Frame Relay Overview
Connection to a frame relay network is done with a local loop from the serial interface of a router to one of a service provider’s frame relay switches. Communication across a frame relay network uses virtual circuits, which are built by a service provider from a router’s serial interface, through a collection of frame relay switches, to another router’s serial interface. Virtual circuits that are programmed into a service provider’s network to stay active all the time are called permanent virtual circuits (PVCs). IOS also supports switched virtual circuits (SVCs), which become active only when they are used; however, SVCs are not yet widely available from frame relay service providers. We use only PVCs in this chapter.
Many PVCs can be built on a single local loop. PVCs are addressed with Data Link Connection Identifiers (DLCIs) at layer 2. From our perspective, each PVC has two DLCIs—one at each end. From a router’s perspective, each PVC needs only one DLCI—the local one. When a router wants to transmit a packet to another router across a PVC, the router must know the local DLCI of the PVC on which the packet is to be transmitted. For this reason, some people say that DLCIs are locally significant.
Figure 1 shows a basic frame relay network. There are three routers—Dallas, FortWorth, and Austin. Each router has a local loop to the frame relay network. There are two PVCs—one from Dallas to FortWorth and one from Dallas to Austin. Let us take a closer look at the PVC between Dallas and FortWorth. The Dallas end of the PVC has DLCI 100 and the FortWorth end has DLCI 101. These DLCIs, since they are on different local loops, do not have to be different, but they usually are, anyway. When Dallas wants to send a packet to FortWorth, Dallas must transmit the packet out the serial interface that contains the PVC, and the frame header must contain the local DLCI, 100. The switch knows that the path of the PVC with DLCI 100 on the Dallas side is supposed to go to the switch connected to FortWorth. The switch connected to FortWorth knows the path of the PVC extends across the local loop to FortWorth and the DLCI of the PVC on the local loop is 101. The switch puts the DLCI 101 into the frame header so the FortWorth router knows, when it receives the frame, that the frame came in on the PVC with local DLCI 101.
Figure 1 - Basic Frame Relay Network
Dallas has two PVCs coming in on the same local loop; therefore, these PVCs must have different local DLCIs. The frame relay topology shown in Figure 1 is called a partial-meshed network because not all of the routers have PVCs to all of the other routers. We could also call this particular topology a hub-and-spoke network because there is one router (the hub) that has a connection to each of the other routers (the spokes), and traffic from a spoke router must go through the hub to reach another spoke router. A fully meshed network has PVCs running between all of the router pairs; to make the Figure 1 frame relay topology into a fully meshed network, we would have to add a PVC between FortWorth and Austin.
Routers and switches maintain contact with each other using Local Management Interface (LMI). About every 10 seconds, routers and switches send an LMI keepalive across the local loop. If a router is receiving LMI keepalives from a switch, the router makes the line protocol of its interface up so the interface state will be up/up. If a Cisco router interface is connected to a frame relay network and the state of the interface is up/up, the router has a communication path to a frame relay switch. This has nothing to do with being able to reach another router on the other end of a PVC. By default, a Cisco router uses LMI to request a status report from the switch every six keepalives (about once a minute). The status report contains a list of each of the local loop’s PVCs, their DLCIs, and their status.
There are two types of LMI that are widely used between routers and switches: Annex D and Gang of Four. Annex D is from the American National Standards Institute (ANSI). The Gang of Four LMI was jointly developed by Cisco, DIGITAL, Intel, and Stratacom. The router and the local switch must agree on the type of LMI they will use between them.
Frame Relay Configuration
There are two ways of configuring frame relay on a Cisco router. The first configuration method uses the classic, frame relay nonbroadcast multiaccess (NBMA) model. In this configuration, the frame relay network is treated as a multiaccess network like a LAN; however, unlike a LAN, a frame relay network has no broadcast capability. There is no frame relay broadcast address. All of the routers connected to the NBMA network share a network address such as an IP subnet address or an AppleTalk cable range. The second configuration method involves treating each of the PVCs as a separate logical point-to-point network, which is done by creating a sub interface for each PVC. The sub interface method requires more network addresses because each PVC has its own network address.
Of the two methods, the sub interface method is usually the recommended one. Both configuration methods are briefly described in the following sections.
We are going to move our IP traffic from the point-to-point serial links to a frame relay network. To do this, we are going to do something that is not normally recommended in a production network. We are going to remove the IP addresses from the point-to-point serial links, and we are going to stop the current IP routing protocol, OSPF, which was configured in
Frame Relay Sub interface Configuration and Verification
The concept of a sub interface is the same now with frame relay; it is just a logical interface that is directly associated with a physical interface. With frame relay sub interface configuration, we can create a sub interface for each of the PVCs coming into a serial interface.
Two types of sub interfaces can be created for frame relay—multipoint and point-to-point. A multipoint sub interface can handle multiple PVCs; its use is similar to that of the NBMA network that we saw in the preceding section. A point-to-point sub interface (the type we are going to use) effectively turns every PVC into a point-to-point network with its own network addressing. Using point-to-point sub interfaces gives us greater control over our frame relay network.
