Flaw Classifier Model for Computing System Performance Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing techniques fail to accurately detect performance-affecting flaws in computing systems, often leading to unclear causes of poor performance, which can be due to hardware resource saturation or flaws, making it difficult for operators to determine whether upgrades or fixes are needed.

Innovation Solution

A system that uses a cascaded approach with an expected resource utilization model and a flaw classifier model to identify performance-affecting flaws by analyzing sensor measurements from computing systems, distinguishing between hardware and software issues, and recommending remediation measures such as workload staggering or hardware upgrades.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If specific flaw signatures are detected in log messages, back-traces and core dumps, then the presence of flaws can be properly detected, but flaws often go unnoticed because they may not manifest in these specific formats

Engineering Contradiction:
Improveflaw detection accuracyVSAvoidflaw detection completeness
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent segments the flaw detection process into multiple independent detection models, each specialized for different types of flaws (e.g., memory leaks, CPU leaks, file descriptor leaks). This segmentation allows the system to detect a broader range of flaws without missing those that don't manifest in traditional log formats.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent creates a universal detection framework that handles multiple types of flaws through a common architecture. The detection system is designed to work with various resource types (memory, CPU, file descriptors) and flaw manifestations, making it both precise for specific flaw types and comprehensive for overall flaw detection.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Adaptability or versatility

If operational resource modeling is used to analyze system performance, then performance can be analyzed over a wide range of hypothetical scenarios, but the system complexity increases due to creating and training multiple resource models

Engineering Contradiction:
Improveperformance analysis coverageVSAvoidmodel creation and training complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent performs preliminary actions by pre-training detection models during system setup or maintenance periods. This allows the models to be ready for immediate use without requiring complex real-time training, reducing the operational complexity while maintaining comprehensive scenario analysis capability.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent uses copying by creating simplified representations of resource models that can be replicated and applied across different scenarios. Instead of training complex models for each hypothetical scenario, the system copies and adapts pre-trained models, significantly reducing training complexity while maintaining analysis versatility.

Inventive Principle:
Principle #26Copying

3Measurement precision

If a cascaded approach with multiple models is used to identify performance-affecting flaws, then accurate classification of computing systems into regimes can be achieved, but the device complexity increases

Engineering Contradiction:
Improvesystem classification accuracyVSAvoidcascaded model system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The cascaded approach is segmented into distinct stages, each handling a specific aspect of flaw detection and classification. This segmentation makes the complex system more manageable by breaking it down into smaller, well-defined components with clear responsibilities.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent inverts the traditional approach by first detecting resource utilization patterns and then inferring flaw presence, rather than directly detecting flaws. This inversion simplifies the detection logic by working from observable metrics to underlying conditions, reducing system complexity while maintaining classification accuracy.

Inventive Principle:
Principle #13The other way round (Inversion)

Data Source

PatentEP3477477B1Identification of performance affecting flaws in a computing system
Publication Date: 2020.02.26 HEWLETT PACKARD ENTERPRISE DEV LP
  • EP3477477B1 patent drawingFigure 1
  • EP3477477B1 patent drawingFigure 2A
  • EP3477477B1 patent drawingFigure 2B

AI summary

A flaw classifier model classifies a computing system as one which contains or does not contain one or more flaws that affect a performance of the computing system. Inputs to the flaw classifier model may include an actual utilization of a resource of the computing system and an expected utilization of the resource of the computing system. The expected utilization of the resource may be determined by an expected resource utilization model. Inputs to the expected resource utilization model may include workload description parameter values and hardware description parameter values.