DCN Midsem — Mock Exam Papers

Message, Sender, Receiver, Transmission Medium, Protocol.

Syntax (structure/format of data), Semantics (meaning of each section), Timing (when to send and how fast).

LAN: privately owned, connects devices in a single office/building/campus, small geographic area. WAN: spans a large geographic area (city/country/globe), often uses third-party carriers to interconnect LANs.

Traffic intensity I = (a·L)/R, where a = average packet arrival rate, L = packet length (bits), R = link bandwidth (bps). It must stay below 1 — as I approaches 1, queuing delay grows without bound.

C = 2·B·log₂(L), where B = bandwidth in Hz, L = number of signal levels.

Maximum throughput = 36.8% (1/e), reached at G = 1.

Any two of: Attenuation (loss of signal energy), Distortion (different frequency components travel at different speeds), Noise (thermal/crosstalk/impulse).

Inserting an extra ESC byte before a FLAG or ESC byte that occurs naturally inside the data, so the receiver can distinguish real data from the frame-boundary flag pattern.

Application, Transport, Network, Data Link, Physical.

FDM: divides the channel into separate frequency bands; all sources transmit simultaneously, each on its own band. TDM: divides time into slots; sources take turns transmitting in their assigned slot on the full bandwidth.

OSI (top→bottom): Application (network-facing services, e.g. HTTP/DNS) · Presentation (encryption, compression, format translation) · Session (dialog control, synchronization) · Transport (process-to-process delivery, ports, TCP/UDP) · Network (logical/IP addressing, routing) · Data Link (framing, MAC addressing, media access) · Physical (raw bit transmission over the medium).

Analogy: Alice writes a letter to Bob. She composes the message (Application), addresses the envelope with Bob's full address (Network layer — stays the same end to end), the postal service routes it through several sorting offices, each relabeling the local delivery instructions at every hop (Data Link — the MAC address changes hop to hop while the IP/postal address does not), and the truck itself just physically carries the paper (Physical).

Encapsulation (OR answer): Data → Segment (Transport layer adds source/destination port numbers) → Packet (Network layer adds source/destination IP) → Frame (Data Link layer adds source/destination MAC + trailer) → Bits (Physical layer converts to signal). Each layer treats the layer above's output as pure payload and wraps its own header around it without inspecting the contents.

(a) Shannon: SNR(dB)=30 ⇒ SNR(linear) = 10^30/10 = 1000. (b) Nyquist: L = 8 ⇒ log2(8) = 3. Note the two answers differ because they answer different questions — Shannon gives the noise-imposed ceiling, Nyquist gives what your chosen number of levels actually achieves on a clean channel; the real usable rate is whichever is lower.

OR answer: Circuit switching reserves a dedicated end-to-end path before data flows (setup phase), guarantees fixed bandwidth, but wastes capacity when idle — e.g. a traditional landline phone call (64 kbps reserved for the whole call). Packet switching sends each packet independently with no reservation, sharing link capacity efficiently for bursty traffic, at the cost of variable delay — e.g. web browsing / WhatsApp messages routed as IP packets.

In Pure ALOHA a station transmits the instant it has data, so a collision occurs if any other station starts anywhere within one frame-time before or after — a window of 2×Tfr. Slotted ALOHA forces all stations to start only at synchronized slot boundaries, so a collision can only happen if two stations pick the exact same slot — the vulnerable window shrinks to just Tfr. Halving the vulnerable window doubles the probability of a clean transmission, which is why slotted throughput is exactly double pure throughput at every G.

OR answer: Reservation — stations reserve a minislot before sending data in the following frame; e.g. VSAT satellite links used by bank branches. Polling — a primary station asks each secondary "Poll" (do you have data?) / "Select" (I have data for you); e.g. old IBM mainframe terminal networks, industrial SCADA systems. Token Passing — a special token circulates among stations and only the current token-holder may transmit; e.g. IBM Token Ring, FDDI.

Mesh gives maximum fault tolerance (a dedicated line per pair) at the cost of the most cabling; Star is simple to manage but the hub is a single point of failure; Bus is cheap to cable but one break anywhere kills the whole segment; Ring passes data/tokens in one direction and typical implementations use a second counter-rotating ring for self-healing. OR answer: Transmission delay = L/R (time to push all bits of a packet onto the link — depends on packet size and link rate). Propagation delay = d/s (time for a bit already on the wire to travel the physical distance, at roughly 2×10⁸ m/s). Example: distance = 2500 km, propagation speed = 2×10⁸ m/s, packet = 5000 bits, link rate = 2 Mbps.