We are going to implement the configuration illustrated in Figure 14. Since Dallas has two PVCs, we are going to create sub interfaces on Dallas. FortWorth and Austin will continue with the original NBMA configuration since they have only one PVC each; however, Austin’s Serial2 IP address will have to change since it will be connected to a different network created by the Dallas point-to-point sub interface.
Figure 14: Frame-relay sub interface internetwork.
Figure 15 shows the commands necessary to change from the NBMA configuration to a point-to-point sub interface configuration. When using sub interfaces, the physical interface normally does not have any network addressing; therefore, we removed the IP address from Serial2 (Line 4). We created the first sub interface, Serial2.1, by referencing it on an interface command (Line 5). The point-to-point keyword is used to create a point-to-point sub interface. The other option is to create a multipoint sub interface by specifying the keyword multipoint. Point-to-point and multipoint sub interfaces cannot be created when the physical interface’s encapsulation is at its default, HDLC; therefore, we had to have the encapsulation frame-relay command on the Serial2 interface to be able to specify that we wanted a point-to-point sub interface. The encapsulation frame-relay command was issued on Serial2 during the NBMA configuration.
Frame Relay Configuration Summary
For frame relay NBMA configuration, all we need is the encapsulation frame-relay command and the appropriate network protocol commands, such as ip address or decnet cost, on the physical interface connected to the frame relay network. In an NBMA environment, we may run into connectivity problems caused by split horizon (see Section 2.2.2.1, “Distance Vector Routing Protocols”).
The recommended frame relay configuration makes use of sub interfaces. All we have to do is issue the encapsulation frame-relay command on the physical interface, create a sub interface for each PVC, use the frame-relay interface-dlci command to assign a DLCI to each sub interface, and issue the appropriate network protocol command(s) on each sub interface. Using point-to-point sub interfaces removes the split horizon problems sometimes experienced in an NBMA environment.
Frame relay is currently very popular because it is usually less expensive than normal leased lines and because it is widely available. Frame relay is usually less expensive than leased lines because when we get a leased line, our service provider provides us with dedicated bandwidth, and when we get a frame relay PVC, our service provider provides us with just a share of their bandwidth (no dedicated). Our traffic has a greater chance of being dropped traversing a frame relay network than a leased line—yet another reason for the lower cost. Most companies are willing to accept these shortcomings because of the monetary savings, and most applications are not greatly affected as long as the frame relay configuration is properly implemented on both the user side and the service provider side
Connection to a frame relay network is done with a local loop from the serial interface of a router to one of a service provider’s frame relay switches. Communication across a frame relay network uses virtual circuits, which are built by a service provider from a router’s serial interface, through a collection of frame relay switches, to another router’s serial interface. Virtual circuits that are programmed into a service provider’s network to stay active all the time are called permanent virtual circuits (PVCs). IOS also supports switched virtual circuits (SVCs), which become active only when they are used; however, SVCs are not yet widely available from frame relay service providers. We use only PVCs in this chapter.
Many PVCs can be built on a single local loop. PVCs are addressed with Data Link Connection Identifiers (DLCIs) at layer 2. From our perspective, each PVC has two DLCIs—one at each end. From a router’s perspective, each PVC needs only one DLCI—the local one. When a router wants to transmit a packet to another router across a PVC, the router must know the local DLCI of the PVC on which the packet is to be transmitted. For this reason, some people say that DLCIs are locally significant.
Figure 1 shows a basic frame relay network. There are three routers—Dallas, FortWorth, and Austin. Each router has a local loop to the frame relay network. There are two PVCs—one from Dallas to FortWorth and one from Dallas to Austin. Let us take a closer look at the PVC between Dallas and FortWorth. The Dallas end of the PVC has DLCI 100 and the FortWorth end has DLCI 101. These DLCIs, since they are on different local loops, do not have to be different, but they usually are, anyway. When Dallas wants to send a packet to FortWorth, Dallas must transmit the packet out the serial interface that contains the PVC, and the frame header must contain the local DLCI, 100. The switch knows that the path of the PVC with DLCI 100 on the Dallas side is supposed to go to the switch connected to FortWorth. The switch connected to FortWorth knows the path of the PVC extends across the local loop to FortWorth and the DLCI of the PVC on the local loop is 101. The switch puts the DLCI 101 into the frame header so the FortWorth router knows, when it receives the frame, that the frame came in on the PVC with local DLCI 101.
Figure 1 - Basic Frame Relay Network
Dallas has two PVCs coming in on the same local loop; therefore, these PVCs must have different local DLCIs. The frame relay topology shown in Figure 1 is called a partial-meshed network because not all of the routers have PVCs to all of the other routers. We could also call this particular topology a hub-and-spoke network because there is one router (the hub) that has a connection to each of the other routers (the spokes), and traffic from a spoke router must go through the hub to reach another spoke router. A fully meshed network has PVCs running between all of the router pairs; to make the Figure 1 frame relay topology into a fully meshed network, we would have to add a PVC between FortWorth and Austin.
