Layering reduces complexity - each layer solving a specific set of issues. Each layer offering a "service" to the one above it. Here are some of the services of the layers in the simplified model:
A process in a particular layer is conceptually communicating with its peers in the same layer in other computers. Except for the lowest layer, this communication is actually directed to some process at the level below, which (again) conceptually communicates with its peer in the other computer.
Example In the TFTP example in the above table, the TFTP server and TFTP client are peers. Say the server wants to send N bytes of data to the client. However, what happens is that the TFTP server actually sends the N bytes down to the next layer. The UDP layer adds a 4 byte header to this data, and sends it down to the layer below. The IP layer adds another 8 bytes of header and sends it to the Ethernet layer which adds 20 more bytes of header and trailer. On the client side, the information is passed up from the Ethernet layer to the IP layer after stripping off the 20 bytes of header and trailer. This continues all the way up to the TFTP client. In short, most communication is between adjacent layers.
Why header and trailers? One use is to store the target address of the data (there will be intermediate targets in addition to the ultimate target). Another is the need for multiplexing and demultiplexing in each layer - remember that there may be several other protocols in each layer.