The IGP in a provider core
The interior routing protocol is the foundation every service rides on.
Analogy: the road map of a city
Delivery vans only reach customers if everyone shares one accurate map.
- The IGP carries loopbacks and link addresses, nothing more; customer routes live in BGP.
- Services find their far end through the loopback, so loopback reachability is the real goal.
- Keep the table small: fewer routes means faster convergence and less memory use.
- Pick one protocol per network and one clear reason, such as IS-IS for flexibility or OSPF for familiarity.
- Everything above the IGP, including labels and VPNs, fails when the IGP fails.
Services depend on the IGP.
A provider carries only 400 internal routes in its IGP while BGP carries a million customer routes.
Common mistake: putting customer routes into the IGP. The table grows and convergence slows.
Bloated IGP
Convergence takes seconds after any change.
Lesson: keep only infrastructure routes in the IGP.
What belongs in the IGP?
Loopbacks and link addresses only.
Key takeaways
- IGP carries infrastructure only.
- Loopback reachability is the goal.
- Keep the table small.
- One protocol, one reason.
- IGP failure breaks everything above.