Make the Surviving Orbit Carry the Failure

A multi-orbit federation has not demonstrated recovery when its controller selects an alternate path, but when that path carries the displaced traffic without losing control of its queues. Qualification must expose the feedback between rerouted payload, transport retransmissions and the signalling needed to complete the move. An idle failover test removes precisely the competition that can make recovery fail.

Create contention before removing the gateway

The failure to inject is a gateway forwarding black hole while its control session remains established, forcing the federation to discover that an apparently available route no longer delivers traffic. Establish admitted sessions across LEO, MEO, GEO and HAPS paths, with destination overlap that makes their traffic compete at the actual surviving bottleneck, whether that is feeder capacity, gateway processing or a scheduler. Increase offered load through successive stable operating points, retaining the production mix of sustained transfers, latency sensitive traffic and burst arrivals, then suppress forwarding at the selected gateway without withdrawing its routes. Configure the local forwarding health watchdog with a fault declaration timeout of 250 ms, explicitly treating this as a laboratory setting for local detection rather than a permissible round trip budget across every orbital layer. Preserve each alternate path's propagation delay, capacity variation and packet scheduling because replacing these with equivalent average throughput conceals the queue occupancy and acknowledgement timing that govern the result.

Let the recovery compete for resources

Keep arrivals running throughout the interruption and restoration, with normal retransmissions and application retries enabled, so that alternate path installation must complete while displaced traffic consumes the same processing, buffers and transmission opportunities that recovery signalling actually uses.

Measure delivery rather than route convergence

Correlate packet sequence identifiers at ingress and destination with timestamped forwarding changes, queue occupancy, scheduler service and controller work backlog, using a common clock and recording its uncertainty. These observations must distinguish detection, alternate path installation and useful delivery, since a successful route update can coexist with a growing destination backlog. The mechanism under examination is a recovery feedback loop: displaced traffic fills the survivor, delayed delivery provokes retransmissions, and the additional work delays signalling or forwarding updates wherever resources are shared. Judge useful delivery as unique application payload received within its service contract, excluding duplicates and late packets that inflate link throughput without restoring service. For cellular access, use the 3GPP family of mobility and session continuity specifications to identify the context and bearer state that must survive, while treating end to end delivery under contention as a separate qualification obligation. Repeat the same forwarding interruption at different phases of the scheduled topology evolution, preserving equivalent pre-fault load, because a survivor with ample capacity at one orbital position may be approaching a constrained interval at another.

Qualify a recoverable admission envelope

Apply one acceptance rule at each admitted load: accept only if every protected traffic class meets its declared interruption and delivery limits, and the survivor's payload queues and control backlog return to their established operating bands within the contracted recovery interval while arrivals continue. A plateau of high aggregate throughput cannot satisfy that rule when one protected class remains starved or the controller still carries deferred work. Repeat a failing operating point with recovery signalling given reserved service at every shared bottleneck, holding the payload admission policy constant, to determine whether resource competition explains the failure. That reservation consumes capacity during normal operation, while reducing admission headroom sacrifices sellable traffic to leave space for displaced sessions and their transient retransmission load. Where the surviving orbit cannot support the original service contract even after queues drain, qualify an explicitly negotiated degraded service separately rather than counting route availability as recovery. The resulting admission limit must follow the capacity available along the evolving alternate path, or the operator is selling protection that disappears with the topology.

Back to NonTerrestrialNetworks · Domains

← Back to NonTerrestrialNetworks
Engineering & Architecture Brief

Make the Surviving Orbit Carry the Failure

A multi-orbit federation has not demonstrated recovery when its controller selects an alternate path, but when that path carries the displaced traffic without losing control of its queues. Qualification must expose the feedback between rerouted payload, transport retransmissions and the signalling needed to complete the move. An idle failover test removes precisely the competition that can make recovery fail.

Create contention before removing the gateway

The failure to inject is a gateway forwarding black hole while its control session remains established, forcing the federation to discover that an apparently available route no longer delivers traffic. Establish admitted sessions across LEO, MEO, GEO and HAPS paths, with destination overlap that makes their traffic compete at the actual surviving bottleneck, whether that is feeder capacity, gateway processing or a scheduler. Increase offered load through successive stable operating points, retaining the production mix of sustained transfers, latency sensitive traffic and burst arrivals, then suppress forwarding at the selected gateway without withdrawing its routes. Configure the local forwarding health watchdog with a fault declaration timeout of 250 ms, explicitly treating this as a laboratory setting for local detection rather than a permissible round trip budget across every orbital layer. Preserve each alternate path's propagation delay, capacity variation and packet scheduling because replacing these with equivalent average throughput conceals the queue occupancy and acknowledgement timing that govern the result.

Let the recovery compete for resources

Keep arrivals running throughout the interruption and restoration, with normal retransmissions and application retries enabled, so that alternate path installation must complete while displaced traffic consumes the same processing, buffers and transmission opportunities that recovery signalling actually uses.

Measure delivery rather than route convergence

Correlate packet sequence identifiers at ingress and destination with timestamped forwarding changes, queue occupancy, scheduler service and controller work backlog, using a common clock and recording its uncertainty. These observations must distinguish detection, alternate path installation and useful delivery, since a successful route update can coexist with a growing destination backlog. The mechanism under examination is a recovery feedback loop: displaced traffic fills the survivor, delayed delivery provokes retransmissions, and the additional work delays signalling or forwarding updates wherever resources are shared. Judge useful delivery as unique application payload received within its service contract, excluding duplicates and late packets that inflate link throughput without restoring service. For cellular access, use the 3GPP family of mobility and session continuity specifications to identify the context and bearer state that must survive, while treating end to end delivery under contention as a separate qualification obligation. Repeat the same forwarding interruption at different phases of the scheduled topology evolution, preserving equivalent pre-fault load, because a survivor with ample capacity at one orbital position may be approaching a constrained interval at another.

Qualify a recoverable admission envelope

Apply one acceptance rule at each admitted load: accept only if every protected traffic class meets its declared interruption and delivery limits, and the survivor's payload queues and control backlog return to their established operating bands within the contracted recovery interval while arrivals continue. A plateau of high aggregate throughput cannot satisfy that rule when one protected class remains starved or the controller still carries deferred work. Repeat a failing operating point with recovery signalling given reserved service at every shared bottleneck, holding the payload admission policy constant, to determine whether resource competition explains the failure. That reservation consumes capacity during normal operation, while reducing admission headroom sacrifices sellable traffic to leave space for displaced sessions and their transient retransmission load. Where the surviving orbit cannot support the original service contract even after queues drain, qualify an explicitly negotiated degraded service separately rather than counting route availability as recovery. The resulting admission limit must follow the capacity available along the evolving alternate path, or the operator is selling protection that disappears with the topology.

Related Architecture Specifications (NONTERRESTRIALNETWORK): airspacesec.com · intersatlink.com · ntnantenna.com

More Technical Articles on nonterrestrialnetworks.com

Delay-Tolerant Routing and Contact Graph Algorithms in Multi-Layer Space Topologies →HAPS as Stratospheric Cellular Relays: Mobile Edge Compute at 20km Altitude →