Routers and switches maintain contact with each other using Local Management Interface (LMI). About every 10 seconds, routers and switches send an LMI keepalive across the local loop. If a router is receiving LMI keepalives from a switch, the router makes the line protocol of its interface up so the interface state will be up/up. If a Cisco router interface is connected to a frame relay network and the state of the interface is up/up, the router has a communication path to a frame relay switch. This has nothing to do with being able to reach another router on the other end of a PVC. By default, a Cisco router uses LMI to request a status report from the switch every six keepalives (about once a minute). The status report contains a list of each of the local loop’s PVCs, their DLCIs, and their status.
There are two types of LMI that are widely used between routers and switches: Annex D and Gang of Four. Annex D is from the American National Standards Institute (ANSI). The Gang of Four LMI was jointly developed by Cisco, DIGITAL, Intel, and Stratacom. The router and the local switch must agree on the type of LMI they will use between them.
Frame Relay Configuration
There are two ways of configuring frame relay on a Cisco router. The first configuration method uses the classic, frame relay nonbroadcast multiaccess (NBMA) model. In this configuration, the frame relay network is treated as a multiaccess network like a LAN; however, unlike a LAN, a frame relay network has no broadcast capability. There is no frame relay broadcast address. All of the routers connected to the NBMA network share a network address such as an IP subnet address or an AppleTalk cable range. The second configuration method involves treating each of the PVCs as a separate logical point-to-point network, which is done by creating a sub interface for each PVC. The sub interface method requires more network addresses because each PVC has its own network address.
Of the two methods, the sub interface method is usually the recommended one. Both configuration methods are briefly described in the following sections.
We are going to move our IP traffic from the point-to-point serial links to a frame relay network. To do this, we are going to do something that is not normally recommended in a production network. We are going to remove the IP addresses from the point-to-point serial links, and we are going to stop the current IP routing protocol, OSPF, which was configured in
Frame Relay Sub interface Configuration and Verification
The concept of a sub interface is the same now with frame relay; it is just a logical interface that is directly associated with a physical interface. With frame relay sub interface configuration, we can create a sub interface for each of the PVCs coming into a serial interface.
Two types of sub interfaces can be created for frame relay—multipoint and point-to-point. A multipoint sub interface can handle multiple PVCs; its use is similar to that of the NBMA network that we saw in the preceding section. A point-to-point sub interface (the type we are going to use) effectively turns every PVC into a point-to-point network with its own network addressing. Using point-to-point sub interfaces gives us greater control over our frame relay network.
We are going to implement the configuration illustrated in Figure 14. Since Dallas has two PVCs, we are going to create sub interfaces on Dallas. FortWorth and Austin will continue with the original NBMA configuration since they have only one PVC each; however, Austin’s Serial2 IP address will have to change since it will be connected to a different network created by the Dallas point-to-point sub interface.
Figure 14: Frame-relay sub interface internetwork.
Figure 15 shows the commands necessary to change from the NBMA configuration to a point-to-point sub interface configuration. When using sub interfaces, the physical interface normally does not have any network addressing; therefore, we removed the IP address from Serial2 (Line 4). We created the first sub interface, Serial2.1, by referencing it on an interface command (Line 5). The point-to-point keyword is used to create a point-to-point sub interface. The other option is to create a multipoint sub interface by specifying the keyword multipoint. Point-to-point and multipoint sub interfaces cannot be created when the physical interface’s encapsulation is at its default, HDLC; therefore, we had to have the encapsulation frame-relay command on the Serial2 interface to be able to specify that we wanted a point-to-point sub interface. The encapsulation frame-relay command was issued on Serial2 during the NBMA configuration.
Frame Relay Configuration Summary
For frame relay NBMA configuration, all we need is the encapsulation frame-relay command and the appropriate network protocol commands, such as ip address or decnet cost, on the physical interface connected to the frame relay network. In an NBMA environment, we may run into connectivity problems caused by split horizon (see Section 2.2.2.1, “Distance Vector Routing Protocols”).
The recommended frame relay configuration makes use of sub interfaces. All we have to do is issue the encapsulation frame-relay command on the physical interface, create a sub interface for each PVC, use the frame-relay interface-dlci command to assign a DLCI to each sub interface, and issue the appropriate network protocol command(s) on each sub interface. Using point-to-point sub interfaces removes the split horizon problems sometimes experienced in an NBMA environment.
Frame relay is currently very popular because it is usually less expensive than normal leased lines and because it is widely available. Frame relay is usually less expensive than leased lines because when we get a leased line, our service provider provides us with dedicated bandwidth, and when we get a frame relay PVC, our service provider provides us with just a share of their bandwidth (no dedicated). Our traffic has a greater chance of being dropped traversing a frame relay network than a leased line—yet another reason for the lower cost. Most companies are willing to accept these shortcomings because of the monetary savings, and most applications are not greatly affected as long as the frame relay configuration is properly implemented on both the user side and the service provider side