Self-learning network correlation for agile fault management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current surveillance platforms in Communications Service Providers (CSPs) rely heavily on human experience and knowledge to infer and identify network and service problems, which is inefficient in rapidly changing Network Function Virtualization (NFV) and Software-Driven Networks (SDN) environments, where traditional expert systems struggle to adapt and operationalize correlation rules in real-time.
Innovation Solution
A self-learning correlation method combining unsupervised and supervised machine-learning techniques to discover and operationalize correlation rules within a surveillance platform, reducing dependency on human experience and enabling automated knowledge creation for real-time network event processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional expert systems with manual correlation rules are used, then system reliability is maintained through human expertise, but adaptability deteriorates as networks change rapidly with NFV and SDN
Solution Approach 1:
The surveillance platform automatically discovers correlation rules by analyzing network events and alarms without human intervention. The system self-learns from historical data, identifies patterns, and operationalizes rules autonomously, eliminating the manual knowledge capture process and enabling rapid adaptation to network changes.
Solution Approach 2:
Manual expert analysis and rule creation is replaced with automated machine learning algorithms. The system uses unsupervised learning to discover patterns in network events and supervised learning to validate and operationalize rules, substituting human mechanical processes with automated computational methods.
2Measurement precision
If more correlation rules are manually created to cover all network scenarios, then measurement precision improves, but device complexity increases
Solution Approach 1:
The correlation rules are made dynamic and adaptive rather than static and fixed. The system continuously learns from new network events and updates its rules automatically, allowing the rule set to evolve with network changes without requiring manual expansion, thus maintaining precision while controlling complexity.
Solution Approach 2:
The system incorporates feedback loops where detected problems and their resolutions are fed back into the learning process. This allows the surveillance platform to continuously refine its correlation rules based on actual network performance and problem patterns, improving accuracy without proportionally increasing system complexity.
3Productivity
If traditional surveillance platforms process all network events, then completeness of monitoring is maintained, but productivity decreases due to processing non-significant alarms
Solution Approach 1:
The system extracts and focuses on significant correlation patterns from the overwhelming volume of network events. By using machine learning to identify and extract only the most relevant event sequences and alarm patterns, the system filters out non-significant noise while preserving critical information about root causes and problem relationships.
Data Source
Figure 1
Figure 2A~2C
Figure 3A~3B
AI summary
Various aspects of the subject technology relate to methods, systems, and machine-readable media for self-correlating network operations. The method includes receiving a stream of network messages, the stream of network messages comprising a variety of network events for various network devices. The method also includes identifying patterns within the stream of network messages, the patterns comprising groupings of the variety of network events. The method also includes determining for each pattern an appropriate operationalization scenario. The method also includes operationalizing the patterns as correlation rules in a correlation engine to automatically detect or predict network alarms from input provided by a fault management system.