BSc CSIT (TU) Science Computer Networks (BSc CSIT, CSC258) Question Paper 2074 Nepal

Layered Network Architecture

A layered network architecture organizes the complex task of network communication into a stack of smaller, manageable layers. Each layer performs a well-defined set of functions, offers services to the layer above it, and uses the services of the layer below it. Layers communicate through interfaces, and peer layers on different machines communicate using protocols. This approach provides modularity, abstraction, interoperability, and ease of maintenance (a layer can be changed without affecting others).

OSI Reference Model

The OSI (Open Systems Interconnection) model was developed by ISO. It is a 7-layer reference model that standardizes communication functions:

Functions of Each Layer

1. Physical Layer – Transmits raw bits over the medium. Defines voltage levels, data rate, pin layout, transmission mode (simplex/duplex), and physical topology.
2. Data Link Layer – Provides node-to-node delivery. Handles framing, physical (MAC) addressing, error control (CRC), flow control, and medium access control.
3. Network Layer – Responsible for source-to-destination delivery across multiple networks. Performs logical addressing (IP), routing, and congestion control.
4. Transport Layer – Provides end-to-end (process-to-process) delivery. Handles segmentation/reassembly, port addressing, connection control, flow control, and error control (e.g., TCP, UDP).
5. Session Layer – Establishes, manages, and terminates sessions. Provides dialog control and synchronization (checkpoints).
6. Presentation Layer – Deals with syntax and semantics of data: translation, encryption/decryption, and compression.
7. Application Layer – Provides the interface and services to the end user/application (HTTP, FTP, DNS, SMTP).

Mnemonic (1→7): Please Do Not Throw Sausage Pizza Away.

OSI vs TCP/IP

The OSI Application, Presentation, Session layers map to the single TCP/IP Application layer; OSI Physical + Data Link map to the TCP/IP Network Access (Host-to-Network) layer.

TCP/IP Protocol Architecture (4 layers)

1. Application Layer – Combines OSI's application, presentation and session layers. Provides protocols that user applications use: HTTP, FTP, SMTP, DNS, Telnet, SNMP. Handles data representation and dialog.

2. Transport Layer – Provides process-to-process delivery using port numbers.

   • TCP – connection-oriented, reliable, ordered, with flow & error control (3-way handshake).
   • UDP – connectionless, unreliable but fast and low-overhead.

3. Internet (Network) Layer – Handles logical addressing and routing of packets across networks. Core protocol is IP; supporting protocols include ICMP, IGMP, ARP/RARP. Provides best-effort, connectionless delivery.

4. Network Access / Host-to-Network Layer – Combines OSI's data link and physical layers. Defines how data is physically sent on the network: framing, MAC addressing, and transmission of bits over media (Ethernet, Wi-Fi, PPP).

Thus TCP/IP is the practical realization of the layering concept that OSI describes theoretically.

Routing

Routing is the network-layer process of selecting the best path for forwarding packets from a source to a destination across an internetwork. The routing algorithm running in routers builds and updates a routing table that maps destinations to next hops. Routing can be static (manual) or dynamic (adaptive, e.g., Distance Vector, Link State).

Link State Routing (Dijkstra's Algorithm)

In Link State (LS) routing, every router:

1. Discovers its neighbors and learns their network addresses.
2. Measures the cost (delay/metric) to each neighbor.
3. Builds a Link State Packet (LSP) describing these costs.
4. Floods the LSP to all routers so every router obtains the complete topology (a graph).
5. Runs Dijkstra's shortest-path algorithm locally to compute the shortest path to every other node.

Dijkstra's Algorithm

Initialize: dist[source]=0, dist[v]=∞ for all v≠source
N' = { source }            // set of finalized nodes
while (not all nodes in N'):
    pick node w not in N' with smallest dist[w]
    add w to N'
    for each neighbor v of w not in N':
        if dist[w] + cost(w,v) < dist[v]:
            dist[v] = dist[w] + cost(w,v)
            prev[v] = w

Example

Consider this weighted graph (find shortest paths from A):

      4        1
  A ----- B ----- C
  |       |       |
 1|      2|      5|
  |       |       |
  D ----- E ----- F
      1        3

Edges: A-B=4, A-D=1, B-C=1, B-E=2, D-E=1, C-F=5, E-F=3.

