitted = msg+CRC → verify at receiver

Subnetting: /24 → 4 subnets → borrow 2 bits → /26 → list all 4 ranges with network and broadcast addresses

Leaky Bucket vs Token Bucket: define both, draw diagrams, compare (constant vs bursty output), applications

RSA: full 8-step algorithm, numerical example with encryption and decryption verification

SNMP+DNS+Compression: ~2.3 marks each — SNMP manager/agent/MIB/PDUs; DNS hierarchy/resolution/records; compression lossless/lossy algorithms

Section C — Q3–Q7 Recommended Choices

Q3a: Slotted ALOHA diagram + P(idle)=0.1 → G=2.303, S=23%. Q3b: Latency = Trans+Prop+Router delays

Q4a: Count-to-infinity (explain problem + split horizon fix). Q4b: DVR table calculation given 3 neighbour vectors

Q5a: IPv6 — 128-bit, fixed 40B header, no checksum, extension headers, SLAAC, no broadcast, flow label. Q5b: RARP+DHCP(DORA)+VPN

Q6a: Silly Window Syndrome + Nagle (sender) + Clark (receiver). Q6b: TCP header fields + 3-way handshake SYN→SYN+ACK→ACK

Q7a: SMTP+MIME (multimedia email) + IMAP4 vs POP3 advantages. Q7b: FTP (2 connections, active/passive) + HTTP (methods, stateless) + Telnet (port 23, insecure)

Draw From Memory — Exam Diagram Checklist

OSI 7-layer model (layers, PDU names, protocols, devices)

TCP/IP 4-layer model vs OSI comparison

Network topologies: bus, star, ring, mesh, tree (all 5)

IPv4 header (all fields with bit sizes)

TCP segment header (all fields with sizes)

TCP 3-way handshake (SYN, SYN+ACK, ACK with seq numbers)

Leaky bucket and Token bucket (both diagrams)

DNS resolution flow (client→local→root→TLD→authoritative)

Slotted vs Pure ALOHA timeline comparison

CRC long division step-by-step

RSA encrypt/decrypt flow (Alice→Bob with keys labeled)

TCP congestion control sawtooth graph (slow start, CA, events)

COMPLETE FORMULA QUICK REFERENCE — ALL 5 UNITS

Mesh links = n(n-1)/2 | Attenuation: dB = 10·log₁₀(Pout/Pin)
Latency = Propagation + Transmission + Queue + Processing
Propagation = Distance/Speed | Transmission = FrameSize/Bandwidth
Pure ALOHA: S = G·e^(-2G), max 18.4% at G=0.5
Slotted ALOHA: S = G·e^(-G), max 36.8% at G=1 | P(idle) = e^(-G) → G = -ln(P_idle)
Go-Back-N efficiency η = W/(1+2a) | a = Tp/Tt, W = window size
Hamming: 2^r ≥ m+r+1 | Subnetting: hosts = 2^hostbits - 2
DVR Bellman-Ford: D(x,y) = min_v{c(x,v) + D(v,y)}
OSPF Cost = 10^8 / bandwidth
RSA: n=pq, φ(n)=(p-1)(q-1), ed≡1 mod φ(n) | C=M^e mod n, M=C^d mod n
TCP throughput ≈ 1.22·MSS / (RTT·√loss) | Effective window = min(cwnd, rwnd)

ComputerNetworks

Computer
Networks