To address the issue of inefficient processing of computational task requests in heterogeneous Computing Power Network, a novel optimal pathfinding algorithm based on adap-tive task division is proposed. By modeling computing network resources and computational tasks as a Directed Acyclic Graph (DAG), a latency prediction model is established. This model, with latency as weights, facilitates adaptive task division within the DAG, and employs a reinforcement learning algorithm for optimal pathfinding. This approach matches computing nodes on the optimal path, enabling edge-to-edge computing and collabo-rative transmission, effectively reducing task processing latency. Simulation experiments demonstrate a significant reduction in the average system latency, with a maximum acceleration ratio of 5.8, showcasing improved performance in complex environments.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Task Scheduling Algorithm for Heterogeneous Computing Power Network


    Contributors:
    He, Guodong (author) / Li, Xiaohui (author) / Lv, Siting (author) / Zhou, Yuanyuan (author) / Ni, ZhiGang (author) / Chen, Xingbo (author)


    Publication date :

    2024-06-24


    Size :

    1702400 byte





    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English




    Solution Space Reduction in Task Scheduling for Heterogeneous Computing Systems

    Lam, Yuet M. | British Library Online Contents | 2013


    Research of task scheduling in network parallel computing environment

    Hu, K. / Zhang, Y. / Hu, J.-p. | British Library Online Contents | 2001


    A Task Scheduling Algorithm for Large Graph Processing in Cloud Computing

    Li, J. / Huang, Q. / Liu, Y. et al. | British Library Online Contents | 2012


    Hybrid genetic algorithm for independent tasks scheduling in heterogeneous computing systems

    Yiwen, Z. / Jiangang, Y. | British Library Online Contents | 2004