Design And Implementation Of Cloud-Edge Collaborative Computing Framework Based On Deep Models

With the rapid development of cloud computing and edge computing technology,the cloud-edge collaborative computing mode has become a new trend in artificial intelligence applications.Cloud-edge collaborative computing can balance the computational load and data transmission by distributing computational tasks to cloud and edge devices,thereby improving the efficiency and performance of artificial intelligence applications.In addition,cloud-edge collaboration can put more computing processing and data storage on edge devices,which can better cope with privacy-sensitive artificial intelligence computing tasks.Based on research on cloud-edge collaborative technology and deep learning technology,this paper designs and implements a cloud-edge collaborative computing framework based on deep models.This paper mainly focuses on the model inference process and achieves cloud-edge collaborative inference acceleration through model segmentation and pipeline scheduling.Firstly,this paper studies the general model segmentation method,discusses the specific split schemes of chain backbone structure and graph backbone structure,and verifies the effect of model segmentation through theoretical analysis and practical experiments.Afterwards,based on the results of model segmentation,this paper abstracts the cloud-edge collaborative inference process as a pipeline scheduling problem at runtime to further optimize the cloud-edge collaborative inference process and reduce the overall computing time.Finally,this paper designs and implements the computing framework,adopts a distributed architecture design to ensure the reliability and availability of the service,uses modern software engineering modular design thinking to decouple the various modules of the system to improve system readability and maintainability,and provides it to system administrators through a graphical interface,reducing operational difficulty and improving user experience.This paper first briefly introduces the research background and the domestic and foreign research status of the relevant fields,then analyzes and introduces the system requirements in detail based on specific application scenarios.Afterwards,this paper studies and discusses the key issues and solutions involved in the system in detail,and finally provides a detailed description of the overall design of the system,giving specific design schemes and test plans and test results of the key modules of the system to demonstrate the overall and usability of the system.