Inferensys

Glossary

Admission Control Simulation

The computational modeling of the network function that decides whether to accept or reject a new bearer request based on available radio resources and the required Quality of Service (QoS) profile.
ML engineer managing model versions on laptop, version history visible, technical Git-like workflow.
RAN RESOURCE MANAGEMENT

What is Admission Control Simulation?

Admission control simulation is the computational modeling of the network function that decides whether to accept or reject a new bearer request based on available resources and the required quality of service.

Admission control simulation is the process of modeling the decision logic that grants or denies a new radio bearer or Protocol Data Unit (PDU) session in a virtualized RAN environment. It evaluates a request against the current resource utilization—including Physical Resource Blocks (PRBs), channel capacity, and backhaul bandwidth—and the specific Quality of Service (QoS) parameters, such as Guaranteed Bit Rate (GBR) or latency budget, to predict whether acceptance would degrade existing sessions.

This simulation is a critical component of a RAN digital twin, allowing operators to safely test and tune admission control algorithms offline before deployment. By replaying real-world traffic patterns or injecting synthetic load, engineers can analyze blocking probability, optimize resource reservation thresholds, and validate the behavior of AI-driven admission policies under stress conditions without impacting the live network.

CORE MECHANISMS

Key Features of Admission Control Simulation

Admission control simulation models the critical decision logic that determines whether a new bearer request is accepted or rejected based on available resources and Quality of Service (QoS) requirements. These key features define how the simulation replicates real-world gNB behavior.

01

QoS-Aware Gating Logic

The core decision engine evaluates incoming bearer requests against the Guaranteed Bit Rate (GBR) and Non-GBR resource pools. The simulation models the exact logic that checks if admitting a new flow would violate the QoS commitments of existing flows.

  • GBR Bearers: Admission is based on strict resource reservation checks
  • Non-GBR Bearers: Uses aggregate maximum bit rate (AMBR) enforcement
  • ARP Priority: Allocation and Retention Priority preemption logic is simulated to determine if a new high-priority bearer can evict a lower-priority one
5QI
QoS Class Identifier
1-15
ARP Priority Levels
02

Resource Block Utilization Modeling

The simulation maintains a real-time model of Physical Resource Block (PRB) consumption across the cell. Each admission decision is validated against the current and projected PRB utilization, accounting for the spectral efficiency of the requesting UE.

  • Tracks downlink and uplink PRB pools independently
  • Models the overhead of control channel elements (CCE)
  • Projects resource consumption based on the requested 5QI characteristics and the UE's reported Channel Quality Indicator (CQI)
273
Max PRBs (100MHz)
0.5ms
TTI Granularity
04

Slice-Aware Admission Control

In a 5G network slicing environment, admission control is performed per Network Slice Instance (NSI). The simulation models the hierarchical admission logic where a request must be admitted by both the slice's resource quota and the underlying cell's physical resources.

  • Models S-NSSAI (Single Network Slice Selection Assistance Information) identification
  • Enforces slice-specific resource quotas defined by the Network Slice Subnet Management Function (NSSMF)
  • Simulates the interaction between the Common Resource Block (CRB) pool and dedicated slice partitions
8
Max Slices per UE
SST
Slice/Service Type
05

Stochastic UE Arrival Modeling

The simulation generates realistic traffic load by modeling UE session arrivals as a Poisson process with configurable inter-arrival times. This allows stress-testing of the admission control algorithm under varying load conditions.

  • Models session holding times with exponential distributions
  • Generates heterogeneous traffic mixes (eMBB, URLLC, mMTC) simultaneously
  • Supports replay of real-world call trace data for scenario replay validation
λ
Arrival Rate Parameter
μ
Service Rate Parameter
06

Admission Rejection Cause Logging

Every rejection is logged with a precise cause code for post-simulation analysis. This telemetry is critical for tuning admission thresholds and identifying capacity bottlenecks in the digital twin before deploying changes to the live network.

  • Logs detailed rejection causes: PRB exhaustion, slice quota exceeded, ARP preemption failure, transport resource unavailable
  • Generates time-series KPIs: Admission Success Rate (ASR), Blocking Probability
  • Correlates rejections with UE location and serving cell load for root cause analysis
>99.9%
Target ASR
<0.1%
Blocking Probability
ADMISSION CONTROL SIMULATION

Frequently Asked Questions

Explore the core concepts behind modeling the network function that decides whether to accept or reject a new bearer request based on available resources and the required quality of service.

Admission control simulation is the computational modeling of the Radio Resource Control (RRC) and bearer management logic that determines whether a new service request can be accepted without degrading the Quality of Service (QoS) of existing sessions. The simulation works by creating a virtual network environment populated with realistic traffic generators and user mobility models, then subjecting a modeled MAC scheduler to a stream of connection requests. The admission control algorithm evaluates each request against the current resource block utilization, channel quality indicators (CQIs), and the Guaranteed Bit Rate (GBR) or Non-GBR requirements of the requested bearer. If the available resources, after accounting for a configurable reservation margin, are insufficient to meet the new request's QoS Class Identifier (QCI) profile without violating existing service level agreements, the simulation logs a rejection event. This allows engineers to tune the admission threshold to balance blocking probability against system overload.

Prasad Kumkar

About the author

Prasad Kumkar

CEO & MD, Inference Systems

Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.

His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.