Wooden Construction
Marika Gille
Civilingenjör, Arkitektur 2017
Luleå tekniska universitet
Institutionen för samhällsbyggnad och naturresurser
This thesis is the closing part of the master program in Architectural Engineering, fo- cusing on building construction, at Luleå University of Technology. The thesis is written during 2017 and corresponds to 30 credits.
The thesis is part of a research project at the division of Industrialized and Sustainable Construction at Luleå Unviersity of Technology. The project is called ’Modularized and energy efficient data centers in wood’. This project has similarities to another project conducted by the same division, called Green Power, that focuses on using the waste heat from data centers to power greenhouse.
I would like to thank my mentor Marcus Sandberg (LTU) who have guided and sup- ported me through my work. I would also like to thank Johan Jatko who have given me support and much valued information about the data center industry. Also thanks to everyone that participated in interviews and site visits.
Lastly I would like to thank my classmates and all the friends I have gotten to know during these years. Without all the support and fun we have shared this degree would not have been possible.
The purpose of this thesis is to investigate the possibility to build a modular data center in wood. The goals is to investigate how to build data centers using building system modules, making it easier to build more flexible data centers and expand the business later on. Investigations have been conducted to find out advantages and disadvantages for using wood in a modularized data center structure. The investigation also includes analysing the moistures effect on the material and if there are any other advantages than environmental benefits in using wood as a building material.
A literature study were conducted to examine where research already have been con- ducted and how those studies can be applicable to this thesis. Although the ICT sector is a rapidly growing industry little research has been published in regards to how to build a data center. Most published information involves electric and cooling, not measurements of the building and how materials is affected by the special climate in a data center. As a complement to the little research interviews were conducted and site visits were made. In- terviews were conducted with Hydro66, RISE SICS North, Sunet and Swedish modules, whilst site visits were made at Hydro66, RISE SICS North, Sunet and Facebook.
As a result of these studies, limitations were identified with regards to maximum and minimum measurements for the building system and service spaces in a data center. These limitations were used as an input when designing a construction proposal using stated building systems and a design proposal for a data center.
During the study, access have been granted to measurements of temperature and hu- midity for the in- and outgoing air of the Hydro66 data center. These measurements have been analyzed with the facts about HVAC systems and the climates effect on wood, for example in regards to strength and stability. This analysis has shown that more data needs to be collected during the winter and that further analysis needs to be conducted, to be able to draw conclusions if the indoor climate of a data center has an effect on the wooden structure.
A design proposal for a data center have been produced with regards to the information gathered by the litterature and empirical studies. The proposal were designed to show how the information could be implemented. The result have increased the understanding on how to build data center buildings in wood and how this type of buildings could be made more flexible towards future changes through modularization.
Keywords:Data center, Modularity, Wooden structure, Indoor climate,
1 Introduction 1
1.1 Background . . . 1
1.2 Purpose and goals . . . 3
2 Method 4 2.1 Literature study . . . 5
2.2 Interviews . . . 6
2.2.1 Site visits . . . 7
2.3 Analysis of climate measurements . . . 7
2.4 Design proposal of a data center . . . 7
3 Theory 9 3.1 Data centers . . . 9
3.1.1 Data center types and usages . . . 9
3.2 Modularization . . . 12
3.2.1 Industrialized building . . . 12
3.2.2 Modular Architecture . . . 13
3.2.3 Building systems . . . 15
3.3 Building in wood . . . 20
3.3.1 Economy and culture . . . 22
3.3.2 Construction . . . 22
3.4 Moisture . . . 22
3.4.1 Strength and stability . . . 23
3.4.2 Biological impact . . . 26
3.5 Heating, Ventilation and Air Conditioning . . . 27
3.5.1 Mechanical cooling . . . 28
3.5.2 Free cooling . . . 28
3.6 Need for research . . . 31
4 Results 32 4.1 Interviews . . . 32
4.1.1 Börje Josefsson, Sunet . . . 32
4.1.2 Christiaan Keet, Hyrdo66 . . . 32
4.1.3 Tor Björn Minde, RISE SICS North . . . 33
4.1.4 Pär Åberg, Swedish Modules . . . 33
4.2 Site visit . . . 34
4.2.1 Sunet . . . 34
4.2.2 Hydro66 . . . 34
4.2.3 RICE SICS North . . . 35
4.2.4 Facebook . . . 35
4.3 Measurements . . . 36
4.4 Design proposals for a data center . . . 40
4.4.3 Construction Propositions . . . 46
5 Analysis and Discussion 48 5.1 Advantages and disadvantages with a modularized building system for data centers . . . 48
5.2 Wooden structures in a data center climate . . . 49
5.2.1 Wood . . . 49
5.2.2 Moisture . . . 50
5.2.3 Heating, Ventilation and Air-Conditioning . . . 50
5.3 Different modules’ impact on building design . . . 52
5.3.1 A design proposal for a modular data center in wood . . . 53
5.4 Critical review of methods . . . 54
5.5 Future investigations . . . 56
6 Conclusion 57
Bibliography 58
A Interviews 64
ICT Information and Communication Technology ASHRAE American Society of Heating, Refrigerating and Air-Conditioning Engineers
LEED Leadership in Energy and Environmental Design
SICS Swedish Institute of Computer Science
ICE Infrastructure- and Cloud research Environment
RISE Research institute of Sweden
RH Relative Humidity in air
VOC Volatile Organic Compounds
PUE Power Usage Efficiency
ERE Energy Reuse Efficiency
MPa Mega Pascal
HVAC Heat, Ventilation and Air Conditioning
UPS Uninterpretable Power Supply
CLT-boards Cross Laminated Timber boards
Data center A data center is a place dedicated to a company’s ICT infrastructure Modularization A course of design of independent units that can be combined
in a number of ways
Industrialized Building Pre-fabricated buildings mostly built in an off-site industry LEED A rating system to evaluate the environmental performance of a building and encourage market transformation towards sustainable design.
Pod A contained environment in a data center, can consist of a set of racks and equipment
In-Situ Construction on site
Off site Construction off site in a factory
Emergency power back up generators that powers the data center in case of a black out
1. Introduction
1.1 Background
A data center is a place dedicated to a company’s IT-infrastructure. It can be as small as a rack or as big as a building. Some data centers are only for one company and some are for multiple companies. In these data centers, information is stored and applications necessary for the companies’ daily business (Fortlax, 2017a).
The ICT (Information and communication Technology) sector is today a big and up- coming industry (Åberg, 2015). In the last 10 years’ data centers have been built rapidly in many places around the world, such as in the Northern of Sweden. Countries can benefit from the development of the ICT-sector. Since in 2015 alone the data center industry had a total economic impact of SEK 13 billion and engaged about 7 000 full-time workers, only in Sweden (Warrenstein et al., 2016).
