Anomaly Ranking via Normalized Fact Quantity Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In large-scale cloud-based deployments, the sheer volume of application performance data generates numerous anomalies, making it difficult for development personnel to prioritize which insights require immediate attention.
Innovation Solution
A method for ranking application insights by determining a normalized fact quantity change using a normalization factor and order factor, allowing for direct comparison and prioritization of anomalies across multiple dimensions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If automated anomaly detection is implemented to track application performance metrics, then detection capability is improved, but the volume of generated insights increases to the point where prioritization becomes difficult or impossible
Solution Approach 1:
The patent applies parameter changes by transforming the raw volume of insights into a normalized ranking metric. Each insight is assigned a rank based on normalized fact quantity change, which converts the unmanageable volume of data into a prioritized sequence that development personnel can effectively review. This resolves the contradiction by changing the parameter from raw count to normalized rank.
2Measurement precision
If multiple dimensions of performance metrics are tracked to improve analysis comprehensiveness, then measurement completeness is improved, but the complexity of prioritizing insights across dimensions increases
Solution Approach 1:
The patent merges multiple dimension-specific insights into a unified ranking system. By calculating normalized fact quantity change for each time-series across different dimensions and then ranking them collectively, the system combines comprehensive multi-dimensional analysis with a single prioritization mechanism. This resolves the contradiction by merging the complexity of multiple dimensions into a unified rank ordering.
Data Source
AI summary
Techniques are provided for ranking time-series including previously detected anomalous fact quantity changes over an associated time interval. Time-series are received, and for each time-series, a normalized fact quantity change is determined, and each time-series is ranked based in part on the normalized fact quantity change. A normalized fact quantity change may be determined by determining a normalization factor over the time interval, and then determining a product of the normalization factor and the absolute value of the fact quantity change of that time interval. Alternatively, a normalized fact quantity change may be the product of the normalization factor, a predetermined order factor, and the absolute value of the fact quantity change. The normalization factor is determined by analyzing the distribution of the fact quantity change over dimension values of the dimension(s) associated with the time-series to determine the number of values in which the fact quantity is concentrated.