Data Center Energy Use Research Articles

Unique New Cooling Technology for Energy Savings in Refineries and Data Centers: Technology and Two Case Studies

Cooling is a major energy use in both refineries and data centers. An invention <xref ref-type="bibr" rid="ref1" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[1]</xref> developed by researchers at Stanford University may alleviate both energy consumption and water usage in numerous applications. This cooling technology uses a band in the atmosphere infrared spectrum that has minimal attenuation of this wavelength. This technology can significantly cool fluid below ambient during the day with incident sun on a panel. The atmosphere is basically “transparent” to radiation from 8 to 13 <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">μ</i> m. This allows light in this wavelength band to escape our atmosphere to outer space, which can be modelled as a blackbody at approximately 2 K. This is done using a “nano-photonic radiative cooling” surface made on a film that can be economically fabricated in square meter sizes. Currently, the system can radiate about 100–400 W/m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> depending on ambient temperatures. Current photovoltaic solar can produce about 200 W/m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> with an input of about 1000 W/m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> . The current prototype systems have the same form factor as solar panels; so much of the “rest of plant” is already designed. This article examines the technology and then two case studies: refinery water and energy usage, and energy use for cooling in computer data centers. This article is a revision of the original article presented at the 2021 IEEE PCIC Conference in San Antonio, Texas in September 2021 <xref ref-type="bibr" rid="ref2" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[2]</xref> .

Energy Use in Quantum Data Centers: Scaling the Impact of Computer Architecture, Qubit Performance, Size, and Thermal Parameters

As quantum computers increase in size, the total energy used by a quantum data center, including the cooling, will become a greater concern. The cooling requirements of quantum computers, which operate at temperatures near absolute zero, are determined by computing system parameters, including the number and type of physical qubits, the packaging efficiency of the system, and the split between circuits operating at cryogenic temperatures and those operating at room temperature. When combined with thermal system parameters such as cooling efficiency and cryostat heat transfer, the total energy use can be determined using a first-principles energy model. These models show that cooling of quantum computers differs in two fundamental ways from conventional data centers: (1) the energy required for cooling is much greater than the energy required for computation, and (2) the cooling loads are sensitive to the computational architecture. The temperature requirements for different qubit types can change energy requirements by orders of magnitude. Power use and computational power, as quantified by quantum volume, are analytically correlated. Approaches are identified for minimizing energy use in integrated quantum systems relative to computational power. Designing a sustainable quantum computer will require both efficient cooling and system design that minimizes cooling requirements.

Design Optimization of a Thermally Integrated Solid Oxide Fuel Cell with Triple Effect Absorption Chiller

