In order to create a scalable EIGRP configuration, the process before the configuration is crucial. Having the correct design before the implementation will save time in the future.
Designing any L3 topology, scalability is one of the major concerns. Filtering and route summarization along with limiting the scope of EIGRP queries can help to design a robust topology. Minimizing the propagation of EIGRP queries ( this happens when no feasible successor is in the topology table) can help as well in the convergence. Even if multiple ASs will not help with this, multiple ASs can help to smaller administrator domains. Also, having zones can reduce the failures domains as well.
Moving forward, a benefit of EIGRP is that it was designed for sub-second convergence. In order to have fast convergence in larger deployments, you should have few considerations in mind:
- Use address summarization
- Have at least one feasible successor
- Configure equal cost paths
- Configure stub routers
BFD is yet another mechanism to help in faster convergence as it helps to detect failure faster.
In advance, if your routers have dual processors and support SSO (stateful switchover) or ISSU (in-service Software Upgrade) can be configured with graceful restart so when a peer is been restarted, the routes are not removed immediately and traffic is been forwarded will control plane is restarting.
However, there are also some challenges and tricky points to bear in mind. When you configure a dual-homed hub and spoke networks, the two hubs should share network information and not a summary address. If you summarize and use a Null route, then definitely this design will save the topology!
