Why is the bandwidth of the pre-established reserved link divided on multiple new connections?

Question

I am reading Kurose Computer Networking book 8e, on page number 57, under 1.3.2 Circuit Switching

Because it link has 4 circuits, for each link used by the end-to-end connection, the connection gets 1/4 of the link's total transmission capacity for the duration of connection.

In the context of circuit switching, each link will be reserved for the network session along with transmission rate but the system allows multiple new connections to establish. For example, even a telephone call is going on, you can still call another person from different phone number 

A <----> B
C <----> D

Series of questions

1. Why does the transmission rate in the reserved link gets divided when more concurrent connections are established?
2. Why does adding more will interfere to that particular link?
3. In the above example let's say if there is a new host establishes from the bottom left side, and it establishes some another link, why does the it will drop the bandwidth of already established link to 1/4? Also shouldn't it be 1/n where n is number of active links?
4. How does the concept of "guaranteed bandwidth" differ from the actual bandwidth used by a single connection?
5. How does reserving bandwidth in circuit switching affect the network's scalability compared to packet switching, where bandwidth is shared dynamically?

Answer 1

1. It's actually the other way around: a connection requires bandwidth L. For the network to support multiple connections, the design bandwidth (on the crucial nodes) is L * N, with N being the maximum number of supported connections (on the crucial nodes).
2. Setting up new connections allocates their bandwidth within the network (along the determined path). If the network bandwidth isn't sufficient, no more connections can be established.
3. New connections do not interfere with already established ones (see 1. and 2.).
4. Each connection allocates a fixed bandwidth from the network. The actually used bandwidth within a connection is irrelevant.
5. One of the major drawbacks in circuit switching is that is reserves a fixed bandwidth in the network each time a connection is established. Whether that connection actually uses that bandwidth doesn't matter. In a packet-switching network, only actually transmitted data is using bandwidth, there's no preallocation.

The difference becomes apparent when you think of phone calls. In a circuit-switched network to support 1 million calls of 64 kbit/s each (duplex), you'd need a capacity of 64 Gbit/s in your network.

If you consider that only one party is talking at any time and there are additional pauses in conversations, you statistically need to actually transmit only perhaps 24 kbit/s per call. Accordingly, you could save 60% and build a network for just 24 Gbit/s or spend the same and support 2.6 million simultaneous calls.

If you also consider connections with a wide variety of bandwidths by type and even along their duration, the packet-switching advantage grows. Yet another plus is the simplicity of its forwarding nodes which may be entirely stateless, in contrast to circuit-switching nodes.

The only real downside of packet switching is its inherent inability to guarantee bandwidth for real-time applications without complex mechanisms (see QoS).