Previous | Next --- Slide 37 of 51
Back to Lecture Thumbnails

The nice aspect about circuit switching is that there is no overhead due to dynamic switching logic during transmissions. I guess it is useful when you have a big amount of data to be transfered.


As said in lecture, a circuit switch network can be thought of as a telephone where the cost is paid upfront and the entire channel of communication is reserved at the beginning. In a packet switching network, messages are divided into smaller units called packets and the packets get sent through the network however the network routes them based on the destination address.


I have a question regarding packet switching. We see that each packet in a message may take a distinct route under this protocol. Since different routes have different lengths, is it possible that the receiver receives packets in the wrong order under this protocol?


@chunagxuean Under this circumstance, each packet should have an number which can be used to restore the original order.

"Before each packet is sent, it is given a 'header' containing the network IP address that it needs to arrive at and also details of the IP address from which it was sent. The header also gives each packet a number and records how many packets the data was split up into."


@chuangxuean Found a nice gif on wikipedia showing what happens:


It seems like packet switching gives us more flexibility in making routing decisions versus the circuit switching which ensures the message will go directly to the receiver, but requires blocking off a whole route from the beginning.


It seems reasonable that large companies may actually want circuit switching for high uptime services that are always between two static locations. For example, if Google wants to keep two data-centers in-sync and there is high consistent traffic, it may be better to use circuit switching. Another example: high frequency trading companies may want a direct circuit switch connection to the stock market servers.

I'm not sure how much circuit switching is ACTUALLY used in reality though...


Going off of @stl's post before: a real-life example of a circuit-switching network is the telephone network, and an example of a packet-switching network is the Internet.


Can someone elaborate on the point of setup/tear down overhead?