Storage Array Anomaly Detection and Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data protection systems and storage array performance monitoring lack effective anomaly detection and classification, particularly in distinguishing between spikes and drops in performance parameters like I/O, bandwidth, and latency, which can impact data integrity and system reliability.
Innovation Solution
A method and apparatus that receive data samples from a storage array, determine anomalies, and reclassify them as spikes or drops using a database of known anomalies and user-defined settings, employing statistical calculations to identify deviations from standard deviations and consecutive anomalies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional anomaly detection methods are used, then data protection systems can identify performance issues, but they cannot accurately distinguish between spikes and drops in performance parameters
Solution Approach 1:
The patent segments anomaly detection into distinct classification categories (spikes vs. drops) with specific criteria for each. Performance parameters are divided into different anomaly types based on directional deviation from baseline, allowing precise classification while maintaining manageable system complexity through structured segmentation of the detection space.
Solution Approach 2:
The system dynamically adjusts anomaly detection by continuously comparing current performance samples against baseline values and standard deviations. The classification methodology adapts to varying performance conditions by recalculating baselines and thresholds, enabling accurate spike/drop distinction without requiring static, overly complex predefined rules.
2Measurement precision
If detailed statistical analysis is performed on each data sample, then anomaly detection accuracy improves, but processing time and computational resources increase
Solution Approach 1:
The patent applies partial statistical analysis by focusing computations only on samples identified as potential anomalies rather than performing exhaustive analysis on all data. The system calculates standard deviations and compares samples to baselines selectively, achieving sufficient detection accuracy without the computational overhead of complete statistical processing of every performance metric.
Solution Approach 2:
The system changes parameters dynamically by adjusting baseline values and standard deviation thresholds based on observed performance patterns. This allows the detection mechanism to adapt its sensitivity and computational requirements, maintaining high accuracy while optimizing processing time by modifying statistical parameters rather than increasing computational effort uniformly across all samples.
Data Source
AI summary
In one aspect, a method includes receiving samples of data generated from a storage array related to a performance parameter; determining, for each sample, whether a sample is anomaly; and determining, for each sample identified as an anomaly, whether the anomaly should be reclassified to a spike or a drop.