Artifact-Based Soft Failure Detection in Computing Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional soft failure detection systems in computing systems often misclassify processes as 'normal' due to relying on excessive artifact thresholds, failing to detect soft failures characterized by abnormal behavior that generates too few artifacts, which are difficult to diagnose and can lead to system degradation.
Innovation Solution
A method that monitors artifacts generated by processes, determines if the number is below a given threshold, analyzes the current state for abnormal behavior, and generates notifications, using historical data and machine learning to predict normal behavior and detect soft failures even when artifact counts are low.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional soft failure detection systems use artifact thresholds to detect failures, then they can detect abnormal behaviors in real time, but they mistakenly classify processes as normal when artifact counts are low due to soft failures
Solution Approach 1:
The patent inverts the conventional approach by not only monitoring when artifact counts exceed thresholds but also monitoring when artifact counts fall below expected thresholds. This inversion enables detection of soft failures characterized by abnormally low artifact generation, allowing the system to identify processes that appear normal by conventional standards but are actually degraded.
Solution Approach 2:
The system continuously monitors artifact generation rates and compares them against expected values derived from historical data and machine learning models. When deviations are detected (either excessive or insufficient artifacts), the system triggers analysis of process states and generates notifications, creating a feedback loop that adapts to changing system conditions and improves detection accuracy over time.
2Reliability
If the system monitors artifact generation to detect soft failures, then detection capability improves, but resource consumption increases
Solution Approach 1:
The system applies partial monitoring by focusing artifact generation monitoring only on specific processes identified as potentially problematic through machine learning analysis. Rather than continuously monitoring all processes equally, the system selectively monitors artifacts from processes showing signs of degradation or abnormal behavior, reducing overall resource consumption while maintaining effective detection capability.
Solution Approach 2:
The system uses machine learning models trained on historical data to automatically determine which processes require monitoring and what their expected artifact generation patterns are. This self-service approach allows the system to adapt its monitoring strategy based on learned patterns, avoiding unnecessary resource expenditure on processes that are likely to be normal or not at risk.
3Ease of manufacture
If conventional systems use fixed artifact thresholds for failure detection, then implementation is simple, but they cannot adapt to changing system behavior patterns
Solution Approach 1:
The system transitions from static fixed thresholds to dynamic adaptive thresholds that automatically adjust based on historical data and machine learning models. The expected artifact generation values are continuously updated based on actual process behavior patterns, allowing the detection system to adapt to changing system conditions, seasonal variations, and evolving operational patterns without requiring manual reconfiguration.
Solution Approach 2:
The system performs preliminary training using historical data to establish baseline expected values for artifact generation before actual failure detection begins. This preliminary action creates the foundation for subsequent adaptive monitoring, allowing the system to learn normal behavior patterns and detect deviations without requiring complex real-time decision-making logic.
Data Source
AI summary
A method detects soft failures as follows. A set of artifacts being generated by at least one process in a system is monitored. A number of artifacts being generated by the process is determined to be below a given threshold in response to the monitoring. The process is monitored in response to the determination. A current state of the process is determined in response to the analyzing. A notification is generated in response to the current state of the process including a set of abnormal behaviors.