Open RAN Distributed Unit Failure Prediction and Graceful Handover
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing radio access networks face challenges in predicting and proactively addressing failures of distributed units, leading to dropped calls and inefficient maintenance procedures, particularly in 5G networks with heterogeneous vendor components.
Innovation Solution
Implementing machine learning models that analyze telemetry data to predict failures and perform graceful shutdowns or upgrades, minimizing service interruptions by intelligently handing over calls and automating maintenance processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are built from telemetry data to predict failures, then failure prediction accuracy is improved, but data processing complexity increases
Solution Approach 1:
The system automatically builds and updates machine learning models using telemetry data without requiring manual intervention. The model self-improves by continuously learning from new failure patterns and network conditions, reducing the need for complex manual data processing and model training operations.
Solution Approach 2:
The system implements feedback loops where predicted failures are monitored and actual outcomes are fed back into the machine learning model. This continuous feedback mechanism refines prediction accuracy over time while automating the data processing complexity through systematic learning from historical failure data.
2Reliability
If graceful shutdown procedures are implemented to hand over calls, then service continuity is improved, but shutdown process complexity increases
Solution Approach 1:
The system performs preliminary actions by predicting failures before they occur and pre-planning graceful shutdown procedures. The handover process is initiated in advance, allowing the system to prepare alternative routing paths and notify relevant parties before the actual shutdown, thereby simplifying the execution of service continuity.
Solution Approach 2:
The system introduces an intermediary coordination mechanism that manages the handover process between distributed units. This intermediary layer abstracts the complexity of coordinated shutdowns from individual components, enabling service continuity through automated mediation between multiple system elements.
3Reliability
If proactive remedial actions are taken to address predicted failures, then network reliability is improved, but response time requirements increase complexity
Solution Approach 1:
The system takes preliminary remedial actions based on predicted failures before actual outages occur. By anticipating potential issues and executing preventive measures in advance, the system eliminates the need for reactive responses under time pressure, thereby simplifying the response time requirements while maintaining high network reliability.
Solution Approach 2:
The system skips unnecessary diagnostic steps by using machine learning predictions to directly identify required remedial actions. Instead of performing comprehensive troubleshooting when failures occur, the system rushes through to the solution by leveraging pre-computed failure patterns and predicted root causes, thereby reducing response time complexity.
4Measurement precision
If continuous telemetry data streaming is implemented, then failure detection accuracy is improved, but data volume and processing load increase
Solution Approach 1:
The system extracts only the critical telemetry data points relevant to failure detection from the continuous data stream. By filtering and selecting only essential parameters such as CPU usage thresholds, memory errors, and connection anomalies, the system maintains high failure detection accuracy while significantly reducing the volume of data requiring processing.
Solution Approach 2:
The system uses feedback mechanisms to dynamically adjust data collection intensity based on network conditions. When the network is stable, data streaming is reduced; when anomalies are detected, data collection intensity increases. This feedback-driven approach maintains failure detection accuracy while minimizing unnecessary data processing load.
Data Source
AI summary
A disclosed method may include (i) building, based on telemetry data from an open radio access network, a machine learning model that predicts when a candidate distributed unit within the open radio access network will experience a failure, (ii) detect, by applying the machine learning model that predicts when the candidate distributed unit will shut down, that a specific distributed unit will experience a specific failure, and (iii) perform, in response to detecting that the specific distributed unit will experience the specific failure, a remedial action that addresses the specific failure. Related systems and computer-readable mediums are further disclosed.