able of content: Introduction Motivation Power utilization models are crucial for developing and improving energy-efficient operations in data centers to reduce unnecessary energy use

Table of content:

1. Introduction
2. Motivation

Power utilization models are crucial for developing and improving energy-efficient operations in data centers to reduce unnecessary energy use. The requirement for energy efficiency in data centers has grown significantly and has been complicated during the past few years. All components of the data center design must carry out their assigned functions to reduce downtime, necessitate the proper energy supply, and assure high data availability.

The technical infrastructure, the backbone of all IT infrastructures, includes technical cooling, power supply, and technical security. Any physical infrastructure failure, no matter how minor, significantly affects how well IT services operate. A green data center’s energy efficiency and minimal environmental impact are crucial components. A data storage, management, and dissemination facility where all systems, particularly the electrical and mechanical frameworks, enhance energy efficiency is referred to as a green or sustainable data center. It reduces carbon emissions, saves money, and boosts productivity. These green data centers assist contemporary businesses in energy conservation and CO2 reduction. Their use is increasing both among large organizations and SMBs worldwide.

Such data centers can effectively achieve the goals for a wide range of corporate data, from data collection to processing, evaluation, and distribution. This manuscript aims to examine recent findings in green cloud computing research and summarize the main issues that have been brought up, such as how it increases data center power consumption. The software components need a lot of electricity. Various strategies can be used at the level of individual software to lower the data center’s energy consumption. When performing computationally or data-intensive tasks, the CPU core is the main resource, followed by cloud storage. When conducting communication-intensive tasks, connected equipment like network cards and routers, switches, and others use significant energy.

The operating system (OS) sits between the layers of physical hardware and applications in the data center architecture. Most research on data center power consumption research is done at the hardware level, but software-level research is just as crucial for reducing power usage. The innovations made at the level of physical components should be exploitable by software. The development of green data centers is only useful if the software used in those facilities is as effective as the hardware technological breakthroughs and uses a lot of resources. While apps create the need for resources, physical hardware is the element that uses IT power.

So it becomes equally vital to examine the specifics of power modeling and energy consumption at the software layer. In addition to the different methods used at the OS, virtualization, and application levels, problem-solving strategies like load balancing, workload categorization, VM placement, and VM migration reduce energy consumption by consolidating physical servers and dynamically altering activities. These methods show promise in reducing energy consumption in high-performance cloud data centers.

This article analyzes the energy consumption of several software layers in data centers at various levels, including OS, virtualization, application, and data center.

• Project Summary (Energy usage in data centers: a system’s perspective)

1. Project Detail
   1. Managing power dynamically in data centers
   2. Problem solving approaches
      1. Virtual machine migration
      2. Load balancing
      3. Workload categorization and prediction
      4. Virtual machine placement

Virtual machine placement is choosing the best PM for a specific VM. Therefore, a VM placement algorithm identifies the optimal VM to PM connection, whether it is a new VM placement or a VM migration for placement re-optimization. Based on the goal of placement, VM placement techniques may be loosely divided into two categories: Power-based and QoS-based. Furthermore, based on the primary approach used to obtain an appropriate VM-PM mapping, VM placement strategies are categorized as follows: Stochastic Integer Programming and constraint programming.

The methods as mentioned earlier cannot forecast the future. Hence they are not appropriate for the current situation. Machine-learning techniques can arrange virtual machines in the best location. They include reinforcement learning, artificial neural networks, and fuzzy reinforcement learning. Population-based approaches start with a group of singular solutions that multiply over time. This category emphasizes investigation and a wider range of search options.

Population-based approaches use the concept of dominance in their screening to find the Pareto optimal solutions. Genetic, Ant Colony Optimization (ACO), Memetic, Firefly, Whale optimization, Sine-Cosine Algorithm, and Salp Swarm Algorithm are the strategies employed. Single solution-based algorithms start with a single solution, which is then altered and changed as the optimization process progresses. These algorithms aim to increase search strength in specific areas and are exploitation-focused. EAGLE algorithm, Imperialist competitive algorithm, and Krill herd algorithm are some of the strategies.

1. Impact on Environment
   1. Renewable energy
   2. E- waste

The problem of e-waste has been exacerbated by the quick rise in consumer electronics use and the development of enterprise-class and hyperscale computers. It is the joint responsibility of consumers, manufacturers, businesses, and governments to guarantee that this trash is minimized, reused, and correctly recycled. To encourage the reuse and recycling of e-waste, WEEE (Waste Electrical and Electronic Component) adopts laws at the national, state, and provincial levels. These regulations aim to reduce the amount of e-waste dumped in landfills and the use of such resources.

LCD monitors, LCD and Plasma televisions, and PCs with Cathode Ray tubes make up most of the e-waste. This does not imply, however, that other electrical and electronic devices do not fall under this heading. E-waste is any old electrical device. Almost all technology-based companies create e-trash. However, the data center’s physical footprint in this area is small. Uninterrupted Power Supplies (UPS), generators, and other elements make up data centers. These parts must first be transformed into non-24/7 roles and have a long primary lifecycle of around 5 to 10 years before they can be recycled.

1. Challenges
   1. Virtual machine migration
   2. Load balancing
   3. Workload categorization and prediction
   4. Virtual machine placement
2. Conclusion

• References