Reinforcement Learning Based Metaheuristic Algorithm For Optimized Load Balancing In Cloud Environment

Abstract

In cloud computing, workload balancing remains a difficult issue, as overburdened or underburdened servers are capable of causing issues with computation speed or even cause an entire network to collapse, subsequently this is a scenario that is never supposed to occur when using the cloud. Therefore, in order to avoid these issues, the system should be load balanced—that is, allocate job across each of the accessible computing assets by taking into account an acceptable schedule of utilization. The Virtual Machines are supposed to be used effectively, in response to the distributed load mechanism. In this work, a multifaceted load balancing optimization based on the Enhanced Firefly algorithm (EFA) with a Partial Order Markov Decision (POMD) process algorithm is proposed as an autonomous job allocation strategy in cloud computing. In order to overcome the limits of concurrent considerations, the suggested solution seeks to maximise VM productivity, streamline task allocation and the use of resources, and provide load balancing amongst virtual machines based on Time to Finish, expenditure, and consumption of resources. Using CloudSim, the performance study of the suggested approach was contrasted with the load balancing algorithms now in use in datasets like Google Cloud Jobs (GoCJ). According to the trial results, the suggested POMD-EFA strategy performed better than the other algorithms in terms of decreasing the Time to Finish, minimizing expenses, and increasing the efficiency in the application of resources.