‘Assessing the attractiveness of port areas to datacenters’

(1)

‘Assessing the attractiveness of port

areas to datacenters’

Hylke Zijlstra

University of Groningen

Master thesis

Business Administration; Operations & Supply Chains

July, 2012

(2)

(3)

Abstract

Datacenters are increasingly locating in seaports, but why they are attracted to seaports is not clear. The aim of this thesis is to identify what makes port areas attractive to datacenters and to develop a model that can assess the attractiveness of port areas to datacenters. Therefore location factors of datacenters were identified through a literature review, a case study of the port authority of the ports in Delfzijl and the Eemshaven and by conducting an expert interview. Thirteen factors were identified and ranked in order of importance by datacenter professionals. Of these thirteen the eight most important were used to calculate weights to the factors. These eight factors formed the base of the final model assessing the attractiveness of specific areas to datacenters. The most important factors are the presence of a connection to an internet exchange, the reliability of the energy net and the presence of cables with sufficient capacity. With this information available, the attractiveness of the three port areas in the Netherlands that host a datacenter, Amsterdam, Rotterdam and the Eemshaven, was assessed. Next to that, the model can be used by other areas that want to attract datacenters to assess on which areas they should try to improve to be seen as a suitable location for datacenters.

Keywords

Datacenters, Location Factors and Seaports.

Supervisor Prof. Dr. I.F.A. Vis

Second assessor

(4)

List of tables

Table 1; Sub questions and methods ... 14

Table 2; List of location factors after literature review ... 22

Table 3; List of location factors after literature review and case study ... 29

Table 4; List of location factors after literature review, case study and expert interview ... 31

Table 5; Final list of location factors ... 33

Table 6; Categorization of location factors ... 33

Table 7; Calculation of importance of categories in conceptual model. ... 34

Table 8; Response characteristics ... 36

Table 9; Survey results ... 37

Table 10; The eight most important factors ... 37

Table 11; Results of AHP method ... 39

Table 12; Final categories and weights ... 39

Table 13; Translation of factors ... 55

Table 14; Results of AHP method (copy of table 11)... 56

(7)

List of figures

Figure 1; Research methodology ... 12

Figure 2; Structure of the thesis ... 15

Figure 3; Porter's diamand (Porter, 2000)... 20

Figure 4; Geographical position of the seaports of Groningen Seaports ... 25

Figure 5; Short description of seaports (www.groningen-seaports.com) ... 25

Figure 6; Conceptual model ... 34

Figure 7; Response characteristics ... 36

(8)

1. Introduction

1.1 Introduction

The market environment in which seaports are operating has changed from a stable to a more dynamic environment. There is a change from an era characterized by economies of scale and mass consumption of standard products, to a new era which is characterized by global corporations, outsourcing, deregulation and technological innovation (Notteboom & Winkelmans, 2001). This change has some implications for port authorities. Port authorities should look beyond their port boundaries in terms of physical investments and managerial capabilities to have different functions and different focuses (Notteboom & Winkelmans, 2001). Many distinctions can be made between ports like, for example, sea-born versus non-sea-born trade or maritime clusters versus industrial clusters (Bichou & Gray, 2005). Within the clusters in a port area a distinction can be made in sectors like cargo handling, transport, logistics, manufacturing and trade which can all be classified as traditional port activities (De Langen, 2004).

When looking at the goals of port authorities, two main goals can be identified which are; ‘to facilitate a sustainable economic development of the port as a whole’ and ‘to become an efficient and effective organization that generates income to cover costs, to make investments and, in some cases, to return to shareholders’ investment’ (Van der Lugt & De Langen, 2007, p. 5). It is surprising to see that there hasn’t been much research done in this area, because generating income might be able by attracting companies with non-traditional port activities as well, thereby stimulating the growth and prosperity of port areas. This research can broaden the view of port authorities, which is necessary because focusing only on traditional port activities means that other opportunities of making profit are not identified.

(9)

increasingly efficient and have more space left to develop other profitable businesses (Paixão & Marlow, 2003).

One of the industries that is not related to traditional port activities, but is developing in port areas, is datacenter computing. Datacenters, also called ‘data hotels’ or ‘data warehouses’, are buildings in which a huge amount of Information Technology (IT) hardware is stacked together. All these servers together use a vast amount of electricity. The hardware in datacenters has a temperature range in which it performs best. Because the hardware produces a large amount of heat itself, one of the biggest issues for datacenters therefore is the cooling of the building.

The Netherlands seems to be attractive to datacenters (Eielts, 2011). When looking at the ports in the Netherlands for example, the port area of Amsterdam facilitates ten datacenters, the port area of Rotterdam four, and the port area of the Eemshaven one. This means that around 10% of all the datacenters in the Netherlands are located in ports (Datacentrumgids, 2012). But also surrounding countries facilitate datacenters in port areas like for example the port area of Antwerp and Hamburg with both three datacenters (Datacentermap, 2012; Van Beekum, 2007).

To the best of our knowledge there is no scientific literature on location factors for datacenters. Because of the specific characteristics of datacenters the general location factors, like for example tax rates, land costs, presence of strong competition and wage rates, will not all apply equally well and other factors are likely to be introduced. In other industries research has been conducted on specific location factors like, for example, for higher educational institutions for which the distance to another competing institution are important (Bajerski, 2011); international resort parks for which zoning limitations are the most specific location factors (Lin & Juan, 2009); the glass industry which depends on the availability of skilled craftsmen (Brown, 1980); the medical device industry which also depends largely on the level of skill and education in the region (Kimelberg & Nicoll, 2012) and ICT firms which need a location that is easily accessible by car (Atzema, 1992).

(10)

datacenters, and when they score low they might be better off focusing on other industries or improve the attractiveness of their area on the factors on which they score low.

Part of the research is a literature review, a case study at Groningen Seaports, the port authority of the Eemshaven and Delfzijl, and an expert interview with a manager of a datacenter. When the location factors are determined, surveys will be used to ask datacenter professionals which factors they perceive as the most important.

