ArticlePDF Available

Cooling Energy Consumption Investigation of Data Center IT Room with Vertical Placed Server

Authors:

Abstract and Figures

As energy consumption by cooling data center IT equipment can be over 40% of total energy consumption, efficient cooling for large data centers is essential for reducing operation costs. Modern data centers are complex systems involving IT facilities, power system, cooling and ventilation systems. In our previous work, literature study was made to investigate available data center energy consumption models; and energy consumption models for data center IT room with distributed air flow control were developed. In this paper, the models are further extended and developed to cover the combined distributed air flow control and vertical placed servers in raised floor ventilation system. Simulation of the three types of ventilation systems with Even load, Idle server and Uneven load scenarios showed that significant cooling energy consumed by a traditional ventilation system can be saved by applying the proposed new concept and method.
Content may be subject to copyright.
1876-6102 © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the scientific committee of the 8th International Conference on Applied Energy.
doi: 10.1016/j.egypro.2017.03.581
Energy Procedia 105 ( 2017 ) 2047 2052
ScienceDirect
The 8
th
International Conference on Applied Energy – ICAE2016
Cooling Energy Consumption Investigation of Data Center IT
Room with Vertical Placed Server
X. Zhang
a
*
, T. Lindberg
b
, N. Xiong
c
, V. Vyatkin
d,e
, A. Mousavi
d
a
Department of Power Device, ABB AB, Corporate Research, Vasteras, Sweden
b
Department of Engineering and Physics, Karlstad University, Sweden
c
School of Innovation, Design and Engineering, Malardalen University, Vasteras, Sweden
d
Department of Computer Science, Electrical and Space Engineering, Lulea University of Technology, Sweden
e
Aalto University, Helsinki, Finland
Abstract
As energy consumption by cooling data center IT equipment can be over 40 % of total energy consumption, efficient
cooling for large data centers is essential for reducing operation costs. Modern data centers are complex systems
involving IT facilities, power system, cooling and ventilation systems. In our previous work, literature study was made
to investigate available data center energy consumption models; and energy consumption models for data center IT
room with distributed air flow control were developed. In this paper, the models are further extended and developed to
cover the combined distributed air flow control and vertical placed servers in raised floor ventilation system. Simulation
of the three types of ventilation systems with Even load,Idle server and Uneven load scenarios showed that significant
cooling energy consumed by a traditional ventilation system can be saved by applying the proposed new concept and
method.
© 2016 The Authors. Published by Elsevier Ltd.
Selection and/or peer-review under responsibility of ICAE
Keywords: data center; server rack; air flow; energy efficiency
1. Introduction
With rapid growth of large data centers worldwide, data centers become energy intensive processes
accounting for over 1% of the world’s electricity usage [1]. Large data centers with capacity up to 120 MW
have been built in recent years. Energy efficiency becomes even more important for these data centers. 
ǤInvestigation showed that
energy consumption by cooling data center IT equipment is between 30% and 55% of the total energy
* Xiaojing Zhang. Tel.: +46 21 323062
E-mail address: xiaojing.zhang@se.abb.com
Available online at www.sciencedirect.com
© 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the scientific committee of the 8th International Conference on Applied Energy.
2048 X. Zhang et al. / Energy Procedia 105 ( 2017 ) 2047 – 2052
consumption. Cooling and ventilation system consumes on average 40% of the total energy consumption
in a data center [2].
A comprehensive data center power consumption model describing IT room, computer room air
handling (CRAH), data center ventilation and cooling characteristics was developed to cover traditional
raised floor ventilation system and the distributed air flow system [3]. The distributed air flow system
showed advantages when IT load or cooling load varies for each server rack. The simulation results showed
that up to 16% of power consumption reduction can be achieved by applying the distributed air flow control
for the data centers with idle server and uneven IT load. Literature study of energy consumption models for
different subsystem components was also made in our previous work with intention of using models to
build a powerful simulation tool and to estimate and simulate different operation scenarios as decision
support for data center design and operation. This paper focuses on modelling air flow with the vertical
placed server and evaluates operation benefits through simulation and comparison between traditional
designs of air flow and the proposed scheme of distributed air flow system combined with vertical place
server.
2. Data center distributed air flow cooling with vertical placed server
In the conventional data center with hot aisle/cold aisle, cold air generated by the cooling system is
supplied through a plenum under the floor and perforated air flow panels. The cold air flows up horizontally
entering the tiny spaces between the servers from one side of the servers and leaving from another side.
Higher flow pressure drop, cold and warm air mixture on the upper side of racks are the main disadvantages.
Most data centers operate under varying and uneven IT loads resulting in different cooling loads over
different racks. Hot spots cannot be avoided when cold air is uniformly supplied to all server racks. To meet
the cooling needs of individual server racks without local overcooling, it requires that cold air being
distributed on-demand across local cooling loads. Figure 1 shows a data center with uneven server
utilization. The concept of distributed air flow control is to divide a data center in pre-defined zones and
different amounts of air flow are supplied to these zones across the data center based on local cooling loads.
It requires special ventilation system design under the raised floor to distribute cooling air through
ventilation ducks and dampers. By this solution, the hot spots and overcooling can be reduced or even
eliminated. A significant energy saving can also be achieved.
Fig. 1. Data center with uneven server utilization
In the vertical placed sever concept, the traditional hot aisle/cold aisle air supply showed in Figure 2 is
replaced by air supply directly from the bottom of each server rack and it is illustrated in Figure 3. The
concept was described in detail in our previous work [2][3]. The distributed air flow control can be applied
in a traditional raised-floor data center or air flow control system with or without vertically placed servers.
X. Zhang et al. / Energy Procedia 105 ( 2017 ) 2047 – 2052 2049
Fig. 2. Traditional hot aisle/cold aisle air supply.
Fig. 3. Air supply with vertical placed server rack.
3. Model for the vertical placed server rack
The data center power consumption models previous developed covered the power consumptions of
servers, racks, CRAH, chiller, cooling tower, UPS and PDU. Comparing to the traditional hot aisle/cold
aisle ventilation system, models to calculate system pressure drop and flow rates are required due to changes
of ventilation system setup. CRAH fan power consumption is affected by ventilation system pressure drop.
For standard type of racks, the pressure drop over the CRAH units was earlier developed [4][5]. In a
standard server rack, servers are stacked on top of each other in parallel. The pressure drops over front
doors, rear doors and servers contribute to the change in pressure over a server rack. For the vertical placed
server rack showed in Fig. 3, the pressure drop models are further developed in this study. The setup consists
of two separate compartments with individual air supply. Each compartment has three rows of racks. Air
streams flow up and warm air is dissipated to hot aisles on both sides between server rows. Since the air
flow enters the bottom of the compartments, the flow rate will be reduced for each row further up. Because
of this change in flow rate, the pressure drop needs to be determined for each row. With three-row and two-
compartment (lower and upper in Fig. 3) design, pressure drops can be described by equations (1) to (6),
οܲ௟௢௪௘௥ǡଵൌܥ௦௘௥௩௘௥
ܸ௦௘௥௩௘௥
ൌܥ௦௘௥௩௘௥
ಿೞ೐ೝೡ೐ೝೝೌ೎ೖ
ಿೝ೚ೢೝೌ೎ೖ
(1)
οܲ௟௢௪௘௥ǡଶൌܥ௦௘௥௩௘௥
ಿೞ೐ೝೡ೐ೝೝೌ೎ೖ
ಿೝ೚ೢೝೌ೎ೖ
(2)
2050 X. Zhang et al. / Energy Procedia 105 ( 2017 ) 2047 – 2052
οܲ௟௢௪௘௥ǡଷൌܥ௦௘௥௩௘௥
ಿೞ೐ೝೡ೐ೝ̴ೝೌ೎ೖ
ಿೝ೚ೢ̴ೝೌ೎ೖ ൅ܥ௕௘௡ௗ
ܸ (3)
οܲ௨௣௣௘௥ǡଵൌܥ௦௘௥௩௘௥
ಿೞ೐ೝೡ೐ೝೝೌ೎ೖ
ಿೝ೚ೢೝೌ೎ೖ ൅ܥௗ௜௦௧
ܸ (4)
οܲ௨௣௣௘௥ǡଶൌܥ௦௘௥௩௘௥
ಿೞ೐ೝೡ೐ೝೝೌ೎ೖ
ಿೝ೚ೢೝೌ೎ೖ
(5)
οܲ௨௣௣௘௥ǡଷൌܥ௦௘௥௩௘௥
ಿೞ೐ೝೡ೐ೝೝೌ೎ೖ
ಿೝ೚ೢೝೌ೎ೖ  (6)
where ܸ1, ܸʹǡ
ܸ͵ are the flow rates through the first, second and third rows of each ܸcompartment,
respectively. Cserver is coefficient for server pressure lossǤܸ௦௘௥௩௘௥ is the flow arte through the servers in the
first compartment. Cbend is coefficient for the pressure loss caused by the roof of the lower compartment,
which forces the flow to bend off its path and leave through the sides. Cdist is the pressure loss coefficient
of the distributor that distributes the air flow equally in the bottom of the upper compartment. Nserver_rack is
the number of servers in each rack. Nrow_rack is the number of racks in each row. In traditional server racks,
supply air flow rate from the CRAH units is based on the racks with the maximum heat load. For zones
with lower utilization, the air bypass increases since the servers need less cooling. Track_in or Track_out can
be used as the set-point parameter, the mixed flow of rack flow and air bypass can be expressed as,
ܶ௥௔௖௞̴௜௡̴௬ ூି௃௏ೝೌ೎ೖ̴೤ೝೌ೎ೖ̴೚ೠ೟̴ೞ೐೟
ଵି௃௏ೝೌ೎ೖା௄ (7)
ܫൌ߳஼ோ஺ுಹಶܥ௠௜௡ܶ஼ோ஺ு೔೙
߳஼ோ஺ுಹಶܥ௠௜௡െܥ௔௜௥
ܬൌ ܰ௥௔௖௞
ܰ஼ோ஺ு൫ͳߜ௟௘௔௞௔௚௘൯ܸ஼ோ஺ு
ܭൌ ܥ௔௜௥
߳஼ோ஺ு̴ுாܥ௠௜௡െܥ௔௜௥
where ߳஼ோ஺ுಹಶ is heat exchanger effectiveness, Cmin and Cair are the lower heat capacity rate and the heat
capacity rate of the air flow respectively,
ܶ஼ோ஺ு೔೙is the temperature of the BCW water entering the
CRAH heat exchanger and ߜ௟௘௔௞௔௚௘ is leakage flow fraction. The server fans are assumed to keep the rack
outlet temperature constant at the set-point value.
ܸ௥௔௖௞ ೝೌ೎ೖ
ೌ೔ೝ೛ǡೌ೔ೝೝೌ೎ೖ̴೚ೠ೟ି்ೝೌ೎ೖ̴೔೙ (8)
The rack flow rate for vertical server racks can be determined directly from the CRAH flow rate since
these supply just the amount of air needed by the racks.
ܸ஼ோ஺ு̴௬ ೝೌ೎ೖೝೌ೎ೖ̴೤
಴ೃಲಹଵିఋ೗೐ೌೖೌ೒೐ (9)
X. Zhang et al. / Energy Procedia 105 ( 2017 ) 2047 – 2052 2051
where ܳ௥௔௖௞ is the heat load of one rack. ߩ௔௜௥ and ܥ݌ǡ௔௜௥ are density and the specific heat of the air.
Nrack and NCRAH are the number of racks and CRAHs.
4. Simulation and Results
A case study of a data center with 208 racks is simulated. Each CRAH has capacity of 452 kW. Cooling
design is with redundancy of N+1. All servers were assumed to be identical, making heat load and cooling
load differences dependent only on the different server utilizations. 13 zones are defined and each rack
contains 42 servers. As the same as our previous distributed air flow investigation, the simulation was made
in MATLAB and covered three scenarios, Even load,Idle server and Uneven load. Both racks and CRAH
units are assumed to operate on the principle of keeping a constant level for the outlet temperature of the
topmost row in a rack compartment, since this is the hottest and thus critical temperature measured with
available sensors. Fig. 4 shows the effect of the rack outlet temperature on the total cooling power
consumption for all three scenarios and with three air supply systems, namely traditional air supply, the
distributed air flow control and the distributed air flow control with vertical placed server rack. When the
server IT load are evenly distributed as shown in Fig. 4 (a), there is no benefit to apply only the distributed
air flow control. From Fig. 4 (b) and 4 (c), it is evident that the distributed air flow control saves energy
and the combined distributed air flow control with vertical placed server rack saves even more energy,
which is mainly ǤBecause cooling air is supplied on demand for
individual zones; cooling air bypass flow is eliminated and vertical cooling air flow has much lower
pressure drop than traditional air supply. Owing to the vertical placed server with vertical air flow through
servers it is favourable to heat transfer with ǡ
Ǥ
Ǥ
Fig. 4. Total cooling power consumption as a function of the rack outlet temperature for traditional air supply, distributed air flow
control and distributed air flow control with vertical placed server; (a) Even load, (b) Idle servers, (c) Uneven load
2052 X. Zhang et al. / Energy Procedia 105 ( 2017 ) 2047 – 2052
Fig. 5 shows the total cooling power consumption for three ventilation systems and three scenarios, Even
load,Idle server and Uneven load with rack outlet temperature set point of 37°C. A base sever utilization
of 0.85 is applied for all three scenarios. In Idle server and Uneven load, 9 and 7 out of the 13 zones have
the basic utilization respectively. The rest of the servers are either idle or with defined utilizations [4].
Depending on system setup, over 29% of cooling energy can be saved with the proposed solution. The
current study is one of the sub-work packages in our modelling and simulation effort. It is limited with
available data center operation measurement and data. There is on-going project for Swedish national data
center test facility in North of Sweden. The next phase of development is to test and verify these results in
this facility.
Fig. 5. Total cooling power consumption for all scenarios.
5. Conclusion
This paper has presented simulation results for the distributed air flow control with the vertical placed
server in comparison with traditional data center ventilation and air flow design. Previous developed data
center energy consumption models have been extended to cover the vertical placed server design. The
distributed air flow control with vertical servers showed the lowest power consumption in all scenarios.
References
[1] J. G. Koomey. Growth in data center electricity use 2005 to 2010.Analytics Press, Oakland, 2011.
[2] Z. Song, X. Zhang, C. Eriksson, Data center energy and cost saving evaluation. The 7th International Confernece on Apllied
Energy, Taipei, Taiwan, 2015.
[3] X. Zhang, T. Lindberg, K. Svensson, V. Vyatkin, A. Mousavi, 7th International Confernece on Computer Modeling and
Simulation, Brisbane, Australia, 2016.
[4] T. Van Giang, D. Vincent, and B. Seddik. Data center energy consumption simulator from the servers to their cooling system.
In PowerTech (POW- ERTECH), 2013 IEEE Grenoble. IEEE, June 2013.
[5] M. Iyengar and R. Schmidt. Analytical modeling for thermodynamic characterization of data center cooling systems. Journal
of Electronic Packaging, 101, June 2009.
Xiaojing Zhang received Ph.D. degree in energy technology in 1997 at Royal Institute of
Technology, Stockholm, Sweden. He joined ABB in 1998 and his career included 18 years of
industrial experiences in process industry and datacenter infrastructure management. Dr. Zhang is
currently Principal Scientist at ABB Corporate Research, Sweden and Adjunct Professor at Lulea
University of Technology, Sweden.
... In a conventional data center cooling cycle, chillers produce chilled water for the computer room air conditioners (CRAC), and the exhaust hot air from servers passes through the cooling coils in CRACs and transfers its heat to the mean fluid. These conventional systems consume about 40% of the total energy of data centers [8,9]. This has, therefore, attracted the attention of researchers to propose effective cooling strategies that can lead to significant energy saving in data centers. ...
... For building a seawater cooling system, knowing this fact is quite crucial since it determines the amount of provided cooling. As it was mentioned, a correlation between the data center supplied air temperature, and the seawater temperature can be established to provide sufficient cooling, which is shown in Eq. (9). Considering different depths for implementing the pipes for the inlet seawater and other affecting parameters (e.g, heat exchanger efficiency, water flow rate, etc.), the supplied air temperature to the data center can be obtained. ...
Article
Direct seawater cooling is a clever, passive, and sustainable method for large-scale free cold supply. This method has been previously applied for limited applications, e.g., district cooling systems. One of the continuously increasing demand kinds of the world energy systems is associated with the cooling of data centers. In this study, the feasibility and the optimal design of seawater cooling systems for supplying the cooling demand of data centers are investigated. For this, the thermodynamic model of the data center and the seawater cooling infrastructure is developed, the optimal sizing of the components over an entire year is found, and the optimal operating conditions are determined. All these are carried out for a case study data center to be integrated with a cooling system fed by the Caspian Sea. It was found that a seawater cooling system, deployed at 700 m depth of the Caspian Sea can provide up to 78% energy saving, comparing to the conventional cooling system. This led to a 45% saving in annual electricity consumption and CO2e emission. Besides, the results of the optimization process indicated that a 60% increase in initial costs of the cooling system will result in 77% lower required pumping power.
... e air cooling system of the data center uses air circulation and air conditioning technology to remove the heat generated by the data center. Zhang et al. expanded and developed the distributed airflow control model of the data center energy consumption model and applied to the distributed airflow control of the raised floor ventilation system and the vertical placed server, and it could reduce the cooling energy consumption of the traditional ventilation system significantly [1]. Yang et al. applied the free cooling mode to the data center air conditioning system to reduce the energy consumption [2], Ham et al. proposed a server model that could simulate the cooling energy consumption of the data center to determine the temperature difference between the server heat production, and it was proposed that when the supply air temperature of the computer room air handler is higher than 19°C, and the cooling energy consumption increases due to the increase of fan energy consumption [3]. ...
Article
Full-text available
Energy consumption is the key factor leading to high operating costs in data center. In this paper, considering the data center service level under the premise of using electricity and gas as energy supply of energy supply and energy consumption from two angles, the data center energy scheduling model is built. The lower model is calculation scheduling model for the data center, and the upper model is data center energy supply scheduling model. Then the particle swarm algorithm is used to simulate the schedule. The results show that using natural gas as supplementary energy supply can effectively reduce the overall energy consumption of data center in considering the service delay level of data center.
... It should be noted that this estimation does not include the carbon emissions of the air conditioning required to cool the server room. Air conditioning can be a significant source of carbon emissions, sometimes exceeding 40% of the total power consumption in typical data centers (Zhang et al., 2017). We did not have a way to measure the energy consumption of the air conditioning system, but we do not expect this to change the order of magnitude of the total emissions. ...
Thesis
Music enhances moving images and allows to efficiently communicate emotion or narrative tension, thanks to cultural codes common to the filmmakers and viewers. A successful communication requires not only a choice of track matching the video's mood and content, but also a temporal synchronization of the audio and visual main events. This is the goal of the music supervision industry, which traditionally carries out the task manually. In this dissertation, we study the automation of tasks related to music supervision. The music supervision problem generally doesn't have a unique solution, as it includes external constraints such as the client's identity or budget. It is thus relevant to proceed by recommendation. As the number of available musical videos is in constant augmentation, it makes sense to use data-driven tools. More precisely, we use the metric learning paradigm to learn the relevant projections of multimodal (video and music) data. First, we address the music similarity problem, which is used to broaden the results of a music search. We implement an efficient content-based imitation of a tag-based similarity metric. To do so, we present a method to train a convolutional neural network from ranked lists. Then, we focus on direct, content-based music recommendation for video. We adapt a simple self-supervised system and we demonstrate a way to improve its performance, by using pretrained audio features and learning their aggregation. We then carry a qualitative and quantitative analysis of official music videos to better understand the temporal organization of musical videos. Results show that official music videos are carefully edited in order to align audio and video events, and that the level of synchronization depends on the music and video genres. With this insight, we propose the first recommendation system designed specifically for music supervision: the Seg-VM-Net, which uses both content and structure to perform the matching of music and video.
... The cost of this power consumption has made information technology (IT) companies, research centers, and scholars propose suitable methods to optimize power consumption. By investigating data center power consumption, it has been Hui shown that cooling has a large part in data center power consumption, which could be over 40% of the total power consumption [2] . This amount of energy is rising due to lousy airflow management, such as ineffective air circulation and the crossing of cold-hot air in hot aisles, which could lead to the creation of hotspots that lead to power consumption increase in computing. ...
Article
Full-text available
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.
... Liu [23] introduced a data center equipped with a modular air conditioning system that can change the fan air volume according to the server load change to reduce energy consumption. Zhang et al. [24] established a data center energy consumption model with distributed air flow control that can significantly reduce the cooling energy consumption of a traditional data center ventilation system. ...
Article
To improve the thermal management of a data center, fin-type water-cooled heat sinks were applied to cool the chips in a server cabinet, and a water-cooled system based on the cooling tower of a data center with a thermal power of 4.8 kW was applied. TRNSYS was used to conduct an effect analysis of the operating condition parameters of the cooling water on the power consumption of the system. Finally, the operating conditions of the cooling water were optimized by minimizing the energy consumption of the water-cooled system. In addition, the factors of different safety chip temperatures and partial thermal loads were considered in optimizing the working conditions and analyzing cooling energy consumption. The results show that the optimal cooling water operating parameters significantly differ corresponding to different safety chip temperatures and thermal loads. Finally, the fitting correlation equations of the optimal flow rate of the primary and secondary cooling water under different safe chip temperatures and partial thermal loads are provided to guide the cooling system design. Considering an ambient temperature of 15 °C and a safety chip temperature of 70 °C as a case study, the optimal primary and secondary cooling water flow rates of 3 and 3.4 L/min, respectively, were obtained, with a minimum power usage effectiveness of 1.146.
Article
Evaporation studies of water using classical molecular dynamics simulations are largely limited due to their high computational expense. This study addresses that issue by developing coarse-grained molecular dynamics models based on Morse potential. Models are optimized based on multi-temperature and at room temperature using machine learning techniques like Genetic Algorithm, Nelder-Mead algorithm, and Strength Pareto Evolutionary Algorithm. The multi-temperature-based model named as Morse-D is found to be more accurate than the single temperature model in representing the water properties at higher temperatures. Using this Morse-D water model, evaporation from hydrophilic nanopores with pore diameter varying from 2 to 5 nm is studied. Our results show that the critical diameter to initiate continuous evaporation at nanopores lies between 3 and 4 nm. A maximum heat flux of 21.3 kW/cm2 is observed for a pore diameter of 4.5 nm and a maximum mass flow rate of 16.2 ng/s for a pore diameter of 5 nm. The observed heat flux is an order of magnitude times larger than the currently reported values from experiments in the literature for water, which indicates that we need to focus on nanoscale evaporation to enhance the critical heat flux.
Article
To develop a highly efficient and energy-saving cooling system in a data center, this study investigates the optimal thermal management both on the internal server cooling cycle and external cold source water cycle under different ambient temperatures based on the principle of minimal power consumption. Taking a server cabinet of 4.8 kW as the case, an internal indirect water-cooled heat sink cycle and an external cooling tower cold source cycle are constructed. A hybrid thermal management model combined with a heat dissipation model and a power consumption model was developed with the aid of MATLAB and TRNSYS. Using this model, the relationship between the chip dissipation performance attached to the fin-type heat sink and the energy consumption performance of the cooling system under different water working conditions and ambient temperatures was analyzed. Finally, to achieve minimal system energy consumption, the optimized inlet temperature and water flow rates corresponding to different ambient temperatures were obtained by fitting expressions that can better serve the optimal design of data center cooling systems. Furthermore, when the optimal value was applied using the year-round temperature of Tianjin, China in the case study, a 21.3% reduction in power consumption and a 4.43% decrease in the power usage effectiveness can be achieved compared to the case without optimization. Therefore, significant power can be saved in a data center cooling system through thermal management optimization.
Article
This study aimed to develop an adaptive artificial neural network model (AAM) for the prediction of the rack inlet temperature and cooling system energy, and the optimal control algorithm for cooling system of a containment-type data center. A cyber-physical system (CPS) framework, that incorporated the AAM and control algorithm, was also proposed for the precise control of the data center cooling system. To develop the AAM model and control algorithm, mathematical modeling of a reference physical model was conducted, and training data were acquired from this model. The performance of the proposed AAM and control algorithm was then compared with that of a non-adaptive ANN model (NAAM) in terms of prediction accuracy and control stability. The analysis results indicated that the optimal control algorithm with the AAM exhibited superior prediction accuracy and control stability than the algorithm with the NAAM. In particular, for the AAM-based algorithm under conditions representing a novel data center environment, the root mean square error (RMSE) and coefficient of variation of the RMSE (CV(RMSE)) for the predicted and actual values were 0.22 °C and 1.02%, respectively, for the inlet rack temperature and 0.19 kW and 0.76% for the cooling system energy. The control was also stable, with an MAE of 0.08 °C and a maximum error of 1.17 °C. Based on this analysis, a CPS-based control strategy incorporating an ANN-based optimal control algorithm is expected to be an effective energy efficiency solution for existing data center without changing IT equipment or cooling systems.
Conference Paper
In this paper we address the problem of modeling and estimating the heat reuse potential of liquid cooled Data Centers (DCs) integrated with local heat distribution system infrastructures and utilities. We provide an abstract mathematical model based on a system-of-systems methodology for representing the heat exchanges between IT equipment and thermal tanks. The behavior of the identified subsystems is modelled mathematically based on thermodynamic equations and physical processes involved in heat exchanges. The models are evaluated on an experimental platform consisting of two liquid-cooled processing units that heat an insulated recipient equipped with temperature sensors, showing a temperature prediction MAPE of 7%. Furthermore, a set of use cases evaluating the heat reuse capability in a residential setup as well as an interior pool heating are presented, showing high potential of heat recycling from IT equipment with liquid cooling systems and the possibility of heating a home with 9 processing units of 210 W power or a pool of size 30 m^3 with a set of 100 processing units of 210 W.