To reduce emissions and energy use in data centers, a novel integration between a solid-oxide fuel cell (SOFC) and a triple-effect absorption chiller (AC) is proposed to meet both the electrical and cooling requirements of server computers. An installation of SOFC units at a data center can operate using natural gas infrastructure that today supplies natural gas or directed renewable fuel and which can increasingly introduce biogas, renewable hydrogen, and/or synthetic fuels in the future. In fact, SOFC systems have already been employed to produce primary power for data centers[i]with high fuel-to-electricity conversion efficiencies and these efficiencies can be further improved by integration with AC waste heat recovery technology. Utilizing the waste heat in the exhaust stream of the SOFC, a triple-effect AC system, using aqueous lithium bromide as a working fluid, can produce the required cooling for the servers and eliminate the requirement for a backup mechanical cooling system. A 200 kW SOFC system is selected for row-level discretization of power production and is coupled to an equivalent 200 kW AC system to provide the necessary cooling. An additional benefit of this discretized installation is the inherent reliability of the distributed power and cooling generation with limited failure domain. Using this system and an n+1 installation strategy, it is possible to eliminate the requirement for backup power systems as well.A numerical simulation investigates the design of the integrated 200 kW SOFC and AC system. The 200 kW SOFC is optimized to produce the required electricity at conversion efficiencies up to 60%[ii], while still producing enough thermal energy to run the 200 kW triple-effect AC system. The numerical simulation indicates that a 200 kW SOFC can produce anywhere from 80 – 130 kW of useful heat in the exhaust, as shown in Figure 1. The AC system is optimized to harness this thermal energy from the SOFC exhaust to meet the cooling demand of the servers and can operate at a coefficient of performance (COP) of up to 1.89. A control strategy for the integrated system is developed based on the power demand of the servers and the ambient conditions of the data center to provide the required power and cooling.Additionally, the numerical study addresses the multiple possible cooling configurations that could be adopted by a data center. Liquid cooled servers utilize higher temperature cooling medium, which allows the AC system to operate at a higher COP. The improvement in AC efficiency when cooling to a higher temperature is explored to further improve the chilling capacity.In conjunction with the numerical simulation, an experimental system has been constructed to demonstrate the efficacy of an integrated SOFC and AC system for data center applications. The experimental system consists of a single server rack, a 12-kW array of SOFC systems for power, and an 18-kW single-effect absorption chiller for cooling. The experimental setup is shown in Figure 2. The server computers are operated at a simulated computational load, which exercises the SOFC array and absorption chiller, testing the system’s ability to produce power and chilling. The experimental system has demonstrated the ability to operate at a COP of up to 0.77 with SOFC exhaust gasses at 235°C, producing 12.5 kW of chilling, equivalent to 47% of the cooling demand of the server rack at full capacity.The experimental system is also used to validate the thermo-physical models used in the numerical simulations. This includes the desorber physics at elevated temperatures created by the utilization of fuel cell exhaust heat, which is much hotter than typical AC heat sources. Additionally, by altering the control strategy of the AC system, it is possible to explore elevated cooling temperatures in the evaporator and absorber, pursuing improved COPs that are possible in a liquid-cooled server configuration.Finally, the financial viability of a large-scale installation of such a system is addressed. The high capital costs associated with SOFC and triple-effect AC systems typically hinder the implementation of such systems. However, our results demonstrate that the integrated system is competitive on a levelized cost basis mainly because of primary energy savings (up to 34% compared to a conventional data center) and ability to displace the installation of backup electrical and chilling systems. [i] eBay turns on first fuel cell powered data centre, has Bloom tech. (2013). Fuel Cells Bulletin, 2013(10), 3–4. https://doi.org/10.1016/s1464-2859(13)70341-9 [ii] Williams, M. C., Vora, S. D., & Jesionowski, G. (2020). Worldwide Status of Solid Oxide Fuel Cell Technology. ECS Transactions, 96(1), 1–10. https://doi.org/10.1149/09601.0001ecst Figure 1

Does not compute: Avoiding pitfalls assessing the Internet's energy and carbon impacts

Does not compute: Avoiding pitfalls assessing the Internet's energy and carbon impacts

Toward Greener Gaming: Estimating National Energy Use and Energy Efficiency Potential

Rising computing power, improved graphics quality, higher-resolution displays, and streaming delivery have rendered computer gaming an increasingly energy-intensive activity. However, the role of gaming-related energy use, and how it varies across platforms, has not been substantively examined by the energy or gaming research communities. We measured the energy consumption of 26 gaming systems representing the spectrum of technology, price, and performance. Among the findings, energy use varied widely by hardware, but equally widely depending on which of 37 game titles or 11 benchmarks were run. Cloud-gaming energy use in datacenters and networks is markedly higher than that for local gaming. Virtual-reality gaming can use significantly more or less energy than gaming with conventional displays, depending on hardware and software choices. In aggregate, we find that gaming represents $5 billion per year in energy expenditures across the United States or 34 TWh/year (2.4% of residential electricity nationally), with 24 MT/year of associated carbon-dioxide emissions equivalent to that of 85 million refrigerators or over 5 million cars. Targeted hardware and software strategies can reduce the gaming energy use by approximately half, while maintaining or improving metrics of user experience. In addition to system designers, gamers and game developers can play a significant role in managing the energy required for gaming.