An international study shows that 700 new data centers in the world are needed, and 200 of them in Europe (Tedestedt, 2017). Sweden has a big chance of getting many of them since they are top four on the list of most qualified countries in the world to develop data centers in, as can be seen in figure 1.1 (Warrenstein et al., 2016). The reason for being such a great country for data centers include cost for energy, risk for natural disasters and political stability to mention a few.
Figure 1.1: 2015 Data Center Qualification Index (Warrenstein et al., 2016).
A big part of attracting new business to Sweden is the reduced taxes for data centers from 2017 (Vattenfall, 2017). Another part of attracting new companies are the big focus and knowledge on climate neutral business. Sweden are skilled in the climate issues today ranking third on the Environmental Performance Index 2016 according to Yale University (2017). The Swedish government wants to be even better by 2025 as stated in their national goals in collaborations with Boverket (Boverket, 2012).
In 2011 Facebook initiated the Open Compute Project as a way to "break open the black box of proprietary IT infrastructure to achieve greater choice, customization, and cost savings" (Park, 2017). Facebook wanted to share their knowledge of building data centers to those who can’t hire design teams and to get feedback on their ideas to improve their future work (Heiliger, 2011). Since this project started in 2011 more information on how to build a data center is now available but there are still some gaps in the research.
Apart from economic and job benefits created by building data centers, The cities in which the companies choose to develop their data centers benefit from new enterprises due to maintenance and subcontractors. Big companies’ data centers is also a marketing tool for the region. Many countries now know where Luleå is located, thanks to Facebook’s establishment (Warrenstein et al., 2016).
The need for flexible solutions has increased in recent years (Data Center Knowledge, 2017; Baines, 2015). When an architect designs a residential building, they predict a life length of approximately 50 years. An industrial company can predict a buildings requirements for 20-30 years ahead. But the ICT industry changes so fast today that the hardware is outdated in 2-3 years. This makes it hard to plan for a data center and in advance predict how the business will look in 10 years (IBM, 2017). When looking at these numbers one can see that there is a need for flexible data centers to grow the company based on demands and the development of the market.
A modular architecture refers to the design of a system composed of separate com- ponents that can be attached together. The opposite of modular architecture is integrated architecture where no separate parts can be identified (Webopedia, 2017). Modularity may therefore enhance the possibility to build flexible buildings that may change with the company. Although much research has been conducted about modularized buildings and industrial building in wood, the research is mostly done in regards to residential buildings, and for the better part villas.
The climate in a data center is fairly constant over the year. The climate has high hu- midity and temperature (ASHRAE, 2016). Wood is a hygroscopic material that is affected by its surrounding environment. Wood strives for equilibrium with its surroundings and easily changes its moisture content. With varying moisture in the wood comes changes in dimensions, strength and durability to mention a few (Svenskt Trä, 2017a). Although a lot of research has been conducted regarding how wood behaves in different climates, little research has been conducted on how a data center’s climate affects wood.
The research about wood is, today, mostly done for villas and residential buildings in regard to climate and moisture or in controlled laboratory environments. For example (Saarman, 1992) and (Angst-Nicollier, 2012) to mention a few. More researches needs
to be conducted regarding data center climates and wood. A data center have a warm and moist climate and larger spans than residential buildings and therefore the residential buildings research may not be applicable on this type of building.
1.2 Purpose and goals
The purpose of this thesis is to investigate the possibility to build a modular data center in wood. The goals is to investigate how to build data centers using building system modules, making it easier to build more flexible data centers and expand the business later on. Investigations have been conducted to find out advantages and disadvantages for using wood in a modularized data center structure. The investigation also includes analysing how the moist inddor climate of a data center may effect a wooden structure and if there are any other advantages than environmental benefits in using wood as a building material in a data center.
The thesis addresses buildings that have the main purpose of hosting a data center.
This thesis will also take into consideration to re-use waste heat from a data center, for example by using a ventilation system that supports re-usage of warm air and usage of the waste heat that is leaving the building.
As a part of the result, a design proposal for a modularized data center in wood will be developed. For the design of the data center, the study has been made with regards to a data center located at an open area in the northern parts of Sweden.
Research questions
• What are the advantages and disadvantages of using a modularized building system for data center building design?
• How will different modular systems affect the data center building design?
• How may the moist climate in a data center affect a wooden structure?
2. Method
The aim of this thesis is to investigate the possibility to build modular data centers in wood. This investigation process includes to find the needs for data centers, identifying different modular building systems and analysing of the climate in a data center and the climates effect on wood.
This thesis has been conducted with an explorative method complemented with a qual- itative method. An explorative method is an investigative process used when little or none knowledge of the studied area existed before the study. An explorative process is used to achieve a basic understanding of the topic (Björklund and Paulsson, 2012). A qualitative study is used when the aim is to get a deeper understanding of a specified topic. A qual- itative study can be compared to a quantitative method where the information collected can be measured and valued (Björklund and Paulsson, 2012).
The explorative method has been chosen to collect information about data centers. The explorative method suits this subject best due to the lack of previously published literature in this subject. When gathering information about modularity and wood as a construction material, literature studies was the only chosen method. This choice was based on the fact that these areas separately are a well-researched areas with many publications. Although, as stated before, most of the research have been conducted on residential villas and not on data centers, literature studies were chosen as the method for gathering information. This is because there are almost no data centers today built in wood.
The analyses were conducted with some previous knowledge. Before this thesis a 7- month long internship on a construction site and 4-month internship as a constructor have been conducted and has affected the analysis. These internships have affected the view on the construction part of this thesis since some knowledge have been gathered in regards to how a construction worker thinks when building.
In figure 2.1, a flow chart over the research process for this thesis is shown.
Figure 2.1: Flow chart over research process.
2.1 Literature study
When conducting literature studies, it is important to be aware of that the information gathered often are subjective and sometimes inconclusive dependent on search words and databases. Strengths with literature studies is that much information can be gathered under short periods of times and it is a good way to investigate existing knowledge (Björklund and Paulsson, 2012).
The literature study has been divided into two parts. The first part addresses data
centers, how they are designed and what equipment is needed. This part was a pre study to the interviews since little knowledge about the subject were known before this thesis.
The second part includes modularity in the construction industry and wood as a structural material as well as moistures effect on wood.
Searches have been conducted, both in English and in Swedish, for following words;
data center, ICT-sector, wood, moisture in wood, free cooling, timber structure, industrial building, wooden construction, and modularization.
Searches have been done in the Luleå University of Technology library, Google Scholar, via the search site for articles available on LTU library website and DIVA.
2.2 Interviews
The interview method chosen were a semi-structured interview where questions were pre- pared but altered depending on the answer given by the interviewee or the organizations business activities. Advantages with interviews is that it gives you fast access to relevant information regarding the subject and can often lead to a deeper understanding than writ- ten text. Disadvantages with interviews is that they are time consuming (Björklund and Paulsson, 2012).
