Scholarlink

Cloud computing is an emerging technology that has a key role in development and deployment of growing number of distributed applications. It offers several benefits, with the most notable being reduced costs, efficient resource re-provisioning, and remote accessibility. Resource allocation deals with assigning appropriate resources to tasks based on consumer's needs, ensuring efficient completion of these tasks. In this paper, Decentralized Multi-Agent Reinforcement Learning (DMARL)-based model which integrates RL system with agents for managing allocation of device and network resources is proposed. This decentralized approach offers improved scalability and flexibility, allowing it to better adapt to changing network sizes and configurations, making it more efficient than centralized methods. The proposed DMARL model offers better results based on energy, fault tolerance, execution time and load balancing when compared to PCRA, DMRO and IMARM models.

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

Decentralized Resource Allocation Model Using Multi-agent Reinforcement Learning for Cloud Environment

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • S. Pandi Prabha,
  • A. Rengarajan

摘要

Cloud computing is an emerging technology that has a key role in development and deployment of growing number of distributed applications. It offers several benefits, with the most notable being reduced costs, efficient resource re-provisioning, and remote accessibility. Resource allocation deals with assigning appropriate resources to tasks based on consumer's needs, ensuring efficient completion of these tasks. In this paper, Decentralized Multi-Agent Reinforcement Learning (DMARL)-based model which integrates RL system with agents for managing allocation of device and network resources is proposed. This decentralized approach offers improved scalability and flexibility, allowing it to better adapt to changing network sizes and configurations, making it more efficient than centralized methods. The proposed DMARL model offers better results based on energy, fault tolerance, execution time and load balancing when compared to PCRA, DMRO and IMARM models.