Bit stuffing: whenever the sender's data contains five consecutive 1s, it inserts an extra 0 immediately after them, purely so a genuine run of six 1s (the flag pattern 01111110) can never appear by accident inside real data. The receiver reverses this: after any five consecutive 1s it sees, it removes the next 0 automatically. OR answer (pick any three): NRZ-L: 0 = high voltage, 1 = low voltage (level itself encodes the bit). NRZ-I: 1 = a transition at the start of the bit interval, 0 = no transition. Manchester: a transition always occurs at the middle of every bit (self-clocking); 0 = high-to-low, 1 = low-to-high. Differential Manchester: always has a mid-bit transition for clocking, but the bit value is shown by whether there's an extra transition at the start of the interval (0 = extra transition, 1 = none). AMI: 0 = zero voltage, 1 = alternating positive/negative voltage.

Performance, Reliability, Security, and (for data communication effectiveness specifically) Delivery, Accuracy, Timeliness, Jitter — the standard set taught is Delivery, Accuracy, Timeliness, Jitter.

d_prop = d/s, where d = distance the signal must travel, s = propagation speed of the medium (≈2×10⁸ m/s in copper/fiber).

2×Tfr — a collision occurs if any other station starts transmitting anywhere from one full frame-time before to one full frame-time after your own transmission.

C = B·log₂(1+SNR), where B = bandwidth in Hz, SNR = signal-to-noise ratio (linear, not dB).

Loss of signal energy/strength as it travels through a medium over distance; compensated for using amplifiers/repeaters.

Insert a 0 after any five consecutive 1s in the data, so the six-1s flag pattern (01111110) marking a frame boundary can never appear accidentally inside real data.

Star: only the failing device's own link is affected — other devices keep working (unless the central hub itself fails, which takes everything down). Bus: any break in the single shared backbone cable takes the entire segment down.

"All People Seem To Need Data Processing" — Application, Presentation, Session, Transport, Network, Data Link, Physical.

Each packet is routed independently through the network with no dedicated path reserved in advance, sharing link capacity efficiently across many flows.

FDM — cable TV (each channel on its own frequency band). TDM — T1 phone trunks (24 calls cycling through fixed repeating time slots).

OR answer: Layering separates concerns — each layer solves exactly one problem (e.g. Physical only pushes bits, Data Link only gets a frame across one link, Network only handles end-to-end addressing/routing) and offers a clean, fixed interface to the layer above/below it. This lets each layer's technology evolve independently (you can swap Wi-Fi for Ethernet without touching the Transport or Application layer), makes the system easier to design, test, and troubleshoot, and lets protocols at different layers be mixed and matched (e.g. the same IP layer running over many different Data Link technologies).

Two stuffed 0s were inserted (one per run that reached five consecutive 1s), so the output grew from 13 to 15 bits.

OR answer: Attenuation — signal weakens with distance, e.g. a long unamplified cable run losing strength. Distortion — different frequency components of a composite signal travel at slightly different speeds and arrive out of phase, e.g. signal smearing over a long analog line. Noise — unwanted energy from an external source corrupts the signal, e.g. crosstalk between adjacent cables or thermal noise in the receiver electronics.

13.5% is lower than Pure ALOHA's own maximum of 18.4% — this shows the maximum only occurs at the optimal load G=0.5; pushing more traffic (G=1) past that point increases collisions faster than it increases successful sends, so throughput falls.

OR answer: Circuit switching first runs a setup phase that reserves a dedicated path with fixed bandwidth end-to-end before any data is sent (e.g. dialing a traditional landline call — the exchange reserves a 64 kbps path for the full duration of the call, whether or not anyone is speaking); after setup, data flows with guaranteed bandwidth and no further routing decisions; a teardown phase releases the reserved resources when the call ends.

Star is easier to troubleshoot because every device has its own dedicated link back to the hub — a fault shows up as exactly one dead link, and testing/isolating it doesn't disturb any other device. In Bus, all devices share one physical cable, so a fault anywhere on it (or a bad connector/tap) can silently affect the entire segment, and locating the exact break usually means physically testing cable sections.

OR answer: Message = the information (e.g. an email's text), Sender = the device that creates and sends it (a laptop), Receiver = the device that gets it (a phone), Transmission Medium = the physical path it travels (Wi-Fi radio waves / fiber backbone), Protocol = the agreed rules both ends follow (e.g. SMTP/TCP/IP) so the message is interpreted correctly at the other end.

I = 0.08 is well below 1, meaning the link is lightly loaded — packets arrive far slower than the link can drain them, so queuing delay stays small. As I approaches 1, the queue (and delay) would grow without bound.

OR answer: Reservation: time is divided into intervals; each interval begins with a reservation frame made of minislots (one per station) where a station can flag that it wants to send, and the primary then grants data slots to only the stations that reserved. Polling: works with one primary and several secondaries; the primary sends a Select frame when it has data for a secondary (to alert/wake it up before sending), and sends a Poll frame to ask a secondary whether it has data to send.

C=2B·log2(L) ⇒ 6000=2×1000×log2(L) ⇒ log2(L)=3 ⇒ L=2³=8 levels.

