Server Anomaly Detection via Regression Gradient Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing information handling systems lack effective anomaly detection mechanisms to prevent user experience impacts from exceeding server capacity, leading to delayed corrective actions.
Innovation Solution
Incorporating linear regression circuitry, gradient circuitry, and support vector machine circuitry to analyze usage data, detect abrupt changes, classify outliers, and update regression lines, enabling proactive management of disk space usage and preventing capacity overruns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional server capacity monitoring is used, then the system structure remains simple, but anomaly detection is delayed and user experience is impacted
Solution Approach 1:
The patent segments the anomaly detection task into three specialized circuitry components: linear regression circuitry for trend analysis, gradient circuitry for abrupt change detection, and support vector machine circuitry for outlier classification. Each component handles a specific aspect of anomaly detection, improving overall detection accuracy while maintaining modular system architecture that manages complexity.
Solution Approach 2:
The system performs preliminary actions by continuously analyzing usage data and updating regression lines in advance of capacity issues. The linear regression circuitry generates baseline trends, and the gradient circuitry prepares abrupt change detection thresholds beforehand, enabling proactive anomaly detection before server capacity is exceeded and user experience is impacted.
2Adaptability or versatility
If multiple types of server usage data are monitored simultaneously, then comprehensive anomaly detection is achieved, but data processing complexity increases
Solution Approach 1:
The patent implements universal circuitry that can handle multiple types of server usage data (disk space, memory, CPU, network) through the same analytical framework. The linear regression circuitry, gradient circuitry, and support vector machine circuitry work together as a multi-functional system that adapts to different data types, achieving comprehensive monitoring without proportionally increasing processing complexity.
Solution Approach 2:
The system manages multiple data types by changing parameters within the unified analytical framework. The linear regression circuitry adjusts to different data characteristics, the gradient circuitry modifies detection thresholds based on data type, and the support vector machine circuitry adapts classification parameters, enabling versatile multi-parameter monitoring while maintaining consistent processing methodology.
3Speed
If real-time anomaly detection is implemented, then timely corrective actions are enabled, but computational resource consumption increases
Solution Approach 1:
The patent applies partial action by using gradient circuitry to detect only abrupt changes that exceed predefined thresholds, rather than analyzing every data point in detail. The support vector machine circuitry classifies only suspicious outliers identified by previous stages, enabling real-time detection with reduced computational overhead by focusing resources on potentially anomalous data points.
Solution Approach 2:
The linear regression circuitry performs preliminary analysis to establish baseline trends and expected variations before the gradient and support vector machine circuitry conduct more intensive analysis. This preliminary action filters out normal variations early, reducing the computational burden on subsequent stages and enabling real-time detection with optimized resource consumption.
Data Source
AI summary
A management server includes linear regression circuitry to receive the data as input data, based on the input data, the linear regression circuitry generates a linear regression line for the input data. Gradient circuitry analyzes angles between two points of data within the input data. In response to a first angle between a first point of data and a second point of data being greater than a threshold angle, the gradient circuitry determines that an interval of data including the first and second points of data is an abrupt interval, removes first data within the abrupt interval from the input data, and updates the linear regression line accordingly. Support vector machine circuitry receives the input data, classifies outliers in the input data, and removes the outliers. The linear regression line is updated based on the removal of the outliers.