1.2 Problem definition

1.2.1 Problem definition

Some ports, like for example the Eemshaven in the Netherlands, do not focus on traditional port activities but do support an industrial area. The kind of economic activities that take place in these areas depends on the location of these ports and what resources are present in these ports. The resources can be tangible as well as intangible and can range from machines to employees. There are more external factors which determine the attractiveness of an industrial area, whether it is located in a port area or not. For example, the wage rate in the region, availability of labor, level of skills of the potential employees, available raw materials, transportation facilities, customers, tax rates, zone requirements and land building costs (Hitt et al., 1983).

As discussed in the previous chapter, datacenters are increasingly locating in port areas. Datacenter computing is growing rapidly (Lam et al., 2010; Viswanathan et al., 2011), but still there is not much published about this specific industry, nor why they are locating in port areas. Because of its specific and unique characteristics it is to be expected that datacenters have very specific requirements which are not comparable with the requirements other companies or industries demand of their location.

(11)

1.2.2 Research goal

The first goal of this master thesis is to provide more insight in the location factors that are of importance to datacenters and the second goal is to assess the attractiveness of port areas to datacenters. This will result in a model which can be used to assess the attractiveness of port areas to datacenters.

1.2.3 Research question

Following the preceding problem definition the research question can be defined: What factors make port areas attractive to datacenters?

1.2.4 Sub questions

To be able to give a complete answer to this question, the following sub questions should be answered as well.

1. What are specific characteristics of datacenters? 2. What are important location factors for datacenters? 3. What are specific characteristics of industrial areas in ports?

4. What is the role of port authorities in attracting new businesses and industries to their area? 5. How can the factors influencing the location decision of datacenters be captured in a model? 6. How does the model function when applying it to Groningen Seaports?

7. Why are datacenters attracted to port areas in the Netherlands?

1.3 Methodology

To be able to answer the research question as presented in Chapter 1.2.3, the sub questions presented in Chapter 1.2.4 should be answered. These sub questions are in many ways different so they need their own approach to be answered. In this chapter those approaches are discussed.

1.3.1 Research methodology

A literature review is conducted, as well as a case study at Groningen Seaports. The case study is of an explorative nature which is a good instrument to discover relevant factors and issues that might apply in other similar situations (Myers, 2009). Next to that, empirical data is gathered through an expert interview with a Chief Technology Officer (CTO) of a datacenter and financial information of datacenters is analyzed. After the literature review and the empirical data gathering, the sub questions and main research question have been answered and conclusions are drawn. Based on that, the model is developed.

(12)

are then used in an Analytic Hierarchy Process (AHP). By using this method weights are assigned to the eight factors, more information on this method can be found in Chapter 5.2.1.

In Figure 1 the research methodology is presented graphically. The blue squares represent the initial inputs, the green rectangles are the first deliverables of this thesis and the red rectangles represent the final deliverables of this thesis. The orange oval is the validation of the first final deliverable. The research will consist of three steps. First of all a literature review is conducted, that will produce a preliminary list of location factors. After that a case study at Groningen Seaports and an expert interview will expand this list and will support some of the factors found in the literature review. After this the list of location factors is completed. These factors will then be ranked by datacenter professionals, which will form the basis for the final model.

Interview with datacenter manager Literature on port areas Literature on location factors Literature on datacenters Case study GSP List of characteristics of datacenters List of characteristics of inudstrial areas in ports List of factors influencing location decisions of datacenters Validation by experts Model assessing the

attractiveness of port areas

Figure 1; Research methodology

1.3.2 Sub questions

(13)

can be found in Chapter 3.2.1. Some (public) financial information of datacenters is analyzed as well, however this did not led to new insights.

The second sub question, ‘What are important location factors for datacenters?’, cannot be answered with the help of literature only. However, some general location factors that seem to apply to datacenters have been identified in literature. Next to this, the case study of Groningen Seaports again led to other factors. The expert interview provided support for some of the location factors that were found in the literature study and the case study.

The third sub question, ‘What are specific characteristics of industrial areas in ports?’, is mainly answered by conducting a literature study. Next to the literature study the case of Groningen Seaports is studied. This case provides helpful information because the Eemshaven is a typical port where non-traditional activities are performed.

The fourth sub question, ‘What is the role of port authorities in attracting new businesses and industries to their area?’, is answered by conducting a literature study and studying the case of Groningen Seaports.

The fifth sub question, ‘How can the factors influencing the location decision of datacenters be captured in a model?’, is answered by using the list of factors produced based on the preceding sub questions as input. From these factors a selection is made, using surveys filled in by datacenter professionals, this survey can be found in Appendix A. The importance of the factors is determined by using the AHP method, which is described in more detail in Chapter 5.2.1. This will then be covered in a model.

The sixth sub question, ‘How does the model function when applying it to Groningen Seaports?’, is focusing on practice. Groningen Seaports is trying to attract more datacenters to the Eemshaven and already attracted one datacenter, next to that, Groningen Seaports was willing to share information and for these reasons is a good case for this thesis. The model produced based on the fifth sub question can therefore be used to assess the attractiveness of the Eemshaven area and provide insight in possible areas for improvement for the Eemshaven. This sub question will test the usability of the model for port authorities.

The seventh sub question, ‘Why are datacenters attracted to port areas in the Netherlands?’, will be answered by comparing the model, and the factors it consists of, with the port areas of Amsterdam, Rotterdam and the Eemshaven.

(14)

Sub question Source

1. What are specific characteristics of datacenters? Literature review, case study and expert interview.

2. What are important location factors for datacenters? Literature review, case study and expert interview.

3. What are specific characteristics of industrial areas in ports?

Literature study and case study. 4. What is the role of port authorities in attracting new

businesses and industries to their area?

Literature study and case study. 5. How can the factors influencing the location decision

of datacenters be captured in a model?

Surveys and use of AHP method to assign weights to factors.

6. How does the model function when applying it to Groningen Seaports?

Information of Groningen Seaports. 7. Why are datacenters attracted to port areas in the

Netherlands?

Information from port areas. Table 1; Sub questions and methods

1.3.3 Literature

As already mentioned, there is not much literature available on datacenters. This means that this thesis will be a welcome addition to fill the literature gap but it also implies that it is hard to answer some of the sub questions with a literature study only.

