Вы находитесь на странице: 1из 103

SHARED UMTS NETWORKS

- A Techno-Economic Study on Mobile Network Operators Possibilities to Deliver Unique Products and Services -

MASTERS THESIS
Stockholm, March 2005

Niklas Agevik

Advisors: Klas Johansson (KTH), Robert Tjernstrm (TeliaSonera) Examiner: Jens Zander (KTH)

Introduction

ABSTRACT
Historically, there has been a one-to-one mapping between a mobile network and the reseller of the networks services. Recent developments challenge this model. Virtual operators have emerged in GSM networks and during the rollout of UMTS networks, many operators built shared UMTS networks. This study analyses the impact of shared networks on the mobile network operators abilities to deliver unique products and services to their customers. Four sharing models that are in use today are analysed and presented in detail. These are site sharing, geographical split networks, shared RAN and spectrum sharing. The research in this study has been done in two steps. In the first step, possible differentiation opportunities are identified through literature studies and interviews with key persons at network operators, equipment manufacturers, business analysts and the Swedish national post and telecom agency. In the second step, this study compares the implementation of these opportunities in the four forms of shared networks with their implementations in a wholly owned network. The opportunities are grouped in three categories: products, quality of service differentiation and multiaccess networks. By setting quality of service attributes on bearers, such as latency and the attention/retention policy, quality of service differentiation is a possibility in common shared networks and geographical split networks. Although unlicensed access technologies, such as WLAN standards, can provide coverage differentiation in one sense, this study shows that this is not economical compared to UMTS pico-cells. The study finds that implementing features in the underlying network requires extensive cooperation between operators in all sharing models except for site sharing. Certain services require quality of service parameters that are mutually exclusive without over-provisioning in terms of network resources. Such parameters include for example latency and capacity. Even though the 3GPP UMTS standard allow for a separation of the service creation layer and the underlying network, this study argues that many services will still be dependent upon network features, thus mitigating the advantage of developing services in the unshared domain.

Introduction

TABLE OF CONTENTS
ABSTRACT....................................................................................................................... I TABLE OF CONTENTS .............................................................................................III ACKNOWLEDGEMENTS.............................................................................................V INTENDED AUDIENCE ..........................................................................................VII PREVIOUS WORK ....................................................................................................... IX 1 2 3 3.1 3.2 3.3 4 4.1 4.2 4.3 4.4 5 5.1 5.2 5.3 5.4 5.5 5.6 5.7 6 6.1 6.2 6.3 6.4 7 7.1 7.2 7.3 7.4 INTRODUCTION .........................................................................................1 BACKGROUND TO NETWORK SHARING IN SWEDEN ......................3 GOAL OF THIS STUDY................................................................................7 PURPOSE ............................................................................................................................. 7 PROBLEM STATEMENT ...................................................................................................... 7 LIMITATIONS TO THIS STUDY........................................................................................... 8 RESEARCH METHODOLOGY ...................................................................9 QUALITATIVE AND QUANTITATIVE RESEARCH ........................................................... 9 DATA ACQUISITION......................................................................................................... 11 QUALITY OF RESEARCH DESIGN ................................................................................... 13 RESEARCH METHOD ........................................................................................................ 15 BUILDING SHARED NETWORKS........................................................... 17 DEFINITION OF SHARED NETWORKS ........................................................................... 17 LEVELS OF NETWORK INFRASTRUCTURE SHARING .................................................... 18 FREQUENT SHARING CONFIGURATIONS ...................................................................... 20 MAPPING OF NEEDS TO SHARING CONFIGURATIONS .............................................. 24 MIGRATION TO INDIVIDUAL NETWORKS.................................................................... 25 COMPARISON BETWEEN SHARED NETWORKS AND VIRTUAL OPERATORS .............. 26 STATUS AND FUTURE FOR ROLL-OUT OF SHARED NETWORKS IN SWEDEN .......... 26 DIFFERENTIATION IN SHARED NETWORKS ...................................29 DIFFERENTIATION OPPORTUNITIES ............................................................................ 29 ANALYSIS OF NETWORK DIFFERENTIATION OPPORTUNITIES ................................ 31 IN-BOUND ROAMING IN SHARED NETWORKS............................................................ 42 SUMMARY OF NETWORK AND ROAMING RELATED OPPORTUNITIES..................... 44 QUALITY OF SERVICE DIFFERENTIATION IN SHARED NETWORKS .................................................................................................47 INTRODUCTION TO QUALITY OF SERVICE DIFFERENTIATION ................................ 47 AGGREGATED AND END-TO-END SERVICE LEVELS ................................................... 49 QUALITY OF SERVICE ARCHITECTURES ........................................................................ 51 QUALITY OF SERVICE IN A SHARED NETWORKS ........................................................ 54

iii

Introduction

8 8.1 8.2 8.3 8.4 9 9.1 9.2 9.3 9.4 10 10.1 10.2 10.3 10.4 10.5

MULTI-ACCESS NETWORKS ................................................................... 57 WLAN AND EDGE INTERACTION WITH SHARED NETWORKS .............................. 57 DIFFERENTIATION BY CAPACITY AND COVERAGE ................................................... 58 NATIONAL AND MULTI-ACCESS ROAMING................................................................... 66 FACILITATING SERVICE ROAMING WITH THE VIRTUAL HOME ENVIRONMENT (VHE)............................................................................................................................... 67 CONCLUSIONS........................................................................................... 71 HOW DO SHARED NETWORKS AFFECT THE DIFFERENTIATION OPPORTUNITIES? 71 IMPLICATIONS OF SHARED NETWORKS ON OPERATORS BUSINESS ....................... 72 SUMMARY OF KEY FINDINGS ........................................................................................ 73 FUTURE WORK ................................................................................................................ 73 BIBLIOGRAPHY ......................................................................................... 75 BOOKS AND REPORTS ..................................................................................................... 75 ARTICLES AND WHITEPAPERS ........................................................................................ 76 TECHNICAL SPECIFICATIONS ......................................................................................... 78 INTERVIEWS ..................................................................................................................... 79 INTERNET RESOURCES ................................................................................................... 79 ACRONYMS..................................................................................... 81 OVERVIEW OF ACCESS TECHNOLOGIES .............................. 85 NODES AND INTERFACES IN UMTS AND GSM .................... 87 INTERVIEW QUESTIONNARE .................................................. 91

APPENDIX A APPENDIX B APPENDIX C APPENDIX D

iv

ACKNOWLEDGEMENTS
This master thesis represents the final part of my education at the Royal Institute of Technology in Stockholm, Sweden. I have studied Electrical Engineering with a major in wireless systems. The thesis work has been carried out at a Swedish network operator, TeliaSonera, in Sundsvall, Sweden from September 2004 until February 2005. I would like to thank my supervisors at TeliaSonera, Lars Liljestam and Robert Tjernstrm, for their valuable insight and guidance. Furthermore, I would like to thank my supervisor at the Royal Institute of Technology, Klas Johansson who has provided valuable feedback during the thesis work. In addition, I would like to thank to my associate, Navid Rostam, who has been a sounding board during the study. Niklas Agevik, Stockholm February 2005

INTENDED AUDIENCE
The intended audience for this report is as follows. My supervisor Klas Johansson and my examiner Jens Zander at the Royal Institute of Technology. My supervisors at TeliaSonera Robert Tjernstrm and Lars Liljestam Other employees at TeliaSonera Employees at other operators that are currently in a UMTS shared network agreement or are considering entering such an agreement.

Reader of this report will learn


The status of shared networks in Sweden How different forms of shared networks are constructed What differentiation possibilities exist for mobile network operators today How these possibilities are affected by a shared UMTS network How shared networks affect operators abilities to deliver unique products and services

TeliaSonera will gain


TeliaSonera will have all knowledge currently available internally in the company about how their shared network affects their businesses gathered in one place. TeliaSonera will also gain an outside view of their differentiation possibilities. They can use this information to focus their resources on the differentiation possibilities that they consider matters the most.

vii

PREVIOUS WORK
When UMTS licenses were granted around the year 2000 in Europe, almost all operators presented whitepapers or similar documents that described how their equipment could be used for network sharing. One such whitepaper is Ericssons Shared networks: An operator alternative to reduce initial investments, increase coverage and reduce Time To Market for WCDMA by sharing infrastructure [24]. Furthermore, 3GPP has written a technical report on the different sharing models used. This technical report is similar to the whitepapers produced by operators and is called Service aspects and requirements for network sharing (TR 22.951) [56]. Both the Ericsson whitepaper and the 3GPP technical report aim to describe how shared networks can be built. In 2001, TIA Europe produced a report on the status of 3G network infrastructure sharing in Europe [34]. It is a good introduction to where infrastructure sharing takes place in Europe, although, a bit out-dated. In 2001, Northstream, a consultancy, produced a report for the Swedish National Post and Telecom Agency entitled Network Sharing: Savings and Competitive Effects [17]. This work analyses the competitive effects of network sharing and is closely related to this study. 3GPP also works on facilitating future shared networks. This work is headed by Ulf Nilsson (who is interviewed in this study) and can be found in technical report 23.851 and technical report 23.251 [65], [66]. The TONIC project (Techno-economics of IP optimised networks and services) is a project within the IST programme which concentrates on techno-economic evaluation of new communication networks and services, in order to identify the economically viable solutions that can make the Information Society [sic] to really take place.. Within their modelling, they have included modelling of shared networks. Their report Deliverable 14: Recommendations and guidelines shows their approach to modelling the economics of wholly owned and shared UMTS networks [14]. All TONIC reports can be found at [82]. The multi-access chapter in this study is based on assumptions from the TONIC project and work form Wireless@KTH. A good introduction to network modelling can be found in the paper Relation Between Base Station Characteristics and Cost Structure in Cellular Systems [30]. The TONIC project also contains economic modelling of operator owned WLAN networks. This study deals with differentiation strategies. A good reference source in this area is Competitive Advantage: Creating and Sustaining Superior Performance (1998) by Michael E. Porter.

ix

Introduction

INTRODUCTION

The way to build mobile networks is changing. In GSM networks, there has been a one-to-one mapping between mobile operators and the access networks that they own and run. During the rollout of UMTS in Europe, it became clear that this is no longer the case. When licenses for 3G networks were sold or handed out in so called beauty contests there were high expectations on the traffic in future UMTS networks. Operators paid high license fees and promised fast rollout combined with good coverage. The total amount of money spent on license fees in Europe is estimated to be at 133 billion [39]. To reduce the amount spent on network rollout, some operators saw a shared UMTS network with other operators as one solution. Since frequency spectrum is a limited resource, only a few licenses are available in each country. Some companies that participated in the auctioning and beauty contests were therefore left empty handed. Their only means of gaining access to the UMTS market was to find a partner willing to share their license with them. In the GSM world a new set of operators has emerged, service providers, enhanced service providers and mobile virtual network operators (MVNO) or virtual operators for short. Common to all of these operators is that they lease usage time in other operators networks and resell it under their own brand to customers. Virtual GSM operators are now an integral part of the telecom industry. Currently, the virtual operators do not have access to the same features as the network operators but it is widely believed that network operators will be forced to grant MVNOs the same opportunities as themselves. As MVNOs get more opportunities to provide features from GSM networks, such as GPRS, they will require the same level of access from WCDMA operators. One way for virtual operators to achieve this is by providing their own core network nodes. If virtual operators own more of the core network nodes instead of leasing them and gain access at a lower level to the network operators nodes there is a clear

Introduction

resemblance between virtual operators and sharing agreements between network operators. The three aforementioned cases (network sharing to reduce costs, sharing of a single WCDMA license and a virtual operator requiring access to the underlying network) require cooperation and infrastructure sharing between rival operators. Furthermore, it challenges the one-to-one mapping between operator-network and creates new problems on how to build and maintain an access and core network. Current equipment and standards available today is not designed for network sharing although some companies are providing proprietary features. For network sharing to run smoothly new features are required. During the rollout of GSM networks, the networks from different operators varied widely in quality. Differences in coverage were evident to the customers and a key selling point was that of network quality. Over time, GSM networks have matured to a point were they for the most part have equal quality. Today, the focus of the competition between operators is on the service level and features available in the mobile network. One of the key differentiation possibilities for operators that completely own their network is that they can implement features and services not available in competing networks or to virtual operators, even those operating in their own network. UMTS licenses in most countries regulate the technical solutions operators can use for network sharing. Apart from this, current technical solutions only have basic support for sharing the equipment. There are different levels of cooperation required between the operators depending on how the agreement between them is arranged. If the operators are equals in the shared network, they are not only fighting amongst themselves for more customers, but also together against other operators. Disagreements between participating operators may result in a loss of total market share for their joint venture. Apart from finding ways to differentiate themselves against each other they must also work together so that their joint venture is differentiated against rival joint ventures and other operators by example being the preferred choice for new virtual operators. Many network operators building and running UMTS networks are incumbent GSM network operators. Some run, or plan to invest in, a WLAN hotspot network. For operators not running a WLAN hotspot network there is the question of if it is a necessary complement to UMTS. Incumbent GSM operators are also in face of upgrading to EDGE. Differentiation is therefore possible by offering products that include joint offerings including GSM, EDGE and WLAN. All these factors put together makes it difficult for a network operator to provide a feature that the other participating operator or virtual operator does not. Operators that for one reason or another find themselves in the position of a shared network will therefore have their possibilities of feature and coverage differentiation weakened. Network operators and virtual operators need to find new ways of differentiating themselves to customers.

Background to Network Sharing in Sweden

BACKGROUND TO NETWORK SHARING IN SWEDEN

The mobile communications industry relies on wireless communications that requires spectrum bandwidth. Available technology limits the supply of spectrum bandwidth available, which has forced governments in all countries to impose regulation of how to divide this spectrum among mobile network operators. By awarding licenses that allows a company to use a given spectrum for a set number of years (15 years in Swedens case) available bandwidth is distributed. This implies that only a few companies can compete for customers in each country. An important role of the regulation agency is therefore to make sure that the companies running mobile networks are indeed competing. During the granting of UMTS licenses around the world, regulation agencies used auctions or beauty contests to distribute the licenses. The Swedish regulation agency, The National Post and Telecom Agency (PTS), awarded licenses according to a beauty contest model. During 2000, they committed to granting four licenses for WCDMA. The Swedish model for awarding licenses required all operators wishing to build a WCDMA network to produce a statement of how they would build their network given that they acquired a license. Operators that to the highest extent fulfilled the terms set by the PTS received a license. The agency set forth terms that required a high coverage rate of the population combined with a fast rollout. They foresaw that these terms would require network sharing for it to be economically viable to build several WCDMA networks in Sweden. Magnus Axelsson, part of the board of directors for the National Post and Telecom agency, said in an interview, We were probably the agency with the most foresight in Europe when it comes to network sharing. If you consider the dispersedly located population and the large geographical area it would not have been sound for the national economy if we forced the rollout of four separate WCDMA 3

Background to Network Sharing in Sweden

networks [71]. To lower the barriers for greenfield operators to entering the UMTS market, the terms also stated that incumbent GSM operators must provide national roaming agreements to greenfield operators in their GSM networks during seven years1. More than four companies applied for the licenses and after consideration by the PTS, they granted all of the available licenses. Two of these were awarded to greenfield operators while two was awarded to incumbent Swedish GSM operators. One of the greenfield operators eventually withdrew and returned their license to the PTS. This license is as of now still unclaimed (February 2005). The remaining greenfield operator entered a sharing and national GSM roaming agreement with one of the incumbents. TeliaSonera (the sponsor of this thesis) is an incumbent Swedish GSM operator that did not receive an UMTS license in the Swedish beauty contest. They entered a sharing agreement with one of the incumbent GSM operators who did acquire a license. This means that all Swedish operators are in a sharing agreement for their UMTS network although the specifics of these agreements differ. Two joint ventures, Swedish UMTS Networks AB (SUNAB) and 3G Infrastructure Services (3GIS) have been formed to administrate the rollout of the two shared networks. As SUNAB consists of TeliaSonera and another incumbent GSM operator, they both have individual full-scale operating GSM networks. Since TeliaSonera did not receive a WCDMA license, they are sharing a single license for their UMTS network. 3GIS on the other hand consists of a greenfield operator together with an incumbent GSM operator, both with individual WCDMA licenses. All Swedish GSM operators have the option of upgrading their GSM network with EDGE, giving nearly the same data speeds as UMTS (See Appendix B). TeliaSonera has said that they will be upgrading their Swedish GSM network to EDGE but as of January 2005, they have not set an exact timeline for the upgrade. They are also the only operator who has a WLAN hotspot network. Since WLAN technologies operate in the unlicensed Industrial, Scientific and Medical spectrum (ISM) all operators have the possibility of building a separate WLAN network. Table 1 lists a summary of the state of the wireless networks run by operators in Sweden as of January 2005. The technical solutions used by virtual operators are similar or in some cases identical to those used in the Swedish shared networks. The Swedish virtual operators have faced the same problems regarding differentiation as the situation for the operators sharing networks. It is therefore interesting to study the Swedish market for virtual operators. All of the virtual operators are as of now focusing on being low-cost providers.

In Swedish this is referred to as Skymningsparagrafen

Background to Network Sharing in Sweden

All incumbent GSM operators have shown unwillingness towards virtual operators that do not clearly target a different segment than themselves. PTS has tried to resolve this issue by introducing new legislation, but the new regulation available has done little or nothing to force the incumbents into accepting virtual operators [18]. This has resulted in that the incumbents own almost of the virtual operators. Djuice is a virtual operator that is a notable exception. It has been accepted by one of the incumbents in a deal that included the Swedish incumbent operator gaining access to Djuice parent companys GSM network in Norway as a virtual operator. Table 2 shows a summary of the status for virtual GSM operators in Sweden. Currently, there are no UMTS virtual operators.

Operator

3 (Tre) Vodafone Sweden Tele2 TeliaSonera Orange

Jointventure name 3GIS 3GIS

Incumbent GSM operator? No Yes

Shares WCDMA license? No No

Committed to EDGE upgrade? N/A No

WLAN operator? No No

SUNAB Yes Yes No No SUNAB Yes Yes Yes Yes Returned license to the National Post and Telecom Agency in November 2004

Table 1: Summary of operators awarded WCDMA licenses in Sweden. Source: [18]

Operator Tango Halebop Campuz Mobile Djuice

Host Network Tele2 Telia Vodafone Sweden Tele2

Owner Tele2 TeliaSonera Vodafone Sweden Telenor (Norwegian GSM and UMTS operator)

Total Market share <1% 2% < 1% 1%

Table 2: Summary of major virtual operators in Sweden (GSM). Source: [18]

goal of this study

3
3.1

GOAL OF THIS STUDY


Purpose

This purpose of this thesis is to analyze the impact of all types of shared UMTS networks on mobile network operators possibilities for differentiation by providing unique products and services. The mobile network operators under study are incumbent GSM operators and greenfield UMTS operators. This thesis will be a mapping of current information on UMTS, GSM, EDGE and WLAN networks to see how these technologies affect the differentiation opportunities in a shared UMTS network. The sponsor of this thesis is TeliaSonera, an incumbent Swedish GSM operator who is in a sharing agreement for their UMTS network. This study will include their sharing agreement in the analysis but not give any specific focus to that specific sharing agreement.

