Graphics Processing Cluster Load Balancing via Virtual Machine Manager

Solution Overview

Problem

Current computer systems face limitations in graphics processing due to hardware constraints, leading to increased operational costs and power consumption, as they require multiple high-performance GPUs for load balancing, which is not efficiently distributed across servers.

Innovation Solution

A computer system and method that allows graphics processing requests to be dynamically assigned to a graphics processing cluster across multiple servers, enabling load balancing and reducing the need for local GPU resources, with a virtual machine manager directing requests to the most suitable server for processing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If multiple high-performance GPUs are deployed in each physical machine to improve graphics processing capability, then the graphics processing performance is improved, but the hardware cost and power consumption increase significantly

Engineering Contradiction:

Improvegraphics processing performanceVSAvoidpower consumption

Core Design Contradiction:

ProductivityVSUse of energy by stationary object

Solution Approach 1:

The patent merges graphics processing resources from multiple physical machines into a unified pool, where a single high-performance GPU can serve multiple virtual machines across different physical hosts. This consolidation reduces the total number of GPUs needed while maintaining or improving graphics processing performance through resource sharing and dynamic allocation.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent enables a single GPU to perform multiple functions by allowing it to serve different virtual machines with different graphics processing requirements. The GPU can dynamically allocate its capabilities to various VMs based on demand, making the hardware resource universal rather than dedicated to a single host.

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Productivity

If multiple high-performance GPUs are deployed in each physical machine to improve graphics processing capability, then the graphics processing performance is improved, but the hardware cost increases

Engineering Contradiction:

Improvegraphics processing performanceVSAvoidhardware quantity

Core Design Contradiction:

ProductivityVSQuantity of substance

Solution Approach 1:

The patent merges graphics processing resources from multiple physical machines into a unified pool, where a single high-performance GPU can serve multiple virtual machines across different physical hosts. This consolidation reduces the total number of GPUs needed while maintaining or improving graphics processing performance through resource sharing and dynamic allocation.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent creates virtual copies of GPU resources through virtualization, allowing a single physical GPU to be divided into multiple virtual GPU instances that can be allocated to different virtual machines. This virtual copying enables multiple VMs to access graphics processing capabilities without requiring proportional physical hardware for each.

Inventive Principle:

Principle #26Copying

3Adaptability or versatility

If graphics processing is confined to local physical machine hardware, then the system architecture remains simple, but the adaptability and resource utilization are limited

Engineering Contradiction:

Improveresource allocation flexibilityVSAvoidsystem architecture complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent introduces a virtualization layer as an intermediary between the physical GPU hardware and the virtual machines. This intermediary manages resource allocation, scheduling, and mapping, enabling flexible and adaptive resource distribution while abstracting the underlying complexity from the VMs and their workloads.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent segments the graphics processing resources and management functions into distinct modular components: physical GPU hardware, virtualization layer, and virtual machine instances. This segmentation allows independent management and allocation of resources to different VMs based on demand, enhancing adaptability while organizing system complexity into manageable segments.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS9183024B2Computer system with a cluster load balancer for graphics processing and graphics processing method thereof

Publication Date: 2015.11.10 WISTRON CORP

AI summary

A computer system and a graphics processing method with a cluster load balancer for graphics processing are provided. The computer system includes at least one physical machine (PM) and a graphics processing cluster. The at least one PM includes at least one virtual machine (VM) and a virtual machine manager (VMM). The graphics processing cluster includes graphics processing servers. Each of the graphics processing servers includes graphics processing units (GPUs). A main graphics processing server out of the graphics processing servers receives a graphics processing request provided by the VMM and the VM for assigning the graphics processing request to a minor graphics processing server out of the graphics processing servers. The minor graphics processing server provides a graphics processing result according to the graphics processing request and transmits the graphics processing result to the VM though the VMM.