How do you choose a routing protocol for an enterprise network?
Show the answer
Say this first
From the requirements: size, topology, vendors, the team's skills and where the administrative boundaries are. Inside the enterprise I default to OSPF because it is an open standard, every vendor supports it and areas give hierarchy. EIGRP is fine in an all-Cisco network. IS-IS appears in service providers and some large fabrics. BGP goes wherever you meet another domain or need policy: internet edge, MPLS provider, cloud, SD-WAN and large data center fabrics. Static covers the stubs.
Go deeper
| Protocol | Fits | Watch out for |
|---|---|---|
| Static | Stubs, default routes, small sites | No failure detection beyond the interface without tracking |
| OSPF | Multi-vendor enterprise IGP | Area design, LSA flooding in large flat areas |
| EIGRP | All-Cisco enterprise, hub and spoke with stubs | Little support outside Cisco (published as informational RFC 7868), query scope in unsummarized designs |
| IS-IS | Service providers, large fabrics | Fewer enterprise engineers know it; NET addressing |
| BGP | Internet edge, provider and cloud peering, SD-WAN, large DC fabrics | Slower by default, policy complexity |
Design principles
- One IGP if you can. Every redistribution point is a place for loops and surprises.
- If you must redistribute, do it at as few points as possible, tag routes and filter on the tags so routes do not loop back.
- Fast convergence comes mostly from design and failure detection (routed links, BFD, summarization), not from which IGP you picked.
- Match the team: a protocol nobody on call understands is a risk, however elegant.
They will ask next
Why do many large data centers run BGP instead of an IGP?
In a spine-leaf fabric, eBGP between tiers (the approach described in RFC 7938) is simple, scales to thousands of switches, gives per-neighbor policy and avoids flooding link-state updates across the whole fabric. ECMP across spines works naturally with multipath. One protocol also covers the fabric, the edge and EVPN.
Why is the internet edge BGP and not OSPF?
The ISP is a different organization. BGP is built for routing between administrative domains with policy on what you send and accept, and it does not expose your internal topology. An IGP trusts every neighbor and shares the whole link-state database.
Traps
- Choosing EIGRP "because it converges faster". With good design OSPF converges just as fast.