Scholarlink

A university equipped with an intelligent network infrastructure capable of ‘learning’ and allocating the necessary network bandwidth required for its different counterparts autonomously would be an effective resource allocation solution for the institution. From a network perspective, a university is a complex organization where each department may have different bandwidth demands that keep evolving temporally. Since the university environment is highly dynamic, static bandwidth allocation schemes are inefficient. Traditionally, implementing dynamic bandwidth allocation strategies was a hassle, as the notion of ‘programmable’ network infrastructure was a far-fetched idea a few decades back. The advent of software-defined networks (SDN) has completely revolutionized the networking domain by bringing in the most needed flexibility and the capability to impart intelligence to network maintenance. In this paper, we propose an reinforcement learning (RL)-based intelligent resource allocation for network bandwidth usage in universities, which continuously learns and adapts the resource allocation (RA) strategy as per the usage pattern. The allocation problem is modeled with the multi-armed bandit (MAB) framework, and the solution is tested with different plausible usage pattern scenarios.

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

Integrating Reinforcement Learning into Software-Defined Networks for Adaptive Bandwidth Allocation in Universities: A Bandit-Based Approach

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • D. Sendil Vadivu,
  • A. S. Syed Shahul Hameed,
  • Narendran Rajagopalan

摘要

A university equipped with an intelligent network infrastructure capable of ‘learning’ and allocating the necessary network bandwidth required for its different counterparts autonomously would be an effective resource allocation solution for the institution. From a network perspective, a university is a complex organization where each department may have different bandwidth demands that keep evolving temporally. Since the university environment is highly dynamic, static bandwidth allocation schemes are inefficient. Traditionally, implementing dynamic bandwidth allocation strategies was a hassle, as the notion of ‘programmable’ network infrastructure was a far-fetched idea a few decades back. The advent of software-defined networks (SDN) has completely revolutionized the networking domain by bringing in the most needed flexibility and the capability to impart intelligence to network maintenance. In this paper, we propose an reinforcement learning (RL)-based intelligent resource allocation for network bandwidth usage in universities, which continuously learns and adapts the resource allocation (RA) strategy as per the usage pattern. The allocation problem is modeled with the multi-armed bandit (MAB) framework, and the solution is tested with different plausible usage pattern scenarios.