When the Business Starts Depending on That Wireless Bridge

Telecommunications technician maintaining rooftop wireless infrastructure at dusk
In this Field Note
  1. Describe demand in both directions
  2. Keep performance and availability separate
  3. Choose the boundary on purpose
  4. Measure from failure to useful service
  5. Test the backup with the demand it must survive
  6. Give the link an operating plan

The wireless bridge started as a quick connection to another building. Then the cameras, phones, business applications, and employees moved onto it. People still call it “that link,” although its failure now stops a respectable portion of the business.

That change in responsibility deserves a change in engineering attention.

A link can be fast, affordable, and entirely suitable for important work. You need to establish what service the installed path actually delivers, which failures it can survive, and how it will be recovered. A good speed test on installation day answers only a small part of that brief.

Describe demand in both directions

List the applications crossing the link, their direction, busy-period demand, and tolerance for delay or interruption. Continuous camera traffic behaves differently from occasional downloads. Interactive applications may suffer when bulk transfers fill a queue even though the average utilization looks reasonable.

Here is a hypothetical branch with traffic concentrated toward the main building:

Demand Example load Question for the design
Cameras toward central recording 80 Mb/s under the measured test conditions What happens during a scene-dependent peak or failover?
Scheduled data transfer in the same direction Up to 50 Mb/s Can it be rescheduled or constrained while protection is reduced?
Business and control traffic Up to 10 Mb/s in this example Which applications must remain usable and how will that be verified?

These example loads can total 140 Mb/s in one direction before allowing for the relevant overhead and bursts. A backup demonstrated to carry 100 Mb/s of usable traffic cannot reproduce that service unchanged. You need a reduced-service plan, additional capacity, or both. Priority handling can protect selected traffic; it cannot make the arithmetic disappear.

Be explicit about capacity claims. A radio’s advertised physical rate, its aggregate bidirectional throughput, and the application throughput of your installed path are different quantities. On time-division systems, verify how airtime allocation fits the traffic mix. Also check the Ethernet handoff, switch uplinks, packet processing, and queues.

Keep performance and availability separate

A link that passes the required load in clear weather has demonstrated performance under those conditions. It has not demonstrated year-round availability.

Plan the path for the selected frequency, antenna positions, terrain, obstructions, interference, and weather assumptions. Visual line of sight is useful evidence but does not, by itself, establish suitable RF clearance. Cambium’s LINKPlanner profile documentation explains how its path view represents clearance and terrain-related considerations.

Review the model’s assumptions in the field. Trees grow, roof equipment changes, and antenna mounts move. The appropriate structural, electrical, grounding, and safe-access work needs qualified people. A path model cannot certify a mount.

Record receive levels, modulation, errors, throughput, and latency at commissioning, together with conditions and test method. Use that baseline to investigate later change. Where the radio reduces its operating rate under adverse conditions, establish what service remains at the lower rate and whether the traffic policy still behaves sensibly.

Spectrum choice belongs in that analysis. Licensed and unlicensed options involve different coordination, interference, cost, and operating considerations. Neither label is an availability guarantee.

Choose the boundary on purpose

A transparent bridge can be appropriate when devices genuinely need Layer 2 adjacency. It also extends the switched failure domain: broadcast traffic and a loop may affect equipment at both ends.

A routed boundary can contain those effects and make preferred and alternate paths more explicit. If a particular service needs an Ethernet handoff, consider whether it can be carried as a deliberately scoped service over a routed network rather than making the entire backbone one large bridge.

Document the reason for the choice and the protection it requires. A design with two physical links is incomplete until the forwarding behavior during normal operation, failure, and recovery is understood. Plugging in the second cable is not an acceptable discovery method.

Measure from failure to useful service

There are several clocks in a failover: detecting the problem, selecting an alternate path, updating forwarding, and the application resuming useful work. A fast first clock does not guarantee a fast last one.

For a supported routed implementation, BFD can help detect loss of forwarding when an Ethernet interface still reports up. In a symmetric example with negotiated 300-millisecond control-packet intervals and a detection multiplier of three, the detection interval is approximately 900 milliseconds. The BFD specification describes negotiation and detection behavior; use the device’s supported implementation rather than assuming every requested timer is attainable.

A physical link-down event can already trigger a rapid response without waiting for that interval. Conversely, congestion, processing load, or an unstable radio path may make aggressive detection produce unwanted route changes. Test the supported settings under representative load and degraded conditions.

The acceptance question is concrete: after the simulated failure, when did the agreed call, transaction, or data flow become usable again? Record that alongside protocol events. Do not report a timer setting as a measured outage duration.

Test the backup with the demand it must survive

Agree on disruptive tests with the affected people and use a lab first where possible. Cover more than a clean cable removal:

  • Loss of the primary path, including a supported test of forwarding loss without Ethernet carrier-down.
  • Operation at the backup’s demonstrated usable rate, with the agreed reduced-service policy.
  • Loss of a relevant shared dependency, where a safe test is possible.
  • Recovery of the primary and the intended decision to return traffic to it.
  • Management access and alert delivery during each condition.

For each case, record what failed, what stayed available, when service recovered, and what the exercise could not establish. If both links share a switch or power source, write that limitation plainly. Two lines on a diagram should not promise independence that the installation lacks.

During the lower-capacity test, inspect the important applications rather than just aggregate traffic. An automatically selected backup that preserves pings while the business cannot work has failed a meaningful acceptance test.

Assign ownership of configuration, monitoring, field maintenance, and recovery. Keep a protected configuration history, compatible spares, and a tested replacement procedure. Trend the measurements that can reveal deterioration before an outage, and review them after material changes in demand.

If the link’s role has grown, start by collecting a busy-period traffic sample and describing the reduced service the business could tolerate. Those two items make a far better brief than “we need a faster radio.”

Oaktown Tech can design and integrate wireless transport around that brief, including routing, capacity, testing, and handoff. The finished result should be something the business can depend on and the next engineer can explain.

Share this article

LinkedInFacebookXEmail

Field discussion

What would you test next?

Ask a question, challenge an assumption, or share a lesson from the field. Keep customer names, live addresses, credentials, and sensitive network details out of public comments. Charles reviews every comment before it appears. See the privacy notice for how comment information is handled.

Join the discussion

Your comment will appear after Charles reviews it. Required fields are marked .

A pseudonym is fine.

Working through a similar problem?

Describe the network, the difficulty, and the outcome you need. Charles reviews the brief and helps establish the right scope.