Scholarlink

<p>Cloud providers are experiencing significant challenges, including energy consumption and QoS, because of the general public’s interest in using cloud services and resources, in addition to the obvious expansion of cloud infrastructure. To address these challenges, several job schedulers have been introduced in recent years; nevertheless, they often lack adaptability and are based on non-RL methods, and their focus has been on batch jobs instead of real-time jobs. This paper proposed a method based on a self-adaptive approach for scheduling real-time jobs in a public cloud environment with an infrastructure as a service (IaaS) model. The proposed method is based on the mape-k feedback loop, which detects changes by continuously monitoring the cloud environment and then adaptively selects the appropriate model for job scheduling. In the proposed method, the plan component is in charge of job scheduling. This component is based on Deep-RL, which can effectively schedule incoming jobs without requiring prior knowledge of the environment or knowing the jobs’ arrival pattern. The proposed method was evaluated in the cloudSimPlus simulator during several experiments. Compared to the state-of-the-art method and basic scheduling policies, the proposed method significantly improves the quality of service (in terms of response time and success rate) and also maintains and reduces energy consumption.</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

Energy-aware self-adaptive real-time job scheduling in a cloud environment using deep reinforcement learning focused on quality of service

  • 期刊论文

  • 发表日期：

    2025年6月16日

查看全文
  • Hossein Zehtab Azari,
  • Saeed Jalili

摘要

Cloud providers are experiencing significant challenges, including energy consumption and QoS, because of the general public’s interest in using cloud services and resources, in addition to the obvious expansion of cloud infrastructure. To address these challenges, several job schedulers have been introduced in recent years; nevertheless, they often lack adaptability and are based on non-RL methods, and their focus has been on batch jobs instead of real-time jobs. This paper proposed a method based on a self-adaptive approach for scheduling real-time jobs in a public cloud environment with an infrastructure as a service (IaaS) model. The proposed method is based on the mape-k feedback loop, which detects changes by continuously monitoring the cloud environment and then adaptively selects the appropriate model for job scheduling. In the proposed method, the plan component is in charge of job scheduling. This component is based on Deep-RL, which can effectively schedule incoming jobs without requiring prior knowledge of the environment or knowing the jobs’ arrival pattern. The proposed method was evaluated in the cloudSimPlus simulator during several experiments. Compared to the state-of-the-art method and basic scheduling policies, the proposed method significantly improves the quality of service (in terms of response time and success rate) and also maintains and reduces energy consumption.