The aim with the interviews were to get a new perspective to the main questions of this thesis and to get some more information about data centers. Questions, that can be found in Appendix A, were asked about the function of a data center, how the interviewee sees their facilities, what works well and not so well. The interviews were conducted at the office or location of the interviewees liking. With Pär Åberg a phone interview were choosen due to that he is working in the south of Sweden. The interview with Pär and Börje were about 30 minutes long and the interviews with Tor Björn and Christiaan were around 1 1/2 to 2 hours long. Only persons that work with data centers have been interviewed, since it is there most knowledge and information were lacking.
Interviews were conducted with the following persons
• Christiaan Keet, CTO, Hydro66
• Börje Josefsson, COO, Sunet
• Tor Björn Minde, CEO, RISE SICS North
• Pär Åberg, CSO, Swedish Modules
Other organizations were also contacted but didn’t have the possibility to participate.
The questions and answers can be found in Appendix A. All interviews have been approved for publishing by the interviewees.
2.2.1 Site visits
Tours were conducted at the following facilities all located in the Northern of Sweden.
• Hydro66, a colocation data center
• Sunet, a data center
• RISE SICS North, research institution
• Facebook (Video tour), Facebook data center
A virtual tour of Facebook’s data center in Luleå was seen, as well as their permanent exhibition at Teknikens Hus in Luleå. According to their site coordinator this is the same information that should have been told in an actual tour of the facility.
These site visits were a possibility to gather more information about data centers, both the design and operation, as well as get inspiration for future design proposals in this thesis.
2.3 Analysis of climate measurements
Luleå University of Technology in association with Hydro66 in Boden Municipality have conducted tests of the air going in and out of the data center, documenting the relative humidity and the temperature of the air. Access has been provided to these measurements from July 2016 to February 2017. The measurements were presented in a chart, this chart as well as two extreme value points have been chosen to be used in this thesis. One between 20-25 of July when there were the warmest in the summer and one between 3-6 of January when it was the coldest.
These measurements have been performed at Hydro66 by Marcus Sandberg and Mikael Risberg from LTU and given to me by Marcus Sandberg. These measurements have not been analysed or published before this thesis. For this thesis, the measurements are anal- ysed with the purpose to see how the climate in a data center changes throughout the year and if this climate could have an effect on the woods strength and stability.
The temperature and humidity for these periods were transfered to tables, stating aver- age day and night values. For the temperature and humidity to have an effect on the wood there needs to be many fluctuations. So the analyses where in regards to large variations in humidity during these periods.
2.4 Design proposal of a data center
When all information had been collected and valued, limitations for a data center were stated. These limitations were a base for the design and dimensions of a data center.
When these dimensions were chosen and a design for the building were decided, two different companies were contacted, Martinsons and Swedish Modules. These companies where contacted to get more design proposals for comparison. Martinsons is a glulam manufacturer that is connected to the research project Modularized and energy efficient data centers in wood. Swedish Modules manufactures complete modules with a steel construction, this company is used as a comparison between steel and wood in regard to constructions solutions.
It is important to produce a design proposal to explore the practical possibilities to build a modular data center in wood and to show both building engineers and data center technicians different ways of thinking in regard to designing a data center.
3. Theory
3.1 Data centers
A data center’s main function is to provide an environment to run computers and electrical equipment on a large scale. Often with focus on availability and security. In a data center many systems need to work together in a perfect cyclic system. A data center needs power, HVAC, shelter, and security but should also be able to remove waste heat (Barroso et al., 2013).
In 2017 the Swedish government decided to give a tax reduction to data center compa- nies that uses more than 0.5 MW for equipment not including fans and cooling (Vattenfall, 2017). This tax reduction aims to enable for more companies to develop data centers.
3.1.1 Data center types and usages
There are different types of data centers that targets different types of clients. CyrusOne (2017) share their version of five different usage areas for data center facilities:
• In-house Data Center
The smallest data centers are the ones located at a company. This room includes servers with enough power to store the company’s information and services. The size and capabilities of an in-house facility depends on how much the company need and is willing to spend.
• Colocation
A colocation data center have multiple tenants that buys rack, cabinets, or cages.
Many companies can benefit from a colocation service since the customer maintain control over their hardware but outsource facility and maintenance to the colocation owner.
• Wholesale Data Center
A wholesale data center can be compared to when a company leasing an office space where the landlord provides maintenance. Often these buildings have fewer customers, otherwise a wholesale data center is much like a colocation.
• Dedicated Hosting
In a dedicated hosting situation, the customer buys full services. The customer has less control in what hardware that are being used and can’t expect to access their servers physically.
• Managed and Shared Hosting
When looking at managed and shared hosting you see that the customer has less control over their equipment. In these types of hosting the hosting company can supply different amount of systems and services to the client. These hosting types suits companies that have little knowledge of IT-equipment and don’t want to have an in-house technician.
Building types
Between the above hosting types there are often little to no difference in building de- sign, (CyrusOne, 2017), and can all be fitted into the envelope of the companies’ choice.
Smaller data centers may be fitted in a room in a building. Some data centers are located in a container, as seen in figure 3.1, that can be transported and relocated easily. The container is often completed with fire safety and backup power. The container often only needs a power source to work (Morgan, 2009).
Figure 3.1: Container data center by IBM (Morgan, 2009).
Then there are stand-alone data centers, whole buildings built only for housing severs and equipment. One example of this is Facebooks data center in luleå, figure 3.2.
Figure 3.2: Visualization of Facebook data center Luleå (Seth, 2014).
One variant of a stand-alone data center is when an old industrial building is rebuilt into a data center, either by changing the layout so that it works for a data center or by putting containers inside the building (Tech Target, 2017).
3.2 Modularization
Modularization, modular, industrialize, and building are defined by (Oxford University, 2017) as:
Modularization - "The action or process of making something modular; construction on modular principles"
Modular - "Employing or involving a module or modules as the basis of design or construction. Relating to an educational course designed as a series of independent units of study that can be combined in a number of ways."
Industrialize - "Build up a system of industries"
Building - "The action of constructing something"
3.2.1 Industrialized building
When looking at the definitions above, an interpretation of the meaning of industrialized building is, to build a system for constructing something. In the construction industry, industrialized building can be referred to as modularity. Modularity is usually the way to build that often uses pre-fabricated walls or whole modules that you assemble on-site.
These modules are often chosen by the client from an assortment of previously designed parts. When searching for modularity in Sweden, you will mostly see modularity by turnkey villas where the contractor has a few different villas to choose from, and the client can make minor changes. Industrialized building intends to reduce cost and time, reduce risks, and enhance quality (Lou and Kamar, 2002; Smith, 2015).
With continual change as a guiding star, flexibility to develop and expand becomes im- portant sales pitches for the future. Many of today’s knowledge companies are fast grow- ing and requires expansion possibilities when choosing location and construction(Fernström and Kämpe, 1998). When new business start-up they do not want to pay a big amount of money at once for big data centers that may not be needed in the future. But with a modularized system they may grow the data center at the same rate as the business.
Another benefit of modularity and pre-fabricated buildings is reduced material and purchase costs to up to 10 %, due to that larger quantities of material can be bought to the industry and less waste material is thrown away. Higher quality can be secured due to a controlled climate and continued testing of modules before delivery and assem- ble(Ericsson and Erixon, 1999; Murtaza et al., 1993).
Figure 3.3: Development of the timber housing industry (Dotted line for trend)(Lennartsson, 2009).
Since Boverket released new building codes in Sweden 1995, that focused more on function rather than technical requirements industrialized building have been more com- mon in Sweden. Multi-story timber buildings is now a possibility that more companies discovers as can be seen in figure 3.3 (Lennartsson, 2009; Boverket, 2012). A problem that Lidelöw et al. (2015) mentions with industrialized building is, for example, that peo- ple in Sweden still think that pre-fabrication in the construction industry is equal to a shoe box on a land area. The construction industry is not so good at implementing the new techniques and methods available in other industries and repeatedly falls back into old footsteps. This comes into correlation with todays market where people are starting to get used to the idea of being able to buy what they want and not what the market has to offer(Lidelöw et al., 2015). To solve that problem Lidelöw et al. (2015) thinks that a change from mass production to mass customization is needed.
Historically, manufacturing has the extremes as either mass production or customiza- tion. Mass production produces large quantities of the same product while customization means smaller quantities of products made just for a certain customer. Module based pro- duction can be seen as customized mass production where many different end results can be produced without many changes to the production (Hjalmarsson and Jonsson, 2010).
Standardization is an important part of industrialization. Although still standardized, the focus have shifted from a standardized end product. Like the Swedish million pro- gramme where the apartment looked exactly the same, towards standardized parts, as elements and fixings. The focus on standardized parts instead of the product gives more variations in floor plans and window placements (Lidelöw et al., 2015).
3.2.2 Modular Architecture
Ulrich (1995) arguments that a modular architecture increases the chances for a compo-
nent to be standardized. A standardized component will be a product that fits in more than one architectural module. Standardization lowers the manufacturers cost in regard to development and production.
The construction industry is dependent on the economic situation in society. Due to this dependency, there is a risk for a company to invest in a system for a small line of business. The risk is that if the market changes the investment will be redundant. But the smaller branches are also often the most profitable since they are customized for their purpose (Lidelöw et al., 2015).
When working with modular products, Ulrich (1995) also defines three types of mod- ular systems. These system types are;
• Slot modularity
– Component sharing modularity – Component swapping modularity – Cut-to-fit modularity
– Mix modularity
• Bus modularity
• Sectional modularity
In figure 3.4 these types can be seen.
Figure 3.4: Different types of modularity (Ulrich, 1995).
Slot modularity is divided into four different subtypes. All subtypes have in common that the module can be connected in a certain position with a standardized interface. Com- ponent sharing modularity is when multiple modules share the same type of components.
Component swapping modularity is when multiple types of components is attached to a base product to create different product variants. Cut-to-fit modularity uses parametriza- tion to change dimensions but still have the same interface. And in the last type, mix modularity combines a various number of components to create a new device.
Bus modularity adds extra modules to an existing product. For example, placing walls on a foundation slab. And sectional modularity is when modules have standardized inter- faces so that the modules can be combined in various manners (Hjalmarsson and Jonsson, 2010).
Key aspects and concerns with modular architecture Sanchez (2012) and The Bone Structure (2016) mentions in their articles are:
• Grater variations in products through smaller changes.
• Faster development times since smaller parts can be upgraded at different times.
• Lower development costs since you do not have to start from the beginning every time.
• Reduced production costs when ordering larger quantities and less education of production staff.
• Reduced influence in floor plan and design.
• Modular buildings does, in some countries, still have the stigma of being of lower quality than stick-built buildings.
3.2.3 Building systems
Regarding modularized building there are several different types of building systems.
These are Stick-built houses, block elements and volume elements. These systems have different advantages and disadvantages, and have been developed through time.
Stick-built homes or beam-post system
Figure 3.5: Stick-built house (Moelven, 2017).
When ordering a stick-built house from a manufacturer the factory delivers construction timber cut to fit the house. Often a stick-built home has a structure of columns and beams, like the structure in figure 3.5. The client then assembles these parts together with help of a drawing. Stick-built home leaves room for more variations for the client to choose floor plan. This is often more economical than other choices but demands more from the client regarding the assembly of the house or engagement of a contractor. By this choice of module, you have more options in regards to furniture and wall cladding, compared to other module types. Another benefit from this building system is that the bearing elements is located in the pillars which makes it easier to change the purpose of the building and alter floor plans in a later stage. One drawback with this system is that with larger spans comes larger beams. These beams may be a limitation if large openings is needed or if a consistent ceiling height is required. Another limitation with pillar-beam structures is that there is a need for many companies to collaborate with installation and assembly on site.
With many parties involved in this process the transfer of information needs to be correct at all times(Alla hustillverkare, 2017; Lidelöw et al., 2015).
Figure 3.6: Building with Trä8 element (Moelven, 2017).
Trä8 is a type of beam-post system developed by Moelven, a Scandinavian industrial group that produces building materials and systems for the construction industry, and Luleå University of Technology (Moelven, 2017). The Trä8 system can be seen in figure 3.6. The aim of the research was to come to a solution for an open space system for multi- story buildings, all in wood. The system can deliver spans up to 8 meters. This system leaves more room for the architect to create new floor plans with the benefits of volume elements and fast erection times (Moelven, 2014).
Wall block systems or load bearing walls
Figure 3.7: Wall block (Small Apartments, 2017).
A wall block system is a system manufactured in a factory off-site. Wall block systems are often the load bearing structure in the building. These wall blocks have dimensions that can be transported with a truck. Figure 3.7 shows an example of a wallblock at delivery. Often these modules are constrained by standards for other building materials, such as plasterboards and plywood. In Sweden, this often means a width in steps of 60 centimeter between joist. The walls are completed with isolation, primed facade, wall cladding, piping for electrics, water, and ventilation. Even windows and sometimes doors will be assembled before delivery. On site these wall systems will be mounted on the slab and connected to each other. This module gives shorter assembly times but still gives the client more choices in regard to floor plans and furniture. Even here the limitations is communication on site (Dala hus, 2017; Lidelöw et al., 2015).
Volume elements
Figure 3.8: Volume element (Svenskt Trä, 2017c).
A volume element consists of four walls, ceiling, and floor. Everything is assembled off- site and then transported to the building site. Often these volume elements, figure 3.8, will be completed with wall cladding, furniture and other installations before delivery. This module system is the most complete choice, here the clients have little or no saying about floor plan and furnishings. This system works well when building multi-story buildings with shorter spans. The point of communication between different parties at the construc- tion site, that other systems have problems with, is not a limitation in volume elements since the elements are almost complete delivery. The disadvantages with this type of modules are the limitations in regard to spans and width for material and transportation.
These limitations in spans also limits the possibilities for floor plans (Elfström and Singh, 2013; Lidelöw et al., 2015).
