Anomaly Detection for Virtualized RAN Infrastructure
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Anomaly detection in virtualized Radio Access Networks (vRANs) is challenging due to high monitoring overhead, complex infrastructure, and rare occurrence of anomalies, making it difficult to collect balanced datasets for training models, especially in resource-constrained edge deployments.
Innovation Solution
Decoupling anomaly detection at the infrastructure layer from the VNF layer allows for tailored techniques and reduces monitoring overhead by training separate models for VNF and infrastructure anomalies, using offline learning and sampling techniques to minimize resource impact and ensure real-time performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If monitoring features for all network functions and infrastructure is implemented, then anomaly detection accuracy is improved, but monitoring overhead increases significantly
Solution Approach 1:
The patent segments the monitoring system into two distinct components: (1) infrastructure monitoring that collects resource utilization data from physical/virtualized infrastructure, and (2) VNF monitoring that collects performance data from virtualized network functions. This segmentation allows each component to be optimized independently, reducing overall monitoring overhead while maintaining comprehensive anomaly detection capability.
Solution Approach 2:
The patent extracts and separates the monitoring of VNFs from infrastructure monitoring. By taking out VNF-specific monitoring metrics and handling them separately from infrastructure resource metrics, the system reduces the complexity of unified monitoring while enabling specialized anomaly detection for each layer.
2Reliability
If monitoring features for multiple collocated vRAN cells is implemented, then anomaly detection coverage is improved, but CPU overhead increases beyond available resources
Solution Approach 1:
The patent applies partial monitoring by selecting a subset of critical features and metrics to monitor rather than all possible parameters. For each vRAN cell, the system monitors only the most relevant performance indicators, achieving sufficient anomaly detection coverage with reduced CPU overhead that fits within resource constraints.
Solution Approach 2:
The patent implements local quality by tailoring the monitoring configuration to each specific deployment scenario. The system adapts which features to monitor and at what frequency based on local resource availability, cell importance, and traffic patterns, rather than applying uniform monitoring across all cells.
3Measurement precision
If comprehensive feature monitoring is implemented for vRAN functions, then anomaly detection capability is improved, but network performance deteriorates due to scheduling deadline violations
Solution Approach 1:
The patent implements periodic monitoring where features are collected at optimized intervals rather than continuously. The monitoring frequency is adjusted based on the criticality of each metric and current network conditions, ensuring anomaly detection capability while allowing the DU to meet its strict scheduling deadlines for real-time signal processing.
Solution Approach 2:
The system monitors only the essential features needed for anomaly detection rather than all available metrics. This partial monitoring approach minimizes the computational burden on the DU, preventing scheduling deadline violations while maintaining sufficient anomaly detection capability.
Data Source
AI summary
To meet the stringent 5G radio access network (RAN) service requirements, layers one and two need to be processed in essentially real time. Thus, prompt anomaly detection is important to prevent negative impacts on customer experience, which is critical for mobile networks to meet the stringent service requirements. However, monitoring networks for anomalies is difficult due at least to (1) the resource constrained edge deployments in which the vRAN resides, (2) the variety of anomaly types and fault locations making anomalies difficult to detect, and (3) the low frequency of anomalies leading to unbalanced data sets for training, among others. The present application addresses these issues by decoupling anomaly detection at the infrastructure layer (servers, NICs, switches, etc.) from anomaly detection at the VNF layer (L1, high-DU, CU). This enables different techniques for identifying anomalies and for reducing the monitoring overhead that is tailored to each layer.