Scholarlink

Cloud environments enable users to compute tasks on Virtual Machines (VMs) while ensuring that they schedule their resources with proper load management. The main motivation of this paper is to evaluate the need of Reinforcement Learning (RL) concerning Load Balancing (LB) in the Cloud by implementing heuristic resource scheduling algorithms First Come, First Serve (FCFS), Shortest Job First (SJF), and Highest Response Ratio Next (HRRN) and computing tasks in ten scenarios in a cloud simulated environment. The experimental results of these heuristic algorithms are compared with each other concerning the load balanced by the VMs in each scenario, along with their deviation percentage against the ideal expected load to balance. Additionally, the performance of these algorithms is tested mathematically by performing an extensive empirical analysis using the R2 analysis model. The experimental results convey that HRRN provides better LB results than FCFS and SJF by computing an aggregate task load of 732.01 tasks across all the ten scenarios compared to FCFS and SJF, which computes aggregately 1874.46 tasks and 1886.91 tasks, respectively. Also, the HRRN algorithm provides better results than FCFS and SJF by providing 49.87% aggregate average deviation from the ideal expected LB value for computing tasks compared to FCFS and SJF, which are 72.66% and 73.50%, respectively. Since no algorithm stands out in all scenarios and every algorithm has certain pitfalls, the RL method has been proposed to add an intelligent LB mechanism to improve the cloud performance.

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

Evaluating the Need of Reinforcement Learning by Implementing Heuristic Algorithms with Its Load Balancing and Performance Testing in Cloud

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Prathamesh Vijay Lahande,
  • Parag Ravikant Kaveri,
  • Vinay Chavan,
  • Kishor Dhole,
  • Prashant Awasthi

摘要

Cloud environments enable users to compute tasks on Virtual Machines (VMs) while ensuring that they schedule their resources with proper load management. The main motivation of this paper is to evaluate the need of Reinforcement Learning (RL) concerning Load Balancing (LB) in the Cloud by implementing heuristic resource scheduling algorithms First Come, First Serve (FCFS), Shortest Job First (SJF), and Highest Response Ratio Next (HRRN) and computing tasks in ten scenarios in a cloud simulated environment. The experimental results of these heuristic algorithms are compared with each other concerning the load balanced by the VMs in each scenario, along with their deviation percentage against the ideal expected load to balance. Additionally, the performance of these algorithms is tested mathematically by performing an extensive empirical analysis using the R2 analysis model. The experimental results convey that HRRN provides better LB results than FCFS and SJF by computing an aggregate task load of 732.01 tasks across all the ten scenarios compared to FCFS and SJF, which computes aggregately 1874.46 tasks and 1886.91 tasks, respectively. Also, the HRRN algorithm provides better results than FCFS and SJF by providing 49.87% aggregate average deviation from the ideal expected LB value for computing tasks compared to FCFS and SJF, which are 72.66% and 73.50%, respectively. Since no algorithm stands out in all scenarios and every algorithm has certain pitfalls, the RL method has been proposed to add an intelligent LB mechanism to improve the cloud performance.