3.2

Problem statement

The basic question asked in this thesis is: How do shared networks affect operators abilities to differentiate themselves by delivering unique products and services? This question can be broken down into smaller parts. This study will analyse the following areas to answer this question.

goal of this study

What differentiation opportunities exist in UMTS and GSM? Differences between shared and wholly owned networks o What technical solutions exist for sharing core and access networks? o How do the different forms of shared networks affect the differentiation opportunities? How are these opportunities affected by network sharing? o Which opportunities are most affected? o Are any opportunities unaffected by network sharing? o Which matters the most? How do shared networks affect operators possibilities for competition?

3.3

Limitations to this study

This study aims to expose and show the technical problems and opportunities affecting operators participating in sharing agreements. Apart from this, there are other difficulties for companies working in joint ventures related to writing agreements and making other business decisions. Furthermore, companies must make decisions on how to market offerings to their customers. These considerations are outside the scope of this thesis. In both GSM and WCDMA, operators are trying to differentiate themselves by providing new and innovative services. This thesis will not try to propose any new, pure service offerings. Instead, this study will expose the possibilities and restrictions for creating services within a shared network. The access technologies for building multi-access networks investigated in this report are GSM, GSM with EDGE, WCDMA, WCDMA with HSDPA and 802.11b/802.11g (WLAN). There are other interesting upcoming access technologies such as standards from the WiMAX, HIPERLAN /1 (IEEE 802.11), HIPERLAN/2 and new improved WLAN standards such as 802.11n. In the global market there are competing standards to GSM and WCDMA such as CDMA, CDMA2000 and CDMA2000 1x (EV-DO). To limit the scope of this thesis the geographical focus is on Europe and on access technologies commercially available today (February 2005). Appendix B gives a broad overview of current access technologies.

Research Methodology

4
4.1

RESEARCH METHODOLOGY
Qualitative and Quantitative Research

There are two main methods for conducting research, qualitative methods and quantitative methods. They are not mutually exclusive but they are suited for attacking different kinds of problems. According to [1] there are two main differences. Quantitative studies convert information into numbers and quantities. The researcher can then perform statistic analyses on these results and form an opinion. In qualitative studies, the focus is on the researchs interpretation and understanding of the information at hand. Quantitative methods are based one the fact that it is possible to measure relevant facts from the object under study. These findings should then be possible to present numerically. The researcher regards his field of research as an object that the researcher measures according to one or more variables. Quantitative methods define before the start of the study what variable to measure. The method does not allow the researcher to be flexible and change the structure of the study during the duration of the investigation phase. Qualitative methods do not assume that it is possible to measure everything. It is hard to define exactly what draws the line between qualitative and quantitative methods but the starting points for all qualitative methods is that every phenomenon under study has unique combinations of qualities and properties. They allow for a greater degree of flexibility during the duration of the study. As the study progresses, the researcher may find new areas of interest and guide the study towards them. The strength of qualitative data is that it gives a comprehensive overview of the problem under study. Therefore, to define which area to research, a qualitative study can precede a quantitative study. 9

Research Methodology

A problem with all research, and especially qualitative research, is that the researchers bias affects the study [3]. This is especially a problem in this study since the company sponsoring the study has an interest in the outcome of the study. I have therefore been careful to avoid TeliaSoneras views to influence the method selection. For this study, it is appropriate to mix qualitative and quantitative methods. Chapter 6 lists the differentiation opportunities available to mobile network operators. The work to find these opportunities is exclusively qualitative. In the analysis section, some differentiation opportunities have been analysed qualitatively while others have been analysed quantitatively.

4.1.1

Analytical Induction

Analytical induction is a qualitative method and is the method used for defining the scope of this study. 1. Formulate and delimit the problem. The researcher must define what aspect of the problem to study. 2. Let the study be guided by the problem formulation but avoid forming a hypothesis. The researcher will have some understanding of the problems character and background but must take care not to let this influence the data acquisition phase. 3. Identify relevant and credible sources. 4. Choose appropriate cases, events and technologies to include in the study. The researcher must make sure that the events and cases are relevant to define the problem in question. 5. The researcher then collects data from the sources. This data should further guide the problem formulation and help the researcher reach a conclusion. 6. Analysis of data is done in two steps: decoding and interpretation. In the decoding step, the researcher must find the relevant parts of the data. In the interpretation step the data is then categorized [3]. I have delimited the problem as according to the purpose section. The original problem has been divided into two separate theses. This thesis contains an analysis of the technical differentiation opportunities while a separate thesis is written on the marketing related opportunities.

10

Research Methodology

Problem definition

Niklas Agevik and Navid Rostam

Find differentiation opportunities

Niklas Agevik and Navid Rostam

Navid Rostam

Market analysis

Technical analysis

Niklas Agevik

Navid Rostam

Conclusions

Conclusions

Niklas Agevik

Recommendations

Niklas Agevik and Navid Rostam

Figure 1: Breakdown of the work done in this study on technical opportunities and in the parallel study on marketing opportunities

4.2 Data acquisition


In all investigations, the means of data acquisition is critical to be able to reach a tenable conclusion. This study uses interviews as primary data and secondary data from existing material such as books and articles [1].

4.2.1

Interviews

In interviews, the interviewer gathers data by asking questions or conducting a dialog with the interviewee. The answers are then gathered and they form the raw data of the study. Interviews can be structured or free. In structured interviews, the interviewer focuses on finding a specific piece of information and has in advance decided what the goal of the interview is. The interviewer must formulate the questions and follow-up questions as to allow for a systematic analysis of the areas the interviewer is interested in. Free interviews are used to find the values and opinions of the respondent as opposed to finding hard facts. The interviewer asks questions that stimulate the respondent to elaborate and formulate his/her opinions. Furthermore, interviews can be standardized, non-standardized or semi-standardized. In a standardized interview, all respondents face the same set of questions in the same order. This is useful when conducting quantitative research and the data collected is hard data such as volume, speeds et cetera. A standardized interview requires planning on part of the interviewer beforehand but the researcher can then insert the data gathered in a model. Non-standardized interviews do not conform to a fixed set of questions. Instead, the interview is complete when the interviewer has found answers to fill his need of information. The interview can be flexible and adapt to the interviewee. Semi-standardized interviews are a mix of standardized and nonstandardized interviews. The interviewer has a fixed set of questions but can expand 11

Research Methodology

on interesting topics and add follow-up questions as necessary to reach the information he/she is interested in [3]. In this study, structured semi-standardized interviews are conducted. Clear goals are set as to what information is requested from each respondent. In some interviews, the interviewee has brought up areas of interest and they have been asked to elaborate and posed follow-up questions on those areas. During the progress of the study, respondents have sometimes been asked follow-up questions later. A typical questionnaire is included in Appendix D.

4.2.1.1 Selection of interview respondents As TeliaSonera sponsors this thesis there is a risk that this will bias the selection of interview respondents. Therefore as many interviews as possible have been conducted with people outside of TeliaSonera. Since many employees inside TeliaSoneras organization has experience of some of the subjects that this study deals with they are also interviewed. Putting together their combined views have been valuable during the hypothesis formulation and data acquisition phases. When their views have been included, extra care has been taken to not let TeliaSoneras views influence this study. During the study, outsiders are interviewed as necessary.

4.2.2

Secondary data

Secondary data is material that already exists and is relevant to the study. This material is usually diverse and not combined in a way that is relevant for the problem under study. The data can be in the form of books, research papers and newspaper articles. It can also come from knowledge found in meeting protocols or discussions on the Internet. Assessing if and where the secondary data can be found is important to avoid conducting research that has already been performed [3]. A number of resources are used in this study for collecting secondary data. The main source has been archives of all mayor professional, technical and scientific journals that relate to the area of interest. Also conference proceedings, mainly from IEEE have been studied. For finding background information, standard works in the literature and training material from Ericsson have been used. The internet has been used to find reports, whitepapers, press releases, 3GPP standards and to keep up to date with the current developments in the 3GPP forum. Also, internal reports from TeliaSonera have been used. In the cases where this material contained information classified as internal, it has been withdrawn it from the public report.

12

Research Methodology

4.3

Quality of research design

In [4] it is stated, as a research design is supposed to represent a logical set of statements, you can also judge the quality of any given design according to certain logical tests. Researchers commonly use four tests to establish the quality of an empirical social research (qualitative research). These tests are construct validity, internal validity, external validity, and reliability. The four methods are however common to all social science methods. Therefore, these tests are used throughout the whole project.

4.3.1

Construct validity

Construct validity involves establishing correct operational measures for the concepts being studied. There are three tactics available to increase construct validity. The first is to use multiple sources, in a manner encouraging lines of inquiry. This tactic is relevant during data collection. The second tactic is to establish a chain of evidence, also relevant during data collection. The researcher can establish a chain of evidence by allowing an external observer the reader of the case study to follow the derivation of any evidence from initial research questions to ultimate case study conclusions. The third tactic is to have the draft case study report reviewed by key informants. All three tactics are used in this project. Furthermore, this study uses several sources of information and several people are interviewed to confirm the findings. Especially the later interviews have focused on confirming or invalidating earlier assumptions and conclusions. A chain of evidence is established by allowing the supervisors at TeliaSonera and Royal Institute of Technology to follow the results constantly. Finally, drafts of the report have been handed in constantly for peer review.

4.3.2

Internal validity

Internal validity is a concern only for causal or explanatory case studies, in which a researcher is trying to determine whether a certain event x led to another event y. If the researcher incorrectly concludes that there is a causal relationship between x and y without the knowledge about some third factor, z, that in fact caused y, the research design has failed to deal with some threat to internal validity. This logic is however inapplicable to descriptive or exploratory studies, no matter if the study is a case study, survey, or experiment. These studies are not concerned with making causal statements, which makes internal validity irrelevant to them. Therefore, internal validity have not been dealt with in this project.

4.3.3

External validity

The third test deals with the problem of knowing whether the findings of a study are possible to generalize beyond the immediate case study. There are different kinds of generalisation. Survey research relies on statistical generalisation, whereas case studies 13

Research Methodology

and experiments rely on analytical generalisation. In analytical generalisation, the researcher strives to generalise a particular set of results to some broader theory. The generalisation is not automatic, though. The researcher must test the theory through replications of the findings in a second or even a third case. The results are far more generalised if the researcher makes such a replication. In my research of different methods for network sharing, I have chosen four sharing cases that are in use among mobile network operators. I have not dealt with external validity explicitly, but these methods of network sharing are to the best of my knowledge the ones possible and the ones in use by operators today.

4.3.4

Reliability

The objective with this final test is to be sure that another researcher will acquire the same results if he would repeat the study. However, the condition is that the other researcher follows the exact same procedures as the earlier investigator, and studies the same case(s), not a similar one by replicating the results. The goal of reliability is to minimize the errors and biases in a study. One prerequisite for allowing other investigators to repeat an earlier study is the need to document the procedures followed in the earlier case. There are two recommended tactics to deal with this issue: 1) using a case study protocol and 2) developing a case study database. In this project, I have shown which cases of network sharing I have studied and why. I have noted down my results in Section 5.3 and drawn my conclusions from these findings. By reading this report, I believe it is possible to repeat my work and reach the same conclusions.
Tests Case study tactic
- Use multiple sources of evidence

Phase of research in which tactic occurs


Data collection Data collection Composition Data analysis Data analysis Data analysis Research design Data collection Data collection

Construct validity

- Establish chain of evidence - Have key informants review draft case study report - Do pattern-matching

Internal validity

- Do explanation-building - Do times-serier analysis

External validity Reliability

- Use replication logic in multiple-case studies - Use case study protocol - Develop case study database

Figure 2: Case study tactics for four design tests. Source: [4] 14

Research Methodology

4.4

Research method

The research in this study has been carried out in four steps. In the first step the research has narrowed the focus of the research and found credible sources by using analytical induction. Selected sources included for example network equipment manufactures, operators, analysts and regulators. Four sharing agreements that are in use by operators today were selected for further study. In the second step, this study looks at the four sharing agreements selected in detail. This work is qualitative descriptive work and has been performed by using secondary data. In the next step, this study focuses on finding the differentiation opportunities available to network operators today. This step was qualitative and achieved by gathering primary data from interviews and secondary data from literature studies. In the fouth step, this study looks at the impact of each sharing agreement on each differentiation opportunity. During all of these steps, the validity steps shown in 4.3 have been used to confirm all findings presented.

15

Building Shared Networks

BUILDING SHARED NETWORKS

To understand how the differentiation possibilities change in shared networks it is important to understand what a shared network is. This chapter describes the differences between four types of shared networks and wholly owned networks. 3GPP initially designed UMTS around the principle one operator, one network [56]. Network sharing was therefore not part of the original standardization process. There is currently ongoing standardization to facilitate network sharing. Therefore, this chapter also deals with the ongoing standardization process of shared networks in 3GPP. One type of sharing model is presented that has been suggested for future shared networks but is not in use today. This model is therefore not analysed further. Appendix C contains the basic structure of GSM and UMTS networks that is necessary to understand this chapter. Since network sharing became a frequent occurrence during the rollout of the UMTS networks, this chapter focuses on the sharing of UMTS networks. The primary focus of the solutions discussed in this chapter is on the solutions used by operators today.

5.1

Definition of shared networks

The actual definition of a shared network varies in different sources. The definition used in this thesis is based on [17], [56] and is defined as follows. A shared mobile network is when two or more operators cooperate in the process of building, maintaining and upgrading a mobile network. They have individual customers but both of their customers must in some cases use nodes belonging to the one of the other operators in the sharing agreement to gain access to the mobile network.

17

Building Shared Networks

5.2

Levels of network infrastructure sharing

The TONIC project [82] has created a model for describing nodes that operators can share in infrastructure sharing agreements. Figure 3 shows this model. In a wholly owned network, an operator owns all levels. In a shared network operators jointly own and/or maintains one or more levels. Although it is possible to mix the levels only certain combinations are feasible, are in use by operators today.

Level 1

Site Passive Elements

Level 2

Antennas

Level 3

Radio Base Stations

Level 4

RNCs

Level 5

MSC

VLR

SGSN

Unshared domain

GMSC

HLR

GGSN

Figure 3: Mobile network with sharing levels. Source: [13] There is no requirement to utilize only combination of levels throughout the whole network. Depending on capacity, coverage and licensing restrictions the operators can use different mixes of sharing levels. Furthermore, the agreement does not have to be between two operators, one operator can have several different agreements with other operators in different areas [24].

5.2.1

Access network sharing

The first three levels constitute the access network. Level one and two are per site while level 3 is on the RNC level. The definition of a site is the physical area where an operator places antennas and radio transmission equipment to provide coverage over a given area.

18

Building Shared Networks

To antenna

TRX (Operator A)

Shared Node B

TRX (Operator B)

To RNC

To RNC

Figure 4: Sharing Node B. Source: own Costs relating to the level one nodes are not specific to any access technology. Examples of these costs are leasing and acquisition, civil engineering, power/cooling systems, fences and housing. The second level is components that are specific to the access technology used such as antennas, feeder cables and couplers. Level three is the sharing of the radio base stations (Node B in UMTS) which connects to the level two elements. By adding more carriers, dividing the area covered into more cells or increasing power an operator can increase the capacity of a single site. For sharing of antennas multi-carrier antennas are required. Sharing a Node B requires special equipment that allows each operator to have their own connection from a Node B to the rest of their network. Logically, the shared Node B acts as two separate radio base stations but physically it is one Node B with connections for two operators with a separation on the software level of what traffic is routed to which operator (See Figure 4). Both operators must use their own carriers (implying that they have a unique MNC) for the base station to route the traffic to the correct RNC. In level four, the RNCs are shared. It follows the same principle as for sharing of Node B and requires specially equipped RNCs and is standardized by 3GPP [49]. It is possible for operators to connect their individual networks to the shared RNC and thereby mix a partly shared RAN with one or more and operators with wholly owned sites. This solution requires that operators have their own carrier frequencies and own MNCs.

To UTRAN

RNC (Operator A)

Shared RNC

RNC (Operator B)

To operator A core network

To operator B core network

Figure 5: Sharing RNC (Compare with Figure 4). Source: own 19

Building Shared Networks

5.2.2

Core network sharing

Level five nodes are the core network nodes (MSC/VLR and SGSN). When a connection has been routed to the MSC, it downloads customer data according to the IMSI number from the respective operators Home Location Register (HLR), which resides in the unshared domain in all sharing agreements. The MSC then stores the data in the Visitor Location Register (VLR).

5.3

Frequent sharing configurations

When sharing a network, operators have different reasons for doing so. Operators can mix any of the sharing levels but certain configurations are more useful at fulfilling these needs. Section 5.4 lists the needs that are the main drivers for network sharing. In a typical scenario, two operators are participating in the agreement. However, it is possible to extend most configurations listed in this section to allow for more operators unless stated otherwise. Sharing agreements between two operators is the most frequent scenario and therefore this is what has most support from equipment manufacturers. 3GPP discusses two solutions for network sharing that are outside the scope of this study. The first the case is when a single operator utilizes one core network for several access networks (E.g., GERAN and UTRAN). This study focuses on interoperator sharing and therefore does not discuss this case. The second case is when one operator covers a specific area and does not attempt to reach national coverage. The regulation agencies in most countries do not allow this and therefore this study does not discuss it. The names of the sharing configuration listed here are not universal. Different equipment manufacturers have different names for the same sharing configurations. This study amalgamates the names given by [24] and [56] tries to choose the most logical name.

