Logically the ether is a single transmission line connecting each node, similar to a bus used in the backplane of a computer system (Chapter ``Computer Architecture''). Information put on the network by a node is broadcast to every other node. Message passing software in an Ethernet system is based on packet switching, in which a message is broken into small pieces that are transmitted individually and reassembled by the receiver. A communication protocol defines the structure of each packet. For example, the first few bits may contain the ID of the receiving node, followed by the packet number and size (e.g. ``packet 4 of 12, length = 100 bytes''), followed by the data itself. Thus when node A wants to send a message to node B, it breaks it into packets, puts node B's address and the packet description in front of each packet, and broadcasts the packets using the ether. Every node continually monitors the network, looking for packets with their node ID. When B sees the packets addressed to it, it saves them in a buffer and reassembles the message as it arrives. The network consists of a single wire, so only one bit of information can be transmitted at a time. The theoretical peak performance of the network is defined by the number of bits per second that can be transmitted by a node. Actual performance depends on the number of bits used for overhead (packet IDs, etc.), the average packet size (including data plus overhead), the number of systems connected to the network, and the amount of time lost due to collisions when two or more nodes trying to transmit at the same time (see also problem 4).
Figure 3: Ethernet Local Area Networks.