AI Cloud Task Scheduling via Regression Path Selection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Traditional task scheduling algorithms in AI computation have limitations due to fixed parameter selection and reliance on prior knowledge, leading to reduced flexibility and accuracy, especially when environmental changes increase task complexity, resulting in increased time complexity and computational inaccuracy.

Innovation Solution

A task scheduling method that decomposes computing tasks into subtasks, extracts feature information from multiple dimensions to calculate the shortest required time using a regression model, and selects the candidate path with the shortest total time, incorporating network topology and hardware features to dynamically adjust scheduling.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional task scheduling algorithms use fixed parameter selection and rely on prior knowledge, then the scheduling process is simple and fast, but the computational accuracy and adaptability to environmental changes deteriorate

Engineering Contradiction:

Improvecomputational accuracyVSAvoidscheduling algorithm complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent changes the parameters from fixed traditional values to dynamic feature vectors extracted from multiple dimensions (network topology, hardware resources, task characteristics). The regression model calculates required time based on these dynamic parameters, enabling accurate adaptation to environmental changes while maintaining computational efficiency

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent replaces traditional mechanical scheduling algorithms with a regression model-based approach. Instead of using fixed formulas and prior knowledge, the system uses machine learning to predict required time, substituting complex rule-based logic with a data-driven model that achieves higher accuracy

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If traditional task scheduling algorithms increase task complexity handling, then more tasks can be scheduled, but the time complexity increases accordingly

Engineering Contradiction:

Improvetask throughputVSAvoidscheduling time complexity

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The patent performs preliminary extraction of feature vectors from multiple dimensions before scheduling. By pre-processing and organizing this information, the system prepares all necessary data in advance, allowing the regression model to quickly calculate required time without increasing time complexity when handling more tasks

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent creates a simplified representation of the complex scheduling problem through feature vectors and regression models. This copying approach allows the system to handle complex tasks by reducing them to manageable feature representations, maintaining fast scheduling even as task complexity increases

Inventive Principle:

Principle #26Copying

3Adaptability or versatility

If the system extracts feature information from multiple dimensions, then the adaptability to changing environments improves, but the computational overhead increases

Engineering Contradiction:

Improveenvironmental adaptabilityVSAvoidcomputational energy consumption

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent segments the feature extraction process into distinct dimensional categories (network topology information, hardware resource features, task characteristics). By organizing features into segments, the system efficiently processes and stores information, reducing computational overhead while maintaining comprehensive environmental adaptability

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent creates a universal feature extraction framework that handles multiple dimensions of information through a single regression model. This multi-functional approach allows the system to adapt to various environmental changes without requiring separate processing mechanisms for each feature type, reducing overall computational energy consumption

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS20240354155A1Task scheduling method and ai cloud computing system

Publication Date: 2024.10.24 MONTAGE TECHNOLOGY CO LTD

AI summary

A task scheduling method and an AI cloud computing system are provided. The task scheduling method comprises: decomposing, via a processor, a computing task into multiple sequent subtasks, and obtaining multiple candidate paths that are capable of processing the sequent subtasks based on a network topology information table, wherein the candidate paths include one or more computing nodes selected from multiple computing nodes, and the computing nodes are configured to process subtasks that match their supported operation types; obtaining feature vectors including feature information of each candidate path and feature information of each subtask, and calculating a total required time for each candidate path to complete the multiple sequent subtasks based on the feature vectors; and selecting the candidate path with the shortest total time to process the multiple sequent subtasks.