Energy consumption of data centers worldwide How will the Internet become green? Ralph Hintemann (Author) Simon Hinterholzer (Author) Borderstep Institute Borderstep Institute Berlin, Germany Berlin, Germany hintemann@borderstep.de hinterholzer@borderstep.de Abstract— The increasing digitalization of the economy and so- Against this background, the present contribution addresses ciety is leading to a dynamic increase in the amount of data that the following questions: is processed and stored in data centers. Some scientific studies • How did the energy consumption of data centers and assume that this growth will also lead to a significant increase in their individual components develop in Germany be- the energy consumption of data centers worldwide. This article deals with the questions of whether and to what extent the energy tween 2010 and 2017? consumption of data centers have increased in the past and what • What are the results of different studies on the devel- future developments can be expected. In addition, future chal- opment of the energy consumption of data centers in a lenges to sustainable data center operation will be analyzed. The global comparison? article deals in particular with the topics of improving the energy • Which current trends influence the energy consump- efficiency of IT hardware and software, the use of regenerative tion of data centers? electricity for the operation of data centers, and the use of waste heat from data centers. • How can the energy consumption of data centers de- Index Terms— Data centers, energy consumption, waste heat, velop up to the year 2030? What forecasts are likely? regenerative electricity, Bitcoin mining, energy efficiency, IT • What are the particular challenges for energy-efficient hard- and software operation of data centers in the future? I. INTRODUCTION The analyses and results presented in this article were pro- duced as part of the project TEMPRO – "Total Energy Man- Video streaming, social media, big data, Bitcoin, artificial agement for professional data centers." TEMPRO is supported intelligence, and digitalization of business processes and pro- through the 6th Energy Research Programme of the German duction flows – these and other trends are leading to more and Federal Government. more data being stored and processed in data centers. Data cen- ter capacity is growing dynamically. The IT company Cisco II. METHODOLOGY assumes in its analyses that the worldwide computing capaci- ties of data centers measured in workloads and computer in- This paper deals with the energy requirements of data cen- stances will more than double between 2016 and 2021 (2.3 ters, presenting the findings from various international studies fold), the data storage capacities in the data centers will even and comparing them with the authors’ own calculations. The grow by a factor of almost 4 to 2.6 ZB in the same period [1]. calculations are carried out with the help of an extensive struc- Whether this significant growth in computing and storage tural model of the data center landscape in Germany which was developed at the Borderstep Institute and is updated annually. capacity will also be accompanied by an increase in energy consumption can hardly be predicted from today's perspective. In this model data centers are defined as all enclosed spatial units such as server cabinets, server rooms, building parts, or However, most available analyses assume a more or less signif- entire buildings in which at least three physical servers are in- icant increase in energy consumption [2]–[14]. Yet the range of stalled. The energy requirements of stand-alone servers are also forecasts is very wide. While Andrae/Edler, for example, pre- calculated. The development of the data center capacities is dict a 15-fold increase in the energy consumption of data cen- calculated in particular on the basis of the IT equipment (serv- ters worldwide between 2010 and 2030 [3], Shehabi et al. ers, storage, networks) in the data centers. The various size found that the energy consumption of data centers in the USA classes of the data centers in Germany are described in the has stabilized in recent years due to increasing energy efficien- model in terms of their equipment with different server types, cy [4]. For more than 10 years, the authors themselves have storage systems, and network infrastructures. The model also been analyzing the development of the energy consumption of takes into account the age structure of the servers and the ener- data centers with the help of a comprehensive structural model gy requirements of the different server types in different oper- of data centers in Germany [9], [14]–[25]. For Germany, an ating states. In addition, the data center infrastructures such as increase of 25% in the energy consumption of data centers in the period from 2010 to 2017 has been calculated [27]. air