When it comes to ports there is more literature available. For example there is literature to be found on current trends in port areas (Notteboom & Winkelmans, 2001; Paixão & Marlow, 2003; Verhoeven, 2010) and characteristics of ports and port authorities (Baccelli et al., 2008; Suykens & van de Voorde, 2006; Van der Lugt & De Langen, 2007; Zondag et al., 2010).

To find the most relevant literature multiple databases have been consulted like ‘Business Source Premier´, ´Web of Knowledge´, ´Porteconomics.eu´ and ´Google Scholar´.

1.3.4 Validation

The validation and the development of the model will be done by asking datacenter professionals to fill in a survey to rank the location factors found according to the importance. The eight most important factors will be used in the model, this way the least relevant factors will not be included in the model and there is consensus on the most important factors.

(15)

1.3.5 Structure of thesis

The thesis starts with an introductory chapter in which the relevance and aim of the research is discussed. In Chapter 2 a more thorough literature review is conducted which also functions as a starting point for the case study at Groningen Seaports, the expert interview with the manager of a datacenters and the analysis of the financial data of some datacenters, which are presented in Chapter 3 and Appendix B. All this information is combined and discussed in Chapter 4, in which the conceptual model is introduced. In Chapter 5 the final model is presented. In Chapter 6 three port areas in the Netherlands are assessed and Chapter 7 will provide a discussion, conclusion and suggestions for further research. The structure is visualized in Figure 2.

Figure 2; Structure of the thesis Empirical research Non-empirical research Ch. 2 Literature review Ch. 1 Introduction Ch. 3.2 Analysis of expert interview Appendix B Analysis financial structure of datacenters Ch. 3.1 Case study GSP Ch. 4 Conceptual model Ch. 5 The model Ch. 6 Implications for port areas

Ch. 7 Discussion, conclusion and

(16)

2. Literature review

This literature review will try to provide an answer to the sub questions regarding specific characteristics of datacenters and industrial areas in ports, to provide insight in location factors and to assess the role of port authorities in the development of their industrial areas.

First of all the characteristics of datacenters are addressed because knowing these might provide a context for the reader when reading the remaining paragraphs. After the characteristics of datacenters, the characteristics of industrial areas in ports are addressed. Then general location factors are discussed where after the role of port authorities in the development of their port area will be explained, the latter is relevant because port authorities might play a role in influencing the attractiveness of an area for a certain industry by, for example, incentives or favorable regulations. The chapter ends with a conclusion.

2.1 Characteristics of datacenters

Datacenter computing is a relatively new industry which became more in use with the advent of the internet and is getting increasingly popular with the current hype of cloud computing. A datacenter typically is a facility supporting a lot of hardware on which data is stored, hence the term ‘data’ center. Much of the popular internet applications and social media have to store their data which should be accessible to their users whenever they want. More often people do not want to store data on their own laptop or personal computer but rather store it ‘somewhere’ online so that they can access it from anywhere in the world. All this data needs to be stored somewhere, and that is where datacenters come in.

There are two kinds of datacenters to be identified. So called single-tenant and multi-tenant datacenters. A single-tenant datacenter has only one customer, so all the space and racks are in use by one specific company. In a multi-tenant datacenter multiple companies use the space in a datacenter.

(17)

All this energy consumption makes the air conditioning inside the datacenter facility very important because besides the high energy it consumes, the hardware in the datacenters produce heat. Therefore datacenters need advanced cooling mechanisms to reduce the temperature in the datacenter to an acceptable level. What is seen as an acceptable level is determined by the temperature in which the hardware performs optimal. This optimal range is between 18 and 27 degrees Celsius and an acceptable range is between 13 and 32 degrees Celsius. The humidity should also be regulated with an optimal humidity range between 25 and 60 percent and an acceptable humidity range between 24 en 75 percent (Vijaykumar, 2011). That cooling is an important aspect in a datacenter can be derived from the power consumptions as well, sixty percent of the consumed energy is needed for the cooling of the datacenter (Robb, 2005).

Since datacenters transfer a lot of information every second, the datacenters need to have a good connection with enough bandwidth. Especially because technologies in the ICT branch are improving in a rapid pace and datacenters are expanding, datacenters have and need to have a good connection which can handle the large amount of data transfer (Lam et al., 2010; Robb, 2006). When for example the processors of the computers in a datacenter are upgraded they can process more information per second, but when the connections of the datacenter do not allow more transfer of data per second the upgrades of the processors have no use.

To keep all the processes going and online, datacenters need good employees which do not only need to have technical abilities but should also have good judgment and communication skills. These are often hard to find which makes the high turnover rate amongst personnel in datacenters even more inconvenient. The high turnover rate is mainly due to the pressure to keep the datacenter up and running no matter what. Due to the increase in the number of datacenters it will probably only be harder in the future to attract qualified employees. (Sullivan, 2008).

The specific characteristics mentioned are presented in the list below as well.

 Large amount of energy used.

 Much heat is produced.

 Advanced cooling mechanisms.

 High demands on the (internet cable) connections of the datacenter.

 Good employees are hard to find and need technical skills as well as communication and judgment skills.

(18)

2.2 Characteristics of industrial areas in ports

Ports can be classified according to a few concepts, namely trade related; sea-borne versus non-sea-borne trade, space related; hinterland versus foreland, national versus international, feeder versus hub and sector-related; direct/indirect versus induced, maritime clusters versus industrial clusters (Bichou & Gray, 2005). The preceding concepts influence the industrial area of a port as well. A port that is mainly a hub and focuses on the hinterland will attract other industries than a port that focuses more on the foreland. Maritime clusters, focusing on activities like shipping, shipbuilding and maritime services, are geographically concentrated in a number of maritime clusters. This concentration in clusters is likely to increase due to internationalization (De Langen, 2002).

