Machine Learning and Deep Learning Algorithms for Green Computing

2. Energy-Efficient Computing Using Machine Learning

Using the scheduling mechanism, the power consumed can be decreased in the software approach. The scheduling approach allocates service requests from clients to virtual machines (VMs) that can fulfil them within the parameters of the service level agreement (SLA). The scheduling approach chooses virtual machines (VMs) running on physical machines (servers) that have the potential to use less power in order to meet the goal of lowering the consumption level of power. Numerous suggested techniques rely on the scheduling or software approach (A. Hameed et al, 2016). Energy efficiency is highly valued in modern cloud computing since it reduces operating costs and adheres to green computing ideals. Resource management in the cloud encompasses many different aspects, such as workload consolidation, job scheduling, virtual machine deployment, and more. Researchers work to establish the best policies for this management. In these attempts, machine learning is essential. In this research, we conduct a comprehensive assessment of the literature on machine learning (ML) in recent works to provide recommendations for energy conservation in cloud computing systems. Large-scale data centres are outfitted with mechanical and electrical gear and sensors that throw millions of data points a day. A machine learning method called “neural network framework” analyses these data points to assess how effectively energy is used. The strategy was tested in Google's data centre. Results indicated that it could result in energy savings (Hatzivasilis, G et al., 2008) (Hassan, M. B et al., 2022). The most important greenhouse gas is carbon dioxide (CO2). According to recent research, five automobiles' worth of on release of carbon dioxide. over course of a Natural Language Processing (NLP) model's development utilising deep neural networks “Deep Learning” (Ning, Z et al., 2019) (Bharany, S et al., 2022) (Luo, T et al., 2023). The mathematical model's initially unknowable parameters are estimated using machine learning. Specifically, we need to ascertain the task performance level. Predict a priori utilisation of resources (e.g., CPU consumption) by different activities under existing workloads and agreements (e.g., reaction periods) given burden characteristics, host features, and competition between jobs on the same host (Raja, S. P., 2021) (Gholipour, N et al., 2021). When paired with precise or approximate schedulers, algorithms based on machine learning can precisely predict system behaviour and assign jobs to hosts in a way that balances power consumption, income, and service quality (Yu, P et al., 2020).