conditioning, power supply, UPS, etc. are modeled for dif- ferent size and redundancy classes. The following sources in particular were used for the calcu- lations: • Study “Entwicklung des IKT-bedingten Strombedarfs in Deutschland“ (Development of ICT-related electrici- ty consumption in Germany) by Fraunhofer IZM and Borderstep on behalf of the Federal Ministry of Eco- nomics and Energy [26], • Current and publicly available results of studies on the development of the data center market from various analyses [16], [17], [28]–[35], • Data from the market research institute Techconsult on market development for servers, storage, and network Fig. 1. Energy consumption of servers and data centers in Germany in the components (eanalyzer) [36]–[38], years 2010 to 2017 (Source: Borderstep) • Data from the market research institutes IDC and EITO Significant further improvements in the efficiency of data on market development for servers, storage, and net- center infrastructures can be expected in the future. Newly built works in Germany and Europe [39], [40], larger data centers today often realize PUE values of 1.3 and • Scientific literature and manufacturer information on lower [27], [33].3 the development of energy consumption for servers, Growth in the data center market is driven primarily by the storage, and network products, and for advanced effi- cloud market (Figure 2). Large international cloud providers in ciency technologies for data centers. particular are currently building up resources in Germany and are thus meeting the requirements of German companies to A detailed description of the model has already been pro- enable the storage and processing of their data in Germany. vided in various publications. [9], [15], [25], [26], [41]. International cloud companies often use the support of coloca- III. DEVELOPMENT OF THE ENERGY CONSUMPTION OF tion providers, thereby driving the colocation market at the DATA CENTERS IN GERMANY same time. The share of colocation data centers in the IT area of all data centers in Germany again increased significantly in Using the structural model of data centers presented above, 2017 and will be 32% in 2017 [27]. it was calculated that despite a significant improvement in the energy efficiency of data centers, their energy consumption in Germany increased by 25% between 2010 and 2017 (Figure 1). The increase was mainly due to an increase in the number and performance of IT components in the data centers. In 2017, around 2.37 million servers were in operation in Germany, 18% more than in 2010. The number of data centers equipped with IT hardware for data storage and network technology also increased significantly – the number of hard disks in Germany more than doubled between 2010 and 2017. As a result of this increase, the power consumption of IT components in data cen- ters rose from 5.8 billion kWh in 2010 to 7.9 billion kWh in 2017. In contrast, the electricity consumption of the data center infrastructure rose only slightly due to the measures imple- mented to increase energy efficiency: from 4.7 billion kWh in 2010 to 5.3 billion kWh in 2017. The average PUE1 value of Fig. 2. Development of the IT area of data centers in Germany with the share of cloud data centers in the years 2010 to 2017 and forecast to 2020 data centers in Germany fell from 1.98 to 1.752 between 2010 (Source: Borderstep) and 2017. This corresponds to an average increase in data cen- ter infrastructure efficiency of 13% [27]. From a regional perspective, the greater Frankfurt area and thus the federal state of Hesse in particular are benefiting from 1 3 The PUE is an indicator often used to assess the energy efficiency of data The Power Usage Effectiveness (PUE) value indicates the ratio of the total energy consumption of a year to the annual energy consumption of centre infrastructures. The advantage of PUE is that it is easy to deter- the IT department of the data center. mine. However, the interdependencies between the energy demand of IT 2 The stand-alone servers, which are normally operated without their own and the energy demand of the entire data center are very complex. If, for air conditioning, are not included in the calculation of these values. Tak- example, the energy consumption of IT is reduced, the PUE can worsen, ing stand-alone servers into account, the average PUE value in Germany even though the total energy consumption of the data center decreases improved from 1.82 in 2010 to 1.58 in 2017. significantly. the growth of the data center market. This applies in particular At the beginning of 2019, Andrae published an update of its to the colocation data center segment [34]. In 2017, there was a calculations and forecasts, which deviates significantly from total IT area of 550,000 m2 in the data centers in Hesse, the values from 2015. For the year 2030, "only" an increase in 250,000 