5.3.1

Site sharing

Site sharing is the most basic form of a shared network. A typical site sharing agreement is for level one costs but site sharing also describes agreements that encompass everything up the Node B level. However, common to all agreements is that operators use their own carrier and MNC, implying that there is no loss of capacity. The operators will have identical coverage in the parts of the network where level two sharing is used. For operators that are incumbent GSM network operators site sharing means that they re-use their existing sites and/or lease them to UMTS operators. Finding good spots for building sites is difficult and time consuming. Operators are therefore often reluctant of sharing sites since this removes one of their competitive advantages [67]. 20

Building Shared Networks

5.3.2

Common Shared Network


Operator A Unshared Domain
Individual WCDMA network

Operator B Unshared Domain


Individual WCDMA network

HLR

HLR

Individual GSM network

Gateway MSC

Gateway GSN

Gateway MSC

Gateway GSN

Individual GSM network

MSC

VLR

SGSN

Level 5

RNC

Level 4

Node B

Level 3

Required Optional

Figure 6: Network sharing using a common shared network. Source: [24] In a common shared network, two operators deploy a RAN and level four core network where all operators jointly own all nodes, with one carrier and one MNC. Figure 6 shows this case. By forming a joint-venture company the operators own Level one to level four together. All operators individually own the nodes in the unshared domain (GMSC, GGSN and HLR). There are two cases where operators can deploy a common shared network. The first case is when two operators are sharing a single WCDMA carrier with one MNC. In the second case, operators have individual WCDMA carriers and unique MNCs for their WCDMA networks. The operators then assign the common shared network a unique MNC but still use carriers from both participating operators. As in the first case, the MSC connects to the operators individual WCDMA networks. Their jointly owned network connects to their individual WCDMA or GSM networks.

21

Building Shared Networks

5.3.3

Geographical Split Network

A geographical split network does not contain any shared nodes. Each operator has their own carrier and MNC and builds separate access networks and core networks covering different parts of the country. There are then three options for providing national coverage. Their first option is to sign a national roaming agreement. For load-sharing purposes, 3GPP has suggested that it should be possible to connect both radio access networks to both core networks but it is as of Rel-5 not standardized. Another solution, which is viable with the current releases, is for the operators to have separate radio access networks and a shared core network (Level 5). This shared core connects to each operators individual core network. Figure 7 shows the three cases. If the operators are GSM network operators (which cover the whole country), they must sign agreements for configurations of their radio base stations into providing handover into their individual GSM networks. E.g., an operator A customer is roaming in operator Bs network. When the connection is weak and a inter system handover is to take place, an operator A customer should be handed over to operator As GSM network. For selecting the correct network, the UE utilizes the list of preferred networks on the USIM. Each operator programs their USIM to have their own MNC as the preferred network. In areas where they both provide coverage the user will then be connected to his/hers home network. Each of their RANs will have a unique MNC. The MSC then sets the name of that network to display in the UE when the users views the list of available networks.
O p e ra to r A C o re N e tw o rk N a tio n a l ro a m in g O p e ra to r B C o re N e tw o rk O p e ra to r A C o re N e tw o rk O p e ra to r B C o re N e tw o rk O p e ra to r A C o re N e tw o rk O p e ra to r B C o re N e tw o rk

O p e ra to r A + B C o re N e tw o rk O p e ra to r A RAN O p e ra to r B RAN

O p e ra to r A RAN

O p e ra to r B RAN

O p e ra to r A RAN

O p e ra to r B RAN

(a )

(b )

(c )

Figure 7: Three ways to build geographical split networks. (a) Uses national roaming, (b) interconnects the two RANs with separate core networks (currently not standardized) and (c) uses a shared core network. Source: [56] Where they only have one network and utilize national roaming both set the MSC to use a common name. In areas where they both provide coverage the MSC is set to send the unique name of the operator. Although this name is only shown when the user manually selects a network from within the terminal. The UE always shows the name of the operator as stored on the USIM regardless of which of the networks it is connected to.

22

Building Shared Networks

5.3.4

Shared RAN

In a shared RAN agreement, the operators use equipment such that the RAN can logically acts as two separate networks, but physically still remains as one network. It uses two carriers and two MNCs, all operating on the same physical equipment. The operators can provide individual WCDMA coverage by connecting their own Node Bs to the shared RNCs.
Operator A Core Network (Unshared) Operator B Core Network (Unshared)

Level 5

Individual RNC (Operator A)

Logical RNC (Operator A)

Shared RNC

Logical RNC (Operator B)

Individual RNC (Operator B)

Level 4

Logical Node B (Operator A)

Shared Node B

Logical Node B (Operator B)

Level 3

Physical Logical

Figure 8: Shared RAN (compare with Figure 4 and Figure 5). Source: [24]

5.3.5

Common Spectrum Network Sharing

There are limitations to all of the aforementioned sharing configurations. For example, it is difficult or in some cases impossible to use any form of load sharing between operators. As is shown in Chapter 6 they are not flexible and create a number of problems. Therefore, 3GPP has suggested that newer releases should allow for a flexible sharing of all RAN resources. In an interview, Mr. Kristensson at Nokia Networks confirmed this view [74]. Ideally, operators would be able to share spectrum in some or all areas. This effectively creates a pool of spectrum. Any operator can then use this pool, even those operators without an individual license. Each customer would then, regardless of which carrier he/she connects to be able to choose between multiple core networks. This requires logic at the RNC level to forward the user to the correct core network. This sharing case is standardized in 3GPP Release 6 [66].

23

Building Shared Networks

5.4

Mapping of Needs to Sharing Configurations

Operators have different reasons of why to share a network. Table 3 shows a list of these needs. 1. Sharing costs of operators who have individual licenses. o Targeting dedicated networks in the near future o No immediate exit but possible if demand requires so in the future 2. Operators sharing spectrum o Long term sharing 3. Virtual operator entering a network o Closely tied to host operator (incumbent owning virtual operator) o External MVNO wishing switch host network as necessary Table 3: Needs for network sharing. Source: own research and [56] The first point implies that the participating operators are voluntarily entering the agreement and are doing so minimize their CAPEX and OPEX. The operators can view this as a short or long-term solution. In the short term, they expect a low initial traffic but eventually traffic will be high enough to justify a wholly owned network. In these cases, moving to individual networks when capacity demand increases is an option. Section 5.5 discusses this further. The operators might not expect the traffic to outgrow the capacity available in the whole network or parts of the network. In these cases, the sharing agreement is a long-term solution and the operators are not interested in a future migration. For the second case, it is assumed that one of the operators participating in the deal do not hold an individual WCDMA license but still wishes to build a UMTS network. The third case applies to pure MVNOs that have no interest in building and maintaining a UMTS radio access network but still wishes to participate in the market. The sharing configurations discussed fulfil different needs.
Need Cost reduction, short term solution Cost reduction, long term solution Spectrum sharing Virtual operator entering a network Configuration Site sharing, geographical split network Site sharing, shared RAN, geographical split network Common shared network Similar to the common shared network case (See Section 5.6)

Table 4 shows the optimal way of fulfilling these needs using the sharing configurations. Table 5 shows examples of where these configurations are in use today.

24

Building Shared Networks

Need Cost reduction, short term solution Cost reduction, long term solution Spectrum sharing Virtual operator entering a network

Configuration Site sharing, geographical split network Site sharing, shared RAN, geographical split network Common shared network Similar to the common shared network case (See Section 5.6)

Table 4: Mapping of sharing configurations to fulfilling operator needs. Source: own research
Sharing configuration Site sharing Geographical split network Common Shared Network Shared RAN Example 3GIS / SUNAB (Sweden), mmO2 and T-Mobile (Germany/UK) Vodafone / HI3G (Sweden) Telia / Tele2 (Sweden) Vodafone / Three (Australia)

Table 5: Shared networks in use today. Source: own research

5.5

Migration to Individual Networks

For operators that are sharing a network for lowering their cost at the expense of lowering the total capacity (fast roll out with low capacity) this can be a permanent or temporary solution. If the operators expect high future traffic demand increasing it must be possible for them to increase capacity after that the initial roll out with low capacity is complete. To maximize the return on investment (ROI), operators that anticipate this must plan their shared network with a future migration in mind. Migration to individual networks is also subject to license terms. The license may allow the operators to run a shared network at first but force them to move into individual networks after a certain time limit. For the common shared network without individual licenses, migrating to individual networks requires that one operator acquires an individual carrier. One operator then migrates into an individual core network and all operators migrate into shared RNCs (Level 4 sharing). The Common Shared Network is then effectively migrated to a shared RAN. In a Shared RAN scenario where there are capacity problems at the RNC level the operators can migrate into individually owned networks (own RNC infrastructure) but still keep site sharing agreements. 25

Building Shared Networks

In Site sharing (Level 1 only) agreements, migration to wholly owned networks is possible by building individual sites and RNC infrastructure.

5.6

Comparison between shared networks and virtual operators

Virtual operator is an umbrella term for three different types of operators. These three types of operators are Service Provider (SP) Enhanced Service Provider (ESP) and Mobile Virtual Network Operator (MVNO). They all have in common that they do not own any radio infrastructure but instead buy wholesale capacity from an MNO [42]. There are no clear distinctions between an SP, ESP and MVNO. Generally, the most basic form of a virtual operator is the SP. It does not own any own switching hardware but instead relies on a host operator to issue SIM/USIM cards, perform AAA, billing et cetera. The ESP is similar and do not add any additional technical differences 2 . The MVNO issue own SIM/USIM cards, owns all switching infrastructure (MSC, SGSN, GMSC, GGSN, HLR), thereby controlling the interconnection with other networks itself. This makes the technical solution identical to a common shared network (Section 5.3.2). Even though the MVNO solution is technically identical, it is important to note that there are important differences between an MVNO and an operator in a common shared network. A full operator controls the building and planning process the network. It is also difficult for the MVNO to offer inbound roaming agreements, which may force it into entering one-way agreements with other operators, which only contain outbound roaming or use the same roaming agreements as its host operator. The full operator can also restrict or delay when and how certain services are made available to the MVNO.

5.7

Status and Future for Roll-out of Shared Networks in Sweden

The two joint ventures, SUNAB and 3GIS both have commercially available UMTS networks running although they as of yet do not cover enough of the population to meet the terms set in the UMTS license. An important Swedish license term is that each licensee must individually cover 30% of the population. The operators participating in 3GIS both have individual licenses and have built a geographical split network. They both provide individual networks in

The main differences is a higher degree of customer ownership and different marketing strategies

26

Building Shared Networks

crowded areas, thereby fulfilling the criterion of individually covering 30% of the population. SUNAB has a common shared network and are therefore unaffected by the criterion. Apart from this, SUNAB and 3GIS have site-sharing agreements in some regions which also are unaffected by the requirement. There are not any capacity problems for any of the operators. In an interview with Mr. Axelsson, member of the board of directors for the National Post and Telecom agency, he stated that all UMTS operators are now building their networks with the lowest capacity allowed by the license terms in almost all parts of their networks [71]. Therefore, any migration to individual networks seems unlikely. Since the fourth license is unclaimed, there is a possibility that a new company would enter the market as a full operator. If they do not wish to build a wholly owned network, they need to collaborate with one of the two ventures for network sharing (and possibly site sharing for the wholly owned part of their network). Since none of the two joint ventures has built a shared RAN that would not been an option for collaborating with one of the joint ventures. They could build a geographical split network with either one or both companies in 3GIS. If they collaborate with SUNAB, they can either build a geographical split network with roaming into SUNAB or be part of the joint venture, creating a three tier common shared network. Both options would create the same network, thereby not affecting their options for differentiation.

27

Differentiation in Shared Networks

DIFFERENTIATION IN SHARED NETWORKS

This chapter shows the differentiation possibilities available to operators in whollyowned networks. The effect of each sharing agreement discussed in Chapter 5 on each of the differentiation opportunities is then presented. With a starting-point in the configurations listed in Section 5.3 a number of factors relating to operator differentiation changes.

6.1

Differentiation Opportunities

It is difficult to specify exactly what creates a differentiation opportunity that will appeal to customers. Even if a mobile network operator can offer a cheap unique service to its customers it will not matter how cheap that service is if a customer has no interest in using the service. In an interview with Mr. Mkitalo he put it as Its doesnt matter how cheaply you sell salted sausages, customers will still not be interested [76]. Therefore, it is important to know that the differentiation opportunities studied are viable. To know that the differentiation opportunities found during this study are viable this study focuses on differentiation possibilities used by operators today. Chapter 4 describes the methodology used for finding these opportunities. The focus of these opportunities is on those available to mobile network operators, although some of them are also available to virtual operators3. Another consideration for finding differentiation opportunities is to specify what customer group the

See Section 5.6 for a definition of the different kinds of virtual operators.

29

Differentiation in Shared Networks

opportunity directs. For mobile network operators, two customer groups exist: end users and virtual operators. For shared network joint ventures, two customer groups exist except for its owners, virtual operators and operators from other regions who wish to sign in-bound roaming agreements. Table 6 shows the result of the research on mobile network operators differentiation opportunities. This table is based on [17] that contain most of the network and marketing opportunities. The network related opportunities are then extended from [67] and [69]. The roaming and virtual operators opportunities comes from [70] and [73] respectively. The quality of service based opportunities are based on [74]. This study focuses on technical differentiation opportunities. Therefore, there is no analysis of the opportunities relating to marketing and customer care. This chapter analyses all network, terminal and in-bound roaming opportunities. Chapter 7 analyses network type opportunities relating to quality of service. Chapter 8 analyses wireless multi-access options.
Differentiation Capacity Coverage Reliability Offered services User differentiation Service differentiation Operator differentiation Bearers Branding Segmentation Pricing Terminal subsidies Terminal features Customer support Bundling with wireless access Out-bound roaming agreements Virtual operators and partnerships In-bound roaming agreements Type Network Network Network Network Quality of Service Quality of Service Quality of Service Quality of Service Marketing Marketing Marketing Marketing Terminal Customer care Bundling Roaming Joint-venture specific Joint-venture specific Directed towards End user End user End user End user End user End user End user End user End user End user End user End user End user End user End user End user Other operators / partners Other operators / partners

Table 6: Differentiation opportunities for mobile network operators and shared network joint ventures. Source: own research

30

Differentiation in Shared Networks

6.2
6.2.1

Analysis of Network Differentiation Opportunities


Coverage and Capacity

Coverage is total area where it is possible to access a given service. Capacity is the maximum total traffic density possible in a given region. When planning the rollout of a mobile network, operators decide their desired coverage and capacity. Figure 9 shows the network planning process in the rollout of a UMTS network. The dimensioning is the core block of the planning process since it is during this process that operators decide what type of network they want to build. During this phase, operators decide what regions to cover and acquire data on the radio propagation conditions of those regions. The expected traffic density and subscriber growth forecast together with the available spectrum determine the capacity available for a given user with a given service. Each cell then becomes either capacity limited or coverage limited. Capacity limited cells have a high traffic density that forces the operator into increasing the capacity in that area. Coverage limited cells on the other hand have a low traffic density (E.g., in rural areas). The operators try to make the cell sizes as large as possible in those areas. Chapter 8 discusses the coverage and capacity trade-off in detail. The coverage and capacity relates to the quality of service of a given cell. Quality of service is defined as the coverage probability, blocking probability and the end user throughput. Chapter 7 further discusses Quality of service. By dividing one cell into several smaller cells (known as cell sectoring) and increasing the number of micro base stations an operator can increase the total capacity of that cell. Each cell can gain higher coverage by increasing output power for a given service. Different operators may want to cover different regions or have different projections on the future traffic density and the bandwidth for a given service. Site sharing agreements that span the entire network removes the opportunity of better coverage by more sites but differentiation by smart antennas or increased power is possible at each site. Several coming technical solutions allow for shared antennas with individual coverage. Shared antennas do not offer any significant cost savings but allows for smaller, less intrusive towers [28].

31

Differentiation in Shared Networks

Input

O utput
R ough num ber of base stations and sites B ase station configurations

D im ensioning R equirem ents for coverage R equirem ents for capacity R equirem ents for quality

S ite selection A rea type / radio propagation B ase station configurations C ell specific param eters for R R M algorithm s C apacity and coverage analysis Q uality of service analysis O ptim isation

C overage and capacity plannin g N etw ork perform ance visualisation

M easured netw ork perform ance

A djustm ent of R R M param eters

Figure 9: WCDMA radio network planning process. Source: [8] In the cases of a common shared network where the operators are sharing a single carrier and a single MNC, providing extra coverage or capacity is not possible without allowing all participating operators customers to gain access to the same features. Therefore, an operator that wishes to differentiate by coverage will have to do so in areas where there are few or none of the other operators customers. This can be achieved in for example offices where the operator can bind a certain companys customers long enough for the investments in coverage of their offices to exceeds the operators investments in infrastructure. A term for this is customer steered coverage [72]. In a shared RAN the operators have individual MNCs and individual carriers. It is not possible to differentiate by coverage and capacity within the shared RAN but it is possible to have individual sites outside of the shared network. Coverage and capacity differentiation is therefore possible. In a geographical split network where only one operator provides coverage it is not possible to differentiate by individual coverage. Where there is overlapping coverage an operator can increase the number of sites and therefore increase the total capacity available over a given area. Hence, capacity differentiation is possible in areas with overlapping coverage (typically coveragelimited areas). The common shared network case where the operators have their own carriers and MNCs is similar to the shared RAN case. The operators can build individual sites with individual coverage. A difference from the shared RAN case is that this separate WCDMA network will have a different MNC than the shared one and roaming is necessary to facilitate the interworking of the two networks. 32

Differentiation in Shared Networks

6.2.1.1 High Speed Downlink Packet Access (HSDPA) High Speed Downlink Packet Access (HSDPA) is an access network upgrade that enables lower roundtrip times and higher downlink speed (10 Mb/s per cell). Appendix B contains a comparison between HSDPA and other access technologies. The main drivers for HSDPA is achieving higher throughput, reducing delay and achieving higher peak rates [63]. As the name implies, the advantages of HSDPA are only available in packet switched connections and only in the downlink. The design of HSDPA makes it most appropriate for streaming, interactive and background quality of service classes (See Section 7.3.2). Therefore, it is primarily a method to speed up non real time services such as mobile internet access. It relies on new type of transport channel, HS-DSCH that terminates in the radio base station. This is different from the Rel-99 packet switched channel that terminates in the SRNC. HSDPA changes the modulation techniques and the protocol that sends the data but all HSDPA cells can interact with non-HSDPA cells. The new air interface requires an upgrade of both the UEs and Node Bs. Since the updated protocol terminates in the Node B instead of in the SRNC, the access network requires an upgrade of both the Node B and the SRNC (Level 3 and Level 4). It does not require any new nodes in the core or transmission network except for an upgrade of the capacity in the transmission network. Since HSDPA increases the capacity ten-fold, operators cannot neglect the effect of this on the core and transmission network. In a common shared network where both operators share spectrum it is difficult for one operator to differentiate by providing HSDPA. In that case, the operator wishes to introduce HSDPA would have to upgrade the access network without the support of the other participating operator. They then need an agreement on how the other operator can allow his customers to access HSDPA if necessary. The operator not interested in HSDPA would effectively mitigate his risk since he can see how the other operator fares with HSDPA. This creates an unfair situation where one operator has mitigated his risk and that operator would therefore have to pay a premium to access it in the future. The same reasoning is applicable to a shared RAN situation although it is possible to introduce HSDPA outside the shared part of the network. In geographical split networks, one operator wishing to introduce HSDPA for the whole region would have to make an agreement so that the other operator also introduces it in his part of the network. HSDPA access is then according to the subscribers IMSI number. In regions where the cells are capacity limited, the participating operators will most likely already have individual networks. An operator 33

Differentiation in Shared Networks

can therefore increase the capacity of his network without affecting the other operator in the sharing agreement in those regions.

