22-Elec-B4 Information Technology Networks · December 2018

Question 1 of 5: Packet Switching and Circuit Switching

Nivaar worked solution (AI-drafted; not reviewed by a licensed engineer)

Notes on this paper

Paper format. Professional Engineers of Ontario — National Examinations, December 2018, 16-Elec-B4 Information Technology Networks. Three hours, closed book; one approved Casio or Sharp calculator permitted. The paper prints five questions of 25 marks each, and any four constitute a complete paper worth 100 marks, with the marks for every sub-part shown in the left margin. All five questions are solved here, because this set is a study resource rather than an exam attempt, and a candidate choosing which four to write benefits from seeing the fifth worked out.

Reference texts. A. Leon-Garcia and I. Widjaja, Communication Networks: Fundamental Concepts and Key Architectures, 2nd ed. (the syllabus reference for this code); J. F. Kurose and K. W. Ross, Computer Networking: A Top-Down Approach, 8th ed.; A. S. Tanenbaum and D. J. Wetherall, Computer Networks, 5th ed.; W. Stallings, Wireless Communications and Networks, 2nd ed.; T. S. Rappaport, Wireless Communications: Principles and Practice, 2nd ed.; S. Sesia, I. Toufik and M. Baker, LTE — The UMTS Long Term Evolution, 2nd ed.

Question 1: Packet Switching and Circuit Switching (25 marks)

Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.

Part (a) — What separates the two switching disciplines (5 marks)

A circuit-switched network reserves resources before any user data moves. A signalling phase walks the path from source to destination and commits a fixed share of every link along the way — a wavelength, a frequency band, or a repeating time slot — to that one conversation. The reservation persists for the whole call whether or not anything is being sent, and it is torn down explicitly at the end. Because the path and the capacity are fixed in advance, every bit follows the same route, arrives in order, and experiences the same constant delay; the network performs no per-message processing once the circuit exists. The cost of this guarantee is that an idle circuit still consumes its share of the link, and that a new call must be blocked outright when no free share remains.

A packet-switched network reserves nothing. The source breaks its message into packets, each carrying enough addressing information to be forwarded independently, and hands them to the network. Every switch stores each packet, reads its header, consults a forwarding table and queues the packet on an outgoing link, so links are shared statistically among all users rather than partitioned among them. There is no blocking — a packet is always accepted if buffer space exists — but there is variable queueing delay, packets may arrive out of order or not at all, and the network must be engineered so that the aggregate of the users' bursts rarely exceeds the link capacity. In short: circuit switching pre-allocates and guarantees; packet switching multiplexes on demand and delivers a best effort.

Part (b) — Why bursty traffic belongs on a packet network (5 marks)

Bursty sources are idle most of the time, so a circuit sized for the peak rate sits unused for the great majority of the call and the link's capacity is wasted in proportion to the peak-to-mean ratio. Packet switching applies statistical multiplexing: because the bursts of independent users rarely coincide, one link can carry many more such users than its capacity divided by their peak rate, and each burst momentarily borrows capacity that the others are not using.

Part (c) — Classifying the four technologies (5 marks)

Each item below is judged by whether it presupposes a reserved end-to-end resource or a shared one.

i. TCP/IP — packet switching. IP is a connectionless datagram protocol: each packet carries a full destination address and is routed independently, and TCP supplies the ordering, retransmission and flow control that the packet network itself does not provide. There is no reservation anywhere in the architecture.
ii. Ethernet — packet switching. Ethernet transports variable-length frames with source and destination MAC addresses over a shared or switched medium; stations contend for the medium (or for a switch's output queue) frame by frame rather than holding a reserved share of it.
iii. Public switched telephony — circuit switching. The PSTN is the canonical circuit network: call set-up reserves a 64 kbit/s channel end to end for the duration of the call, delivering constant delay and no loss, and a call that cannot be given a channel is blocked with a busy tone.
iv. FDMA — circuit switching. Frequency-division multiple access assigns each user a distinct frequency band for the duration of its transmission, which is a reservation in the frequency domain and is exactly what a circuit is. (Strictly, FDMA is a multiple-access method rather than a switching discipline, so it can carry packets — but the resource it hands out is a dedicated channel, so its natural use is circuit switching.)

