Data Center Power Research Articles

Czy plastik może rozpocząć nową erę w archiwizacji danych?

Wraz z szybkim rozwojem technologii informatycznych wiele aspektów naszego życia przechodzi cyfrową transformację. Z roku na rok w sieci pojawia się coraz więcej użytkowników, a ciągle udoskonalana sztuczna inteligencja zyskuje na popularności, co prowadzi do wzrostu produkcji informacji. In formacje te przechowywane są głównie w centrach danych, które – przy nieustającym napływie nowych treści – są zmuszone powiększać swoją powierzchnię. Łączy się to z coraz większymi kosztami utrzymania i zapotrzebowaniem na energię. W 2021 r. zużycie energii elektrycznej przez centra danych wyniosło 220–320 TWh, co odpowiada około 0.9–1.3% globalnego zapotrzebowania. Bazy danych wymagające ciągłego dostarczania energii są czynnikiem generującym około 1% całkowitej emisji dwutlenku węgla. Ponadto raportowano już, że przy rosnącym wykładniczo zapotrzebowaniu, za 10 lat ilość dostępnego krzemu klasy mikroprocesorowej nie będzie wystarczająca do przechowywania wszystkich danych. Dlatego naukowcy poszukują alternatyw dla obecnych nośników danych i opracowują nowe technologie z wykorzystaniem cząsteczek chemicznych, a w ostatnim czasie plastiku. W artykule przedstawione zostały przykłady rozwijających się technologii przechowywania danych cyfrowych w polimerach. Omówiono polimery jako nośniki danych i porównano do obecnie stosowanych rozwiązań. Przedyskutowano także, czy plastik może stać się przyszłym materiałem do archiwizacji informacji. Can plastic start a a new era in data archiving? With the rapid development of information technology, many aspects of our lives are undergoing a digital transformation. An increasing number of users are going online every year, and constantly improving artificial intelligence is gaining popularity, which leads to the growing production of information. Nowadays, information is usually stored in data centres, which will be forced to increase their space with the constant flow of new bits of information. Together with the increase in their space, energy consumption and associated maintenance costs are escalating. In 2021, global data centre power consumption was 220–320 TWh, which is about 0.9–1.3% of global power consumption. Continuous power supply for database operations is responsible for about 1% of total carbon dioxide emissions. Furthermore, it has already been reported that with the exponentially growing amount of data, in about 20 years, the amount of silicon for microprocessors will no longer be sufficient to store all the information. Therefore, scientists are looking for alternatives to the currently used data storage solutions and are developing new technologies using chemical molecules. Recently, even plastic has been explored as a data carrier. In this work, we present examples of new technologies for data storage in polymers. We have discussed polymers as data carriers in comparison with currently used solutions and deliberated whether plastic can become a future material for information archiving.

Output analysis and capacity configuration of green backup power supply in data center

The backup power supply ensures the stability and reliability of the power supply for a data center. Starting from green backup power supply, this paper studies the selection and configuration method of energy storage mode of backup power supply according to the backup power demand of data center and peak regulation demand of power grid, and analyzes different forms of green backup power supply and the output characteristics of hydrogen fuel cell. Aiming at improving the economy of green backup power assisted peak regulation, the capacity configuration principles of hydrogen energy storage and electrochemical energy storage technology are clarified, and then the capacity configuration of standby power in data center is simulated. The results show that the capacity configuration requirements of lithium battery, fuel cell and hydrogen storage tank corresponding to engineering life can be obtained through algorithm optimization.

Overview of Voltage Regulator Modules in 48 V Bus-Based Data Center Power Systems

The intermediate dc bus voltage in modern data center backend power supply is evolving from conventional 12 V to 48 V. It still requires the voltage regulator modules (VRM) to feed the terminal loads such as memory and computing units operating with very high current (> 100 A/module) and very low logic voltage (0.8 V-1.8 V). This makes it challenging to optimize the design of load-side VRMs with quadrupled input voltage. This paper comprehensively reviews the state-of-the-art 48 V VRMs and categorizes them according to passive component utilization. The first category is inductive solution which is further divided into coupled-inductor-based converters and transformer-based converters. The second category named capacitive solution is further divided into resonant switched-capacitor-based converters (Resonant SCC) and hybrid switched-capacitor-based converters (Hybrid SCC). Typical topologies are discussed, analyzed and summarized to perform a comprehensive performance comparison, such that the characteristics of different VRMs can be manifested. Some design considerations are also given to facilitate the design of the practical prototypes. Moreover, opportunities and challenges in the future data center power system are presented to provide technical insights.

