Controlled Bottleneck Formation via Evidence-Audited PRB Allocation in Virtual 5G/O-RAN Testbeds
Year: 2026
Type: preprint
Abstract: The presence of radio link control (RLC) backlog does not, by itself, establish overload. We study deliberate formation of an observable performance bottleneck through confirmed physical resource block (PRB) quota adjustments and distinguish three operating regions: negligible persistent backlog (R1), detectable backlog under predeclared acceptable service (R2), and verified violation of service constraints (R3). The framework combines queue conservation, a hybrid representation of E2-confirmed control, a reconstructed Adaptive C2 decision policy, conditional protective bounds, integral PRB accounting and constrained multiobjective optimisation. Independent raw-data audits of one 40-second FIXED26 episode and one corrected Adaptive C2 episode each match 46,630 unique UDP transmission and reception identifiers, with zero registered deficit. Applied PRB quota-time decreases from 1040.000 to 726.899 PRB s (30.106%); the sampled native RLC peak increases from 3.165 to 123.435 kB and packet delivery-time P95 from 10.947 to 93.717 ms. C2 makes 21 confirmed adaptive quota changes but receives no live UE goodput feedback. Source-level checks also identify matching hashes for 16 of 17 selected static configuration fields. The observations establish a bounded, traceable single-workload resource - traffic-performance trade-off; they do not certify R2 operation, statistical superiority, measured energy savings or a global Pareto frontier. Open reproducibility and prospective QoS threshold requirements are specified.
Author: Khoshaba, Oleksandr
Institution: Vinnytsia National Technical University
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7235299000