Article
This elaboration presents the configuration of the cooling system of the POLCOM Data Center which utilizes commercially available components, proposes steering strategy and analyzes their functionality. The designed architecture of cooling and dedicated control system is presented to demonstrate the novelty and customization with respect to the demanding outdoor temperatures existing in the climate of the Małopolska Province. The cooling system control and operation is illustrated by a comprehensive analysis of compressor and freecooling modes. The discussion on embedded functionality of chillers is undertaken. According to features and limitations of the complex solution, the cooling system reached the annual average coefficient of performance of 8.63 in 2015 (increase of 110% compared to the year 2014), operating 65% of the time during the year in compressor mode and 35% in freecooling mode. This coefficient in compressor mode amounted to 4.39 while in freecooling mode totaled 16.50. It was proved that in the real case under consideration they generated losses in electricity consumption amounting to 557MWh per year. The real-time experimental data collected from the commercial Data Center installation are used to present a unique operation of a such complex system.
Article
Full-text available
In data centers, about 40% of the total energy is consumed for cooling the IT equipment. Cooling costs are thus one of the major contributors to the total electricity bill of large data centers. This paper studies two factors affecting data center cooling energy consumption, namely air flow management and data center location selection. A unique rack layout with a vertically cooling air flow is proposed. Two cooling systems, computer room air conditioning (CRAC) cooling system and airside economizer (ASE), have been studied. Based on these two cooling systems, four cities have been selected from the worldwide data center locations. A number of energy efficiency metrics are explored for data center cooling, such as power usage effectiveness (PUE), coefficient of performance (COP) and chiller hours. By analyzing the effects of chiller hours and economizer hours, comparative economic results of cooling power consumption are provided in both systems. The results show that the cooling efficiency and operating costs vary significantly with different climate conditions, energy prices and cooling technologies. As climate condition is the major factor which affects the airside economizer, employing the airside economizer in the cold climate yields much lower energy consumption and operation costs.
Conference Paper
Efficient energy use has become a worldwide issue for designing and managing the datacenters. Behavioral modeling is a massive task and essential for researching and building our datacenter energy management system (part of a project called Energetic-FUI, France). This paper presents a general expression in the development of our Datacenter Workload Energy Simulation tool (DCWES) using Matlab/Simulink. All modules of the DCWES tool have been converted to Modelica and Java formats to integrate into the Energetic-FUI software.
Article
The increasingly ubiquitous nature of computer and internet usage in our society has driven advances in semiconductor technology, server packaging, and cluster level optimizations in the IT industry. Not surprisingly this has an impact on our societal infrastructure with respect to providing the requisite energy to fuel these power hungry machines. Cooling has been found to contribute about a third of the total data center energy consumption and is the focus of this study. In this paper we develop and present physics based models to allow the prediction of the energy consumption and heat transfer phenomenon in a data center. These models allow the estimation of the microprocessor junction and server inlet air temperatures for different flows and temperature conditions at various parts of the data center cooling infrastructure. For the case study example considered, the chiller energy use was the biggest fraction of about 41% and was also the most inefficient. The room air conditioning was the second largest energy component and was also the second most inefficient. A sensitivity analysis of plant and chiller energy efficiencies with chiller set point temperature and outdoor air conditions is also presented.
Growth in data center electricity use
  • J G Koomey
J. G. Koomey. Growth in data center electricity use 2005 to 2010. Analytics Press, Oakland, 2011.
degree in energy technology in 1997 at Royal Institute of Technology He joined ABB in 1998 and his career included 18 years of industrial experiences in process industry and datacenter infrastructure management. Dr. Zhang is currently Principal Scientist at ABB Corporate Research
  • Xiaojing D Zhang Received Ph
Xiaojing Zhang received Ph.D. degree in energy technology in 1997 at Royal Institute of Technology, Stockholm, Sweden. He joined ABB in 1998 and his career included 18 years of industrial experiences in process industry and datacenter infrastructure management. Dr. Zhang is currently Principal Scientist at ABB Corporate Research, Sweden and Adjunct Professor at Lulea University of Technology, Sweden.