UNIT 4: COMPUTER NETWORKS - EXAM-FOCUSED SHORT NOTES

1. NETWORK MODELS & FUNDAMENTAL ARCHITECTURES

ISO-OSI Reference Model (Frequent)

A 7-layer architectural framework for network communication. Each layer provides services to the layer above it.

Critique: Theoretically elegant but historically slow to implement; protocols were hard to design and deploy. TCP/IP became the practical de facto standard.

TCP/IP Protocol Suite / Model (Frequent)

A 4/5-layer practical model that underpins the modern Internet.

Comparison:

• Layer Mapping: TCP/IP's Application layer is broader; its Link layer is narrower.

• Functionality: OSI strictly separates services, interfaces, and protocols. TCP/IP assumes a robust network layer and focuses on interoperability.

• Implementation: TCP/IP protocols (IP, TCP, UDP) are the Internet's foundation. OSI remains a teaching/reference model.

Network Types & Components (Common)

• PAN (Personal Area Network): Very short range (e.g., Bluetooth). <10m.

• LAN (Local Area Network): Single building/campus (e.g., Ethernet). High speed.

• MAN (Metropolitan Area Network): City-wide (e.g., Cable TV network).

• WAN (Wide Area Network): Country/continent (e.g., Internet). Uses public infrastructure.

• Internetwork: Connection of multiple networks (e.g., The Internet).

• Components: Hosts (end systems), Links (wired/wireless), Intermediate devices (Hubs, Switches, Routers, Gateways).

• Relationships: Peer-to-peer (equal status, e.g., two hosts). Primary-secondary (master-slave, e.g., client-server).

2. PHYSICAL LAYER & TRANSMISSION MEDIA

Transmission Media (Frequent)

Guided (Wired):

Unguided (Wireless):

Channel Capacity & Performance (Frequent)

• Nyquist Formula (Noiseless): Maximum bit rate for a noiseless channel of bandwidth B Hz using L signal levels.

$$R_{max} = 2B \log_2 L \text{ (bps)}$$

> **Example:** B=3kHz, L=4 (2 bits/signal) → R = 2*3000*2 = 12,000 bps.

• Shannon's Theorem (Noisy): Maximum theoretical capacity C of a channel of bandwidth B Hz with signal-to-noise ratio SNR.

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

> **Key:** SNR is often in dB: $$\displaystyle \text{SNR}_{\text{linear}} = 10^{\text{SNR}_{\text{dB}}/10} $$.

> **Example:** B=3kHz, SNR=30dB → SNR_lin=1000 → C ≈ 3000 * 9.97 ≈ 29.9 kbps.

Multiplexing (Common)

• FDM (Frequency Division): Each signal gets a dedicated frequency band. Used in radio/TV broadcasting, traditional telephony.

• TDM (Time Division): Each signal gets a dedicated time slot (time slice) in a repeating frame.

  • Synchronous TDM: Fixed slots, even if a source has no data (inefficient).

  • Statistical TDM: Dynamic slot allocation based on demand (more efficient).

• WDM (Wavelength Division): Fiber optic version of FDM. Multiple light wavelengths (colors) on same fiber. DWDM (Dense WDM) packs channels very closely.

Line Coding & Digital Transmission (Common)

Purpose: Convert digital bits (1s/0s) into physical signals (voltage/light patterns) for transmission over a medium.

TDM Frame Calculation: For n channels each at rate R bps, frame size = n * (bits per channel). Frame rate = R / (bits per channel). Frame duration = 1 / frame rate.

3. DATA LINK LAYER

Functions (Common)

1. Framing: Encapsulating network layer packets into frames (add header/trailer).

2. Physical Addressing: Using MAC addresses in frame header.

3. Error Control: Detection (CRC) and correction (Hamming).

4. Flow Control: Prevent fast sender overwhelming slow receiver.

5. Link Management: Establishing, maintaining, releasing link.

Error Detection & Correction (Frequent)

Detection:

• Parity: Single bit (detect odd # errors). Double parity (2D) for burst errors.

