Network Packet Loss Detection Using Hierarchical Statistical Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Detecting and locating failed or malfunctioning network components in large computer networks is challenging due to complexity and variability in network topology and fault characteristics, leading to inconsistent results from statistical techniques.
Innovation Solution
A system that employs multiple statistical models and cost functions to analyze event data, generating a hierarchical aggregation graph to identify potential network error sources, and combining model outputs to converge on accurate entity identification, minimizing errors and noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single statistical technique is used to locate failed network components, then the analysis process is simple, but the accuracy and reliability of fault identification deteriorates due to network variability
Solution Approach 1:
The patent combines multiple statistical techniques (e.g., hypothesis testing, anomaly detection, machine learning models) into an integrated analysis system. Each technique processes network event data independently and their results are aggregated to identify failed components, thereby improving accuracy while managing complexity through modular architecture
Solution Approach 2:
The system creates a composite analytical approach by layering different statistical methods and weighting their outputs based on their performance characteristics. This composite methodology adapts to varying network conditions and fault types, maintaining high precision across diverse scenarios
2Measurement precision
If multiple statistical models are employed to improve fault detection accuracy, then the precision of error source identification is improved, but the system complexity increases
Solution Approach 1:
The patent segments the network into hierarchical levels (e.g., network segments, subnets, individual components) and applies different statistical models at each level. This segmentation allows complex multi-model analysis to be broken down into manageable stages, improving precision while controlling overall system complexity
Solution Approach 2:
The system employs a universal framework that can accommodate multiple statistical models through a common interface and standardized processing pipeline. This multi-functional architecture allows different models to be swapped or combined based on specific needs without redesigning the entire system
3Quantity of substance
If statistical techniques are applied to large networks with hundreds or thousands of components, then comprehensive coverage is achieved, but the difficulty of detecting and locating failures increases
Solution Approach 1:
The patent introduces hierarchical aggregation as an additional dimension, organizing thousands of network components into nested groups and layers. Statistical analysis is performed at multiple hierarchical levels, allowing failures to be detected and located more easily by progressively narrowing down from network-wide patterns to specific component identification
Data Source
AI summary
A computer implemented method includes determining a hierarchical graph of a network of entities. The method further includes determining event data from a plurality of events. The method also includes applying a plurality of statistical models on the event data corresponding to each of the plurality of the events and the hierarchical graph. Each of the statistical models generates a model output data for each of the plurality of events. The method further includes generating a model output data based on the model output data corresponding to each of the plurality of statistical models by using a first cost function. The method also includes generating a set of data based on the model output data by using a second cost function. The method further includes determining one or more entities of the network experiencing packet loss based on the set of data.