Scholarlink

<p>In the context of cloud computing, efficient virtual machine (VM) reallocation is a cornerstone of VM consolidation, playing a pivotal role in optimizing energy consumption and maintaining Service Level Agreements (SLAs). This paper presents a novel approach utilizing Reinforcement Learning (RL) to address the VM placement problem, focusing on energy efficiency and SLA compliance. Effective VM reallocation is critical for consolidating workloads, which in turn reduces the number of active physical servers and consequently the overall energy consumption. Our proposed RL approach dynamically adjusts VM placements based on real workload demands and energy usage patterns, ensuring an optimal balance between energy efficiency and SLA violation. The approach uses a reward function designed to minimize energy use while ensuring VMs are placed in a way that maintains SLA commitments. Experimental results demonstrate the effectiveness of our method, achieving up to 25% reduction in energy consumption and more than 90% improvement in SLA fulfillment compared to benchmark algorithm. This research advances sustainable and cost-effective cloud infrastructure management by highlighting the critical role of VM placement in achieving efficient consolidation.</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

Reinforcement learning for efficient VM reallocation: A novel approach for sustainable cloud computing

  • 期刊论文

  • 发表日期：

    2025年9月29日

查看全文
  • Somayeh Rahmani,
  • Vahid Khajehvand,
  • Mohsen Torabian

摘要

In the context of cloud computing, efficient virtual machine (VM) reallocation is a cornerstone of VM consolidation, playing a pivotal role in optimizing energy consumption and maintaining Service Level Agreements (SLAs). This paper presents a novel approach utilizing Reinforcement Learning (RL) to address the VM placement problem, focusing on energy efficiency and SLA compliance. Effective VM reallocation is critical for consolidating workloads, which in turn reduces the number of active physical servers and consequently the overall energy consumption. Our proposed RL approach dynamically adjusts VM placements based on real workload demands and energy usage patterns, ensuring an optimal balance between energy efficiency and SLA violation. The approach uses a reward function designed to minimize energy use while ensuring VMs are placed in a way that maintains SLA commitments. Experimental results demonstrate the effectiveness of our method, achieving up to 25% reduction in energy consumption and more than 90% improvement in SLA fulfillment compared to benchmark algorithm. This research advances sustainable and cost-effective cloud infrastructure management by highlighting the critical role of VM placement in achieving efficient consolidation.