OSPF Operation in a Single Area NBMA Setting

A non-broadcast multi-access environment is one where there are more than two routers without a native broadcast capability. Frame Relay is a popular technology that commonly uses a NBMA implementation. Other examples of NBMA technologies are X.25 and Asynchronous Transfer Mode (ATM).

The most popular NBMA topology is frame relay, where a core router at a central site uses a single interface to connect several virtual circuits to several other sites over a frame cloud. A representation of this and other NBMA topologies is shown in Figure 4.9.

Figure 4.9. NBMA topologies.

OSPF approaches an NBMA environment just like any other broadcast network such as Ethernet. Full- and partial-mesh topologies can become costly, however, because of the formula mentioned earlier in the "OSPF Broadcast Multi-Access Operation" section. To emulate broadcast behavior with a full mesh of 15 sites, for example, would entail 15x(151)=210 permanent virtual circuits (PVCs). Although a partial mesh is often used to alleviate some of these costs, the star topology is the most common and economically feasible logical configuration for frame relay.

NBMA Mode

There are two formal modes of running OSPF over a NBMA topology according to RFC 2328: NBMA and point-to-multipoint. With NBMA mode, routers are usually configured in a full-mesh topology to establish the proper adjacencies and emulate broadcasting. In a full-mesh design, all the routers have virtual circuits established with all the other possible destinations in the network. Although this solution provides redundancy, it may be cost-prohibitive for some organizations. OSPF then imitates a broadcast environment by electing a DR and BDR for the NBMA topology, using the OSPF techniques discussed earlier. LSA updates and link-state acknowledgement (LSAck) packets are then flooded to all neighbors listed in the neighbor tables. This is actually a very efficient way to configure OSPF over NBMA when you have a limited number of neighbors. It offers more manageable link-state database sizes and generally decreased levels of traffic. It can be CPU and bandwidth intensive , however, in larger environments.

Please note that in some resources, you'll see both link-state advertisement and link-state acknowledgement represented as LSA. To clarify, we represent link-state advertisement as LSA and link-state acknowledgement as LSAck.

With OSPF running in NBMA mode, all routers are in a single subnet and neighbors must be statically defined at the outset to get the DR election process started. After the DR router is chosen , it generates an initial LSA for the network to get the ball rolling. There is one DR and one BDR per network segment so that when a change occurs it can be processed and flooded throughout the single OSPF area. The OSPF flooding process on a multi-access link happens according to the following four-step process:

1. If a change occurs in the link state, the router that detects the change multicasts an LSU to the multicast address of all OSPF DRs and BDRs at 224.0.0.6.

2. The DR acknowledges the modification and floods an LSU to other routers via 224.0.0.5. Because each LSA must be acknowledged , each router responds with an LSAck packet back to the DR.

3. A router connected to another subnetwork also floods the update to the link-state to the DR of the other subnetwork (or an adjacent router in point-to-point mode).

4. The routers update their link-state databases when the updates are received. It uses the new data to calculate a new routing table from the shortest path first algorithm. Don't forget that OSPF routers wait for a default interval of five seconds every time an LSU is received before recalculating, as previously mentioned.

If your NBMA-mode routers are not configured in a full-mesh topology, you must manually select the DR and BDR and have full connectivity to every other router.

Subinterfaces

It is also possible to break an NBMA network into additional logical subnets by using subinterfaces, particularly when a full mesh is not feasible because of campus design limitations, bandwidth constraints, or IP addressing issues. Using subinterfaces tends to be administratively complex and error-prone , so you might be better off running your OSPF over NBMA in point-to-multipoint mode. This scenario is addressed later in this chapter in the "Point-to-Point Mode" section.

Subinterfaces are often employed to partition a physical interface into several logical interfaces in NBMA environments. This allows each subinterface to operate as its own point-to-point or point-to-multipoint interface. On non-broadcast networks where a direct link does not exist between all the routers, you can use subinterfaces to create logical subnetting schemes and group router interfaces accordingly . With subinterfaces, each subnet could be in its own NBMA or point-to-point subnetwork because each virtual circuit is defined in a separate logical subnet. Another advantage to subinterfaces is network stability. If you were simply using a single, physical interface for a permanent virtual circuit and the interface went down, the router would indicate that the link had failed. With subinterfaces, however, you get the benefit of the physical interface remaining up, even if the subinterface has a problem.

