Edge Network Observability Using ML Summary Data Pipelines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Customers of cellular networks, such as enterprise customers, lack the ability to access and utilize network data and metadata generated by the network resources in a meaningful and insightful way, due to the vast volume of probe data produced, which is computationally prohibitive and difficult to analyze, and existing systems suffer from high costs, data transport issues, and data quality problems.
Innovation Solution
An adaptive observability platform with a software defined data pipeline control module that processes probe data at edge computing devices close to data generation, utilizing machine learning models to generate summary data, including performance indicators and corrective actions, reducing the burden of data transport and analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If customers access raw probe data from network resources, then they can obtain comprehensive network information, but the data volume becomes computationally prohibitive and difficult to analyze
Solution Approach 1:
The system extracts only the essential and relevant features from the vast probe data using machine learning models. Instead of providing customers with all raw data, the system identifies and extracts key performance indicators and meaningful patterns, delivering a curated subset of information that maintains analytical value while dramatically reducing complexity.
Solution Approach 2:
The patent introduces an intermediary processing layer between the probe data sources and customers. This intermediary system, utilizing machine learning models, acts as a mediator that transforms raw probe data into meaningful insights, bridging the gap between comprehensive data collection and manageable information delivery to customers.
2Loss of information
If comprehensive probe data is collected from network resources, then complete network visibility is achieved, but data transport costs and issues increase
Solution Approach 1:
The system extracts and transmits only the essential features and key insights derived from probe data rather than transporting all raw data. This extraction approach maintains complete network observability through sophisticated feature selection while dramatically reducing data transport volume and associated costs.
Solution Approach 2:
The patent segments the data processing function from data collection. Instead of collecting all data centrally and then transporting it, the system divides the workflow into local probe collection and centralized feature extraction, allowing selective transmission of only the most important extracted features rather than complete datasets.
3Measurement precision
If machine learning models process all probe data, then accurate network insights are generated, but computational costs become prohibitive
Solution Approach 1:
The system applies machine learning models to extract only the most relevant features and insights from probe data rather than processing all data comprehensively. This selective extraction approach maintains measurement precision for key network performance indicators while significantly reducing the computational energy required by focusing processing power on the most critical data elements.
Solution Approach 2:
The patent implements partial processing where machine learning models focus on extracting specific, high-value features from probe data rather than performing exhaustive analysis on all data. This partial action approach achieves sufficient measurement precision for network optimization while avoiding the prohibitive computational costs of complete data processing.
4Speed
If real-time network data is provided to customers, then network optimization responsiveness is improved, but data processing complexity increases
Solution Approach 1:
The system performs preliminary feature extraction and processing of probe data in advance, preparing meaningful insights before they are needed for network optimization decisions. This preliminary action creates a ready stream of processed information that can be delivered in real-time to customers, improving responsiveness while the actual complex processing occurs beforehand.
Solution Approach 2:
The patent extracts and delivers only the most critical real-time features and metrics that enable network optimization, rather than providing complete real-time data streams. This selective extraction approach maintains responsiveness for time-sensitive optimization while reducing the complexity of real-time processing by focusing on essential parameters only.
Data Source
AI summary
Technologies for providing insight data from an edge of a cellular network are described. One method includes first probe data in association with one or more components of the network. The method further includes providing the first probe data to a first model configured to determine first summary data based on the first probe data. The method further includes determining that the first summary data has been requested by a first northbound application. The method further includes providing the first summary data to a central server responsive to determining that the first summary data has been requested by the first northbound application.