Purdue E-Pubs

In depth analysis of Marine Corps University publication of essays on the historical and emerging aspects of amphibious warfare

Chapman, Bert

On Contested Shores:  The Evolving Role of Amphibious Operations in the History of Warfare

Liu, Jianghui

Jiang, Jinglin

Patra, Satya

Ding, Xiaosu

Huang, Chunxu

Cross, Jordan

Boor, Brandon

Jung, Nusrat

English

Indoor Nanoparticle Emissions and Exposures during Heat-Based Hair Styling Activities

Agrawala, Prabhav

Goyal, Anurag

Grid-interactive Efficient Buildings Using Thermal Energy Storage For Electric Heat Pumps

Grekin, Rebecca

de Chalendar, Jacques

Benson, Sally

Impacts of Relaxing Humidity Constraints on Chilled Water Demand in a Commercial Building

Chen, Dongyu

Qiao, Yiyuan

Sun, Qun Zhou

Physics-based Dynamic Bayesian Network For Fault Detection And Diagnostics In Building HVAC Systems

Wu, Feng

Yadav, Sourabh D.

Alkandari, Sarah A.

Ma, Jie

Dhillon, Parveen

Liu, Haotian

Braun, James E.

Karava, Panagiota

Ziviani, Davide

Horton, W. Travis

Design and Commissioning of a Facility for Studying Next-Generation High Performance Building Technologies and their Interactions with Occupants

