Information Technology Reference
In-Depth Information
the remote sites being used as transit for reachability between the hub routers. As previously
discussed, this is undesirable because of PVC sizing and isolating impact during failure
scenarios. This could also result in routing loops.
The solution is to not summarize prefixes and redistribute them directly into BGP. The
MED is set to the IGP metric, providing the core with topological information about which
hub router to send the traffic to for each prefix. This allows R3 to use R1 as primary and R4
to use R2 as primary. The hub routers do not need to be directly connected, nor do they need
iBGP between them.
As a general rule, it is desirable to summarize as much as possible whenever possible, but
in some scenarios summarization can create a great deal more complexity with little realized
benefit. If there are 5000 prefixes, summarizing these into a single prefix in BGP results in
a savings of less than 5 MB of RAM. It is much more important to reduce prefix count in
the IGP than in BGP.
Case Study: BGP Core Deployment
This case study examines a typical enterprise network and seeks to address the challenges
that network engineers face as they work to scale their network. The design requirements
are detailed, and a BGP architecture is selected based on the requirements identified. After
the appropriate architecture is determined, the different components of the network are
examined individually to determine how best to integrate them. A migration strategy is then
developed and executed. The case study finishes with a look at the final configurations.
This case study was created with complex requirements and a large number of routers. The
intent is to bring together the concepts discussed in this chapter in a realistic scenario.
BGP Core Design Scenario
The current network consists of a single EIGRP AS, AS100. It is very common to see Stuck
In Active (SIA) messages and very high CPU loads for the EIGRP process. There are
approximately 7000 EIGRP prefixes in the routing table and 900 Layer 3 devices in the
network. The entire network is currently numbered using the RFC 1918 address space of
10.0.0.0/8. Prefix assignment was not performed in a consistent manner, allowing for
hierarchy and summarization. Figure 5-20 shows the initial topology for the network.
Design Requirements
The current instability has reached the point at which it is no longer acceptable, because it
is visibly affecting productivity. The CIO has sponsored an initiative to increase reliability;
however, the project is being done on a shoestring budget, which does not permit a whole-
sale upgrade. Service-Level Agreements (SLAs) have been created between IT and the
operating divisions in the company. These agreements will be active in 30 days.
