Virtual Machine CPU Load Calculation via Direct NIC Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In information processing systems with virtual machines, estimating CPU load accurately is challenging due to increased I/O data transmission and the ability of virtual machines to perform direct data transmission via physical NICs without hypervisors, complicating network configuration management across multiple hypervisors.

Innovation Solution

A load calculation program that acquires processor usage and data transmission information, detects virtual network interfaces not routing through the hypervisor, and calculates CPU loads based on network configurations to estimate CPU usage for each virtual machine with high accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If virtual machines perform direct data transmission via physical NIC without hypervisor involvement, then I/O data transmission capability is improved, but CPU load estimation accuracy deteriorates

Engineering Contradiction:
ImproveI/O data transmission capabilityVSAvoidCPU load estimation accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent segments virtual network interfaces into two categories: those that route through the hypervisor and those that perform direct transmission via physical NIC. By detecting and distinguishing between these two types of interfaces, the system can separately calculate CPU loads for each category, thereby maintaining accurate CPU load estimation even when some virtual machines use direct transmission paths that bypass the hypervisor.

Inventive Principle:
Principle #1Segmentation

2Adaptability or versatility

If multiple different hypervisors are mounted on physical machines, then system versatility is improved, but network configuration management complexity increases

Engineering Contradiction:
Improvehypervisor compatibilityVSAvoidnetwork configuration management
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements a universal detection mechanism that can identify virtual network interface characteristics across multiple different hypervisor types. The detection method focuses on observable behavior (whether data transmission routes through the hypervisor) rather than hypervisor-specific configurations, enabling the system to adapt to various hypervisor implementations without requiring separate management approaches for each type.

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

3Adaptability or versatility

If manual management of virtual NIC network configurations is performed, then configuration flexibility is improved, but management time and effort increase

Engineering Contradiction:
Improveconfiguration flexibilityVSAvoidmanagement time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent implements an automated detection mechanism that self-identifies virtual network interfaces performing direct transmission without requiring manual configuration or intervention. The system automatically detects the transmission path characteristics, categorizes interfaces accordingly, and calculates CPU loads based on detected configurations, thereby eliminating the need for manual management while preserving configuration flexibility.

Inventive Principle:
Principle #25Self-service

Data Source

PatentUS10169102B2Load calculation method, load calculation program, and load calculation apparatus
Publication Date: 2019.01.01 FUJITSU LTD
  • US10169102B2 patent drawing
  • US10169102B2 patent drawing
  • US10169102B2 patent drawing

AI summary

A load calculation program that causes a computer a process includes acquiring processor usage information including usage of a processor of an managed computer, which is a management object, and usage of the processor for each of a plurality of virtual machines generated by a hypervisor executed on the managed computer; acquiring data transmission information including a data transmission amount for each virtual network interface used by the plurality of virtual machines; detecting a first virtual network interface that performs data transmission without routing through the hypervisor among the plurality of virtual network interfaces, based on the processor usage information and the data transmission information; and calculating load information including processor usage for data transmission of each of the plurality of virtual machines, based on whether or not each of the virtual machines uses the first virtual network interface.