Ethernet

Networks consisting of only physical devices are vulnerable to collisions. Collisions occur when more than one device transmits packets on a network segment at the same time. The main purpose of this layer is to reduce collisions on the physical network.

By far, the most prominent and widespread technology used to connect devices together on the Link Layer today is called Ethernet. Ethernet allows us to form logical boundries around physically connected devices via the concept of network switches or bridges. Switches essentially reduce the amount of machines that can collide in a large network by dividing it up into smaller networks of networks. Any device on the network can reach any other device's networking interface by invoking its MAC address.

MAC addresses are constructed by concatenating six bytes (8-bit hexidecial numbers), for example, "11:22:33:44:55:66".

This means there are 2^48, or over 281 trillion possible MAC addresses. Because there are so many potential MAC addresses, they are theoretically globally unique. The first half of the MAC address doubles as an Organizationally Unique Identifier (OUI), which also helps to ensure uniqueness.

When one device wants to send information to another device on the network, it includes its own MAC address as well as that of the intended receiver in each frame. This is all fine and well, but we quickly encounter a problem: machines do not inherently know the MAC addresses of other machines on their network! This is where the Address Resolution Protocol (ARP) comes in, and we'll learn a little bit about how it works below.

Relevant Note(s): Network Protocols