Bayesian Root Cause Analysis for VM Resource Anomalies

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current systems fail to effectively identify and address the root cause of excessive resource usage in virtual machine systems, particularly in cloud environments, where hypervisors and guest machines consume substantial resources, leading to inefficiencies and potential issues like memory leaks or spinning processes.

Innovation Solution

A computer-implemented method using Bayesian acyclic graphs to monitor resource usage data for hypervisors and guest machines, identify outliers, and generate a root cause report by executing a root cause model that determines whether excessive resource utilization is due to a spinning process or genuine workload.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If resource usage monitoring is implemented for hypervisors and guest machines, then system efficiency and resource management are improved, but system complexity and computational overhead increase

Engineering Contradiction:

Improveresource management efficiencyVSAvoidsystem complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent introduces a Bayesian acyclic graph model as an intermediary computational framework that processes resource usage data from multiple sources (hypervisor and guest machines). This model acts as a mediator between raw monitoring data and root cause identification, systematically evaluating correlations and causal relationships without requiring complex custom analysis code for each scenario.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The monitoring system is segmented into distinct functional components: resource usage collection modules for hypervisors and guest machines, an outlier detection module, a Bayesian graph execution engine, and a root cause identification module. This segmentation allows each component to be optimized independently and facilitates modular deployment and maintenance.

Inventive Principle:

Principle #1Segmentation

2Measurement precision

If outlier detection and root cause analysis are performed on resource usage data, then accuracy in identifying spinning processes is improved, but computational time and processing overhead increase

Engineering Contradiction:

Improveoutlier detection accuracyVSAvoidcomputational time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary outlier detection on resource usage data before executing the full Bayesian acyclic graph analysis. By pre-identifying anomalous data points and filtering obvious outliers, the system reduces the scope and complexity of subsequent root cause analysis, thereby decreasing computational time while maintaining detection accuracy.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The Bayesian graph execution focuses computational resources on evaluating only the relevant causal pathways identified through preliminary outlier detection. Rather than exhaustively analyzing all possible causal relationships in the system, the method applies partial action by concentrating analysis on the most probable root causes, thus reducing overall computational overhead.

Inventive Principle:

Principle #16Partial or excessive action

3Measurement precision

If comprehensive resource usage data is collected from multiple sources, then root cause identification accuracy is improved, but data processing complexity and storage requirements increase

Engineering Contradiction:

Improveroot cause identification accuracyVSAvoiddata processing complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent merges resource usage data from multiple sources (hypervisor-level metrics and guest machine-level metrics) into a unified dataset that feeds into the Bayesian acyclic graph model. This consolidation is performed with standardized schemas and correlated timestamps, reducing the need for separate processing pipelines for each data source and simplifying overall data management.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The Bayesian acyclic graph model serves as a universal analytical framework that can process various types of resource usage data (CPU, memory, disk, network) from multiple sources using the same causal relationships and evaluation logic. This multi-functionality eliminates the need for source-specific analysis code and reduces data processing complexity.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS10970126B2Outlier and root cause determination of excessive resource usage in a virtual machine environment

Publication Date: 2021.04.06 KYNDRYL INC

AI summary

A computer-implemented method includes: monitoring, by a computing device, resource usage data for a hypervisor; monitoring, by the computing device, resource usage data for a guest machine; identifying, by the computing device, outliers in the resource usage data for the hypervisor and outliers in the resource usage data for the guest machine; executing, by the computing device, a root cause model based on the resource usage data for the hypervisor and the resource usage data for the guest machine; generating, by the computing device, a root cause report identifying results of the executed root cause model; and providing, by the computing device, the root cause report for display.