International Journal of Network Management Computer Networks and Communications
Experimental Characterization of PCF‐Driven Network Slicing in Virtualized 5G SA Cores
Abstract
ABSTRACT The increasing demand for differentiated services in 5G networks necessitates efficient infrastructure utilization through network slicing (NS). While orchestration frameworks and machine learning approaches have been extensively studied, the direct impact of policy control function (PCF) policy parameters on observable slice performance remains underexplored in practical deployments. This paper presents a systematic experimental evaluation of PCF‐driven quality‐of‐service enforcement in a virtualized 5G Standalone core network using Open5GS and UERANSIM. We implement three representative network slices (URLLC, eMBB, and mMTC) with differentiated 5G QoS indicator (5QI), allocation and retention priority (ARP), and session aggregated maximum bit rate (Session‐AMBR) configurations and evaluate their behavior across five controlled scenarios: baseline capacity, congestion‐based prioritization, dynamic preemption, policy persistence, and scalability under device density. All reported quantitative results are specific to the evaluated Open5GS–UERANSIM VirtualBox deployment, traffic workload, and controlled bottleneck configuration. Under these conditions, the lower priority flow experienced 89.7% packet loss, whereas the post‐reconnection throughput varied by 0.9%. The co‐located three‐VM testbed, provisioned with one vCPU and 1‐GB RAM per VM, approached CPU saturation at 10 emulated UEs; this observed threshold is a testbed limitation rather than a general 5G SA capacity limit.