A method and system for scheduling computing power slicing in heterogeneous computing power resource pools

By constructing task requirements and resource environment models and combining them with multi-objective scheduling strategies to generate optimal slice combinations, the problem of low computing power utilization in heterogeneous resource pools is solved, achieving efficient computing power scheduling and resource management, and improving task execution efficiency and resource utilization.

CN122137887APending Publication Date: 2026-06-02HUASHU CLOUD TECH CO LTD
View PDF 5 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
HUASHU CLOUD TECH CO LTD
Filing Date
2026-05-08
Publication Date
2026-06-02

AI Technical Summary

Technical Problem

Existing computing power scheduling systems cannot effectively integrate heterogeneous resource pools, resulting in low computing power utilization, long task waiting times, and high operation and maintenance costs. In particular, they are inefficient in resource allocation to meet the performance differences of different nodes in large-scale training tasks or multi-node inference tasks.

Method used

This paper presents a computing power slicing scheduling method for heterogeneous computing power resource pools. By receiving task requests, it constructs a task requirement model and a resource environment model, generates a task slice set by combining a multi-objective scheduling strategy, selects the optimal slice combination for scheduling, monitors the task execution status in real time to optimize the scheduling strategy, and supports breakpoint resume computing and dynamic resource adjustment.

Benefits of technology

It achieves efficient management and intelligent scheduling of heterogeneous computing resources, improves resource utilization and task completion efficiency, reduces computing power fragmentation and energy consumption costs, and supports fine-grained management and elastic supply for multiple scenarios such as large model training, AI inference and HPC computing.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122137887A_ABST
    Figure CN122137887A_ABST
Patent Text Reader

Abstract

This application relates to the technical field of artificial intelligence computing power management, and in particular to a computing power slice scheduling method and system for heterogeneous computing power resource pools. The method includes: receiving and parsing task requests submitted by users to construct a task requirement model; obtaining feature information of each computing power node in the heterogeneous computing power resource pool to construct a resource environment model; generating a task slice set based on the task requirement model and the resource environment model, combined with a multi-objective scheduling strategy, and selecting the optimal slice combination to configure a scheduling plan; allocating each computing power task slice in the optimal slice set to computing power nodes for execution based on the scheduling plan; real-time monitoring of task execution status and resource usage information to generate execution results, and collecting execution results to provide feedback and optimize the multi-objective scheduling strategy. This application achieves efficient management and intelligent scheduling of heterogeneous computing power resources.
Need to check novelty before this filing date? Find Prior Art