Variation in the delay of arrival of packets belonging to the same flow — even if average delay is fine, uneven arrival spacing (jitter) degrades real-time traffic like voice/video.

All stations are synchronized to a common clock dividing time into fixed-length slots equal to one frame's transmission time; a station with data must wait until the next slot boundary to begin sending, rather than sending the instant it becomes ready.

ports = n-1 = 5.

LAN (Local Area Network), WAN (Wide Area Network), and the Internet (a network of networks).

NRZ-L encodes the bit value directly as a voltage level (0=high, 1=low); NRZ-I encodes it as the presence/absence of a transition at the start of the interval (1=transition, 0=no transition), independent of absolute voltage level.

In (synchronous) TDM, a source's reserved time slot goes unused (wasted) if it has nothing to send at that moment, since slots are fixed per source rather than shared dynamically.

A special control frame (the token) circulates among stations in a defined order; only the station currently holding the token is permitted to transmit data.

The total delay experienced at one node (router/switch) before a packet leaves toward the next hop: processing delay + queuing delay + transmission delay + propagation delay.

Because the stuffing rule forces a 0 to be inserted the instant five consecutive 1s occur, capping every run in the transmitted data at five — so a genuine run of six 1s can only ever be the flag pattern (01111110) marking a frame boundary.

OR answer: In packet switching, each packet carries its own destination address and is routed independently hop by hop as a self-contained datagram — no path is reserved beforehand, so packets belonging to the same message may take different routes and arrive out of order. This is efficient for bursty traffic since idle sources don't waste reserved bandwidth, unlike circuit switching where a fixed path (and its bandwidth) is locked for the entire session regardless of whether data is actively flowing.

The e^-kG term models the probability that no other station transmits during the vulnerable window; since Pure ALOHA's window (2Tfr) is twice as wide as Slotted ALOHA's (Tfr), the exponent carries a factor of 2 for Pure ALOHA and 1 for Slotted ALOHA, directly reflecting "twice the exposure to collision, per unit of offered load G." Both are already well past their individual optimum G (0.5 and 1 respectively), so both are declining from their peaks, but Pure ALOHA collapses far more sharply because its wider vulnerable window punishes higher load much more severely.

OR answer: TCP/IP's single Application layer absorbs OSI's Presentation and Session layers because, in practice, formatting/encryption (Presentation) and dialog/session management (Session) ended up being handled inside application software itself (e.g. TLS and session cookies are implemented by the application/library, not by a separate universal network layer) — TCP/IP was designed pragmatically from working protocols rather than as a theoretical seven-layer model, so it only keeps the layer boundaries that mapped to real, separately-implemented pieces of the stack.

Bus: all devices tap into one shared backbone cable; if a device fails, the rest of the network is unaffected since each tap is a passive connection — but if the cable itself breaks, the whole segment goes down. The two ends of the backbone need terminators to absorb signal and prevent reflections bouncing back down the wire and corrupting data.
Ring: devices are connected in a closed loop, each repeating the signal to the next; if one device (or its ring interface) fails and there's no redundancy, the entire ring is broken because the signal path is a single closed chain — this is why practical ring networks (e.g. FDDI) often use a second counter-rotating ring for self-healing. Ring needs no separate terminators since the loop is inherently closed.

OR answer: A browser request: Application layer creates the HTTP GET request; Transport layer wraps it in a TCP segment with source/destination ports (encapsulation); Network layer wraps that in an IP packet with source/destination IP addresses; Data Link layer wraps that in a frame with MAC addresses for the next hop; Physical layer sends raw bits. At the server, each layer strips (decapsulates) exactly the header its peer layer added, until the original HTTP request is recovered — each layer only ever reads the header meant for it, never the payload above.

Trade-off: Mesh needs far more cabling and ports (45 links vs 10) but survives any single link failure without affecting other pairs — no single point of failure. Star needs drastically less cabling and simpler device hardware (1 port each), but the entire network depends on the central hub; if the hub fails, every device loses connectivity.

OR answer (example — Differential Manchester & AMI): Differential Manchester always has a transition at mid-bit for self-clocking, and encodes the bit value using whether there's an additional transition at the start of the bit interval (0=extra transition, 1=none) — used in classic Token Ring (IEEE 802.5) networks. AMI (Alternate Mark Inversion) represents 0 as zero voltage and 1 as alternating positive/negative voltage pulses — used in T1 telephone trunk lines.

Bit stuffing operates at the level of individual bits — it's used with bit-oriented protocols where the frame boundary flag is a specific bit pattern (01111110); a 0 is stuffed after every five consecutive 1s in the data. Byte stuffing operates at the level of whole bytes — used with byte-oriented protocols where the flag is a specific reserved byte value; an ESC byte is inserted before any byte in the data that accidentally matches the FLAG or ESC byte value. Both exist for the same reason at different granularities: guaranteeing the receiver can never mistake real data for a frame-boundary marker.

OR answer: