aroma ist-4-027567 d17 economic evaluation of novel aroma ... · in mbms case study, the target of...

D17: Economic evaluation of novel AROMA RRM/CRRM algorithms and solutions Page i

Copyright 2006-2007 AROMA Consortium /03/12/2007

AROMA

AROMA

AROMA IST-4-027567 D17

Economic evaluation of novel AROMA RRM/CRRM algorithms and solutions

Contractual Date of Delivery to the CEC: 30-10-2007

Actual Date of Delivery to the CEC: 03-12-2007

Editor: Robert Farotto (TI)

Author(s): See list

Participant(s): TI, TID, TEL

Workpackage: WP2

Est. person months: 7

Security: PU

Nature: R

Version: 001

Total number of pages: 68

Abstract: This deliverable describes the potential economic advantages of using specific AROMA RRM/CRRM algorithms and solutions. The economical analysis is carried out taking into account selected scenarios, also providing specific business case based on potential market demands. Keyword list: CAPEX, OPEX, Techno-economic issues, CRRM algorithms, Mobile TV, MBMS,

HSDPA

D17: Economic evaluation of novel AROMA RRM/CRRM algorithms and solutions Page ii


DISCLAIMER

The work associated with this report has been carried out in accordance with the highest technical standards and the AROMA partners have endeavoured to achieve the degree of accuracy and reliability appropriate to the work in question. However since the partners have no control over the use to which the information contained within the report is to be put by any other party, any other such party shall be deemed to satisfied itself as to the suitability and reliability of the information in relation to any particular use, purpose or application.

Under no circumstances will any of the partners, their servants, employees or agents accept any liability whatsoever arising out of any error or inaccuracy contained in this report (or any further consolidation, summary, publication or dissemination of the information contained within this report) and/or the connected work and disclaim all liability for any loss, damage, expenses, claims or infringement of third party rights.

D17: Economic evaluation of novel AROMA RRM/CRRM algorithms and solutions Page iii


DOCUMENT HISTORY

Date Version Status Comments

2007-10-02 001 Int First draft containing a proposed ToC

2007-10-31 002 Int Second draft

2007-11-13 003 Int Third draft

2007-11-28 004 Int Final version for PCC approval

2007-12-03 001 Apr Document approved and submitted to E.U.

D17: Economic evaluation of novel AROMA RRM/CRRM algorithms and solutions Page iv


Authors List

Andrea Barbaresi (TI) Massimo Barbiero (TI) Per Emanuelsson (TEL) Robert Farotto (TI) Giuseppe Minerva (TI) Tarapiah Saed (TI Consultant) Marco Tosalli (TI) Avelina Vega (TID)

D17: Economic evaluation of novel AROMA RRM/CRRM algorithms and solutions Page v


EXECUTIVE SUMMARY This deliverable reports an analysis of the potential economic impacts tied to the RRM/CRRM solutions investigated within the AROMA project. The economical analysis is carried out taking into account specific exemplary scenarios, and also making use of potential market demands. After giving some general information related to the methodology followed for the techno-economic evaluations, the document focuses on a couple of relevant study cases:

• the First one dealing with a selected CRRM algorithm based on a “fittingness factor” (it is a particular metric that helps in selecting the most suitable RAT/cell to be used in a heterogeneous scenario), whereas

• the second one is related to the mobile TV over MBMS versus HSDPA. In both the case studies, the techno-economic evaluations have been carried out by assuming a short or medium term increase of data traffic and by analyzing the potential savings offered by the addressed solutions with respect to the total investment (CAPEX+OPEX) needed to increase the capacity of a pre-existing network. Notice that the analysis has not been based on revenues because they, often, are not proportional to the load generated in the network and rely on different mechanisms (marketing based) with respect to the technical ones. More precisely, in the CRMM study case related to “fittingness factor” the savings offered by the CRRM algorithm with respect to the case when the algorithm is not present are evaluated, assuming that new investments are needed for upgrading the already existing UTRAN sites with the introduction of additional frequency carriers, in order to fulfill the assumed traffic increase for the next 5 years. In MBMS case study, the target of the techno-economic analysis is to compare the different investments needed to enhance an existing 3G network, in order to fulfill the requirements due to the introduction of massive (in the long-term) services based on TV and video, on a mobile terminal, comparing two alternatives: 1) exploitation of HSDPA connections and 2) MBMS introduction to provide broadcast/multicast. The techno-economic analysis on MBMS is completed by a short but comprehensive overview of Mobile Operators strategies and market trends of TV Mobile services in Europe. The document is organized as follows: After a brief introduction an overview of the (Common) Radio Resource Management solutions envisaged in AROMA is presented to select the most appropriate algorithms to be analyzed from a techno-economic viewpoint. Then the methodology to be followed in the analysis is presented and a complete techno-economic analysis for the two selected study cases (the CRRM “fittingness factor” and mobile-TV over MBMS) is done. Finally some conclusions are also addressed. The document also includes two annexes: one devoted to discus the envisaged analytical model of the CRMM algorithm based on the fittingness factor framework, whereas the other is devoted to provide a dimensioning model for the mobile-TV scenario.

D17: Economic evaluation of novel AROMA RRM/CRRM algorithms and solutions Page vi


Table of Contents

1 INTRODUCTION ............................................................................................................................. 2 2 AN OVERVIEW OF AROMA RRM/CRRM ALGORITHMS AND SOLUTIONS............................. 2

2.1 TABLE OF AROMA ALGORITHMS/SOLUTIONS AND INVESTIGATIONS............................................... 2 2.2 ECONOMIC IMPACT OF AROMA ALGORITHMS AND SOLUTIONS ..................................................... 7

3 CRRM AROMA ALGORITHMS AND SOLUTIONS....................................................................... 8 3.1 OVERVIEW OF AROMA CRRM ALGORITHMS .............................................................................. 8

3.1.1 Common congestion control............................................................................................ 8 3.1.2 Coverage-based CRRM for Voice Traffic........................................................................ 8 3.1.3 Fittingness factor algorithm ............................................................................................. 9 3.1.4 CRRM perceived throughput ........................................................................................... 9 3.1.5 Opportunistic CRRM........................................................................................................ 9 3.1.6 CRRM Cost Function..................................................................................................... 10 3.1.7 MPLS based mobility management and IP QoS........................................................... 10

4 METHODOLOGY FOLLOWED FOR THE TECHNO-ECONOMIC EVALUATIONS................... 11 4.1 INVESTMENTS VERSUS REVENUES VALORIZATION....................................................................... 11 4.2 DEPENDENCE OF THE RESULTS FROM THE TIME BASED TRAFFIC HYPOTHESES............................. 11 4.3 MARKET PENETRATION OF MULTI-MODE TERMINALS ................................................................... 12

5 TECHNO-ECONOMIC EVALUATION OF FITTINGNESS FACTOR CRRM ALGORITHM........ 16 5.1 INTRODUCTION ........................................................................................................................ 16 5.2 ALGORITHM DEFINITION ............................................................................................................ 16 5.3 ALGORITHM MODELING AND MARKOV CHAIN DESIGN.................................................................. 18

5.3.1 Flowchart of the CRRM algorithm ................................................................................. 18 5.3.2 Analytical model of the CRRM algorithm....................................................................... 20

5.4 SCENARIO AND QOS CONSTRAINS ............................................................................................ 21 5.4.1 CAPEX and OPEX for UTRAN Carrier upgrade ........................................................... 23

5.5 RESULTS ................................................................................................................................. 23 5.6 CONCLUSION ........................................................................................................................... 28

6 TECHNO-ECONOMIC EVALUATION OF MOBILE TV OVER MBMS........................................ 29 6.1 INTRODUCTION ........................................................................................................................ 29 6.2 DVB-H VERSUS MBMS FOR MOBILE TV............................................................................ 31 6.3 OVERVIEW AND MARKET TRENDS OF MOBILE TV SERVICE .......................................................... 33

6.3.1 Mobile TV Market: European scenario .......................................................................... 34 6.4 METHODOLOGY FOR THE TECHNO-ECONOMIC EVALUATION OF MOBILE TV OVER MBMS.............. 36 6.5 ECONOMIC IMPACTS OF MBMS ................................................................................................ 37

6.5.1 Market Assumptions for Mobile TV services ................................................................. 38 6.5.2 Service mix .................................................................................................................... 38 6.5.3 Usage ............................................................................................................................ 38 6.5.4 European users’ forecasts at national level from 2008 to 2018 .................................... 38

6.5.4.1 Total annual traffic for the whole country ................................................................................ 39 6.5.4.2 Traffic projection for an “average European town” .................................................................. 39

6.6 DIMENSIONING MODEL.............................................................................................................. 40 6.7 CAPEX AND OPEX VALORIZATION........................................................................................... 40

6.7.1 Investment related to the upgrade of HSDPA and MBMS technology.......................... 40 6.7.2 Investment related to the deployment of new sites ....................................................... 40 6.7.3 Investments related to the introduction of the second UTRAN carrier .......................... 41

6.8 RESULTS OF TECHNO-ECONOMIC EVALUATION........................................................................... 41 6.9 CONCLUSIONS ......................................................................................................................... 44

7 CONCLUSIONS ............................................................................................................................ 45 8 ANNEX A: ANALYTICAL MODEL OF THE CRRM ALGORITHM BASED ON THE FITTINGNESS FACTOR FRAMEWORK ............................................................................................. 46

8.1 ANALYTICAL MODEL OF THE CRRM ALGORITHM ........................................................................ 46 8.1.1 State characterization and Space State dimension....................................................... 46

D17: Economic evaluation of novel AROMA RRM/CRRM algorithms and solutions Page vii


8.1.2 Set of Events ................................................................................................................. 47 9 ANNEX B: DIMENSIONING MODEL FOR THE MOBILE-TV SCENARIO ................................. 59

9.1 TRAFFIC BASED DIMENSIONING MODEL ...................................................................................... 59 9.1.1 Evaluation of the number of second UTRAN carries and new nodeB .......................... 60

9.1.1.1 Uplink ...................................................................................................................................... 60 9.1.1.2 Downlink ................................................................................................................................. 61 9.1.1.3 Downlink transmission power requested for dedicated channels in downlink ......................... 62 9.1.1.4 Mobile TV on HSDPA: downlink transmission power requested for HS-DSCH....................... 63 9.1.1.5 Mobile TV on MBMS: downlink transmission power requested for S-CCPCH ........................ 64

10 REFERENCES .......................................................................................................................... 66 ACRONYMS.......................................................................................................................................... 68

D17: Economic evaluation of novel AROMA RRM/CRRM algorithms and solutions Page 1


ABSTRACT This deliverable reports an analysis of the potential techno-economic impacts tied to the RRM/CRRM solutions investigated within the AROMA project. The document, after a general introductory overview, focuses on a couple of relevant case studies, one dealing with the mobile TV over MBMS and the other one analyzing a specific CRRM algorithm based on a “fittingness factor” (it is a particular metric that helps in selecting the most suitable RAT/cell to be used in a heterogeneous scenario). The economical analysis is carried out taking into account specific exemplary scenarios, and also making use of potential market demands. In both the case studies, the techno-economic evaluations have been carried out by assuming a short or medium term increase of data traffic and by analyzing the potential savings offered by the addressed solutions with respect to the total investment (CAPEX+OPEX) needed to increase the capacity of a pre-existing network. The analysis has not been based on revenues because they, often, are not proportional to the load generated in the network and rely on different mechanisms (marketing based) with respect to the technical ones. More precisely, in the CRMM case study related to “fittingness factor” the savings offered by the CRRM algorithm with respect to the case when the algorithm is not present are evaluated, assuming that new investments are needed for upgrading the already existing UTRAN sites with the introduction of additional frequency carriers, in order to fulfill the assumed traffic increase for the next 5 years. The approach developed in order to implement and evaluate the “fittingness factor” algorithm, is based on Markov Chain theory. The fictive scenario considered is representative of a European city characterized by a geographical dimension of 150 km2 with about 1 million of inhabitants. Three different services (voice, video and data) are considered with a traffic distribution over tri-sectorial sites with GSM and UMTS co-located cells. In MBMS case study, the target of the techno-economic analysis is to compare the different investments needed to enhance an existing 3G network, in order to fulfill the requirements due to the introduction of massive (in the long-term) services based on TV and video, on a mobile terminal, comparing two alternatives: 1) exploitation of HSDPA connections and 2) MBMS introduction to provide broadcast/multicast. In order to evaluate the number of carriers per cell and the number of base stations needed to support the assumed traffic, the dimensioning model takes into consideration both the coverage extension of each cell and its capacity requirements, which are derived from the interference limit in uplink and from the transmission power limit in downlink. The techno-economic analysis on MBMS is completed by a short but comprehensive overview of Mobile Operators strategies and market trends of TV Mobile services in Europe. Main alternative mobile technologies which the Mobile Operators worldwide are reported and commented, with a full list of DVB-H operating networks or running trials in Europe. Some examples of value chains that may be enforced or weakened by different implementations are depicted, by distinguishing, in a rough way, pros and cons for mobile operators, TV broadcasters, content provider, service provider, etc. Another important aspect taken into account in the work related to both the two techno-economics investigations carried out consists in the analysis of the market penetration of multi-mode terminals. How much relevant this aspect could be within the context of a heterogeneous network scenario is clear: by means of the CRRM mechanisms addressed by the AROMA projects, different services are supposed to be offered by means of different radio access networks and technologies, in a transparent way for the users, with the aim of improving the QoS and optimizing the network. It is evident that this objective can be accomplished only if a not negligible percentage of users own terminals capable of using most of the radio technologies taken into account. The obtained results show the economic benefits due the adoption of load balancing strategies in a heterogeneous scenario and the advantages of using MBMS with respect to HSDPA to offer mobile TV. It is remarked that, more than the specific economic figures related to the analyzed scenario, the added value is represented by the analysis methodology that has been applied to both the case studies.



1 INTRODUCTION This deliverable reports an analysis of the potential economic impacts tied to the RRM/CRRM solutions investigated within the AROMA project. The following section 2 is devoted to sketch an high level general overview of AROMA RRM/CRRM algorithms and solutions which will be the basis for a techno-economic analysis detailed in subsequent chapters. A rationale of the ideas behind the choice of the most appropriate algorithms is also presented. In section 3 CRRM algorithms only are taken into account, showing for each of them (always in a qualitative way but, with a higher detail with respect to section 2 the economic impact they may have. The methodology followed for the detailed techno-economic evaluations for the selected case studies is shown in section 4 where some forecast about market penetration of multi-mode terminals are provided too. Section 5 reports the techno-economic evaluation of the fittingness factor CRRM algorithm while the techno-economic evaluation of mobile TV over MBMS is reported in section 6. The topic of section 5 concerns a CRRM algorithm that has been chosen among other ones, because it represents a general framework for implementing many different policies, so it has a lot of possible applications. Concerning section 6, the choice of developing a business case centered on MBMS depends on the fact that this technology, which is very promising from a commercial and evolutionary point of view, has also been widely studied in AROMA deliverables, and many different algorithms are based upon it. For both solutions a complete techno-economic analysis will be developed in subsequent sections, after giving some introductory economic background information, especially for MBMS. Some final consideration are drawn in the conclusions reported in section 7.The technical models used in the two case studies are reported in sections 8 and 9.

2 AN OVERVIEW OF AROMA RRM/CRRM ALGORITHMS AND SOLUTIONS

2.1 Table of AROMA algorithms/solutions and investigations As a preliminary work an high level overview of all RRM/CRRM related AROMA algorithms/solutions and investigations has been done collecting in a table all the related relevant information. The collected information includes:

• The short name of the algorithm/solution or investigation used for quick references • The category to which the developed algorithm or solution belongs to: RRM, CRRM, QoS and

Auto-tuning • A more detailed categorization is also given in a Sub-category explanation • A brief description • The involved RAT and the linked scenarios where the algorithm or solutions are applied.

Moreover, some brief information, which is related to the expected economic impacts these solutions or algorithms may determine, is reported too:

• Technical KPI (Key Performance Indicators)1 • Brief description of potential techno-economic impacts2 • Potential techno-economic KPI3

The described table is presented in the following four pages

1 Short description of the most relevant sources foreseen for greater revenues and/or optimization useful to reduce investments (CAPEX and/or OPEX). 2 The impacts are expressed in terms of CAPEX, OPEX, revenues or a mixed combination of these categories. 3 A short description of a possible way of evaluating the economic impacts, comprising a qualitative relationship with other typical marketing quantities, such as willingness to pay, ARPU, etc.



2.2 Economic impact of AROMA algorithms and solutions At a very early stage of our analysis we realized that two main subjects were of particular interest for techno-economic analysis among the topics which have been investigated within AROMA project:

1) the CRRM , due to their future foreseen widespread, because of the already set trend for mobile operators to build and put together different mobile access networks.

An important starting consideration is to underline that, as a lot of studies already clearly demonstrated, the 3G network investments are concentrated on the radio access part in a percentage of 70% [38], so that an optimization on this portion of the network is crucial to have an effective reduction of the overall costs. Also, with the introduction of data oriented services in wireless access systems, we see that traffic demand will be increasingly heterogeneous. Applications have differentiated quality of service requirements and usage patterns may vary significantly over the service area (e.g. in an airport or station vs. a residential area). At the same time, traffic volumes per connection are significantly higher than for voice services, which will put additional requirements on cost efficient network deployments. As a means to solve this problem, operators may exploit multiple radio access technologies, composed in so-called “heterogeneous networks” in order to match network designs to non-uniform spatial traffic distribution. This choice must also be enforced by using some algorithm and procedures which may take into account and take advantage of distinguish properties or different radio access technologies. For example, as a general rule a macro cellular system should primarily be used to provide a wide area coverage whereas pico cells may be deployed in “hot spots”, that is in small areas with high traffic density. The distribution of the traffic between the two systems may be optimized by the introduction of specific CRRM strategies, taking into account both the geographic split of the traffic and the different type of users which show different usage of a mix of services. Even when introducing different emerging technologies in an already existing 3G network, the development of CRRM algorithm within the radio access network is vital for a proper functioning of a heterogeneous network topology.