An Improved Swiss Rectifier and Its Nonlinear Control for Lower THD

Three-phase power factor correction (PFC) converters capable of step-down voltage are attractive in lower components voltage stress, and optimal design of following dcdc stage, which is an alternative for next-generation datacenter power conversion. An improved three-phase step-down PFC converter (swiss rectifier) based on harmonic-current-injection (HCI) concept is proposed. It improves input current quality by eliminating switching dead zone of HCI network and avoids short circuit fault from hardware level. According to one-cycle control (OCC), a novel nonlinear control strategy (termed as closed-loop OCC) is presented, which reduces the impact of dc inductor current ripple on input current. The principles of the improved swiss rectifier and closed-loop OCC are analyzed in detail, verified by simulation and on an 80 kHz, 300 V/2 kW prototype with digital controller. At rated condition, input current THD < 2%.

A Three-Port LCC Resonant Converter for the 380-V/48-V Hybrid DC System

The dc power distribution has become a promising solution to powering datacenters, which can reduce power conversion steps and total power losses. However, both of the two most promising options, 380 V and 48 V dc power distributions, have their own defects that cannot be ignored, such as 380 V dc power distributions have lower reliability and 48 V dc power distributions have lower efficiency. In this article, a three-port <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">LCC</i> resonant converter is proposed to eliminate the above-mentioned defects by combining the aforementioned dc distributions to form a 380-V/48-V hybrid dc system. A full-bridge <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">LCC</i> converter is used to transfer power to the 380 V load, whereas a bidirectional <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">LCC</i> converter is used to charge the 48-V battery packs, and the two are magnetically coupled by a three-port transformer. The converter realizes zero-voltage switching of all switches and zero-current switching of rectifying diodes through resonance. The battery charging current is controlled by changing the frequency of all switches to achieve the switch of constant-current charging and constant-voltage charging, and the phase-shift control is adopted to keep the output voltage constant. Simulations and experiments are carried out to verify the validity of the theory.

Utilizing landfill gas (LFG) to electrify digital data centers in China for accelerating energy transition in Industry 4.0 era

The volume of landfill gas (LFG) produced during municipal solid waste (MSW) decomposition in local landfills has increased rapidly in recent years. To promote circular economy (CE) applications through resource recovery, this article investigates the feasibility of utilizing LFG as a sustainable energy to power data centers (DCs) in Xiamen (China) and evaluates the implications of waste heat reuse on the energy consumption of DC. The objective of this study was to investigate the environmental impacts and economic benefits of this scheme based on a SFC model. With respect to its novelty, LFG is utilized as an energy supply to DCs, while the waste heat generated in the energy production is recovered. The socio-economic benefits generated in this process are also elaborated. It was found that a large number of MSW was completely recycled to generate energy. The LFG enterprise could benefit from energy sale, while the energy satisfied the annual demand for DC power consumption. Utilizing this energy output resulted in the reduction of primary energy and less CO₂ emissions. By reusing the waste heat for domestic heating, the energy efficiency could be improved, while primary energy could be further saved, avoiding more CO₂ emissions. This implies that utilizing LFG output and waste heat to power DCs not only facilitate resource recovery from unused resources to improve energy efficiency and obtain socio-economic benefits, but also promote the digitalization of DCs and sustainable development of urban cities in the framework of CE.

Task scheduling to a virtual machine using a multi‐objective mayfly approach for a cloud environment

SummaryCloud computing has been progressively popular in the arenas of research and business in the recent years. Virtualization is a resource management approach used in today's cloud computing environment. Virtual Machine (VM) migration algorithms allow for more dynamic resource allocation, as well as improvement in computing power and communication capability in cloud data centers. This necessitates an intelligent optimization approach to VM allocation design for an improved performance of application. In this article, a multi‐objective optimal design approach is proposed to tackle the tasks of VM allocation. Multi‐Objective Optimization (MOO) is a strategy adopted by several methods to handle tasks and workflow scheduling issues that deal with numerous opposing goals. In the cloud computing context, effective task scheduling is critical for achieving cost effective implementation as well as resource utilization. To address the optimal solution, this article proposes an entropy‐based multi objective mayfly algorithm is assessed using a convergence pattern in MOO. The model is tested by implementing in a cloud simulator and results prove that the recommended model has an improved performance with regard to factors such as time and utilization rate.

