How would the setup phase work? Using packet routing? Without an agreed-upon protocol, how do we agree upon who gets certain links?
sam
I feel it works the same way as the locks work. Sender will keep on probing the link and if it sees it free it occupies it and may be set some flag to indicate that link is occupied. So setup phase just consist of probing and setting the status for the link.
BigFish
Can someone help explain why packet-based switch cannot have arbitrary message size? It seems like we can break an arbitrary long message into many packets and transmit them.
cgjdemo
I am still confused why circuit-switched routing does not need buffering. If a switch receives too many packets to transfer, does it just discard the packets which it is unable to transfer?
grose
@BigFish, where does it say they can't?
@cgjdemo Precisely what makes circuit-switched routing, circuit-switched, and not packet-based, is that it does not have packets. Only one person can talk at a time, and it goes directly to the other person. So there's no buffering needed on the network. The clients on either end might still need it though.
BigFish
@grose In this slide it says one benefit of circuit-based routing is "Arbitrary message size (once path is set up, send data until done)". So I think since packet-based routing does not set up the path at the beginning, it cannot have arbitrary message size. Maybe this is an incorrect inference.
grose
@BigFish, oh, well, since messages are broken up into packets, I thought you could send an arbitrary-sized message via just splitting it into enough packets.
I think the issue here might be semantic.
BigFish
@grose Agree, thanks a lot!
jcarchi
@BigFish @grose This is the way I saw it. I thought about it like a phone call. Circuit switching we're basically saying the phone call can last an arbitrary amount of time. We don't know how long it will last, but we can keep the connection open for as long as we would like. In packet switching, we sort of need to know how long the phone call will last before putting it though.
HLAHat
So, keep in mind how flow control works for packet switching. Depending on link contention, packets may have to be buffered in a router until a link becomes available. If we have arbitrary sized packets, it is possible that a packet will be too big to fit in a router. So in a packet switched network, we need to know how big a packet can be so we can design routers with enough storage for any packet that may come through the network.
Packet switched networks have no such limitation. The messages being transferred can be as large as they need to be because there is no buffering between the nodes.
toutou
What I think is in previous slice, buffer is needed when there is a contention between 2 packets. In circuit-switched routing, however, there is no contention. Therefore there is no need for buffering.
How would the setup phase work? Using packet routing? Without an agreed-upon protocol, how do we agree upon who gets certain links?
I feel it works the same way as the locks work. Sender will keep on probing the link and if it sees it free it occupies it and may be set some flag to indicate that link is occupied. So setup phase just consist of probing and setting the status for the link.
Can someone help explain why packet-based switch cannot have arbitrary message size? It seems like we can break an arbitrary long message into many packets and transmit them.
I am still confused why circuit-switched routing does not need buffering. If a switch receives too many packets to transfer, does it just discard the packets which it is unable to transfer?
@BigFish, where does it say they can't?
@cgjdemo Precisely what makes circuit-switched routing, circuit-switched, and not packet-based, is that it does not have packets. Only one person can talk at a time, and it goes directly to the other person. So there's no buffering needed on the network. The clients on either end might still need it though.
@grose In this slide it says one benefit of circuit-based routing is "Arbitrary message size (once path is set up, send data until done)". So I think since packet-based routing does not set up the path at the beginning, it cannot have arbitrary message size. Maybe this is an incorrect inference.
@BigFish, oh, well, since messages are broken up into packets, I thought you could send an arbitrary-sized message via just splitting it into enough packets.
I think the issue here might be semantic.
@grose Agree, thanks a lot!
@BigFish @grose This is the way I saw it. I thought about it like a phone call. Circuit switching we're basically saying the phone call can last an arbitrary amount of time. We don't know how long it will last, but we can keep the connection open for as long as we would like. In packet switching, we sort of need to know how long the phone call will last before putting it though.
So, keep in mind how flow control works for packet switching. Depending on link contention, packets may have to be buffered in a router until a link becomes available. If we have arbitrary sized packets, it is possible that a packet will be too big to fit in a router. So in a packet switched network, we need to know how big a packet can be so we can design routers with enough storage for any packet that may come through the network. Packet switched networks have no such limitation. The messages being transferred can be as large as they need to be because there is no buffering between the nodes.
What I think is in previous slice, buffer is needed when there is a contention between 2 packets. In circuit-switched routing, however, there is no contention. Therefore there is no need for buffering.