Heterogeneous Accelerator Allocation for Mixed Deadlines and Traffic

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing technologies fail to efficiently manage accelerator resources in heterogeneous configurations, leading to insufficient processing capacity and responsiveness when dealing with varying data amounts and deadlines in computer systems.

Innovation Solution

An accelerator state control device that includes a recording unit for performance data, a prediction unit for traffic and deadline estimation, and a determination unit for allocating accelerators based on predicted performance and deadlines, ensuring efficient resource utilization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If accelerators with different processing capabilities are used in a heterogeneous configuration, then processing efficiency for specific workloads is improved, but resource allocation complexity increases

Engineering Contradiction:

Improveprocessing efficiencyVSAvoidresource allocation complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent implements dynamic resource allocation where the accelerator selection and configuration change based on real-time workload characteristics. The system continuously monitors workload parameters and adjusts accelerator assignment accordingly, transitioning from static to dynamic allocation to handle heterogeneous accelerator environments effectively

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system employs feedback mechanisms by monitoring workload characteristics, processing performance, and resource utilization metrics. This feedback is used to continuously optimize accelerator allocation decisions, allowing the system to learn from past performance and improve resource distribution in heterogeneous configurations

Inventive Principle:

Principle #23Feedback

2Device complexity

If static resource allocation is used, then system simplicity is maintained, but responsiveness to varying data amounts and deadlines deteriorates

Engineering Contradiction:

Improvesystem simplicityVSAvoidresponsiveness

Core Design Contradiction:

Device complexityVSProductivity

Solution Approach 1:

The patent transforms static resource allocation into a dynamic system that automatically adjusts accelerator assignment based on real-time workload characteristics. The allocation changes dynamically in response to varying data amounts and processing deadlines, maintaining system responsiveness without requiring complex manual configuration

Inventive Principle:

Principle #15Dynamics

3Productivity

If accelerator resources are increased to handle peak traffic, then processing capacity is improved, but resource utilization efficiency during low traffic periods deteriorates

Engineering Contradiction:

Improveprocessing capacityVSAvoidresource utilization efficiency

Core Design Contradiction:

ProductivityVSLoss of energy

Solution Approach 1:

The system implements dynamic scaling of accelerator resources by adjusting the number and type of active accelerators based on real-time traffic volume and workload characteristics. During peak traffic, additional accelerators are activated to handle the load, while during low traffic periods, accelerators are deactivated or reassigned, optimizing both capacity and efficiency

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent changes operational parameters such as accelerator clock speeds, power states, and allocation configurations based on workload demands. By dynamically adjusting these parameters, the system maintains high processing capacity when needed while reducing resource consumption during low-utilization periods

Inventive Principle:

Principle #35Parameter changes

4Measurement precision

If prediction of traffic amount and processing deadline is implemented, then resource allocation accuracy is improved, but computational overhead increases

Engineering Contradiction:

Improveresource allocation accuracyVSAvoidcomputational overhead

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs preliminary prediction of traffic patterns and processing deadlines using historical data and current workload characteristics. By predicting future resource needs in advance, the system can proactively allocate accelerators before peak demands occur, improving allocation accuracy without requiring complex real-time calculations during critical processing periods

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20260046332A1Accelerator state control device, accelerator state control system, accelerator state control method and program

Publication Date: 2026.02.12 NT T INC

AI summary

An accelerator state control device includes a plurality of accelerators having different processing performance, and controls a state of the accelerators when arithmetic processing is performed by offloading specific processing of an application to the accelerators. The accelerator state control device includes: when data in which different processing deadlines are mixed is input, an arithmetic device performance collection/recording unit that collects and records performance information of the accelerators; a traffic amount/processing deadline prediction unit that predicts a traffic amount and a processing deadline; and an arithmetic device allocation determination unit that obtains a data amount corresponding to the processing deadline, and determines an accelerator that satisfies the performance on the basis of the data amount.