Dynamic Model and Converter-Based Emulator of a Data Center Power Distribution System

Data centers have become a widespread power electronics (PE) load, which has significant impact on the power grid. In order to investigate the data center load characteristics, this article proposes a complete dynamic model for a typical data center ac power distribution system. A generalized model with mode transition is proposed to coordinate different power stages in the data center power system. Meanwhile, to help evaluate the grid dynamic performance and transient stability, an all-in-one load data center power emulator is developed on a reconfigurable PE converter-based hardware testbed (HTB). The dynamic power model is digitized and simplified to be implemented in two local voltage source inverters on the HTB. This proposed data center power emulator has been verified experimentally in a regional network. Dynamic performances during voltage sag events and server load variations are emulated and discussed. The article details the design, development, and verification of the data center model and power emulator. The proposed model and emulator provide an effective, easy-to-use tool to better design data centers and study the interaction with the power system.

A superconducting magnetic energy storage based current-type interline dynamic voltage restorer for transient power quality enhancement of composited data center and renewable energy source power system

Simultaneous protection of sensitive renewable power generators and sensitive loads under transient voltage disturbances is an urgent issue in current power systems. Most existing solutions are based on separate custom power devices and energy storage systems. To efficiently utilize renewable energy under voltage sags and reduce energy storage capacity, a current-source-inverter interline dynamic voltage restorer (CSI-IDVR) based on superconducting magnetic energy storage (SMES) is proposed. The current source topology is designed for the IDVR to obtain a more appropriate current rise limitation and fewer switches. The CSI-based voltage dual control is designed and implemented to suppress the voltage oscillations, ensuring the transient stability of the sensitive renewable power generator and sensitive loads during asymmetric grid faults. An internet data center (IDC) combined with a typical renewable energy generator, i.e., doubly-fed induction generator (DFIG), is analyzed. Under transient voltage disturbances, the proposed CSI-IDVR can maintain the appropriate voltage profiles of the DFIG and IDC. Compared to the separate DVRs, the required SMES capacity in the CSI-IDVR can be substantially reduced. The feasibility and performance of the proposed CSI-IDVR are verified by numerical simulation.

Sustainable Data Centers in Southeast Asia: Offshore, Nearshore, and Onshore Systems for Integrated Data and Power

While data centers are increasingly important for modern society owing to the expansion in data processing demand, they also contribute to emissions and thus global warming. The design of such social infrastructure requires sociotechnical evaluation, including the identification of the different needs of multiple stakeholders. This study designs and proposes concepts for data center systems as integrated infrastructure consisting of both data center and green power plant in response to rapid regional growth and potential of offshore and nearshore infrastructure in Southeast Asia. To design such an infrastructure involving multiple stakeholders with differing needs, a systems approach is employed to analyze stakeholder values and system functions leads to the identification of key stakeholders and system requirements. Subsequently, a quantitative multi-objective model based on prioritized stakeholder needs and system requirements was developed to comprehensively evaluate the data center design tradespace. Simulation results of multiple location across the region imply that Singapore is an attractive location for investment given an offshore data center with an offshore fixed wind power plant that can contribute to a 50% reduction in scaling difficulty and a 50% reduction in lifetime CO2 emissions compared to onshore data centers with solar plants, while 5% increase of total investment.

Feasibility study of a combined system of electricity generation and cooling from liquefied natural gas to reduce the electricity cost of data centres

Global data centre power demands would expand from 286 TWh in 2016 to around 321 TWh in 2030. The cooling system represents electricity consumption of approximately 40–50% of the total energy used. The global LNG trade has reached 356.1 MTPA. Cold energy equal to 89025 GW is released into the ocean. Therefore, this study focused on the technical and economic feasibility of an LNG receiving terminal combined with a data centre using a direct expansion cycle (DEC), a Rankine cycle (RC) and a combination of Rankine cycle and direct expansion cycle (RC + DEC) under different natural gas distribution pressures to produce a supply of cooling water and electricity to reduce electricity consumption and greenhouse gas emissions. According to the study, the RC + DEC produces the maximum cold water at 7 °C, with the total cold energy of 44.23 MW which is sufficient for cooling a data centre with a capacity of 5345 racks, to reduce the electricity for conventional cooling system is 13521 kWh and generated electricity form turbine is 9968 kWh. This research has the potential to reduce the operating costs of data centres by more than USD 23.87 million per annum as well as CO2 emissions by 83859 t per annum with exergy efficiency of 78.94%. In an economic study, indicated a payback period of 1.60 years with an IRR of 62%.