There are some specific entry barriers for industries in seaports which can be classified in economic entry barriers, regulatory and institutional entry barriers and locational entry barriers. Economic entry barriers are due to the fact that existing companies will most likely already have the best positions in the port and when the market in which the company operates has a large minimum efficient scale (MES), which is the smallest amount of production of a company while still taking advantage of economies of scale, compared to market size, entrants will face a competitive disadvantage. Finally, existing tenants may benefit from accumulated public investments. Regulatory and institutional barriers are caused by government involvement. Port authorities or (local) governments can issue special regulations which can make it hard for companies to locate in that specific area, in Chapter 2.4 the role of port authorities will be discussed in more detail. Locational entry barriers are caused by the unavailability of land. In many ports there is no space left to facilitate more companies (De Langen & Pallis, 2007). When land is scarce, this will drive up the prices of available land in port areas as well.

Another characteristic of industrial areas in ports is the availability of water and mooring opportunities. This means that industries needing a large amount of cooling water or access to water will have more benefits to locate themselves in a port area. Also, to be able to transfer the cargo or passengers most seaports will have good road and railway connections. Furthermore it has been mentioned that the land price in industrial areas in ports are relatively high and all the best positions in the area are already taken most of the time.

The specific characteristics of industrial areas in ports are presented in the list below as well.

 Best positions in the area are already occupied.

 Lack of free space to build.

 High land price.

 Mooring and water opportunities.

(19)

 There is one responsible authority that governs the area. In the next paragraph general location factors will be discussed.

2.3 Location factors

Location factors are factors that influence the decisions of companies about where to locate their business. These factors can be for example the prevailing wage rate in a region, the availability of labor and required skills, proximity to raw material sources, transportation facilities and customers, tax rates, zone requirements and land and building costs (Hitt et al., 1983).

There are many publications on location factors, even originating from the 19th century. Marshall (1890) wrote about how companies benefit from external economies of scale by locating in the vicinity of other companies. He made a distinction in three types, the first are economies resulting from access to a common labor market and shared public goods like the infrastructure and educational institutions. The second type are economies saved from transportation and transaction costs due to the regional proximity of firms along the supply chain and the third type are economies resulting from knowledge spillovers (Falck & Heblich, 2008). Although these external economies of scale still seem to apply in some way, the globalization and demand for more customized products makes an addition to the theory of Marshall welcome.

Since the publication of Marshall (1890) standard production factors such as raw materials and skilled workers are becoming less important (Falck & Heblich, 2008). More modern location factors include the availability of venture capital, an overall supportive business infrastructure and institutions and highly skilled workers who can think cross-disciplinary and are able to think ‘out of the box’ and exploit their creativity (Falck & Heblich, 2008).

(20)

good employees (regional level). When the decision for a region has been made the company has to choose between different industrial areas, this choice will for example be based on the land price, nearby companies and congestion rates of the traffic (local level). So companies use different kinds of location factors in different stages of their search for a good location. (Atzema et al., 2002)

The diamond model, as presented in Figure 3, describes four categories of factors influencing the competition within an industry but also influencing the location decisions of companies (Porter, 2000). The factor (input) conditions can contain natural resources, human resources, capital resources, physical infrastructure, administrative infrastructure, information infrastructure and scientific and technological infrastructure. Related and supporting industries describe the presence of capable, locally based suppliers and the presence of competitive related industries. Demand conditions are determined by sophisticated and demanding local customers, customer’s needs that anticipate those elsewhere and unusual local demand in specialized segments that can be served globally. And the context of firm strategy and rivalry is determined by a local context that encourages appropriate forms of investment and sustained upgrading together with vigorous competition among locally based rivals (Porter, 2000).

When summarizing the diamond model in a few lines, the most important factors influencing the location of a company is the presence of important resources and important infrastructure, the presence of capable suppliers,

presence of demand, and the presence of competition.

Location factors also depend on the type of industry, one factor will be of more importance for an industry than another factor might be. However, some general important factors are presence of important resources and important infrastructure, the presence of capable suppliers, the presence of venture capital, the availability of qualified employees, the presence of demand and the presence of

competition. Adding to this, the wage rates and tax rates will be of importance as well but mainly for international companies since they will differ more between countries than within countries.

Figure 3; Porter's diamand (Porter, 2000) Factor (input) conditions Demand conditions Related and supporting industries Context for firm

(21)

In this research the regional and local location factors will be most important.

2.4 Port authorities and their role in area development

A port authority is responsible for the management of their specific port area. An important issue in dealing with port authorities is the role of the governments of the countries in which the port authority is located. The port authorities can receive support of the state on for example the financial level. Governments might want to help or influence port authorities for different reasons like economic protection, economic development or even the public good (Suykens & van de Voorde, 2006).

To the best of our knowledge there are not many publications on the role of port authorities. One classification of the role of port authorities that is known is the one by Verhoeven (2010) which will be the main source for this paragraph.

The role that port authorities have in the development of their area depends heavily on the governance structure of the specific port authorities. A classification can be made of three main functions of port authorities (Verhoeven, 2010). The first function is that of the landlord. Port authorities in this category will be responsible for the management, maintenance and development of the port estate and are responsible for the provision of proper infrastructure and facilities as well. Typical issues for these port authorities are on the level of what kinds of infrastructure should be available, like dedicated or non-dedicated, and how to develop their area in terms of real estate. The second function a port authority can have is that of the ‘regulator’. These port authorities play a much more passive role in the development of their port area. The main focus is on controlling, surveillance and in some cases policing functions. These port authorities focus on ensuring the safety and security of ship and cargo operations and enforcing applicable laws and regulations on these fields but also on for example environmental protection and labor regulations. In some cases they can develop their own regulations and employ their own police force. The last function is the ‘operator’. This group of port authorities provide port services like the physical transfer of goods and passengers between sea and land and the provision of technical-nautical services. However, these services are increasingly privatized and the role of the operator port authorities is decreasing (Verhoeven, 2010).

(22)

However, the port authorities with a ‘regulator’ function can also have much influence, albeit more indirect by for example issuing rules or policies on the level of labor or environment, favoring or opposing certain industries (Verhoeven, 2010).

2.5 Conclusion

Some specific characteristics of datacenters are the high energy usage, demanding cooling needs of the facility, a high demand on the connection infrastructure due to the large amount of data transfer, a relatively low number of employees and there are large amounts of sensitive hardware present in the buildings.