m2 of which was in colocation data centers. As a re- the energy demand of data centers worldwide to 1,929 billion sult, the capacities in the colocation data centers – measured by kWh/a is expected. For the year 2018 it assumes 211 billion the IT space in Hesse between 2010 and 2017 – have increased kWh/a - in contrast to 539 billion kWh/a in the estimate from by 60% (Figure 3). In terms of the available IT connection re- the year 2015 [45]. sources, capacities even increased by 100%. [34]. A publication by Belkhir and Elmeligi [14] also assumes a very significant increase in the energy requirements of data centers. Based on a calculation by Vereecken et al. [46], they assume a worldwide energy demand for data centres of 275 billion kWh/a in 2009 and anticipate an annual growth rate of 10% by 2020. This growth rate was determined by the market research company Technavio [47]. This results in an energy demand of 659 billion kWh/a for data centers worldwide in 2018. Malmodin/Lunden calculate worldwide energy consump- tion of data centers to be 240 billion kWh in 2015 [44]; accord- ing to van Heddegdem et al. no less than 270 billion kWh were needed in 2012 [48]. Bitterlin assumes that data centers world- wide required 416 billion kWh of energy in 2015 [12]. Fig. 3. Development of data center capacities (IT area in m2) in Hesse in the According to a Borderstep Institute estimate, worldwide years 2010 to 2017 and forecast to 2020 (Source: Borderstep) energy consumption of server data centers increased by about 30% to 287 billion kWh between 2010 and 2015 [9]. This in- Despite the dynamic development in the data center market crease accelerated once again in the last two years. A current in Germany, it is possible that it will become less important in TEMPRO project estimate concludes that between 2015 and international comparison as a data center location. Especially in 2017 the energy consumption of data centers worldwide in- the rapidly growing segment of hyperscale data centers, which creased by approx. 20% to 350 billion kWh (Figure 4). Increas- are very large, significantly more capacity is being built ing digitalization and the sometimes very high energy demand worldwide, also elsewhere in Europe, for example, the Nether- of new applications such as Bitcoin mining were identified as lands, Ireland and Scandinavia. [33], [34]. reasons for this accelerated growth in energy consumption [27]. Operators of hyperscale data centers shy away from Ger- Even though there is little reliable information on the energy many as a business location primarily because of the high elec- consumption of Bitcoin mining to date, it can be assumed that tricity costs and the sometimes lengthy approval processes. The the worldwide annual energy consumption of this application increasing shortage of skilled personnel for data centers in increased by more than 30 billion kWh between the beginning Germany is also becoming an increasingly disadvantageous of 2017 and the end of 2018 [6], [49]. location factor [33]. Rapidly growing applications such as Bitcoin mining [6] cannot be operated in Germany at break- even due to the high electricity prices. IV. ENERGY CONSUMPTION OF DATA CENTERS IN EUROPE AND WORLDWIDE Several scientific studies examine the energy consumption of data centers. In the following, the results of selected and well-known studies are briefly presented and compared. The results of a 2015 paper by Andrae/Edler [3] have achieved a relatively high level of attention in the discussion about the energy consumption of data centers [13], [14], [42]– [44]. This is certainly due to the fact that Andrae/Edler forecast a strong increase in energy consumption. In the "expected" scenario, according to their calculations, the energy consump- Fig. 4. Energy consumption of servers and data centers worldwide in the tion of data centers worldwide will increase by a factor of 15 years 2010 to 2017 (Source: Borderstep) from approx. 200 billion kWh/a in 2010 to almost 3,000 billion kWh/a by 2030. Even in the "best" scenario, an increase by a The different assessments of the development of the energy factor of almost 6 to 1,337 billion kWh/a is calculated. In the consumption of data centers are also reflected in regional anal- "worst" scenario, the energy consumption of data centers will yses. For the USA, Shehabi et al. calculated that the energy increase by a factor of 40 to 7,933 billion kWh by 2030 [3]. consumption of data centers has hardly increased since 2010; they calculate energy consumption of 70 billion kWh in 2014, V. DISCUSSION: HOW WILL THE ENERGY compared to approx. 