Towards Controlled Performance Bottleneck States in Virtual 5G/O-RAN: A Three-Region Queue-Aware Framework

Throughput and resource indicators alone do not reveal the state of an individual queue in a virtual 5G/O-RAN system. This preprint develops an analytical and measurement-oriented framework for deliberately inducing a bounded performance bottleneck state and, eventually, regulating it without buffer-overflow-relatedrejection. At a declared service boundary, criterion C1 is the dimensionless ratio of arriving workload to potentially available service. A deterministic, finite-buffer, work-conserving fluid queue motivates three operating regions: unsaturated operation (R1), controlled backlog (R2), and verified queue overflow (R3). An additional UNDETERMINED label prevents unsupported experimental classification when telemetry is insufficient.We derive conditional finite-horizon overload thresholds and phase-explicit sinusoidal backlog trajectories.Reproducible numerical examples demonstrate that identical C1 extrema can yield different backlog peaks and overflow outcomes when the excitation period or initial phase changes. The influence of non-zero backlog is examined analytically rather than in the 18 numerical scenarios. A measurement protocol specifies offered-arrival accounting, service-rate estimation, queue-state freshness, discard attribution and temporal alignment. An evidence-gated research sequence proceeds from open-loop load characterisation to queue feedback and prospective joint constrained regulation. The numerical results are model predictions, not new 5G/O-RAN measurements; neither queue-control performance nor experimental region verification is claimed here.