Scholarlink

<p>Cloud computing environments encounter significant challenges in resource management through queueing and scheduling systems, as traditional methods struggle with dynamic workload optimization. This research introduces an innovative AI-enhanced framework combining deep workload prediction and reinforcement learning for dynamic scheduling. The framework features a dual-layer neural network architecture with a hybrid decision engine that merges conventional queueing metrics with learned policies. Experimental results in a simulated cloud environment showcase remarkable improvements: a 30% decrease in average waiting time, 25% optimization in queue length, and 91% peak resource utilization compared to traditional approaches. The AI-enhanced model demonstrates 20–35% higher throughput rates across various workload intensities, with the reinforcement learning scheduler maintaining steady performance under high loads. Statistical confidence levels exceed 95%, validating the approach's effectiveness. The research provides practical solutions for cloud service providers, enabling implementation of efficient, adaptive resource management systems that reduce operational costs while enhancing service quality. The modular architecture ensures scalability and seamless integration with existing cloud infrastructure, making it particularly valuable for large-scale production environments.</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

AI-enhanced modelling of queueing and scheduling systems in cloud computing

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年3月29日

查看全文
  • Himani Chaudhary,
  • Geetanjali Sharma,
  • D. K. Nishad,
  • Saifullah Khalid

摘要

Cloud computing environments encounter significant challenges in resource management through queueing and scheduling systems, as traditional methods struggle with dynamic workload optimization. This research introduces an innovative AI-enhanced framework combining deep workload prediction and reinforcement learning for dynamic scheduling. The framework features a dual-layer neural network architecture with a hybrid decision engine that merges conventional queueing metrics with learned policies. Experimental results in a simulated cloud environment showcase remarkable improvements: a 30% decrease in average waiting time, 25% optimization in queue length, and 91% peak resource utilization compared to traditional approaches. The AI-enhanced model demonstrates 20–35% higher throughput rates across various workload intensities, with the reinforcement learning scheduler maintaining steady performance under high loads. Statistical confidence levels exceed 95%, validating the approach's effectiveness. The research provides practical solutions for cloud service providers, enabling implementation of efficient, adaptive resource management systems that reduce operational costs while enhancing service quality. The modular architecture ensures scalability and seamless integration with existing cloud infrastructure, making it particularly valuable for large-scale production environments.