6.2.2

Reliability

Operation and maintenance (O&M) routines is an important factors for achieving a high reliability [17]. O&M interfaces are available in all nodes throughout the network and affect the sharing agreement. In spectrum sharing cases, as in the common shared network, the operators jointly own the whole WCDMA network and it is therefore plausible that they jointly own O&M nodes. Therefore, there is no room for differentiation by reliability. The same is partly true for the shared RAN case. In shared RANs, the operators jointly own some parts of their network and as in the shared spectrum case they will need a joint centre for O&M. Differentiation by reliability is therefore reduced in the shared RAN case. Operators participating in other shared network agreements must agree on the same standards for maintenance of the network. This is especially important in geographical split networks where one operator do not own any nodes in some regions. However, in capacity-limited areas where both operators have individual networks it is possible for one operator to differentiate by having higher reliability. In a common shared network, operators can compete with their reliability of the nodes in the unshared domain. In geographical split and shared RAN they can compete with these nodes and the Level 5 nodes.

6.2.3

Offered Services

Providing new services is one of the key differentiation possibilities for operators today [18]. The service layer handles all connections between the end user and the nodes running the service. This layer is separate from the network connection functions such as establishing and controlling connections. Table 7 lists some of the most important services or platforms for offering services in UMTS. This list is not exhaustive and a complete list of service aspects in UMTS is given by 3GPP specification series 22 and series 23 [77], [78].

34

Differentiation in Shared Networks

3GPP Rel-99 Circuit-switched voice service Location based services IMS based services OSA based services Mobile Application Execution Environment (MExE)

Table 7: List of some UMTS services and UMTS service platforms. Source: own research, based on [2], [77] and [78].

6.2.3.1 WAP Server and WAP Gateway The operator or a third party can run a WAP server that handles WAP requests. Both operator run WAP servers and third party WAP servers exists in the unshared domain. Third party run servers are typically not aware of what features are available in the underlying network and are therefore independent of network features [54]. They are therefore unaffected by network sharing. Operator run WAP servers can have specific applications that utilize features in the underlying network. Even though operator WAP servers exist in the unshared domain, the features available in the shared domain affect their capabilities. An example of such a feature is positioning where the user access a WAP site for acquiring positioning information. A WAP Gateway translates between WAP and HTTP requests. This enables an internet or intranet web-server to serve WAP requests [54]. An operator or a third party can provide the server. The servers reside in the unshared domain and therefore network sharing does not affect it.

6.2.3.2 Circuit-switched Voice Service The basic voice is a circuit switched service in both UMTS and GSM. It is possible to offer the same service through IMS (See Section 6.2.3.4), turning the service into a packet switched service instead of circuit switched service. It is therefore important to define that this section considers the circuit switched voice service. Table 8 lists the codecs that 3GPP has standardized for usage in GSM and UMTS. Some of these are multi rate, meaning that they can increase their bandwidth depending on the channel conditions. Increasing or decreasing the bit rate lowers/raises the quality of service. The actual codec and bit rate used is determined by the UTRAN where the decision parameters for changing codec or bit rate is set by some control parameters. Historically, speech codecs have been narrowband with a sampling rate of 8 kHz. This has also been the standard for GSM and UMTS codecs such as AMR [52]. Most 35

Differentiation in Shared Networks

of the energy of a speech signal is located above 7 kHz and is therefore lost when using narrowband speech codecs.. Wideband codecs use a higher sampling frequency. Therefore, they and can capture more of the energy and recreate a more natural speech [40]. 3GPP and ITU-T has standardized AMR-WB that operates at a sampling frequency of 16 kHz [53]. 16 kHz is the standard for wideband speech coding. Since the originating and terminating network of a call can differ, transcoders are required in all PLMNs. The transcoder in the originating PLMN decodes the voice signal and the terminating PLMN encodes the signal again. In UMTS networks, a media gateway (MGw) performs the actual transcoding. The MGw is a level 5 core node. 3GPP has introduced support for tandem free operation, which removes the transcoding. This approach allows direct MS-MS, UE-UE or MS-UE without any transcoding [57]. Therefore, changing which codecs to implement has an effect on the UEs and the media gateway. In a common shared network, an operator can differentiate by providing higher quality speech codecs that consume more bandwidth to all or certain customers. In a geographical split network, it is also possible but the operators must agree on setting the same control parameters.

ID / Codec GERAN GMSK GERAN 8PSK UTRAN ID / Codec GERAN GMSK GERAN 8PSK UTRAN

TDMA EFR n/a n/a n/a GSM FR YES n/a n/a

UMTS AMR2 n/a n/a YES OHR AMRWB n/a YES n/a

UMTS AMR n/a n/a YES OFR AMRWB n/a YES n/a

(GSM) HR AMR YES n/a n/a OHR AMR n/a YES n/a

(GSM) FR AMR YES n/a n/a UMTS AMRWB n/a n/a YES

GSM EFR YES n/a n/a FR AMRWB YES n/a n/a

GSM HR YES n/a n/a PDC EFR n/a n/a n/a

Table 8: Speech codecs in GSM and UMTS. Wideband codecs are listed as WB. Source:[51]

6.2.3.3 Location Based Services Location based services allows the user or another party to find the location of an UE. Depending on the use case and the quality of service specified by the operator, different levels of accuracy are required. 3GPP has standardized positioning with the following use cases in mind [46]:

36

Differentiation in Shared Networks

Value added services to subscribers Enhance O&M, supplementary services, create additional IN services Emergency situations Lawful interception (support of legally sanctioned positioning by external parties)

Standalone LMU (Location Measurment Unit), optional

SAS (Standalone SMLC)

Node B (associated LMU)

RNC

UE
Node B (associated LMU)

Core Network

RNC

Optional Required

Figure 10: Schematic view of location based services implementation in UTRAN 3GPP has standardized three methods that the UTRAN must support for performing positioning with various degrees of accuracy. These are cell ID based, Observed Time Difference of Arrival (OTDOA) and assisted GPS. The calculations required to determine the exact position for any of these methods (or a combination of several) can be network based, UE based, network based with UE assistance or UE based with network assistance. Figure 10 shows the structure of a positioning enabled UTRAN [48]. Three new nodes specific to positioning are available, stand-alone LMU, associated LMU and stand-alone SAS. The LMUs provide measurements for estimating the UE position and reports these to the CRNC. A stand-alone LMU is connected via the Uu interface (over the air) to the RNC and an associated LMU is logically (and possibly also physically) separate from the RNC and connected via the Iub interface. The stand-alone SMLC (SAS) provides GPS assistance data and acts as a calculation server to the RNC. The LMUs and SAS are not necessarily specific to WCDMA.

37

Differentiation in Shared Networks

H LR

MSC

GMSC

RAN

VLR

G a te w a y M L C

SGSN

GGSN

L e v e l 1 -4

Level 5

U n s h a re d d o m a in

Figure 11: Positioning enabled UMTS core network

The Cell ID based method determines the position of a UE by requesting the UEs cell ID from the SRNC. If the cell ID is unknown to the SRNC, it pages the UE. In case it is unreachable the last known position can be used. If the UE is in the soft handover state, it is possible to combine the measurement information from all RNCs in the UEs active set to improve accuracy. To improve accuracy, the SRNC can request measurements and calculations from one or more LMUs. OTDOA uses the timing difference between received frames in the UE. When the core network receives a request for positioning an UE it requests data from the SRNC. The SRNC sends measurements of the timing differences from the UE. The timing difference from several RNCs is necessary to maintain high accuracy. These measurements and the position of the Node Bs are then sufficient to calculating the UEs position. The CN can calculate the position of an UE either in UE-assisted or UE-based mode. In the UE-assisted mode, the UE performs the measurements but forwards them to the SRNC or SAS for calculating the position. In the UE-based mode the Node Bs send their exact geographical position to the UE, which measures their power and calculates its own position from these measurements. An UE may also be equipped with a GPS receiver. Instead of only using the measurements from the GPS the measurements are used together with data from the network to increase the accuracy. For the UE-assisted method, the SRNC requests measurement data from the UE and performs the calculations together with measurements from an LMU in the SRNC or at the SAS. In the UE-based method, the UE requests GPS Assistance data from the UTRAN and uses it to calculate its position. Common to all of these methods is that they require a Gateway Mobile Location Centre (GMLC) server in the core network. It connects via the Lg interface to the 38

Differentiation in Shared Networks

SGSN and MSC/VLR and the Lh interface to the HLR and is located in the unshared domain as shown in Figure 11. Furthermore, for any of the aforementioned methods to be accurate the Node Bs need exact measurements of their geographical position. By using GPS or external LMUs equipped with GPS placed at fixed positions it is possible to measure the position of the Node B. It is then possible to calculate the surrounding Node Bs positions. For all positioning solutions, the operator must know the location of the Node Bs. It is the SRNCs duty to translate the measurement information into a SAI or geographical coordinate, possibly with the assistance of a SAS. A SAS improves the measurement accuracy with the use of GPS and offloads the UTRAN and UE by performing computations. Table 9 shows the levels affected by the various types of positioning. An SAS is optional for all types of positioning shown in Table 9 and affects Level 4 sharing agreements (RNCs).
Type / Sharing level Cell ID based without LMU Cell ID based with LMU Observed Time Difference of Arrival Stand-alone GPS measurements in UE UE-based, network assisted GPS measurements UE-assisted, network based GPS measurements UE No No Yes Level 1 No No No Level 2 No No Optional Level 3 Yes Yes Yes Level 4 No No Yes Level 5 Yes Yes Yes

Yes

No

No

Yes

Yes

Yes

Yes

No

Optional

No

Yes

Yes

Yes

No

Optional

Yes

Yes

Yes

Table 9: Effects of positioning features on sharing levels and UEs. A Yes indicates that the Level influences the sharing agreement. Source: own research

6.2.3.4 IP Multimedia subsystem (IMS) One of the mayor changes in the upcoming releases of the UMTS standard is the adoption of the IP protocol. Adopting the IP protocol allows operators a greater degree of flexibility when developing new services and the development of new services that are not cost-effective to implement using the current structure for packet switched services in 3GPP Release 99. Providing these services requires an upgrade of the terminals and the core network. The IP multimedia subsystem is a standardized way for operators to build new 39

Differentiation in Shared Networks

services upon IP. Push-over-cellular (PoC) is one of the first services under study in 3GPP to implement as a pure IMS based service [64]. Eventually, all services offered operators, including the real-time voice service, could be IMS-based. It is therefore interesting to study the implications of IMS for service offering. IMS relies upon the Session Initiation Protocol (SIP). SIP is a light-weight application-level protocol that can create, modify and terminate sessions between one or more participants. Examples of sessions are voice service and multimedia streaming. It also encompasses functionality to connect to a user by a single unique address. SIP is not a subject of this study but an understanding of it is necessary to understand IMS at a lower level. A thorough explanation of SIP is available in [86] and [87] . IMS consists of a number of logical nodes that can all reside in the unshared domain or Level 5. The main node is the Call Server Control Function (CSCF) that can act as three different roles. During all roles, it generates CDRs for any activities. The ProxyCSCF (P-CSCF) is the first node that an UE has contact with in the home network. It authorizes the use of quality of service resources and forwards the message either internally or to another network. The Interrogating-CSCF (I-CSCF) is the first contact point for an incoming connection from another network. It assigns a ServingCSCF to the connection. The Serving-CSCF (S-CSCF) performs registration, charging and maintains one or more sessions for an UE. When a user registers with an S-CSCF he/she is assigned an S-CSCF as the S-CSCF handling the session may be different than the one handling the registration. Depending on the type of session, the CN assigns the UE different S-CSCFs. During an originating session terminating in another network it forwards the session to the correct I-CSCF. The S-CSCF forwards terminating sessions to the appropriate P-CSCF. IMS is implemented by upgrading the core network nodes and therefore affects geographical split networks and common shared network. IMS is not part of the virtual home environment (VHE) but IMS can carry out any services running on an SCS (See Section 8.4). Hence, a fully implemented IMS network in a common shared network or all parts of the geographical split network provides full roaming of all VHE based services.

6.2.3.5 Open Service Access The open service access is a flexible framework for implementing new services without affecting the underlying network functionality. One of the goals of OSA is to provide a flexible framework. It is then be possible to build new services without affecting already existing applications. OSA consists of three parts: applications, 40

Differentiation in Shared Networks

framework and service capability servers. Applications access network features through sets of Service Capability Features (SCFs). Typical SCFs are call control (initiate, disconnect et cetera) and data session control. The interface between the applications and SCFs is a standardized OSA API. Service Capability Servers (SCSs) support the OSA API by executing the commands issued. The framework contains internal mechanisms for authentication and service discovery. The applications and application servers can belong to a third party. For example, in shared networks with roaming agreements (E.g., geographical split networks) it is possible for one operator to own and run all application servers while accessing the SCFs of the other participating operator. Since applications are external, the operator providing the OSA API must authorize them. This authorization is done offline and/or online. The offline authorization includes an SLA where the application provider and operator agree on the cost for accessing different SCFs. Online authorization is more complicated and implies that OSA authorizes the application at the time of access by the end user. In the online authorization, the application must authorize itself to the framework to ensure that the application belongs to a trusted party.
Application Application server

Discovery

Open service architecture Framework User location Call control Service capability servers

HLR

CSE

S-CSCF

Servers

OSA internal API

Figure 12: OSA Architecture. Source: [61] Users that access the applications must also authenticate themselves to limit their access to the applications that they subscribe. The home environment must have a policy list that defines a list of services that a certain user can access. Figure 12 shows an OSA enabled core network. Since the applications and their respective servers exist in the unshared domain they can theoretically belong to a third party. However, operators can be reluctant to expose their interfaces to a third party. The OSA API accesses the SCSs and SCFs in the unshared domain. These 41

Differentiation in Shared Networks

connect to Level 5 shared nodes. Some of these nodes are also unshared and add another layer of abstraction before accessing the Level 5 nodes. In a common shared network, a single operator can implement OSA but it requires added OSA awareness in the shared CN. In geographical split networks, it is possible to implement OSA in one of the operators domain while still exercising some control of the connections going is through that operators domain. The visited network can provide some applications without the visited network being aware of the OSA applications.

6.2.3.6 USIM Application Toolkit (USAT) USAT is a standardized environment for applications stored on the USIM Integrated Circuit Card (UICC). It contains functions to access and execute certain logic in the supporting mobile equipment (ME). One such application is the Universal Subscriber Identity Module (USIM), which contains data that allows the user to enter the network. The USAT also contains a transport mechanism to provide new applications to the UICC. USAT specifies some functions relating to security. It assumes that the user has a trusted relationship with the issuer of the UICC. The UICC can interact with the ME (and vice versa) to perform certain functions such as initiating calls and data sessions. A complete description of the interactions with the ME is available in [48]. The USIM and ME applications can be upgraded via the mobile network. For example, it can change the IMSI number of the subscriber. An USAT server is located in the shared core domain (Level 5) [32]. This service capability is therefore affected by all network-sharing agreements that include core network sharing, i.e., a common shared network and some forms of geographical split networks.

6.2.3.7 Mobile Station Application Execution Environment (MExE) MExE is a standardized environment for external applications or applets to negotiate the capabilities of an UE and the services available in the network. Applications that run on the UE can interact with the network to provide additional services to the user. Nodes that affect the service environment are both in the unshared domain (such as WAP servers) and in the shared domain (such as positioning servers) [55]. The nodes available in the shared domain therefore affect the applications that an operator can offer.

6.3

In-bound Roaming in Shared Networks

The traditional use of roaming is when a subscriber is visiting another operators network, for example when travelling abroad. A customers home operator can have roaming agreements with several operators in the visited region with one being the 42

Differentiation in Shared Networks

preferred network operator. The USIM contains a list of networks to which the customer has roaming agreements. The operator can set one of them as a preferred network when the operator has roaming agreements with several operators in the region. When the terminal can see several networks to which it can connect the differentiating factor is the output power or quality of the networks. The terminal will try to connect to the network from which it can get the best quality of service. These roaming customers are highly profitable for operators and it is therefore desirable to acquire these customers [70]. When a roaming customer has agreements with two or more operators, as often is the case, the operators can do their best to compete against each other to acquire the customer. However, when operators share the same network this creates a problem. It is common for operators to acquire roaming customers at specific spots where they enter the region such as airports or ferry terminals [70]. Therefore, operators have high marketing costs in those areas for being the preferred operator for customers that enter the region

6.3.1

Acquiring Roaming Customers

In sharing cases where the operators have their own carriers and unique MNCs such as site sharing or shared RAN this is not a problem since the assignment of a roaming customer is done according to the same rules as for wholly owned networks. In geographical split and common shared networks, operators must agree on how to divide the revenue from the roaming customers when a visiting subscriber enters the shared network. In a geographical split network, the participating operators can try to divide the area that they cover separately in such a way that there is an equal amount of roaming customers in those areas. In cities, airports and train stations or other places where there is a large amount of roaming customers they can each provide individual coverage. In this situation, they will compete in the same way as for the case with no sharing or only a site sharing agreement. The problem with this approach is that it requires all participating operators to have a good idea of the amount of roaming customers in areas where they only get coverage by national roaming agreements. If the operators can agree on sharing data on all their roaming customers they can both analyze and estimate which areas outside of their own coverage has a large amount of roaming customers. They can then proceed with building individual coverage in those areas. This approach is difficult since it requires sharing of data between operators and it may be difficult for one operator to provide coverage in an area that is far away from the remaining access network of that operator.

6.3.2

Dividing Revenue from Roaming Customers


43

Differentiation in Shared Networks

The simplest solution for all shared networks is to split the revenues equally among the participating operators. A solution, applicable to both common shared networks and geographical split networks, is that the participating operators sign an SLA on how to divide the roaming customers. They must agree on the value of each roaming customer and the cost of the utilization of the resources of having that subscriber in each operators network. This cost should ideally be the same for both operators networks. If a subscriber enters that has an agreement with both participating operators they can split the revenue equally among themselves. If one operator has a high number of in-bound roaming agreements where the operator is the preferred operator, the aforementioned solutions can be se seen as disadvantageous. Both of the solutions are disadvantageous if one of the operators has acquired a high number of the GSM roaming customers where the agreements also cover UMTS roaming. For example, consider a major incumbent operator that is sharing the network with a small upstart greenfield operator. The major operator will have several GSM roaming agreements, which it can extend to cover UMTS. In a common shared network, both operators are acquiring a circuit- or packet switched service from a joint venture owning the level 1 to level 5 network infrastructure [30]. If the operators are prepared to split the roaming revenues evenly one option is then to let the revenue from roaming customers go to the joint venture and utilize it for maintaining the shared infrastructure. Two intrinsic problems exist with all solutions that split the revenues equally among the operators. As mentioned earlier it is almost inevitable that one operator will feel like being at a disadvantage. Another problem with splitting revenues equally is that operators may rely on the other operator to attract roaming customers. Hans Eijvergrd from TeliaSonera mentioned in an interview that one way for operators to compete for roaming customers is to assign the roaming customer to the operator that can provide the best service level features. For example, if a customer enters and has a CAMEL subscription with several value added services the operator that best can fulfil these services will receive the customer.