67 billion kWh in 2010. The small in- CONSUMPTION OF DATA CENTERS DEVELOP IN THE crease in the USA is due to a general improvement in the ener- FUTURE? gy efficiency of IT systems and infrastructure with lower PUE As the results presented in the previous section show, there values. In addition, there has been a significant shift in compu- are great uncertainties in determining the energy consumption ting power to particularly efficient hyperscale data centers with of data centers worldwide even for the current point in time. PUE values below 1.2 [4], [10]. Forecasts beyond 2020 are even more difficult due to the un- If one compares the development of the energy consump- clear development of technologies and the extent to which data tion of the data centers in the USA calculated by Shehabi et al. centers and the services they provide will be used in the future with the development in Germany calculated by Borderstep, it [4], [10]. In particular, the possible end of the efficiency ad- is striking that energy consumption in Germany seems to in- vances made to date, and thus the end of Moore's Law, could crease significantly more strongly, although the increases in the lead to a significant increase in energy consumption in the fu- server inventories are comparable. The basic structure of ture. Andrae/Edler, for example, have modeled an end to Shehabi et al.'s model is comparable to that of the Borderstep Moore's Law in their calculations by 2022 and thus also an end Institute. Therefore, it was possible to carry out a detailed com- to the significant increases in data center efficiency [3]. This is parison of the different model parameters. The differences in a major reason for the very significant increase in energy con- the developments can essentially be attributed to two factors. sumption in their scenarios. It can be discussed if and when On the one hand, there are hardly any hyperscale data centers Moores Law will end. It is clear that technical development is in Germany – therefore, computing power has hardly shifted to approaching the physical limits of conventional CMOS silicon this particularly efficient type of data center. Secondly, the as- technology. The structures currently consist of only a few sumptions in the models differ with regard to the average de- atomic layers. Leading companies such as Intel and AMD ex- velopment of the maximum power consumption of a server at pect Moore's law to come to an end around 2023 [50]. Howev- full load. While Shehabi et al. assume that maximum power er, alternative materials and 3D architectures will presumably consumption is constant [4], the Borderstep model assumes an contribute to further increases in performance and reliability increase in maximum power consumption due to a significant despite the small structures. An abrupt end to improvements in increase in the average amount of RAMs and multiprocessor the energy efficiency of computing operations is unlikely. systems. These assumptions are also confirmed for Europe by Assuming a continuation of Moore's Law, the Andrae/Edler other studies [7], [11]. Whether the differences in the model model yields much more moderate increases. Another reason assumptions are justified by the real differences in the USA and for the significant increases in energy consumption in the mod- Europe or Germany is yet to be verified by further investiga- el is the assumption that the performance of data centers is de- tions. termined by the IT traffic between the user and the data center. The results of various studies on the energy consumption of According to Cisco forecasts, this increases by 23% annually data centers in Europe are relatively similar. The Ecodesign [3], [51]. If the assumption of increasing data center perfor- Preparatory Study on Enterprise Servers and Data Equipment mance is modified in such a way that the number of workloads determines energy consumption of 78 billion kWh for data and compute instances is chosen as the measure, the calculated centers in Europe by 2015 [11]. Prakash et al. calculate energy increase in energy consumption in the model would be lower. consumption of 52 billion kWh for 2011 and forecast an in- According to Cisco, the number of workloads and compute crease to 70 billion kWh by 2020 [7]. According to estimates instances is currently growing at an average annual rate of by Borderstep, energy consumption of data centers in Western 18.6% [1]. Europe also rose significantly between 2010 and 2017. Based To illustrate the differences between the different analyses, on data on the development of workloads and server numbers figure 5 shows various forecasts of the energy consumption of in data centers of the IT company Cisco [30], [31], the authors data centers worldwide through 2030. Studies have been se- assume that energy consumption has increased from 56 billion lected which allow the presentation of the development up to kWh in 2010 by a good 30% to 73 billion kWh in 2017 [27]. the year 2030. Beside the investigations of Andrae/Edler [3], In summary, the various studies on the development of the Andrae [45] and Belkhir/Elmeligi [14] a “best case” is present- energy consumption of data centers do not provide a uniform ed. This development would occur using the Andrae/Edler picture. While some studies, e.g. Andrae/Edler, assume a very model if Moore's Law continued and moderate growth rates in strong increase, other studies assume low to moderate growth data center performance were assumed. Furthermore, according in the last ten years. to Borderstep's calculations, the trend in the development of the However, the energy efficiency of the data center infra- worldwide energy consumption of data centers between 2010 structure and thus the PUE values have improved significantly and 2017 will continue through 2030. in recent years, and the share of IT components in the energy consumption of the data centers has thus increased. tral role in the future [18]. Innovative new solutions must be implemented here. This applies to both classic air-cooled and new innovative liquid-cooled IT systems. In Germany and pre- sumably in other countries too, however, the framework condi- tions must change in order to promote the use of waste heat in data centers. So far, electricity prices in Germany have been so high that the operation of heat pumps that can raise the waste heat level of data centers to a usable level is uneconomical. Often it is economically more favorable to burn oil or gas for heat generation instead of using existing waste heat. Another future challenge for environmentally friendly data center operation is the supply of electricity generated largely from renewable sources because the supply of continuously Fig. 5. Energy consumption of servers and data centers worldwide – forecasts available electricity generated from hydropower or biomass is to 2030 limited. This means that intelligent use of fluctuating wind and solar power in data centers is becoming increasingly important. A comparison of the various forecasts shows that the possi- Here, too, there are promising technological approaches [18], ble future development of the energy consumption of data cen- [54], [55]. ters has a wide range. In the “best case” the energy consump- tion of data centers can remain largely constant. If Moore's VII. SUMMARY AND CONCLUSION Law ends and the performance of the data centers increases This paper discusses the development of the energy con- significantly, annual energy consumption may increase to al- sumption of data centers, and the results of various studies are most 3,000 billion kWh/a (Andrae/Edler 2015 "expected"). presented. Although the various studies presented assume Andrae 2019 and Belkhir/Elmeligi expect an energy consump- greater or smaller increases in energy consumption of data cen- tion of approx. 2000 billion kWh/a for 2030. If the current de- ters in recent years, estimates of both the absolute amount of velopments determined by Borderstep continue, the energy energy consumed and the increases in energy consumption consumption of data centers will double by 2030 compared to differ significantly. today. The forecasts for the development of future energy con- However, the analysis of the development makes one thing sumption of data centers differ even further. The article briefly clear: energy-efficient operation of data centers will continue to presents and discusses possible scenarios for the energy con- be of great importance in the future. However, the challenges sumption of data centers worldwide through 2030, ranging are changing. This will be discussed in the following section. from keeping energy consumption constant to an increase by a factor of 40. VI. FUTURE CHALLENGES: ENERGY-EFFICIENT IT All the uncertainty and variation of the forecasts notwith- HARDWARE AND SOFTWARE, USE OF WASTE HEAT AND standing, a further significant increase in the energy consump- RENEWABLE ENERGIES tion of data centers seems likely. Improving energy efficiency As shown above, the PUE values of data centers are im- will therefore continue to be of great importance. The focus proving continuously. As a result, the share of the data center here will be more on improving the energy efficiency of IT infrastructure in the energy consumption of data centers is be- components in the future, as significant improvements in infra- coming smaller. In the future, efforts to improve the energy structure such as cooling and uninterruptible power supply efficiency of data centers will have to focus even more on IT have already been achieved in the past. Measures that do not components. Further improvements to IT hardware, energy- directly affect the energy efficiency of data centers, such as the efficient software, and efficient software deployment models use of waste heat and operation with (fluctuating) regenerative such as virtualization and container technology offer opportuni- energy, will also become increasingly important in the future. ties for optimization. Completely new technologies