Scholarlink

<p>Fog computing can mitigate IoT latency and backbone bandwidth, although the scheduling of diverse workloads under variable loads continues to result in energy inefficiency and SLA breaches. We present a multi-stage AI-driven scheduling pipeline that concurrently manages task classification, urgency-based prioritization, adaptive fog-node scheduling, and load-aware virtual machine selection. The pipeline categorizes tasks as either fog or cloud and prioritizes fog tasks based on an emergency score. Subsequently, it organizes fog tasks with a bipartite-graph-assisted Deep Q-Network (SUPER DQNET) and chooses cloud virtual machines utilizing a binary owl-inspired optimizer (BOON). The proposed framework, implemented in iFogSim2 and assessed against EEIOMT, EaDO, and baseline schedulers (Random, RR, FCFS), demonstrates up to 35% reduction in energy consumption, approximately 40% increase in throughput, and over 40% decrease in SLA violations, facilitating adaptive scheduling for next-generation IoT applications.</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

A novel artificial intelligence based dynamic task scheduling and load awareness

  • 期刊论文

  • 发表日期：

    2026年2月16日

查看全文
  • Hamed Altalhoni,
  • Noraida Haji Ali,
  • Farizah Yunus,
  • Saleh Atiewi,
  • Amal Alshardan,
  • Muder Almiani,
  • Shadi AlZu’bi

摘要

Fog computing can mitigate IoT latency and backbone bandwidth, although the scheduling of diverse workloads under variable loads continues to result in energy inefficiency and SLA breaches. We present a multi-stage AI-driven scheduling pipeline that concurrently manages task classification, urgency-based prioritization, adaptive fog-node scheduling, and load-aware virtual machine selection. The pipeline categorizes tasks as either fog or cloud and prioritizes fog tasks based on an emergency score. Subsequently, it organizes fog tasks with a bipartite-graph-assisted Deep Q-Network (SUPER DQNET) and chooses cloud virtual machines utilizing a binary owl-inspired optimizer (BOON). The proposed framework, implemented in iFogSim2 and assessed against EEIOMT, EaDO, and baseline schedulers (Random, RR, FCFS), demonstrates up to 35% reduction in energy consumption, approximately 40% increase in throughput, and over 40% decrease in SLA violations, facilitating adaptive scheduling for next-generation IoT applications.