Optimization of power consumption in data centers using machine learning based approaches: a review

&lt;span&gt;Data center hosting is in higher demand to fulfill the computing and storage requirements of information technology (IT) and cloud services platforms which need more electricity to power on the IT devices and for data center cooling requirements. Because of the increased demand for data center facilities, optimizing power usage and ensuring that data center energy quality is not compromised has become a difficult task. As a result, various machine learning-based optimization approaches for enhancing overall power effectiveness have been outlined. This paper aims to identify and analyze the key ongoing research made between 2015 and 2021 to evaluate the types of approaches being used by researchers in data center energy consumption optimization using Machine Learning algorithms. It is discussed how machine learning can be used to optimize data center power. A potential future scope is proposed based on the findings of this review by combining a mixture of bioinspired optimization and neural network.&lt;/span&gt;

A Six-Level Flying Capacitor Multilevel Converter for Single-Phase Buck-Type Power Factor Correction

This article investigates behavior of flying capacitor multilevel (FCML) converters in single-phase buck-type power factor correction (PFC) applications. Recent developments in FCML converters using gallium nitride (GaN) transistors are leveraged to improve power density of a single-phase 240 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\text{V}_{\text{RMS}}$</tex-math></inline-formula> ac to direct 48 V dc conversion stage in data center power delivery applications. Here, we experimentally demonstrate this concept in a digitally controlled six-level FCML converter hardware prototype. The experimental prototype can deliver 4.5 A average output current at 48 V, resulting in 216 W output power. A key contribution of this article is experimental demonstration of an FCML buck converter in a single-phase PFC application where the flying capacitor voltages follow fractions of the rectified input voltage by swinging at twice-line frequency.

Leveraging renewable oversupply using a chance-constrained optimization approach for a sustainable datacenter and hydrogen refueling station: Case study of California

This work evaluates the sustainability of using curtailed renewable energy to power data centers and provide hydrogen to hydrogen refueling stations. An optimization problem is formulated to minimize the annual system costs for various case scenarios with a chance-constrained optimization approach to handle uncertainties. System components included natural-gas–powered combined cooling, heating, and power systems, electrolyzers, hydrogen fuel cells, heat pumps, hydrogen tanks, and battery energy storage systems. The results revealed that powering data centers with a mix of curtailed renewable energy and fossil fuels (Case 2) was the most economical option. Using only renewable sources (Case 3) allowed for a zero-emission alternative but was the most expensive; the data center powered by fossil fuels had 49,009 tons of annual CO2 emissions. From a sizing perspective, combined cooling, heating, and power systems dominated in Case 2, and for Case 3, a fuel cell was the major power source and a heat pump was the only cooling option. Sensitivity analysis demonstrated that the annual costs were most sensitive to the technology costs and may reduce capital expenditures in the near future. Overall, this work presents a comprehensive step toward economic and environmentally friendly data centers comprising both operator and investor perspectives.

Making distributed edge machine learning for resource-constrained communities and environments smarter: contexts and challenges

The maturity of machine learning (ML) development and the decreasing deployment cost of capable edge devices have proliferated the development and deployment of edge ML solutions for critical IoT-based business applications. The combination of edge computing and ML not only addresses the development cost barrier, but also solves the obstacles due to the lack of powerful cloud data centers. However, not only the edge ML research and development is still at an early stage and requires substantial skills normally missed in resource-constrained communities, but also various infrastructure constraints w.r.t. network reliability and computing power, and business contexts from the resource-constrained environments require different considerations to make edge ML applications context aware through smart and intelligent runtime strategies. In this paper, we analyze representative real-world business scenarios for edge ML solutions and their contexts in resource-constrained communities and environments. We identify and map the key distinguished contexts of distributed edge ML and discuss the impacts of these contexts on data and software components and deployment models. Finally, we present key research areas, how we should approach them, and possible tooling for making edge machine learning solutions smarter in resource-constrained communities and environments.

Enhancing dynamic energy network management using a multiagent cloud-fog structure

Smart Grid benefits from information and communications technology (ICT) to integrate data from different sources across the network. The growing market for electric vehicles (EV) and electric devices is increasing energy demand. To manage a large amount of complex intelligent equipment, EV charging and discharging, and data-intensive devices, a reliable modern smart grid with a service-oriented, secure, efficient, and cost-effective system is compelling. Cloud computing is a promising approach to achieve that objective as monitoring power grid data flow and data center management are the key. This paper proposes an optimization model for energy management within the power cloud. Our energy system model enables accessible services to computing resources and real-time data stream processing within an integrated environment. Using digital technology, the proper implementation of our model reduces costs and energy consumption and improves the smart grid reliability for customers and energy providers in a distributed manner. This paper presents the implementation of a three-procedure optimization algorithm within a novel Multi-Agent Cloud-fog Structure (MACS) to meet the requirements raised by smart grid communication and distribution. Our model promotes reliable energy consumption adjustments by end-users who can choose power supplied by solar, wind, geothermal, biomass, or other renewable sources or from non-renewable sources in ways that reveal opportunities for demand-supply balance and energy saving.