Combined building systems
A couple of companies have started to experiment with combining different building sys- tems. Examples are where you have volume modules for bathroom and kitchen and the larger, more open, areas are built with a pillar-beam system (Lidelöw et al., 2015).
3.3 Building in wood
A study in Canada amongst architects, structural engineers and others involved in the construction industry summarizes motivations and barriers for using wood as a structural component in non-residential buildings (Gosselin et al., 2017).
Figure 3.9: Motivations for adapting wood as structural material for non-residential build- ings (Gosselin et al., 2017).
As can be seen in figure 3.9 sustainability is the main motivation to use wood as a structural material. But also, things as cost and building speed is well represented. The largest barrier, figure 3.10, for using wood as a structural material were the building codes.
Many engineers found the building codes inadequate and hard to read. Also in the barrier pyramid, figure 3.10, you can see the cost factor. The same argument that are used as motivation are also used as a barrier. This may be due to shorter building times saves money whilst longer design processes demands more resources.
Figure 3.10: Barriers for adapting wood as structural material for non-residential build- ings (Gosselin et al., 2017).
Forests covers a large percent of the land area in Canada as well as in Sweden so therefore material availability should not be a barrier (Hidalgo, 2017).
Downsides with wood as a construction material is that it can crack and creep with time. Wood is also flammable and can contribute to a faster fire flow. When exposed to water and moisture wood can begin to rot or grow mould that can be harmful to humans (Jacobsson, 2007).
Figure 3.11: Carbon dioxide emissions in the production of building materials (Svenskt Trä, 2017b).
A study by Svenskt Trä (2017b) concludes that it is hard to compare the environmen- tal effects of different materials. Figure 3.11 shows the least amount of carbon dioxide manufacturing of massive wood release, and this number is without the factor that wood can capture carbon the whole life cycle (Hynynen, 2016; Hidalgo, 2017).
3.3.1 Economy and culture
Building in concrete is more time consuming and advanced than building in wood. Cast- ing the concrete in-situ requires knowledge, wooden frames, time, and a lot of reinforce- ment bending. Risks with this technique include quality issues due to unfavourable site conditions, bad weather, or the lack of skilled competence (Lou and Kamar, 2002). To lower costs many timetables in the construction industry are very unrealistic and does not allow for proper dry times for concrete, resulting in moisture damages, whilst timber construction is based on prefabricated dry elements and fast assembly times (Hynynen, 2016).
The construction industry is scared of changes which can be seen in figure 3.10. Ac- cording to Jonas Hjalmar, psychologist, change is unknown and humans rather wants a certain present situation than an uncertain future. Even though the future may be better in the long run (Blom, 2017). For example, culture in the industry and lack of experience is all factors of a strong tradition of building more in concrete and steel, even though it is easier to work in and assemble wood (Dolby et al., 1988).
3.3.2 Construction
The majority of a timber construction is produced off-site, which improves both qual- ity and productivity. In Germany, off-site construction improved the quality as well as provided bigger variety and flexibility in design. In Finland, off-site manufacturing repre- sented 70 % of total building construction and offered effective and rapid site assembly as well as improvement of quality and productivity. In Malaysia, they have stated their own term ’Industrialized building system’ which is to manufacture components in a controlled environment and then assembly the construction with minimal additional site work. This system has had good benefits on time and cost as well as quality (Lou and Kamar, 2002).
3.4 Moisture
According to Judge (2015), a common opinion in the data center industry is that you need to keep a high humidity and as low temperature as possible in a data center to lengthen the lifetime of the servers. ASHRAE (American Society of Heating, Refrigerating and Air- Conditioning Engineers) has conducted a study showing that lower temperatures, from 30 to 20◦C, have insignificant effects on the lifetime of the servers. A concern for low humidity in the data center has been that the risks of static electricity from clothes and
other materials would increase. ASHRAE have now concluded that there is an insignifi- cant higher risk between 8 % RH (Relative Humidity) and 20 % RH where the previous recommendations were stated (Judge, 2015). This can be compared to the Swedish rec- ommendations of a residential humidity of 30-70 % RH Socialstyrelsen (2005).
3.4.1 Strength and stability
Wood is a hygroscopic material which means that the material absorbs water (Woodprod- ucts.fi, 2017). Wood tends to achieve equilibrium with the surrounding air. This means that the effect of the humidity needs to be considered for when designing a building in wood (University of Cambridge, 2015a).
In a freshly felled tree the ratio between water and dry wood mass is between 60 to 200 %. After a drying process before going out to the market, the wood contains approximately 10 % water. When dry timber is exposed to water the cell walls fills with water and expands, causing an increase in dimension. Apart from dimension changes the cell walls also weakens with an increased water percentage. The hydrogen forms stronger bonds with water which results in that cellulose–cellulose bonds are replaced by cellulose- water bonds. These bonds are weaker making the timber more sensitive to stresses and decreases the stiffness (University of Cambridge, 2015a).
When wood is dried from fresh to 12-15 % humidity content the compression and bending strength double. The tensile strength of wood is at its largest at 6-12 % humidity content (Woodproducts.fi, 2017). Wood begins to suffer damage if its moisture content is over 20 % for longer periods of times and 70 % RH is a critical value since over that, wood begins to rot and decay. With water induced stresses the crack will start in the center of the beam instead of at the surface as the case of a dry beam (Svenskt Trä, 2015).
To estimate the moisture content in the wood Woodproducts.fi (2017) have published a diagram showing the relationship between temperature, relative humidity of air and moisture content in wood, as shown in 3.12.
Figure 3.12: Chart over moisture content in wood in regards to humidity (Woodprod- ucts.fi, 2017)
Although (Woodproducts.fi, 2017) claims that wood begins to suffer damage at a moisture content over 20 %, University of Cambridge (2015a) shows in figure 3.13 that the maximum compressive strength diminishes approximately linearly from 70 MPa to 22 MPa when the moisture content of wood increases from 0 % to 20 %.
Figure 3.13: The relation between moisture content in wood and the maximum compres- sive stresses that can be distributed (University of Cambridge, 2015a).
Over time wood tends to creep, meaning that the material loses some of its strength over time. This phenomenon may also occur with varying moisture content. During two different studies by Angst-Nicollier (2012), samples of wood is exposed to a varying moisture content, wetting and drying the wood, and then stated the average stresses in the wood. The wetting and drying were between 40-80 % RH respectively 50-90 % RH over a period of 12 or 21 days. During wetting the stresses in the cross section is larger than the characteristic tensile strength of 0.5 MPa. The characteristic value is when the wood begins to suffer damages.
Figure 3.14: Average moisture induced stresses after 12 and 21 days of wetting (a) and drying (b)(Angst-Nicollier, 2012).
From these studies the average stresses were plotted in a diagram shown in figure 3.14 showing that when wood is whetted the tensile stresses is exceeding the characteristic tensile strength of 0.5 MPa. Also in figure 3.15 it can be seen that the moisture exposure have a great impact on the tensile strength (Angst-Nicollier, 2012).
Figure 3.15: Modeled stress range after 12 days of wetting in comparison to a reference object(Angst-Nicollier, 2012).