such as With a view to the future, it can be assumed that the focus neuromorphic processors or the use of artificial intelligence to of sustainable data center operation will shift from the sole improve efficiency in data centers also offer high potential consideration of energy requirements to other sustainability [18]. categories. Initial approaches, e.g. for the consideration of the Even if all possible future improvements in the efficiency 17 Sustainable Development Goals of the UN, already exist of IT components and infrastructure are taken into account: today [56]. almost everywhere, the electricity used in data centers is con- verted into heat and then released into the environment – most- ACKNOWLEDGMENTS ly using additional energy for ventilation and cooling. The ex- This contribution is a result of the project TEMPRO – Total ample of Sweden in particular shows that it is possible to use Energy Management for professional data centers, which is the waste heat from data centers under appropriate conditions funded by the German Federal Ministry of Economics and En- [52], [53]. In view of the increasing energy consumption of ergy. We would like to thank all project partners and especially data centers, the topic of waste heat utilization will play a cen- project manager Alexandra Pehlken for their constructive and http://www.independent.co.uk/environment/global-warming- excellent cooperation. data-centres-to-consume-three-times-as-much-energy-in-next- decade-experts-warn-a6830086.html. [Accessed: 28-May- 2018] REFERENCES [13] N. Jones, “How to stop data centres from gobbling up the world’s electricity - The energy-efficiency drive at the [1] Cisco, “Cisco Global Cloud Index: Forecast and Methodology information factories that serve us Facebook, Google and 2016-2021,” 2018 [Online]. Available: Bitcoin.,” nature, Sep. 2018 [Online]. Available: https://www.cisco.com/c/en/us/solutions/collateral/service- https://www.nature.com/articles/d41586-018-06610- provider/global-cloud-index-gci/white-paper-c11-738085.pdf. y?WT.ec_id=NATURE- [Accessed: 07-Feb-2018] 20180913&utm_source=nature_etoc&utm_medium [2] G. A. Brady, “Energy Efficiency in Data Centres and the =email&utm_campaign=20180913&spMailingID=57356 Barriers to Further Improvements: An Interdisciplinary 737&spUserID=MTQyOTQwOTY4MDcS1&spJobID=14816 Investigation,” The University of Leeds, Leeds, 2016 [Online]. 95484&spReportId=MTQ4MTY5NTQ4NAS2. [Accessed: Available: 15-Sep-2018] http://etheses.whiterose.ac.uk/12359/1/Gemma%20Brady%20 [14] L. Belkhir and A. Elmeligi, “Assessing ICT global emissions PhD%20Thesis%20(1).pdf. [Accessed: 15-Apr-2018] footprint: Trends to 2040 & recommendations,” J. Clean. [3] A. S. G. Andrae and T. Edler, “On Global Electricity Usage of Prod., vol. 177, pp. 448–463, 2018. Communication Technology: Trends to 2030,” Challenges, [15] K. Fichter and R. Hintemann, “Beyond Energy: Material vol. 6, no. 1, pp. 117–157, Apr. 2015. Stocks in Data Centers, Taking Resource Efficiency into [4] A. Shehabi et al., “United States Data Center Energy Usage account in Green IT Strategies for Data Centers,” J. Ind. Ecol., Report,” Ernest Orlando Lawrence Berkeley National no. im Erscheinen, 2014. Laboratory, Berkeley, CA, LBNL-1005775, 2016 [Online]. [16] R. Hintemann, “Consolidation, Colocation, Virtualization, and Available: https://eta.lbl.gov/sites/all/files/publications/lbnl- Cloud Computing – The Impact of the Changing Structure of 1005775_v2.pdf. [Accessed: 19-Feb-2018] Data Centers on Total Electricity Demand,” in ICT [5] J. Koomey, “Growth in data center electricity use 2005 to Innovations for Sustainability. Advances in Intelligent 2010,” Rep. Anal. Press Complet. Req. N. Y. Times, 2011 Systems and Computing, L. M. Hilty and B. Aebischer, Eds. [Online]. Available: Berlin, Heidelberg: Springer, 2014. http://www.missioncriticalmagazine.com/ext/resources/MC/H [17] R. Hintemann, K. Fichter, and D. Schlitt, “Adaptive ome/Files/PDFs/Koomey_Data_Center.pdf. [Accessed: 22- computing and server virtualization in German data centers - Jul-2014] Potentials for increasing energy efficiency today an in 2020,” [6] A. de Vries, “Bitcoin’s Growing Energy Problem,” Joule, vol. in Proceedings of the 28th Conference on Environmental 2, no. 5, pp. 801–805, 2018. Informatics - Informatics for Environmental Protection, [7] S. Prakash, Y. Baron, L. Ran, M. Proske, and A. Schlösser, Sustainable Development and Risk Management, Oldenburg, “Study on the practical application of the new framework 2014, pp. 477–484 [Online]. Available: methodology for measuring the environmental impact of ICT - http://enviroinfo.eu/sites/default/files/pdfs/vol8514/0477.pdf. cost/benefit analysis,” European Commission, Brussels, [Accessed: 25-Jan-2015] Studie, 2014. [18] R. Hintemann and S. Hinterholzer, “Technology radars for [8] M. Avgerinou, P. Bertoldi, and L. Castellazzi, “Trends in Data energy-efficient data centers: A transdisciplinary approach to Centre Energy Consumption under the European Code of technology identification, analysis and evaluation,” in Conduct for Data Centre Energy Efficiency,” energies, Sep. Sustainable Technologies. World Congress. 