WAN and Branch Connectivity Design
Choosing the transport
WAN design balances cost, bandwidth, latency, and availability. Know the options: MPLS (private, SLA-backed, expensive), Internet/VPN (cheap, variable), dedicated fiber/Ethernet, broadband, LTE/5G as primary or backup, and SD-WAN overlays that abstract multiple transports with policy-based path selection. Understand hub-and-spoke versus full mesh versus partial mesh topologies, and how SD-WAN enables direct internet breakout at branches.
Resilience and quality
Design for path diversity (separate providers, separate physical routes), failover (active/standby or active/active), and QoS to protect voice/video across variable links. Know SLA concepts (latency, jitter, packet loss, MTBF) and how to map business-critical applications to transport classes. Branch design also covers local survivability: local internet breakout, local DHCP/DNS, and PoE for branch APs and phones.
Traps to watch
- Assuming two circuits from the same ISP are fully diverse - they often share conduit or a central office.
- Forgetting asymmetric routing and MTU/MSS issues on tunnels (encapsulation overhead causes fragmentation).
- Treating SD-WAN as a transport - it is an overlay and policy layer, not a circuit.
- Ignoring latency budgets for real-time apps when choosing internet versus MPLS.
Rule of thumb
Match transport to application tolerance: real-time voice needs low jitter and loss (private or well-managed path with QoS), bulk data tolerates internet. Worked example: a branch needs 200 Mbps and 99.9% availability with voice. A single internet link fails the availability test; a primary internet link plus an LTE backup with SD-WAN failover and QoS meets both cost and resilience goals.