22-Elec-B4 Information Technology Networks · December 2016
Nivaar worked solution (AI-drafted; not reviewed by a licensed engineer)
Paper format. Professional Engineers of Ontario, Annual Examinations — December 2016, 07-Elec-B4 Information Technology Networks. Three hours, closed book, a PEO-approved non-programmable calculator permitted. Five questions of 25 marks each; any four constitute a complete paper worth 100 marks, and the marks are printed in the left margin against every sub-part. All five questions are solved here, because this set is a study resource rather than an exam attempt.
Reference texts.
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) — the structural difference (5 marks). A circuit-switched network reserves resources before any user data moves. A call proceeds in three phases: setup, during which signalling propagates end to end and each switch on the chosen path commits a share of its capacity (a frequency, a time slot, a wavelength, a code); transfer, during which the reserved path belongs exclusively to that call; and teardown, which returns the capacity to the pool. Because the path is dedicated, there is no per-message address information to carry, no queueing inside the network, and therefore an essentially constant end-to-end delay. The price is that the reservation is held whether or not the source is sending: a talker who pauses, or a terminal whose user is reading rather than typing, wastes the full reserved rate.
A packet-switched network reserves nothing. The message is divided into packets, each carrying a header with enough addressing for the network to forward it independently, and each switch or router accepts a whole packet, looks up the destination, queues it, and transmits it on the outgoing link when that link is free. This is statistical multiplexing: many bursty sources share one link and the peak of one fills the silence of another, so a link can carry an aggregate whose summed peak rates exceed its capacity. The price is that queueing is now real — delay varies from packet to packet (jitter), buffers can overflow so packets can be lost, and every packet pays a header overhead. Circuit switching trades utilisation for predictability; packet switching trades predictability for utilisation.
Part (b) — what LTE's packet-only design affects (5 marks). The services most affected are the ones that GSM and UMTS carried on a dedicated circuit-switched bearer: voice telephony, and with it circuit-switched SMS, circuit-switched video calling, and emergency calling. In one sentence: because LTE's radio and core network expose only IP bearers, a voice call can no longer be a reserved 64 kbit/s circuit and must instead be re-engineered as real-time IP media (VoLTE, carried over a dedicated QCI-1 guaranteed-bit-rate bearer with the IP Multimedia Subsystem providing call control), or else be handed back to the legacy 2G/3G circuit domain by circuit-switched fallback.
Part (c) — classifying four technologies (5 marks).
The pattern worth naming is that the physical-layer multiple-access schemes (CDMA, TDMA, and FDMA alongside them) are mechanisms for dividing a medium into dedicated channels, and so are the natural substrate of circuit switching, whereas the protocols that carry addressed, variable-length units (TCP/IP, Ethernet) are packet switching. Modern systems layer one on the other — HSPA and LTE schedule packets on top of a code- or time-divided air interface — but the classification asked for here is the native one.
Part (d) — traffic with no service guarantee (5 marks). Packet switching. Traffic with no delay or rate guarantee is almost always bursty — file transfers, web fetches, backups, telemetry — and a reserved circuit sized for its peak sits idle for most of the connection, so the circuit-switched design would either waste most of the network or block most of the users. Statistical multiplexing lets the network carry many such sources on the same link and hand each of them the whole link when it happens to be the only one active, which is both a far higher utilisation and a far shorter transfer time for the individual burst. The variable delay and occasional loss that packet switching introduces cost nothing here precisely because no guarantee was required, and end-to-end retransmission repairs the loss.
Part (e) — a constant rate held for a long time (5 marks). Circuit switching — or, more precisely, circuit switching is the better fit and packet switching can only equal it by imitating it. A constant-bit-rate source that runs for hours fills its reservation continuously, so the statistical multiplexing gain that justifies packet switching disappears: there is no silence for another source to exploit. Against that, the circuit gives constant delay with no jitter buffer, no per-packet header overhead, and no possibility of congestion loss, all of which a packet network must work to recover. A packet network can serve such a flow well, but only by reserving capacity along the path (an LTE guaranteed-bit-rate bearer, an MPLS label-switched path with bandwidth reservation, a DiffServ expedited-forwarding class), which is a circuit re-created in software.
