ARP operates at the Data Link layer of the OSI model, and its purpose is often overlooked. When Host A wants to send data to Host B, we know Host A must have a destination IP address for Host B (the Network layer address). What we tend to overlook is that Host A must also have a Layer 2 destination address for Host B -- in other words, it's got to have a destination MAC address for Host B. That's where ARP comes in.

Host devices, as well as switches and routers, keep a table that maps IP addresses to MAC addresses. This is an ARP Cache, and in our scenario Host A would first look in its own ARP Cache to see if it has a corresponding MAC address for Host B. If not, Host A will send out an ARP Request. This request is a broadcast, which means that every single device on the network segment will get it. (The broadcast MAC address is ff-ff-ff-ff-ff-ff.)

The ARP Request contains the IP address of the destination host, in this case Host B. The Request is basically saying, "Here's the IP address of the host I need to talk to. Will the device with this IP address send me its MAC address?"

Every host on the segment will receive the Request, but only the device with the IP address contained in the Request will respond to it. In this example, Host B will send a unicast ARP Response to Host A, telling Host A what its MAC address is. Host A now has the IP and MAC address for Host B, and can now send data successfully to Host B.

Since the ARP Request is a broadcast, there is no problem if there is a hub, repeater, or switch physically between Host A and Host B, since those devices forward broadcasts. Routers do NOT forward broadcasts, however, so if there is a router between Host A and Host B, there could be a problem. There is an ARP feature that addresses this problem, though, and we'll talk about that in my next Network+ exam tutorial!