2) the MBMS business case, especially in comparison with other solutions for the deployment of a Mobile TV service.

Mobile TV in increasingly getting a hot topic, which has been widely discussed among different players in the telecommunications and media industry. Mobile operators are facing the saturation of voice services and a declining ARPU so that they hope that the TV concept in the mobile phone will be a potential way to revitalize a quite stable market in the developed countries. The development of mobile Television in the context of European policies for Information Society and for the Media is strongly enforced as it is clearly stated in this recent statement from Viviane Reding [39], Member of the European Commission responsible for Information Society and Media: “Mobile TV seems set to become the next high growth consumer technology. It is at the crossroads of two powerful social trends: greater mobility, and new forms of accessing media content. Mobility is a powerful driver of growth: in early 1990 even the most optimistic forecasts for mobile phones were 40 million users worldwide by the turn of the millennium. Today over 1.5 billion people use GSM phones worldwide (with approximately 430 Million in Europe). This growth continues; one million new users sign up to GSM services every day. European industry and business have clearly benefited greatly from these developments: a study by Deutsche Bank estimates that GSM contributes about 2% to Europe’s GDP. This is a significant achievement. We all know it from our own lives. The new found mobility and freedom of communication that GSM has given us has changed the way we work and our daily life. Many applications were a surprise: think how SMS has created entirely new social networks, particularly among youngsters. Mobility and mobile systems are also helping to bridge the Digital Divide. In Asia, Latin America and Africa we see massive take-up of mobile networks. Media content consumption is also changing. New diverse audiovisual services are emerging, based on the internet, outside of the traditional triangle of TV, radio and written press. A content revolution is in the making. Internet diversifies information channels and – more far reaching - it has a significant economic impact on business models and advertising budgets. We can expect more personalized; time shifted; on-demand; non-linear services. Consumers will expect more choice and more individual treatment. On-line content markets are predicted to double in size in the next three years to reach nearly 3bn€. The amazing growth rate of blogs (a new blog is being created every second) is another example that points in that direction. The success of video on demand services is another indicator. If we use this opportunity right this trend offers not only new growth opportunities, also a new channel for creativity, diversity and democracy. “



3 CRRM AROMA ALGORITHMS AND SOLUTIONS

3.1 Overview of AROMA CRRM algorithms In order to identify the most interesting CRRM solution to be analyzed in depth from the economic point of view, we picked from the general table of the investigated topics the seven solutions dealing with CRRM and, for each of them, we expanded the qualitative techno-economic description. These descriptions are provided in the next seven sub-sections.

3.1.1 Common congestion control This algorithm addresses the topic of how to solve congestion situations in UTRAN/GERAN networks by means of executing vertical handover procedures between both RATs and also by means of RAT-specific procedures, like, for instance, the bit rate reduction procedure in UTRAN. The proposed algorithm has an impact both on the QoS that can be offered to the user and on the network capacity. The objective of a network operator that decides to introduce such procedures in an heterogeneous network, comprising different radio access techniques, is to support its customers with the required QoS in a profitable way to derive new revenues. In AROMA D12 deliverable [8] there is an implementation of the formulas (based on an analytical markovian model which has been developed on many AROMA deliverables) that may determine, in an analytical way, the congestion probability in a combined UTRAN/GERAN scenario. These formulas are helpful to solve many congestion cases, nowadays still based on some procedures (defined in AROMA D09 deliverable [7]) of congestion detection and resolution (or reduction). The percentage of time that the network is in congestion is an indication of the Perceived QoS (users' Quality of Experience) and eventually Assessed QoS (risk of churn).

3.1.2 Coverage-based CRRM for Voice Traffic As reported in section 5.9 of [7], the Coverage-based CRRM algorithm may determine an increase of capacity with respect to other basic CRRM algorithms such as for example the Load Balancing algorithm. The simulations show that this algorithm may increase the capacity of an heterogeneous network (based on co-sited GSM and UMTS radio base stations) of 31% without compressed mode (which is the case when we have multimode terminal with many transceivers in both radio access technologies) and of 27% with compressed mode (which is the case when we have multimode terminal with only one set of transceivers). From an economic point of view this may be evaluated by comparing the different investments (necessary to face a defined increase of voice traffic capacity):

1) without interworking between GSM and UMTS network (reference case). 2) with interworking between GSM and UMTS network and a CRRM load balancing algorithm. 3) with interworking between GSM and UMTS network and a CRRM coverage based algorithm.

In 2) and 3) a slight difference may also be evaluated taking into account the need of using the terminal in compressed mode (reference case) with respect of NOT using the terminal in compressed mode (but in the second case some assumptions about the effective availability of multimode terminals with multiple transceivers must be taken into account). In general the migration by GSM operators of a portion of their subscribers to UMTS is used to ensure increased quality of service for GSM/GPRS users, but it may also reduce the cumulative investment needed for both the GSM and UMTS access layer, providing that an interworking between the two access networks and a robust and tailored CRRM algorithm is adopted.



3.1.3 Fittingness factor algorithm This algorithm influences both the initial RAT selection and the inter-RAT handover decisions on the base of the value of a “fittingness factor” which reflects the degree of adequacy of a given RAT to a given user. This particular kind of metric takes into account the terminal capabilities, the RAT suitability with respect to the service and user requirements and the network state (i.e., path loss, load and interference state). It is then evident that, both macroscopic aspects at cell level and microscopic aspects at local level are considered so as to select the optimal RAT in any phase of a call. The peculiarities of the algorithm make clear that it impacts first of all on the perceived quality of the connection. This feature represents a key enabler for the operator interested in an increase of the revenue by providing the adequate QoS level to the customer according to his specific needs. This helps also in terms of general customer satisfaction reducing then the potential risk of churn. Moreover, being the algorithm based also on macroscopic level consideration, the choices driven by the fittingness factor also influence the overall heterogeneous network capacity. If any ongoing call during its lifetime is always making use of the most suitable RAT, it should happen rarely that non high performance demanding services are exploiting the most performing RATs. This turns into an optimal exploitation of the available RATs according to the operator’s requirements. As usual, the optimal resource exploitation turns into CAPEX savings (avoidance or delay in network upgrades/expansions).

3.1.4 CRRM perceived throughput The legacy EVEREST project developed a simulator which may evaluate the global performances (in terms of overall throughput of a defined system) of A) a defined CRRM algorithm, taking into account the traffic steering that this algorithm realizes on different radio access techniques in a defined scenario and with a certain mix of services, comparing it with a B) manual RAT selection case, where no CRRM algorithms are available. Economic evaluations of the downlink were already done in previous works, but the simulator has been extended in order to take into account the overall performances of the system in the uplink, so similar economic considerations might be repeated also for that case. Using the techno-economic calculation tool based on the environment, traffic and service mix described for the target scenario of hotspot within urban area we get a total network architecture estimated from the need of both coverage and capacity. Different CRRM algorithms will result in different total usage of the RATs and this study analyses the financial impact based on the required network resources. Hence, different CRRM algorithms will generate different network investment requirements to the network operator.

This analysis may show the potential gain of using CRRM within high traffic areas such as hotspot environments. Exact figures in investment costs can not be given due to the constant market variation, but the overall conclusions regarding the possibilities for CRRM optimization impact on the radio access network cost may be identified.

3.1.5 Opportunistic CRRM This algorithm can be applied for Mobile Services which belong to QoS Background class according to 3GPP definition, which tolerates a delay (Round Trip Time) without any particular limit. As a consequence, the mobile network can maintain the same service session and manage the Radio access bearers with more flexibility. As regards the mobile network architecture and radio access, a Mobile Operator might use this algorithm in case of massive deployment of HSDPA network in order to have good radio coverage in the main dense urban area. This scenario reflects the real status of HSDPA market deployments since many Mobile Operators are deploying HSDPA network with a high penetration rate in terms of population radio coverage. In these conditions this algorithm can provide Mobile Operators with more flexibility in the managing radio bearers with great benefits to reduce the OPEX with the same investments expected for the radio access.



3.1.6 CRRM Cost Function Heterogeneous networks may be characterized by several parameters that determine their performance, being known as the Key Performance indicators (KPI). This CRRM algorithm tries to integrate a set of KPIs into a single one by using a Cost Function that takes these KPIs into account, providing a single evaluation parameter as an output. Such an output reflects the different network conditions and CRRM strategies performance. In other words, this solution enables the implementation of different CRRM policies by manipulating KPIs according to users or operators perspectives. Due to the peculiarities of the algorithm, that can allow accomplishing different objectives, it is quite obvious that the economic implications can be very different as well. Delay only policies or Block only policies can provide some kind of privilege to packet switched service and to circuit switched services respectively. The economic impact in this case is strictly tied to the QoS experienced by the users and then, to the user category that the operators would like to prioritize. Other kind of choices in terms of considered KPIs in the cost function can lead to a more balanced heterogeneous network, with an evident impact on capacity (in this case the implications affect, as usual, the CAPEX more than the revenues). In conclusion, it is possible to understand a more precise economical impact of this solution only when the KPIs taken into account in the cost function are known.

3.1.7 MPLS based mobility management and IP QoS The algorithm is actually a RRM enhancement algorithm because it is completely based on LTE architecture. The performance of some of the algorithms that were introduced are also evaluated in presence of vertical handovers among different RATs, but all the reported algorithms concern the core network (centralized QoS management via a broker) and NOT the access network. All the data that concern vertical handovers are considered as input coming from other radio level simulations. Techno-economic analysis may be addressed by estimating the economic impacts of the benefits involved in maintaining the sessions while performing horizontal and vertical handovers by MPLS based mobility management and the improvements in service quality by QoS routing. MPLS is at the moment supposed to save the providers money by letting them consolidate multiple overlay services over a single network infrastructure, one based on IP/MPLS. Some doubts are coming because MPLS has proven to be an OPEX challenge for the carriers and it turned out, as much as everyone always complained about the inflexibility and high cost of SONET/SDH, that the cost of moving away from those technologies is anyway quite high. A techno-economic analysis based on the introduction of such a technology in the core network for mobility management purposes is helpful to understand the real benefits coming from its adoption. As operators move toward a consolidation of all their service networks the demands on technology increase. The common packet core network needs to be multiservice and able to transport TDM, packet voice, Frame Relay, ATM and IP traffic. As this packet network is carrying mission-critical and revenue-generating traffic, it needs to be easily upgraded. The migration path from an ATM-centric core network to a R5 next-generation IP/MPLS network needs to be as smooth and CAPEX-friendly as possible. For all of the above reasons, the best and safest option for operators is to deploy this packet core network on multiservice switches. In addition, an IP VPN solution implemented on multiservice switches yields more than 60% CAPEX savings compared with an overlay router solution, as proven by a real Nortel Networks implementation for a very large West-European wireless operator. Extending this multiservice packet core in the RAN will further improve the return on investment. According to Nortel network research [6] cost savings of more than 60% were achieved when implementing a multiservice switch-based solution offering business-grade IP VPN services versus an overlay IP router solution.



4 METHODOLOGY FOLLOWED FOR THE TECHNO-ECONOMIC EVALUATIONS

4.1 Investments versus revenues valorization In sections 5 and 6 two different solutions among the ones addressed by the AROMA project have been evaluated from a techno-economic point of view, by analyzing the potential economic impacts relying on the technical aspects already fully investigated by the project. More in detail, in section 5 the advantages of adopting a CRRM RAT selection algorithms in a 2G and 3G co-site scenario have been analyzed whereas in section 6 the mobile TV over MBMS scenario have been compared with respect to the mobile TV over HSDPA one. It is worth noting that even tough the two above mentioned solutions are intrinsically different (in the first case a RAT selection CRRM algorithm is evaluated whereas in the second case two different technological options are compared) the general approach followed to evaluate these solutions is basically the same. In both the two cases the techno-economic evaluations have been carried out by assuming a short- or medium- term increase of the data traffic and by analyzing the potential savings offered by the addressed solutions with respect to the total investment (i.e. CAPEX and OPEX) needed to increase the capacity of the network. Hence, these evaluations are based only on the estimation of the total costs faced by a network operator for upgrading the already deployed network in order to support the expected amount of traffic. In principle, an alternative way of calculating the economic value of the solutions taken into account could be based also on the estimation of the extra revenues related to the additional data traffic supported by the network. Within the context of the work, this approach has been considered less appropriate, since it would require as much exact as possible assumptions on the revenues deriving from the services. Unfortunately, market forecasts on revenues could be very subjective and are usually affected by an higher degree of uncertainly with respect to the estimation of the network investments, since these are strictly related to the willingness to pay of the users for new services. Moreover, revenues from the offerings of new services are strictly dependent also on specific marketing strategies carried out as far as the end-user pricing policies is concerned, that can find justifications on many reasons (e.g. promotion of a specific new service by means of flat-rate prices, volume discounts to boost the usage, etc.). Within the context of a complex scenario like the one made possible by the all IP heterogeneous network, pricing policies are evidently a very complex issues (especially compared to the voice only traditional scenario), since different pricing and charging schemes are possible (e.g. daily, weekly, monthly flat rate, per subscription rate, data traffic rate, etc. Moreover service revenues not always are proportional to the load generated in the network, because these respond to different (marketing based) mechanisms with respect to the technical ones. Thus, the value of the service can be not related at all the traffic generated. As a consequence, the revenues typically do not grow linearly with the amount of data exchanged, and this make hard to estimate the economic impacts of the addressed solutions on the basis of the capacity increase achieved within the heterogeneous network. For the above mentioned reasons, considerations on potential revenues have been avoided and are outside the scope of this document (detailed information concerning end-user pricing and revenues forecasts can be found in IST TONIC Deliverable D11 [28].

4.2 Dependence of the results from the time based traffic hypotheses As already mentioned in the previous section, the techno-economic investigations reported in this document are based on the assumptions that a not negligible increase of data traffic will be demanded by users of mobile heterogeneous network in the next years, especially in dense populated areas. Several market forecasts agree on this assumption on the basis of the recent trends observed in European countries where 3G systems are more diffused nowadays. An example of this trend in data traffic grown requested by mobile subscribers in Finland is depicted in Figure 4-1 [33].



Figure 4-1 - Example of observed data traffic increase from 2005 to 2006 [33]

Even though the increase of data traffic demands is evident, it is however very difficult to say to what extent, and when, the potential savings related to this market trend can be realized. This uncertainty rise from the difficulty to make reliable traffic forecasts on a per year base, and thus it is difficult to say how big the demand for future network capacity will be. In order to limit the sensitivity of the results on the traffic forecasts, the approach based on the comparison of the solution with respect a “reference case” has been followed. In this way, economic impacts have been highlighted according to a “what-if” approach, apart from the absolute values achieved. This means that the results reported in the next section should not be considered relevant in an absolute way but should be considered useful to compare the different scenarios taken into accounts. In any case, it is worth noting that in both the two carried out investigations, the Net Present Value (NPV) of the investments estimated during the reference period of times has been taken into account. This calculus has been made in order to have a general idea of the actual economic value of the addressed solutions, even if should be clear that it strictly depends on the specific time assumptions considered.

4.3 Market penetration of multi-mode terminals Another important aspect taken into account in the work consists in the thorough analysis of the market penetration of multi-mode terminals. How much relevant this aspect could be within the context of a heterogeneous network scenario is clear: by means of the CRRM mechanisms addressed by the AROMA projects, different services are supposed to be offered by means of different radio access networks and technologies, in a transparent way for the users, with the aim of improving the QoS and optimizing the network. It is evident that this objective can be accomplished only if a not negligible percentage of users own terminals capable of using most of the radio technologies taken into account. For instance, in the case of CRRM algorithms devoted to allocate the traffic between GSM and UMTS (ref. section 5.3.2), it is clear that the more dual-mode terminals widespread, the more efficient can be the final effect of this type of algorithm, as the achieved results demonstrate. Similarly, when evaluating the economic implications on the best way to accommodate the mobile TV subscribers (ref. section 6), it can be guessed how much important is the considered market availability of HSDPA and MBMS terminals, respectively. For the above mentioned reasons, as a starting point of the work, an in deep investigation concerning the actual mobile terminal penetration (differentiated with respect to the different technologies available nowadays) as well as the expected short-term evolution of them has been carried out. On this concern the following information has been collected from public sources [28], [29], [30], [31] and further elaborated for the specific scenarios taken as reference. According to Forrester Research [29] Europe’s operators estimate the UMTS adoption only by 10% of the European mobile users in 2007. This research also shows that UK and Italy are in the lead for 3G



adoption: these countries are expected to see 3G penetration rates of 68% and 72%, respectively, by the end of 2010. As depicted in Figure 4-2, Forrester Research predicts 60% of Europeans will have 3G mobile phones by the end of 2010, as depicted in Figure 4-2.

Figure 4-2 – Penetration of UMTS terminals [28].

