Network Monitoring Through Topological Alarm Dependency Summaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Complex communication networks generate overwhelming amounts of alarm logs, making it difficult for human operators to distinguish real fault symptoms from noisy alarms and locate root causes effectively.
Innovation Solution
A data processing device and method that organize network resources by topology, analyze alarm dependencies, and summarize these dependencies to provide training data for classifiers, enabling automated fault identification and root cause analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If human operators manually analyze alarm logs to identify faults and locate root causes, then measurement precision and reliability can be maintained, but productivity decreases due to overwhelming data volumes and long analysis times
Solution Approach 1:
The alarm dependency structure is segmented into multiple levels through topological analysis. The system divides the network topology into hierarchical levels and identifies alarm dependencies at different levels, allowing the classifier to process and analyze alarms in a structured, hierarchical manner rather than processing all alarms uniformly, thus improving efficiency while maintaining analysis precision
Solution Approach 2:
The system performs preliminary topological analysis and alarm dependency identification before classification. By pre-processing the alarm data to establish dependency relationships and topological contexts, the system prepares structured training data that enables the classifier to make faster, more accurate fault detection decisions without requiring manual analysis of all alarm details
2Measurement precision
If the system processes detailed alarm dependency structures with specific topological information, then measurement precision and fault detection accuracy improve, but device complexity increases due to the need for topological analysis and data processing
Solution Approach 1:
The patent introduces an intermediary data processing layer that transforms raw alarm data into structured training data. This intermediary layer performs topological analysis and alarm dependency identification, acting as a mediator between the complex network topology and the classification task, thereby maintaining high detection accuracy while organizing complexity into manageable processing stages
Solution Approach 2:
The system changes the representation parameters of alarm data by transforming detailed alarm dependency structures into summarized training data formats. By converting complex topological alarm relationships into standardized training data with key features, the system maintains measurement precision while reducing the complexity burden on the classification system
3Reliability
If the system generates and processes large volumes of alarm data and dependency structures, then the ability to distinguish real faults from noisy alarms improves, but loss of information increases due to data volume overwhelming operators
Solution Approach 1:
The system extracts and separates the essential information from the alarm data through topological analysis and dependency identification. By extracting only the critical alarm relationships and topological contexts needed for fault detection, the system creates condensed training data that maintains reliability while preventing information overload for both the system and operators
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A device for providing training data to a classifier for monitoring a network of resource instances is disclosed. Resource instances are organized according to a topology, have respective resource types and may generate alarms having alarm types. The device comprises means (1501, 1502) for obtaining a first alarm dependency structure, ADS, (400) comprising dependencies (401) between alarms, means (103) for summarizing the ADS, and means (103) for providing classifier training data comprising an ADS network state class and the summarized alarm dependency structure, SADS, resulting from said assigning, replacing and merging. Also disclosed is a device for monitoring a network - this device comprises means for obtaining an ADS, summarizing an ADS and obtaining a network state class from a classifier based on the summarized ADS. Methods corresponding to both devices are also disclosed.