Question 5 of 5: Packet Switching and Circuit Switching

Question 5: 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) — the difference. In a circuit-switched network the resources needed by a conversation are reserved before any data flows. A setup phase selects a path and dedicates capacity on every link along it — a frequency band, a recurring time slot, a wavelength — and that capacity belongs to the call for its whole duration whether or not anything is being sent; the call is torn down at the end. Because the path and the rate are fixed, the delay is constant, jitter is negligible, the bits arrive in order and there is no possibility of congestion once the call is admitted. The cost is that idle capacity is wasted, that a call may be blocked outright when no free circuit exists, and that setup takes time.

In a packet-switched network no reservation is made. Data is divided into packets, each carrying a destination address in its header, and each packet is forwarded independently, queued behind other traffic and served on a store-and-forward basis at every node. Capacity is shared statistically, so a link can carry many more intermittent users than it has nominal channels, but delay becomes variable, packets may be reordered or dropped when a buffer fills, and heavy load degrades everyone's service rather than blocking new users. Circuit switching therefore trades efficiency for predictability, and packet switching does the reverse.

Part (b) — which needs centralized scheduling. Circuit switching requires it. Reserving a specific channel on every link of an end-to-end path can only be done by an entity that knows which channels are already committed, so signalling and admission control must schedule the resource centrally before the call is admitted; packet switching needs no such authority, because each router makes an independent local forwarding decision for each packet as it arrives.

Part (c) — classifying the four techniques.

FDMA and TDMA are worth contrasting deliberately: they divide the medium along different axes, frequency against time, but both hand a user a guaranteed share that persists until the call ends, and that persistence — not the axis — is what makes them circuit-mode techniques. Modern systems blur the line, since a packet-switched service can be carried inside TDMA slots that are re-allocated frame by frame, but in the plain form named here both are circuit techniques.

Part (d) — traffic requiring quality-of-service guarantees. Circuit switching, because reserving capacity end to end before the data flows is what makes a guarantee possible: the bandwidth, delay and jitter are then fixed by construction and cannot be eroded by other users. A packet-switched network can approximate the same guarantee only by adding reservation and scheduling machinery on top — admission control with RSVP, or differentiated-services queueing — which is in effect re-introducing a circuit.

Part (e) — bursty traffic. Packet switching, because a bursty source is idle most of the time, and a reserved circuit sized for its peak rate would sit unused for most of the call while still blocking other users. Statistical multiplexing lets many such sources share the same link, since their peaks rarely coincide, and it is the reason data networks converged on packet switching in the first place.