An industrial internet computing interference avoidance method, storage medium and processor

By optimizing the deployment order and node selection of model replicas in the Industrial Internet, network bottlenecks and hardware interference issues were resolved, improving the efficiency and performance of model inference.

CN117938777BActive Publication Date: 2026-07-14SHENZHEN INST OF ADVANCED TECH CHINESE ACAD OF SCI
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
SHENZHEN INST OF ADVANCED TECH CHINESE ACAD OF SCI
Filing Date
2023-12-18
Publication Date
2026-07-14

AI Technical Summary

Technical Problem

In industrial internet environments with limited bandwidth, deploying multiple replicas of models can lead to network bottlenecks and hardware interference, impacting model inference performance.

Method used

A serverless inference system that supports collaborative optimization of network transmission bottlenecks through channel residual control employs a hybrid model placement strategy and scheduling delay planning strategy to optimize the deployment order of model replicas and node selection, thereby avoiding network bandwidth contention.

Benefits of technology

The execution efficiency of model inference tasks has been optimized, network bandwidth contention and resource interference have been reduced, and overall performance and response speed of inference tasks have been improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117938777B_ABST
    Figure CN117938777B_ABST
Patent Text Reader

Abstract

The application is suitable for the technical field of synchronous motor control, and provides an industrial internet computing interference avoiding method, a storage medium and a processor. The method is used for model placement in optimization of industrial internet edge model reasoning, can identify bottlenecks in network communication in a concurrent multi-model deployment process, and dynamically adjusts the order of to-be-downloaded copies and the priority order of to-be-deployed nodes, so that the copies with large data volumes are preferentially matched with nodes with small residual data volumes, thereby avoiding competition for network bandwidth. In addition, the strategy can also optimize the order of model placement and the transmission sequence, so that the reasoning task can be more efficiently executed.
Need to check novelty before this filing date? Find Prior Art