Making the Telco Look Like One Big Whopping Switch

By the end of this first section of this chapter, you'll know the basics of Frame Relay and how it behaves like a LAN switch in many ways. Before a LAN switch can switch frames between PCs or routers, the frames have to be physically connected to the switch. I'll start by explaining the equivalent in Frame Relay. The next section explains how the telephone company (telco) creates a Frame Relay service that switches frames to the routers that are connected to the Frame Relay network. The details of Frame Relay do differ from Ethernet switching, but many of the same general concepts apply.

Cabling a Router to the Big Frame Relay Switch

To make Frame Relay work, each router needs a physical cable between itself and a device called a Frame Relay switch. The telco uses Frame Relay switches in its local central offices (COs) that together switch the data to the correct sites. Like a PC attaches to a LAN switch with an Ethernet cable, a router needs a physical connection to a Frame Relay switch to use Frame Relay.

When a router physically connects to a Frame Relay switch, it is connecting to a Frame Relay service. The company that sells Frame Relay services is called a Frame Relay service provider. Often, a Frame Relay service provider is also a telco, but in most cases, that company works with other telcos to create the Frame Relay network. The service that the provider is selling is the ability for a router to send a Frame Relay frame and have it be delivered to another router that is also connected to the same Frame Relay network.

To create the physical connection between a router and the Frame Relay service provider, the provider needs to run a cable from all your sites to their local COs. Sound familiar? It should. Frame Relay is a set of protocol specifications, all matching the functions of OSI Layer 2. For Layer 1 functionsthings like the cabling and how the bits are actually transmitted over the wireFrame Relay uses the same standards that serial links do, as shown in Figure 15-1.

Figure 15-1. Physical Parts of Frame Relay

To physically connect to a Frame Relay switch, the provider runs the physical equivalent of a leased line between the router and a nearby Frame Relay switch. The telco needs to be able to send and receive data to and from each router, and leased lines do that. Because Frame Relay doesn't define standards for Layer 1 features, such as basic physical transmission and cabling, Frame Relay relies on the same standards that point-to-point WAN links do for the physical cabling and electrical details. Frame Relay standards refer to this physical serial link between a router and a Frame Relay switch as a Frame Relay access link.

Although the cabling and CSU/DSUs are the same as with a leased line, the telco does something different in the CO: It connects the cable to a Frame Relay switch. A Frame Relay switch is any equipment that understands Frame Relay and can forward traffic based on Frame Relay protocols. The provider's collective set of Frame Relay switches, along with the other equipment between them, form that provider's Frame Relay network.

Basic Logic Used by the Big Whopping Frame Relay Switch

To understand LAN switching, you first had to understand something about the set of protocols and standards that together form what we call Ethernet. Likewise, to see how Frame Relay works, you need to know a little about Frame Relay protocols.

The original Frame Relay protocols were defined by a vendor consortium called the Frame Relay Forum. A vendor consortium is a group of vendors that get together and agree to make their products work a particular way, while they wait on standards bodies to formalize a standard. The Frame Relay Forum is one such consortium. Later, the ITU formalized Frame Relay, as did the American National Standards Institute (ANSI).

Like its LAN cousin Ethernet, some of Frame Relay's most important features are addressing, framing, and switching. Figure 15-2 gives me a good backdrop from which to talk about the basics of all three.

Figure 15-2. Frame Relay Switching Using Frame Relay Addresses

When a telco sells you a Frame Relay service, in essence the telco promises to forward Frame Relay frames sent by one router to one of many other routers. Before a router can send a packet, it must add the correct data link header and trailer to the packet. In Figure 15-2, R1 starts by putting an IP packet into a Frame Relay frame, between a Frame Relay header and trailer.

When building the frame, R1 must put the correct address in the header. Each Frame Relay header holds an address field called a data-link connection identifier (DLCI). The DLCI is a 10-bit number, usually written as a decimal number between 0 and 1023. R1 puts a particular DLCI into the Frame Relay header, expecting that the frame will be forwarded by the Frame Relay network to the other routerR2 in this case.

