Author(s): Sweta S Munnoli and Devi Venkatesh Gowtham

Abstract: The effects of cloud computing on energy consumption and environmental degradation are rising to the forefront as the industry grows in size. Half of data centers' operational expenses are attributable to energy use, according to statistics. The increasing demand for energy puts significant strain on the environment and necessitates a substantial amount of energy. People in the IT industry are paying special attention to the problem of cloud data centers' excessive energy usage. This issue likewise requires immediate resolution. Two factors are now contributing to excessive energy use. First, a low server usage ratio is caused by a resources scheduling method that prioritizes completion time, and little jobs consistently use a lot of resources. Secondly, data centers' existing cooling systems are built on the peak value approach, which results in an oversupply of cooling, higher operational costs, and massive energy waste. This study proposes a new paradigm for green cloud data centers, taking into account the reasons for a great deal of energy of data centers. As a means of assessing the efficacy of optimal elastic scaling for actual data center traces, this study delves into workload and information center modeling to aid in the prediction of workload as well as data center operation. Working with an analytical approach, we simulate the workload and information center operations using three different machine learning algorithms. In order to determine the optimal design and scaling activities, we first utilize a mathematical framework as a predictor to assess and test the set of optimization solutions. Only then will we apply it to the actual data center. The findings demonstrate that analysis-oriented machine learning may aid in determining optimal workload demand forecast values and assessing the necessary provisioning of scaling and resource capacity. Finding the best setup and solving the elasticity scaling boundary values are both done via machine learning. We propose an intelligent refrigerating engine and a scheduling control engine that use AI-related strategies to reduce energy usage. Also, we validate the framework's viability, construct the scheduling control engine, and design a platform for environmentally friendly cloud data centers. It shows that the framework is capable of making the cloud platform run efficiently and with little power consumption and that data centers can run with little power waste.

DOI: 10.33545/27075907.2024.v5.i2b.74

Pages: 93-96 | Views: 643 | Downloads: 257

Download Full Article: Click Here