• Checksum: Sum of data words (1's complement). Used in higher layers (TCP/IP).

• CRC (Cyclic Redundancy Check): Most powerful. Uses polynomial division.

  Steps:

  1. Append r zeros to data, where r = degree of generator polynomial G(x).

  2. Divide augmented data by G(x) (modulo-2 division).

  3. Remainder (CRC bits) replaces appended zeros.

  4. Transmit Codeword = Data + CRC.

  5. Receiver divides received codeword by same G(x). Remainder = 0 → No error.

Example: Data 1101011011 (D(x)), G(x)=10011 (degree 4).

1. Append 4 zeros: 11010110110000.

2. Divide by 10011 → Remainder = 1110.

3. Final Codeword: 11010110111110.

Correction:

• Hamming Codes: Add k parity bits to m data bits to correct single-bit errors. Distance d_min=3.

• FEC (Forward Error Correction): Receiver corrects errors without retransmission. Used in noisy/long-delay links (satellite).

Data Link Protocols & Flow Control (Frequent)

Stop-and-Wait:

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

• Efficiency (Link Utilization): $$\displaystyle U = \frac{1}{1 + 2a} $$, where $$\displaystyle a = \frac{\text{Propagation Time}}{\text{Frame Transmission Time}} $$.

• Limitation: Very inefficient for long propagation delays or high bandwidth.

Sliding Window Protocols:

• Go-Back-N (GBN):

  • Sender window size N, receiver window size 1.

  • Cumulative ACKs: ACK n means all frames ≤ n received correctly.

  • Timeout: If ACK not received, sender re-transmits frame n and all subsequent frames in window.

  • Utilization: $$\displaystyle U \approx \frac{N}{1 + 2a} $$ (for large N, no errors).

• Selective Repeat (SR):

  • Sender & receiver window size N (typically ≤ half sequence space).

  • Individual ACKs: Can ACK out-of-order frames.

  • Timeout: Retransmit only the missing frame.

  • Window Constraint: $$\displaystyle N \leq 2^{m-1} $$ (where m = sequence # bits) to avoid ambiguity.

• Piggybacking: Attaching ACK to a data frame going in the reverse direction (instead of separate ACK frame). Improves efficiency in full-duplex links.

Multiple Access Protocols / MAC Sublayer (Frequent)

Static Allocation: FDM, TDM, WDM. Efficient if traffic is constant, inefficient if bursty.

Random Access (Contention-Based):

Controlled Access: Token Passing (Token Ring, Token Bus), Reservation, Polling.

LAN Standards & MAC Protocols (Frequent)

Framing & Stuffing (Common)

• Bit Stuffing: (Flag-based framing). Insert a 0 after any five consecutive 1s in data to avoid confusion with flag 01111110. Receiver removes stuffed 0.

  Example: Data 01111110111 → After stuffing: 01111110 0 111 (flag 01111110 not in data).

• Byte Stuffing: (Character-oriented). Use special escape character (e.g., DLE) before any flag-like character in data.

Data Link Layer Protocols (Common)

• HDLC (High-Level Data Link Control):

  • Frame: Flag (01111110) | Address | Control | Info | FCS | Flag

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

  • Operation: Uses I-frames (data), S-frames (control), U-frames (management).

• PPP (Point-to-Point Protocol):

  • Frame: Flag | Address (0xFF) | Control (0x03) | Protocol | Data | FCS | Flag

  • Phases: Link Establishment (LCP), Authentication (PAP/CHAP), Network Layer Protocol (NCP), Termination.

  • Use: Over serial links (dial-up, DSL).

• SLIP (Serial Line IP): Predecessor to PPP. No error detection, no protocol field, no authentication. Limitation: Only supports IP, no dynamic IP assignment.

4. NETWORK LAYER

Functions & Design Issues (Common)

Packetizing, Routing (path selection), Congestion control, Internetworking (logical addressing), Fragmentation/Reassembly, Error handling (ICMP).

