Scholarlink

<p>In this paper, we address the problem of dynamic scheduling for distributed computing tasks on complex heterogeneous embedded information processing systems with stringent real-time constraints. The heterogeneity of hardware resources and the complexity of task workflows lead to a vast scheduling solution space, posing significant challenges for achieving efficient and timely task deployment. We propose a hierarchical and segmented heuristic scheduling (HSHS) method, which substantially reduces the solution space by first layering and segmenting both computational tasks and hardware clusters and then employing a greedy heuristic to dynamically optimize task allocation. Experimental evaluation, based on a prototype system comprising 139 computing tasks distributed across 4 heterogeneous chassis clusters, demonstrates that our algorithm achieves superior real-time performance with a deployment time of 61.95 ms–significantly faster than baselines such as HEFT-based schedulers–while attaining higher resource utilization (91.50%). Furthermore, scalability tests show that HSHS maintains efficient scheduling performance as the number of tasks increases, and optimality gap analysis confirms that it produces near-optimal solutions with greatly reduced computational overhead. The method also exhibits effective fault tolerance, enabling rapid task remapping in response to runtime node failures.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hierarchical heuristic scheduling for real-time distributed workflows in heterogeneous embedded systems

  • 期刊论文

  • 发表日期：

    2025年12月4日

查看全文
  • He Li,
  • Xiaofeng Li,
  • Long Chen,
  • Xiaoping Li

摘要

In this paper, we address the problem of dynamic scheduling for distributed computing tasks on complex heterogeneous embedded information processing systems with stringent real-time constraints. The heterogeneity of hardware resources and the complexity of task workflows lead to a vast scheduling solution space, posing significant challenges for achieving efficient and timely task deployment. We propose a hierarchical and segmented heuristic scheduling (HSHS) method, which substantially reduces the solution space by first layering and segmenting both computational tasks and hardware clusters and then employing a greedy heuristic to dynamically optimize task allocation. Experimental evaluation, based on a prototype system comprising 139 computing tasks distributed across 4 heterogeneous chassis clusters, demonstrates that our algorithm achieves superior real-time performance with a deployment time of 61.95 ms–significantly faster than baselines such as HEFT-based schedulers–while attaining higher resource utilization (91.50%). Furthermore, scalability tests show that HSHS maintains efficient scheduling performance as the number of tasks increases, and optimality gap analysis confirms that it produces near-optimal solutions with greatly reduced computational overhead. The method also exhibits effective fault tolerance, enabling rapid task remapping in response to runtime node failures.