6.4

Summary of Network and Roaming Related Opportunities

Table 10 shows a summary of network related differentiation opportunities. As the table shows, for most sharing agreements differentiation opportunities exist in the service area. The opportunities for geographical split networks differ if, in a certain region, the network is capacity-limited or coverage-limited. When operators have individual networks (such as capacity-limited areas), there are no effects on the differentiation opportunities in the network layer when compared to a wholly owned network. However, any differentiation of services in capacity-limited areas will not affect the rest of the shared network. 44

Differentiation in Shared Networks

IMS based and OSA based services requires an upgrade of both the unshared part of the CN and the Level 5 elements. OSA allows operators to develop services in the unshared domain. Operators should therefore cooperate to introduce these elements as they benefit all operators in the sharing agreement. Figure 12 shows a UMTS network where OSA and VHE are used to separate the service creation layer from the underlying network. Another consideration for operators in sharing agreements is how to handle roaming customers. Since both participating operators are interested in making as much as possible from their previous roaming agreements there is a risk that their joint venture becomes uncompetitive. The operators must therefore take care to divide the roaming customers in such a way that they both have incentives to increase the number of roaming agreements and at the same time provide a competitive offering to those customers.
Common Shared Network Limited None Limited None Shared RAN Site sharing4 Geographical split network5 None None Limited None

Capacity Coverage Reliability Circuitswitched voice service Roaming Location Based Services IMS based services OSA based services

Full Full Limited Full

Full Limited Full Full

Limited Limited Limited Full

Full Limited Limited Full

Full Full Full Full

Limited Limited Full / limited Full

Table 10: Summary of network related differentiation opportunities. Source: own research and [17].

Given that the site sharing agreement covers all sites. If the operators only share some sites, they will have the same differentiation opportunities as in a wholly owned network. 5 This is valid for coverage-limited areas. In capacity-limited areas, the operators will most likely have individual networks and it is therefore no difference between a shared and a wholly owned network. Although, this means that some services might only be available in the capacity-limited parts of the networks.
4

45

Differentiation in Shared Networks

Figure 13: Separation of service development layer and the underlying UMTS network. Source: [75]

46

Quality of Service Differentiation in Shared Networks

QUALITY OF SERVICE DIFFERENTIATION IN SHARED NETWORKS

This chapter discusses two uses for quality of service in shared networks. First, a brief introduction to quality of service is given and the IP and UMTS quality of service architectures are explained (Section 7.1 - 7.3 ). Quality of service is then analysed form two perspectives in shared networks. The first part shows the problems with having a consistent quality of service in roaming based geographical split networks. The second part discuss different methods for allowing operators to provide the same service over the same access network but with differentiated quality of service. Quality of service is a wide term that encompasses many different meanings. In circuit switched services, quality of service usually means full service or no service at all. In GSM, most services utilize circuit-switched connections, but in UMTS packetswitched services are becoming more frequent.. This development is expected to continue as operators introduce IMS. All CDRs generated in 3GPP Release 99 compatible UMTS networks save the quality of service parameters used for each bearer, therefore making quality of service based charging possible.

7.1

Introduction to Quality of Service Differentiation

Before implementing any schemes to achieve quality of service differentiation operators must understand what they are trying to achieve. Quality of service differentiation is a way of providing a better service to certain users or classes of 47

Quality of Service Differentiation in Shared Networks

services in congested networks. If there is spare capacity in the network there is little or no reason to degrade a users experience. For fibre networks it has been suggested that new technology will make bandwidth abundant and therefore make quality of service schemes a solution to a problem that does not exist (See [21] and [22]). While it is possible to upgrade all parts of the access and core network with new capacity as discussed in Section 6.2.1 the air interface remains a bottleneck [41]. This is especially true in shared networks involving spectrum sharing where one of the operators participating in the agreement may not have full control of when to increase the capacity. For shared RANs and geographical split networks, the operators will most likely have individual networks in capacitylimited areas. Table 11 lists broad quality of service groups for differentiation between operators in shared networks. User differentiation means that a certain group of users receives a higher or lower quality of service. Operator differentiation differentiates the quality of service level of all services to users belonging to a certain operator, while service differentiation differentiates each service to all users. These quality of service groups are not mutually exclusive. For example, distinguishing all users from a given operator that are using a certain service is possible.

Operator differentiation User differentiation Service differentiation

Table 11: Quality of Service differentiation possibilities in GSM and UMTS. Source: [74]

Since quality of service lowers the quality for a certain group of users, it is important to define what lower quality of service means for a given service. In circuit-switched connections lower quality of service means an increased number of blocked or dropped calls. For the circuit-switched voice service in GSM and UMTS users may tolerate very few dropped or blocked calls. A degradation of the quality of service will therefore be unacceptable.

First thought of the operator: We want quality of service differentiation. But what IS quality of service differentiation? Mr. Kristensson, Nokia Research [74]

48

Quality of Service Differentiation in Shared Networks

In packet switched networks, lowering bandwidth during busy hours can be an acceptable degradation of the quality of service. Therefore, quality of service is more interesting in packet switched networks. Quality of service is also of strategic importance to network operators since it affects how they build their network [2]. As shown in Figure 14 there is a relationship between the delay in the network and the investments in the network. Depending on what services the operator expects to provide, lowering investments for a longer delay can be a good trade off. Operators participating in shared network agreements must therefore reach an agreement on how to optimize the network.

Figure 14: Relationship between delay and the requirements on the network. Source: [2]

7.2

Aggregated and end-to-end service levels

For circuit switched services such as those that the telecom world historically has provided QoS has been a measure of the number of dropped and blocked connections. Since the traffic in wireless networks also has been circuit switched, the same measures have been applied in those networks. New measures have also been relevant such as total operator coverage. UMTS and IMS will make it easier to introduce new packet switched services. In packet switched networks, the network handles flow control and state mechanisms on a higher level in the stack and it is therefore not relevant to discuss connections. The measures of dropped and blocked connections are hence not relevant. Instead, four attributes affect the quality of service: delay, jitter, packet loss/reliability and available peak bandwidth [2], [21]. Delay is the time passing between a node sending a packet and another node receiving the packet. Jitter is the variation in this delay over time. Peak bandwidth is 49

Quality of Service Differentiation in Shared Networks

the maximum amount of packets that a network can send during a certain period. The packet loss measures the percent of the packets the networks drops. For any of these attributes to be useful there must be a definition of how they are measured. There are two levels on how to measure the quality of service: end-to-end service and aggregated QoS [35]. Aggregated QoS is defined by agreements stating high-level QoS requirements between the user and the network provider. Typical control parameters are percent of packets dropped and average delay. This means that there are no measurements on the individual traffic flows. Instead, network traffic is measured over aggregated traffic flows during a time specified by the SLA. Figure 15 (b) shows the differentiated services model that is a typical model for achieving aggregated QoS. The edge routers shape the traffic flows to ensure that they conform to the SLA. The core routers then do not contain any logic for shaping quality of service. End-to-end quality of service on the other hand specifies per-flow attributes and therefore sets higher restrictions on the underlying network. Hence, for the network provider to be able to guarantee end-to-end quality of service the provider must own or control all nodes between the applications communicating.

Core router Router Router Router Edge router Core router

Core router Edge router

(a)

(b)

Figure 15: Integrated services (a) and differentiated services (b) for guaranteeing a certain quality of service. Source: [21], [43]

In geographical split networks, both operators must set demands for the quality of service of the areas where their partner provides roaming. In jointly owned networks such as shared RAN and common shared networks all participating operators must agree on common standards for the quality of service in the shared network. Since the agreements runs over long times and it is impractical to measure the quality of every connection aggregated quality of service is more applicable in these cases. For operator, user and service differentiation the quality of each individual connection matters. Therefore, the operators must measure end-to-end quality of 50

Quality of Service Differentiation in Shared Networks

service. For end to end quality of service to possible it is necessary to have full control of both the originating and terminating point in the connection chain. Figure 15 (a) shows the integrated services model where a protocol reserve capacity in each node in the routing chain before the actual traffic flow starts. Section 7.3.2 describes how this is achieved in UMTS.

7.3
7.3.1

Quality of service architectures


IP

Quality of service in UMTS and IP networks are similar because of two reasons. Packet switched services are for the most cases accessed through an IP network, such as an intranet over VPN or internet services. In the core network, IP is being introduced in the form of IMS and the current evolution path for UMTS networks are towards an all-IP core [25]. The reasons for operators to introduce all-IP cores are outside the scope of this study. Providing a specific quality of service whether measured through aggregated flows or in an end-to-end manner in UMTS is therefore dependant on the QoS of the connected IP network(s). The same requirements and level of control is necessary whether this IP network resides in the core domain or through a third party. It is therefore necessary to first study IP quality of service before studying UMTS quality of service. Regular IP networks do usually not have QoS mechanisms implemented. Two systems have been proposed for implementing QoS, integrated services (IntServ) and differentiated services (DiffServ). Integrated services are based on the idea that for end-to-end quality of service to be possible all nodes from the originating to the terminating application must reserve resources. Before opening up a stream, resources are reserved in all nodes between the applications as shown in Figure 15(a). State information is therefore necessary in each node for each stream. This puts a high strain on the nodes involved in the process [21]. The differentiated services model assumes that packets travel through different domain owned by different network operators. At the edge of each cloud of routers an edge router performs packet shaping and policing, prioritizing packets from different customers according to SLAs.

7.3.2

UMTS

Figure 16 shows the end-to-end quality of service architecture in UMTS. The UMTS network transports all services carried over bearer services. One bearer service is a connection between two ends, for example between two terminals in different PLMNs.

51

Quality of Service Differentiation in Shared Networks

TE

MS

UTRAN

SGSN

GGSN

TE

End-to-end service
TE/MS local bearer service External bearer service
CN bearer service Backbone network service

UMTS bearer service

Radio acces bearer service

Radio bearer service

Iu bearer service

UTRA FDD/ TDD service

Physical bearer service

Figure 16: UMTS end-to-end quality of service architecture. Source: [50]

It is important to note the distinction between CN bearer services and the RAN bearer services. The design of UMTS separates the core network protocols from the air interface, allowing for a variety of radio access techniques. The bearer service allowing communication from a GGSN to an UE is named the UMTS bearer service. When the originating or terminating end of the connection is in another network (e.g., the internet or another PLMN) the connection has to rely on the external bearer service. A connection from one originating UE to a terminating UE is called UMTS end-to-end service. As the operator has full control over the nodes in the CN and UTRAN any quality of service problems will most likely happen in the radio bearer service and the external bearer service. Table 12 lists the different categories of applications and their characteristics as defined by 3GPP. Some of the classes are asymmetric, i.e., there is an uneven distribution of the traffic density in the receiving and sending channel. A thorough discussion on how these classes are realized in the different bearer services can be found in [2]. Note that there is no requirement on using a specific class for a specific service (E.g., using the streaming class for streaming multimedia). The possible applications listed are recommendations and the operator can configure the network to assign all services with conversational bearers for example. When an UE initiates a connection it requests the UMTS bearer service from the CN with a number of parameters including specified maximum bit rate, guaranteed bit rate, transfer delay and if these demands are negotiable or not. The CN then proceed by requesting a radio access bearer with the appropriate properties. The UTRAN establishes the radio link and the end-to-end service between network and UE is complete. In the packet switched domain, a Packet Data Protocol (PDP) such as IP carries the actual data. 3GPP defines one connection over a PDP protocol as a PDP context. Each UE can have several PDP contexts with varying quality of service. The PDP context has all parameters as set by the UE during the activation state. However, it is possible to re-negotiate the quality of service during an ongoing session.

52

Quality of Service Differentiation in Shared Networks

Table 13 lists the parameters that an operator can set for each bearer. Apart form these parameters, the actual traffic class is also a parameter when requesting a bearer. With the traffic class as a parameter, the network can make assumptions about the traffic stream. A detailed explanation of each attribute is available in [60].
Traffic class Conversational Characteristics Possible applications Low delay and jitter, no buffering, Voice service, video symmetric, guaranteed bit rate telephony, video games Streaming Low delay, moderate jitter, buffering Streaming multimedia allowed, asymmetric, guaranteed bit rate Interactive Moderate delay and jitter, buffering Web browsing, network allowed, asymmetric, no guaranteed bit games rate Background High delay and jitter, buffering allowed, Background synchronizing asymmetric, no guaranteed bit rate of e-mail Table 12: The QoS classes in UMTS and their possible applications. Source: [7] and [60]

Traffic class Maximum bitrate Delivery order Maximum SDU size SDU format information SDU error ratio Residual error ratio Delivery of erroneous SDUs Transfer delay Guaranteed bit rate Traffic handling priority Allocation/retention priority Source statistics descriptor Signaling Indication

Conversational class X X X X X X X X X

Streaming class X X X X X X X X X

Interactive class X X X

Background class X X X

X X X

X X X

X X X X X X X X

Table 13: UMTS QoS bearer parameters. Source: [60]

53

Quality of Service Differentiation in Shared Networks

7.4

Quality of Service in a Shared Networks

The external bearer service is in the unshared domain of the operator and is therefore unaffected by all sharing agreements. When the external bearer service relies on nodes outside of the operators control, SLAs are required for end-to-end quality of service Studies on MVNOs suggest that a differentiation opportunity for virtual operators is to provide their own switching and backbones interconnections [5]. Instead of routing traffic over the internet, the MVNO can configure the GGSN to route traffic for a certain group of customers through leased lines onto for example an intranet. Since the GGSN is in the unshared domain, MNOs in any sharing agreements have the same differentiation opportunity. As Table 14 shows, implementing quality of service measures in UMTS without cooperation between operators is difficult. Achieving end-to-end quality of service without cooperation is even more difficult since it requires control of all nodes in the connection chain. Therefore, operators in most sharing agreements will be forced to cooperate on defining quality of service parameters for their respective customers. This is a problem since some of the parameters listed in Table 13 are difficult to accommodate simultaneously. For example, as shown in Figure 14 a network optimized for low latency is built differently than a network optimized for high delay (in exchange for, e.g., high capacity). If the operators are delivering services that require high delay and high capacity they will need to build a network that is over provisioned for both types of services, which in turn creates high costs. This means that if operators will be launching services with mutually exclusive quality of service attributes they will have a much higher costs than an operator with a wholly owned network or a shared network where the operators plan to launch the same types of services.
Type of network Geographical split network, coverage limited Geographical split network, capacity limited Shared RAN, coverage limited Shared RAN, capacity limited Common Shared Network Site sharing Bearers affected by a sharing agreement All except the external bearer service None All except the external bearer service None All except the external bearer service None

Table 14: Bearers in UMTS affected by the different sharing agreements. Source: own

54

Quality of Service Differentiation in Shared Networks

7.4.1

Quality of Service differentiation

Without quality of service differentiation, the network handles all users in a besteffort class, indicating that all services compete for the same resources on equal terms. As shown in Section 7.3.2 a number of parameters affect the end user experience relating to quality of service. In UMTS, it is useful to group some of these parameters into four different classes to realise different usage scenarios. UMTS can then differentiate quality of service by allowing all services to run in any of the classes defined. Apart from the 3GPP QoS classes, a number of classes are defined for IP networks in techniques such as differentiated services. These classes group similar parameters together and a number of different service classes for end users have been proposed. A differentiated services approach proposes the following quality of service classes: Expedited Forwarding Assured Forwarding Best Effort

In UMTS, the background class corresponds to a best-effort approach while streaming and conversational equals the expedited forwarding class. Interactive equals the assured forwarding class [22]. Classes for differentiating users who are using the same services have been proposed in [5], [21], and [22]. Table 15 lists these classes. Premium service Olympic service (gold, silver, bronze) Table 15: Two broad classes for quality of service differentiation between users The premium service is for applications requiring a higher reliability than the besteffort service. The gold, silver and bronze services are intended for different levels of the same group of attributes. For example, if the gold class has lowest latency available the silver class users have the double of this latency. This implies that during network congestion the network shifts resources away from users or applications without assured services. The actual implementation of this depends on the differentiating attribute. For example, Figure 17 shows a scheme that differentiates users in a mobile broadband application with different guaranteed bit rates based on olympic-classes. Another example is for the attention/retention policy, which specifies at what priority the network drops bearers when they are scarce. One operator could have a higher priority for its bearers in certain services. This would be an example of a premiumclass service.

55

Quality of Service Differentiation in Shared Networks

Operators can develop similar schemes that differentiate on for example latency or maximum bit rate.

Figure 17: Figure (a) shows a congested cell with only one user accessing packet-switched services. Figure (b) shows the same cell with several packetswitched users without QoS planning. Figure (c) shows the same cell with olympic quality of service differentiation (the users are assumed to be using the service internet access as in figure a and b). Source: own research

56

Multi-Access Networks

MULTI-ACCESS NETWORKS

As the Chapter 6 and Chapter 7 show there are few differentiation opportunities on the network level for operators participating in any of the shared network agreements discussed in Chapter 5. During the interview phase of this study, many respondents emphasised the use of GSM/EDGE and WLAN for differentiation [67], [69], [71], [73], [72]. Therefore, this chapter studies the usage of multi-access networks for operators to differentiate themselves by capacity and coverage. This topic is large and is therefore given a special chapter. Incumbent GSM operators already differentiate themselves by providing a GSM network and they can use this opportunity to extend their network by introducing EDGE in their RAN. Both incumbent GSM operators and greenfield UMTS operators can also build a WLAN hot-spot network. This chapter evaluates where multi-access operators should extend their UMTS network with EDGE and where it is profitable to extend it with WLAN. There are currently a number of competing standards for WLAN. This study focuses on technologies commercially available today and therefore focuses on IEEE 802.11g.

8.1

WLAN and EDGE Interaction with Shared Networks

A number of technical problems exist to provide seamless roaming between GSM, EDGE, UMTS and WLAN. Apart from the availability of terminals that can move between different accesses the operators must also be able to provide a seamless service experience between the different networks. The problems with service provisioning across heterogeneous networks is similar to the problems of service provisioning for operators in roamingbased geographical split networks. Section 8.3 analyses the new uses for roaming and what features a UMTS core network provides for a seamless service provisioning for both geographical split networks and multi-access networks.