Industrial areas in ports have some specific characteristics like the lack of available land, the high costs of land, the presence of water and mooring facilities, the presence of good road and railway connections and the high entry barriers for new companies, mainly due to the fact that already present companies have the most favorable positions.

The role of port authorities in the development of their area seems to differ between port authorities. Some port authorities have more responsibility and power than others, mainly due to the role governments gave them.

After this chapter an intermediate list of location factors can be developed as visualized in Table 2. In the next chapter the empirical research will be discussed and more factors and columns will be added to the list.

Location factors

Availability of internet cables with sufficient capacity Availability of land (to build and, perhaps later on, expand) Availability of skilled employees

Distance to educational institutions Favorable zone requirements Land price

Prevailing wage rates

Redundancy of energy supply Reliability of the energy net

(23)

3. Empirical research; Case study & expert interview

In this chapter the empirical part of the thesis will be discussed. Starting with the case study at Groningen Seaports, after which the expert interview will be presented. After each main paragraph the list of location factors, as presented in Chapter 2.5, will be extended with new factors. Another part of the empirical research is the analysis of the financial structure of datacenters. This analysis provided no new insights and has therefore not been included in this chapter but can be found in appendix B for those who are interested.

3.1 Case study

Part of the case study are meetings with four employees of Groningen Seaports, who are responsible for the development of the area and the advent of new companies or industries and, more specifically, datacenters. Three of these employees are part of a project group focusing on datacenters in the Eemshaven area. Next to that, documents are analyzed concerning the development of the area and the specific companies. The initial contact person at Groningen Seaports is Ms. M. Zwerver, the marketing director of Groningen Seaports. The case study will be used to provide an answer to the following sub questions;

 What are important location factors for datacenters?

 What are specific characteristics of industrial areas in ports?

 What is the role of port authorities in attracting new businesses and industries to their area?

3.1.1 Case study protocol

In this section the case study protocol will be presented which follows the main topics as suggested by Yin (2009). Therefore the protocol will start with a short introduction where after the data collection procedures and the outline of the case study report will be described, after which the main questions of the case study are introduced.

3.1.1.1 Introduction to the case study

(24)

in the area. Concluding, the case of Groningen Seaports can help to find more location factors for datacenters and provide more insights in industrial areas in ports and the role of a port authority in those areas.

3.1.1.2 Data collection procedures

The case is described by making use of multiple sources. First of all the marketing manager, Ms. M. Zwerver, was interviewed. During this interview open questions were asked about the role of Groningen Seaports in their industrial area and about the typical characteristics of their industrial area. Next to the interview a meeting of the project group on datacenters was attended. This project group is responsible for the datacenter in the Eemshaven area and for attracting new datacenters. During this meeting location factors of datacenters were discussed. The project group consists of Ms. L. Pigge, Mr. T. Smit and Mr. R. van Tuinen. To get more information about the role that Groningen Seaports has in the development of their area a survey of Groningen Seaports has been analyzed. This survey is conducted by Groningen Seaports in 2010 to assess the satisfaction of the companies in their industrial areas.

3.1.1.3 Outline of case study report

In Chapter 3.1.2 Groningen Seaports will be described in more detail. The remainder of the case study is based on the sub questions of the main research as presented in Chapter 1.2.4. In Chapter 3.1.3 the industrial areas of Groningen Seaports will be described in more detail, in Chapter 3.1.4 the role of Groningen Seaports in their industrial area is described and Paragraph 3.1.5 will cover information about datacenters.

3.1.1.4 Case study questions

The three main questions in this case study are adopted from the sub questions presented in Chapter 1.2.4, being:

 What are typical characteristics of the industrial areas of Groningen Seaports?

 What is the role of Groningen Seaports in their industrial area?

(25)

3.1.2 Introduction Groningen Seaports

Groningen Seaports is the port authority of two Dutch seaports which are located in Delfzijl and the Eemshaven, both in the province of Groningen as can be seen in Figure 4. The ports are situated right in the middle of the two main ports

Hamburg and Rotterdam and therefore have a favorable position in the Northwestern part of Europe. The port authority is responsible for the port area as well as the industrial sites, and it provides a complete package of services, like for example preparing the right infrastructure for their respective customers. Next to the two seaports, Groningen Seaports is also responsible for two inland ports, Farmsumerhaven and Oosterhornhaven, and a rail terminal in Veendam.

In the remainder of this chapter the main focus will be on the Eemshaven because a datacenter is located there. First the industrial areas of both the Eemshaven and Delfzijl are introduced in more detail after which the role of Groningen Seaports in the development of these areas will be discussed. After that more information on datacenters in the Eemshaven will be provided and the chapter will end with an intermediate conclusion.

3.1.3 Industrial area Groningen Seaports

As already mentioned in the previous section, Groningen Seaports is responsible for two ports with each their own industrial area. The two seaports of Delfzijl and the Eemshaven differ in many ways. The seaport of Delfzijl mainly focuses on chemical and metal industries, with the chemical industry in Delfzijl producing roughly 15% of the total chemical production in the Netherlands. The seaport of the Eemshaven has a different focus, namely energy. The energy park in the Eemshaven, when finished, will produce more than 33% of all the

energy usage of the Netherlands. Next to that, a hydro-energy cable _{Figure 5; Short description of seaports} (www.groningen-seaports.com) Figure 4; Geographical position of the seaports of Groningen Seaports

Delfzijl

Industry focus; Chemicals and metal Transshipment; 4,868,000 tons

Surface area; 1,469 hectares Environmental zoning; category 1-5

Current depth; 9 meters

Eemshaven

Industry focus; Energy supply Transshipment; 2,754,000 tons

Surface area; 1,129 hectares Environmental zoning; category 1-5

(26)

is coming in from Norway.

Delfzijl has the largest surface area and transships almost twice the amount that the Eemshaven transships, namely 4,9 million tons versus 2,8 million tons. The port of the Eemshaven has a depth of 11 meters, and planned to increase it to 14 meters, and the port of Delfzijl has a depth of 9 meters. Both port areas have a high ranking on environmental zones, meaning that both can support all kinds of industries. The Eemshaven for example can even build a nuclear power plant, although there are no plans to do so.