Applying Dijkstra from A:

Shortest distances from A: B=4, C=5 (A→B→C), D=1, E=2 (A→D→E), F=5 (A→D→E→F).

Link state protocols (e.g., OSPF, IS-IS) converge quickly and avoid the count-to-infinity problem of distance-vector routing.

Guided vs Unguided Transmission Media

Guided (wired) media use a physical conductor/path to carry signals; unguided (wireless) media propagate signals through air/space without a physical conductor.

Summary: Guided media (bounded) physically direct the signal and offer reliability and security; unguided media (unbounded) offer mobility and wider coverage at the cost of security and noise immunity.

LAN vs MAN vs WAN

Networks are classified by their geographical span.

Summary: A LAN connects devices over a small area, a MAN spans a city (often interconnecting several LANs), and a WAN interconnects LANs/MANs over very large distances — the Internet being the largest WAN.

Bandwidth

Bandwidth is the range of frequencies a channel can transmit, or equivalently the maximum data-carrying capacity of a link. In the analog sense it is measured in Hertz (Hz); in the digital sense (channel capacity) it is measured in bits per second (bps).

Nyquist's Theorem (Noiseless Channel)

For a noiseless channel, the maximum bit rate is:

where $B$ = bandwidth in Hz and $L$ = number of discrete signal levels.

Example: A 3000 Hz noiseless channel with 2 levels gives $C = 2 \times 3000 \times \log_2 2 = 6000$ bps.

Shannon's Theorem (Noisy Channel)

For a noisy channel, the maximum capacity depends on the Signal-to-Noise Ratio (SNR):

where $S/N$ is the linear SNR (often given in dB, where $\text{SNR}_{dB} = 10\log_{10}(S/N)$).

Example: $B = 3000$ Hz, SNR = 30 dB ⇒ $S/N = 10^{3} = 1000$, so

Note: Shannon gives the theoretical upper bound (accounting for noise); Nyquist tells how many signal levels are needed to reach a target rate. In practice the lower of the two limits applies.

Network Topologies

Topology is the physical or logical arrangement of nodes and links in a network.

1. Bus Topology

All devices connect to a single common backbone cable.

• Advantages: Cheap, easy to install, less cable.
• Disadvantages: Single backbone failure breaks the whole network; collisions; hard to troubleshoot.

2. Star Topology

All devices connect to a central hub/switch.

• Advantages: Easy to add/remove nodes; failure of one link doesn't affect others; easy fault isolation.
• Disadvantages: Central device is a single point of failure; more cable; cost of the hub/switch.

3. Ring Topology

Devices form a closed loop; data travels in one (or both) direction(s).

• Advantages: No collisions (token passing); orderly access; good for moderate load.
• Disadvantages: A single node/link failure can break the ring; difficult to reconfigure.

4. Mesh Topology

Every node connects to every other node ($n(n-1)/2$ links for full mesh).

• Advantages: Highly reliable/fault-tolerant; dedicated links give privacy and no traffic contention.
• Disadvantages: Very high cabling and I/O cost; complex installation.

5. Tree (Hierarchical) Topology

Star networks connected in a hierarchy via a backbone.

• Advantages: Scalable; easy to expand; supports many devices.
• Disadvantages: Failure of the backbone/root affects large segments; heavy cabling.

Summary table:

Framing

Framing is the data-link-layer function of dividing the stream of bits received from the network layer into manageable units called frames, so the receiver can identify where each frame begins and ends. The DLL adds a header and trailer (with addresses and a checksum) around the payload.

Frame boundaries can be marked using character count, byte (character) stuffing, bit stuffing, or physical-layer coding violations.

