BlueHour LNGO
Live demonstration

The LNGO Lab

Break a hypothetical Traditional Operating Model with a router failure inspired by the November 13, 2026 FAA telecommunications outage. Then run the same failure, minute for minute, against an LNGO architecture and see whether the essential business outcome survives.

The essential outcome here is simple: operations continue. No system-wide stop is declared, and no operations centre is without reliable information for longer than five minutes.

What happened, and what is assumed

FactThe FAA incident, as reported

Established facts as published by the FAA and reported by the press, as of November 18, 2026. The root cause is still under investigation.

  • At about 3:30 p.m. ET on November 13, 2026, the FAA had a technical problem and issued a nationwide ground stop. Air traffic resumed by 8:15 p.m. ET. FAA
  • Legacy FAA equipment became unstable when the FAA Telecommunications Infrastructure network was flooded with about 1,000 connection attempts every four to five seconds, disrupting aircraft communications and radar feeds. FAA
  • The initial investigation points to two sites: a radio communications site in El Dorado, Arkansas, where network provider L3Harris had been deploying a new router, and a radar site in Amarillo, Texas. Blocking both locations stabilized the network. FAA
  • The Transportation Secretary said the Amarillo router had been replaced about three weeks earlier. Reported
  • Instability affected 17 of the FAA’s 22 Air Route Traffic Control Centers and spread to several Terminal Radar Approach Control facilities. Some radio frequencies and radar feeds stayed operational, and all airborne aircraft landed safely. FAA
  • The FAA reports no sign of a cyberattack. Operators cancelled about 12,462 flights that day, not all of them directly because of the ground stop. FAA
  • As a precaution, the FAA disabled the functionality involved on the originating routers and on more than 60 others, and paused their installation. FAA

HypotheticalWhat the Lab assumes

The Lab is inspired by the failure pattern above. It is not a model of the FAA network, and BlueHour has not modeled or validated that network.

  • A stylized network of four regions: four core routers, eight remote sites and sixteen operations centres. Every number, delay and threshold is an illustrative assumption, not calibrated to the event.
  • The failure: a router change at a remote site floods the control plane with connection attempts. Optionally, a second router, changed weeks earlier, starts contributing a few minutes later.
  • A centre needs a healthy path to its core router and its remote feeds. Below 70% reliability it is counted as without reliable information.
  • A hypothetical safety policy declares a system-wide stop when four or more centres lack reliable information for five minutes, and lifts it after twenty stable minutes with the sources contained.
  • Runs are deterministic. The same settings and seed always produce the same result, identified by its run fingerprint.

How the model works and its limitations

Inject the failure

Both architectures face exactly the same failure. Change the failure, or change either architecture’s controls, and the run recomputes.

Multiples of one router’s control-plane capacity
Only changes small load noise
Run
Before the failure

Traditional Operating Model

Ready.

Architecture controls

LNGO architecture

Ready.

Architecture controls
Outcome comparison
OutcomeLeft architectureRight architecture

Left architecture: event log

    Right architecture: event log

      What each control does in the model

      Every control changes a specific term in the simulation, not a colour on the screen. Turn them on and off one at a time to see which ones matter for which failure. In this model no single control covers every failure: rate limiting contains a flood but does nothing for a hard failure, and failover rescues a hard failure but not a flood that takes out the remote feeds.

      ControlEffect in the simulationBlueHour60 MOMs
      Staged change with automatic rollbackA change being made now is watched by a canary check and rolled back three minutes after it misbehaves. It cannot catch a latent fault from an earlier change.MOM-202 Change Impact Analysis; MOM-205 Third-Party Change Absorption; MOM-206 Reversibility
      Segmentation and dependency isolationFlood traffic and reconvergence churn are not relayed across segment boundaries. Saturated routers settle in 5 minutes instead of 18.MOM-408 Dependency Mapping; MOM-304 Complexity Ceiling Management
      Control-plane rate limitingCaps the control-plane traffic any single link can deliver to a router.MOM-305 Operating Risk
      Redundant pathsA shared path runs through another core router on the same control plane. An independent path runs through a separate backbone.MOM-310 Continuity & Resilience
      Automatic failoverCentres move to a healthier path in 2 minutes. Without it, people move them 30 minutes after an incident is opened.MOM-310 Continuity & Resilience
      Continuity monitoring with a live dependency mapDetects abnormal control-plane traffic within a minute and locates the source in 4 minutes. Without it, an incident opens after centres report problems and the source takes the time you set to find.MOM-409 Whole-System Visibility; MOM-408 Dependency Mapping
      Pre-delegated containment authorityIsolation runs as soon as the source is located, and people are notified with the evidence. Without it, isolation waits 20 minutes for escalation.MOM-307 Delegated Authority; MOM-308 Escalation & Human Override; MOM-505 Governance & Accountability

      How the model works

      Each minute, fault sources emit control-plane traffic. Routers relay part of what they receive, and saturated routers add reconvergence churn for a while. A router’s health falls as its load passes 80% of capacity and reaches zero at 130%. A centre’s reliability is the health of the path it is using, reduced if it loses remote feeds. Detection, location, authorization, isolation, rollback and failover each follow the delays shown above. A crashed router returns after 90 minutes; an isolated one is repaired 90 minutes after isolation.

      The simulation code is separate from this page and has its own automated tests, including that results are reproducible, that a stronger flood never improves the outcome, and that LNGO does not hold in the beyond-design-basis scenario.

      Limitations

      • The topology and every parameter are hypothetical. Results show architectural differences, not predictions for any real network.
      • Durations are not calibrated to the FAA event and should not be compared with it.
      • Real networks have many more failure modes, human factors and interactions than this model represents.
      • The beyond-design-basis scenario shows that LNGO reduces, but does not eliminate, the possibility of interruption.