Comparing datasets of volume servers to illuminate their energy use in data centers

As data centers proliferate, their energy intensity deserves close attention. Always-on operations and growing usage for cloud and other backend processes make servers the fundamental driver of data center energy use. Yet servers’ power draw under real-world conditions is poorly understood. This paper explores characteristics of volume servers that affect energy use, quantifying differences in power draw between higher-performing Standard Performance Evaluation Corporation (SPEC) and ENERGY STAR servers and that of a typical server. First, we establish general characteristics of the US installed base, before reporting hardware configurations from a major online retail website. We then compare idle power across three datasets (one unique to this paper) and explain their differences via the hardware characteristics to which power draw is most sensitive. We find idle server power demand to be significantly higher than benchmarks from ENERGY STAR and the industry-released SPEC database, and SPEC server configurations—and likely their power scaling—to be atypical of volume servers. Next, we examine power draw trends among high-performing servers across their load range to consider whether these trends are representative of volume servers, before inputting average idle power load values into a recent national server energy use model. Lastly, results from two surveys of IT professionals illustrate the incidence of more efficient equipment and operational practices in server rooms/closets. Future work should include server power field measurements in data centers of different sizes, accounting for variations in configurations and setting changes post-purchase, as well as investigating the linkage between time and server energy efficiency.

Automatically Exploring Tradeoffs Between Software Output Fidelity and Energy Costs

Data centers account for a significant fraction of global energy consumption and represent a growing business cost. Most current approaches to reducing energy use in data centers treat it as a hardware, compiler, or scheduling problem. This article focuses instead on the software level, showing how to reduce the energy used by programs when they execute. By combining insights from search-based software engineering, mutational robustness, profile-guided optimization, and approximate computing, the Producing Green Applications Using Genetic Exploration ( PowerGAUGE ) algorithm finds variants of individual programs that use less energy than the original. We apply hardware, software, and statistical techniques to manage the complexity of accurately assigning physical energy measurements to particular processes. In addition, our approach allows, but does not require, relaxing output quality requirements to achieve greater non-functional improvements. PowerGAUGE optimizations are validated using physical performance measurements. Experimental results on PARSEC benchmarks and two larger programs show average energy reductions of 14% when requiring the preservation of original output quality and 41% when allowing for human-acceptable levels of error.

The energy and greenhouse-gas implications of internet video streaming in the United States

The rapid growth of streaming video entertainment has recently received attention as a possibly less energy intensive alternative to the manufacturing and transportation of digital video discs (DVDs). This study utilizes a life-cycle assessment approach to estimate the primary energy use and greenhouse-gas emissions associated with video viewing through both traditional DVD methods and online video streaming. Base-case estimates for 2011 video viewing energy and CO2(e) emission intensities indicate video streaming can be more efficient than DVDs, depending on DVD viewing method. Video streaming benefits from relatively more efficient end-user devices than DVD viewing, though much of that savings is lost when accounting for the additional energy from network data transmission. Video streaming appears distinctly favorable when compared against any DVD viewing that includes consumer driving, which significantly increases the energy and CO2(e) emissions per viewing hour. Total US 2011 video viewing required about 192 PJ of primary energy and emitted about 10.5 billion kg of CO2(e). Shifting all 2011 DVD viewing to video streaming reduces the total primary energy use to about 162 PJ and the CO2(e) emissions to about 8.6 billion kg, representing a savings equivalent to the primary energy used to meet the electricity demand of nearly 200 000 US households each year. Sensitivity analysis indicates that results are most influenced by the end-user DVD player power demand, data transmission energy, and consumer travel for store DVDs. Data center energy use—both operational and embodied within the IT equipment—account for <1% of the total video streaming energy use. Results from this study indicate that designers and policy makers should focus on the efficiency of end-user devices and network transmission energy to curb future increases in energy use from the proliferation of video streaming.

A critical analysis of Power Usage Effectiveness and its use in communicating data center energy consumption

Data centers represent an increasingly popular construction project type, supported by the continued growth in internet-based services. These facilities can, however, consume large amounts of electricity and—especially if growth trends continue—put strain on utility grids and energy resources. Many metrics have been proposed to evaluate and communicate energy use in data centers. In many cases, the goal is that these metrics will be used to develop energy conscious behavior and perhaps data center energy rating systems or building codes to reduce average energy use. In this paper, we examine one of the more popular metrics, Power Usage Effectiveness (PUE), and discuss its shortcomings toward effectively communicating energy consumption. Our inference is that PUE is an instantaneous representation of electrical energy consumption that encourages operators to report the minimum observed values of PUE. Hence, PUE only conveys an understanding of the minimum possible energy use. Instead, we propose the use of energy-based metrics or average PUE over a significant time period—e.g., a year—to better understand the energy efficiency of a data center and to develop energy rating/ranking systems and energy codes

