Packet and circuit switching

Overview

Recall questions on packets and their headers, packet switching and circuit switching compared, out-of-order and lost packets, variable latency, why packet switching suits data traffic, when circuits are better, and hop limits.

What you need to know

Answer each one before you reveal it.

Questions from this subtopic

Each one with the model answer underneath.

1. 1

   What is a packet, and what does its header contain?

   Model answer

   A packet is a small, fixed-limit block into which a larger message is divided for transmission. It has a header, a payload and, at the link layer, a trailer. The header carries the addressing and control information needed to deliver it and put it back together: the source address, the destination address, the sequence number saying where this packet belongs in the original message, the length or protocol identifier, and a counter limiting how many hops it may make before being discarded. The payload is the slice of the actual data being sent. The trailer carries an error-checking value, such as a checksum, so the receiver can tell whether the packet was corrupted in transit.

2. 2

   Describe how packet switching works.

   Model answer

   The sending machine divides the message into packets and gives each one a header containing the source and destination addresses and a sequence number. Each packet is then sent into the network independently. At every router, the packet is examined and forwarded onto whichever outgoing link the routing table currently indicates is best towards the destination, so different packets from the same message can take different routes depending on congestion and link failures. The links are shared: packets belonging to entirely different conversations are interleaved on the same wire. At the destination, the packets are checked for corruption, put back into order using their sequence numbers and reassembled into the original message, with any missing or corrupt packet requested again.

3. 3

   Describe how circuit switching works.

   Model answer

   Before any data is sent, a complete physical path is established from sender to receiver through the network, and the capacity along that path is reserved exclusively for this one connection. The data then flows along that fixed circuit as a continuous stream, in order, taking the same route from start to finish, with no per-packet addressing needed because the path itself determines the destination. When the communication ends, the circuit is torn down and the capacity is released for other users. The traditional telephone network is the classic example.

4. 4

   Compare packet switching and circuit switching.

   Model answer

   Setup: circuit switching needs a call setup phase before any data moves, and a teardown afterwards; packet switching sends immediately with no setup. Route: a circuit uses one fixed path for the whole session, while packets are routed independently and may each take a different path. Capacity: a circuit reserves bandwidth exclusively, so it is guaranteed but wasted whenever the parties are silent; packet switching shares links between all users, so capacity is used efficiently but is not guaranteed. Delivery: circuit-switched data arrives in order with constant delay; packets can arrive out of order, with varying delay, and can be lost. Resilience: if a link on a circuit fails, the whole call drops; packet switching can route around a failed link because the routing decision is made afresh at each hop. Overhead: every packet carries a header, so packet switching wastes some capacity on addressing, while a circuit needs no per-message addressing. Cost: circuit switching needs enough reserved capacity for peak demand and so is expensive; packet switching serves far more users over the same infrastructure.

5. 5

   Why can packets arrive out of order, and how is the original message recovered?

   Model answer

   Because each packet is routed independently. If congestion or a link failure causes the router to send later packets by a shorter or less busy path, those packets can overtake earlier ones; packets also spend different amounts of time waiting in router queues, and a packet that was lost and retransmitted arrives long after the ones that followed it. Recovery is the job of the transport layer at the destination: every packet carries a sequence number, so the receiver buffers what arrives, places the packets in sequence order regardless of arrival order, notices any gap in the numbering, requests retransmission of the missing packet, and only passes the completed, correctly ordered data up to the application.

6. 6

   Why does packet switching produce variable latency, and what kind of application suffers most?

   Model answer

   Because the links are shared and nothing is reserved. A packet's delay depends on how long it waits in the queue at each router, and that depends on how much other traffic happens to be present at that moment; different packets also take different-length routes. The result is jitter - packets arriving at irregular intervals rather than a steady stream. Applications that need data to arrive at a constant rate suffer most: live voice and video calls, and live streaming, where late data is useless because the moment it belongs to has already passed. They compensate with a jitter buffer, deliberately holding a fraction of a second of data before playing it so that late packets still arrive in time, at the cost of a small delay.

Other ways to work this

The same content, four other ways.

What is inside Levo

Six ways to work the same material.