Referring to [30], Figure 4-3 provides an assumed long-term growth rate for 3G penetrations up to 2021:

Figure 4-3 – Expected penetration of UMTS terminals up to 2021. [30]

With respect to the penetration of HSDPA capable terminal, as reported in [28] according to Wireless Intelligence, mobile broadband connections are expected to reach 40 billions. worldwide by the end of 2008 and WCDMA HSDPA is expected to represent around 45% of total WCDMA cellular connections by 2010. Similarly, Pyramid Research outlooks that about 19% of global mobile phone subscribers (40% of WCDMA connections) will have HSDPA capable terminals by 2011, ad depicted in Figure 4-4 [28].



Figure 4-4 – Terminal penetrations versus different technologies [28].

Also according to a recent forecast of Juniper Research [31], HSPA technologies will dominate the high-end mobile broadband market over the next five years, accounting for about 70% of global 3G deployments. More in detail, Figure 4-5 shows the expected HSDPA penetration growth for 2008, 2009 and 2010.

Figure 4-5 – Forecast HSDPA growth (Source: Strategy Analytics). [31]

Finally, according to a white paper from the UMTS forum [32], by 2012, there will be almost 1 billion users of HSPA technology worldwide, as reported in Figure 4-6 [32].

Figure 4-6 – Expected HSPA users worldwide (Source: UMTS forum) [32]



From the data reported above, we have derived the subscription penetration forecast for different mobile systems in Western Europe for the next coming five years including the running year, reported in Table 4-1.

Table 4-1 - Penetration forecasts of GERAN vs. UTRAN subscribers for the next 5 years. YEAR Percentage of GERAN

subscribers in the system Percentage of UTRAN

subscribers in the system 2007 64 % 36% 2008 57% 43% 2009 48% 52% 2010 38% 62% 2011 35% 65% 2012 33% 68%

Similarly, Table 4-2 reports the considered penetration rates of HSDPA terminals with respect to the UMTS R99 ones, for the next 5 years.

Table 4-2 - Penetration forecasts of HSDPA vs. R99 in UTRAN for the next 5 years.

YEAR Percentage of UMTS HSDPA terminals

Percentage of UMTS R99 terminals

2007 10% 90% 2008 18% 82% 2009 25% 75% 2010 33% 67% 2011 45% 55% 2012 65% 35%

From the last two tables we can derive also the distribution of GSM, UMTS R99 and UMTS HSDPA terminals that have been considered in the evaluations reported in sections 5 and 6, as shown by Table 4-3.

Table 4-3 - Forecast of distribution of GSM, UMTS R99 and UMTS-HSDPA terminals

YEAR Percentage of GSM terminals

Percentage of UMTS R99 terminals

Percentage of UMTS HSDPA terminals

2007 64 % 32.00 % 3.60 % 2008 57% 35.26 % 7.74 % 2009 48% 39.00 % 13.00 % 2010 38% 41.54 % 20.46 % 2011 35% 35.75 % 29.25 % 2012 33% 23.80 % 44.20 %



5 TECHNO-ECONOMIC EVALUATION OF FITTINGNESS FACTOR CRRM ALGORITHM

5.1 Introduction In this section potential economic benefits of applying traffic steering policies between GSM and UMTS systems have been investigated. In the considered case study, the traffic steering strategy is supposed to be realized by means of a specific implementation of a RAT selection algorithm based on the Fitness Factor framework4. This framework has been developed by the AROMA project and assessed from a technical point of view in AROMA Deliverable D12 [8]. The techno-economic evaluation has been carried out by assuming an increase of the data traffic demands for the next 5 years which requires new investments on network resources and by analyzing the potential savings offered by the CRRM algorithm with respect to the case when the algorithm is not present. In both these two cases (when the algorithm is present and when it is not), the specific investments considered in order to increase the capacity of the network to fulfill the traffic increase consists in the upgrade of the already deployed UTRAN sites by activating the second UTRAN carrier (using the same approach described in [13]). This choice can be understood by considering that, as it is well known, in WCDMA there is a limit over the capacity increase achievable by means of the introduction of new sites in a dense urban area. As a matter of fact, several UTRAN cells not so much spaced in a small area are not able to offer much higher capacity due to the mutual interference that every cell causes to each other. On the other hand, in a capacity limited downlink scenario, it is evident that the additional frequency is able to offer a direct capacity increase5. As a results, in capacity limited urban areas adding more spectrums is a very cost efficient way to increase capacity whereas the alternative of increasing capacity by deploying new macro base stations is significantly more expensive.

5.2 Algorithm definition The scenario where the CRRM algorithm is supposed to operate consists in a fictive already deployed 2G/3G heterogeneous network which offers the radio coverage in a specific area be means of GSM and UMTS co-site cells. The GSM coverage is supposed to be complete in the area of interest whereas the UTRAN coverage is not complete (i.e. about 90% of the total area is covered). Within this scenario, the following three main categories of mobile terminals diffused in the market nowadays are supposed:

• Category 1: single-mode”GSM-only” mobile terminals (i.e. 2G terminals) • Category 2: dual-mode “GSM /UMTS-R99” mobile terminals (i.e. 3G terminals, not HSDPA

capable) • Category 3: multi-mode “GSM/UMTS-R99/HSDPA” mobile terminals (i.e. 3G terminals,

HSDPA capable) In the considered case study case, it was assumed that subscribers having GSM terminals request only the voice service (neglecting the case of requests of data services over GPRS or EDGE, due to the fact that packet data services are provided in a best effort way on these technologies and this does not impact on the voice capacity). On the other hand, 3G users are supposed to be able to request voice, video call as well as web browsing services. When the CRRM algorithm is absent (reference case), the following prearranged camping strategy6 is performed by the network (this strategy corresponds to what often happens nowadays in real network scenarios, where both 2G and 3G systems coexist in the same area): 4 Specific details on the algorithm, in addition to what already described in section 2 and section 3 are provided in the next section 5.3 5 Under certain conditions it is possible to see the additional carrier as being a parallel network. Limitations on the power amplifier means that the capacity gain could be lower than two, but it will still be significant. 6 Information about how to implement this strategy has been fully investigated in AROMA Deliverable D12 [8].



• 2G terminals camp (obviously) on GSM system • 3G terminals always camp on the UMTS system when a suitable UTRAN cell is available

(from a radio quality point of view), so that they camp on GSM only in lack of 3G coverage conditions

Moreover, in the reference case it is assumed that users camped on UMTS which request the www services are always allocated on HSDPA, on condition that they own a HSDPA capable terminal7. Instead, video call service is always allocated on UMTS R99, since this real-time service requires a fix amount of bandwidth both in uplink and downlink. In the second case, when the CRRM algorithm is present, the selection among GSM, UTRAN dedicated transport channels and HSDPA is always performed in accordance of the fitness factor values evaluated by the CRRM algorithm for each RAT, on the basis of the cell load of the cells. As already evaluated in AROMA Deliverable D12 [8], by taking into account the fitness factors, it is possible to implement load balancing strategies among the involved RATs efficiently. It is worth noting that the more the 3G terminal penetration the more the negative consequences of the absence of load balancing mechanisms: as a matter of fact, when 3G terminal penetration increases, all the traffic concentrates in the UTRAN network causing potential over-loaded conditions, whereas the GERAN is more and more under-loaded with respect to the optimum. Table 5-1 summarizes services and RAT involved in the scenario taken into account and specifies the traffic channels and radio access bearers’ characteristics:

Table 5-1 – Services, RAT and radio bearer characteristics. GSM UMTS DCH UMTS HSDPA

Voice TCH Full Rate traffic channel

12.2/12.2 kbit/s UL/DL CONVERSATIONAL CS RAB Considered as not supported

Video call Considered as not supported

64/64 kbit/s UL/DL CONVERSATIONAL PS RAB Considered as not supported

WWW Considered as not supported

64/384 kbit/s UL/DL INTERACTIVE PS RAB

HS-DSCH channel with proportional fair packet scheduling algorithm

According with the fitness factor framework, the behavior of the CRRM algorithms depends (also) on the definition of the network-centric suitability ( )NFηδ associated to each RAT, that represents a function that reduces the fittingness factor of the RAT depending on the amount of non-flexible load. The specific functions chosen for GSM, UMTS-R99 and HSDPA, reported with greater details in 5.3.2, have been identified with the aim of implementing the following high-level load balancing strategy:

1. Voice calls is preferably allocated in GSM, when possible (only when the GSM cell has no more radio resource available the user requesting voice is allocated in the UTRAN co-located cell, if available)

2. WWW connections are preferably allocated in HSDPA: only when a high number of contemporary HSDPA users are present within a cell, the CRRM algorithm may select dedicated channels for allocating the new www request.

Advantages of the load balancing mechanism implemented by the CRRM algorithm on the basis of he two mentioned strategies have been investigated by evaluating year per year how many cells are no more able to support the assumed traffic so that they need to be upgraded with a second UTRAN carrier. Results reported in section 5.5, demonstrate that when the CRRM algorithm is present, it is possible to have a smaller number of second carrier activations with respect to the reference case.

7 This implies that all the UTRAN cells of the scenario are supposed to support HSDPA.



5.3 Algorithm modeling and Markov Chain design

5.3.1 Flowchart of the CRRM algorithm The CRRM algorithm which implements the load balancing strategy between GERAN and UTRAN cells is fully specified by the following flowchart:

Start

Voice call requestcoming from single-mode

terminal (GSM-only)

Voice call requestcoming from dual-mode

terminal (GSM/R99)Video call request

WWW connection requestcoming from dual-modeterminal (R99/HSDPA)

WWW connection requestcoming from single-mode

terminal (R99-only)

1 2 3 4 5

Start

Voice call requestcoming from single-mode

terminal (GSM-only)

Voice call requestcoming from dual-mode

terminal (GSM/R99)Video call request

WWW connection requestcoming from dual-modeterminal (R99/HSDPA)

WWW connection requestcoming from single-mode

terminal (R99-only)

1 2 3 4 5 Figure 5-1 – CRRM algorithm flowchart.

With the aim to simplify the global flowchart, we explain each case from 1 to 5 as a stand-alone item.

• Voice call request by a single-mode terminal (GSM-only)

Voice call requestsby a single-mode

terminal (GSM-only)

CAPGSMvo GSMN <

Voice call requestaccepted on

GSM resources

Voice call requestblocked for GSM-only

voice terminals

Yes No

1Voice call requestsby a single-mode

terminal (GSM-only)

CAPGSMvo GSMN <


GSM resources

Voice call requestblocked for GSM-only

voice terminals

Yes No

1

Figure 5-2 - Management of a voice call request (GSM-only terminal)



• Voice service request by a dual mode terminal (GSM/R99)

Voice call requestsby a dual-mode

terminal (GSM/R99)


GSM resources

No

2

Evaluation of

[ ]voiceRP 99

Probability that a dual-mode terminalis covered by R99 voice coverage

GSM available

Dual-mode terminal coveredby R99 and GSM voice coverages

Dual-mode terminal coveredonly by GSM voice coverage

GSM&R99 available GSM-only available R99-only available


R99 resources


GSM resources

Voice call request blockedfor dual-mode voice

terminals with probability

[ ]voiceRP1 99−99RGSM δδ ≤


GSM resources


R99 resources



[ ]voiceRP 99

Yes

No

Yes

No

Yes

NoYes

Yes

No

Voice call requestsby a dual-mode

terminal (GSM/R99)


GSM resources

No

2

Evaluation of

[ ]voiceRP 99

Probability that a dual-mode terminalis covered by R99 voice coverage

GSM available

Dual-mode terminal coveredby R99 and GSM voice coverages

Dual-mode terminal coveredonly by GSM voice coverage

GSM&R99 available GSM-only available R99-only available


R99 resources


GSM resources



[ ]voiceRP1 99−



[ ]voiceRP1 99−99RGSM δδ ≤


GSM resources


R99 resources



[ ]voiceRP 99



[ ]voiceRP 99

Yes

No

Yes

No

Yes

NoYes

Yes

No

Figure 5-3 - Management of a voice call request (dual mode terminal)

• Video call request

No

Video call request

R99 resourcesavailable

Voice call blocked withprobability equal to 1

Evaluation of

Yes

3

[ ]videoRP OK 99

Probability that a terminalis covered by R99 video coverage

Video call requestaccepted on R99

resources with probab.

[ ]videoRP OK 99

Video call requestblocked with probability

[ ]videoRP1 OK 99−

No

Video call request


Voice call blocked withprobability equal to 1

Evaluation of

Yes

3

[ ]videoRP OK 99

Probability that a terminalis covered by R99 video coverage



[ ]videoRP OK 99



[ ]videoRP OK 99




[ ]videoRP1 OK 99− Figure 5-4 - Management of a video call request



• Data service request from dual mode terminal (R99/HSDPA)

Yes

WWW connection requestby a dual-mode terminal

(R99/HSDPA)

No

4

Evaluation ofProbability that a dual-mode terminalis covered by R99/HSDPA coverage

Dual-mode terminal covered byR99/HSDPA coverage

R99&HSDPA available

WWW connectionrequest acceptedon R99 resources

WWW connectionrequest accepted

on HSDPAresources

Yes

No

[ ]dataRP OK 99

Dual-mode terminal not coveredby R99/HSDPA coverage

WWW connection requestblocked for dual-mode data

terminals with probability[ ]videoRP1 OK 99−

HSDPA available

HSDPAR δδ ≥99


on HSDPAresources

Yes

NoR99 available


Yes


terminals with probability[ ]dataRP OK 99

No

Yes

WWW connection requestby a dual-mode terminal

(R99/HSDPA)

No

4

Evaluation ofProbability that a dual-mode terminalis covered by R99/HSDPA coverage

Dual-mode terminal covered byR99/HSDPA coverage

R99&HSDPA available



on HSDPAresources

Yes

No

[ ]dataRP OK 99

Dual-mode terminal not coveredby R99/HSDPA coverage





HSDPA available

HSDPAR δδ ≥99


on HSDPAresources

Yes

NoR99 available


Yes





No

Figure 5-5 - Management of a WWW connection request (dual mode terminal)

• Data service request by a single-mode terminal (R99-only)

No

WWW connection requestby a single-mode terminal

(R99-only)


WWW connectionblocked with

probability equal to 1

Evaluation of

Yes

5

Probability that a terminalis covered by R99 data coverage

WWW connection requestaccepted on R99


[ ]videoRP OK 99

WWW connection requestblocked with probability


[ ]dataRP OK 99

No

WWW connection requestby a single-mode terminal

(R99-only)


WWW connectionblocked with

probability equal to 1

Evaluation of

Yes

5

Probability that a terminalis covered by R99 data coverage



[ ]videoRP OK 99



[ ]videoRP OK 99





[ ]dataRP OK 99

Figure 5-6 - Management of a WWW connection request (R99-only terminal)

5.3.2 Analytical model of the CRRM algorithm

In order to predict the behavior of the GSM/UMTS heterogeneous network with and without the CRRM algorithm described in the flowcharts shown in paragraph 5.3.1, two analytical models of the network based on a Markov chain have been developed.



Mathematical details of the network’s model when the CRRM algorithm is present can be found in Annex A (ref. section 8), whereas the network’s model when the load balancing strategy is not applied is not reported in the document since the approach is the same and it is a much more simple case than the previous one.

5.4 Scenario and QoS constrains

The scenario considered in order to apply the fittingness factor algorithm to a well-defined case study, concerns a fictive European town characterized by a medium relevance both in term of geographical dimension (about 150 km2) and in term of number of citizens. We take into account 24 tri-sectorial sites with GSM and UMTS co-located cells. Three different services (voice, video and data) are considered with a realistic traffic distribution. The distribution of the offered traffics permits to manage a non homogeneous importance of the various districts in which the urban area is divided.

The global traffic scenario is constituted by five years (2008-2012) to analyze. During this period, an appropriate evolution of some parameters is taken into account. As a consequence of this, we consider:

• a constant voice traffic

• a constant video traffic

• a growth (respect to year 2007) of the data traffic (see Table 5-2)

Table 5-2: data traffic growth during period 2008-2012

Year Data traffic growth 2008 1.03 2009 1.06 2010 1.09 2011 1.12 2012 1.15

• a variation of dual-mode GSM/UMTS terminals penetration (see Table 5-3))

Table 5-3: Dual-mode GSM/UMTS terminals during period 2008-2012

Year Only-GSM terminals GSM/UMTS terminals 2008 57% 43% 2009 48% 52% 2010 38% 62% 2011 35% 65% 2012 32% 68%

• a variation of HSDPA capable terminals penetration (see Table 5-4)

Table 5-4: HSDPA capable terminals during period 2008-2012

Year R99-only terminals HSDPA-capable R5 terminals

2008 82 18 2009 75 25 2010 67 33 2011 55 45 2012 35 65

The behavior of the co-located cells (GSM and UMTS cells are considered, in the markovian model, like a unique system) is evaluated in term of performances using a set of threshold able to put in



evidence if a generic pair of co-located cells works or not in an efficient way. These thresholds are summarized in Table 5-5:

Table 5-5: threshold related to performances of a pair of co-located-cells Performance Threshold value

Voice loss 2% Video loss 5% Data Loss 5%

HSDPA throughput 400 kbps Data throughput 350 kbps

A generic pair co-located cells that does not respect one or more performances shown in Table 5-5, is equipped with an additional transceiver using the second carrier on the UTRAN cell in order to guarantee better performances in terms of losses and data throughputs. The CRRM fittingness factor algorithm is compared with the reference case (where the pre-arranged camping strategy already described in 5.2 is assumed), on the base of the number of co-located cells that during period 2008-2012 need to be increased in term of R99 equipment.