Challenges and Approaches in Green Data Center

Cloud computing is a fast evolving area of information and communication technologies (ICTs)that hascreated new environmental issues. Cloud computing technologies have a widerange ofapplications due to theirscalability, dependability, and trustworthiness, as well as their abilityto deliver high performance at a low cost.The cloud computing revolution is altering modern networking, offering both economic and technologicalbenefits as well as potential environmental benefits. These innovations have the potential to improve energyefficiency while simultaneously reducing carbon emissions and e-waste. These traits have thepotential tomakecloud computing more environmentally friendly. Green cloud computing is the science and practise of properlydesigning, manufacturing, using, and disposing of computers, servers,and associated subsystems like displays,printers, storage devices, and networking and communication systems while minimising or eliminatingenvironmental impact. The most significant reason for a data centre review is to understand capacity,dependability, durability,algorithmic efficiency, resource allocation, virtualization, power management, andother elements. The green cloud design aims to reduce data centre power consumption. The main advantageof green cloud computing architecture is that it ensures real-time performance whilereducing IDC’s energyconsumption (internet data center).This paper analyzed the difficultiesfaced by data centers such as capacityplanning and management, up-time and performance maintenance, energy efficiency and cost cutting, realtime monitoring and reporting. The solution for the identified problems with DCIM system is also presentedin this paper. Finally, it discusses the market report’s coverage of green data centres, green computingprinciples, andfuture research challenges. This comprehensive green cloud analysis study will assist nativegreen research fellows in learning about green cloud concerns and understanding future research challengesin the field.

Research on Server Green Computing in 5G Data Center

The global 5G commercial era is approaching. The throughput and number of connections of 5G will be 100 times that of 4G. The processing power of massive data will determine the experience of users surfing the Internet. In order to cope with the geometric growth of network data, the computing power and storage capacity of 5G data centers need to be correspondingly improved. With the surge of servers, the energy consumption of the 5G data center has also been greatly improved. How to effectively improve the performance of the server and reduce the energy consumption has become an extremely urgent problem. This paper focuses on the management method of the data center server in the context of 5G network transformation, and proposes effective solutions in combination with the application scenarios of large-scale 5G data centers.

Research advances on AI-powered thermal management for data centers

Given the complex nature of data centers' thermal management, which costs too many resources, processing time, and energy consumption, thermal awareness and thermal management powered by artificial intelligence (AI) are the targeted study. In addition to a few research on AI techniques and models, other strategies have also been introduced in recent years. Data center models, including cooling, thermal, power, and workload models, and their relationship are factors that need to be understood in the optimal thermal management system. Simulation approaches have been proposed to help validate new models or methods used for scheduling and consolidating processes and virtual machines (VMs), hotspot identification, thermal state estimation, and power usage change. AI-powered thermal optimization leads to improved process scheduling and consolidation of VMs and eliminates the hotspot from happening. At present, research on AI-powered thermal control is still in its infancy. This paper concludes with four issues in thermal management, which will be the scope of further research.

Switching Transition Analysis and Optimization for Bidirectional CLLC Resonant DC Transformer

The demand for isolated bidirectional dc transformer (DCX) is driven by the fast development of energy storage system, data center power supply and transportation electrification. Due to zero voltage switching (ZVS) and small rms current, the open-loop CLLC resonant converter operating at the resonant frequency is considered a promising candidate for DCX with a constant voltage transfer ratio. To solve unsmooth bidirectional power flow and current distortion in traditional CLLC-DCX with synchronization rectification (SR) modulation, a dual-active-synchronization (DAS) modulation is adopted with identical driving signals on both sides. Firstly, the switching transition of this modulation is fully analyzed with the consideration of large device output capacitances. After comparison of different transitions, a so-called Sync-ZVS transition is found more desirable with ZVS, no deadtime conduction loss, and load-independent voltage gain. In order to achieve this switching transition, an Axis and Center Symmetric (ACS) method is proposed. Based on this method, an overall design procedure of CLLC-DCX with DAS modulation is also proposed. Finally, Sync-ZVS transition and the proposed ACS method are both verified by three 18 kW 500 kHz prototypes with a 98.8% peak efficiency. This article is accompanied by one video demonstrating the load changing test.