3.4.2 Biological impact
The study of Fürhapper (2016), conducted on pine wood, shows that varying moisture cycles affects the behaviour of emissions of VOC (Volatile organic compounds) when the wood is untreated. In the beginning higher levels of emissions were found but by 21-28
days the affected specimen had lower emissions. Treated wood on the other hand had no significant effects on the emissions. The treated wood was covered in a water-based product in different thickness to create diffusion-open or diffusion-closed. This means that the untreated wood, during the first 20 days after built, can send out deleterious particles. But after that period moisture does not have a bad effect of VOC any more (Fürhapper, 2016).
Another risk with a moist humidity, when building in wood, is the higher risk of mould, fungus and insects. Due to todays’ possibilities of preservative treatment of wood that protects against these harmful species, the durability of wood can be improved, even in moist air (Tsoumis, 1991).
3.5 Heating, Ventilation and Air Conditioning
As stated above the right temperature and humidity is important in a data center. As a part of this, a good ventilation system is important to maintain the right climate. It is important to choose the right cooling system since cooling costs can stand for more than half of a data centers total annual operating costs (Nortek Air Solutions, 2015; Delta Electronics Inc, 2016).
In the data center industry the energy efficiency of the data center is measured in PUE (Power Usage Efficiency).
PUE = Total Facility Energy
IT Equipment Energy (1)
This is a measurement of how much energy the whole building is using in reference to the servers. Many global data centers have a PUE of over 2.0. The goal is to be close to 1.0, the lowest number on the scale (Åberg, 2015).
One way to lower the PUE is using free cooling. Free cooling is a method of using as much outside air as possible to cool the facility without having to cool it further, see subsection 3.5.2. Facebook for example claims that they have a PUE of 1.07, and that is mostly due to their efficient free cooling system (Facebook, 2017b; Åberg, 2015).
The PUE standards have one negative feature, it can’t be lower than 1.0. So more and more companies are starting to also use ERE (Energy Reuse Efficiency).
ERE = Total Energy − Reuse Energy
IT Equipment Energy (2)
This system takes the reuse of waste heat into account. An example of re-usage is district heating (Åberg, 2015). The ERE is also what this thesis addresses, in the way to minimize the use of energy for ventilation and cooling and that the waste heat will be reused for example in the process of drying wood and to power greenhouses.
3.5.1 Mechanical cooling
When cooling and ventilating a larger area or house, mechanical cooling can be an alter- native. In mechanical cooling the outside air is drawn in by fans inside the building where it is distributed by another set of fans. Mechanical cooling may pressurize the building to make the air flow better. Mechanical ventilation systems provide controlled, uniform ventilation throughout the building. Mechanical ventilation may either have the air flow- ing from outside, cool it down and push it through the building and then directly out. But it may also mix inside air with outside air to minimize the need for heating or cooling mechanically (U.S. Department of Energy, 2017).
3.5.2 Free cooling
McFarlane lists three ways of free cooling that right now are the mostly used in data centers right now, according to McFarlane (2017). All three are variations of free cooling that are designed to reduce costs and energy usage.
These three types are:
1. Air-sided free cooling 2. Water-sided free cooling 3. Adiabatic cooling
Air-sided cooling is when the outside air is taken in, via a filter, directly in to the data center and cools down the room.
Water-sided cooling is when the air inside the data center is brought to a room where cold water, often from a lake bed, runs in pipes. The hot air is cooled down by the cold water and the water is heated, exchanging energy.
Adiabatic cooling is a variation of air-sided cooling where you add a little bit of water to the air to lower the temperature.
All three ways sound simple to perform but there are large volumes of air required to cool a data center (McFarlane, 2017; Nortek Air Solutions, 2015).
Figure 3.16: Model of Facebook adiabatic free cooling system(Heiliger, 2011).
Using free cooling makes chillers and compressors redundant. No need for extra piping due to that free cooling uses plenums for distribution, with less piping there are less risk for leakage or installation errors. With free cooling the whole system is located inside the building. The risk of damages on the equipment due to weather changes then disappears (Heiliger, 2011).
Operational advantages with a free cooling system is that there is less likelihood of mixing exhaust air with intake air since they are on the opposite sides of the building (see figure 3.16). As can be seen in figure 3.16, not all hot return air is relieved immediately, the hot air is mixed with the new cold air to maintain the right temperature in the data center without cooling. The amount of air coming in and going out depends on the tem- perature outside and inside of the building. Sensors are measuring the temperature and changes the intake and outtake with an automatic system.
Even though the system is called free cooling this ventilation system is not entirely free. A free cooling system still needs fans, filters, and water to cool the air when the outside temperature is warmer than 25◦C (Fortlax, 2017b). But as stated in the previous section, many construction and operation systems can be removed and reduced. Free cooling can reduce energy costs by up to 52 % and enabling the possibility to become an LEED (Leadership in Energy and Environmental Design) gold certified facility (Nortek Air Solutions, 2015). LEED is a rating system that classifies buildings that are resource efficient. LEED buildings should use less water and energy and reduce greenhouse gas emissions (US Green Building Council, 2017).
To furthermore maximize the capacity of free cooling, hot aisles and cold aisles are preferred to minimize the risk of mixing the air. In Taipei, Taiwan, Delta power solu- tion Inc conducted an experiment of isolating these aisles (figure 3.17). By this method, the efficacy of free cooling is higher and no chillers or row coolers are needed (Delta
Electronics Inc, 2016).
A rule of thumb for data centers is that an increase of temperature in the cold aisle of 1 ◦C leads to an energy saving of about 2-3 %. So, a temperature increase from 18 ◦C to 25◦C, which is in ASHRAES new guidelines (Judge, 2015), can save the data centers energy consumption by 10-14 % (Delta Electronics Inc, 2016).
Figure 3.17: Model of Delta Solutions Inc’s row cooling system(Delta Electronics Inc, 2016).
3.6 Need for research
During the literature studies the lack of research for this subject were found. When search- ing for wood and data center or modular and data center little or no information was found.
As to the information which were found, there are only one data center built in wood, in the world, today. This data center is Hydro66, located in Boden, Sweden.
As stated above, there is a need for flexibility in the data center sector. However, there seems to be little known information about modular data centers. The ones built so far are containers or scalability in a larger, already built location Morgan (2009).
When researching modular architecture, most published material have a focus on res- idential buildings (Lidelöw et al., 2015). The modules produced for residential buildings may not always be applicable on industrial buildings due to bigger buildings with larger spans.
The studies conducted on how moist affect wood is mostly conducted in laboratories in a controlled climate (Angst-Nicollier, 2012; University of Cambridge, 2015a) and also conducted on small specimens and not on larger beams in a construction.
4. Results
4.1 Interviews
Following sections summarizes the interviews conducted. The transcripts from the inter- views can be found in Appendix A.
4.1.1 Börje Josefsson, Sunet
The location visited with Börje is a data center located in a container. This container contains everything needed to run the servers, except from emergency power that were located in an annexed container.