It is clear that both the seaports focus more on industrial clusters than traditional port activities like container shipping. Only 7 percent of the shipped cargo of Groningen Seaports in 2011 were containers and 89 percent of the shipped cargo was bulk which is mainly processed in the Delfzijl port’s chemical park. Groningen Seaports is also not focusing on container shipment because of the lack of a direct rail connection to the rail port in Veendam and the fact that the port of Rotterdam is relatively nearby and companies rather want to ship their containers to Rotterdam.

From the survey conducted by Groningen Seaports, investigating the satisfaction of all the companies in the industrial areas, it can be found that the most attractive aspect of the industrial areas of the Eemshaven and Delfzijl is the geographical location. After that the quality of the infrastructure and the land price were mentioned the most. For this survey conducted in 2010, 131 companies were contacted with a response rate of 88%.

3.1.4 Role of GSP

As described in detail in Chapter 2.4, ports can have three different functions or roles. The classification of these functions depends on the governance structure and the involvement of the port authority in the port area. In short the role of the landlord has the most influence in the area, the regulator has less influence and the operator has the least influence, for a more thorough description of the roles we would like to redirect the reader to Chapter 2.4.

(27)

Groningen Seaports can also help new or other industries to come to their areas by trying to influence governments to change the zoning plan (bestemmingsplan in Dutch), allowing Groningen Seaports to expand their area. Groningen Seaports is actively involved in the acquisition of new companies in industries they would like to attract, like for example datacenters. Next to that they are also looking at the interest of the other companies, they for example try to avoid companies that might do better in another area by advising them to look for another location.

According to the survey mentioned before in Chapter 3.1.3, in 2010 the majority of the companies located in the Eemshaven or the Delfzijl area were satisfied with the role of Groningen Seaports.

3.1.5 Datacenter(s) in the Eemshaven

3.1.5.1 Current situation

At the moment of writing this thesis, July 2012, there is one datacenter located in the Eemshaven and although no more details are available a second datacenters seems to be coming (Zijm, 2012). The datacenter already located in the Eemshaven is a so called single-tenant datacenter facility with Google as a customer (De Boer, 2007). The datacenter started operating in 2008. Groningen Seaports is constantly trying to get more datacenters in the Eemshaven and is actively approaching potential new datacenter customers.

3.1.5.2 Sustainability

One of the trends in the datacenter industry is to create a more sustainable business concept (Duy et al., 2011; Viswanathan et al., 2011). This can be done internal, by designing a more cooling efficient datacenter and using the least power consuming hardware, but it can also be done externally. This more external approach focuses mainly on the way the energy is produced. The more renewable and sustainable the energy mix of the datacenter is, the ‘greener’ it is.

As already mentioned in chapter two, the availability of a reliable power supply is important to datacenters. Therefore the available energy mix in the Eemshaven is also a point of interest for the potential new datacenters. In the Eemshaven a broad energy mix, supplying around 35 percent of the total Dutch energy production, is available with for example energy produced by biomass, coal, gas and wind. From Norway a cable from a hydro powered plant is also connected to the Eemshaven. (Energyvalley.nl, 2012; Groningen-Seaports.nl, 2012a).

(28)

3.1.5.3 Possibilities in the Eemshaven

Another important factor for datacenters is the presence of a connection to an internet exchange and internet cables. Although there is some distance between the internet exchanges of Amsterdam and Hamburg, the Eemshaven is connected to both and also has a data cable coming ashore connecting New York City with the Eemshaven.

Since the datacenter industry is still growing rapidly, datacenters often look for expansion after they have started operating meaning that there should be enough land available to expand. In the Eemshaven they have enough opportunities to expand, or if necessary, build another facility.

An aspect that is taken in consideration by potential new datacenters is the risk factor. These risks can be any type of natural disaster or human errors. Since the Eemshaven is next to the sea, one danger seems to be flooding, however in the Eemshaven this is minimal because the location for datacenters is, contrary to the biggest part of the west and northern part of the Netherlands, above sea-level. Another risk might be that one of the wings of a wind mill will come down and destroy everything underneath. These risks are all investigated before a decision on the location will be made by datacenters.

3.1.6 Conclusions

(29)

Location factors L* C**

Availability of internet cables with sufficient capacity X Availability of land (to build and, perhaps later on, expand) X X

Availability of skilled employees X X

Availability of sustainable energy X

Distance to educational institutions X

Distance to internet exchange X

Favorable zone requirements X

Land price X X

Likelihood of a disaster (either caused by human errors or nature) X

Presence of connection to internet exchange X

Prevailing wage rates X

Redundancy of energy supply X X

Reliability of the energy net X X

Table 3; List of location factors after literature review and case study * Literature review

** Case study

3.2 Expert interview

To acquire information from the datacenter field itself, an expert interview with a manager of a datacenter was conducted. Because of the sensitivity of the information the name of the datacenter and the manager will not be mentioned in this thesis. The expert interview will be used to answer the following sub questions;

 What are specific characteristics of datacenters?

 What are important location factors for datacenters?

3.2.1 The interview

The interview was conducted by e-mail, mainly to reduce the time needed by the interviewee to answer the questions. Because of the fact the interview was taken by e-mail the interview can be regarded as a structured interview. The questions asked were of an open nature and the number of questions was kept low while trying to have the questions asked, cover the most relevant issues. This because a small amount of important questions results in a better quality of the answers than when asking a lot of questions (Myers, 2009).

(30)

3.2.2 Answers

The results of the interview are structured in some categories which are Risks, Energy issues, Sustainability and Employees. These categories are based on the answers of the manager of the datacenter.

3.2.2.1 Risks

One of the most important issues is risk management for datacenters. Potential customers want the datacenters to be as reliable as possible so datacenters should always be online and functioning. The manager made a clear distinction between different types of risks. There are three types to be identified; geographical risks, economical risks and location related risks.