The number of GSM carriers of the cells in the scenario guarantees, in year 2007, a voice loss of the co-located cells equal to ≅2%. The GSM carriers’ distribution is shown in Figure 5-7:

0

16 16

20

10

43

1 12

0

5

10

15

20

25

1 2 3 4 5 6 7 8 9 10

Number of carriers

Cel

ls

Figure 5-7: distribution of the GSM carriers in year 2007

The other scenario parameters8 are reassumed in Table 5-6:

Table 5-6: further scenario parameters

Parameter Value Ortogonality factor 0.9

Codes with spreading factor 128 available for R99 calls 87 Codes with spreading factor 128 used by a R99 voice call 1 Codes with spreading factor 128 used by a R99 video call 4 Codes with spreading factor 128 used by a R99 data call 16

Rate of users in macro diversity 0.2 Geometrical Factor (G) 6 dB

Power used for signaling common channels 36 dBm Noise power in UL section -105 dBm

Maximum power transmitted by the mobile terminal 21 dBm Maximum load for R99 cell in UL section 0.7

BSC Gain in UL section 18 dB Eb/N0 required in UL section for voice calls 8.5 dB

8 Considered values come from the AROMA scenarios described in deliverables D05 and deliverable D11.



Eb/N0 required in UL section for video calls 5 dB Eb/N0 required in UL section for data calls 2.8 dB

Bit rate in UL section for voice service 12.2 kbps Bit rate in UL section for video service 64 kbps Bit rate in UL section for data service 64 kbps

Service activity factor rate in UL section for voice service 0.9 Service activity factor rate in UL section for video service 0.9 Service activity factor rate in UL section for data service 0.3

UL Inter-cell interference 0.6 Noise power in DL section -100 dBm

R99 average path loss in DL section 100 dB Maximum load for R99 cell in DL section 0.95

Eb/N0 required in DL section for voice calls 8.5 dB Eb/N0 required in DL section for video calls 7 dB Eb/N0 required in DL section for data calls 4.2 dB

Bit rate in DL section for voice service 12.2 kbps Bit rate in DL section for video service 64 kbps Bit rate in DL section for data service 384 kbps

Service activity factor rate in DL section for voice service 0.9 Service activity factor rate in DL section for video service 0.9 Service activity factor rate in DL section for data service 0.6

5.4.1 CAPEX and OPEX for UTRAN Carrier upgrade The CAPEX and OPEX for UTRAN Carrier upgrade are the same described in techno-economic evaluation of MBMS, please refer to sections 6.7.2 and 6.7.3.

5.5 Results The analytical model reported in the previous section has been exploited to derive technical indicators concerning the performance of the heterogeneous network with and without the presence of the load balancing mechanism implemented by the CRRM algorithm. Achieved results deal with the blocking probability experienced by the uses when they request services to the network as well as the mean per user perceived throughput for the data service, making the distinction between mean throughput offered by the network to the HSDPA users and the mean throughput offered to all the users requesting data and allocated both over HSDPA and over DCH. Also the users distribution over each RATs (in terms of mean number of contemporary active users) has been evaluated in order to fully investigate the final results of the allocation decisions implemented by the CRRM algorithm and to compare them with respect to the users distribution deriving by the pre-assigned mechanisms considered in the reference case. The above mentioned KPIs have been evaluated for all the couples of 2G and 3G cells located in the area of interest, and per each of the 5 years taken into account. In Table 5-8 it is reported an example of the achieved results with respect to the input data summarized in Table 5-7, as far as the first 10 couples of 2G/3G cells is concerned. Input values refer to the first year of the analysis (2008) where the penetration of 2G terminals is 57% and the 3G one is 33% and where only 16% of the 3G terminals support HSDPA. In the table reporting input values (ref. Table 5-7), per each row it is specified the traffic offered to each 2G/3G co-located cells for each of the three services considered (in terms of Erlang for voice and video call services and in terms of mean number of contemporary RAB connections for www service), as well as the number of Traffic CHannels (TCHs) owned by the GSM cell.

Table 5-7 – Scenario related input data for the analytical model.



Input Data 2G/3G co-located cells

Info Traffic load Terminal penetration

cell ID GSM TCHs Voice [Erl]

Video [Erl]

WWW [# RABs] GSM only

not HSDPA capable

1 21 14.2 0.4 1.5 0.57 0.82 2 35 27.0 0.7 2.8 0.57 0.82 3 12 6.4 0.2 0.7 0.57 0.82 4 13 8.0 0.2 0.8 0.57 0.82 5 10 4.8 0.1 0.5 0.57 0.82 6 11 5.5 0.1 0.6 0.57 0.82 7 12 6.4 0.2 0.7 0.57 0.82 8 13 8.0 0.2 0.8 0.57 0.82 9 10 4.8 0.1 0.5 0.57 0.82

10 49 40.7 1.0 4.2 0.57 0.82 In the table reporting results (ref. Table 5-8) KPIs with and without the CRRM algorithm are reported. In general, these results demonstrate how much important could be the load balancing mechanism implemented by the CRRM algorithm. As a matter of fact, when the algorithm control the RAT allocation of the users, lower system level blocking probabilities can be achieved, as well as higher values of offered throughput. As example, referring to the results related to the first couple of 2G/3G cells, a lower blocking probability is obtained with the CRRM algorithm with respect to the video call service (1.9% versus 3%) and also an higher throughput is offered by the network to each user requesting the www service (with respect to the HSDPA users, 943 kbit/s versus 717 kbit/s).

Table 5-8 – KPIs results derived by using the analytical model Blocking probability Per user perceived th-put Users distributions on RATs

Total Voice Video WWW HSDPA

users [kbps]HSDPA+DCH users [kbps]

voice users on

GSM

voice users on UTRAN

video users on UTRAN

www users on UTRAN

without CRRM 1.33% 1.32% 3.05% 3.06% 717 420 10.78 5.92 0.03 0.12 1 with CRRM 1.33% 1.33% 1.92% 1.92% 943 431 14.30 0.14 0.04 0.12 without CRRM 2.02% 2.00% 5.39% 4.61% 558 409 24.40 11.09 0.06 0.22 2 with CRRM 1.78% 1.78% 1.98% 1.98% 887 430 28.1 0.35 0.07 0.22 without CRRM 1.15% 1.09% 2.49% 2.69% 695 422 3.66 2.70 0.16 0.53 3 with CRRM 1.03% 0.99% 1.99% 2.01% 782 427 6.3 0.05 0.16 0.53 without CRRM 1.24% 1.17% 2.66% 3.07% 645 419 4.58 3.33 0.19 0.65 4 with CRRM 1.74% 1.72% 2.03% 2.08% 739 426 7.7 0.10 0.19 0.66 without CRRM 1.77% 1.72% 2.52% 3.43% 724 562 1.5 3.16 0.12 0.19 5 with CRRM 1.08% 1.02% 2.32% 2.40% 758 425 2.72 2.00 0.12 0.39 without CRRM 1.11% 1.05% 2.40% 2.52% 728 423 3.15 2.32 0.14 0.46 6 with CRRM 0.91% 0.86% 1.97% 1.98% 809 428 5.4 0.03 0.14 0.46 without CRRM 1.63% 1.62% 3.80% 3.78% 624 414 17.58 8.70 0.05 0.17 7 with CRRM 1.83% 1.83% 1.95% 1.95% 911 430 21.3 0.28 0.05 0.17 without CRRM 1.61% 1.60% 3.75% 3.74% 629 414 17.53 8.56 0.05 0.17 8 with CRRM 1.64% 1.64% 1.95% 1.95% 914 430 21.2 0.25 0.05 0.17 without CRRM 1.52% 1.51% 3.51% 3.52% 653 416 15.56 7.77 0.05 0.15 9 with CRRM 1.51% 1.51% 1.94% 1.94% 923 431 19.1 0.21 0.05 0.16 without CRRM 1.36% 1.35% 3.12% 3.13% 707 419 11.68 6.20 0.04 0.12 10 with CRRM 1.25% 1.25% 1.92% 1.92% 940 431 15.13 0.14 0.04 0.12

By looking at the values of the users’ distribution in Table 5-8, it is easy to understand that the good performance achieved by the CRRM algorithm is due to a better exploitation of the GSM system:



users requesting voice owning a dual mode terminals are allocated to the GSM cell and this permits to have more capacity for data services in the UTRAN cells. Of course, this strategy is implemented by taking into account the constrains imposed by the scenario. As already mentioned, UTRAN coverage is not the same with respect to the GSM one, so that only a part of users having dual-mode terminals can be allocated to the UTRAN cell when they request the voice service. On this concern, as already described in section reporting the analytical model details (ref. Annex A), the coverage of UTRAN cells also depends on the load of the cell, so that it decreases when the load of data service increase, as can be deducted from the pole capacity formula. The cell breathing phenomenon can have some impacts on the results and for this reason it has been taken into account. As an example, Table 5-9 reports the percentage of UTRAN coverage with respect to the GSM one, for the 10 couple of cells extracted from the entire scenario taken as reference:

Table 5-9 - Percentage of coverage offered by the UTRAN cell with respect to the GSM one

cell ID

UTRAN coverage (% with respect to the

GSM one)

1 93.80% 2 93.06% 3 96.01% 4 95.02% 5 96.42% 6 96.16% 7 95.08% 8 94.09% 9 96.27%

10 92.98%

As mentioned in 4.2, the KPIs described above have been used to estimate how many cells cannot respect the QoS constrains specified in Table 5-5, due to an excessive amount of traffic offered by the users. In this way we derived the number of UTRAN second carriers that has to be considered in order to support the increase of data traffic during the five years. Results of this elaboration are reported in Table 5-10 and depicted in Figure 5-8 (percentage with respect to the total number of UTRAN cells).

Table 5-10 – UTRAN cells which require the 2nd carrier activation due to each QoS constrain. 2008 2009 2010 2011 2012 Percentage of

UTRAN 2nd carrier activations

without CRRM

with CRRM

without CRRM

with CRRM

without CRRM

with CRRM

without CRRM

with CRRM

without CRRM

with CRRM

Voice loss 22.2% 13.9% 50.0% 4.2% 77.8% 0.0% 80.6% 0.0% 86.1% 0.0% Video loss 23.6% 0.0% 31.9% 0.0% 50.0% 0.0% 52.8% 0.0% 56.9% 0.0% Data loss 18.1% 0.0% 31.9% 0.0% 47.2% 0.0% 65.3% 0.0% 76.4% 0.0% HSDPA Thr-put 2.8% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% Data Thr-put 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% Total 23.6% 13.9% 56.9% 4.2% 79.2% 0.0% 86.1% 0.0% 90.3% 0.0%



23.6%

13.9%

56.9%

4.2%

79.2%

0.0%

86.1%

0.0%

90.3%

0.0%0%

10%20%30%

40%50%60%

70%80%90%

100%

2008 2009 2010 2011 2012

Percentage of 2nd UTRAN carrier activation

without CRRMwith CRRM

Figure 5-8 – Percentage of UTRAN cells which require the 2nd carrier activation.

Results related to the number of UTRAN cells that should be upgraded by introducing the second UTRAN carrier demonstrate clearly the benefits of using the CRRM algorithm. It is worth noting that, when the CRRM algorithm is used, the number of activations estimated by considering the QoS constrains decrease in the years following the first one. This issue can be understood by considering that the penetration of 3G terminals (as well as the penetration of HSDPA ones) increases during the time. For this reason, also the overall network performance increase due to the higher degrees of freedom of the CRRM algorithm, as already anticipated in section 5.1. Even if this phenomenon exists, it should be considered that in practice the investments related to the upgrade of UTRAN cells to introduce the second carrier are not alienable. This practical consideration has been taken into account in the estimation of the total investments, which has been done by considering that the investments related to the ten UTRAN cells required in the first year of the analysis cannot be recovered in the following years. In this sense, the considered profile of second UTRAN carrier activations for the estimations of the investments is reported in Figure 5-9.

23.6%

13.9%

56.9%

13.9%

79.2%

13.9%

86.1%

13.9%

90.3%

13.9%

0%10%20%

30%40%50%60%

70%80%90%

100%

2008 2009 2010 2011 2012

Percentage of 2nd UTRAN carrier activation(Hyp. not alienable investments)

without CRRMwith CRRM

Figure 5-9 – Percentage of UTRAN cells which require the 2nd carrier activation

(hypothesis of not alienable investments). The total investments (CAPEX + OPEX) needed to introduce the additional transceivers using the 2nd UTRAN carrier with and without the CRRM algorithm are depicted in Figure 5-10.



Total 2nd UTRAN carrier investments

€ 0

€ 200,000

€ 400,000

€ 600,000

€ 800,000

€ 1,000,000

€ 1,200,000

€ 1,400,000

2008 2009 2010 2011 2012

without CRRM

with CRRM

Figure 5-10 – Total investments (CAPEX + OPEX) needed to introduce the 2nd UTRAN carrier

(hypothesis of not alienable investments).

Figure 5-10 clearly demonstrates the positive economic impacts offered by the CRRM algorithm in terms of investments savings. In any case, concerning this aspect, it is worth noting that the cost of the introduction of the algorithm has not considered, since it is very difficult to estimate and it strictly depends on the specific implementations and technological choices which can be very different case by case. In any case, also by considering that an extra cost should be included for the introduction of the CRRM algorithm, the economic benefits for an operator in terms of investments savings is not put in jeopardy. Finally, Table 5-11 reports how the investments should be spread over the considered five years (note that with the CRRM algorithm, investments are needed only in the first year, in order to upgrade the ten most critical cells with the second UTRAN carrier).

Table 5-11 – Per year investments to introduce the 2nd UTRAN carrier. Per year 2nd UTRAN carrier investments

2008 2009 2010 2011 2012

without CRRM € 306,000 € 432,000 € 288,000 € 90,000 € 54,000with CRRM € 180,000 € 0 € 0 € 0 € 0

By means of the figures reported in table above, it is possible also to estimate the Net Present Value9 for the two considered cases, which represents the actual value of the entire investments made over the five years (ref. Table 5-12).

Table 5-12 – Net Present Value of the five years 2nd UTRAN carrier investments. 5 years 2nd UTRAN carrier investments

NPV

without CRRM € 1,118,669with CRRM € 176,471

The difference in terms of NPV between the two considered cases (delta NPV) is equal to 942,199 Euro. This figure can be considered as the last key indicator of the positive economic impacts offered by the CRRM algorithm.

9 A WACC of 2% is considered.



5.6 Conclusion The results of the techno-economic analysis reported in section 5.5 clearly demonstrate the positive potential economic impacts of the introduction of the CRRM algorithm taken into account, in terms of investments savings. Even if a specific algorithm was taken as reference, the load balancing strategies implemented by the algorithm for voice and data services (ref. section 5.2) can be considered very general. For this reason, results similar to the ones reported in this work can be expected also with different CRRM algorithm implementations, on condition that they are devoted to put into practice the considered load balancing mechanisms. In this sense, the validity of the carried out techno-economic investigation can be considered quite general, too. On the other hand, it cannot be forgotten how much sensitive could be the results achieved with respect to the amount of traffic assumed, as mentioned in section 5.1. It is evident that, if the hypothesis on the increase of traffic demanded by the mobile users will not come true, no benefits from whatever RAT selection CRRM algorithms can be expected in economic terms, since at the present the most number of mobile networks present in the main European countries have been deployed to support adequately all the actual traffic requested by users. It should be also considered that the economic analysis was based on the assumption that the network operator already has an important asset consisting in the already deployed GSM network. All the advantages offered by the CRRM algorithm is due to a better exploitation of this asset for voice users, which cannot be exploited adequately when 3G terminals are camped on UTRAN by default. In this sense, the validity of this analysis is limited to the case where the network operator has different radio access technologies and intends to put into practice appropriate strategies to exploit all of them within the context of an all-IP heterogeneous network.



6 TECHNO-ECONOMIC EVALUATION OF MOBILE TV OVER MBMS

6.1 Introduction The choice of developing a business case focused on MBMS depends on the fact that this technology, which is very promising from a commercial and evolutionary point of view, has also been widely studied in AROMA deliverables, and many different algorithms are based upon it. As reported in section 6.3.1, a new kind of traffic generated by mobile TV subscribers is expected for the near future. Due to the characteristics of this service, the amount of data generated by these users can be significant and it impacts the capacity of the network. Operators have multiple technology options to choose from to provide mobile TV and video services. Many of these options are either non available today or are currently being traded around the world. In this period before deployment of mobile broadcast or multicast networks, operators typically offer a streaming unicast/video download service or an early-stage mobile TV broadcast offering. A mobile broadcast TV network, such as for example, a network based on DVB-H, is a separate network from a cellular one, so it will require a more significant infrastructure than providing either a multicast or streaming/unicast offering over a cellular network. That is the reason why operators will typically not make this investment alone, but either partner with a media company or lease the service from the mobile TV provider. Moreover, most of the traffic for Mobile TV is generated by users which requires also voice and data services offered by means of UMTS network, mobile network operators have instead the opportunity to offer also the TV service by upgrading and improving the capacity of their UMTS networks, which is an alternative that may benefit from saving in investments and on a large scale basis economy.

