Your Redundant Network Has Never Actually Failed Over

No, it hasn't. Unless someone has deliberately pulled the primary link under real production load and watched the backup take over cleanly, the failover has never been tested, only assumed. Dual-gateway SD-WAN, VRRP handoff, GRE tunnels between Meraki devices, none of that architecture proves anything about the physical channel it depends on. And that channel is almost never validated once the rack is live.

Control plane vs. physical layer

The demo problem

A failover demo proves the control plane works. It doesn't prove the cable does. Two gateways sharing WAN uplinks with a virtual IP handing control to the backup the instant the primary drops is genuinely solid engineering. None of it matters if the copper underneath is already compromised.

The Cat6 spec gap

Cat6 is rated in isolation, not in your rack

Cat6 is rated for 250MHz, but that spec applies to the cable on its own, not the jack, the patch panel, the cross-connect and every termination in between meeting the same standard. Cabinets get opened where every patch cord looks new and every termination behind it is rough, or undertwisted, or barely passing a visual check. On a dual-uplink setup, a degraded channel on one path doesn't show up as an outage. It shows up as a fault that only appears under load, at 2am, exactly when the failover is supposed to quietly do its job.

Failure signature

Why it stays invisible

The failure mode that catches people out isn't dead, it's intermittent. A hard failure gets fixed fast because it's obvious. A degraded channel that drops packets only under load can sit there looking healthy on every dashboard for months.

Months of an invisible failure add up to real money once you put a number on it — our free outage cost calculator builds an estimate from your own assumptions about downtime and lost trade.

Under pressure

Labelling is part of the fault tolerance, not an afterthought

When a carrier link drops and someone is tracing which cord terminates which uplink under pressure, a mislabelled port turns a five-minute fix into an hour of guesswork. Redundancy at the gateway level assumes integrity at the physical layer underneath it. Most SD-WAN deployments never test that assumption until the day both links are needed at once. Keep Trading Connectivity is built to test it deliberately, with a live failover test on the connections the business actually depends on, before the untested day arrives on its own.

What the field confirms

The physical layer keeps showing up as the blind spot.

Brian F., Systems Architect, IT/OT Infrastructure, Cybersecurity & Resilience

"Split-brain is the flip side of this failure. The wrong node thinks it's in charge because the heartbeat broke, not the uplink. The control plane is only as trustworthy as whatever channel it's using to talk to itself."

Adam Crisp, Telco & Network Specialist

"The expensive assumption is usually supply chain. You design a bulletproof dual-path network and still get taken out because the replacement part for a failed component is eight weeks out, and the vendor who sold you 'next business day' support quietly stopped stocking local spares."

Jeremy Vasquez, Data Center & Networking

"'Did someone unplug it' used to be step one. Now, with this much abstraction, it's step 52, after you've ruled out the SDN controller, the overlay and the underlay, and had a meeting about it. The cable was still fine the whole time."

Olgert T., Network Engineer, Cisco & Fortinet, Industrial/OT Networking

"Redundancy isn't real until failover is tested end to end. Most orgs never actually run that test. They fail over a router or a link in a change window, tick the box, and call it validated. Pulling the primary at 2am under real load is a different test entirely, and it's the one that never gets scheduled."

Kria Elinarbur, IT Specialist

"Observability catches it once the symptom shows up on the network side, which is most of the way there. The gap sits one layer down, monitoring signal quality and error counters off a channel nobody actually validated against spec in the first place."

Who we are

AAA Communications — Telstra Partner since 1992.

AAA Communications has operated as a Telstra Partner since 1992, delivering Layer 1 infrastructure validation, structured cabling, and in-building 4G/5G, Wi-Fi and SD-WAN assurance for enterprise, industrial and critical environments across Australia. We test the physical layer under real-world load before it becomes the reason a redundant system fails, because redundancy at the network layer is only as good as the channel it depends on.

Common questions

SD-WAN failover and physical layer testing, explained.

Does SD-WAN failover test the physical layer underneath it?

No. SD-WAN failover logic tests the control plane, whether the backup path activates when the primary drops. It doesn't validate the physical cabling, termination quality or channel performance under load, which is a separate, physical-layer test.

Why did my redundant network still go down during an outage?

Most redundant network failures trace back to a degraded physical channel that was never tested under real load, not a failure of the redundancy logic itself. A cable or termination that passes a visual check can still fail intermittently under production traffic.

How do you actually test failover properly?

Pull the primary link under real production load, not in a scheduled change window, and confirm the backup path carries full traffic cleanly. Partial black-holing or null-routing one path is a stronger test than simply unplugging a cable.

Is Cat6 cabling enough for a redundant SD-WAN setup?

Cat6 rated at 250MHz applies to the cable in isolation. The jack, patch panel, cross-connect and every termination in the channel all need to meet spec too, and that's rarely audited once a cabinet is live.