In large networks, a standard OSPF configuration can lead to instability. Routers with limited CPU and memory may struggle with frequent Shortest Path First (SPF) calculations, causing slow convergence and network flaps. To build a resilient and fast-reacting network, we need to fine-tune OSPF's behavior.

Two key mechanisms for this are LSA throttling and SPF tuning. These timers control how often Link State Advertisements (LSAs) are generated and how frequently the SPF algorithm runs. By adjusting them, we can dampen the effect of an unstable link, preventing it from overwhelming every router in an area.

The command timers throttle spf sets three values: start, hold, and max-wait.

• start: The initial delay in milliseconds before running an SPF calculation after a topology change. This gives the router a moment to see if more changes are coming.
• hold: The minimum time to wait between two consecutive SPF calculations.
• max-wait: The maximum time to wait between two SPF calculations.

The LSA throttling timers work similarly, controlling the rate of LSA generation. These adjustments ensure the network doesn't overreact to minor, transient issues, leading to a much more stable routing domain.

Another powerful tool for optimization is prefix suppression. In hub-and-spoke topologies, advertising every single transit link's prefix into the OSPF domain is often unnecessary. Suppressing these prefixes reduces the size of the Link State Database (LSDB) and simplifies the routing table.

By configuring ip ospf prefix-suppression on an interface, you prevent OSPF from advertising the IP prefix for that interface while still allowing the interface to participate in OSPF for neighbor discovery.

While OSPF handles routing within an Autonomous System (AS), BGP manages routing between them. As an AS grows, the requirement for every internal BGP (iBGP) speaker to peer with every other iBGP speaker creates a full mesh of connections. This quickly becomes unmanageable.

The number of required sessions follows the formula $n(n-1)/2$, where $n$ is the number of routers. An AS with 10 routers needs 45 iBGP sessions. An AS with 100 routers needs 4,950. This is where Route Reflectors and Confederations come in.

A Route Reflector (RR) acts as a focal point. Routers within a cluster peer with the RR, which then reflects routes to them. This drastically reduces the configuration overhead. A client router only needs one peering session with its RR instead of with every other client.

Confederations are another solution. A confederation divides a large AS into smaller, private sub-ASes. Within each sub-AS, a full iBGP mesh or route reflectors can be used. eBGP is run between the sub-ASes. To the outside world, the entire collection of sub-ASes appears as a single, large AS. This approach provides better scalability and policy containment but adds complexity.

Beyond scalability, expert-level network design requires precise control over traffic paths. BGP's path selection algorithm is deterministic, but we can influence its decisions by manipulating attributes like Local Preference, AS-Path, and Multi-Exit Discriminator (MED).

Route maps are the primary tool for this. They function like conditional statements, allowing you to match routes using prefix lists or ACLs and then modify their attributes.

For example, to ensure outbound traffic to a specific partner network always uses a primary link, you can create a route map that sets a higher Local Preference for routes learned from that link. Since Local Preference is the first attribute checked after Weight, it's a powerful way to influence outbound path selection within your AS.

This configuration tells the router to set a Local Preference of 200 on routes matching the PARTNER_NET prefix list when they are received from neighbor 203.0.113.1. All other BGP routers in the AS will honor this higher preference (default is 100) and choose this path.

Conditional advertisement takes this a step further. Using advertise-map and non-exist-map, you can advertise certain prefixes to a neighbor only if another prefix exists (or doesn't exist) in the BGP table. This is critical for building resilient multi-homed networks where backup paths should only be advertised if the primary path is down.

Ready to test your knowledge of these advanced routing concepts?

By mastering these techniques, you can design and manage large-scale networks that are not only functional but also stable, resilient, and highly optimized for performance.