The decision behind what type of broadcasting technologies can better fit Mobile Operator capacity needs and strategies also strongly depends on both regional regulatory constraints (standard and frequency band) and industry lobbies. The GSM Association has presented some recommendations for efficiently introducing Mobile TV in European countries. The matter of costs involved for providing such services for these areas makes its future rather questionable [14]. As alternative, in principle mobile TV service can be offered by means of a DVB-H infrastructure, which is a network based on different infrastructures with respect to the mobile network. In this case, investments required to deploy a DVB-H system able to offer good levels of coverage are very high. In a statement of July 2007 the Commission of the European Communities said it would “encourage the implementation of DVB-H” for mobile TV reception in Europe and warned it could mandate DVB-H as a common standard. One of the disadvantages related to a full usage of this standard all over Europe, is the fact that in some European countries, DVB-H will have to wait up to five years for spectrum to become available. Generally speaking, ambiguities regarding spectrum allocation for DVB-H can delay wide spread launch of DVB-H Mobile TV until 2008 [34] The status of the diffusion of DVB-H operating networks or running trials in Europe is reported in the following Table 6-1 [15].



Table 6-1 – Diffusion of DVB-H networks and trials in Europe [15]

Full Service

Launch Trial Service

(O = completed or results available)

ALBANIA

DigitALB LAUNCHED

AUSTRIA Nationwide 2008 Vienna/Salzburg X

BELGIUM

Ghent / Brussels / Mechelen X

DENMARK Copenhagen X

FINLAND

Mobiili-TV LAUNCHED Helsinki O

FRANCE

Metz X Paris (TDF) (Phase 2) X Paris (CANAL+) X Nationwide Q3 2008

GERMANY

Berlin (bmco) O Berlin (T-Systems) X Erlangen X Nationwide Q1/Q2 2008

HUNGARY

Budapest X

IRELAND Dublin X

ITALY

3 Italia LAUNCHED TIM TV LAUNCHED Vodafone LAUNCHED Turin O

NETHERLANDS

IBC 2007 - Amsterdam Nationwide (KPN/Digitenne) Q1 2008 The Hague O

POLAND

Warsaw O

PORTUGAL Lisbon (TVI & RETI) X Lisbon (SGC Telecom) O

SPAIN

Barcelona / Madrid O Nationwide 2008 Seville - Axion Technical Trial X Seville / Valencia O Zaragoza / Gijón X

SWEDEN

Gothenburg & Stockholm (Teliasonera) X Stockholm (Teracom) O Stockholm (Viasat) X

SWITZERLAND

Bern - (Customer Acceptance) X Bern - (Technical) O Nationwide 2008

UKRAINE

Kiev X

UNITED KINGDOM Cambridge X Oxford O



In the European countries where mobile TV is actually offered by means of DVB-H, in order to reduce the requested investments, mobile operators have established alliances with traditional TV broadcasters. In any case, the resulting business model of this kind of strategy is yet unclear and the revenues from mobile TV over DVB-H at the moment are not satisfactory. DVB-H mobile TV services are not likely to become a major source of revenues, let alone the next ‘killer application’, for mobile operators or broadcasters [17] Finally, from a technical point of view, indoor penetration capability of DVB-H signal is very poor and this requires a significant increase of antennas inn order to guarantee good quality of service also inside buildings. For the reasons mentioned above, mobile operators are considering the possibility to increase the capacity of their UMTS network in order to accommodate also the mobile TV traffic.

6.2 DVB-H VERSUS MBMS FOR MOBILE TV There are actually different options in order to delivery Mobile TV service, basically three different ones:

• Mobile TV on UMTS This option consists on delivering unicast streaming video on demand and live streaming via GPRS and UMTS (point to point).

• Mobile TV on MBMS (point to multipoint) • Mobile TV on DVB-H (broadcast network)

The following Figure 6-1 shows the market for handheld television services:

Figure 6-1 - Market for handheld television services [35]

So, around 40-60% of mobile phone users would be interested in receiving TV on their mobile phone and willing to pay up to 10 Euros per month for the service. When technology is introduced, a number of constraints come into play against the various business models which are based on the availability of Spectrum and the availability and maturity of the particular technology at infrastructure and handset levels. It is important to understand these variables as they influence the business models availability and time to market capability. Also, there is the question of which of the business models will succeed, which in turn, will drive down costs of both infrastructure and handsets through volume demand and delivery. Clearly a combination of technology and spectrum harmonization provide the fastest route to lower cost and provide common user navigation of Mobile TV Services. The generic value chain for Mobile TV can be drawn as in the figure below:



Figure 6-2 – Generic value chain for Mobile TV [35]

The different roles that the different agents involved in the value chain can adopt may be observed in the next table: Content

provider Service provider

Billing Downlink network operator

Return network operator

Content provider

Yes Yes Off chance Improbable Improbable

Service provider

Yes Yes Yes Improbable MVO

Broadcast operator

Yes Yes Off chance Yes MVO

Mobile Operator

Yes Yes Yes ?¿ Yes

Business model for MBMS is the simplest one.

Figure 6-3 - UMTS evolution business model [35]

While DVB-H overlays current 2G/3G generating a risky model:

Figure 6-4 - DVB-H business model [35]



6.3 Overview and market trends of Mobile TV service A lot of European and USA Telecom operators are already offering Mobile TV as a part of their advanced multimedia services, or are at least completing their trials for this service. As a general rule, the basic value chain for mobile TV over 3G is essentially similar to other types of content services delivered over 3G, with video platform management being a key difference. Few operators are seeking to dominate the supply chain (among them 3 Italy and 3 UK), whereas most of them are outsourcing the content adaptation, aggregation and video platform management to their partners including Mobile TV, Wonder Phone and Kelyan Lab [36]. A dedicated mobile broadcast network operates according to a model that assumes collaboration between mobile operator and broadcaster, because of high costs of deploying a dedicated mobile broadcast network alone is high. A wholesale model is an attractive option used by some countries, which also enables a multi mobile broadcast technology approach with a single, centralized platform and multiple delivery networks. In UK, BBC has offered the most important channel television (with the exception of some sports) and many radio channels via the networks of 3, Orange and Vodafone. Sky has launched its mobile television service at the end of 2005 and has currently 230,000 viewers out of more than 8 million subscribers [22]. A less significant number of customers are also using Virgin Mobile service, started in October 2006. In this context the major interest on Mobile TV is coming form BBC itself, which considers Mobile TV as an important part of its future strategy of content distribution, both for retention of existing users, but also to increment the market share and promotes trials in order to understand consumer behavior on the mobile media. The BBC has also launched an enhanced web service for users with mobile phones with web browsers [23]. Sky also offers BSkyB Mobile TV, which is a portal accessible via GPRS/UMTS (by all operators except 3) which makes available all the contents of some Sky channels: Sky Sports, Sky News, Sky Weather, and a personalized access to a portal of sport news and related bets. A specific downloadable client is needed, available for a small subset of Nokia commercial cellular phones. BT is offering a wholesale platform (based on DAB-IP, which is very widespread in UK) for the entire content provider that do not want to develop their own network. Virgin is the first company giving a mobile TV service based on this BT wholesale platform. Orange is the most important success case of Mobile TV in Europe, with an offer of 60 live channels and thousands of short video clips (increasing every week) on 3G and DVB-H (known as HD Mobile TV) networks [24]. The business model in Germany consists of a Mobile Network Operator Consortium which uses the DVB-H network of some network operators, buys the content from some Content Providers and resells wholesale services to Mobile Network Operators and other resellers. A group of common base (access fee) and premium channels (pay per view) is shared among different operators, whereas some exclusive channels are specific to every mobile operator [36]. The same principle has also been followed by Spain where there is only one distribution network managed by the national TV broadcaster Abertis Telecom, which gives a wholesale service (partly common channels and partly operator specific channels) to the three mobile operators. The second phase of the trial has finished in February 2007 [36]. In USA Verizon Wireless is offering services from February 2007 on dual mode terminals using FLO standard developed by Qualcomm [25]. The service is available in numerous US cities. Also AT&T Cingular has signed an agreement with Qualcomm to use the MediaFlo TV service for Cingular Mobile TV offer to be launched in late 2007, so the two biggest US carriers will probably dominate the market, even if there are other initiatives coming from Modeo (a cellular tower operator) which is making a trial based on DVB-H and Windows Mobile 5.0, mainly aiming at the market for mobile TV on laptops and other portable devices. Another forerunner to Mobile TV service offerings in USA arena is MobiTV which has reached 2 million users in February 2007 [25]. At the moment South Korea is the world’s most advanced mobile TV market with a good penetration rate of Mobile TV service based on Mobile Broadcasting technologies: T-DMB and S-DMB. The total number of mobile customers is about 7 million: 2 million for S-DMB technologies and 5 million for T-DMB. The Korean players involved in the deployment of S and T-DMB declare to be far from break even. In any case, as usual, a key factor for the success of the Korean Mobile TV service is the availability of Mobile terminals supporting DMB technologies which include GPS navigator. [37]



The main characteristics of the principal broadcast Mobile TV technologies currently used in European pilots and/or commercial launches, also listed in a separate table are here reported [40]:

Figure 6-5 – Mobile TV Broadcast Technologies in Europe [40]

Figure 6-6 – Mobile TV Broadcast Technologies used in EU Member States [40]

6.3.1 Mobile TV Market: European scenario

According to the main analyst and info providers Mobile TV services will allow Mobile Operator revenues to increase significantly over the next 3-5 years. Informa Telecom, for example, expects that the DVB-H customers will be 120 millions worldwide by the end 2011 and up from 100.000 by the end of 2007. [17] In addition to that the revenues from the traditional Voice Service are falling (down) due to a strong competition on pricing and marketing offer. In this scenario it’s becoming increasingly Mobile Operator, both incumbent and new comer, to come up with Mobile Services which increase revenues so as to have a return on investment in a short term.

At the moment Mobile Operators have a wide range of mobile technologies at its disposal in order to provide a Mobile TV service. The Figure 6-7 [17] illustrates the main mobile technologies which the Mobile Operators worldwide are investigating and evaluating.



Figure 6-7 – Main mobile technologies investigated and evaluated [17]

In the European market MO are focusing on DVB-H and DMB technology as regards pure broadcasting technologies, but all mobile Operators have already launched a Mobile TV services based on UMTS technology10 and the migration towards HSDPA can guarantee higher bit rate on the radio interface and as a consequence a better QoS and customer experience.

Up to now there are only (http://www.dvb-h.org/services.htm) five Commercial Services based on DVB-H technology in Europe, three of which are in Italy where Telecom Italia and Vodafone share the same DVB-H network owned by Mediaset, the main Italian private Media Company. In parallel there are a lot of ongoing trials (about fifty) and some MOs in France, Germany and Spain and have declared the commercial launch in a short time. The mobile operators in time being are also investigating two key issue regarding the service delivery

• What customer behavior for Mobile TV services will be adopted? • What type of mobile or broadcasting technology will be more suitable for the Mobile

Market?

As regards to the customer behavior it is expected that the overall time spending on consuming media will be spread over different media platform both on fixed and access domain. Community, file sharing, peer to peer communications are some example of new ways of delivering and retrieving digital contents on fixed domain and it’s not clear what will be the customer behavior on mobile domain. In the case of personal content such video on demand, Podcasting and Personal TV the unicast mobile technology will provide a good answer from technological point of view. In this scenario MBMS technology can provide Mobile Operators with a good flexibility in the managing and delivering mobile content in the case of traffic load in certain dense urban areas o in the case of peak traffic in certain time. In the case of delivering media content with same paradigm of traditional analogical TV services, which means delivering 10-12 TV channels at the same time for a mass market, the broadcasting technology is the only ways of providing a satisfactory service. But also in this service scenario MBMS can play a role of complementary technology in order to deliver mobile content with a reduced duration in limited area or to target market segments such as Business TV. Each technology has its strengths. DVB-H and broadcasting technology handle a heavy usage of Mobile TV services with many TV channels broadcast all the time. Cellular technologies such as HSDPA and MBMS handle a low medium usage in a complementary way and for customer behavior more similar to what happens for Video Service on Internet domain.

10 Mobile TV as streaming over dedicated channel.



The main disadvantage of broadcasting technology is that a Mobile Operator needs to deploy a more significant infrastructure investment. Then Operators either partner with a media company or lease the service from the mobile TV provider such as BT (Movio), MediaFLO or Mediaset. In Italy there is also an exception for 3 which acquired a regional TV broadcaster which includes the frequency license and high power transmitter for about 200 million of euros [36]. It’s clear that the deployment of broadcasting technologies strongly depends on the market penetration of Mobile TV services as well as how long time each user watches Mobile TV on average.

6.4 Methodology for the techno-economic evaluation of mobile TV over MBMS

The main objective of the proposed techno-economic analysis is to investigate the total investment (i.e. CAPEX and OPEX) needed to increase the capacity of the UMTS network in order to be able to support also the traffic generated by the expected Mobile TV service subscribers comparing HSDPA and MBMS technologies. As mentioned in 4.1, we decided not to estimate the revenues in a complete development of a full business case mainly because it’s very difficult to have a significant forecast of this economic value in a long-term analysis. Usually the pricing of the mobile data and VAS services tends to get flat as the related service becomes widespread and due to promotional offers, so that the simple evaluation of the investments is a good way of comparing the two ways (HSDPA and MBMS) of delivering Mobile TV over 3G network. As well-known (results in this sense have been provided also by the AROMA project in [8]), the capacity of UMTS network can be effectively increased by means of the HSDPA technology. Even if HSDPA can be very useful to increase the network’s capacity, in the case of Mobile TV, also the MBMS technology can be taken into account. As described in [8], MBMS has been defined just in order to use in efficient way the radio resources when the same content must be distributed among different users. For this reason the analysis has taken in consideration these two different technologies in order to see in which condition one of the two technologies may be considered more desirable with respect to the other (e.g. after how many years, under certain assumptions, MBMS may be considered more profitable than HSDPA to support Mobile TV subscribers, considering that in for MBMS it is necessary to reserve a fixed amount of the cell transmitted power). With respect to this goal, the methodology taken into account is based on the comparison between two different cases:

• “Mobile TV over HSDPA” • “Mobile TV over MBMS”

For both the two cases considered, starting from the already deployed network suitable for the actual traffic level, the considered investments in order to support the expected traffic increase consist first in the upgrade of the already existing cells to activate the 2nd UTRAN carrier and then in the deployment of new sites11 (using the same approach described in [13]). The following figure shows the logic modules of the evaluation model:

11 Deployment of new sites is considered starting from the condition that all the already existing cells of the scenario have been upgraded with the second UTRAN carrier.



Figure 6-8: Logic modules of the evaluation model.

6.5 Economic impacts of MBMS The analysis was made by taking into account a 10 year period starting from 2008 to 2018, because it is commonly accepted that the adoption of Mobile TV services will be gradual and not disruptive. For that reason the main economic results will be evident on a long-term scale, whereas in a short-term analysis, no meaningful outcomes may be pointed out. The target of the techno-economic analysis is to compare the different investments needed to enhance an existing 3G network, in order to fulfill the requirements due to the introduction of massive (in the long-term) services based on TV and video, on a mobile terminal, in a broadcast/multicast context. The two alternatives taken into account are: 1) HSDPA connections, which set up a new connection for each user requesting the service 2) MBMS, which may guarantee some broadcast/multicast capability in an easy and direct way, when the number of Mobile TV users within a cell exceeds a specific threshold, as fully investigated in [8]. In both cases, the supposed growth of users which have a significant daily usage of that service will cause an increase of the overall traffic which may be served (with an acceptable quality of service), by correctly dimensioning the radio access part. When the traffic demand grows, it is possible to gradually introduce a second carrier for some cells, up to a full duplication of the carrier for every cell (at least in the dense urban and urban sites, which are the most affected by the traffic increase). When the duplication is completed, new sites are added to the original network structure. The objective of our analysis is to demonstrate that in the first years of the plan, when the growth of Mobile TV users is still in its raising phase, the adoption of a MBMS network is not so relevant and perhaps even more expensive with respect to the HSDPA case12, whereas in the following years, when the number of users is bigger and the service is best-established, the MBMS solution is the most preferable one. In evaluating the investments which are needed for each solution, only the more expensive changes are determined. For example, it is assumed that the hardware and software upgrades costs relative to the introduction of HSDPA or MBMS are neglected because their order of magnitude is lower with respect to the planning of a second carrier or of new sites in order to serve the new traffic requirements. Moreover, these upgrades are commonly integrated in new versions of software (and

12 Due to the fixed amount of cell power that has to be necessarily reserved for MBMS.

Offerred Traffic Scenarios

Dimensioning model for MBMS

Dimensioning model for HSDPA

CAPEX Valorization model

CAPEX Valorization model

Unit CAPEx

For MBMS

Unit CAPEx

For MBMS

Comparison of HSDPA and MBMS



hardware) for the network nodes, so that it’s not convenient to separate their cost from the general cost of a new software and/or hardware release comprising many add-on features.

6.5.1 Market Assumptions for Mobile TV services The market figures which constitute the input of dimensioning model are both derived from some basic working assumptions which are considered to be very likely for an average European country and from some publicly available market data which are extended in the immediate future by using some well known forecasting methodologies. The number of total connections and related country penetration are used to derive the number of users for a generic operator, also taking into account its actual and foreseen market share, and the penetration of 3G handsets with respect to the whole number of connections, coming from the afore-mentioned technology split. The dimensioning is based on peak hour quantities, so the global annual value, mixed with the typical daily usage, must be integrated with some other information, widely available on literature, on the daily distribution of traffic and its relation with peak hour. At the end of the processing of the market forecast and the average daily usage assumptions, we obtain the number of the traffic figures expressed in annual Voice end Video telephony Erlangs in peak hour, and Interactive and Background Mbit/s, similarly referred to peak hour. The quantities related to Mobile TV and Video services are also expressed in Mbit/s and they are kept apart from the other input quantities because in the following dimensioning we may decide to use HSDPA channels or a MBMS transport, and then compare the different investments due to these two different options. Mapping of the different services on different RAB is reported in Table 9-1and Table 9-3.

