Network Fault Origin Assessment via Trace Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Detecting and locating failed network components in large computer networks is challenging due to the complexity and number of components, as existing methods may be limited by issues such as local minima and local maxima in statistical calculations.
Innovation Solution
The approach involves using network traces to collect information on successful and unsuccessful packet transmissions, attributing responsibility to network components through an initial assessment and refining it using a statistical calculation like expectation maximization, while employing a weighting function to address local minima issues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If statistical techniques like expectation maximization are used to locate faulty network components, then fault location capability is improved, but the method is susceptible to local minima and local maxima problems that reduce accuracy
Solution Approach 1:
The patent applies preliminary action by performing an initial assessment of network component fault likelihood before running the full statistical calculation. This initial assessment uses a simplified model to generate starting values that guide the expectation maximization algorithm, preventing it from converging to incorrect local minima or maxima. The preliminary action sets the stage for more accurate final results.
Solution Approach 2:
The patent changes parameters by introducing multiple initial assessment models with different assumptions and methodologies. Instead of relying on a single statistical model, the system evaluates fault likelihood using multiple parameter sets and initial conditions, then combines or selects the best results. This parameter diversity helps avoid local optimization traps.
2Measurement precision
If network traces are collected and analyzed for all components, then fault detection capability is improved, but the complexity of analyzing large networks increases
Solution Approach 1:
The patent applies segmentation by dividing the network into smaller analysis units or groups of components. Instead of analyzing all network components simultaneously, the system segments the network topology and processes traces for different segments separately. This reduces the computational complexity of analyzing large networks while maintaining comprehensive fault detection coverage across all segments.
Solution Approach 2:
The patent applies partial action by focusing initial analysis on a subset of network components or paths that are most likely to be faulty, based on trace patterns and network topology. Rather than uniformly analyzing all components with equal depth, the system performs partial detailed analysis on high-probability candidates while using lighter analysis on other components, reducing overall complexity while maintaining detection capability.
Data Source
AI summary
A network endpoint may issue network traces directed to another endpoint. Completed traces may provide information pertaining to possible paths between the endpoints. In response to a failed trace, a component along a possible path between the endpoints may be associated with a value that indicates an assessed contribution of the component to the network failure. The value may be used to initialize a parameter to a statistical calculation that iteratively refines the estimate to form a probability estimate regarding a likelihood that the component is contributing to a network fault.