Ham, Se Hyeon

Kwon, Soonbum

Kim, Yongchan

Purdue University 
Purdue e-Pubs 
International High Performance Buildings 
Conference School of Mechanical Engineering 
2024 
Comparative Energy Performance Evaluation of Data Centers 
using Economizer at Various Regions 
Se Hyeon Ham 
Korea University, un2179@korea.ac.kr 
Soonbum Kwon 
Korea University, bum8891@korea.ac.kr 
Yongchan Kim 
Korea University, yongckim@korea.ac.kr 
Follow this and additional works at: https://docs.lib.purdue.edu/ihpbc 
Ham, Se Hyeon; Kwon, Soonbum; and Kim, Yongchan, "Comparative Energy Performance Evaluation of 
Data Centers using Economizer at Various Regions" (2024). International High Performance Buildings 
Conference. Paper 485. 
https://docs.lib.purdue.edu/ihpbc/485 
This document has been made available through Purdue e-Pubs, a service of the Purdue University Libraries. 
Please contact epubs@purdue.edu for additional information. 
Complete proceedings may be acquired in print and on CD-ROM directly from the Ray W. Herrick Laboratories at 
https://engineering.purdue.edu/Herrick/Events/orderlit.html 
3561, Page 1 
Comparative Free Cooling Performance Evaluation of Water-Side and Air-Side 
Economizers for Data Centers in Various Regions 
Se Hyeon Ham 1, Soonbum Kwon 2, Yongchan Kim 3 * 
1 Graduate School of Mechanical Engineering, Korea University, Anam-ro 145, Seongbuk-gu, Seoul, 
Korea, un21 79@korea.ac.kr 
2 Graduate School of Mechanical Engineering, Korea University, Anam-ro 145, Seongbuk-gu, Seoul, 
Korea, bum 889 l @korea.ac.kr 
3 School of Mechanical Engineering, Korea University, Anam-ro 145, Seongbuk-gu, Seoul, Korea 
yongckim@korea.ac.kr 
* Corresponding Author 
ABSTRACT 
Data centers are growing rapidly, and their power demand is increasing significantly owing to their crucial role in 
data processing and storage. Recently, the adoption of free cooling methods utilizing economizers has emerged as a 
strategy to mitigate power consumption for data centers. However, limited research has been conducted on 
achieving optimal energy efficiency and reducing power consumption for data center operations. Furthermore, 
owing to the weather-dependent nature of free cooling methods, a regional analysis is essential. The objective of this 
study is to compare and optimize the free cooling methods utilizing economizers across different regions. TRNSYS 
was used to model data center systems. Two types of economizers, namely water-side economizer (WSE) and air-
side economizer (ASE) systems, were modeled with three weather datasets in Cairo, Seoul, and Toronto. The ASE 
outperformed the WSE owing to its wider operating temperature range and utilization. In Seoul, when the ASE was 
applied, the annual PUE decreased from 1.275 to 1.209. Notably, the economizer performance was better in Toronto 
than in Cairo. When the ASE was applied, the annual power usage effectiveness (PUE) decreased by 0.033, 0.066, 
and 0.078 in Cairo, Seoul, and Toronto, respectively. This phenomenon was particularly noticeable in regions with 
cooler climates, where economizers showed enhanced efficiency. 
1. INTRODUCTION 
As information and communication technology (ICT) is growing, the demand for data centers is increasing rapidly 
owing to their crucial role in data processing and storage. Moreover, the data center industry is growing at the 
largest rate among the ICT industries, and their power demand is increasing by more than 30% per year 
(International Energy Agency, 2021). Accordingly, achieving high energy efficiency has become an important issue 
for data center operations. In data centers, a lot of computing equipment is densely located and has a high heat 
generation density. A cooling facility, consisting of a cooling tower, chiller, pump, and computer room air handler 
(CRAH), is installed to dissipate the high heat generation. Approximately 30% of the total power of a data center is 
used for cooling and especially large amounts of power are consumed in chillers, CRAHs, and pumps (Cho, 2015). 
Therefore, it is important to reduce the power consumption of chillers, CRAHs, and pumps. 
There are many ways to reduce power consumption in data centers, one of which is the free cooling method through 
an economizer. The free cooling method is a method of cooling the server without operating the chiller. In the free 
cooling method, various types, mainly including water-side economizer (WSE) and air-side economizer (ASE), are 
commercialized. The WSE operates using a heat exchanger to cool chilled water through the heat exchange between 
cooling and chilled water. It can be applied either in series or parallel with the chiller. Meanwhile, the ASE utilizes 
outside air to cool the server room, thereby reducing the workload on the chiller. 
8th International High Performance Buildings Conference at Purdue, July 15 - 18, 2024 
3561, Page 2 
The free cooling system has been studied extensively. Zhang (2014) provided a comprehensive summary of 
advancements in data center free cooling and established criteria for free cooling systems. Li (2020) conducted 
model-based optimization of the free cooling system, and Deymi-Dashtebayaz (2019) compared the performance of 
a WSE and an ASE simultaneously. A regional analysis on the free cooling methods is essential because their 
performance can vary depending on climate conditions. However, limited research has been conducted on the 
weather-dependent nature of free cooling methods. 
The objective of this study is to compare and optimize the free cooling methods utilizing economizers in different 
regions. The models for the ASE and WSE were developed using TRNSYS in a data center with high heat 
generation density. Three weather datasets for Cairo, Seoul, and Toronto were utilized to assess the effectiveness of 
the free cooling. Finally, the annual power consumption and power usage effectiveness (PUE) of the ASE and WSE 
were compared based on the three weather datasets. 
2.METHOD 
2.1 Data center cooling system 
Figure l(a) illustrates the heat transfer process in a typical data center. This process involves the operation of the 
cooling tower, chiller, and CRAH to extract heat from the servers. The server fan dissipates heat from the server 
processor into the server room, where the CRAH uses the chilled water to cool the air. The chilled water then 
transfers heat between the CRAH and chiller, while the chiller exchanges heat with the cooling water. The cooling 
water carries heat to and from the chiller and cooling tower, where the cooling tower releases heat to the outside air. 
The pumps circulate both the cooling and chilled water. Figure 1 (b) shows the heat flow with parallel WSE 
application, utilizing a water-to-water heat exchanger for heat exchange between the cooling and chilled water 
without the need for a chiller. Figure l(c) shows the heat flow with indirect ASE application, employing an air-to-air 
heat exchanger to cool the server room air with the outside air. 
2.2 Numerical modeling 
A simulation on the energy consumption of a data center over a year was conducted using TRNSYS. An actual data 
center, located in Incheon, Korea, was considered based on information provided by the previous study (Jang, 2022). 
The building comprises nine floors, with a total floor area of 32,400 m squared. The 1st to 3rd floors are used as a machine 
room, and the 4th to 9th floors are used as a server room. The air-conditioning area is 21,600 m2 and the server room 
operates for 24 hours throughout the year. 
The specifications of the server were assumed as follows: 2U size Dell server unit, which has 1,400 W maximum 
power consumption, operating constantly for 24 hours; each floor accommodates a 40U size rack containing 20 units, 
resulting in a total of 120 server racks per floor; the power density of the server rack is 927 wm superscript -2. The indoor 
temperature was set at 18-27 °C according to the ASHRAE data center guidelines (ASHRAE, 2019). 
For the cooling system, TRNSYS default map data for the cooling tower and chiller were used. The performance of 
the pump and CRAH was predicted using similarity laws under partial load conditions. Weather datasets for Cairo, 
Seoul, and Toronto were employed to evaluate the effectiveness of the free cooling, with annual temperatures of 
18.7, 9.8, and 5.3 C, respectively. The detailed specifications of the data center and its components are listed in 
Tables 1 and 2, respectively. 
2.3 Data reduction 
The power efficiency of the server unit was assumed to be 100%, and the amount of heat generated by the server 
was calculated using Equation (1 ). 
         (1) 
The total power consumption for server cooling was calculated by Equation (2), where P subscript CT, P subscript CH, P subscript pump, and P subscript CRAH 
represent the power consumption in the cooling tower, chiller, pump, and CRAH, respectively. 
(2) 
The PUE was used to measure the cooling efficiency of the data center, as given in Equation (3). 
8th International High Performance Buildings Conference at Purdue, July 15 - 18, 2024 
 
  
 
     
   
 
 
 
    
  
 
 . . 
 