6.5.2 Service mix There is a particular emphasis on TV and video connection forecasts, but the dimensioning is made as a whole on a mix of services which is consistent with AROMA scenarios assumptions. It’s anyhow impossible (and not realistic at all) to dimension the network elements in an independent way for the different type of services. Also voice and video telephony services are considered in their foreseen growth.

6.5.3 Usage The input data coming from external sources concern the number of users, whereas the usage of each service in terms of number of sessions per day, typically given both for interactive and background services, is a parameter which has been estimated.

6.5.4 European users’ forecasts at national level from 2008 to 2018 The Mobile Broadcast TV forecasts are taken from recent forecast already available in the public domain [16]. The forecasts relating to the increase in Mobile TV use vary and they are typically unclear about which Mobile TV services they include. According to Strategy Analytics, by the end of 2006 there will be 8 million Mobile TV devices globally and by 2010 there will be 120 million Mobile TV service subscribers (3G 2006) [16]. Also IMS Research indicates that there will be 120 million Mobile TV service subscribers in 2010 (Wickham 2005) [16]. Informa telecoms and media estimates that the number of Mobile TV broadcast service users is expected to grow from 130,000 in 2005 to 124.8 million in 2010 (McQueen and Reid 2005). The following table presents an estimation of European and worldwide Mobile TV users according to Informa telecoms and media. [16].

Figure 6-9 – Mobile TV user forecasts (millions)



Please note that the methodology adopted to derive forecast data for Mobile TV users in the mentioned source only considers the users which effectively already own a terminal which is suitable for the fruition of Mobile TV services. We don’t have to apply any MBMS or HSDPA penetration rate to these data, because it is already implicit and embedded in them. The forecasts are then extended to the following years of the analysis by using a Bass diffusion model, which describes the process how new products get adopted as an interaction between users and potential users. The model is widely used in forecasting, especially product forecasting. In its original formulation the formula to be used is [26]:

Where:

• f(t) is the rate of change of the installed base fraction • F(t) is the installed base fraction • p is the coefficient of innovation (also known as external influence or advertising effect) • q is the coefficient of imitation (also known as internal influence or word of mouth effect)

The coefficients have been chosen equal to the average value for a typical long term adoption of a new technology.

6.5.4.1 Total annual traffic for the whole country The forecast related to users growth, mixed with the usage of a mix of different services, determine the overall total annual traffic (for the whole country) which is represented in Table 6-2. The figures have been obtained by processing the forecasts reported in [16], [17] and [33]. Please note that the mix of services has already been grouped by taking into account the different requirements for each specific service, so that only four categories, the ones which are relevant for the dimensioning model, are considered. All the data related to Voice and Videotelephony services are expressed in Erlang, whereas both the Interactive and Background Data are expressed in terms of Mbps. Please note that we neglect the Uplink contribution for the Mobile TV service, assuming that it is non present at all or of an order of magnitude lower with respect to other services.

Table 6-2 – Global annual traffic for the whole country

6.5.4.2 Traffic projection for an “average European town” The total annual traffic, subdivided in the different types of traffic services, so expressed in Erlang or in Mbit/s, is related to an average European country, but we decided to extend our business case to a single average town. We have decided to consider a one million inhabitants town, which is a sort of a prototype for an average industrialized European city. The distribution takes into account both the population of that town referred to the overall country population, and the area which must be covered in order to have a full coverage of the town.



The splitting of the town surface in dense urban, urban and rural areas has been put for the sake of simplicity, equal to (respectively) nearly 20%, 40%, 40%, which represent reasonable figures for a typical town. A rounded number of sites for each area (respectively, 130, 20 and 1) was considered on the basis of typical figures for the scenario taken into account.

6.6 Dimensioning model

Results of the techno-economic evaluation reported in section 6.7 have been derived by applying a specific dimensioning model on the scenario described in previous section. Technical details concerning this model are reported in Annex B (ref. section 9).

6.7 CAPEX and OPEX valorization

CAPEX and OPEX figures considered in the present work and reported in this section are examples based on [12], [13], [41] and further elaborated by the partners of the AROMA consortium, also taking into account a 5% per year depreciation factor as suggested in [27]. With respect to the investments needed to introduce the second UTRAN carrier in a cell, the CAPEX and OPEX figures have been estimated taking into account that change in the number of transceiver equipments from one to two is only needed.

6.7.1 Investment related to the upgrade of HSDPA and MBMS technology These investments where both neglected because their effective value is considered to be not relevant to the scope of our analysis which focuses on the advantages and disadvantages of the two technologies in order to fulfill the requirements due to a large and progressive extension of the Mobile TV service usage in an average town of a European country. In this scenario, the focus cannot be to make a real choice between two different technologies, from the operator point of view, because we can easily assume that both technologies will be by any means introduced by any European “big” mobile operator in the long term as a natural consequence of the common upgrading of their existing networks. In particular:

• The introduction of HSDPA (which consist both in hardware and software upgrades in different network nodes) is linked not only to the Mobile TV itself, but concerns the need of offering a quite wide range of data services, both interactive and background, directly connected to the supposed increasing usage of browsing and streaming applications. Many European operators have already or are quickly introducing HSDPA in their network as a natural upgrade of GPRS, so it is not difficult to suppose that HSDPA will be the rule for every mobile operator which offers a significant bouquet of data services, that definitely means every average mobile operator in the long term

• Also scouting of the solutions which different vendors, main players in the European market, put in their future roadmaps, shows that the changes related to the introduction of MBMS (mainly software updates in NodeBs, RNCs and in the Core Network), are considered as a software upgrade, which will be given in any case as a standard evolution of their RRM component.

• For similar reasons, we may assume that all the new equipment which must be added in the Core Network in order to connect with TV and video content sources, both proprietary or more likely belonging to a Media company linked with a partnership agreement, are an investment which is of a lower order of magnitude, which is probably the same for the two technologies, so that no differences are to be taken in consideration for the single technology, and which will be by any means done, because it is needed for many different applications based on video and streaming fruition (not necessarily only for Mobile TV)

6.7.2 Investment related to the deployment of new sites

CAPEX



The valorization of the network quantities provided by the dimensioning model needs the availability of two classes of information:

• the cost of a single macro site installation. • the investments for the site upgrade from 1 TRx to 2 TRx in order to use the second UTRAN

carrier.

Total CAPEX related to traditional macro site installation were assumed equal to 140 k€, including both site related costs (e.g. civil works, energy, etc.) and equipment related costs (antennas, RF components, base station, etc.). At the moment the price of a new equipment is reducing during the years, we may assume (ref. [27]) that the reduction per year is 5% (see ref. [15]), but this assumption only concerns the first year of the plan because, as the already mentioned study underlines, this tendency is rapidly coming to a saturation point where the CAPEX of a new equipment may be considered as a fixed quantity without a significant reduction nor increase of the costs. These reduction percentages must be applied to the average cost per site of 140k€ already mentioned, in order to derive the final value effectively used in our analysis.

OPEX

OPEX evaluation model takes into account the following items cost: • Energy Cost • Site rental • Transmission cost (leased lines) • Maintenance

Total OPEX per year with reference a single macro site installation were assumed equal to 30 k€, including energy cost, site rent costs and maintenance. When introducing a new carrier we must only consider the 5 k€ which derive from the transmission cost of the rental of a new CDN connection, the site rent is not affected and energy cost and maintenance increase may be typically neglected. No significant reduction or increases are foreseen for the whole period of our analysis.

6.7.3 Investments related to the introduction of the second UTRAN carrier In previous works, total investments related to the introduction of the second UTRAN carrier was estimated in 20 k€ per single site, also including costs related to backhauling. Taking into account the 5% of reduction due to the actualization to 2006, a value of 18 k€ is considered for the present analysis.

6.8 Results of techno-economic evaluation The number of transceivers and sites for each year in the “Mobile TV on HSDPA” case is reported in the following charts:



UTRAN TRx dimensioning(Mobile TV on HSDPA)

0100200300400500600700800900

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

Year

#TR

x Dense UrbanUrbanRural

Figure 6-10 – Transceivers dimensioning (mobile TV on HSDPA).

UTRAN cells dimensioning(Mobile TV on HSDPA)

020406080

100120140160

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

Year

#cel

ls Dense UrbanUrbanRural

Figure 6-11 – Cells dimensioning (mobile TV on HSDPA)

As depicted by the figures above, taking into account our assumptions, it is quite evident that only in the first years of analysis the starting number of transceivers and sites is able to support the assumed traffic. Then, by taking into account the foreseen growth of mobile TV subscribers, results show that from 2011 new investments are needed in terms of 2nd carrier activations. It is also worth noting that, in order to make a fair comparison between HSDPA and MBMS, in the case study we assumed that HSDPA is used only for Mobile TV service. In a realistic case, it should be considered that HSDPA is also exploited to offer high data rate services to the users. In this sense, it should be expected that the capability to support Mobile TV users by means of HSDPA decreases when also other services are allocated on HSDPA and some QOS constrains have to be guaranteed for them. In this sense, the year of the introduction of the second UTRAN carrier will of course anticipate with respect to our analysis (2011). Finally, as depicted in Figure 6-11, it can be observed that the number of sites is constant all over the period for Urban and Rural areas whereas in the Dense Urban Area we have a significant increase of sites only in the two last years. We found similar graphs for MBMS dimensioning:



UTRAN TRx dimensioning(Mobile TV on MBMS)

0100200300400500600700800

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

Year

#TR

x Dense UrbanUrbanRural

Figure 6-12 – Transceivers dimensioning (mobile TV on MBMS).

In the “Mobile TV on MBMS” case and under our assumptions, the number of transceivers already slowly grows starting from the first years, in order to fulfill the foreseen traffic increase, so it can be argued that in the first years of our analysis MBMS is less convenient with respect of HSDPA13. The reason is that the point-to-multipoint transmission offered by MBMS is not exploited until there are only few contemporary mobile TV users per cell and HSDPA is able to offer a greater spectral efficiency with respect to the MBMS, when point-to-point DCH channels are used. Similarly to the previous case, also MBMS is supposed to be used only for Mobile TV service. On the other hand, this assumption is more realistic than in the HSDPA case, so the introduction of a new carrier should not anticipate in a meaningful way with respect to our analysis (2011). Finally, achieved results also show that the number of sites when MBMS is used is constant all over the period for each areas.

UTRAN cells dimensioning(Mobile TV on MBMS)

020406080

100120140

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

Year

# si

tes

Dense UrbanUrbanRural

Figure 6-13 – Cells dimensioning (mobile TV on MBMS).

13 This fact is confirmed by Figure 6-14 where the total amount of investments is reported.



6.9 Conclusions The cumulative CAPEX and OPEX for the two different technologies is compared in the following graph, which is relative to an estimation of the investments needed for an average town in a European country, for a “main” mobile operator (e.g. incumbent or second operator, with a 40% market share).

Cumulative CAPEX and OPEX

0123456789

10111213

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

Year

Euro

(mill

ions

)

HSDPAMBMS

Figure 6-14 – Cumulative CAPEX and OPEX for MBMS and HSDPA

In the first years of the analysis MBMS is less convenient with respect to HSDPA because of the reasons already mentioned in previous section. When the number of Mobile TV subscribers continues to grow, the MBMS solution becomes more convenient. The break even point in our analysis is reached at around 2013. The number of users per cell (and their average usage) which is obtainable in this year may be considered the threshold of the data service usage that makes the introduction of MBMS more advantageous with respect to HSDPA. At the end of the observed period, the ratio of the cumulative investments for the two technologies is equal to 1.90, so the investments for supporting Mobile TV on HSDPA are nearly doubled with respect to the MBMS ones. In our analysis no new sites are added for MBMS solution, whereas their introduction is foreseen for HSDPA starting from 2016 (which corresponds to the knee of the curve of cumulative CAPEX and OPEX for this technology that can be observed in Figure 6-14). In any case, as already mentioned in section 4.2, the achieved results strictly depend on the assumed hypotheses in terms of traffic, deployment and scenario).



7 CONCLUSIONS In this report issues related to the potential economic impacts of the solutions addressed by the AROMA project have been discussed and investigated. First of all, a general methodology based on the evaluation of the total investments (CAPEX and OPEX) needed to increase the capacity of a heterogeneous network and to support the expected increase of data traffic has been proposed. Then this methodology has been applied in two complete case studies. In the first investigation, the economic benefits of realizing load balancing strategies between GSM and UMTS have been clearly demonstrated, by taking as reference a CRRM algorithm which can be implemented according with the Fitness Factor framework, already proposed by the AROMA project. In the second study, advantage of using MBMS with respect to the HSDPA to offer the mobile TV service has been evaluated starting from the most widely accepted hypotheses on the increase of mobile TV subscribers foreseen for the next years. Even tough, as already mentioned in section 4.2, it should be considered that the achieved results strictly depend on the assumed hypotheses in terms of traffic, deployment and scenario).



8 ANNEX A: ANALYTICAL MODEL OF THE CRRM ALGORITHM BASED ON THE FITTINGNESS FACTOR FRAMEWORK

8.1 Analytical model of the CRRM algorithm

In this section, the analytical model of the CRRM algorithm described in detail in section 5.3 is shown. By this point of view, the “fittingness factor” CRRM algorithm can be represented by means of the following relation:

)(,,,,,, NFjsijsijsi QC ηδ××=Ψ [Eq. 1]

where the product jsijsi QC ,,,, × returns a boolean value with the aim to decide if a request can be assigned to a specific RAT or not, while the availability of resources over this RAT is another condition to consider in order accepting a new connection.

In general, the logical process to choose a generic RAT is structured as following:

Arrival process (request of a connection)related to specific mobile terminal

for a specific service

Individuation of the subset of RATssatisfying, for each RAT, following criteria:

• it has available resources• it is compatible with requested service• it is compatible with mobile terminal

capabilities requesting the connection

Selection of the RAT characterised by themaximum value of the parameter δ(ηNF). Ifthis value is associated to more than oneRAT, the selection is performed according

to a predefined set of priority level

Arrival process (request of a connection)related to specific mobile terminal

for a specific service







Selection of the RAT characterised by themaximum value of the parameter δ(ηNF). Ifthis value is associated to more than oneRAT, the selection is performed according

to a predefined set of priority level

Figure 8-1 – Global process for RAT selection phase

The approach developed in order to evaluate the “fittingness factor” algorithm, is based on Markov Chain theory, as explained in the following paragraphs.

8.1.1 State characterization and Space State dimension

The generic state of the model can be described by the following set of state variables:

{ }HSDPAd

NFRd

FRd

Rvi

Rvo

GSMvo NNNNNN ,,,,, 99 999999

Where the meaning of each variable is the following:

• [ ]CAPGSMvo GSM0N ,∈ - number of voice calls (both flexible and non-flexible types) allocated on the

GSM cell resources;

• [ ]9999 , Rvo

Rvo K0N ∈ - number of voice calls (only flexible type) allocated on the R99 cell resources;



• [ ]9999 , Rvi

Rvi K0N ∈ - number of video calls (only non-flexible type) allocated on the R99 cell resources;

• [ ]FRd

FRd K0N 9999 ,∈ - number of data connections (WWW service from flexible users) allocated on the

R99 cell resources;

• [ ]NFRd

NFRd K0N 9999 ,∈ - number of data connections (WWW service from non-flexible users) allocated

on the R99 cell resources;

• [ ]HSDPAd

HSDPAd K0N ,∈ - number of data connections (WWW service from flexible users) allocated on

the HSDPA cell resources.

Parameter CAPGSM corresponds to the maximum number of voice calls that can be managed by GSM RAT; it depends on the number of time slot (TCH channels) available in the cell. Parameters related to the capacity of the R99/HSDPA cell ( 99R

voK , 99RviK , FR

dK 99 , NFR

dK 99 and HSDPA

dK ) depend on the capacity region, strictly related both to the number of user of each service allocated on the R99 cell and to the current per user throughput over HSDPA cell.

8.1.2 Set of Events

The events modeled by Markov Chain are summarized as follows:

• Request (birth) of a voice call from single mode terminal (GSM only);

• Request (birth) of a voice call from dual mode terminal (GSM/R99);

• Request (birth) of a video call from terminal supporting only R99 RAT;

• Request (birth) of a WWW session from dual mode (R99/HSDPA);

• Request (birth) of a WWW session single mode terminal (R99 only);

• End (death) of a voice call over GSM RAT;

• End (death) of a voice call over R99 RAT;

• End (death) of a video call over R99 RAT;

• End (death) of a flexible WWW session over R99 RAT;

• End (death) of a non-flexible WWW session over R99 RAT;

• End (death) of flexible WWW session over HSDPA RAT.