57

Multi-Access Networks

WLAN poses an additional problem of service provisioning. Although theoretically possible, WLAN-interworking with 3GPP systems is problematic. 3GPP has set a number of use cases for WLAN interworking. The simplest case is when the two networks only share customer care and a common billing architecture. Although a 3GPP core network for common authentication, authorization and accounting (AAA) is standardized, achieving service provisioning in both networks and continuity is difficult. If a user is to access services in the UMTS packet-switched domain such as IMS based services, positioning services and multi-media messaging this requires a higher level of technical interworking. Since IMS based services are IP based these have the lowest barriers for WLAN interworking [44]. Most of these services rely only on nodes in the unshared domain and are therefore unaffected by any sharing agreements. An exception is the positioning service that is dependant on the UTRAN (See Section 6.2.3.3).

8.2

Differentiation by Capacity and Coverage

An access network can be either capacity or coverage limited. There is always a tradeoff between increasing capacity and increasing coverage. Smaller cells result in total higher capacity while large cells cover a large area but decrease the total capacity. In coverage-limited areas, the base stations operate at the largest possible reception area and do not reach their maximum capacity. In capacity-limited areas, the cell sizes are as small as technically possible and operate at their maximum capacity. The cells covering the largest possible areas are named macro-cells. Smaller cells are named micro-cells or pico-cells. Figure 18 shows a comparison between the different cell sizes.

Macro-cell

Pico-cell

d (a)

d (b)

Figure 18: A coverage-limited region covered by a macro-cell and a capacitylimited area covered by pico-cells

8.2.1

Analysis of Sharing Agreements and Multi-Access differentiation

58

Multi-Access Networks

Capacity differentiation is most interesting in urban areas where there are a high number of customers over a small area. In urban areas where many cells are capacitylimited, operators participating in geographical split networks agreements will most likely already have individual networks to accommodate the large number of customers. Operators sharing spectrum will not have this opportunity and capacity differentiation by multi-access networks is therefore of most interest to them. Geographical split network operators can benefit from an increased capacity if the UMTS pico cells do not provide enough capacity for the services they wish to provide. In both cases, capacity differentiation is most interesting by adding an equivalent of pico-cells. WLAN has cell sizes similar to those of pico-cells and is therefore most interesting to study in the context of capacity differentiation.

As show in Section 6.2.1, differentiation by coverage is not possible for operators in geographical split networks and common shared networks. Since capacity limited areas are the most profitable areas, operators will already have full coverage in capacity-limited regions. Differentiation by coverage is therefore most interesting in coverage-limited areas. The replacement technology for UMTS macro cells in coverage-limited areas under study is GSM upgraded with EDGE. Although there are other replacement technologies available such as standards by the WIMAX forum, the focus of this study is on mobile network operators and their options available today. WIMAX is therefore left for further study.

8.2.2

Evaluation Theory

The focus of this evaluation is on Sweden. To extend the calculations for an arbitrary country it is necessary to find the limits of rural, suburban and urban areas for that country. The definitions in this study are from the Swedish National Post and Telecom Agency and Statistics Sweden [89], [90]. Table 18 lists the population distribution in Sweden and the limits for classifying areas as urban, suburban and rural.

Geotype Area of Population of definition genotype (km2) geotype Urban Suburban Rural 2 250 90 000 357 750 4 500 000 4 500 000 1 000 000

Average population density (pop/km2) 2 000 50 2,8

Table 16: Population distribution in Sweden. Source: adapted from [89]

A number of parameters define what cell size to use for the rollout of UMTS or GSM in a given area. The operator must first estimate the total capacity necessary. The number of subscribers in that area and the average data rate of the services they use determine the necessary capacity. By using a mix of macro, micro or pico cells, the operator can then achieve the required capacity. The cost of each type of base station must then be compared to the base stations performance in terms of maximum range and the maximum capacity possible within that range.

59

Multi-Access Networks

A general cost model for a heterogeneous access network is according to Equation 1. As shown it is necessary to estimate the number of base stations of each type and the cost of each base station. This calculation assumes that the cost of each base station is stated in net present value (NPV).

C system = c macro N macro + c micro N micro + c pico N pico


Equation 1: Relation between the cost of a heterogeneous network and the number of sites. Source: [30] For all investments, it is necessary to estimate the initial fixed costs and the running costs. The initial cost of an investment is referred to as capital expenditure (CAPEX) while the running costs are referred to as operating expenditure (OPEX) [19]. The CAPEX of a single base station is determined by the equipment cost, site build out and site installation. Each type of base station (macro, micro and pico) has a specific CAPEX and OPEX. Equation 2 shows the method for calculating the cost of a single base station in NPV terms. The TONIC project has estimated the time for writing off an access network such as UMTS or GSM at 10 years [13]. According to [19] the average discount rate in companies is at approximately 10%. Therefore, the net present value calculations in this study have used this as the discount rate.

cbs =

K 1

ck k k = 0 (1 + d )

Equation 2: Net present value of a single base station. Source: [30]

A N W Nbs = service , user user 2 Wmax R max


Equation 3: Number of base stations when accommodating for capacity and coverage limited areas [30]

8.2.3

Evaluation Results

Table 17 shows a comparison of cost and performance for the three types of UMTS base stations, WLAN (802.11g) and GSM EDGE. Given a certain data-speed, Equation 3 gives the total cost per-user and square kilometre to build a mobile network that has enough capacity to achieve the given speed. It is then possible to compare the cost of using different access technologies in different areas. Figure 19 shows the economics of a UMTS network in a highcapacity scenario with an average data rate of 10 kbit per second in the busy hour per user. Equation 4 shows that this is equal to 4500 kB per hour. The cost for each additional pico-cell when the UMTS network consisting of macro and micro cells is capacity limited is at 3 712.

60

Multi-Access Networks

Figure 22 shows a comparison between building GSM-EDGE macro base stations and UMTS macro, micro and pico base stations (compare with Figure 19). The economics for GSM-EDGE is identical to that of UMTS macro cells until there are more than 71 users per square kilometres at 10 kbps. GSM-EDGE can therefore be used for coverage differentiation in some areas. It should be noted that EDGE has a higher latency than UMTS for data services(approximately 700ms compared to 100ms) which will make some services impossible in EDGE (packet-switched videoconferencing and speech for example). Figure 23 and Figure 24 shows a comparison between UMTS macro cells and GSM EDGE macro cells for lower-capacity scenario of 1 kbps in the busy hour per user. Similarly as shown in Equation 4, it can be seen that 1 kbit per second in the busy hour equals 450 kB per hour.

3600 10 10 3 = 4500 10 3 bytes 8


Equation 4: 10 kbps busy hour traffic corresponds to 4500 kB per hour

Figure 20 shows the same scenario but with WLAN 802.11g access points instead of UMTS micro cells. The cost for each additional WLAN access point when the UMTS network is capacity limited is at 34 000, 9.3 times the cost of an UMTS pico cell. It can clearly be seen that the price decline for WLAN comes only at extreme hotspots.

It should be noted that this is a basic evaluation, which leaves out some additional costs (e.g., marketing), and cost savings that are possible for operators (e.g., shared transmission). For a more profound analysis of the economics of building and running WLAN and UMTS networks see the TONIC project [82]. Many operators are reluctant to give figures on their exact costs. The figures for UMTS and EDGE are taken from [30]. The figures for WLAN can be seen as a worst-case scenario and are from internal informal interviews at TeliaSonera.

61

Multi-Access Networks

UMTS Macro BS (3 Sectors) Maximum cell range Minimum cell range Maximum capacity Equipment Site build out (no previous site) Site installation Operation and maintenance Site lease Last mile transmission, leased lines6 Radio Transmission Site Total discounted cost, incumbent GSM operator (10 years) Total discounted cost, no previous site (10 years) 1000 m 250 m 6250 kbps 110 000 70 000 30 000 3 000 10 000

UMTS Micro BS (1 Sector) 250 m 100 m 1250 kbps 20 000 0 15 000 1 000 3 000

UMTS Pico BS (1 Sector) 100 m 25 m 1750 kbps 5 000 0 10 000 1 000 3 000

WLAN 802.11g hotspot 25 m 10 m 56000 kbps 500 500 1 500 300 500

GSM EDGE BS (macro) 35000 m 1000 m 500 kbps 10 000 70 000 0 1 000 3 000

5 000 130 277 28 203 167 590

5 000 26 759 28 203 35 277

5 000 11 759 28 203 30 277

6 000 2 528 33 843 5 380

5 000 16 759 28 203 90 277

256 070

90 239

70 239

41 250

65 239

326 070

90 239

70 239

41 750

135 239

Table 17: Base station performance and cost. Source: [30], [13]

Includes 5% year-on-year price erosion

62

Multi-Access Networks

10,00 kbps

1 000 000 100 000 10 000 1 000 100 10 1 10 100 1 000 Subscribers / km 2 Macro only

Cost per subscriber

Cost per subscriber (NPV)

Additional micro BS Additional pico BS

10 000

100 000

Figure 19: UMTS Macro, micro and pico-cells. Source: own

10,00 kbps

1 000 000 100 000 Cost per subscriber (NPV) 10 000 1 000 100 10 1 10 100 1 000 Subscribers / km 2 Macro only

Cost per subscriber

Additional micro BS Additional WLAN AP

10 000

100 000

Figure 20: UMTS Macro, micro and WLAN 802.11g access points. Source: own

63

Multi-Access Networks

10,00 kbps

49 000

Cost per subscriber (NPV)

39 000

29 000

Additional cost per subscriber

19 000

9 000

- 1 000 10 100 1 000 Subscribers / km 2 10 000 100 000

Figure 21: Additional costs occurred by replacing UMTS pico cells with WLAN 802.11g (observe the change of scale compared to the previous figures). Source: own

10,00 kbps

1 000 000 100 000 10 000 1 000 100 10 1 10 100 1 000 Subscribers / km 2 Additional micro BS point for UMTS (271 subscribers /km2) GSM-EDGE Macro only Additional micro BS (71 subscribers/km2)

Cost per subscriber


Additional UMTS-pico cells

Cost per subscriber (NPV)

10 000

100 000

Figure 22: GSM-EDGE, UMTS micro-cells and UMTS-pico cells. Compare with Figure 19. Source: own

64

Multi-Access Networks

1,00 kbps

20 000 18 000 Cost per subscriber (NPV) 16 000 14 000 12 000 10 000 8 000 6 000 4 000 2 000 0 10 100 1 000 Subscribers / km 2 10 000 100 000 capacity-limited UMTSl macro BS

Cost per subscriber

Figure 23: Cost per subscriber at a 1kbps data rate with UMTS macro cells. Source: own

1,00 kbps

20 000 18 000 Cost per subscriber (NPV) 16 000 14 000 12 000 10 000 8 000 6 000 4 000 2 000 0 10 100 1 000 Subscribers / km 2 10 000 100 000 capacity-limited GSM EDGE BS

Cost per subscriber

Figure 24: Cost per subscriber at 1kbps data rate with GSM EDGE macro cells. Source: own

65

Multi-Access Networks

8.3

National and multi-access roaming

Roaming is the ability to use the mobile station (MS) or the user equipment (UE) in other GSM or WCDMA networks other than the home network of the subscriber. It was a key requirement for the success of GSM networks and it is widely believed that it will continue to be so for UMTS. For traditional circuit-switched services such as voice, there is a profound support for roaming but for packet-switched services, this is not the case [26]. With the introduction of new services, based on for example IMS, operators must extend roaming agreements, making roaming an even more complex issue. Roaming is interesting in shared and multi-access networks because of a number of reasons. It is a requirement for building geographically shared networks and is closely related to the case for a green field WCDMA network operator that wishes to provide fallback into another operators GSM network [79]. The scenarios here are based upon two operators sharing a network but they could all be extended to allow for more operators depending on the features provided by the equipment manufacturers.

8.3.1

Technical aspects for service provisioning

There are three blocks of functionality required to support roaming: mobility management, authentication and billing procedures. The D and Gr interfaces are open and interconnected between operators that have roaming agreements, allowing an operator to download the subscription from the correct HLR [8].
HSGSN HGGSN

Inter operator GPRS backbone

Internet

UE

VSGSN

VGGSN

Figure 25: Two ways of performing GPRS roaming, via HGGSN (dotted line) and VGGSN (sold line). Roaming connections can de divided into circuit-switched and packet-switched (GPRS) connections. For GPRS there are two roaming scenarios possible as shown in Figure 25 [26]. In the first scenario all GPRS traffic is routed via the subscribers home networks GGSN which enables the operator to provide the same portals and identical access as when the subscriber is in the home network. The second scenario is to route the subscriber via the visited networks GGSN. The advantage of this solution is that there is no need for a backbone connecting the GPRS traffic to the home network, which decreases lag times.

66

Multi-Access Networks

Presently, operators sign roaming agreements that establish the commercial terms for the roaming. For international roaming this is a tedious solution since it requires each operator to have individual agreements with all other operators to which it wishes to enable roaming with. As the mobile networks continue the evolution path towards all-IP, there will be more issues that complicate the roaming agreements. An alternative solution to this problem is to use a GPRS Roaming Exchange (GRX) network [80]. A GRX entity signs agreement with individual network operators and new operators will then only need to sign an agreement with the GRX. Since there may be multiple GRX inter-GRX agreements are also necessary. In shared networks the interesting aspects of roaming is how the same services can be provisioned throughout the whole network, irrespective if roaming is done to accommodate a green field operator or enabling geographical split sharing. For both cases, a technology that facilitates the service provisioning is necessary. If this functions correctly the user will then usually not know which operators network he/she is currently connected to. The GRX solution is useful for international roaming but for national roaming in geographical split networks individual agreements is more appropriate since the networks involved are to function as one and the agreement spans a longer time. When travelling abroad the subscriber will typically accept that it is noticeable that the network has changed but for national roaming this is a problem. Since the UE displays the name of the core network operator that it is registered with the operators will need to name their core networks in such a way that this is not noticeable to the subscriber. The operators will require CAMEL (Section 8.4.1) or similar functionality for providing seamless IN functionality.

8.4

Facilitating Service Roaming with the Virtual Home Environment (VHE)

A virtual home environment allows the same set of personalised features and services to be presented to a subscriber regardless of the network or terminal utilized [62]. The virtual home environment contains a user profile which stores the services that the users has access to and personalized settings. This is functionality is important to operators that plan to use multi-access networks, geographical split networks or otherwise use advanced roaming functionality as a differentiation opportunity. For a seamless VHE all networks between which the user roams must have the necessary intelligence and flexibility to implement the VHE. The visited network must support user access to the home environment and provide a transparent communication between clients, servers, terminals and PLMNs. The networks must also exchange user and billing data such as CDRs. Operators may be reluctant to exchange such data but in geographical split networks this data will be exchanged regardless of if a VHE is implemented. The same is valid for greenfield operators that have roaming into another network. It is therefore interesting to study how to implement these in a shared UMTS network. The VHE is not a service in itself but a platform for offering services across

67

Multi-Access Networks

different networks. Depending on the service, VHE interacts with all service capabilities discussed in Section 6.2.3. Especially important is OSA, which needs special service capabilities for facilitating the VHE, although none of these affect the basic structure of how OSA is implemented in a network [61]. Customized Applications for Mobile network Enhanced Logic (CAMEL) is a part of the intelligent network features in an UMTS network and VHE relies heavily on its functionalities for providing services. It is therefore necessary to study how a shared network affects a CAMEL implementation.

8.4.1

Customized Applications for Mobile network Enhanced Logic (CAMEL)

CAMEL is not a service in itself, but a platform for offering services. It resembles the IN (Intelligent Network) that exists today in mobile networks but provides mechanisms that allows operator specific services (OSS) to be consistently and independently served regardless of the serving network. That is, operator specific services are independent from the HPLMN [45]. Therefore, CAMEL is referred to as evolved IN. CAMEL will also evolve with the network and in the later releases of 3GPP receive a wider role than current IN services [2]. Figure 26 shows the nodes and interfaces of CAMEL in the CN. All CN nodes require updates for CAMEL support. The new nodes are gsmSSF (GSM Service Switching Function), gsmSCF (GSM Service Control Function) and gsmSRF (GSM Specialized Resource Function). CAMEL subscribers have subscription data specific to CAMEL stored in the HLR. The name for this data is CSI (CAMEL Subscription Information) and the users home network (HPLMN) provisions it when the user is in another operators network (VPLMN). The CSI requested by the VPLMN is sent to the corresponding VLR and stored in the VPLMNs VLR. When a VPLMN requests the user subscription data from the HPLMN the GMSC receives a CSI. If the VPLMN supports CAMEL it will invoke the CAMEL service logic in the VMSC for the session. The GMSC also informs the gsmSSF, allowing it to maintain state information and be notified about events in the session. The exact states that trigger events that can be supervised by the gsmSSF are named detection points. A complete list of detection points is available in [45]. When the CSI is downloaded into the VLR it informs the MSC which in turn invokes the CAMEL functionality in the MSC. The MSC informs the gsmSSF and on requests sends state and event information as in the GMSC case. This allows for a flexible structure where applications can control the state of ongoing calls and data sessions via the triggers set by the gsmSSF. One such service that has been proposed using only CAMEL nodes is real-time charging [27]. Since CAMEL requires features in the core network nodes it is affected by Level 5 sharing agreements such as common shared networks. All operators, regardless of

68

Multi-Access Networks

sharing levels will need to upgrade their nodes in the unshared domain (GMSC, HLR and GGSN).
gsm SC F U nshared

C AP

CAP M AP

gsm SS F Unshared
CAP

gsm SSF U nshared HLR U nshared


M AP

V LR Level 5

M AP

M SC Level 5

G M SC U nshared

gsm SR F U nshared

Figure 26: CAMEL nodes and interfaces in UMTS [45].

8.4.2

Inter system handovers between GSM and UMTS

The UTRAN is designed to allow for inter system handovers between GSM and UMTS [58]. Early GSM BSSs in are not designed to interact with a WCDMA RNC. WCDMA RNCs on the other hand are designed to interact with GSM BSSs. The UTRAN relies on measurements from the terminal to measure the surrounding GSM cells. When performing these measurements the UE enters slotted mode which lowers the spreading factor and therefore leaves time available for the measurements. The UE measures the GSM Broadcast Control Channel (BCCH) and forwards it to the RNS. Although the UE performs the measurements, the handover is then evaluated by the network. The RNC will recognize the need for a handover to GERAN depending on a number of control parameters and the measurement information provided. Typical control parameters as described by 3GPP are signal level, connection quality, power level propagation delay, current traffic loading per cell, interference levels, and maintenance requests. Depending on the control parameters of the serving BSS the GERAN initiates the handover procedure by initiating a connection to the 3G MSC [58]. The exact control parameters are not defined by 3GPP and are proprietary both in the sense of the algorithms provided by the equipment manufacturer and the parameters set by the operator. Switching users from GERAN to UTRAN requires upgraded functionality in the GERAN.