Experimental evaluation of a controlled hybrid two-phase multi-microchannel cooling and heat recovery system driven by liquid pump and vapor compressor

The energy use in data centers is on an accelerating rise due to both demand and technological limitations. Today, the most widely used cooling strategy for data centers is refrigerated air-cooling. Unfortunately, air-cooling presents phenomenally low efficiencies. Therefore green computing paradigms are needed to improve energy efficiency by several orders of magnitude and allow a continued chip scaling for tackling the energy crisis in future-generation data centers. A promising solution would be implementing direct on-chip two-phase cooling technology, which not only improves the heat removal efficiency but also permits the reuse of waste heat since the two-phase coolant can cool CPUs effectively at 60 °C. In the present work a specific cooling cycle using micro-evaporation technology has been experimentally evaluated considering different aspects such as cooling cycle and energy recovery efficiencies and controllability. In resume, this novel cycle shows strong competence in energy usage, heat recovery and controllability towards green data center.

Prolog to "Estimating the Energy Use and Efficiency Potential of U.S. Data Centers"

This paper offers a concise new mathematical modeling approach for estimating data center energy use and efficiency potentials at different geographic scales. It improves analytical cohesiveness and transparency, and it is accessible and open to refinements. It also adds new in sights into electricity saving potentials and efficiency measures by allowing the estimation of potentials associated with discrete measures applied to different classes of IT devices and infrastructure equipment in varying space types. Finally, it applies an improved model to generate the most recent (and more relevant) estimates of demand and savings potentials associated with nationwide efficiency improvements using recent data on the installed base of IT devices and efficiency measures in U.S. data centers.

Green Cloud Computing: Balancing Energy in Processing, Storage, and Transport

Network-based cloud computing is rapidly expanding as an alternative to conventional office-based computing. As cloud computing becomes more widespread, the energy consumption of the network and computing resources that underpin the cloud will grow. This is happening at a time when there is increasing attention being paid to the need to manage energy consumption across the entire information and communications technology (ICT) sector. While data center energy use has received much attention recently, there has been less attention paid to the energy consumption of the transmission and switching networks that are key to connecting users to the cloud. In this paper, we present an analysis of energy consumption in cloud computing. The analysis considers both public and private clouds, and includes energy consumption in switching and transmission as well as data processing and data storage. We show that energy consumption in transport and switching can be a significant percentage of total energy consumption in cloud computing. Cloud computing can enable more energy-efficient use of computing power, especially when the computing tasks are of low intensity or infrequent. However, under some circumstances cloud computing can consume more energy than conventional computing where each user performs all computing on their own personal computer (PC).

Data center design and location: Consequences for electricity use and greenhouse-gas emissions

The rapidly increasing electricity demand for data center operation has motivated efforts to better understand current data center energy use and to identify strategies that reduce the environmental impact of these buildings. This paper builds on previous data center energy modeling efforts by characterizing local climate and mechanical equipment differences among data centers and then evaluating their consequences for building energy use. Cities in the United States with significant data center activity are identified. Representative climate conditions for these cities are applied to data center energy models for several different prototypical space types. Results indicate that widespread, effective economizer use in data centers could reduce energy demand for data centers by about 20–25%, equivalent to an energy efficiency resource in the US of ∼13–17 billion kWh per year. Almost half of the potential savings would result from better airflow management and proper control sequences. The total energy savings potential of economizers, although substantial, is constrained by their limited potential for use in server closets and server rooms, which together are estimated to account for about 30% of all data center energy demand. Incorporating economizer use into the mechanical systems of larger data centers would increase the variation in energy efficiency among geographic regions, indicating that as data center buildings become more energy efficient, their locations will have an increasing effect on overall energy demand. Differences among regions become even more important when accounting for greenhouse-gas emissions. Future data center development could consider site location, along with efficiency measures, to limit the environmental impact attributable to this increasingly prominent economic sector.