9th International Conference on ICT for Sustainable Development

Authors - Prathamesh Vijay Lahande, Parag Ravikant Kaveri, Jatinderkumar R. Saini
Abstract - The resource optimization process in the cloud is crucial and can be achieved through the ideal Load Balancing (LB) mechanism. The cloud under-goes several challenges with resource optimization due to poor LB mechanism, where its Virtual Machines (VMs) are either overloaded or idle. The main aim of this experimental-based research is to enhance the LB mechanism of the cloud by implementing and comparing the performance of novel hybrid LB algorithms RLFCFS and RLSJF to optimize the resources. The RLFCFS and RLSJF novel LB algorithms are designed by combining the Reinforcement Learning (RL) technique with the heuristic FCFS and SJF algorithms. The proposed algorithms improve resource optimization in terms of cost and time by facilitating enhanced LB mechanism through RL intelligence mechanism. The performance of RLFCFS and RLSJF LB algorithms is compared with respect to the average (avg.) load managed by the VMs and the avg. percentage (perc.) of deviation observed against the expected load in each experimental stage. The experimental throughput conveys that the RLFCFS LB algorithm managed an aggregate avg. load of 968.77 tasks against the RLSJF LB algorithm, which managed 999.08 tasks aggregately across all experimental stages. Concerning the avg. perc. of deviation, the RLFCFS LB algorithm deviated by 63.44 % against the ideal expected load to manage against the RLSJF LB algorithm, which deviated by 64.60 %. This shows that the RLFCFS LB algorithm gave better resource optimization results than the RLSJF LB algorithm. Lastly, these results are mathematically validated using the Simple Linear Regression model.