Shared UMTS network between operator A and B

U E

Operator B GSM network

Figure 27: Inter system and inter operator handover.

69

Multi-Access Networks

These parameters can be set to increase or decrease the usage of the GERAN or UTRAN. For example, a greenfield operator (Operator A) has a roaming agreement with operator Bs GSM network. An operator A customer is in the GSM network and is on the edge of the shared UMTS network as shown in Figure 27. Operator B would then have to set a level for when the power of the UEs signal is low enough to request it to be switched back to the shared UMTS network. It can be favourable for one operator to keep a user in their wholly owned network instead of sharing some of the revenue from that customer with the other participating operator in the shared network. This can happen in two cases. The first case is when the customer is a UMTS subscriber that belongs to the other participating operator, who may or may not be a green field operator but in either case does not have GSM coverage. Instead of switching that subscriber into the shared network when he/she enters WCDMA coverage the operator providing fallback will have an incentive to keep the customer in the GSM network. The second case is for roaming customers that have agreements with both customers in the shared network. Instead of sharing that revenue it would be more advantageous for one operator to have the roaming customer in their wholly owned network.

70

Conclusions

9
9.1

CONCLUSIONS
How do shared networks affect the differentiation opportunities?

This study has analysed four shared network agreements implications on a number of differentiation opportunities. The most basic sharing agreement studied is site sharing. Since it is usually only used in parts of the mobile network and operators provide their own equipment to each site, a pure site sharing agreement does not affect any differentiation opportunities. All other sharing agreements have some effect on each of the differentiation opportunities found in this study. How important the different opportunities are will depend on what services the operators plan to provide. On a number of opportunities, the operators must work in close cooperation and understanding of each others goals to avoid losing market share of their total joint venture. Something that can be harmful to competition since it forces the operators to disclose their respective plans. For example, in all agreements except site sharing it is not possible for one operator to deliver location-based services without an agreement with the other participating operator. These problems may cause operators to delay the provisioning of new innovative services that require changes on the network level. If the operators cannot come to terms on how to deliver such services, their joint venture will suffer. Fortunately, standards such as OSA and IMS paint a bright picture on the future for the possibility of operators to deliver unique services. If the operators can agree on implementing for example OSA, it is possible to build services that interact with the core network but exist in the unshared domain. Even if it possible to implement the service logic in the unshared domain, some services are depend on features in the underlying network (e.g. positioning). These

71

Conclusions

services will then suffer from the same problems as any other service implemented in a shared core network. Other factors that can hinder the service provisioning are capacity and latency. If the operators have different views on what capabilities are important in the network, their range of services will suffer. For example, if one operator wishes to provide services that require high capacity bearer, such as a mobile internet access service, but the other operator does not, they must find an agreement where it is possible for one operator to force capacity upgrades as necessary without affecting the other operator. For example, Push-Over-Cellular and a mobile broadband access service are mutually exclusive since PoC is optimized for low latency but capacity is not an issue while for a mobile broadband service, high capacity is a necessity while latency is less important. The operators will also have to agree on the price for usage of the resources. Services that require a high capacity or a low latency will consume a high amount of the resources in the network. It is therefore a strong effect on the operators abilities to deliver services dependent on those network features. As shown in Chapter 8 WLAN is a possible to use for capacity differentiation but remains most economical at specific hot spots. UMTS remains the best option for providing a high capacity over a large area. It is possible to achieve handover between WLAN and UMTS either on the network level (WLAN-interworking) or on the application level. The latter option is more interesting to operators since it does not require any special configuration to the underlying network. Handover between GSM/EDGE and UMTS is in use today and it is therefore possibly for incumbent operators differentiate by coverage with a capacity somewhat lower than UMTS. All sharing agreements except for site sharing affect the terms of both inbound and outbound roaming and how operators divide in-bound roaming revenues. This study shows that quality of service differentiation is a possibility for users from the different operators in the sharing agreement is a possibility. This can for example mean that certain users always get a guaranteed bandwidth.

9.2

Implications of Shared Networks on Operators Business

As operators are forced into closer cooperation, operators will cooperate on what products and services to deliver. The support from manufacturers for network sharing will continue to increase, thereby creating an excellent opportunity for virtual operators to get access to the network on the same level as the incumbents. Regulators will also be keen on this idea since it increases competition among operators. It might be difficult for network operators to stop this evolution as virtual operators can show to regulators that going from a shared network into a network that supports multiple virtual operators at the same level as the incumbents is a small step.

72

Conclusions

This leaves the incumbents in a hard place. They can either focus on going the same way as virtual operators or try to become a pure bit-pipe operator. If an operator chose to differentiate themselves by the same way as virtual operators, the operator would stop focusing on differentiation on the network level and instead just try to have a network that is good enough for the lions-share of their customers. They would then only focus on producing products and services. If they chose to become a bit-pipe operator they will then instead concentrate on making it easy and worthwhile for virtual operators to use their network, thereby having as many virtual operators as possible in their network. In Sweden for example, there are as of yet no UMTS virtual operators. As operators upgrade their equipment to support network sharing it would be no stretch for one operator to focus on becoming a bit-pipe if their current strategy of trying to attract enough customers to their UMTS network does not work. For this strategy to succeed, the service-offerings that the virtual operators in such a network can resell must be on par with those provided by the incumbents.

9.3

Summary of Key Findings

Key findings: Multi-access networks is an important differentiation opportunity, especially for incumbent GSM operators who can upgrade to EDGE IMS, OSA and VHE are especially important in shared networks since it allows operators to deploy new services in the unshared domain, but since many new services rely on features in the shared domain the advantage of developing services in the unshared domain will be mitigated. International roaming remains a problem until 3GPP Release 6 is implemented in operators networks. If the operators cannot divide their roaming revenue in such a way that incentives remains for the operators to acquire roaming customers the joint-venture will suffer. Delivering unique products and services will be difficult. Operators will be forced to cooperate on many products and services.

9.4

Future Work

Much of the findings in this study are related to the legacy inherent nature of a oneto-one mapping between network operator and service provider. Therefore, as the market continues to evolve towards a clearer distinction between service providers and network operators, more research is necessary to see what requirements such a breakdown would require of the UMTS network. Some of this work is already completed in 3GPP Release 6 where there will be a clear distinction between the operators providing a core network and the operator(s) running the radio access network.

73

Conclusions

This study also shows how quality of service differentiation can be used to differentiate between operators in a shared network. This area should be studied further, for different services and see if any other service offerings are mutually exclusive. One such problem area is how to configure the network to handle handovers between different operators appropriately. Operators must be able to individually configure shared base stations such that they perform handover to competing GSM or UMTS networks at appropriate levels of quality of service. More techniques used for multi-access differentiation should also be analysed, for example WiMAX. This study has focused on services available today. Future work should also include any upcoming new services in the analysis.

74

Bibliography

10
10.1
[1] [2]

BIBLIOGRAPHY
Books and reports
Andersen V. and Gamdrup P. (1994) Forskningsmetoder, in Andersen H. (ed.) Vetenskapsteori och metodlra: en introduktion, Lund, Sweden: Studentlitteratur. Kaaranen H., Ahtiainen A., Litinen L., Naghian S. and Niemi V. (2001) UMTS Networks: Architecture, Mobility and Services, West Sussex, United Kingdom: John Wiley & Sons, Ltd. Lundahl U. and Skrvad P-H. (1999) Utredningsmetodik fr samhllsvetare och ekonomer, 3rd edition, Lund, Sweden: Studentlitteratur. Yin, Robert K. (1994) Case Study Research: Design and Methods, 2nd edition, California, USA, Sage Publications Heicker R. (2003) 3G: Strategier och drivkrafter, Lund, Sweden: Studentlitteratur. Toskala A. (2001) Background and Standardisation of WCDMA, in Holma H. and Toskala A. WCDMA for UMTS, West Sussex, United Kingdom: John Wiley & Sons, Ltd. Holma H, Toskala A., Lappalainen U. (2001) Introduction, in Holma H. and Toskala A. WCDMA for UMTS, West Sussex, United Kingdom: John Wiley & Sons, Ltd. Holma H., Zhi-Chun H., Hmlinen S., Laiho J., Sipil K., Wacker A. (2001) Radio Network Planning, in Holma H. and Toskala A. WCDMA for UMTS, West Sussex, United Kingdom: John Wiley & Sons, Ltd. Introduction to UMTS (1999) Stockholm, Sweden: APIS Training & Seminars. UMTS SIP Signaling and Voice over IP/ATM (2004): APIS Training & Seminars. The 3G Business Case: bid book versus reality (2001): Northstream Research and Publications. 3G rollout status (2002), version 1.2, Northstream AB. Varoutas D., Sphicopoulos T., Loizillon F., Harno J., Kalhagen K-O., Stordahl K, Torre Olsen B., Welling I. (2002): IST-2000-25172 TONIC

[3] [4] [5] [6]

[7]

[8]

[9] [10] [11] [12] [13]

75

Bibliography

[14]

[15] [16]

[17] [18]

[19]

Deliverable number 9: Final results on the economic viability of 3G MVNOs, in Katsianis D. (ed.), University of Athens, Nokia, Telenor and France Telecom R&D. Stordahl K., Kalhagen K-O., Elnegaard N., Lhteenoja M., Harno J., Welling I., Cadro P., Loizillion P. D. F., Katsianis D., Varoutas D., Monath T., Braune M., Duarte M de O., Castro E., Rocha J., Pego P., Santhina G., Borrego J. P., Budry L., Demierre E., Roy J., (2002): IST-2000-25172 TONIC Deliverable number 14: Recommendations and guidelines, in Olsen B. T. (ed.), Telenor, Nokia, France Telecom R&D, University of Athens, University of Aveiro, T-Systems, Nova, Swisscom. Olsson A., Freese E., Franzn A., Igeskog R., Shafiee A., Heinegrd A. (2003), IP Multimedia Subsystem Network Study, TeliaSonera. Kviselius N., Thorngren B., Valiente P., Hultell J., Markendahl J., Werding J. (2004), Novel Access Provisioning project Market Candidates and Business Models, Working paper version 0.1.0 2004-07-05, Wireless@KTH and Stockholm School of Economics. Northstream AB (2001), Network Sharing: Savings and Competitive Effects, PTS Rostam N. (2005), Differentiation strategies for operators participating in shared network agreements, Masters Thesis, Royal Institute of Technology, Stochkolm, Sweden Pindyck S. R. and Rubinfeld L. D. (2001), Microeconomics, 5th edition, New Jersey, USA, Prentice Hall

10.2
[20]

Articles and whitepapers


Ricardo M., Dias J., Carneiro G., Rueia J. (2002) Support of IP QoS over UMTS Networks, The 13th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, vol. 4, pp. 15-18. Xipeng X., Lionel M. (1999): Internet QoS: A Big Picture, IEEE Network Magazine, vol. 4, pp. 8-18. Feng R., Song J. (2002) Some QoS Issues in 3G wireless networks, Proceedings of IEEE TENCON, pp. 724-727. Daoud F. and Mohan S. (2002) Strategies for provisioning and operating VHE services in multi-access networks, IEEE Commun. Mag., vol. 40, January, pp. 7888. White paper Shared networks: An operator alternative to reduce initial investments, increase coverage and reduce Time To Market for WCDMA by sharing infrastructure, Ericsson AB. Yang J.and Kriaras I. (2000), Migration to all-IP based networks, 3G Mobile Communication Technologies Conference Publication No. 471, Portugal, pp. 1909-1913. Roos A., Hartman M., Dutnall S. (2003), Critical Issues for Roaming in 3G, IEEE Wireless Communications, February, pp. 29-35. IlGu Jung, GilHaeng Lee (2003), Studies on Services with CAMEL features in 3G, Personal, Indoor and Mobile Radio Communications, 2003. PIMRC 2003. 14th IEEE Proceedings on, vol. 3, September, pp. 2358-2362.

[21] [22] [23]

[24]

[25]

[26] [27]

76

Bibliography

[28]

[29]

[30]

[31] [32] [33]

[34] [35]

[36] [37]

[38]

[39] [40] [41] [42] [43]

Village J. A., Worrall K. P., Crawford D. I. (2002), 3G Shared Infrastructure, 3G Mobile Communication Technologies Conference Publication No. 489, May, pp. 10-16. Eriksson M., Furuskr A., Kronestedt F., Lindheimer C., Mazur S., Moln J., Tidestav C. (2000), System overview and performance evaluation of GERAN The GSM/EDGE Radio Access Network, Wireless Communications and Networking Conference, vol. 2, September, pp. 902-906. Johansson K., Furskr A., Karlsson P., Zander J. (2004): Relation Between Base Station Characteristics and Cost Structure in Cellular Systems,15th IEEE International Symposium on Pesonal Indoor and Mobile Communcations Leivo J. (2002), 3G Network Sharing: Managing Cost of Ownership in 3G Network Sharing, COMDEX Nordic. Whitepaper: How do UMTS, UICC, USIM and USAT-(I) fit together?, IM Technologies. Tamura T., Takahashi T., Morita T., Ohtaki K., Takeda H. (2003), IMT2000 Core Network Node Systems, IEEE Wireless Communications, vol.10, February, pp. 15-21. 3G Network Infrastructure Sharing in EU: Status in some EU countries at 15 October 2001 (2001), TIA Europe. Leroy S., Bos L., De Vriendt J. (2002), End-to-end UMTS quality of service architecture for the support of real-time IP multimedia services in UMTS R5, 3G Mobile Communication Technologies, Publication No. 489, May, pp. 234238. Telephonys complete guide to WiMAX: The business case service provider deployment (2004), Wimax Forum. Sbire B., Bysted T., Pedersen K. (2003) IP Multimedia Services Improvements in the GSM/EGE Radio Access Network, The 57th IEEE Semi annual Vehicular Technology Conference, Finland, vol.1 , pp. 403-407. Skorin-Kapov, L., Huljenic, D., Mikic, D., Vilendecic, D. (2004), Analysis of end-to-end QoS for networked virtual reality services in UMTS, Communications Magazine, IEEE, vol. 42, April, pp. 49-55. Sidenbladh T. (2002), License and regulatory update on UMTS, UMTS Forum. Whitepaper: Wideband Speech Coding Standards and its Applications, Voiceage (2004) Laukkanen J. (2000), UMTS Quality of Service Concept and Architecture, University of Helsinki (Presentation) Curley D. (2001), Part IV: Mobile Virtual Network Operators, Presentation at 3G Technologies and Applications, Bonn, Germany Berghoff G (2002): Introduction to QoS and Differentiated Services Applications to Nokias IP RAN, Nokia Networks R&D

77

Bibliography

10.3
[44]

Technical specifications
3GPP (2003): TR 22.934 v6.2.0 Technical Specification Group Services and System Aspects; Feasibility study on 3GPP system to Wireless Local Area Network (WLAN) interworking (Release 6). 3GPP (2004): TR 22.078 v7.1.0 Technical Specification Group Services and System Aspects; Customised Applications for Mobile network Enhanced Logic (CAMEL; Service Description; Stage 1 (Release 7). 3GPP (2004): TR 22.071 v7.0.0 Technical Specification Group Services and System Aspects; Location Services (LCS); Service description; Stage 1 (Release 7). 3GPP (2004): TR 25.305 v6.1.0 Technical Specification Group Radio Access Network; Stage 2 functional specification of User Equipment (UE) positioning in UTRAN (Release 6). 3GPP (2004): TR 22.038 v7.1.0 Technical Specification Group Services and System Aspects; USIM Application Toolkit (USAT); Service description; Stage 1 (Release 7). 3GPP (2002): TS 23.236 Technical Specification Group Services and System Aspects; Intra-domain connection of Radio Access Network (RAN) nodes to multiple Core Network (CN) nodes (Release 5). 3GPP (2004): TS 23.207 Technical Specification Group Services and System Aspects; End-to-end Quality of Service (QoS) concept and architecture (Release 6). 3GPP (2004): TS 26.103 Technical Group Services and System Aspects; Speech codec list for GSM and UMTS (Release 6). 3GPP (2002): TS 26.071 Technical Specification Group Services and System Aspects; Mandatory speech CODEC speech processing functions; AMR speech CODEC; General description (Release 5). 3GPP (2002): TS 26.171 Technical Specification Group Services and System Aspects; Speech Codec speech processing functions; AMR Wideband Speech Codec; General Description (Release 5). WAP Forum (2001): Wireless Application Protocol Architecture Specification 3GPP (2002): TS 22.057 Mobile Execution Environment (MExE), Service Description Stage 1 (Release 7) 3GPP (2002): TS 22.951 Technical Specification Group Services and System Aspects; Technical Specification Group Services and System Aspects; Service aspects and requirements for network sharing; (Release 6). 3GPP (2004): TS 28.062 Technical Specification Group Services and System Aspects; Inband Tandem Free Operation (TFO) of speech codecs; Service description; Stage 3 (Release 6) 3GPP (2004): TS 22.129 Technical Specification Group Services and System Aspects Service aspects; Handover Requirements between UTRAN and GERAN or other Radio Systems (Release 6) 3GPP (2004): TS 23.009 Technical Specification Group Core Network; Handover procedures (Release 5). 3GPP (2004): 23.107 Technical Specification Group Services and System Aspects; Quality of Service (QoS) concept and architecture (Release 6). 3GPP (2004): 23.127 Technical Specification Group Services and System Aspects; Virtual Home Environment (VHE) / Open Service Access (OSA) (Release 6)

[45]

[46] [47]

[48] [49]

[50] [51] [52]

[53]

[54] [55] [56]

[57]

[58]

[59] [60] [61]

78

Bibliography

[62] [63] [64]

[65] [66]

3GPP (2004): 22.121 Technical Specification Group Services and System Aspects; Service aspects; The Virtual Home Environment (Release 5) 3GPP (2004): 25.308 Technical Specification Group Radio Access Network; High Speed Downlink Packet Access (HSDPA); Overall description; Stage 2 (Release 6) 3GPP (2004): 23.979 Technical Specification Group Services and System Aspects; 3GPP enablers for Open Mobile Alliance (OMA) Push-to-talk over Cellular (PoC) services; Stage 2 (Release 6) 3GPP (2004): 23.251 Technical Specification Group Services and System Aspects; Network Sharing; Architecture and functional description; (Release 6) 3GPP (2004): 23.851 Technical Specification Group Services and System Aspects; Network sharing; Architecture and Functional Description; (Release 6)

10.4
[67] [68] [69] [70] [71] [72] [73] [74]

Interviews
Nilsson U. (2004), TeliaSonera, Networks and Production Mobile Networks, interviewed by telephone on 20 September 2004. Nilsson U. (2004), TeliaSonera Networks and Production Mobile Networks, interviewed in person on 10 October 2004. Norberg p. (2004), TeliaSonera, Networks and Productions Division, interviewed by telephone on 23 September 2004. Eijvergrd H. (2004), TeliaSonera, interviewed in person and by telephone on 13 October 2004 and 11 January 2005 respectively. Axelsson M. (2004), member of the board of directors, The Swedish Post and Telecom Agency, interviewed in person on 8 November 2004. Bergljung C. (2004), Manager Mobile Networks Research and Development, interviewed by telephone 24 September 2004. Ragnevad J.(2004): Strategic Advisor, Northstream, interviewed in person on 3 December 2004 Kristensson M. (2004), Solution Manager End-to-End Service Performance, Nokia Networks, interviewed in person and through e-mail on 13 December 2004 and 16 December 2004 respectively. Laakso P. (2004), Senior Marketing Manager, Nokia Networks, interviewed in person and through e-mail on 13 December 2004 and 16 December 2004 respectively. Mkitalo . (2004), Senior Vice President, Mobile Business, TeliaSonera, interviewed in person 14 December 2004.

