Scholarlink

This article presents an integrated Gaussian Mixture Model and Multi-Agent Reinforcement Learning (GMM-MARL) framework for optimizing controller placement in banking Software-Defined Wide Area Networks (SD-WAN). Traditional static placement methods fail to address the dynamic transaction patterns and stringent compliance requirements of financial institutions. Our approach combines GMM clustering for initial controller positioning based on geographical and banking hierarchy constraints with MARL agents for continuous adaptation to varying network conditions. Experimental evaluation on real-world banking topologies demonstrates 45% reduction in transaction latency, improved load balancing, and maintained regulatory compliance compared to existing methods, while reducing operational costs by 34% annually.

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

Multi-objective Controller Placement in Banking SD-WAN: A GMM-MARL Approach for High-Frequency Transaction Networks

  • 章节

  • 发表日期：

    2026年

查看全文
  • Abdulrahman M. Abdulghani,
  • Azizol Abdullah,
  • A. R. Rahiman,
  • Nor Asilah Wati Abdul Hamid,
  • Bilal Omar Akram

摘要

This article presents an integrated Gaussian Mixture Model and Multi-Agent Reinforcement Learning (GMM-MARL) framework for optimizing controller placement in banking Software-Defined Wide Area Networks (SD-WAN). Traditional static placement methods fail to address the dynamic transaction patterns and stringent compliance requirements of financial institutions. Our approach combines GMM clustering for initial controller positioning based on geographical and banking hierarchy constraints with MARL agents for continuous adaptation to varying network conditions. Experimental evaluation on real-world banking topologies demonstrates 45% reduction in transaction latency, improved load balancing, and maintained regulatory compliance compared to existing methods, while reducing operational costs by 34% annually.