Chapter 7. Area Design

Areas, as discussed in Chapter 2, are used to control the size of the link state database and the negative impact of flooding and SPF calculations on network bandwidth and router resources. They accomplish this control by breaking up the network into individual SPF domains. That is, rather than having a network in which every router has an identical link state database and in which the SPF process computes a single shortest-path tree spanning the entire routing domain, the routing domain is broken up into multiple SPF domains. The link state databases must be identical only within these individual areas, and the area borders are defined by the scope of the individual SPF trees calculated from these area databases.

In order to communicate routing information between two areas, at least one router must be connected to both of the areas, have a copy of the link state database for each the two areas, and run separate SPF calculations for each area. OSPF calls this router the Area Border Router (ABR). The IS-IS L1/L2 router is analogous to the OSPF ABR, but area boundaries are a bit trickier in IS-IS, as you will see later in this chapter.

The ABR sends the routing information it learns from one connected area into the other connected area. And in this behavior is a significant but often misunderstood characteristic of OSPF and IS-IS: The two protocols are always called link state protocols, but in fact they are link state only within an area.

Because shortest-path trees do not cross area borders, the inter-area routing information sent into the area by the ABR basically says, "I can reach prefix X at a cost (or distance) of Y." The routers within the area have no visibility beyond the ABR, and must trust the ABR to accurately advertise the prefixes it can reach. This behavior is distance vector, not link state.

You learned in Chapter 2 that the nature of distance vector protocols (a sequence of routers telling their neighbors, "You can get there from here. Trust me.") makes them susceptible to information corruption, possibly leading to routing loops. Distance vector protocols use several strategies and mechanisms to avoid routing loops, such as split horizon and route poisoning. OSPF and IS-IS must also take steps to prevent potential loops from being created due to the distance vector behavior at area borders.

The most basic way to avoid loops is to impose a loop-free inter-area architecture, as depicted in Figure 7.1(b). In this architecture, inter-area routing information can only be exchanged through the backbone area; nonbackbone areas are not allowed to exchange routing information directly. Understanding this concept is key to designing multi-area OSPF and IS-IS networks. But beyond this concept are a number of factorsboth general and protocol specificto consider when designing area topologies. This chapter focuses on these factors.

Figure 7.1. Allowing routing information to be freely distributed among areas (a) can lead to routing loops. You can remedy this by requiring routing information to be exchanged over loop-free paths (b).