Before we look at some basic Frame Relay configurations, we need to go over a few important topics and terms. Figure as below shows two routers that connect to each other through a Frame Relay network.

Here are some terms you'll need to understand when working with Frame Relay:

Frame cloud

When connecting to a public frame network, you only know about your points of entry to the network; the interior of the network (the "cloud") is invisible to you. The network itself is often called a cloud, because you can't see what happens to your packets after they enter it.

Virtual circuit (VC)

A virtual circuit is a logical connection created by the frame provider from Point A to Point B across a frame cloud.

Data Link Connection Identifier (DLCI)

A DLCI is a value assigned by the frame provider to identify a virtual circuit. In other words, it's the Frame Relay equivalent of an address. DLCIs are unique only locally. That is, your router will have a unique DLCI for each virtual circuit it uses from one Frame Relay provider. However, as far as the Frame Relay provider is concerned, DLCIs are just numbers; the provider can reuse DLCIs throughout its network.

The router maps IP addresses to DLCIs so that it can communicate with a remote router by using the appropriate DLCI. There are two ways to map a DLCI to an IP address. First, you can allow the router to discover the DLCI by using inverse ARP, which is enabled by default. Second, you can explicitly map an IP address to a DLCI.

Local Management Interface (LMI)

The LMI is based on the type of Frame Relay switch you are connecting to. Your provider will give you this information. The LMI types are Cisco, Ansi, and q933a; Cisco is the default. Note that the routers at each end of the link may have different LMI settings, because they are connected to different types of switches.

Point-to-point

A point-to-point connection is a single virtual circuit that connects two points. In Figure 6-1, Router 1 connects to Router 2 with a frame network between them. On either side of the frame cloud is a router that knows that there is only one router at the other end. This kind of configuration is similar to connecting two routers directly over a serial line.

Multipoint

In a multipoint network, a single interface is connected to multiple virtual circuits with multiple DLCIs. Each virtual circuit is still point-to-point, but many logical point-to-point connections share the same physical interface. Subinterfaces should be used for each fully-meshed portion of the multipoint network. Remember that subinterfaces use the X.Y notation, where X is the interface and Y is the subinterface.

There are two types of multipoint networks: partially-meshed and fully-meshed. In a fully-meshed network, all the routers have direct connections to each other. In contrast, in a partially-meshed network, each router is connected to at least one other router, but may not have a direct connection to all the routers in the network. For example, you might have three routers, A, B, and C; Routers B and C are connected to Router A, but do not have a direct connection to each other.

Split horizon

Split horizon is a technique commonly used in routing protocols; it means that the router will not send information about a route out the same interface from which it learned the route. Split horizon is normally used to prevent routing loops. However, it can cause problems in a partially-meshed multipoint Frame Relay network. More than one router may be listening at the other end of any interface. Therefore, we don't want to suppress route announcements. For example, assume that we have three routers (i.e., three virtual circuits) connected to our multipoint interface. If a route comes to our interface from any of those points, we want to announce the route to the other two points. If split horizon is enabled, we can't send the route out our interface because that is where the route originated. However, split horizon should be enabled on a fully-meshed multipoint Frame Relay network.