   
 
 
    
  
 
    
   
 
 
  
  
    
 
  
  
     
  
 
 
  
 
 
 
 
     
  
 
 
        
 
 
 
 
 
 
       
 
(3) 
3561, Page 3 
Figure 1: Heat flows in: (a) a typical data center, (b) WSE applied data center, and (c) ASE applied data center. 
Table 1: Specifications of data center. 
Category 
Area 
Power 
Value 
21,600 m squared 
20,000 kW 
8th International High Performance Buildings Conference at Purdue, July 15 - 18, 2024 
  
  
      
    
                                 
         
  
     
  
   
    
  
 
 
    
 
    
3561, Page 4 
Table 2: Specifications of components. 
Category Capacity Power 
Cooling tower 38,700 RT 222kW 
Chiller 33,720 kW 3,960 kW 
Pump 1,141 kg/s 328kW 
CRAH 223 m3/s 3,216 kW 
ASE 223 m3/s 1,608 kW 
3. RESULTS AND DISCUSSION 
3.1 Seasonal System Performance 
Figure 2 shows the annual trends of system operating temperature and component power consumption. The heat 
transfer occurred sequentially through the cooling water, chilled water, and server room air. As shown in Figure 2(a), 
the operating temperature was controlled according to the variation in the outdoor temperature, when the server 
room temperature was maintained in the range of23 .8- 31.4 °C. In Figure 2(b), the overall power consumption was 
lower in the winter season (December- February), resulting in significant reductions in the power consumption of the 
cooling tower, chiller, and CRAH. However, the energy performance of the data center decreased in the summer 
season (June- August) owing to the decrease in the cooling tower' s effectiveness. The cooling water in the cooling 
tower is brought down to the dew point temperature, which serves as its lower limit. As the outside temperature 
increased, the operating temperature of the entire system increased, resulting in decreased cooling tower 
performance and increased power consumption in the chiller and CRAH to maintain the target room temperature. 
Figure 2: Annual performance trends of the data center: (a) system operating temperature and (b) component power 
consumption. 
3.2 Annual Energy Performance 
Figure 3 shows the monthly energy trends of the data center. As shown in Figure 3(a), the PUE was higher in the 
summer season and lower in the winter season. The annual PUEs of without-free cooling (none), WSE, and ASE 
systems were 1.275, 1.239, and 1.209, respectively. At their lowest points, the monthly PUEs were 1.155, 1.116, and 
1.080, respectively. The ASE outperformed the WSE owing to its wider operating temperature range and utilization. 
The monthly PUE was higher in the summer season and lower in the winter season. This fluctuation was attributed 
to seasonal variations in outdoor conditions, which directly influenced the performance of the cooling system. 
8th International High Performance Buildings Conference at Purdue, July 15 - 18, 2024 
 
 
 
  
 
 
    
    
 
 
 
            
 
  
  
 
 
  
 
   
  
 
            
 
 
 
  
  
 
 
  
 
   
  
 
   
    
    
            
 
   
 
 
  
 
 
  
 
 
 
 
 
 
 
 
  
 
 
 
 
 