Point-to-Multipoint Mode

In a nutshell , point-to-multipoint is really a collection of point-to-point links where routers identify their neighbor routers but do not elect DRs or BDRs. This provides an additional advantage of no added election traffic or LSAs being propagated for the network. In large networks, you get the added benefit of a reduction in the number of virtual circuits that need to be established as well. Point-to-multipoint is commonly used in partial mesh and sometimes star topologies (refer to Figure 4.9). Subinterfaces also come in handy with point-to-multipoint links. For example, assume that you are the network engineer for an organization. You have allocated a single subnet for your frame relay network and you would also like to avoid DR and BDR elections , subinterfaces, or manual neighbor configurations. Your internetwork is made up of four sites that are linked in a partial-mesh Frame Relay topology, as shown in Figure 4.10.

Figure 4.10. Four sites connected in a hub-and-spoke, partial mesh frame relay topology using a single IP subnet.

Point-to-multipoint mode has the following benefits and characteristics:

• No DR or BDR is elected.

• There are no LSAs for the network, but rather additional LSUs to describe neighboring routers.

• Non-adjacent neighbors can route through a common router configured for point-to-multipoint mode.

• Multicast Hello packets discover neighbors dynamically so that static neighbor configuration is not necessary.

• All the routers are on a single subnet.

• LSA and LSAck packets are duplicated and sent to each interface's neighbor.

In Figure 4.10, we use point-to-multipoint mode on all four routers on single interfaces (no subinterfaces) in a single subnet. This enables us to avoid having to manually configure neighbors as well as prevent DR and BDR elections. This mode can also be configured with the ip ospf network point-to-multipoint interface configuration command. Broadcast mode would require a DR/BDR election. Point-to-point mode with subinterfaces would not demand a DR/BDR election; however, it would need a separate subnet for each configured virtual circuit (VC).

NBMA (frame relay) clouds can be fully meshed or arranged in many variations of hub-and-spoke and/or partially meshed topologies. OSPF, however, sees the NBMA cloud as a broadcast medium, so you must correctly configure OSPF to operate within the frame relay cloud. Cisco has defined additional NBMA modes to explain OSPF neighbor relationships that go beyond the standards. Cisco also defines a point-to-multipoint broadcast mode, broadcast mode, and point-to-point mode.

Point-to-Multipoint Non-Broadcast Mode

Point-to-multipoint non-broadcast mode is a Cisco-defined mode that extends the RFS 2328-compliant point-to-multipoint mode covered in the previous section. Some point-to-multipoint networks use non-broadcast media such as IP over ATM and frame relay switched virtual circuit (SVC) routers that do not perform dynamic neighbor discovery. Although you must define your neighbors with static configurations, you have the added flexibility of modifying the costs of the links to represent different bandwidths. The partial-mesh or star topologies are used in these environments and neighbors must have the same subnet number. The interface is configured with the ip ospf network point-to-multipoint nonbroadcast command.

Broadcast Mode

Broadcast mode is another Cisco-defined method for getting around having to list all the neighbors with a static configuration. The interface is configured with an ip ospf network broadcast command so it acts just as it does on a LAN; therefore the neighbors must belong to the same subnet. You should use a full-mesh topology here or else rig the DR election with an interface priority configuration. This is the default setting for a broadcast multi-access network such as Ethernet.

Point-to-Point Mode

The point-to-point mode is typically implemented when you have only two nodes on your NBMA (frame relay) network. However, you can also implement it by using subinterfaces on NBMA networks with more than two nodes. Subinterfaces are configured with the ip ospf network point-to-point command and each point-to-point connection is one IP subnet. Table 4.3 reviews the key details for each mode.

Table 4.3. OSPF over NBMA Modes (RFC- and Cisco-defined)

OSPF over NBMA	IP Subnets Needed	Description
NBMA	1 (same subnet number)	Fully meshed; manually configured adjacencies; DR/BDR elected; RFC2328-defined
Point-to-Multipoint	1 (same subnet number)	Partial mesh or star; automatic adjacency ; no DR/BDR elected; RFC2328-defined
Point-to-Multipoint non-broadcast	1 (same subnet number)	Partial mesh or star; manually configured adjacencies; no DR/BDR elected; RFC2328-defined
Broadcast	1 (same subnet number)	Fully meshed; automatic adjacency; DR/BDR are elected; Cisco-defined
Point-to-Point	Separate subnet for each subinterface	Partial mesh or star; subinterfaces; automatic adjacency; no DR/BDR elected; Cisco-defined

To help ensure your success on the BSCI exam, memorize the preceding table until you have it down cold.