Single, Dual and Multihomed Designs
Count the providers, then count the links to each provider. Finally, check which routers and physical paths are still shared.
Single-homed
One link connects the customer to one provider. It is simple to operate, but that connection has no alternate exit.
For straightforward Internet access, a static default route may be sufficient. BGP can still be used when the routing requirements justify it; the connection design does not itself select the routing protocol.
Dual-homed
Two or more links connect the customer to the same provider. Start with two links between the same router pair.
Two links, shared routers
Either independent link can survive the other failing, but both endpoint routers remain common dependencies.
Separate provider routers
This example removes a single provider-edge router as a common dependency, but R1 and the ISP remain shared. Adding another customer router can improve device resilience; separate circuits should also avoid the same physical path.
Separate customer routers
Two customer borders can share one ISP edge. This removes one customer-router dependency but still leaves the provider router shared.
Separate routers at both ends
Each customer border uses a different ISP edge. A single border-device failure can leave the other connection available, provided internal routing reaches it. The provider remains shared.
Single multihomed
The customer has one connection to each of two providers. This adds a second provider, rather than just a second circuit from the first provider.
R1 remains a common failure point here. Using separate customer routers can remove that dependency, provided the internal network can reach the surviving router.
Separate customer borders
R1 uses ISP-A, while R2 uses ISP-B. A border failure no longer removes both provider connections. The LAN still needs a working path to the surviving border.
Dual multihomed
The customer has redundant connections to each provider. This diagram uses two circuits per provider, but still only one customer router.
More circuits alone do not protect against R1 failing. A design with multiple customer and provider routers can cover more device failures, at the cost of more operational complexity.
Two customer borders
Here each customer border has two links to its provider. If R1 fails, ISP-A is lost but R2 can still reach ISP-B. If R2 fails, the reverse applies. Other layouts can distribute the links differently; examine the actual failure coverage rather than assuming the name guarantees it.
Compare failure coverage
| Design shown | Providers | Links per provider | Remaining dependency |
|---|---|---|---|
| Single-homed | 1 | 1 | Link, router, provider |
| Dual-homed | 1 | 2 | Shared endpoint routers in some variants; one provider in every variant |
| Single multihomed | 2 | 1 | Customer router in the first variant; one circuit per provider |
| Dual multihomed | 2 | 2 | Customer router in the first variant; shared provider edge per ISP |
Key Takeaways
- Single-homed means one provider connection; dual-homed adds connections to the same provider.
- Multihoming adds independent providers; it does not automatically add customer-router redundancy.
- Dual multihoming combines provider diversity and redundant provider connections.
- Protecting a link, a router and a provider are separate design goals.
- Surviving equipment must remain reachable from the LAN, and circuits must avoid shared physical failure points.
References and diagram notes
Further design context: service-provider connectivity guide.