Finally, the provider's Frame Relay network forwards the frame to the other router. To accomplish the task, each Frame Relay switch forwards the frame, based on the DLCI, through the network, until it gets to the router on the other side. It's similar in concept to how a PC might send an Ethernet frame, with a destination MAC address, and the LAN switch forwards the frame to the right destination.

Although the general ideas behind Frame Relay switching and Ethernet switching are similar, the processes do differ a lot when you look at the details. For instance, Frame Relay has a single address field, which is only 10 bits long, as opposed to Ethernet, which has a source and destination address field, each 48 bits (6 bytes) long. And Frame Relay switches must be configured to know where to forward frames with particular DLCIs in their headers, instead of automatically learning addresses and their locations like Ethernet LAN switches do.

When comparing this Frame Relay example with a serial link between two routers, the details are different, but the end result is the same. When R1 wants to forward a packet to R2, regardless of whether you use a leased line and PPP, or Frame Relay, when R2 gets the frame, it soon discards the data link header and trailer and leaves them with the packet. The routers are happy because they can forward packets.

If Two Sites Are Good, Three (or More) Must Be Better

Although you might have found the past few pages fascinating, the real advantage that Frame Relay holds over leased lines isn't obvious until you see an example with at least three Frame Relay sites. With Frame Relay, a router can use its single physical access link to forward traffic to multiple remote routers. If you're hanging with me on the analogy with Ethernet switches, that's an easy concept because the same thing happens with Ethernet. With Ethernet, a PC sends frames with varying destinations, over the one Ethernet cable that connects it to the LAN switch, and the LAN switch forwards the frames based on the destination MAC address. With Frame Relay, the same thing happens, except the Frame Relay switch forwards the frames based on the DLCI. Figure 15-3 shows the basic concept.

Figure 15-3. Frame Relay Switching to Multiple Remote Sites

In Figure 15-3, R1 uses its single physical access link to send data to both R2 and R3. When you order the Frame Relay service, you tell the telco that you want to be able to send data from the Atlanta router (R1) to both Cincinnati (R2) and Boston (R3). To make it work, the telco gives you some documentation that tells you how it will program the Frame Relay switch. A shorthand version of that documentation is shown in the lower-left section of the figure. According to the provider, if R1 sends a frame with DLCI 102, the Frame Relay network forwards the frame to R2 (Cincinnati); if the provider puts 103 in the header, the Frame Relay network forwards the frame to R3 (Boston).

It's Virtually Like a Leased Circuit, So Let's Call It a Virtual Circuit

In Frame Relay lingo, the ability for R1 to send data to R2 over the Frame Relay network is called a virtual circuit (VC). A VC is like a leased circuit, in that exactly two devices can send and receive data using it. It's called virtual to contrast it with a physical leased circuit. And because a VC is typically predefined to always be there, VCs are often called permanent virtual circuits (PVCs).

When you're drawing network diagrams, it is useful to show PVCs separately from the physical cabling, as shown in Figure 15-4. (Frankly, there's no real standard for how to draw a PVC, so I tend to draw a single PVC with an odd-style of linereally three parallel linesas shown in the figure.)

Figure 15-4. Frame Relay PVC Concepts

Typically, the service provider preconfigures all the required details of a PVC. For instance, in Figure 15-3, the switches knew to forward frames with DLCI 102 to Cincinnati, and frames with DLCI 103 to Boston.

One reason that drawing the PVCs is important is that when you order Frame Relay service, you do not have to have a PVC between each pair of routers. For instance, Figure 15-3 shows three sites, and frames going from one site (R1) to the other two sites (R2 and R3). That network does not have to have a PVC between R2 and R3. The choice of which sites need to have a PVC depends on where the network engineers think that traffic needs to flow in the network. When routers use Frame Relay, and there is a PVC between each pair of routers, the PVCs are in a full mesh, as shown in Figure 15-5; when not all routers have a PVC, it's called a partial mesh, as was shown in Figure 15-3.