2018. (WCST 2017. 2018), Cambridge, 2018 [Online]. Available: [9] R. Hintemann and J. Clausen, “Green Cloud? The current and https://www.researchgate.net/publication/330359801_Technol future development of energy consumption by data centers, ogy_radars_for_energy- networks and end-user devices,” in Proceedings of ICT for efficient_data_centers_A_transdisciplinary_approach_to_tech Sustainability 2016, Amsterdam, The Netherlands, 2016 nology_identification_analysis_and_evaluation. [Accessed: [Online]. Available: http://www.atlantis- 24-Jan-2019] press.com/php/pub.php?publication=ict4s-16. [Accessed: 05- [19] K. Fichter, S. Beucker, J. Clausen, and R. Hintemann, “Green Sep-2016] IT: Zukünftige Herausforderungen und Chancen,” [10] A. Shehabi, S. J. Smith, E. Masanet, and J. G. Koomey, “Data Umweltbundesamt, Dessau, 2009 [Online]. Available: center growth in the United States: decoupling the demand for http://www.umweltbundesamt.de/publikationen/green-it- services from electricity use,” Environ. Res. Lett., vol. 13, no. zukuenftige-herausforderungen-chancen. [Accessed: 05-Jun- 12, 2018 [Online]. Available: 2014] http://iopscience.iop.org/article/10.1088/1748-9326/aaec9c. [20] BITKOM, “Leitfaden Energieeffizienz in Rechenzentren. Ein [Accessed: 02-Jan-2019] Leitfaden zur Planung, zur Modernisierung und zum Betrieb [11] Bio by Deloitte and Fraunhofer IZM, Ecodesign Preparatory von Rechenzentren.,” Berlin, 2008 [Online]. Available: Study on Enterprise Servers and Data Equipment. Brussels, http://www.bitkom.org/de/themen/54930_53432.aspx. 2016 [Online]. Available: [Accessed: 09-Jul-2014] https://publications.europa.eu/en/publication-detail/- [21] BITKOM, “Energieeffizienz-Analysen in Rechenzentren. /publication/6ec8bbe6-b8f7-11e5-8d3c-01aa75ed71a1. Messverfahren und Checkliste zur Durchführung.,” [Accessed: 22-Jan-2018] Bundesverband Informationswirtschaft, Telekommunikation [12] T. Bawden, “Global warming: Data centres to consume three und neue Medien e.V (BITKOM), Berlin, Leitfaden 3, 2008 times as much energy in next decade, experts warn,” The [Online]. Available: Independent, Jan. 2016 [Online]. Available: http://www.bitkom.org/files/documents/Energieeffizienz- [34] R. Hintemann and J. Clausen, “Potenzial von Analysen_in_RZ_web.pdf. [Accessed: 09-Jul-2014] Energieeffizienztechnologien bei Colocation Rechenzentren in [22] R. Hintemann and K. Fichter, “Energy consumption and Hessen,” Borderstep Institut für Innovation und quantities of materials in German data centers,” in Electronics Nachhaltigkeit, Berlin, 2018 [Online]. Available: Goes Green 2012+ (EGG), 2012, 2012, pp. 1–6. https://www.digitalstrategie-hessen.de/rechenzentren. [23] R. Hintemann, “Energy Consumption of Data Centers in [Accessed: 26-Apr-2018] 2014,” Borderstep Institut für Innovation und Nachhaltigkeit, [35] M. Howard-Healy, “Co-location Market Quarterly (CMQ) Berlin, 2015. brief - Vortrag auf dem BroadGroup’s Knowledge Brunch in [24] R. Hintemann, “Energy consumption of data centers continues Frankfurt,” Broadgroup, 2018. to increase – 2015 update,” Borderstep Institut für Innovation [36] Techconsult, “Daten des eanalyzers,” 2014. [Online]. und Nachhaltigkeit, Berlin, 2015 [Online]. Available: Available: www.eanalyzer.biz www.borderstep.de/publikationen [37] Techconsult, “Daten des eanalyzers,” 2015. [Online]. [25] R. Hintemann, S. Beucker, J. Clausen, L. Stobbe, M. Proske, Available: www.eanalyzer.biz and N. F. Nissen, “Energy efficiency of data centers - A [38] Techconsult, “Daten des eanalyzers,” 2016. [Online]. system-oriented analysis of current development trends,” in Available: www.eanalyzer.biz EGG 2016+ Proceedings, Berlin, 2016. [39] EITO/IDC, “EITO Costumized Report for Borderstep.,” [26] L. Stobbe, R. Hintemann, M. Proske, J. Clausen, H. Zedel, and EITO/IDC, Berlin, 2014. S. Beucker, “Entwicklung des IKT-bedingten Strombedarfs in [40] IDC, “Server Market and Enterprise Storage Systems By Deutschland - Studie im Auftrag des Bundesministeriums für Country 2014-2017,” 2018. Wirtschaft und Energie,” Fraunhofer IZM und Borderstep [41] R. Hintemann, K. Fichter, and L. Stobbe, “Materialbestand der Institut, Berlin, 2015 [Online]. Available: Rechenzentren in Deutschland-Eine Bestandsaufnahme zur http://www.bmwi.de/BMWi/Redaktion/PDF/E/entwicklung- Ermittlung von Ressourcen-und Energieeinsatz,” Stud. Im des-ikt-bedingten-strombedarfs-in-deutschland- Rahm. UFO-Plan-Vorh. “Produktbezogene Ansätze Inf.