Tunable Predicate Discovery for Cloud Performance Anomaly Diagnosis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Complex systems, such as cloud computing platforms, face challenges in managing and improving performance due to heterogeneity and complexity, making it difficult to track key performance indicators and diagnose anomalies effectively.
Innovation Solution
The implementation of tunable predicate discovery, which involves a method to obtain a dataset, determine anomaly scores, and generate a ranked list of predicates to identify conditions associated with anomalies, allowing for efficient diagnosis of performance issues without relying on explicit supervision signals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional monitoring and tracking methods are used in complex cloud computing systems, then complete performance data can be collected, but the complexity and cost of diagnosing anomalies increases significantly
Solution Approach 1:
The patent extracts and focuses only on the most relevant features and predicates from the vast cloud computing performance data. Instead of analyzing all collected performance data, the system identifies and extracts key predicates that are most indicative of anomalies, thereby reducing diagnostic complexity while maintaining detection accuracy.
Solution Approach 2:
The patent introduces an intermediary layer of predicate discovery and anomaly scoring that mediates between raw performance data and final anomaly diagnosis. This intermediary processing layer transforms complex multi-dimensional performance data into simplified anomaly scores and ranked predicate lists, making the diagnosis process more manageable.
2Loss of information
If comprehensive performance monitoring is implemented across multiple data centers, then system visibility is improved, but the computational overhead and cost of analysis increases
Solution Approach 1:
The system extracts only the most salient performance indicators and anomalies from the comprehensive multi-data center monitoring data. By identifying and focusing on key predicates rather than processing all monitoring data equally, the system maintains complete system visibility while reducing computational overhead for analysis.
Solution Approach 2:
The patent applies local quality by treating different performance metrics and data sources with different levels of analysis intensity. Not all performance data requires the same computational resources for analysis - the system dynamically adjusts the depth of analysis based on the local characteristics and anomaly scores of different predicates and data sources.
3Measurement precision
If detailed anomaly analysis is performed on all performance data, then detection accuracy is improved, but the time required for diagnosis increases
Solution Approach 1:
The system extracts and prioritizes the most likely anomaly-causing predicates by generating ranked lists based on anomaly scores. This allows the system to achieve high detection accuracy by focusing detailed analysis only on the top-ranked predicates rather than performing equally detailed analysis on all possible causes.
Solution Approach 2:
The patent applies partial action by performing detailed anomaly analysis only on the most promising predicates identified through initial scoring and ranking, rather than conducting exhaustive detailed analysis on all performance data. This selective approach maintains detection accuracy for critical anomalies while reducing overall diagnosis time.
Data Source
AI summary
The described implementations relate to tunable predicate discovery. One implementation is manifest as a method for obtaining a data set and determining anomaly scores for anomalies of an attribute of interest in the data set. The method can also generate a ranked list of predicates based on the anomaly scores and cause at least one of the predicates of the ranked list to be presented.