Consider the diagrams below:
- Single-homed: One link connected from one router to one ISP
- Best-suited for sites with no critical applications, like a home office
- Most often configured with a default route outbound
- May advertise internal routes to the ISP
- Outage occurs with the loss of a single link
- Dual-homed: Two scenarios, one ISP
- One router, one ISP
- Provides link redundancy
- No device redundancy (connectivity lost if the router fails)
- No ISP redundancy (connectivity lost if the ISP fails)
- Primary and backup connectivity, depending on protocol
- Active/active, may load-balance/share, depending on protocol
- Uses static or dynamic routing
- Two routers, one ISP
- Same as previous scenario, but provides device redundancy if one enterprise router should fail
- One router, one ISP
- Multi-homed: Two enterprise routers operating in-tandem, with one link to a disparate ISP, each
- Provides link redundancy between routers
- Provides device redundancy
- Provides ISP redundancy
- Better performance than the previous examples if path determination configured correctly
- Usually uses BGP for exterior routing protocol, where enterprise advertises internal addresses to ISP, and BGP determines the best path toward a destination
- With BGP, becoming a transit AS is a risk and must be mitigated
- Dual multi-homed: Two enterprise routers operating in-tandem, with redundant links to disparate ISPs
- Same as the multi-homed use-case, but with added link redundancy for most optimal (and most expensive) deployment
Stuck-in-Active: Journal of an IT-Network Administrator 