Scholarlink

<p>This paper introduces an innovative and integrated technique, which can be used to optimize the successful processing management of workloads in clouds, based on Genetic Algorithms (GA) and Q-Learning, both powerful optimization tools. The global search optimization capability of GA is used to thoroughly search the solution space to find the best configuration of resources which provide a balance between load distribution and system efficiency. In parallel, Q-Learning—a reinforcement learning method—is employed to optimize these configurations by learning from feedback, given by the cloud system’s performance, such that the model can adapt itself to the variable workload with time. By combining the GA and Q-Learning, we achieve a system that has a stronger resilient capability against possible failures, changes its strategy according to the dynamic demands and also learns over historical decisions to perform more resource optimizations. The method is intended to avoid inefficiency, lower operational costs and guarantee high availability and performance in the cloud. Experimental results show that the proposed hybrid approach achieves the highest resource utilization of 90%, 150 ms (−) response time with 85% system efficiency, thus improving the operational efficiency of cloud systems.</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

Resilient and adaptive workload optimization for cloud infrastructures: a GA-Q-Learning hybrid approach

  • 期刊论文

  • 发表日期：

    2025年10月25日

查看全文
  • Sandeep,
  • Ankit,
  • Bhanumati Panda,
  • V. S. N. Murthy,
  • Kunchanapalli Rama Krishna,
  • Siddharth Arora

摘要

This paper introduces an innovative and integrated technique, which can be used to optimize the successful processing management of workloads in clouds, based on Genetic Algorithms (GA) and Q-Learning, both powerful optimization tools. The global search optimization capability of GA is used to thoroughly search the solution space to find the best configuration of resources which provide a balance between load distribution and system efficiency. In parallel, Q-Learning—a reinforcement learning method—is employed to optimize these configurations by learning from feedback, given by the cloud system’s performance, such that the model can adapt itself to the variable workload with time. By combining the GA and Q-Learning, we achieve a system that has a stronger resilient capability against possible failures, changes its strategy according to the dynamic demands and also learns over historical decisions to perform more resource optimizations. The method is intended to avoid inefficiency, lower operational costs and guarantee high availability and performance in the cloud. Experimental results show that the proposed hybrid approach achieves the highest resource utilization of 90%, 150 ms (−) response time with 85% system efficiency, thus improving the operational efficiency of cloud systems.