The evolution of the system depends on the choices taken by the CRRM algorithms when a new connection request arrives. This kind of evolution can be modeled by a-continuous time Markov chain, where the transitions from a generic state Aσ to a generic state Bσ occur as the following events happen:

Birth events

- voice call request accepted on GSM resources:

{ }HSDPAd

NFRd

FRd

Rvi

Rvo

GSMvoA NNNNNN ,,,,, 99 999999=σ → { }HSDPA

dNFR

dFR

dRvi

Rvo

GSMvoB NNNNNN ,,,,,1 99 999999+=σ

- voice call request accepted on R99 resources:

{ }HSDPAd

NFRd

FRd

Rvi

Rvo


dNFR

dFR

dRvi

Rvo

GSMvoB NNNNNN ,,,,1, 99 999999 +=σ



- video call request accepted on R99 resources:

{ }HSDPAd

NFRd

FRd

Rvi

Rvo


dNFR

dFR

dRvi

Rvo

GSMvoB NNNNNN ,,,1,, 99 999999 +=σ

- WWW connection request (flexible) accepted on R99 resources:

{ }HSDPAd

NFRd

FRd

Rvi

Rvo


dNFR

dFR

dRvi

Rvo

GSMvoB NNNNNN ,,1,,, 99 999999 +=σ

- WWW connection request (non-flexible) accepted on R99 resources:

{ }HSDPAd

NFRd

FRd

Rvi

Rvo


dNFR

dFR

dRvi

Rvo

GSMvoB NNNNNN ,1,,,, 99 999999 +=σ

- WWW connection request accepted on HSDPA:

{ }HSDPAd

NFRd

FRd

Rvi

Rvo

GSMvoA NNNNNN ,,,,, 99 999999=σ → { }1,,,,, 99 999999 += HSDPA

dNFR

dFR

dRvi

Rvo

GSMvoB NNNNNNσ

Death events

- end of a voice call on GSM resources:

{ }HSDPAd

NFRd

FRd

Rvi

Rvo


dNFR

dFR

dRvi

Rvo

GSMvoB NNNNNN ,,,,,1 99 999999−=σ

- end of a voice call on R99 resources:

{ }HSDPAd

NFRd

FRd

Rvi

Rvo


dNFR

dFR

dRvi

Rvo

GSMvoB NNNNNN ,,,,1, 99 999999 −=σ

- end of a video call on R99 resources:

{ }HSDPAd

NFRd

FRd

Rvi

Rvo


dNFR

dFR

dRvi

Rvo

GSMvoB NNNNNN ,,,1,, 99 999999 −=σ

- end of a WWW connection (flexible) on R99 resources:

{ }HSDPAd

NFRd

FRd

Rvi

Rvo


dNFR

dFR

dRvi

Rvo

GSMvoB NNNNNN ,,1,,, 99 999999 −=σ

- end of a WWW connection (non-flexible) on R99 resources:

{ }HSDPAd

NFRd

FRd

Rvi

Rvo


dNFR

dFR

dRvi

Rvo

GSMvoB NNNNNN ,1,,,, 99 999999 −=σ

- end of a WWW connection on HSDPA:

{ }HSDPAd

NFRd

FRd

Rvi

Rvo

GSMvoA NNNNNN ,,,,, 99 999999=σ → { }1,,,,, 99 999999 −= HSDPA

dNFR

dFR

dRvi

Rvo

GSMvoB NNNNNNσ

In order to model the CRRM algorithm different parameters should be calculated, for example to find the probability that a mobile terminal has R99 coverage for a specific service. This probability is evaluated as shown below:

[ ] [ ]SOKR99PavailableR99P =

[Eq. 2]

Where S is related to different services (voice, video, data).

Notice that to find such probability, the mobile terminal should be covered by that RAT (i.e. R99 or HSDPA) for the specific service in both uplink and downlink sections. So we assume that:

[ ] [ ] [ ]( )SOK

SOK

SOK DLPULPRP ,min99 =

[Eq. 3]



Where:

[ ]SOKULP Is the probability that R99 RAT cover the mobile terminal in the uplink for service S,

[ ]SOKDLP Is the probability that R99 RAT cover the mobile terminal in the downlink for service S.

Finding such parameter depending on the following formula:

[ ]GSM

RSOK L

LULP

max_

99max_=

[Eq. 4]

Where:

GSMLmax_ Is the maximum allowed path loss in the GSM RAT (it is an input of the algorithm),

99max_RL Is the maximum path loss in R99 RAT; it can be calculated by means of the following formula:

( )UL

ibi

bN

TR 11

RNEW

PPL η−

+

=

,0

max99_max

[Eq. 5]

To find the uplink load factor, which depends on the current state of the model, it is possible to use the following formula:

( ) ( )f1WNE

Ri

ibiiUL +⋅⋅= ∑ 0ν

η

[Eq. 6] It is important to notice that the formula used to calculate the uplink load factor is a summation depending on the number of active users in the R99 RAT. The number of voice, video and WWW users allocated in the cells is strictly related with the currents state σ, as shown below:

{ }HSDPAd

NFRd

FRd

Rvi

Rvo

GSMvo NNNNNN ,,,,, 99999999 .

As a consequence of this, the uplink load factor can be expressed as:



( )

( )

( )

+⋅

⋅+

+⋅

⋅+

+⋅

⋅=

ULdata

bULdata

databRd

ULvideo

bULvideo

videobRvi

ULvoice

bULvoice

voicebRvoUL

fWNE

RN

fWNE

RN

fWNE

RN

1

1

1

0,

99

0,

99

0,

99

υ

υ

υη

[Eq. 7]

Where:

NFRd

FRd

Rd NNN 999999 +=

[Eq. 8]

As a result of the previous calculation 99max_RL can be obtained, which depends on the current state. On the other hand, we can find out the probability that R99 RAT covers the mobile terminal in the

downlink direction for Sth service, [ ]SOKDLP . This evaluation is performed as follow:

[ ] [ ]TotDLS

SOK PPPPDLP −≤= max

[Eq. 9]

Where:

DLSP Is the power needed to accept a new request of Sth service; it depends on the path loss;

maxP Is the maximum available power in the downlink section; it constitutes an input;

TotP Is the total transmitted power in the downlink section, it depends on the state.

Total power is calculated by means of formula:

( )∑ ⋅++=usersR

DLS

DLS

RSCCHTot PPNPP

99max

99 λ

[Eq. 10]

Where:

CCHP Is the common channel power,

DLSP Is the average downlink power per each service,



The term DLSP is expressed by the formula:

( ) NDL

RSHDLS

DLS

DLS

PLKP ..1.. 99max_+= χλ

99R

SN Is the number of services [Eq. 11]

Where the set of parameters DLSλ , DL

Sχ , SHK and NP represents an input of the model.

Notice that in general, DLRL 99max_ should be an averaged value but in our study we consider only the

worst case.

The term DLSλ is expressed by the formula:

( )( )DLSH

DLS

DLS

DLS fK +−+= αχγλ 1.1..

[Eq. 12]

Where DLf is the downlink inter-cell over intra-cell interference ratio that can be expressed by the relation:

Gf DL 1

=

[Eq. 13]

and G is the geometrical factor (G and α are given in input to the model).

With the aim to evaluate the power needed to accept a new request of Sth service (it depends on the current path loss), it is possible to use the following formula:

( ) NDLRSH

DLS

DLS

DLS PLKP ..1.. 99+= χγ

[Eq. 14]

It is important to notice that DLSP is similar to DL

SP , where in case of DLSP the path loss is not averaged.

The term DLSγ is expressed by the formula:

( )DL

S

bDLsb

DL

S

bDLsb

DLS

NERw

NER

−+

=

0,

0,

..1

.

α

γ

[Eq. 15].

To find out the probability

[ ] [ ]TotDLS

SOK PPPPDLP −≤= max

[Eq. 16]

we can notice that the condition TotDLS PPP −≤ max depends on the current state. Relating to the previous

formulas we can re-write the condition as:

( ) TotNDLRSH

DLS

DLS PPPLK −≤+ max99..1..χγ



[Eq. 17].

In this case, we can write the path loss DLRL 99 as:

( ) NSHDLS

DLS

TotDLR PK

PPL.1..

max99 +

−≤

χγ

[Eq. 18]

Assuming that

[ ]GSMDLR LL max_99 ,0∈

[Eq. 19]

so

[ ] [ ] ( ) GSMNSHDLS

DLS

TotTot

DLS

SOK LPK

PPPPPPDLPmax_

maxmax

1..1.. +

−=−≤=

χγ

[Eq. 20]

with values of [ ]SOKDLP and [ ]S

OKDLP comprise between 0 and 1.

Network-centric suitability ( )NFηδ is a function that reduces the fittingness factor of flexible traffic depending on the amount of non-flexible load; it captures the suitability from an overall RAT perspective and depends on the non-flexible load NFη ; the first proposed empirical definition of this function was the following (already mentioned in the previous chapter):

( )( )

( ) ( )

−>

−

−<

=flexibleistrafficANDDif

D

flexiblenonistrafficORDif

NFNF

NF

NF ,min12

,min1

_,min11

ηηηη

ηηηδ

[Eq. 21]

A new and more evolved formulation introduces the parameter b that permits to well conform the shape of the function ( )NFηδ :

( )( )

( ) ( )

⋅−>

⋅−

⋅−<

=flexibleistrafficANDDbif

Db

flexiblenonistrafficORDbif

NFNF

NF

NFev

,min12

,min1

_,min1 1

ηηηη

ηηηδ

[Eq. 22]

When b = 1, formula [Eq. 22]collapses in the first version [Eq. 21]], as illustrated in Figure 8-2



Delta for b = 1

0

0.2

0.4

0.6

0.8

1

1.2

0 0.2 0.4 0.6 0.8 1

Figure 8-2 - ( )NFηδ for b = 1

A different value (b → ∞) implies the diagram shown in Figure 8-3.

Delta for b = 15

0

0.2

0.4

0.6

0.8

1

1.2

0 0.2 0.4 0.6 0.8 1

Figure 8-3 - ( )NFηδ for b = 15

It is worth to notice that HSDPA RAT does not have non-flexible load, so we consider ( )NFηδ always equal to one. In our application, we set the value of parameter b in order to drive new data connection requests to use HSDPA resources, if mobile terminal supports this kind of cell. Besides this, we consider parameter D of the Fittingness Factor CRRM algorithm equal to one, because of this choice permits to apply the algorithm with a hard policy.

Also in case of GSM which may contain non-flexible load, we can use the following formulas to find ( )NFηδ value. It depends on the current state and on the condition that R99 cell covers or not the

mobile terminal.

In order to calculate ( )NFηδ for GSM cell, it is necessary to evaluate the average normalized load, expressed by the formula:

GSMCAP

GSMvoN=η

[Eq. 23]

To estimate the non-flexible load, we can notice that, in order to have non-flexible load over GSM, the generic terminal should be characterized either by single mode capabilities (GSM-only) with probability P or by an inadequate R99 coverage (in this case the terminal is considered out of R99 coverage) for



voice service, with probability [ ]( )voiceOKRP 99\1− . In this case, the non-flexible load in the GSM can be

expressed as:

[ ]( )( )voiceOKNF RPP 991−=ηη

[Eq. 24]

To achieve such out of coverage probability, the term [ ]voiceOKRP 99 should be equal to zero, so we can

model the non-flexible load over GSM as shown below:

PNFηη =

[Eq. 25]

Using the formula explained above, we can find ( )GSMNFηδ . The calculations illustrated now, are valid for

both uplink and downlink sections, because of GSM has symmetric connections, in which each user occupies one time slot in both sections.

R99 RAT, instead, can contains non-flexible load and the evaluations depend on the following formulas in order to find the parameter ( )NFηδ , depending on the current state. By a general point of view, R99 cell has asymmetric connections, so it is required to distinguish between network suitability in the uplink or downlink section and the final suitability in both directions can be estimated based on the suitability in each section. By the downlink side point of view, DL

R99δ can be defined as the network suitability in the downlink direction for R99 RAT. Its evaluation depends on the current averaged normalized load and the non-flexible load in the R99 RAT.

The average normalized load DLη in the downlink section can be expressed as the ratio between the total transmitted power in the downlink TotP , which depends on the state, and the maximum available power in the downlink maxP , which is an input of the model. This load can be calculated as shown in the following formula:

maxPPTotDL =η

[Eq. 26].

In order to find the non-flexible load in the downlink direction for R99 RAT, we can observe that in our scenario, we have non-flexible load either when the request for data session (WWW) coming from a mobile terminal supporting only R99 RAT, or when the request concerns a video call, because of video service can only be carried by over R99 resource in the considered scenario. In this context, the non-flexible load in the downlink direction of R99 cell, can be expressed as the ratio between the total power used by common channel and non-flexible users, depending on the state, and the maximum available power in the downlink section maxP .

The non-flexible load DLNFη is evaluated by means the relation:

( ) ( )max

max99

max99 ....

PPPNPPNP DL

dDLd

NFRd

DLvi

DLvi

RviCCHDL

NFλλη ++++

=

[Eq. 27]

where:

DLviP is the average power needed for a video call carried in the downlink section,



DLdP is the average power needed for a data session carried in the downlink section,

DLviλ is the average arrival rate of video calls in the downlink section,

DLdλ is the average arrival rate of non-flexible data sessions in the downlink section.

The terms DLviλ and DL

dλ are evaluated with the formulas [Eq. 28] and [Eq. 29]:

vi

viDLvi T

a=λ

[Eq. 28]

where:

via is the video traffic in the cell, which is given as an input in Erlang unit;

viT is the average duration of video call,

qTa

d

dDLd .=λ

[Eq. 29]

where:

q is the probability that terminal can support R99 RAT,

da is the data traffic in the cell, which is given as an input in Erlang unit,

dT is the average duration of a data session.

We can estimate the network centric suitability in the uplink section by estimating the flexible and non-flexible load in the uplink direction. Flexible load depends on the state and it can be calculated as shown in [Eq. 7]. As a function of the definition of the non-flexible load in R99, we can calculate UL

NFη as shown below:

( )

( )ULfWNobEU

datadatabR

FNRdN

fWNE

RN

data

ULvideo

bULvideo

videobRvi

ULNF

+⋅

⋅+

+⋅

⋅=

1,99

10,

99

υ

υη

[Eq. 30].

Applying equations shown above, we can find the network centric suitability in both uplink and downlink section, but – as mentioned – network suitability in R99 depends on both uplink and downlink direction. So we merge the suitability of the two directions in only one parameter, with the aim to reflect the general suitability of R99 cell; the global parameter is evaluated as follows:



( ) LDR

ULRR 999999 1 δϑδϑδ −+=

[Eq. 31]

where ϑ represents the weight of the uplink section with respect to the downlink section in R99 cell; it constitutes an input of the model.

As far as HSDPA per user throughput is concerned, HSDPA RAT can support only data service (WWW sessions), where the RAB performances are variable. They depend on the total available power for HSDPA connections, as underlined by the following relation (always mentioned in a previous chapter):

99_max_ RTTHSDPAT PPP −= .

Now, we can evaluate the throughput per user in HSDPA context, based on available power and current SNR (Signal to Noise Ratio), depending on the current state:

( )HSDPAd

HSDPA NSNRfTh =

[Eq. 32]

where ( )SNRf is a function that returns the total available throughput in HSDPA based on the current SNR. This function can be implemented as an input table.

The parameter SNR can be estimated as follows:

GIE

IE

SNR

C

C

C

C

11 +−=

[Eq. 33]

where:

G is the geometrical factor. It represents an input of the model.

C

CI

E is the ratio between chip energy and interference energy. It is calculated using the formula:

maxP

PI

E HSDSCH

C

C =

[Eq. 34]

where:

HSDSCHP is the power of high speed dedicated channel. It is calculated as:

( )TotTot

CCHHSDSCH PP

PPP −= max.

[Eq. 35]

and

CCHP is the common channel power; it is an input of the model



maxP is the maximum available power in the downlink section; it is an input

TotP is the total transmitted power in the downlink section, depending on the state; it can be calculated as shown in [Eq. 10].

By the Admission Control point of view, the capacity of a generic RAT to accept a new request for a specific service S (voice, video, WWW) can be denoted as the availability of that RAT. For example, when GSM resources are available, one or more time slots are usable to allocated a new voice user. This condition can be expressed by means of the formula:

≤

elseNoGSM

GSMvo

available

CAPNifyesGSM

[Eq. 36]

In a R99 cell, three conditions are checked to accept a new session. In more detail, the uplink load factor after user acceptance for a specific service must be lower than a given uplink threshold, the downlink load factor after user acceptance for a specific service must be lower than a given downlink threshold and there must be a spreading code sequences available.

The availability of R99 to accept new request for a generic Sth service can be expressed as:

≤

≤

availablecodesspreadingDLarethere

ANDif

ANDif

R99 DLSDL

ULSUL

savailable

max

max

ηηηη

[Eq. 37]

Based on the new state characteristics after accepting the new service request, the uplink load factor sULη can be calculated as shown in [Eq. 7]. Same consideration can be done for the downlink load

factor sDLη that can be calculated as the ratio between the total transmitted power in the downlink

section TotP after accepting the new request, and the maximum power maxP in the same section (it is

an input of the model):

max

TotsDL P

P=η

[Eq. 38].

Based on the new state characteristics after accepting the new service request, TotP can be calculated using formula [Eq. 10]. Also the availability of a spreading code sequence can be checked based on the condition of the cell, depending on both the new state after accepting the new request and the spreading factor associated to the service requested by the new connection. This condition can be expressed as follows:

( ) ( ) ( ) max128128

9999128

99128

99 SFSFNNSFNSFN viFRd

FRd

viRvi

voRvo ≤⋅++⋅+⋅

[Eq. 39]

where:

voSF128 is the spreading factor needed for one voice call over R99 resources; it is an input

viSF128 is the spreading factor needed for one video call over R99 resources; it is an input



viSF128 is the spreading factor needed for one data session over R99 resources; it is an input

max128SF is the maximum number of available spreading factor over R99 resources; it is an input



9 ANNEX B: DIMENSIONING MODEL FOR THE MOBILE-TV SCENARIO

9.1 Traffic based dimensioning model

Results of the techno-economic evaluation reported in section 6.7 have been derived by applying a specific dimensioning model on the scenario described in previous section. The dimensioning model has been defined by extending the one already implemented for the previous techno-economic activities.