Geographical risks are best described as the risk of a disaster. This can be natural disasters as well as disasters caused by human errors or dangerous industries. Examples are risks of flooding and earthquakes and the presence of dangerous companies like companies active in the chemical sector or fireworks. Dangerous transports by train or airplanes in the neighborhood of the datacenter can be a risk as well. Economic risks are more focusing on the potential customers and the day to day operation of the datacenter. Important aspects are the distance to potential customers and the availability of employees in the region. Location related risks can be seen as the characteristics of the physical location. The reliability of the energy net, the availability of enough energy for an expansion in a later stage, the availability of extra cooling sources like the presence of water, the accessibility (by road but also the distance to an airfield for international customers), and local regulations like permissions and zone requirements are examples of location related risks. The presence of internet cables is still important but is becoming less important because in the Netherlands almost all areas have good access.

3.2.2.2 Energy issues

As already mentioned, energy and energy supply are very important for datacenters. Due to the high power consumption rates of datacenters, the location of a datacenter needs to have enough possibilities to facilitate this power consumption. This energy should be available as quick as possible and the connections need to be reliable. For example, it shouldn’t be the case that some minor construction operation in the vicinity of a datacenter can cause a blackout for the datacenter. 3.2.2.3 Sustainability

(31)

3.2.2.4 Employees

It is very difficult for datacenters to find suitable employees. In general there already is a lack of technical employees in the Netherlands. Next to that, employees of datacenters should have knowledge of more than one different technical aspect and they should typically be all-rounders. For example, they should have knowledge about electro technique, cooling functionalities and fire prevention. The manager compared a datacenter with a large ship or small power plant that needs to be in operation 24 hours a day, and then consider that the datacenter can’t be offline for a day for maintenance.

3.2.3 Conclusion

The information presented in this paragraph is not leading to new factors in the list of location factors but are confirming some of the factors that were already introduced in the previous sections, as can be seen in Table 4.

Although the analysis of financial information of datacenters was also part of the research, it was not included in this chapter because no new insights were found and it also did not support any of the factors already found. For those who are interested, the financial analysis was included in Appendix B.

Location factors L* C** E***

Availability of internet cables with sufficient capacity X X Availability of land (to build and, perhaps later on, expand) X X

Availability of skilled employees X X X

Favorable zone requirements X X

Land price X X

Likelihood of a disaster (either caused by human errors or nature) X X

Presence of connection to internet exchange X

Reliability of the energy net X X X

Table 4; List of location factors after literature review, case study and expert interview * Literature review

(32)

4. Conceptual model

In this chapter the conceptual model, consisting of the thirteen factors that have been introduced in the preceding chapters, is presented. The list of factors is in fact an answer to the second sub question; ‘What are important location factors for datacenters?’. In this chapter the conceptual model will be introduced and the eight most important factors will be determined, these eight factors will be used in the next chapter. In Chapter 4.3 more information is provided about the selection of the eight factors.

4.1 Introduction

The basis for the conceptual model was built throughout Chapters 2 and 3. Starting with the literature review and followed by the case study and the interview with the CTO of a datacenter in the Netherlands. The literature review provided nine of the factors in the model of which around 78% was also supported by the empirical part, the case study and expert interview, of the research. The empirical part provided four additional factors, resulting in a total of thirteen factors.

4.2 The conceptual model

4.2.1 Categorization and importance

(33)

Final list of location factors L* C** E***

Availability of internet cables with sufficient capacity X X Availability of land (to build and, perhaps later on, expand) X X

Availability of skilled employees X X X

Favorable zone requirements X X

Land price X X

Likelihood of a disaster (either caused by human errors or nature) X X Presence of connection to internet exchange X

Reliability of the energy net X X X

Table 5; Final list of location factors * Literature review

** Case study *** Expert interview

To come to a clear conceptual model the factors in Table 5 have been categorized, this can be found in Table 6. The categorization is based on the categories that were used in Chapter 3.2.2 and were expanded by clustering other related factors.

Location factors L C E Category

Availability of internet cables with sufficient capacity X X Connections

Availability of land (to build and, perhaps later on, expand) X X Land issues

Availability of skilled employees X X X Employees

Availability of sustainable energy X Sustainability

Distance to educational institutions X Employees

Distance to internet exchange X Connections

Favorable zone requirements X X Land issues

Land price X X Land issues

Likelihood of a disaster (either caused by human errors or nature)

X X Disasters

Presence of connection to internet exchange X Connections

Prevailing wage rates X Employees

Redundancy of energy supply X X Energy issues

Reliability of the energy net X X X Energy issues

Table 6; Categorization of location factors

(34)

Category Number of factors Number of sources Weights Connections 3 4 1.33 Land issues 3 6 2 Employees 3 4 1.33 Sustainability 1 1 1 Disasters 1 2 2 Energy issues 2 5 2.5

Table 7; Calculation of importance of categories in conceptual model.

4.2.2 The conceptual model

Now the relative importance of the factors is determined, the conceptual model can be drawn as in Figure 6. The categories can be found around the central concept, which is the location of a datacenter. All the categories influence the location of a datacenter. The larger the size of the circles in which the names of the categories are written, the more factors this category represents. The thickness of the arrows in the conceptual model are determined by the weights introduced in Table 7.

Employees

Land issues

Energy issues Disasters

Location of

datacenters

Connections

Sustain ability

Figure 6; Conceptual model

4.3 Most important factors

(35)

method, which is introduced in more detail in Chapter 5. With this method weights are calculated through pair-wise comparisons of the factors. By using more than eight factors, the AHP method becomes less effective.

4.3.1 The survey

The eight most important factors have been determined by using a survey. In this survey managers from datacenters were asked to rank the thirteen factors that are presented in Table 5. The eight most important factors will be used in the next chapter to develop the final model which will also add weights to the factors.

The survey was designed to have a simple drag and drop list of the thirteen factors. Because all the managers that participated were Dutch, the questions and factors have also been translated to Dutch. The translation can be found in appendix D.

The factors were initially presented in alphabetical order to reduce the chances of any kind of bias.

4.3.2 Responses

4.3.2.1 Participants

For the survey 53 datacenters were contacted to fill in the survey mentioned before. These datacenters were located scattered around the Netherlands and were of small as well as larger size to have a representative sample. The total population of datacenters in the Netherlands is around 75 (www.datacentrumgids.nl). There are more datacenter facilities but around 75 unique datacenter companies that are focusing on storing data.

