Fictitious Computing Node for Dynamic Resource Allocation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Large computing clusters face density limitations due to fixed arrangements of CPU, GPU, and storage elements within servers, leading to inefficient resource utilization and increased physical space and cost requirements.

Innovation Solution

The deployment of a method that presents a fictitious computing node with available resources, dynamically composing a target computing node with available resources to execute jobs, and decomposing the node after job completion, allowing for flexible and efficient use of computing resources.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If fixed server arrangements with CPU, GPU, and storage elements are used, then device structure is simplified, but resource utilization efficiency deteriorates

Engineering Contradiction:

Improvedevice structureVSAvoidresource utilization efficiency

Core Design Contradiction:

Device complexityVSProductivity

Solution Approach 1:

The patent segments computing resources by separating compute nodes from storage elements. Compute nodes can be dynamically allocated to different storage elements based on job requirements, rather than being permanently bound to specific hardware configurations. This segmentation enables flexible resource allocation and improves utilization efficiency.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent implements dynamic compute node allocation where compute nodes can be assigned to different storage elements depending on the job being executed. The system transitions from static server arrangements to dynamic resource allocation, allowing compute nodes to serve multiple storage elements sequentially and improving overall resource utilization.

Inventive Principle:

Principle #15Dynamics

2Productivity

If more servers are deployed to accommodate increasing job traffic, then productivity is improved, but physical space requirements worsen

Engineering Contradiction:

Improvejob handling capacityVSAvoidphysical space requirements

Core Design Contradiction:

ProductivityVSArea of stationary object

Solution Approach 1:

The patent makes compute nodes universal by enabling them to serve multiple storage elements across different jobs. A single compute node can be allocated to different storage elements based on job requirements, eliminating the need for dedicated servers for each storage element and reducing physical space requirements while maintaining productivity.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent merges compute resources and storage resources into separate pools that can be dynamically combined based on job requirements. Instead of having fixed server units combining specific CPU, GPU, and storage elements, the system allows independent compute nodes to be temporarily combined with different storage elements, reducing the total number of physical servers needed.

Inventive Principle:

Principle #5Merging (Combining)

3Productivity

If compute nodes are dynamically allocated to storage elements, then resource utilization efficiency is improved, but device complexity worsens

Engineering Contradiction:

Improveresource utilization efficiencyVSAvoidsystem management complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent introduces a workload manager as an intermediary between compute nodes and storage elements. The workload manager handles the complexity of dynamic allocation by automatically assigning compute nodes to storage elements based on job requirements, thereby improving resource utilization without requiring direct complex management between individual compute nodes and storage elements.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent implements feedback mechanisms where the system monitors job requirements and storage element availability, then dynamically adjusts compute node allocations accordingly. This feedback loop enables automated optimization of resource utilization while managing system complexity through intelligent control rather than manual configuration.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS20250068472A1Job target aliasing in disaggregated computing systems

Publication Date: 2025.02.27 LIQID INC

AI summary

Deployment of arrangements of computing components coupled over a communication fabric are presented herein. In one example, a method includes presenting an indication of a fictitious computing node capable of handling execution jobs, the fictitious computing node presented with a first set of computing resources available to service the execution jobs without regard to an availability to service the execution jobs. Responsive to a job directed to the fictitious computing node, the method includes composing a target computing node comprising a second set of computing resources with a present availability to service the job. The method includes initiating execution of the job by the target computing node, and responsive to completion of the job, de-composing the target computing node and presenting the indication of the fictitious computing node.