Application Fault Detection via Dynamic Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Detecting application faults in complex systems is challenging due to the variability of application instances and hardware, making it difficult to isolate faults and predict performance degradation, which can impact upstream applications and hardware resource availability.
Innovation Solution
An application fault detection architecture that monitors performance metrics, determines performance statuses, and sends alerts when degradation thresholds are exceeded, allowing for real-time fault detection and forecasting based on individual application behavior.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a universal set of rules is applied to detect faults, then the detection process is simplified, but it cannot accurately detect faults due to variability of application instances and hardware
Solution Approach 1:
The system changes the parameters used for fault detection from universal fixed thresholds to dynamic, instance-specific parameters. It collects performance metrics from individual application instances and uses machine learning models to determine customized performance status thresholds for each instance, thereby achieving accurate fault detection while maintaining operational simplicity through automation.
2Measurement precision
If performance metrics are monitored for every application instance individually, then fault isolation accuracy is improved, but system complexity increases
Solution Approach 1:
The system creates a multi-functional monitoring architecture that serves multiple purposes: it collects performance metrics for individual instance monitoring, trains machine learning models using aggregated data from multiple instances, and provides both individualized fault detection and system-wide fault isolation. This universal approach reduces overall complexity by sharing infrastructure and data across functions.
Solution Approach 2:
The system creates simplified copies or representations of application performance through performance metrics and machine learning models. Instead of directly analyzing complex application behaviors, it monitors copied performance data that captures essential fault indicators, reducing monitoring complexity while maintaining detection accuracy.
3Ease of manufacture
If fault detection relies on generic rules, then implementation is easier, but downstream faults cannot be isolated from upstream application performance degradation
Solution Approach 1:
The system segments fault detection into multiple levels: individual application instance monitoring, downstream application monitoring, and upstream impact analysis. By dividing the detection process into separate monitoring streams and using machine learning to correlate them, it can identify whether performance degradation originates from upstream applications or occurs independently in downstream applications, thereby achieving accurate fault source identification.
Data Source
AI summary
An application monitoring device that includes a memory operable to store an application and a fault detection engine implemented by a processor. The fault detection engine is configured to obtain a set of application metric values for the application. Each application metric value indicates a performance level of the application. The fault detection engine is further configured to compare each application metric value to a set of application metric value ranges and to determine a performance status value for each application metric value based on the comparison. The fault detection engine is further configured to determine a warning level for the application and to determine that the warning level exceeds the fault detection threshold value. The fault detection engine is further configured to trigger an alert indicating a fault has been detected in the application in response to the determination.