4.3.2.2 Respondents

To be able to get as much responses as possible, three conditions should be met in the survey. These conditions are: the participant must believe filling in the survey is pleasant and satisfying, the survey is important and a worthwhile use of his or her time and the participant must dismiss any mental reservations that he or she might have about participation (Cooper & Schindler, 2008). In order to comply with these conditions as much as possible, the relevance of the survey for the research and the goal of the research itself were mentioned to the potential participants. Next to that it was mentioned that the survey would not take more than five minutes, the answers would be kept anonymous and the name of the companies would not be mentioned in the thesis.

(36)

the 53 potential participants, 10 could not be contacted, mainly due to the fact that the responsible person was out of office, this group is called ‘Not available’ in Table 8.

To calculate the response rate, the number of companies that were not available is not included because the request to fill in the survey did not reach the responsible person. There is not one generally accepted minimum response rate, some researchers set 20% as the minimum and some use 50% (Karlsson, 2009). The response rate of the survey therefore was 21% as visualized in Figure 7.

Response characteristics Number

Responses 9 Non-responses 34 - No response (27) - Refused (7) Not available 10 Total 53

Table 8; Response characteristics

4.3.3 Results

The results of the surveys are presented in Table 9. The nine responses, the average position and the final positions are presented. The results seem to be quite consistent which is also indicated by the fact that of the eight factors that had the highest average position, seven were also mentioned the most in the top eight of the individual respondents.

(37)

Factor 1 2 3 4 5 6 7 8 9 Avg. Position 1 Availability of internet cables with sufficient

capacity

4 7 5 1 2 1 1 2 1 2.67 1

2 Reliability of the energy net 1 1 3 9 4 5 4 3 4 3.78 2

3 Presence of connection to internet exchange 3 8 6 4 3 2 8 5 3 4.67 3

4 Distance to internet exchange 2 9 7 5 6 4 11 4 2 5.56 4

5 Land price 7 2 2 7 12 6 2 7 10 6.11 5

6 Availability of land 10 11 1 6 11 7 3 1 12 6.89 6

7 Redundancy of energy supply 5 12 12 2 5 3 10 9 6 7.11 7

8 Availability of sustainable energy 6 4 4 8 7 13 5 8 11 7.33 8

9 Likelihood of a disaster 8 3 13 3 1 12 12 6 9 7.44 9

10 Availability of skilled employees 11 5 11 11 10 8 7 11 5 8.78 10 11 Favorable zone requirements 9 10 8 10 8 11 6 10 8 8.89 11 12 Distance to educational institutions 13 6 9 13 9 10 9 13 13 10.56 12 13 Prevailing wage rates 12 13 10 12 13 9 13 12 7 11.22 13

Table 9; Survey results

The eight most important factors that will be used in the remainder of the research are determined by the highest average positions. In Table 10 these eight factors are presented. The factors that were ranked from six to nine have almost similar scores (6.89 to 7.44). Therefore the ninth factor might be of importance as well, in the discussion this will be discussed in more detail.

Position Factors

1 Availability of internet cables with sufficient capacity 2 Reliability of the energy net

3 Presence of connection to internet exchange 4 Distance to internet exchange

5 Land price

6 Availability of land

(38)

5. Model assessing the attractiveness of areas to datacenters

5.1 Introduction

In this chapter the final model of this research will be presented. This model will show the most important location factors for datacenters and can therefore be used to assess the attractiveness of an area to datacenters. This is done by adding weights to the eight factors presented in the previous chapter. To come to these weights the AHP method is used. More information about this method is presented in the next paragraph.

5.2 The AHP method

5.2.1 Introduction

The Analytic Hierarchy Process (AHP) is designed to calculate weights to elements. The method uses a number of inputs in which experts in a field do a pairwise comparison of the different factors. Using some mathematical formulas the weights can then be calculated. Because weights can be calculated relatively easy, this method is better than standard ranking methods. With the weights it is possible to express to what extent one factor is more important than another.

By using only the eight most important factors in this part of the research, the least relevant factors are filtered out. Besides, by adding more factors the number of comparisons would become very high. Including a ninth factor would increase the number of comparisons needed to complete the method with 32% and a tenth factor with 67%. Next to that, the higher the number of factors the more likely the results will be inconsistent and therefore less valuable, according to Dr. Klaus D. Goepel from http://bpmsg.com/, an expert in using the AHP method.

By using the AHP method, weights can be added to the different location factors and with that information, industrial areas that would like to attract datacenters can assess their area on the attractiveness to datacenters.

(39)

5.2.2 Results

In Table 11 the results of the calculations are presented. The calculations are done by an excel spreadsheet developed by Dr. Klaus D. Goepel (BPMSG, 2012). With a consistency ratio of 2.9% the results can be regarded as consistent. A consistency ratio of 0% would mean perfect consistency and a ratio of more than 10% would indicate inconsistent results.

Factor

1 Availability of internet cables with sufficient capacity 22% 2 Presence of connection to internet exchange 18%

3 Reliability of the energy net 14%

4 Distance to internet exchange 12%

5 Availability of land 10%

6 Land price 9%

7 Redundancy of energy supply 8%

8 Availability of sustainable energy 7%

Table 11; Results of AHP method

5.3 The model

In Figure 8 the final model is presented, with the location of a datacenter as the central concept. The same categories have been used as in the conceptual model. The categories are introduced in Chapter 4.2.1. In the final model however, some categories represent a smaller amount of factors and the ‘disaster’ and ‘employees’ categories are not represented in the model anymore because the factors from this category were not amongst the eight most important factors. The thicker the line from the category to the central concept is, the more important that category is. The lines are drawn in millimeters and represent their weights, so for the category connections the line is 5,2mm. The basis for the model can be found in Table 12. From the model it is clear that the ‘connections’ category is by far the most important, while sustainability is the least important.

Category Number of factors Total weight

Connections 3 52%

Energy issues 2 22%

Land issues 2 19%

Sustainability 1 7%

(40)

Energy

issues

Location of

datacenters

Connections

Sustain

ability

Land issues

52% 22% 7% 19%

‘Assessing the attractiveness of port areas to datacenters’