vGPU-Aware Virtual Machine Placement in Distributed Cloud Systems

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current virtual machine placement techniques in cloud computing environments fail to optimally utilize virtual graphic processing unit (vGPU) resources, leading to underutilization and compatibility issues, particularly during initial placement and subsequent deployments, which requires manual intervention and power-off operations to resolve.

Innovation Solution

A system and method that considers vGPU requirements for placing virtual computing instances on hosts, using a graphics resource management module to select suitable hosts based on vGPU needs, ensuring efficient resource utilization and avoiding heterogeneous profile restrictions on NVIDIA GRID cards.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If current placement techniques consider only memory and CPU utilization for VM placement, then placement simplicity is maintained, but vGPU resource utilization becomes non-optimal

Engineering Contradiction:

Improveplacement simplicityVSAvoidvGPU resource utilization

Core Design Contradiction:

Ease of operationVSProductivity

Solution Approach 1:

The system performs preliminary classification of VMs into graphics-oriented and non-graphics-oriented categories before placement. This advance categorization enables the placement algorithm to apply appropriate strategies for each type, ensuring optimal vGPU resource utilization from the initial placement without requiring complex real-time adjustments later.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent applies different placement strategies to different types of VMs based on their specific characteristics. Graphics-oriented VMs are placed using vGPU-aware algorithms that consider GPU resource availability and compatibility, while non-graphics-oriented VMs use traditional CPU and memory-based placement. This localized approach optimizes vGPU utilization without unnecessarily complicating the overall placement system.

Inventive Principle:

Principle #3Local quality

2Speed

If VMs are placed randomly with respect to vGPU resources, then placement speed is maintained, but vGPU underutilization occurs

Engineering Contradiction:

Improveplacement speedVSAvoidvGPU resource utilization

Core Design Contradiction:

SpeedVSProductivity

Solution Approach 1:

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent changes the placement parameters based on VM type. For graphics-oriented VMs, the system considers vGPU profile requirements, available GPU resources, and compatibility constraints as placement parameters. For non-graphics-oriented VMs, traditional CPU and memory parameters are used. This parameter adaptation enables efficient placement that optimizes vGPU utilization while maintaining placement speed.

Inventive Principle:

Principle #35Parameter changes

3Productivity

If administrators manually power off VMs to resolve vGPU underutilization, then vGPU resource allocation is optimized, but operational complexity and time consumption increase

Engineering Contradiction:

ImprovevGPU resource utilizationVSAvoidtime for manual intervention

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The system implements automated monitoring and management of vGPU resource utilization. When vGPU underutilization is detected, the system automatically identifies affected VMs, determines suitable target hosts, and performs migration operations without requiring administrator intervention. This self-service approach maintains optimal vGPU utilization while eliminating the time loss associated with manual operations.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent incorporates continuous monitoring of vGPU resource utilization that provides feedback to the placement and migration system. When utilization patterns indicate underutilization, the system automatically triggers corrective actions such as VM migration to appropriate hosts. This feedback mechanism ensures optimal vGPU resource allocation while eliminating the need for manual administrator intervention and associated time losses.

Inventive Principle:

Principle #23Feedback

4Adaptability or versatility

If VMs are placed without considering vGPU requirements, then placement flexibility is maintained, but VM compatibility for virtualization operations deteriorates

Engineering Contradiction:

Improveplacement flexibilityVSAvoidVM compatibility for virtualization operations

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS10789668B2Intelligent provisioning of virtual graphic processing unit resources

Publication Date: 2020.09.29 VMWARE INC

AI summary

A system and method for placing virtual computing instances in a distributed computer system utilizes virtual graphic processing unit (vGPU) requirements of the virtual computing instances to place the virtual computing instances on a plurality of hosts of the distributed computer system. Each virtual computing instance with vGPU requirements is placed on one of the plurality of hosts in the distributed computer system based on the vGPU requirements of that virtual computing instance. Each virtual computing instance without vGPU requirements is placed on one of the plurality of hosts in the distributed computer system without any vGPU consideration.