When designing a data center, Börje believes that there should be at least 100 cm in front of and behind the servers for maintenance. The height of the building should be designed with regards to the height of the racks and additional spacing for cables. The electrical cables should be separated due to the risk of interference.
Börje mentions two different ways to cool the servers, downside up or front to back.
Börje recommends front to back due to higher efficiency. This has higher efficiency due to that, when cooling downside up, the air is heated on the way and the servers at the top will be reached by warm air. When cooling front to back the distance for cooling is shorter and the air will not be as warm.
4.1.2 Christiaan Keet, Hyrdo66
Hydro66 is a London based company with a colocation data center located i Boden. The reason for choosing Sweden as base for their colocation is the cheap ground and building prizes as well as green and cheap electricity. Sweden also has a climate that makes it possible to use free-cooling as their ventilation system. Free cooling lowers the power used for keeping a good climate in the building. In Sweden and with a free-cooling system, additional cooling is only needed above 25◦C.
When starting a small company as Hydro66, modularity was an important part of their business plan. Starting small with one building and, as the company grows, extend the building. Hydro66 has a collaboration with Vittjärvshus, a company that manufactures houses, so the design of the building is set. With that system, the longest part of the con- struction process is the time it takes to lay the foundation. In Hydro66 facilities the office building is the most expensive, in this part they not only have meeting and surveillance rooms, they also have bedrooms and a kitchen for their staff.
Christiaan recommends not to have a square data center due to the risk of hot spots that needs extra ventilation. In Sweden, there are low ground prizes which makes it easier to expand sideways and not upwards, Christiaan think this is good for the design of a data center. He also thinks that a data center should be built around a ventilation system and not to adapt a ventilation system for a building.
Hydro66 is uncommon in their view on emergency power. By connecting to two isolated power grids, they do not need emergency power, in the form of diesel generators, since the risk of both sources having problems at the same time is considered minimal.
With that said, if a company demands emergency power it can be arranged outside the building. Hydro66 focuses a lot on the environment and a big marketing point for them is the green power delivered in Sweden by water.
4.1.3 Tor Björn Minde, RISE SICS North
SICS is a research institute belonging to the ICT-sector of RISE. The purpose of SICS North is to develop and research the ICT-sector for a long-term competence development.
Tor Björn thinks that it is hard to know if it is the user that drives the development of the of the sector or if it is the technology. The human never settles and always want more advanced and cheaper equipment. This development will continue for a long time to come.
Tor Björn believes that there are two ways to build a data center. Either you built big at once, like Facebook in Luleå, or you build smaller, modular buildings. He believes that both ways are equally good and the choice is dependent on the company’s economy and goals.
When building a data center, Tor Björn thinks that, switchgear, emergency power, and batteries should be separated from the servers, due to fire safety. Also regards should be taken to the work area in front of and behind the server racks, this distance should be at least 120 cm.
4.1.4 Pär Åberg, Swedish Modules
Pär Åberg considers modularity to be the possibility to prefabricate buildings where you need flexible solutions and minimal work on site. Swedish Modules mount all technical equipment in the factory before delivery, this leads to shorter production times and gives
the possibility to perform quality testing before shipment. Swedish Modules have a de- livery time between 8 and 12 weeks and on-site work of around 4 days depending on the size of the building. This can be compared to about 2-3 month if all work is done on-site.
When choosing the size of the building and the modules, transportation is a limit. An ordinary size for Swedish Modules is 4.1 x 4.5 x 13.6 m since this is the largest module that can be transported all over the world without any bigger problems in transportation.
Pär thinks that the industry needs to be more standardized and modular in the future.
He compares the data center industry with the car industry where the car industry manu- factures Teslas whilst the data center industry still manufactures the T-ford. This is due to the fact that in the data center industry every building is usually customized.
4.2 Site visit
Three tours were made at Sunet, Hydro66 and SICS, and also a virtual site visit of Face- books data center in Luleå. All four facilities are different and have different purposes.
4.2.1 Sunet
Sunet stands for Swedish University computer Network and is a company that is respon- sible for providing Sweden’s universities and colleges with access to national and inter- national well-developed data communication (Sunet, 2017).
The location Sunet let me visit is a data center located in a container. Everything except power, UPS and emergency power is in one container. The power equipment is placed in an annexed container. The container is made in steel and is delivered ready to use from a manufacturer. One problem in this container is that there are too small spaces to be able to maintain the servers properly.
4.2.2 Hydro66
Hydro66 is a colocation company that supplies high end customers with a sustainable, green and secure location for their data information. Hydro66 data center is designed to address the big changes in the sector for the next 20 years. It is built with environment and customer in mind and it is scalable and efficient at a low cost (Hydro66, 2017).
Hydro66 is among the first in the world to use a wooden structure as their data cen- ter. Although London based, Hydro66 uses local businesses to build and develop their Swedish colocation. The building envelope is a house ordered from a turnkey manufac- turer that is specialized in villas. Using free cooling and local sustainable energy supplies, such as water power, the company focuses a lot on the environmental footprints. With no emergency power as standard, but the using of two different power sources, this building
is unique in the data center industry. Emergency power could be connected if demanded by the customer.
The data center is relatively small compared to known companies as Facebook and Apple for example. They have chosen to start small and expand by building one house, at first only containing one pod, and expanding as customers joins Hydro66. The customers can rent from one rack to a full house or more. Hydro66 first server hall is now half full but expanding with new pods. They are also in the midst of building a new server hall.
One smaller problem that has been discovered is that the ceiling is a bit low to easily maintain the power cords above the server racks.
4.2.3 RICE SICS North
RISE SICS North is a part of RISE SICS that is a leading research institute for applied information and communication technology. SICS is a non-profit organization that per- forms advanced research in collaboration with Swedish and international industry and academia (RISE SICS, 2017).
At SICS North they now have two smaller server rooms (Modules) for research. One module is used as a data center where they monitor different IT-loads, energy use, tem- peratures, and equipment. The second one were still under construction. In the second server room test of different airflow methods will be conducted. Raised floors makes it possible for cooling from underneath. Raised floor technique that was used a couple of years ago and RISE SICS North now want to conduct more research on.
These two rooms are built inside a larger building that used to be a logistic industry.
Around the server rooms there are equipment as UPS batteries and fire safety and a display area for demonstration of the processes. The industry hall is big enough to accommodate one or two additional server rooms.
4.2.4 Facebook
Facebook is one of the largest and most known companies in the world. Facebook’s mission is to give people the power to share and make the world more open and connected (Facebook, 2017a).
The data center in Luleå is Facebook’s first data center outside the United States.
Luleå were chosen for the cold climate, suitable for free cooling, and the green electricity from water power (Waugh, 2011).
The data center in Luleå is a 300 x 100 m big building using free-cooling as their cool- ing system. (Facebook, 2017b) Facebook’s data center is built with the same standards around the world and they are optimized for Facebook’s usage. Facebook is also one of the first data center companies that have published their research and technical specimens for the world to read and learn from (Heiliger, 2011).
