Network infrastructure evolution begins with the store and forward networking model. This model was how early internet adopters (1960s – 1980s) would send messages back and forth to host computers. While being able to send a message across a network infrastructure was a revolutionary computing breakthrough, big deficiencies did not go unnoticed. With this model, a message got sent one at a time. They would get sent through a series of hops from one computer to the next. When a message was received by an intermediary computer, it would be stored there, and then forwarded on to the next computer once the line was open. A big problem was that a long message would clog the system, and drastically slowdown the delivery of other messages waiting in que. Another problem is that there was not a built-in method for dynamically addressing outages in the network.
After more than 20 years of researching ways to address problems in store and forward networking, the idea of packets was innovated. With the notion of packet-switching, a message is broken into small packets. The packets get sent out on the internet to find their way. These packets would also have to traverse a series of hops. However, because messages are broken into smaller packets, it leads to better sharing of resources for transmission of data. Further, packets of the same message are not required to take the same series of hops to reach their final destination. The packets have no regard as to how they find their way, but they do know when all the packets of the message have arrived, and how to assemble back to the complete message.
This notion of packet-switching lead to the shared network infrastructure that we use in our TCP/IP networks today. With this notion, the network of big computers evolved to a shared network of small routers. The main purpose of the routers is to forward packets. Moreover, the existence of a single router would become less relevant than one computer in store and forward networking. In that model, one computer played a critical role in the whole reliability of the network. However, with much more routers setup everywhere with the sole purpose of forwarding packets, it was to become not so critical if one router went offline. There would be other paths available for the packet to be routed through.
However, this problem of reliability was still a big problem. The way you solve a big problem is to break it down to a subset of smaller problems. Then you can focus on solving each smaller problem. Breaking down this problem lead to the layered network model. There were several variations as to how many layers the problem got broken into, but the model that became most popular is the TCP/IP (Internet Protocol Suite) model.
The TCP/IP model consists of four layers. They are Application, Transport, Internet, and Link. So to solve the whole problem of internet reliability, you can focus on one layer at a time. Each layer presents a difficult problem in itself, but it is manageable.
When discussing the evolution of our shared network infrastructure, it must be noted that there is also a model called the 7 layer OSI model. The Open System Interconnection model competed with the TCP/IP model as the preferred model for building out the internet. TCP/IP has won the mind share, but the OSI model remains valid.