[75]

[76]

10.5
[77] [78]

Internet Resources
3GPP Specification Series 22, http://www.3gpp.org/ftp/Specs/htmlinfo/22-series.htm, available November 2004 3GPP Specification Series 23, http://www.3gpp.org/ftp/Specs/htmlinfo/23-series.htm, available January 2005

79

Bibliography

[79]

[80]

[81] [82] [83] [84] [85] [86] [87] [88]

[89] [90]

GSM World, GSM Roaming: http://www.gsmworld.com/roaming/index.shtml, available November 2004 GSM World, GPRS Roaming: http://www.gsmworld.com/technology/gprs/grx_prov.shtml, available November 2004 http://www.ambient-networks.org/ Tonic project http://www-nrc.nokia.com/tonic/ UMTS Forum, http://www.umts-forum.org WAP Forum, http://www.wapforum.org Total Telecom, http://www.totaltele.com/ Rosenberg J. et al (2002), RFC 3261: SIP Session Initiation Protocol, http://www.ietf.org/rfc/rfc3261.txt, available December 2004 http://www.cs.columbia.edu/sip/, available December 2004 http://www.startsidan.telia.se/com/telia/ics/portal/apps/press/PressRele asePage.html?id=102622&disclaimerId=&referer=PressIngressPage.html&l ang=EN&year=2004, Pressrelease TeliaSonera, available January 2005 The Swedish National Post and Telecom Agency (PTS), http://www.pts.se Swedish Statistics, http://www.scb.se

80

APPENDIX A
1G

ACRONYMS

2G

2.5G

2.75G

3G

3GPP

AAA API BCCH BSS CAMEL CAPEX

CDMA CDR CN CRNC CS CSCF EDGE ESP GERAN GGSN GMLC

First generation mobile telecommunications technology. The first analogue standards for mobile communication. The two most wide-spread systems are NMT, used in Europe, and AMPS, used in the United States. Second generation mobile telecommunications technology. Umbrella term for the digital successors of 1G. The most widely used 2G standard is GSM. Second and a half generation mobile telecommunications technology. 2.5G is a stepping-stone between 2G and 3G, and is often placed on par with GPRS. 2.75G is the term used for systems that do not meet the 3G requirements, but are marketed as if they do, or vice versa. The term is often placed on par with EDGE. Third generation mobile telecommunications technology. Services associated with 3G offer the possibility to transfer voice and data on high transmission rates. Operators that use GSM generally prefer the UMTS standard (based on WCDMA) for their 3G networks. UMTS is therefore concentrated to Europe. 3rd Generation Partnership Project. A collaboration agreement between different standards bodies to produce globally applicable technical specifications and technical reports for a 3rd generation mobile system. Authentication, Authorization and Accounting Application Programming Interface GSM Broadcast Control Channel Base Station System Customised Application for Mobile network Enhanced Logic Capital expenditure. Capex are expenditures used by a company to acquire of upgrade physical assets such as equipment, property, or buildings. Code Division Multiple Access Charging Data Record Core Network Controlling RNC Circuit-Switched Call Server Control Function Enhanced Data rates for GSM Evolution. EDGE is a technology that enhances data rates for 2G or 2.5G networks. Enhanced Service Provider. See Service Provider (SP). GSM-EDGE Terrestrial RAN. Gateway GPRS Support Node. The GGSN maintains the connections towards other packet-switched networks such as the internet. Gateway MLC. See MLC

81

GMSC

GPS GSM GPRS

GRX HLR HSCSD HSDPA IP IMS IN LBS LMU MExE MGW MLC MNC MNO MS MSC MVNO

NDC

NMT O&M OSA RNS SAS SDU SIP SLA OPEX OTDOA PLMN

Gateway MSC. The GMSC is the switch at the point where the UMTS network is connected to external circuit-switched (CS) networks. All incoming and outgoing CS connections are routed through the GMSC. Global Positioning System. Global System for Mobile communications (originally Groupe Speciale Mobile). The most common 2G standard, used mostly in Europe. General Packet Radio Service. GPRS allows packet-switched services such as internet access, with lower resource usage than in circuitswitched connections. GPRS Roaming Exchange Home Location Register. The HLR is a database that contains subscriber information. High-speed circuit-switched data. An extension to GSM that allows high speed data in the circuit-switched domain. High Speed Downlink Packet Access Internet Protocol. A routing protocol used in e.g., the internet. IP Multimedia Subsystem Intelligent Network Location Based Services Location Measurement Unit Mobile Station Application Execution Environment Media Gateway Mobile Location Centre Mobile Network Code Mobile Network Operator Mobile Station. A GSM terminal with a SIM card. Mobile services Switching Centre. The MSC controls the BSSs and interconnects calls to e.g., PSTN. Mobile Virtual Network Operator. MVNOs are operators that do not own any telecommunication license. They work as resellers, buying capacity from a network operator and reselling it to end-customers. National Destination Code. The NDC consists of 1-3 digits specifying the destination within a given region. A telephone or fax number consists of a country code, an NDC, and a subscriber number. Nordic Mobile Telephony. A first generation mobile telecommunications network, used mainly in the Nordic countries. Operations and Maintenance Open Service Access Radio Network Subsystem Stand-alone SMLC Service Data Unit Session Initiation Protocol Service Level Agreement Operational expenditure. Observed Time Difference of Arrival. A method for positioning an UE relative to a site. Public Land Mobile Network. The mobile network, containing MSC service areas.

82

PoC PS PSTN PTS PTT QoS RAN SCF SCS SP SRNC RAN RNC SGSN SMS UE UMTS

USAT USIM UTRAN VAS VHE VLR WAP WLAN

WLAN-I WCDMA

Push-over-Cellular. A walkie-talkie like feature. Packet-Switched. Public Switched Telephone Network. PSTN is a term for a fixed telephony network. Post och Telestyrelsen. The Swedish National Post and Telecom Agency. Push-To-Talk. See PoC. Quality of Service Radio Access Network Service Capability Feature Service Capability Server Service Provider. An entity selling services through another operators UMTS or GSM network. Serving RNC. Radio Access Network Radio Network Controller. An RNC is an entity in UMTS networks, which controls the base stations. Serving GPRS Support Node. SGSN is an element providing services similar to that of the MSC/VLR in the packet-switched domain. Short Message Service. User Equipment. A UMTS terminal with a USIM card. Universal Mobile Telecommunications System. A European telecommunications standard fulfilling the requirements of an IMT2000 3G system. USIM Application Toolkit Universal Subscriber Identity Module. A UMTS SIM card. UMTS Terrestrial RAN Value Added Services Virtual Home Environment Visitor Location Register Wireless Application Protocol. WAP is an open international standard for applications that use wireless communication. Wireless Local Area Network. An umbrella term for a number of unlicensed wireless access technologies. WLAN provides high capacity over a small area, typically less than a radius of 40 metres from an access point. WLAN-Interworking. To interconnect WLAN access point with UMTS base stations (Node Bs). Wideband Code Division Multiple Access.

83

APPENDIX B

OVERVIEW OF ACCESS TECHNOLOGIES

There is a multitude of access technologies available for both operators and home users. Some are geared towards high speed but cover a small area and others cover a larger area with a loss of speed. To make any meaningful comparison between the different technologies an understanding of how these access technologies works and what their intended uses are is required. While some of these technologies are examined in greater detail in the report, others are not. This chapter aims to illustrate current and future technologies and give an understanding as to why some of these have been selected for further research. It is also intended as a quick reference when reading the report.

B. Telecom standards
Mobile telecommunication standards have a long history from the first analogue cellular systems such as TACS in Europe or AMPS in the US. This comparison is focused on growing or emerging technologies and these 1G systems are therefore left out. From these standards there has been an evolution to the 3G standards available today. The different standards bodies over the world have tried to unify the 3G standards but have not succeeded. Although the standards differ on some levels, this work has resulted in keeping the architectural principles similar. The two 3G standards available today is UMTS in Europe and IMT-2000 in Japan and the US. A thorough explanation of the evolution from 2G to 3G standards is available in [6]. The comparison given here should be regarded as a quick overview and not a complete comparison. A profound comparison of all standards is difficult given that there are a large array of parameters to consider such as cost of deployment and coverage area et cetera. Even comparing the speeds as shown below is not entirely fair. The speed shown is the theoretical speed as given by the manufacturers. Some technologies will deal better with interference as the user moves while being connected to the network. All speeds are given as the total capacity of the system unless stated otherwise. If two users are connected their respective speed will be half of the theoretical maximum. If three users are connected their respective speed will be a third of the theoretical maximum and so on.

85

Generation7

System

2G

D-AMPS

Main geographical area USA

Speed (downlink) N/A

Evolution to

2G

GSM

Global

2G

GSM with GPRS8 IS-95 (cdmaOne) EDGE Phase1 (E-GPRS) EDGE Phase2 (HSCSD)9 WCDMA WCDMA with HDSPA CDMA2000 1x CDMA2000 1xEV-DO CDMA2000 1xEV-DV

Global

2G 2.5G 2.5G 3G 3G 3G 3G 3G

USA Global Global Global Global USA USA USA

9.6 Kbps or 14.4 Kbps (per user) 9.6 Kbps or 14.4 Kbps (per user) 64 Kbps 384 Kbps 384 Kbps 750 Kbps 1.5 Mbps 144 Kbps 2 Mbps > 2 Mbps

Being phased out in favour of GSM GPRS

EDGE Phase1 CDMA2000 1x EDGE Phase2 No clear upgrade path HSDPA No clear upgrade path CDMA2000 1xEV-DO CDMA2000 1xEV-DV No clear upgrade path

Table 18: Comparison matrix of mobile telecommunication standards. Source: own research

B. Unlicensed Access Technologies


WLAN is an umbrella term for a number of access technologies that all operate in the unlicensed ISM spectrum. This means that any individual or entity can deploy them without requiring a license.
Name 802.11b 802.11a 802.11g Hiperlan/1 Hiperlan/2 Bluetooth Theoretical Maximum Bandwidth 1-11 Mbps 6-54 Mbps 6-54 Mbps 2-20 Mbps 6-54 Mbps 720 kbps Multiplexing technique DSSS OFDM OFDM/CCK CSMA/CA OFDM FHSS Frequency 2.4 GHz 5 GHz 2.4 GHz 2.4 GHz 5 GHz 2.4 GHz

The concept of different generations is a somewhat undefined concept. The generations listed here are the commonly accepted definitions. 8 While speeds remain the same, GPRS introduces the packet switched domain in addition to the circuit switched domain. 9 EDGE Phase2 provides the same speeds as in EDGE Phase1 in the circuit switched domain.
7

86

APPENDIX C

NODES AND INTERFACES IN UMTS AND GSM

To understand how shared networks are built, it is necessary to compare them with wholly owned UMTS or GSM networks. Figure 28 shows a GSM network, equipped with GPRS. This is a typical wholly owned network by an incumbent GSM operator today. The network contains two main parts: the core network and the access network. The access network consists of the Base Station Subsystem (BSS), and mobile stations (MS). The core network consists of three subsystems: the Network Subsystem (NSS), the GPRS Packet Core, and supporting nodes such as Home Location Register (HLR), Authentication Centre (AuC), and Equipment Identity Register (EIR). The core network also contains services. These are divided into two groups: Value Added Services (VAS) and Intelligent Network (IN). Figure 28 also shows the interfaces between the nodes.

C. GSM
Mobile Station Mobile Station is the combination of a Subscriber Identity Module (SIM) and a Mobile Equipment (ME). The SIM is the small Smart Card in the mobile equipment, containing information about the users subscription. The mobile equipment is the actual handset. Without the SIM-card, it is only possible to use the mobile station for emergency calls. Base Transceiver Station The Base Transceiver Station (BTS) maintains the actual connection with the mobile station. It decodes and processes all messages sent to and received from the MS. Each BTS covers a certain geographical area ranging between 1 km and 35 km. The area covered by one BTS is called a cell. To avoid neighbouring cells from causing interference with each other, they use different frequencies. Frequency reuse is possible only if the cells are not adjacent. The BTS has a fixed capacity divided between subscribers in range of the BTS. Hence, the larger area it covers, the more subscribers will share its capacity. It is therefore necessary to vary the cell size to accommodate the capacity need in a given area. Base Station Controller The Base Station Controller (BSC) controls and supervises the BTSs in the radio network. Every BSC controls several BTSs. While the BTS takes care of the actual radio communication, the BSC is in charge of all action that is taken. The BSC tells the BTSs what to do, when to transmit, what power to use, etc. From the core network point of view, the BSC maintains the connection with the MS. Transcoding and Rate Adaptation Unit The Transcoding and Rate Adaptation Unit (TRAU) handles speech transcoding. It can convert one speech-coding format to another coding format.

87

Mobile services Switching Centre BSCs are connected to a Mobile services Switching Centre (MSC). An MSC sets up, routs, and releases calls. It controls the BSS and can interconnect calls to e.g., the Public Switched Telephony Network (PSTN) via a GMSC. When a subscriber places a call, the MSC analyses the actual digits and routes the call accordingly. The MSC also handles billing data. One MSC controls a large number of BSCs, and each operator has only a few MSCs. Visitor Location Register Each MSC is connected to a Visitor Location Register (VLR), a database that keeps information about all the subscribers currently served by the MSC. The VLR can fetch this data either from the Home Location Register (HLR) or from the Gateway MSC (GMSC) for roaming customers. Gateway MSC The Gateway MSC (GMSC) handles mobility management, communication management, and connections to other networks. When a call is routed to the GSM network, it first connects to the GMSC. The GMSC questions the HLR of where the MS is located, and routes the call to the appropriate MSC. Outbound calls from the GSM network are routed via the GMSC to other networks. The GMSC acts as a gateway for interfacing with external networks such as a PSTN and packet data networks. Home Location Register The HLR is a database that contains subscriber information. It also contains routing information that the GMSC needs in order to route calls to a subscriber. Normally, each operator has only one HLR. Other support nodes The Authentication Centre (AuC) and the Equipment Identity Register (EIR) handles security-related information. The AuC handles authentication and encryption towards the MS. The EIR is a database that contains information about the mobile equipment, i.e., the actual hardware, in the network. Through the EIR, the operator can track e.g., stolen or unapproved equipment. In short, the AuC authenticates the user while the EIR authenticates the hardware. Service platforms The Value Added Service (VAS) platform provides a certain type of service in the GSM network. The minimum VAS contains typically two pieces of equipment: Short Message Services Centre (SMSC) and Voice Mail System (VMS). From the service evolution point of view, VAS is the very first step in generating revenue with services and partially tailoring them. The Intelligent Network (IN) enables more individually tailored services, and makes many new services possible. For example, prepaid subscriptions are implemented with IN technology. Serving GPRS Support Node The Serving GPRS Support Node (SGSN) provides services similar to that of the MSC/VLR, but in the packet-switched domain. 88

Gateway GPRS Support Node The Serving GPRS Support Node (SGSN) provides services similar to that of the GMSC but in the packet-switched domain.
Um A

BSS
MS BTS BSC TRAU MSC/VLR

NSS

ISDN
GMSC

PSPDN PSTN X.25 CSPDN

HLR/AuC/EIR

V A S

I N

Gb

GPRS Packet Core

Other data network Internet

SGSN

GGSN

Figure 28: A GSM network equipped with GPRS (3GPP Release 99). Source: [2]

C. UMTS
In the context of UMTS, the term User Equipment (UE) denotes terminals or mobile stations. The Base Stations (BS) have similar functionalities as the BTSs, and the Radio Network Controllers (RNC) are comparable with the GSM BSCs. The RNCs and BSCs are interoperable, meaning that seamless handover is possible between a GSM and a UMTS network. A new radio access method, Wideband Code Division Multiple Access (WCDMA), is also introduced. In the early releases of the UMTS specifications, the core network remains similar to a GSM network with a packet-switched domain. However, the IN features evolve into Customised Applications for Mobile network Enhanced Logic (CAMEL). CAMEL makes it possible to transfer service information between different UMTS networks. This implies that subscribers can use their services when they are roaming in other networks.

89

APPENDIX D

INTERVIEW QUESTIONNARE

This appendix contains the typical set of questions that were given to interview respondents. As necessary, the respondents have been asked to clarify or expand on subjects that were especially interesting to this study. This questionnaire was mostly used in the first part of the thesis were the differentiation strategies are identified. 1. The technical solutions that operators have chosen for shared networks are as of yet only rudimentary. In what direction do you feel that the standard should move? What should the standard focus on to facilitate network sharing? 2. What differentiation strategies exist for operators today? 3. Which differentiation strategies are lost when operators enter a network sharing agreement? 4. Is it reasonable to compare MVNOs with operators that share networks? What are the similarities and differences? 5. MVNOs have a clear (technical) similarity to network operators in shared network agreements. What differentiation strategies do they successfully use? Can the network operators copy these? 6. MVNOs in Sweden have had difficulty to gain market share. Should the network owners open up for a closer cooperation with virtual operators or should they try to stop them in their tracks? 7. In Hong Kong, the regulation agency has forced a clearer split between network owner and the reseller of its services. Who gains from this? Should operators try to focus on making a split between the network owner part and the operator part (reseller of services)? 8. Some reports suggest that WLAN is a complement to UMTS while other regards it as a competitor. What is your view on this? How can operators gain the most from WLAN? 9. What are, in your opinion, operators possibilities for differentiation in shared networks? How should they differentiate themselves?

91

Вам также может понравиться