A circuit can stay up while voice calls break up and card payments time out. The branch still has connectivity, but the link is no longer good enough for the traffic using it. That kind of partial failure is where an SD-WAN management platform can earn its place.
SD-WAN uses policies to steer traffic over available connections. The physical or provider links form the underlay. The logical paths built across them form the overlay. The management platform helps operators define policy, distribute it to branch devices, and investigate how those devices chose a path.
Write the policy from the application’s point of view
A branch may have fiber and cellular connectivity. A voice application needs a usable latency, jitter, and loss profile. A large overnight transfer can tolerate delay but may consume enough cellular data to become expensive. Payment terminals may have strict requirements for which networks they can reach.
Write a separate policy for each traffic class. For voice, decide how much degradation triggers a move and how long the original path should stay healthy before traffic returns. For bulk transfers, decide whether the job should pause when fiber fails. For payments, verify that the alternate path preserves the routing and security controls the application depends on.
Probe targets matter. A healthy measurement to a nearby endpoint provides evidence about that path. It may say little about a SaaS application reached through different transit or inspection points.
Compare how each platform expresses branch policy
Cisco Catalyst SD-WAN documents application-aware routing with SLA-based path selection. It is a useful starting point for teams that need to express steering policy across Cisco WAN edges. During a pilot, have an operator trace a policy from the central definition to the decision made at a specific branch. Cisco application-aware routing documentation
Fortinet Secure SD-WAN exposes performance SLA measurements through FortiGate. It is a practical candidate when the branch already uses FortiGate for security. Evaluate policy with inspection enabled and check the resulting performance on the intended appliance. Fortinet performance SLA documentation
Versa Networks documents configurable overlay topologies through Director workflows. Evaluate it when segmentation and topology are central to the design. Ask the team to create a branch policy with a deliberate exception, then verify which routes are visible to each segment. Versa overlay documentation
Make a bad link part of the pilot
Test a clean disconnect, then introduce sustained loss or jitter while the interface remains up. Observe new and existing sessions separately. A policy may steer new traffic successfully while an existing session still reconnects because its source address or stateful inspection path changed.
Restore the circuit gradually and watch for repeated switching. Then ask an on-call engineer to explain the event from the available telemetry. The platform should make it clear which path carried the application, why policy moved it, and what conditions caused it to move back.