Scholarlink

<p>Cloud computing provides a robust, scalable, and flexible infrastructure necessary to support the demanding computational needs of various applications. Additionally, it facilitates optimization strategies that enhance performance and resource utilization, thereby driving innovation and efficiency. Optimizing workflow scheduling is crucial in cloud environments to achieve peak performance. Effective scheduling techniques are essential to minimize execution time, enhance speedup and efficiency, and manage parallel overload. This paper proposed a model to optimize workflow scheduling performance in cloud environments using a reinforcement learning technique called Q-learning. The primary goal of the model is to reduce the makespan. Q-learning is combined with the three heuristic scheduling algorithms as As late as possible (ALAP), Critical path on processor (CPOP), and performance-effective task scheduling (PETS) are combined with the Q-learning method, which generates very effective performance in terms of reduction of the makespan. The proposed algorithms are named as the QL-ALAP, QL-CPOP, and QL-PETS. Experimental results are based on four real scientific workflows, including Montage, SIPHT, Cybershake, and AIRSN. The average reduction of the makespan is 13.97% as compared with the state-of-the-art algorithms such as ALAP, CPOP, and PETS. Using four real scientific workflows with different datasets and also giving better results of other QoS parameters. We have also performed a statistical analysis indicating that the proposed algorithm outperforms traditional heuristic algorithms. This paper demonstrates a successful approach to optimising workflow scheduling performance in cloud 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

Enhancing heuristic workflow scheduling algorithms using Q-learning technique in cloud computing

  • 期刊论文

  • 发表日期：

    2025年6月25日

查看全文
  • Kanchan Namdev,
  • Ranjit Rajak

摘要

Cloud computing provides a robust, scalable, and flexible infrastructure necessary to support the demanding computational needs of various applications. Additionally, it facilitates optimization strategies that enhance performance and resource utilization, thereby driving innovation and efficiency. Optimizing workflow scheduling is crucial in cloud environments to achieve peak performance. Effective scheduling techniques are essential to minimize execution time, enhance speedup and efficiency, and manage parallel overload. This paper proposed a model to optimize workflow scheduling performance in cloud environments using a reinforcement learning technique called Q-learning. The primary goal of the model is to reduce the makespan. Q-learning is combined with the three heuristic scheduling algorithms as As late as possible (ALAP), Critical path on processor (CPOP), and performance-effective task scheduling (PETS) are combined with the Q-learning method, which generates very effective performance in terms of reduction of the makespan. The proposed algorithms are named as the QL-ALAP, QL-CPOP, and QL-PETS. Experimental results are based on four real scientific workflows, including Montage, SIPHT, Cybershake, and AIRSN. The average reduction of the makespan is 13.97% as compared with the state-of-the-art algorithms such as ALAP, CPOP, and PETS. Using four real scientific workflows with different datasets and also giving better results of other QoS parameters. We have also performed a statistical analysis indicating that the proposed algorithm outperforms traditional heuristic algorithms. This paper demonstrates a successful approach to optimising workflow scheduling performance in cloud environments.