Section 5.2. Areas and Router Types
5.2. Areas and Router TypesThroughout this chapter, I have pointed out topics that are discussed in more detail in the context of LS database synchronization in Chapter 6. With the completion of the discussion of flooding, we are almost there. But there are still three basics you need to know about before tackling synchronization: metric types, and the details of the OSPF LSAs and IS-IS TLVs used in the synchronization process. These three topics depend on some knowledge of OSPF and IS-IS router types, which are best understood in the context of areas. Chapter 7 discusses areas in depth, but the very basic concepts (beyond what was introduced in Chapter 2) are covered here for the purpose of defining router types. You already know the purpose of areas from Chapter 2: They allow link state domains to grow large by limiting the scope of flooding. The limitation of flooding in turn limits the size of LS databases, which in turn limits the computational complexity of the SPF calculations. The end result is a limitation on the amount of network resources needed for flooding, the amount of memory needed for the LS database, and the amount of CPU needed for SPF calculation. The bottom line is that rather than having a single shortest-path tree spanning the entire link state domain, we use areas to break the domain up into multiple shortest-path trees. An area boundary delineates the boundary of a single shortest-path tree. Both OSPF and IS-IS employ a two-level area hierarchy that provides for multiple possible lower-level areas and a single upper-level area. All inter-area traffic between lower-level areas must pass through the upper-level area, to avoid looping. 5.2.1. OSPF Areas and Router TypesFigure 5.8 shows a basic OSPF area structure, with lower-level areas attached to a single upper-level area. OSPF calls the upper-level area the backbone area, and this area is always identified by the AID 0.0.0.0 (or just area 0). Nonbackbone areas can have any 32-bit identifier except 0.0.0.0. Architecturally, area borders are defined by the routers that connect areascalled, logically enough, area border routers (ABRs). A router becomes an ABR, and an area border is created, when at least one of its interfaces is connected to a different area than at least one other interface. That is, an OPSF router is not an ABR if all of its interfaces connect to the same area. At least one of the ABR interfaces must connect to the backbone area. This enforces the rule that traffic between two nonbackbone areas cannot transit a third nonbackbone area. Figure 5.8. Area border routers connect the backbone area to all other areas. Figure 5.8 is a simplistic illustration of ABRs. In any network design, two general principles will always serve you well: (1) Always have a backup, and (2) get the most out of what you have. Figure 5.9 shows how ABRs can adhere to those principles: (1) An area can be connected to the backbone by more than one ABR, and (2) a single ABR can connect more than one area to the backbone. Chapter 7 says much more about ABRs and area design, but for now pointing out these two very basic capabilities of ABRs will suffice. Figure 5.9. An area can have more than one ABR, and one ABR can connect more than one area to the backbone. Some IP networks are completely self-contained and never speak to the "outside world." OSPF serves such a network just fine. But the great majority of networks need connectivity outside their own IGP domainwhether to the worldwide Internet or just to another private routing domain. When such connectivity is necessary, external routesroutes to destination prefixes outside of the local domainmust be advertised into the domain. This might be a single default route or it might be a subset of the global Internet routing table.[6] The route might be learned from another routing protocol, or it might be a statically configured route. The routers that advertise external prefixes into an OSPF domain are called autonomous system boundary routers (ASBRs). As Figure 5.10 shows, an ASBR can be located anywhere within an OSPF domain. An ASBR can be in the backbone area or in a nonbackbone area,[7] and an ABR can also be an ASBR.
Figure 5.10. An autonomous system boundary router advertises the prefixes of destinations outside of the OSPF domain. 5.2.2. IS-IS Areas and Router TypesAreas are one of the few IS-IS concepts that are difficult to grasp at first. A big part of the reason, in my opinion, is that most people encountering IS-IS for the first time have at least some previous experience with OSPF. As a result, we want to force IS-IS to fit into our understanding of OSPF. You have already seen at least one example of where there is not a direct parallel between the two protocols (LSPs and LSAs). Although the basic function of areas is the same in the two protocols, their structure differs distinctly. Like OSPF, IS-IS uses a two-level hierarchy, and all inter-area traffic must pass through the upper-level area to avoid loops. IS-IS terminology fits nicely here: The upper-level area is the level 2 (L2) area, and the lower-level (nonbackbone) areas are level 1 (L1) areas. However, there is no reserved AID for the L2 area the way there is for the OSPF backbone area. An IS-IS L2 area can be any legitimate AID. You already know that the IS-IS AID is a part of the NET address, and the NET is assigned to the entire router, not to a specific interface. What this tells you is that unlike OSPF, where different interfaces can be assigned to different areas, an IS-IS router and all of its interfaces belong to a single area.[8] An IS-IS router in an L1 area that can form adjacencies only with routers having the same AID is an L1 router. Likewise, a router that is in the L2 area and can form adjacencies only with routers having the same AID is an L2 router. These capabilities are determined by configuring all interfaces on the router as L1-only or L2-only.
There must, of course, be a way to connect L1 areas to the L2 area. This is done by configuring certain interfaces on certain routers to accept an adjacency to a router with a different AID. These routers are called L1/L2 routers. Here's where things get confusing. We want to "see" IS-IS areas depicted the same way we depict OSPF areas: as neat boundaries encompassing some set of routers, as in Figure 5.11. The problem is, IS-IS does not always fit this picture easily. In some cases, both an L1 and L2 adjacency can exist between the same two routers. And when multiple IS-IS routers are connected to a broadcast link, there can be a jumble of L1 and L2 adjacencies on the same link. Where do the area boundaries really lie? Figure 5.11. L1/L2 routers are roughly analogous to OSPF ABRs in that they provide connectivity between areas. You will be well served if, instead of thinking of IS-IS areas in terms of physical links and logical boundaries, you think in terms of adjacencies, as shown in Figure 5.12. The AID still describes a group of routers, but relates more to the individual routers than some sort of logical "real estate." An IS-IS area, then, is a contiguous set of adjacencies between routers with the same AID. Although this is a bit more difficult to visualize than OSPF areas, the approach permits a surprising amount of flexibility, as Section 7.4.1 will show. Figure 5.12. IS-IS areas are best understood as sets of adjacencies. If we define IS-IS areas as a set of adjacencies, one of the most important factors in understanding the areas is understanding how and when adjacencies are formed. Table 5.3 is a summary of Table 4.2, showing the different possible relationships between two routers and the resulting adjacencies that result depending on whether their AIDs are the same or different. You will see the table again in Chapter 7 when area architectures and design are discussed in much greater detail.
IS-IS does not have a named router type for routers that advertise external prefixes into the domain, as does OSPF. But certain rules must be followed, and these are also discussed in Chapter 7. |
EAN: 2147483647
Pages: 111