Part (d) — Why quality-of-service guarantees favour circuits (5 marks)

Because the capacity is committed before the first bit is sent, a circuit's bandwidth, delay and jitter are known at set-up time and cannot be degraded by any other user's behaviour; no queue builds, so nothing is lost to congestion and the delay is simply the propagation and framing delay. A packet network can only approximate such guarantees, by admission control and scheduling, because its delays depend on traffic offered by everybody else.

Part (e) — Building a circuit on a TDMA carrier (5 marks)

Given. A TDMA air interface in which time on the carrier is divided into frames of duration $T_f$, each frame being split into $N$ equal slots; a GSM-style frame with $T_f = 4.615$ ms and $N = 8$ is used below as a concrete illustration.

Find. How a connection with a fixed, guaranteed bit rate is realised on such a carrier, and what the resulting slot period is.

Circuit emulation on a TDMA carrier: the call is granted slot 3, and slot 3 of every frame belongs to it until the call is released.

Approach. Reserve one slot position for the whole call, so that the reservation recurs at the frame rate.

Reserve a slot index at call set-up. The base station's call-admission function assigns the connection a slot number $n$ and keeps that assignment for the life of the call. The terminal transmits only in slot $n$ of every frame and listens only in slot $n$; the assignment is released at call tear-down, and a new call is blocked if all $N$ slots are already assigned. This is a reservation in the time domain and is exactly what a circuit is.
The recurrence period fixes the bit rate. One slot arrives once per frame, so the slot duration and the rate at which the circuit is served are $$T_{slot} = \frac{T_f}{N} = \frac{4.615\ \text{ms}}{8} = 0.5769\ \text{ms}, \qquad f_{slot} = \frac{1}{T_f} = 216.7\ \text{slots/s}$$ Because $T_f$ and $N$ are constants of the air interface, the user sees a constant bit rate and a constant delay — the defining property of a circuit: $$\boxed{\text{a circuit} = \text{slot } n \text{ of every frame, for the duration of the call}}$$

In one sentence: allocate the same time slot in every successive TDMA frame to the same connection for the whole call, so the connection receives a fixed number of bits per frame at a fixed instant — a circuit built out of periodic slots instead of a continuous frequency band. GSM does precisely this, and adds a guard time at each slot boundary so that bursts from terminals at different ranges do not overlap.

Topic. Switching disciplines and multiplexing: reservation versus statistical sharing.

Key relations. Statistical-multiplexing gain $G = C/(N_{peak}R_{peak})$ for a link of capacity $C$; Erlang-B blocking for a circuit trunk group; TDMA slot period $T_{slot} = T_f/N$ and per-user rate $R_{user} = b_{slot}/T_f$ for $b_{slot}$ bits per burst.

Why this works. The whole question turns on one trade: a reservation converts an uncertain delay into a certain refusal. A circuit network answers overload by blocking new calls, so the calls already admitted are untouched; a packet network answers overload by queueing and eventually discarding packets, so every flow degrades a little. Which is preferable depends entirely on the traffic. A telephone call offers a steady 64 kbit/s and cannot use more, so a reservation wastes nothing and buys a hard guarantee. A web session offers milliseconds of megabits separated by seconds of silence, so a reservation at the peak rate would leave the link idle for most of the session while statistical multiplexing packs many such sessions onto it. Modern networks blur the line deliberately: MPLS and LTE bearers push circuit-like reservations into a packet core, precisely so that voice can keep its guarantee while data keeps its efficiency.

Common pitfall. Calling FDMA or TDMA a “switching” method: they are multiple-access methods, and the marks come from saying which switching discipline each one naturally supports and why. A second pitfall is asserting that packet switching is simply more efficient — it is more efficient for bursty traffic, and it is measurably worse for a constant-rate, delay-sensitive stream, which is why the same exam asks part (d).

Source. Leon-Garcia & Widjaja, Communication Networks, 2nd ed., ch. 4 (circuit and packet switching) and ch. 7; Tanenbaum & Wetherall, Computer Networks, 5th ed., §2.5.4.

← Paper overview Next: Question 2 →