Routing Algorithms (Frequent)

Static vs. Dynamic: Static (manually configured, simple, no adaptation). Dynamic (adapt to topology/load changes, complex).

Distance Vector Routing (DVR - Bellman-Ford):

• Principle: Each router knows distance (cost) to neighbors. Shares its entire distance vector with neighbors periodically. Uses Bellman equation:

$$D_x(y) = \min_{v \in \text{neighbors}} \{ c(x,v) + D_v(y) \}$$

where `D_x(y)` = cost from `x` to `y`, `c(x,v)` = cost to neighbor `v`.

• Process: Iterative, asynchronous. Each router updates its table based on received vectors.

• Problems: Count-to-infinity (slow convergence), routing loops, bad news travels slowly.

• RIP: Uses DVR. Metric = hop count (max 15). Updates every 30 sec. Slow convergence, limited scale.

Link State Routing (LSR - Dijkstra):

• Principle: Each router discovers entire network topology (via flooding of link state packets). Builds a graph.

• Algorithm: Dijkstra's Shortest Path First (SPF).

  1. Start with source node S. Set S cost=0, others=∞.

  2. Pick lowest-cost node N not in tree. Add N to tree.

  3. Update costs of N's neighbors via N.

  4. Repeat until all nodes in tree.

• Advantages: Fast convergence (no loops), global view.

• Disadvantages: High memory/CPU overhead for large networks. Requires reliable flooding.

Comparison:

IP Addressing (Frequent)

Classful Addressing (Legacy):

Limitations: Wasted addresses (e.g., a company needing 500 hosts gets a Class B with 65k hosts), no flexibility, routing table explosion.

CIDR & Classless Addressing:

• Notation: a.b.c.d/n (prefix length n). e.g., 192.168.1.0/24.

• Address Block: All addresses with same first n bits. Size = $$\displaystyle 2^{32-n} $$.

• Aggregation/Supernetting: Combining multiple contiguous prefixes into one larger prefix to reduce routing table size. e.g., 192.168.0.0/24 + 192.168.1.0/24 → 192.168.0.0/23.

Subnetting (Frequent):

• Concept: Divide a large network (e.g., /24) into smaller logical subnets.

• Subnet Mask: Borrow k bits from host part. New prefix = original + k. Mask = 255.255.255.0 → 255.255.255.192 for 2 subnets (k=1).

• Design Problem Steps (Given Network/OriginalPrefix):

  1. Determine k needed: 2^k >= N (required subnets).

  2. New prefix = OriginalPrefix + k. New mask.

  3. Each Subnet Range:

     • Network Address: Base + (i * BlockSize), where BlockSize = 2^{HostBits}.

     • First Host: Network + 1.

     • Last Host: Broadcast - 1.

     • Broadcast Address: Network + BlockSize - 1.

  4. Usable Hosts/Subnet: $$\displaystyle 2^{\text{HostBits}} - 2 $$ (subtract network & broadcast).

Example: 192.168.10.0/24 → 4 subnets (k=2, /26, mask=255.255.255.192).

BlockSize = $$\displaystyle 2^{6} = 64 $$.

Subnet 0: Net=192.168.10.0, Hosts=192.168.10.1 - 192.168.10.62, BC=192.168.10.63.

Subnet 1: Net=192.168.10.64, Hosts=65-126, BC=127.

Subnet 2: Net=192.168.10.128, Hosts=129-190, BC=191.

Subnet 3: Net=192.168.10.192, Hosts=193-254, BC=255.

VLSM (Variable Length Subnet Masking): Apply subnetting recursively. Allocate addresses efficiently (larger subnets first).

IPv4 vs. IPv6 (Frequent):

Network Layer Protocols & Devices (Frequent)

• ICMP (Internet Control Message Protocol): Network layer protocol for error reporting & diagnostics.

  • Messages: Destination Unreachable, Time Exceeded (TTL expired), Echo Request/Reply (ping), Redirect, Source Quench (deprecated).

  • Use: ping (ICMP Echo), traceroute (ICMP Time Exceeded).

• ARP (Address Resolution Protocol): Maps IP address → MAC address on a local network.

  • Operation: Host broadcasts ARP Request: "Who has IP X? Tell Y (MAC)". Owner replies with ARP Reply (unicast).

  • ARP Cache: Stores recent mappings (timeout-based).

  • Need: IP addresses are logical; frames need physical (MAC) addresses for delivery on LAN.

• RARP (Reverse ARP): Maps MAC address → IP address. Used by diskless workstations to discover their IP at boot. Replaced by BOOTP and DHCP.

Network Devices Comparison:

Remote Bridging: Connecting two LANs via a bridge over a WAN link (e.g., leased line). Challenges: High latency (breaks real-time protocols), loop prevention (spanning tree over WAN is slow), cost.

Congestion Control (Frequent)

• Congestion vs. Flow Control: Flow control is point-to-point (sender→receiver). Congestion is global (network-wide resource depletion).

• Causes: High load, slow processors, low bandwidth, bad routing (packet loops).

• Principles:

  • Open-loop (Prevention): Policies to prevent congestion (good routing, admission control).

  • Closed-loop (Feedback & Reaction): Detect congestion, feed back to senders to reduce rate.

• Techniques:

  • Load Shedding: Discard packets when buffer full. Policies: Random, Priority-based (discard low-priority first).

  • Traffic-aware Routing: Avoid congested paths.

  • Admission Control: For virtual circuits (e.g., ATM, MPLS). Deny new connections if network congested.

  • TCP Congestion Control (Closed-loop):

    1. Slow Start: Start with cwnd=1 (congestion window). Double cwnd every RTT until threshold ssthresh.

    2. Congestion Avoidance: After ssthresh, increase cwnd by 1 per RTT (linear).

    3. Fast Retransmit: After 3 duplicate ACKs, retransmit missing packet (no wait for timeout).

    4. Fast Recovery: Set ssthresh = cwnd/2, cwnd = ssthresh + 3, then congestion avoidance.

QoS (Quality of Service) (Common)

Techniques to guarantee performance (delay, jitter, bandwidth, loss).

• Integrated Services (IntServ): Per-flow resource reservation (RSVP). Fine-grained but not scalable.

• Differentiated Services (DiffServ): Per-class service. Mark packets with DSCP (6 bits in IP TOS field) for PHB (Per-Hop Behavior). Scalable.

• Traffic Shaping: Regulate traffic rate (Leaky Bucket, Token Bucket).

• Resource Reservation: Reserve bandwidth/buffers for critical flows.

5. TRANSPORT LAYER

Functions & Services (Common)

Process-to-process delivery (ports), Segmentation/Reassembly, Connection control (estab/term), Flow control, Error control (retransmission), Congestion control.

Transport Layer Protocols (Frequent)

UDP (User Datagram Protocol):

• Connectionless, unreliable. No flow/congestion control, no retransmission.

• Header (8 bytes): Source Port (16), Dest Port (16), Length (16), Checksum (16).

• Applications: DNS, streaming media, VoIP, TFTP, SNMP. Where speed > reliability.

TCP (Transmission Control Protocol):

• Connection-oriented, reliable. Full duplex, byte-stream.

• Header Format (20-60 bytes):

  • Source Port (16), Dest Port (16)

  • Sequence Number (32): Byte number of first byte in segment.

  • Acknowledgment Number (32): Next expected byte (cumulative ACK).

  • Data Offset (4): Header length in 32-bit words.

  • Reserved (3)

  • Flags (9): URG, ACK, PSH, RST, SYN, FIN.

  • Window Size (16): Receiver's advertised window (bytes).

  • Checksum (16)

  • Urgent Pointer (16) (if URG set)

  • Options (variable)

• Connection Management:

  • Three-Way Handshake (Establishment):

    1. Client → Server: SYN=1, seq=x

    2. Server → Client: SYN=1, ACK=1, seq=y, ack=x+1

    3. Client → Server: ACK=1, seq=x+1, ack=y+1

    Why 3-way? Prevents old duplicate connection initiations (security). Both sides agree on initial sequence numbers.

  • Four-Way Handshake (Termination / Graceful Release):

    1. Client → Server: FIN=1, seq=u

    2. Server → Client: ACK=1, ack=u+1 (may still send data)

    3. Server → Client: FIN=1, seq=v

    4. Client → Server: ACK=1, ack=v+1

    Why 4-way? Each side closes independently (full duplex). Graceful release ensures all data is delivered before connection closed (no data loss). TIME_WAIT state (2MSL) handles delayed packets.

Comparison: TCP vs. UDP (Frequent)

6. APPLICATION LAYER

Domain Name System (DNS) (Frequent)

• Role: Hierarchical, distributed database mapping domain names (e.g., www.rgpvonline.com) to IP addresses.

• Components:

  • Resolvers: Client-side library/software (e.g., in OS).

  • Name Servers: Authoritative for a zone. Types:

    • Root Servers: Know TLD servers.

    • TLD Servers: (.com, .org, .in). Know authoritative for domains in TLD.

    • Authoritative Servers: Know IPs for specific domains.

  • Zones: Administrative domain (e.g., rgpvonline.com).

• Resolution Process:

  1. Resolver queries local DNS server (usually ISP's).

  2. If not cached, local server may perform recursive query (asks chain on behalf of client) or iterative query (client/root follows referrals).

  3. Typical iterative: Root → TLD → Authoritative.

  4. Caching: At every level (resolver, local server) with TTL.

• Resource Records (RR): A (IPv4), AAAA (IPv6), CNAME (alias), MX (mail server), NS (name server).

Electronic Mail (Frequent)

• Architecture:

  • MUA (Mail User Agent): User interface (Outlook, Thunderbird).

  • MTA (Mail Transfer Agent): Server-to-server transfer (Sendmail, Postfix). Uses SMTP.

  • MDA (Mail Delivery Agent): Local delivery to mailbox (procmail).

• SMTP (Simple Mail Transfer Protocol):

  • Push protocol. Client (MUA) → Server (MTA) on port 25.

  • Commands: HELO, MAIL FROM:, RCPT TO:, DATA (then message, ends with .), QUIT.

  • Message Format: Headers (From:, To:, Subject:, Date:) + blank line + Body.

  • ESMTP (Extended SMTP): Adds STARTTLS (encryption), AUTH (authentication), SIZE, 8BITMIME for binary/MIME.

• Email Access (Pull): POP3 (download & delete), IMAP (server-side folders, sync).

World Wide Web & HTTP (Frequent)

• WWW Architecture: Client (Browser), Server (Apache, Nginx), Proxy/Cache.

• HTTP (Hypertext Transfer Protocol):

  • Application layer protocol. Connectionless (each request/response uses new TCP connection in HTTP/1.0), stateless (server doesn't remember clients → use cookies).

  • HTTP/1.1: Persistent connections (default, multiple requests/responses on same TCP). Pipelining (send multiple requests without waiting).

  • Request Methods: GET (fetch), POST (submit data), HEAD (headers only), PUT, DELETE.

  • Response Status Codes:

    • 1xx (Informational), 2xx (Success: 200 OK), 3xx (Redirection: 301 Moved Permanently),

    • 4xx (Client Error: 404 Not Found), 5xx (Server Error: 500 Internal Server Error).

File Transfer (Common)

• FTP (File Transfer Protocol):

  • Architecture: Separate control & data connections.

    • Control Connection: TCP port 21. Commands (USER, PASS, LIST, RETR, STOR) and replies.

    • Data Connection: TCP port 20 (active mode: server connects to client's port). Passive Mode (PASV): Client connects to server's high port (firewall-friendly).

  • Operation: Client initiates control connection, sends commands. Data connection opened/closed per transfer.

Network Management (Common)

• SNMP (Simple Network Management Protocol):

  • Components:

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

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

    • MIB (Management Information Base): Database of manageable objects (variables) on agent.

  • Operations (PDU Types): GET, GETNEXT, SET, GETBULK, TRAP (asynchronous alert from agent to manager).

  • Versions: SNMPv1 (no security), SNMPv2c (community string), SNMPv3 (user-based security, encryption).

7. SWITCHING & INTERCONNECTING DEVICES

Switching Techniques (Frequent)

*   **Datagram:** No connection. Each packet routed independently (best-effort). **Network layer** (IP).

*   **Virtual Circuit:** Connection established. All packets follow same path (VC#). **Data link/network layer** (Frame Relay, ATM). | Moderate (per-packet) | No (shared) | **Internet (IP - Datagram).** |

Network Devices (Frequent)

Key: Bridge/Switch = MAC-based, same network. Router = IP-based, different networks. Hub = dumb repeater.

📚 EXAM TIPS & COMMON PITFALLS:

[!TIP] CRC Calculation: Always show polynomial division steps. Remember: Append r zeros, divide by G(x), remainder is CRC. Final codeword = Data || CRC. If remainder ≠ 0 at receiver → error.

[!TIP] Subnetting: Never forget to subtract 2 for network & broadcast addresses when calculating usable hosts. For N subnets, borrow k where 2^k >= N. New prefix = old + k.

[!TIP] Routing Algorithms: In DVR (Bellman-Ford), update happens when a neighbor's vector changes. Show the iteration table. In LSR (Dijkstra), draw the graph, mark tentative/permanent nodes step-by-step.

[!TIP] TCP Header from Hex Dump: Convert hex to binary/decimal. Remember: Source/Dest Port (16 bits each), Seq/ACK (32 bits), Flags (9 bits: URG, ACK, PSH, RST, SYN, FIN), Window (16 bits). Data Offset = header length in 32-bit words.

[!TIP] Stop-and-Wait Efficiency: $$\displaystyle U = \frac{1}{1+2a} $$. For efficiency ≥ 50%, $a \leq 0.5$ → Propagation Delay ≤ Frame Transmission Time.

[!TIP] CSMA/CD Minimum Frame Size: Must be ≥ $2 \times \text{Propagation Delay} \times \text{Bandwidth}$ (in bits). Ensures collision detected before transmission ends.

[!TIP] ALOHA Throughput: Pure: $$\displaystyle S = G e^{-2G} $$ (max 18.4% at G=0.5). Slotted: $$\displaystyle S = G e^{-G} $$ (max 36.8% at G=1). G = offered load (attempts/frame time).

[!TIP] OSI vs TCP/IP: OSI has 7 layers, strict separation. TCP/IP has 4/5 layers, combines OSI's 5-7 into Application, 1-2 into Link. TCP/IP is implementation-focused, OSI is theoretical.

[!TIP] IPv4 vs IPv6: IPv6 has 128-bit addresses, no header checksum, no fragmentation by routers, mandatory IPsec, extension headers. IPv4 has broadcast, fragmentation by routers, optional security.

[!TIP] DNS Resolution: Usually iterative from local server. Root knows TLD servers, TLD knows authoritative for domain. Caching is critical for performance. nslookup/dig are tools.

[!TIP] SMTP vs HTTP: SMTP is push (client pushes to server on port 25). HTTP is pull (client pulls from server on port 80/443). SMTP uses 7-bit ASCII (MIME for binary). HTTP is binary-friendly.

[!TIP] Three-way vs Four-way Handshake: Three-way for connection establishment (both sides agree on ISN). Four-way for termination because TCP is full-duplex; each direction closed independently. TIME_WAIT (2MSL) ensures last ACK received and old duplicates expire.

[!TIP] Switching Device Layer: Hub=1, Bridge/Switch=2, Router=3, Gateway=7. This is a frequent comparison question.