Neural Network Task Allocation for Distributed Computing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for allocating computing tasks in distributed systems face challenges in accurately predicting resource requirements, leading to inefficiencies such as increased latency and energy consumption due to overload situations and the need for reactive re-orchestration.

Innovation Solution

A computer-implemented method using a neural network to predict resource requirements of computing tasks, allowing for proactive allocation to optimize energy consumption and latency by selecting the most suitable computing devices based on predicted resource needs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If reactive re-orchestration is used to handle overload situations, then system reliability is improved, but latency and energy consumption increase nonlinearly

Engineering Contradiction:
Improvesystem reliabilityVSAvoidlatency
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent applies preliminary action by using a neural network to predict resource requirements of computing tasks before they are allocated. This allows the system to proactively assign tasks to computing devices that will have sufficient resources, preventing overload situations before they occur and avoiding the need for reactive re-orchestration, thereby reducing latency while maintaining reliability

Inventive Principle:
Principle #10Preliminary action

2Reliability

If pre-defined static resource requirements with large safety margin are used, then reliability is improved, but resource waste increases

Engineering Contradiction:
Improvesystem reliabilityVSAvoidresource waste
Core Design Contradiction:
ReliabilityVSLoss of substance

Solution Approach 1:

The patent applies parameter changes by transitioning from static, pre-defined resource requirements to dynamic, neural network-predicted resource requirements. The neural network learns optimal resource allocation parameters from historical data, enabling the system to accurately predict actual resource needs without requiring large safety margins, thus reducing resource waste while maintaining reliability

Inventive Principle:
Principle #35Parameter changes

3Productivity

If pre-defined static resource requirements are used for task allocation, then allocation speed is improved, but accuracy of resource requirements decreases

Engineering Contradiction:
Improveallocation speedVSAvoidaccuracy of resource requirements
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies mechanics substitution by replacing the mechanical approach of manually defining static resource requirements with an intelligent neural network system. The neural network automatically learns and predicts accurate resource requirements from historical data, maintaining fast allocation speeds while significantly improving the accuracy of resource requirement predictions compared to static pre-defined values

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP4571508A1Method and system for deploying a computing task to a distributed computing system
Publication Date: 2025.06.18 SIEMENS AG
  • EP4571508A1 patent drawingFigure 1
  • EP4571508A1 patent drawingFigure 2~3
  • EP4571508A1 patent drawingFigure 4

AI summary

A computer-implemented method for deploying a computing task (1) to a distributed computing system (200) comprising a plurality of computing devices (201-207) comprises: acquiring (S10) a task description (3) of the computing task (1); using a neural network device (111) to predict predicted resource requirements (4) of the computing task (1); allocating the computing task (1) to a selected one of the plurality of computing devices (201-207) based on the predicted resource requirements (4); and deploying the computing task (1) to the selected computing device (201). The method allows solving a task allocation problem for the distributed computing system such that power consumption, latency, and the like are further optimized.