 
3561, Page 5 
During the winter season, the utilization of free cooling led to the most significant decrease in the PUE for both the 
WSE and ASE. As shown in Figures 3(b) and (c), the reduction in the chiller power consumption was significant in 
the WSE, whereas the decrease in the chiller and CRAH power consumption was substantial in the ASE. 
Figure 3: Monthly performance trends for the data center: (a) PUE according to free cooling type, (b) power 
consumption of WSE, and ( c) power consumption of ASE. 
3.3 Effects of climatic conditions 
Figure 4 shows the annual PUEs of the data center in different regions under three free cooling conditions. In the 
without-free cooling system, the annual PUEs for Cairo, Seoul, and Toronto, were 1.292, 1.275, and 1.259, 
respectively. Despite Cairo and Toronto having annual temperatures of 18.7 and 5.3 C, respectively, their annual 
PUEs did not significantly differ. This can be attributed to Cairo's hot and dry climate, which can maximize the 
cooling tower' s performance. When the WSE was applied, the annual PUEs for Cairo, Seoul, and Toronto decreased 
to 1.290, 1.239, and 1.210, respectively, resulting in PUE reductions of 0.002, 0.036, and 0.049, respectively, 
compared to that of the without-free cooling system. In Cairo, the annual PUE decreased by only 0.002 compared to 
the without-free cooling system because its outdoor temperature was too hot to utilize the WSE effectively. 
However, in Toronto, the substantial annual PUE decrease was attributed to its high WSE utilization with the lower 
outdoor temperature. When the ASE was applied, the annual PUEs for Cairo, Seoul, and Toronto decreased to 1.259, 
1.209, and 1.181 , respectively, resulting in PUE reductions of 0.033, 0.066, and 0.078, respectively, compared to 
that of the without-free cooling system. Overall, the ASE demonstrated effectiveness in reducing the PUE for all the 
regions because it effectively cooled the server exhaust air using the outdoor air. 
Figure 4: Annual PUEs of the data center for various regions. 
8th International High Performance Buildings Conference at Purdue, July 15 - 18, 2024 
3561, Page 6 
4. CONCLUSIONS 
In this study, the seasonal energy consumption for data centers and the performance of free cooling systems were 
analyzed in different climatic regions. TRNSYS was used to estimate the performance of the components under 
various operating conditions. The without-free cooling, WSE, and ASE systems were modeled with three weather 
data sets including Cairo, Seoul, and Toronto. The power consumption of the cooling tower, chiller, and CRAH 
were lower in the winter season and higher in the summer season. The free cooling methods significantly improved 
the performance in the winter season. The ASE outperformed the WSE owing to its wider operating temperature 
range and utilization, reducing the annual PUE from 1.275 to 1.209 in Seoul. Moreover, the performance 
improvement of the economizer also varied depending on the climatic region. With the ASE application, the annual 
PUE decreased by 0.033, 0.066, and 0.078 in Cairo, Seoul, and Toronto, respectively. This was particularly 
noticeable in the regions with cooler climates, where economizers showed enhanced efficiency. 
p power consumption (W) 
P subscript ASE power consumption of ASE (W) 
P subscript CH power consumption of chiller (W) 
P subscript cooling power consumption of cooling (W) 
P subscript cRAH power consumption of CRAH (W) 
P subscript cT power consumption of cooling tower (W) 
P subscript pump power consumption of pump (W) 
eta subscript DC data center power usage efficient (-) 
Acronyms 
ASE air-side economizer 
CRAH computer room air handler 
ICT information and communications technology 
PUE power usage effectiveness 
WSE water-side economizer 
NOMENCLATURE 
REFERENCES 
International Energy Agency (2021). Data centres and data Transmission networks, International Energy Agency. 
Cho J., Yang J., Lee C., Lee J. (2015). Development of an energy evaluation and design tool for dedicated cooling 
systems of data centers: sensing data center cooling energy efficiency, Energy Building 96, 357-372. 
Jang Y., Lee D., Kim J., Ham S. H., Kim Y. (2022). Performance characteristics of a waste-heat recovery water-
source heat pump system designed for data centers and residential area in the heating dominated region, J. Building 
Engineering 60, 105416. 
ASHRAE (2019). ASHRAE 90.4 standard for data center, Atlanta, Georgia: ASHRAE. 
Zhang H., Shao S., Xu H., Zou H., Tian C. (2014). Free cooling of data centers: a review, Renewable and 
Sustainable Energy Reviews 35, 171-182. 
Li J., Li Z. (2020). Model-based optimization of free cooling switchover temperature and cooling tower approach 
temperature for data center cooling system with water-side economizer, Energy and Buildings 227, 110407. 
Deymi-Dashtebayaz M., Namanlo S. V., Arabkoohsar A. (2019). Simultaneous use of air-side and water-side 
economizers with the air source heat pump in a data center for cooling and heating production, Applied Thermal 
Engineering 161, 114133. 
ACKNOWLEDGEMENT 
8th International High Performance Buildings Conference at Purdue, July 15 - 18, 2024 
3561, Page 7 
This work was supported by a National Research Foundation of Korea grant funded by the Korean Government 
(MSIT) (Grant number 2023RlA2C2003124). 
8th International High Performance Buildings Conference at Purdue, July 15 - 18, 2024 


Comparative Energy Performance Evaluation of Data Centers using Economizer at Various Regions

Gluesenkamp, Kyle

Akamo, Damiola

Li, Zhenning

Hirschey, Jason

Liu, Xiaobing

Shen, Bo

Exploring the Power Demand and Efficiency Performance Limits of Heat Pumps with Thermal Storage

Margraf, Hendrik

Meral, Fatih

Lonardi, Federico

Luke, Andrea

Detailed Analysis of Energy Demand and COVID-19 Impacts on Hotel Buildings

Montibeller, Cassiano

Thaler, Gabriel

Flesch, Rodolfo César Costa

Machado, João Paulo Zomer

Ludwig, Guilherme Henrique

Automatic Test Stand to Impose Operating Suction and Discharge Conditions on Compressors

Purdue University West Lafayette

Not a member yet

On Contested Shores: The Evolving Role of Amphibious Operations in the History of Warfare