Byte (Character) Stuffing

Used in byte-oriented protocols. Each frame starts and ends with a special FLAG byte. To prevent a flag pattern occurring inside the data from being mistaken for a frame boundary, an ESC (escape) byte is inserted before any FLAG or ESC byte appearing in the data. The receiver removes the inserted ESC bytes (de-stuffing).

Example (FLAG, ESC are special bytes):

Data:           A FLAG B ESC C
Stuffed (sent): FLAG A ESC FLAG B ESC ESC C FLAG

Bit Stuffing

Used in bit-oriented protocols (e.g., HDLC). The flag is the bit pattern 01111110. To ensure this pattern never appears in the data, the sender inserts a 0 after every five consecutive 1s in the data. The receiver, on seeing five 1s followed by a 0, removes (de-stuffs) that 0.

Example:

Data:    0110111111111100  (six+ consecutive 1s)
Stuffed: 011011111 0 11111 0 100   (a 0 inserted after each run of five 1s)

Key difference: Byte stuffing works on whole bytes using ESC bytes (variable-length, multiple-of-8 overhead); bit stuffing works at the bit level and is independent of character codes.

Stop-and-Wait vs Sliding Window

Both are flow-control / ARQ protocols at the data-link (or transport) layer.

Summary: Stop-and-Wait sends one frame at a time and is simple but wastes capacity on long/fast links. Sliding Window allows several outstanding (unacknowledged) frames, achieving pipelining and much higher utilization, at the cost of larger buffers and sequence-number management.

Connection-Oriented vs Connectionless Services

Analogy: Connection-oriented service is like a phone call — you dial, talk, then hang up, with a guaranteed ordered connection. Connectionless service is like the postal system — each letter (packet) is addressed and sent independently with no guarantee of order or delivery.

UDP (User Datagram Protocol)

UDP is a connectionless, unreliable transport-layer protocol. It provides process-to-process delivery using port numbers but adds no flow control, error recovery, or ordering. It is fast and low-overhead, making it suitable for real-time applications such as DNS, DHCP, VoIP, video streaming, and online gaming where speed matters more than reliability.

Key features:

• No connection establishment (no handshake).
• No acknowledgements or retransmission.
• No sequencing — datagrams are independent.
• Small fixed 8-byte header.

UDP Header Format (8 bytes)

The header has four 16-bit fields:

 0              15 16             31
+----------------+----------------+
| Source Port    | Destination Port|
+----------------+----------------+
| Length         | Checksum        |
+----------------+----------------+
|            Data (payload)        |
+----------------------------------+

Because the header is only 8 bytes (versus TCP's 20+), UDP has minimal overhead.

CIDR (Classless Inter-Domain Routing)

CIDR is an IP addressing scheme that replaces the rigid classful (A/B/C) system with variable-length subnet masking. Instead of fixed network/host boundaries, CIDR allows the prefix length to be any value, written in slash notation as IP/n, where n is the number of bits in the network prefix.

Why CIDR?

• Reduces IPv4 address exhaustion by allocating only as many addresses as needed.
• Enables route aggregation (supernetting), shrinking routing tables.
• Removes the inefficiency of fixed classes.

Notation

For 200.10.0.0/22:

• Prefix length = 22 bits ⇒ host bits = $32 - 22 = 10$.
• Subnet mask = 255.255.252.0.
• Number of addresses = $2^{10} = 1024$ (usable hosts $= 1024 - 2 = 1022$).

Example (Supernetting / Aggregation)

Suppose an ISP must route to four contiguous /24 networks:

192.168.0.0/24
192.168.1.0/24
192.168.2.0/24
192.168.3.0/24

These can be aggregated into a single CIDR route:

because the first 22 bits are common (192.168.000000xx). A router now advertises one route instead of four, reducing routing-table size.

Summary: CIDR uses IP/prefix notation to allocate address blocks of any size and to aggregate multiple networks into a single advertisement, improving address utilization and routing efficiency.