4.3 Measurements
Where the sensors were mounted in Hydro66 and the measured data are shown below.
The sensors measure temperature and humidity for intake and waste heat. For this thesis sensors ID 6 and ID 7 has been analyzed in all graphs, where ID 6 represents intake air and ID 7 represents waste heat. In figures 4.1 and 4.2 the placement of the sensors is shown.
Figure 4.1: Placement of sensor at intake (Sandberg, 2017).
Figure 4.2: Placement of sensor at waste heat (Sandberg, 2017).
In July, figures 4.3 and 4.4, the intake humidity and temperature varied a lot. Humidity were between 90-100 % RH and a temperature between 24-26◦C during the day and 35- 65 % RH and 8-16 ◦C during the night at the intake. Nevertheless, the waste heat was quite steady at 20-40 % RH and a temperature between 34-38◦C.
Figure 4.3: Humidity 19-26 July (Sandberg, 2017)
Figure 4.4: Temperature 19-26 July (Sandberg, 2017)
In January, figures 4.5 and 4.6, the intake humidity and temperature were a lot more stable during the day and night. Humidity were between 70-80 % RH and temperature between -8 - -2◦C during the day and 80-95 % RH and – 28 - -12 ◦C during the night at the intake. The temperature and humidity of the waste heat varied more. During the time the waste heat temperature were between 10◦C to -10◦C, the relative humidity was around 15-20 % RH. When the temperature then dropped to -10 - -15 ◦C the relative humidity rose to 60-66 % RH.
Figure 4.5: Humidity 2-8 January(Sandberg, 2017)
Figure 4.6: Temperature 2-8 January(Sandberg, 2017)
The values for July and January have been summarized in table 1 and 2 below.
Table 1: Table over air intake
Period Day/Night Temperature◦C Relative Humidity %
20 - 25 July Day 24 - 26 90 - 100
20 - 25 July Night 8 - 16 35 - 65
3 - 6 January Day -8 - -2 70 - 80
3 - 6 January Night -28 - -12 80 - 95
Table 2: Table over waste heat air
Period Day/Night Temperature◦C Relative Humidity %
20 - 25 July Day 34 - 38 20 - 40
20 - 25 July Night 34 - 38 20 - 40
3 - 6 January Day -10 - 10 15 - 20
3 - 6 January Night -15 - -10 60 - 66
Looking at the whole period, in figures 4.7 and 4.8, there are a lot of variations during the day and nights. Although a trend is visible in both temperature and humidity. The intake humidity varies around 80-100 % RH through the whole year whilst the waste heat humidity varies more. The trend is between 20-40 % RH with some big peaks of variations of up to 60 % RH during a day or two. Looking at the temperature diagram these peaks of variation seems to be due to sudden temperature drops, deviant from the trend. The temperature of the intake air is from July to November steadily decreasing from 20◦C to -12◦C. The variations during this period is about 4-6◦C from one day to another. During the winter from November to February, when the study is summarized for this thesis, the temperature is varying more. From one day to another the temperature can differ up to 14 ◦C, varying from -12◦C to 2◦C. With a low temperature peak in the beginning of January at -25◦C. The waste heat temperature is following the intake curve well, steadily decreasing from 36◦C to 26 ◦C from July to November. From November
to February the waste heat temperature is anything but steady, varying from 28◦C to -8
◦C from one day to another.
Figure 4.7: Humidity July to February (Sandberg, 2017)
Figure 4.8: Temperature July to February (Sandberg, 2017)
4.4 Design proposals for a data center
4.4.1 Limitations
When designing a data center, some specifications needs to be considered before building.
These limitations are based on information given during the interviews and own values
gathered during the site visits. Some of the limitations are also supported in Telecommu- nications Industry Association (2012) and ASHRAE (2016) that are American national standards related to how to design a data center.
• Choose the cooling system before designing the building because the cooling is a big issue in data centers and the building should be designed to minimize costs of ventilation and maximize the cooling. A quadratic building is harder to cool, resulting in hot spots that can be devastating for the servers.
• The choosing of building system may limit the building design in the regards to floor plan solutions and cannot be built with the spans of at least 8 meters that is required for a data center.
• In front of and behind the servers a minimum distance of 1200 mm is required to be able to serve the servers and other electrical equipment.
• Extra ceiling height should be added to the building. Often the cables are wired above the racks where they need to be maintained. Also, the high voltage cables and low voltage cables needs to be separated due to interference. Internet fibre cables should be placed separate from the electric cables to streamline installation processes.
• A rack is 600 x 1000 x 2200 mm (w x d x h).
• To get a reduced tax from the Swedish government the effect must exceed 0.5 MW for equipment not including fans and cooling.
• Apart from the server hall an annex, in a different fire zone, is needed for switch gear, back up batteries and possibly generators. Also, offices for supervision and surveillance as well as break room is needed for the 24/7 personnel needed to main- tain security.
4.4.2 Chosen modules and measurements
Dimensions for the building were decided with regards to the theory about building sys- tems, data center types and HVAC as wells as the limitations stated in chapter 3 and subsection 4.4.1. Also discussions with two data center technicians (Jatko and Larsson, 2017) were conducted during the design process to conclude a relevant result. Placing 8 server racks in a row and 16 racks in a pod and then placing 4 pods in one data cen- ter module, to keep the building small but still cover the governments requirements for tax reductions at 500 kW (Vattenfall, 2017). These 64 server racks have an approximate capacity of 650 kW (one server demands approximately 250 W).
The with of the building is the width of 8 racks, each with a width of 600 mm, ventila- tion ailes and service spaces. All leading to a minimum width of 11 m inside. The length of the building is 8 racks, each with a depth of 1000 mm plus 9 service spaces of 1200
mm, giving a minimum length of 20 m. This gives a house module with minimum inner measurements of 11 x 20 m.
Figure 4.9: Floor plan description
The design of the building will be pre-fabricated in smaller building around 11 x 20 meters and where you can expand with the same modules to the companies liking.
Every module has its own ecosystem, the first built module can supply the second module with power and water (for ventilation). The modules will be connected to each other by a service corridor. For this example, the modules are placed in a row, which will be the easiest design, but this can be changed to the customers liking and to best suit the building plot.
Figure 4.10: General plan
The height of the building is limited by the height of the racks, 2200 mm, and the space needed for cables and maintenance. To maintain a high air pressure to minimize the use of fans the ceiling should be at the same level all through the building. A change in ceiling heights makes it harder to keep a constant airflow since a change in volume demands a change in air volume. This makes the minimum height of the outer walls 4 m, giving 1.7 m for cables and maintenance, see figure 4.11. After communication with (Lindgren, 2017) the ceiling height was altered to an industry standard of 4.5 m. This choice is made in regard to that the building could be used by another company if the data center decides to move.
Figure 4.11: Section
Figure 4.12: 3D View
Figure 4.13: 3D View
