Scholarlink

<p>The proliferation of multi-cloud architectures has intensified challenges in Service-Level Agreement (SLA) management due to resource heterogeneity, policy conflicts, and dynamic workloads. Existing solutions relying on static templates or isolated machine learning models lack adaptability to real-time coordination and fail to resolve cross-provider conflicts. This paper proposes MSACM-RL, a reinforcement learning (RL)-driven framework integrating game theory and hierarchical admission control to optimize SLA compliance. Key innovations include a Q-learning-based dynamic policy adaptation mechanism for multi-cloud resource coordination, a Bayesian Nash bargaining model resolving provider conflicts under incomplete information, and a three-tier admission control pipeline leveraging predictive resource modeling. Experimental results show that the proposed framework achieves an 18% improvement in SLA compliance rates and a 33% reduction in policy conflicts compared to state-of-the-art frameworks. Also, it can significantly reduce negotiation rounds, computational overhead and latency violations under high-intensity workloads, demonstrating superior scalability and adaptability in dynamic 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

Hierarchical admission control in multi-cloud systems: integrating reinforcement learning, nash bargaining, and predictive state modeling

  • 期刊论文

  • 发表日期：

    2025年11月11日

查看全文
  • Xiao Peng,
  • Zeng Saifeng

摘要

The proliferation of multi-cloud architectures has intensified challenges in Service-Level Agreement (SLA) management due to resource heterogeneity, policy conflicts, and dynamic workloads. Existing solutions relying on static templates or isolated machine learning models lack adaptability to real-time coordination and fail to resolve cross-provider conflicts. This paper proposes MSACM-RL, a reinforcement learning (RL)-driven framework integrating game theory and hierarchical admission control to optimize SLA compliance. Key innovations include a Q-learning-based dynamic policy adaptation mechanism for multi-cloud resource coordination, a Bayesian Nash bargaining model resolving provider conflicts under incomplete information, and a three-tier admission control pipeline leveraging predictive resource modeling. Experimental results show that the proposed framework achieves an 18% improvement in SLA compliance rates and a 33% reduction in policy conflicts compared to state-of-the-art frameworks. Also, it can significantly reduce negotiation rounds, computational overhead and latency violations under high-intensity workloads, demonstrating superior scalability and adaptability in dynamic environments.