- abschlussbericht,property=pdf,bereich=bmwi2012,sprache=de Kommun. “Förderkennzeichen 370 893 302 Beauftr. Vom ,rwb=true.pdf Umweltbundesamt, 2010. [27] R. Hintemann, “Boom führt zu deutlich steigendem [42] J. Pohl and M. Finkbeiner, “Digitalisation for sustainability? Energiebedarf der Rechenzentren in Deutschland im Jahr Challenges in environmental assessment of digital services,” 2017,” Borderstep Institut für Innovation und Nachhaltigkeit, Inform. 2017, 2017. Berlin, 2018. [43] J. Malmodin and D. Lundén, “The energy and carbon footprint [28] CBRE, “Nachgefragte Leistung europäischer Rechenzentren of the ICT and E&M sector in Sweden 1990-2015 and übersteigt erneut 100 MW-Wert,” Nachgefragte Leistung beyond,” Pap. Publ. Present. ICT Sustain. ICT4S Amst. Neth., europäischer Rechenzentren übersteigt erneut 100 MW-Wert, pp. 30–31, 2016. 21-Mar-2018. [Online]. Available: [44] J. Malmodin and D. Lundén, “The Energy and Carbon http://news.cbre.de/nachgefragte-leistung-europaischer- Footprint of the Global ICT and E&M Sectors 2010–2015,” rechenzentren-ubersteigt-erneut-100-mw-wert. [Accessed: 04- Sustainability, vol. 10, no. 9, p. 3027, 2018. Apr-2018] [45] A. S. G. Andrae, “Projecting the chiaroscuro of the electricity [29] CBRE Global Corporate Services, “European Data Centres use of communication and computing from 2018 to 2030,” Market Review. Q4 2016,” London, 2017 [Online]. Available: 2019. https://www.cbre.de/de-de/research/European-Data-Centres- [46] “Vereecken: Energy efficiency in thin client solutions - MarketView-Q4-2016. [Accessed: 10-Jun-2017] Google Scholar.” [Online]. Available: [30] Cisco, “Cisco Global Cloud Index: Forecast and Methodology https://scholar.google.com/scholar_lookup?title=Energy%20ef 2014-2019,” 2015 [Online]. Available: ficiency%20in%20thin%20client%20solutions&author=W.%2 http://www.cisco.com/c/en/us/solutions/collateral/service- 0Vereecken&publication_year=2009. [Accessed: 15-Apr- provider/global-cloud-index- 2019] gci/Cloud_Index_White_Paper.pdf [47] Technavio, “Global Data Center Market 2015-2019,” 2015. [31] Cisco, “Cisco Global Cloud Index: Forecast and Methodology [48] W. Van Heddeghem, S. Lambert, B. Lannoo, D. Colle, M. 2015-2020,” 2016 [Online]. Available: Pickavet, and P. Demeester, “Trends in worldwide ICT https://www.cisco.com/c/dam/en/us/solutions/collateral/servic electricity consumption from 2007 to 2012,” Comput. e-provider/global-cloud-index-gci/white-paper-c11- Commun., vol. 50, pp. 64–76, 2014. 738085.pdf. [Accessed: 10-Oct-2017] [49] Digiconomist, “Bitcoin Energy Consumption Index,” [32] R. Hintemann, “Energieeffizienz und Rechenzentren in Digiconomist, 2019. [Online]. Available: Deutschland – weltweit führend oder längst abgehängt?,” https://digiconomist.net/bitcoin-energy-consumption. Netzwerk energieeffiziente Rechenzentren - NeRZ, Berlin, [Accessed: 12-Apr-2019] 2017 [Online]. Available: https://www.borderstep.de/wp- [50] M. Peckham, “The Collapse of Moore’s Law: Physicist Says content/uploads/2017/07/NeRZ-Studie-Rechenzentrumsmarkt- It’s Already Happening,” Time [Online]. Available: 30-06-2017.pdf. [Accessed: 25-Oct-2017] http://techland.time.com/2012/05/01/the-collapse-of-moores- [33] R. Hintemann and J. Clausen, “Bedeutung digitaler law-physicist-says-its-already-happening/. [Accessed: 15-Apr- Infrastrukturen in Deutschland. Sozioökonomische Chancen 2019] und Herausforderungen für Rechenzentren im internationalen [51] Cisco, “Cisco Global Cloud Index: Forecast and Methodology Wettbewerb. Berlin. Verfügbar unter,” Berlin, 2018 [Online]. 2013-2018,” 2014. Available: https://www.eco.de/wp- [52] U. Ostler, “Server-Hitze für 10.000 warme Stuben - Ein content/uploads/dlm_uploads/2018/06/DI_Studie.pdf. weiteres Datacenter mit Anschluss ans Fernwärmenetz – in [Accessed: 14-Jun-2018] Stockholm,” DataCenter Insider, 18-Sep-2017 [Online]. Available: https://www.datacenter-insider.de/ein-weiteres- datacenter-mit-anschluss-ans-fernwaermenetz-in-stockholm-a- fuehrt-den-rechenzentrumsbetrieb-in-eine-klimaneutrale- 644074/. [Accessed: 10-Oct-2017] zukunft-a-778745/. [Accessed: 13-Dec-2018] [53] S. Reveman and U. Ostler, “Die Energie in deutschen [55] A. Rüdiger and U. Ostler, “Das erste Rechenzentrum in einer Datacenter verpufft zu 100% - Rechenzentren jagen das Geld Windmühle läuft und die Kundschaft ist happy,” 14-Sep-2018. zum Fenster raus,” DataCenter Insider, 12-Jul-2016 [Online]. [Online]. Available: https://www.datacenter-insider.de/das- Available: http://www.datacenter-insider.de/die-energie-in- erste-rechenzentrum-in-einer-windmuehle-laeuft-und-die- deutschen-datacenter-verpufft-zu-100-a-541729/. [Accessed: kundschaft-ist-happy-a-752383/. [Accessed: 13-Dec-2018] 12-Jul-2016] [56] R. Hintemann, “Nachhaltige Rechenzentren: Viel mehr als nur [54] U. Ostler, “Windcloud 4.0 führt den Rechenzentrumsbetrieb in Energieeffizienz,” Funkschau, Nov. 2018 [Online]. Available: eine klimaneutrale Zukunft,” 18-Nov-2018. [Online]. https://www.funkschau.de/datacenter/artikel/159905/. Available: https://www.datacenter-insider.de/windcloud-40- [Accessed: 04-Jan-2019]