RAN Application Capacity Prediction with MEC Compute Forecasting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current mechanisms for managing applications in multi-access edge computing (MEC) devices via radio access networks (RANs) fail to optimize computing resources efficiently, leading to network congestion, reduced throughput, and network disruptions due to reactive monitoring and rebalancing of traffic, which consumes significant computing and networking resources.
Innovation Solution
A monitoring system that utilizes a machine learning model to predict application capacity and compute requirements by preprocessing logs from user devices, RANs, and MEC devices, generating labeled and normalized images of RAN coverage areas, and optimizing resource allocation based on these predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If reactive monitoring and rebalancing of traffic is used, then network disruptions can be responded to, but computing and networking resources are significantly consumed
Solution Approach 1:
The system performs preliminary actions by predicting application capacity requirements and generating labeled and normalized images of RAN coverage areas before network disruptions occur. The machine learning model proactively identifies potential capacity issues and triggers preventive rebalancing, eliminating the need for reactive monitoring and resource consumption during disruption response.
2Reliability
If reactive traffic rebalancing is performed, then network issues can be addressed, but network disruptions and lost traffic occur
Solution Approach 1:
The system performs preliminary analysis by processing logs from user devices, RANs, and MEC devices to predict application capacity requirements before network issues manifest. The machine learning model generates labeled and normalized images of RAN coverage areas in advance, enabling preventive traffic rebalancing that avoids network disruptions and maintains continuous throughput.
3Adaptability or versatility
If current mechanisms for handling applications are used, then applications can be provided to user devices, but computing resources become overloaded
Solution Approach 1:
The system performs preliminary prediction of application capacity requirements using machine learning models before applications are provisioned. By analyzing historical and real-time data from logs and generating labeled images of RAN coverage areas, the system predicts future resource needs and optimizes resource allocation in advance, preventing computing resource overload while maintaining application provision capability.
Solution Approach 2:
The system implements feedback mechanisms by continuously processing logs from user devices, RANs, and MEC devices to train and update machine learning models. This feedback loop enables the system to learn from actual resource usage patterns and improve its predictions, optimizing computing resource allocation dynamically while supporting application versatility.
Data Source
AI summary
A device may receive user device application performance logs, radio access network (RAN) performance logs, and multi-access edge computing (MEC) compute logs, and may train a machine learning model with the user device application performance logs, the RAN performance logs, and the MEC compute logs. The device may identify application capacity requirements and MEC compute requirements associated with images of RAN coverage areas of RANs, and may generate labeled and normalized images of RAN coverage areas. The device may receive real time MEC compute data, RAN performance data, and user device application data, and may identify one of the labeled and normalized images of the RAN coverage areas that matches a RAN coverage area of a particular RAN. The device may predict an application capacity and MEC compute requirements for the particular RAN, and may perform one or more actions based on the application capacity and the MEC compute requirements.