The original model was designed to derive only the number of UTRAN cells (macro-cells or micro-cells) needed to support the considered traffic, on the basis of the coverage extension of each cell and the capacity requirements. Instead, for the present analysis, before assuming investments for the introduction of new base stations in the area of interest, it is assumed that the already existing base stations can be upgraded by activating the second UTRAN carrier in each cell.

According to this approach, the two outputs of the model for the entire scenario are the following:

• cellpercarriers __ : that represents the mean number of UTRAN carriers supported by the cells of the scenario14 ;

• nodeBofnumber __ : that represents the number of tri-sectorial nodeB needed to support the assumed traffic15.

With respect to the outputs of the dimensioning model, three different situations can occur:

1.

==

nodeBofnumberinitialnodeBofnumbercellpercarriers

_____1__

when the total amount of traffic can be supported by the initial network deployment (i.e. no investments are needed);

2. [ ]

=∈


_____2;1__

when the total amount of traffic can be supported by activating the second UTRAN carrier in some cells of the scenario (i.e. investments reported in 6.7 related to the introduction of the second carrier have to be taken into account);

3.

≥=


_____2__

when the total amount of traffic can be supported only by activating the second UTRAN carrier in all the cells of the original scenario and by deploying additional sites (i.e. investments reported in 6.7.1, related to the introduction of the second carrier as well as the deployment of new sites have to be taken into account);

In order to evaluate the number of carriers per cell and the number of nodeB needed to support the assumed traffic, the dimensioning model takes into consideration the interference limit in uplink and the transmission power limit in downlink, as described in the following sections.

14 This quantity can range from 1 (i.e. all the cells of the scenario support only one UTRAN carrier) to 2 (i.e. all the cells of the scenario support two UTRAN carriers). 15 This quantity can range from the pre-existing number of node-B assumed in the area of interest to higher values.



9.1.1 Evaluation of the number of second UTRAN carries and new nodeB

9.1.1.1 Uplink

According to the scenario taken into account, in uplink all the services are mapped onto dedicated transport channels (DCHs). WCDMA pole capacity equation (ref. [9]) can be used in order to evaluate the cell load in uplink, when the services are mapped onto DCHs:

MAXs

ULssn ηλ <⋅∑

Where ns denotes the number of connections per cell for every service (“s” pedix) , whereas ULsλ can

be assumed equals to:

( )ULULs

ULs

ULs FAF +⋅⋅= 1γλ

The quantity

ULs

ULs

ULs

ULsUL

s NoEbRWNoEbR

)/()/(

⋅+⋅

=γ

represents the cell load contribution of the individual connection of the user.

In the economic analysis described later on, both voice and data services have been taken into account. Let Tvoice be the offered voice traffic per cell (Erl) and S

dataT be the offered traffic per cell (kbit/s) associated to the data service s. For the voice service, the number of voice connections per cell can be obtained by means of the Erlang-B whereas the number of data connections of service s can be obtained as the ratio between the total amount of traffic per cell associated to the service and the bit rate of each connection:

( )

=

=sUL

SULdata

sdata

voiceblockvoice

RTn

TPBn

/

,

,

Where Pblock represents the per cell blocking probability for the voice service and was assumed of 2%.

The service related parameters used to estimate the capacity in uplink for the services considered in the techno-economic evaluation have been taken from Table 16 and Table 17 of AROMA deliverable D05 [21], and reported here in section 5, whereas the maximum cell load ( MAXη ) has been assumed

equals to 0.7 and the inter-cell/intra-cell interference ratio ( ULF ) is assumed equal to 0.55.

Table 9-1: Parameters for capacity constrain estimation in uplink [21].

UL

Bit Rate kbit/s

Eb/N0 dB AF

Voice Circuit 12.8 5.9 0.6 Video telephony Circuit 64.0 3.5 0.9

Streaming RT Packet 16.0 5.5 0.1 Interactive NRT Packet 64.0 3.5 0.3

Background NRT Packet 64.0 3.5 1.0



9.1.1.2 Downlink

In order to evaluate the number of carriers per cell and the number of nodeB needed to support the assumed traffic, the following logic has been implemented in the dimensioning model for the downlink:

1. IF cellseragepP favailableTOT _cov1 ⋅≤

==


_____1__

2. IF cellserageppPcellseragep favailable

favailableTOT

favailable _cov)(_cov 211 ⋅+<≤⋅

=⋅

⋅−+=

nodeBofnumberinitialnodeBofnumbercellseragepcellseragepP

cellpercarriers favailable

favailableTOT

______cov

_cov1__ 2

1

3. IF cellserageppP favailable

favailableTOT _cov)( 21 ⋅+>

⋅+=

=

cellserageppP

nodeBofnumber

cellpercarriers

favailable

favailable

TOT

_cov)(__

2__

21

Where:

• 3____cov ⋅= nodeBofnumberinitialcellserage :represents the total number of cells with respect to the initial deployment assumed in the are of interest (apart from the support of a specific amount of initial traffic, these cells guarantee the network coverage)

• )(TPP TOTTOT = : represents the total power requested with respect to the total amount of traffic T assumed in the area of interest;

• 111 fcch

fMax

favailable PPp −= : is the power made available by each cell of the scenario in the

UTRAN carrier f1 and it depends on the maximum available cell power and the power reserved for common channels for carrier f1.

• 111 fcch

fMax

favailable PPp −= : is the power made available by each cell of the scenario in the

UTRAN carrier f1 and it depends on the maximum available cell power and the power reserved for common channels for carrier f2 16.

In the two considered scenarios, the mobile TV service is assumed to be offered by means of the HS-DSCH channel (HSDPA case, taken as reference) or by means of the MTCH channel (i.e. MBMS case, to be compared with respect to the reference case). As a consequence, in order to evaluate the total power requested to accommodate the traffic assumed in the area of interest, the following equations which take into account the presence of HSDPA or MBMS have to be considered:

• Mobile TV on HSDPA: )()( HSDPAHSDPADCHDCHTOT TPTPP +=

• Mobile TV on MBMS: MBMSDCHDCHTOT PTPP += )(

16 Since in the evaluations the total cell power available has been assumed of 20 W and the cell power reserved for common channels has been assumed of 4 W for both the two UTRAN carriers, the total power made available be in each cell when two carriers are used is equal to 2(Ptot-Pcch).



where:

• )( DCHDCH TP : represents the total downlink power requested to accommodate the total traffic allocated onto dedicated channels in the are of interest (ref. 9.1.1.3);

• )( HSDPAHSDPA TP : represents the total power requested to accommodate mobile TV users o HSDPA (ref. 9.1.1.4 for details);

• MBMSP : represents the total power requested to accommodate mobile TV users on MBMS (ref. 9.1.1.5 for details).

As explained with more details in 9.1.1.4 and 9.1.1.5, it is worth noting that when mobile TV users are allocated on HSDPA, the power that has to be allocated for the HS-DSCH channel of every cell depends on the total amount of HSDPA traffic (THSDPA) offered by these subscribers in the are of interest. On the other hand, when mobile TV users are allocated on MBMS, taking advantage of the point-to-multipoint transmission, the power for the MTCH channel of every cell does not depends on the traffic and it is determined only on the basis of coverage considerations. As a consequence, starting from a specific amount of mobile TV subscribers, the MBMS solution become profitable with respect to the case where mobile TV users are allocated on HSDPA, as it will be described with more details in section 6.8 devoted to report the results of the carried out techno-economic evaluation.

9.1.1.3 Downlink transmission power requested for dedicated channels in downlink

In downlink, it is useful to consider the following approximated equation for the evaluation of the mean power transmitted by each cell on the dedicated channels:

∑ ⋅+⋅+=−

s

DLsMAX

DLssCCH

icellDCH PpnPP )( λ

Where ns denotes the number of connections per cell for every service (“s” pedix) , whereas DLsp and

DLsλ can be assumed equals to:

+−⋅+⋅⋅=

⋅⋅+⋅⋅=

)1()1(

)1(

DLSHDLs

DLs

DLs

NoiseSHDLs

DLs

DLs

iKAF

PLKAFp

αγλ

γ

And

( )( )DLsDL

s

DLs

DLsDL

s NoEbRWNoEbR

/)1(/⋅⋅−+

⋅=

αγ

System and cell level parameters used to estimate the power requested by DCH channels in downlink are reported in Table 9-2whereas service based parameters used to estimate the power requested by DCH channels in downlink are reported in Table 9-3:

Table 9-2: System and cell level parameters for DCH power estimation in downlink.

Parameter Symbol Value

Downlink cell total power [dBm] PMAX 43

Inter-cell interference factor iDL 0.7

Ortogonality factor α 0.8

Mean path loss [dB] L 100



% SH overhead KSH 30%

Common channels TX power [dBm] PCCH 34

Thermal noise power [dBm] Pnoise -100

Table 9-3: Service level parameters for DCH power estimation in downlink [21] DL

Bit Rate kbit/s

Eb/N0 dB SAF

Voice Circuit 12.8 6.5 0.65 Video

telephony Circuit 64.0 5.3 0.80 Streaming RT Packet 64.0 5.3 1.00 Interactive NRT Packet 384.0 5.2 1.00

Background NRT Packet 64.0 5.3 1.00

. Finally, the total downlink power required to support the total amount of traffic assumed in the area of interest can be easily evaluated as a sum of the power transmitted by each cell on the dedicated channels:

∑=

−=cellsofnum

i

icellDCHDCHDCH PTP

__

1)(

9.1.1.4 Mobile TV on HSDPA: downlink transmission power requested for HS-DSCH The following approach is followed in order to evaluate the total power requested to accommodate mobile TV users over HSDPA:

• Single-user HSDPA throughput that has to be guaranteed in each cell (i.e. total HSDPA throughput offered by each cell) is evaluated starting from the total amount of traffic generated by Mobile TV subscriber in the scenario. This mapping between the input traffic figures over the overall scenario and the input traffic per each cell of the scenario is performed taking into account the different number of cells assumed in three different areas: dense urban, urban and rural, according to the traffic distribution represented in the table below:

Table 9-4 – Traffic distribution among dense urban, urban and rural areas. Traffic distribution Town

DENSE URBAN URBAN RURAL

Referenced Town 0.6 0.4 0.2

• Then, the total power requested to accommodate mobile TV users on HSDPA in the area of interest ( HSDPAP ) is evaluated by estimating the transmission power that should be allocated to S-DSCH in every cell, by taking into consideration the relation depicted in Figure 9-1(ref. [10]) and also reported in Table 9-5.



Figure 9-1 – HSDPA single user cell throughput versus allocated HSDPA power [10].

Table 9-5 – Single user HSDPA cell throughput versus total cell power percentage.

Single user HSDPA cell throughput

(kbps) % cell power

250 8% 400 11% 550 16% 700 20% 800 25% 840 28% 880 29% 920 30% 960 40% 1000 47% 1040 53% 1080 58% 1120 62%

• If the HSDPA throughput to be guaranteed by each cell in the three areas of the scenario exceeds the maximum HSDPA throughput achievable17, then the number of nodeB is increased consequently in order to be able to accommodate all the mobile TV subscribers by means of HSDPA.

9.1.1.5 Mobile TV on MBMS: downlink transmission power requested for S-CCPCH In WP3 of the AROMA project (ref. AROMA deliverable D12 [8], section 4.3.2), a very detailed analysis of the relation between the MBMS transmitted power versus coverage has been carried out,

17 The maximum value reported in the table assumes that five HS-PDSCH codes per cell are allocated and it takes into account the ITU Vehicular A multipath channel profile.



by means of simulations. Moreover, a general analytical framework has been also developed and proposed with the aim of estimating the required total transmitted power in downlink when several services, relying on both p-t-p (i.e. using DCHs) or p-t-m (i.e. using S-CCPCH) transmission modes. This model can be used in order to evaluate the transmission power of S-CCPCH that has to be allocated in every cell, in order to accommodate in an appropriate way the mobile TV service, with respect to the different radio conditions experienced by the users (taken into account by means of the G factor). As a consequence, the results reported in deliverable AROMA deliverable D12 [8], can be fully exploited in order to perform a very accurate dimensioning of the mobile TV service over MBMS. In any case, taking into account the final goal of the dimensioning model (devoted to estimate the total investment required to support the mobile TV service over the next years in a very large area, with the typical extension of an European city), it was decided to dimension the power for MBMS service in a simpler way. According to the followed approach, the figures from section 4.4.7 of [11] (corresponding to vehicular A channel at 3km/h), reported in Table 9-6, have been considered:

Table 9-6 - MBMS power allocation (% with respect to the total power

MBMS power allocation

(% with respect to the total power) No MBMS 64 kbit/s 128 kbit/s 256 kbit/s

0.00% 38.00% 53.7% 60%

Since in the techno-economic analysis the mobile TV service is supposed to be offered by means of a 256 kbit/s channel, the cell power reserved for MBMS corresponds to 60% of the maximum cell power available. As fully investigated in AROMA deliverable D12 [8], according to the number of MBMS subscribers within a cell, p-t-p or p-tm transmission mode can be selected. As already mentioned in section 9.1.1.3, when point-to-multipoint transmission mode is used by the MBMS framework, the percentage of power dedicated to S-DSCPCH is able to offer the mobile TV service to all the users inside a cell. As a consequence, the advantage in terms of radio resources becomes evident only when many cells of the scenario have a mean number of mobile TV users greater than the switching threshold. The techno-economic evaluation reported in section 6.7 is mainly devoted to evaluate when this condition occur, according to the more consolidated traffic forecasts for the next years, also estimating the economic advantage with respect to the investments needed to support mobile TV over HSDPA.



10 REFERENCES [1] Huawei Technologies, PR China – Huawei UMTS1800 Solutions – 2nd Meeting of the APT

Wireless Forum – Shenzen, PR China, September 2005

[2] Harmantzis, Gunasekaran – Cost Analysis of WLAN Integrated with GSM/GPRS Networks – School of Tecnology Management, Hoboken USA 2004

[3] Motorola White Paper – Minimizing the cost of UMTS/HsxPA Networks – July 2007

[4] Nokia White Paper – Business benefits of WCDMA technology – December 2004

[5] A. Barbaresi et alii, “Economic evaluation of legacy IST-EVEREST RRM/CRRM algorithms and solutions” – IST AROMA project D08 deliverable, October 2006

[6] Nortel White Paper – Multiservice Packet Core Consolidation – January 2003

[7] O. Sallent et alii, “First report on AROMA algorithms and simulation results” – IST AROMA project D09 deliverable, October 2006

[8] O. Sallent et alii, “Intermediate report on AROMA algorithms and simulation results” – IST AROMA project D12 deliverable, March 2007

[9] Radio Network Planning and Optimization for UMTS, edited by J. Laiho, A. Wacker and T. Novosad, Wiley, 2002.

[10] Holma, Toskala, “HSDPA/HSUPA for UMTS: High Speed Radio Access for Mobile Communications”, Wiley, 2006.

[11] 3GPP TR 25.803, “S-CCPCH performance for MBMS (Release 6)”, v .6c0,

[12] K. Johansson, “Cost Efficient Provisioning of Wireless Access”, licentiate Thesis in Telecommunications, KTH Stockholm, Sweden 2005.

[13] HiQ data AB, “Potential Value of Additional Spectrum for an UMTS Operator”, Stockholm, Sweden, 2004.

[14] http://news.softpedia.com/news/Mobile-TV-Success-on-the-European-Market-Is-Questioned-57218.shtml

[15] DVB project office official site http://www.dvb-h.org/index.htm

[16] Piia Karhu – “Emerging Mobile Service Innovation Markets: The Case of the Finnish Mobile TV Service Market” – Dissertation of the University of St. Gallen, January 2007

[17] Tuukka Autio - "Broadcast Mobile Television Service In Finland: A Techno-Economic Analysis" - Thesis form HELSINKI UNIVERSITY OF TECHNOLOGY - March 2007

[18] Per Wirdemark – Report Coverage Requirements for UMTS - HiQ Data AB, September 2004

[19] Hannu Verkasalo - TKK COIN - FINNISH SMARTPHONE STUDY 2006, Helsinki University of Technology, December 2006

[20] Heikki Hämmäinen, Antero Kivi, Hannu Verkasalo - Usage Patterns Of Mobile Subscribers: Regulatory Observations In Finland - Helsinki University of Technology, February 2006

[21] Rickard Ljung and Anders Dahlen – “Target Scenarios specification: vision at project stage 1 “ – IST AROMA project D05 deliverable, March 2006

[22] http://informitv.com/articles/2007/03/29/bbcmobiletv/

[23] http://www.bbc.co.uk/pressoffice/pressreleases/stories/2007/03_march/29/3g.shtml



[24] http://www.mobiletelevisionreport.com/businessvalue_proposition_reiters_mobile_tv_report/index.html

[25] http://mobileanalystwatch.blogspot.com/2006_05_01_archive.html

[26] Bass, Frank (1969). "A new product growth model for consumer durables". Management Science 15 (5): p215-227.

[27] Quotient associates – Spectrum Pricing in the Nederland, Final Report, March 2004

[28] IST-TONIC, Deliverable #11, "Final results on seamless mobile IP service provision economics”, Oct. 2002

[29] http://nordicwirelesswatch.com/wireless/story.html?story_id=2296

[30] J. Björkdahl, E. Bohlin, "Competition Policy and Scenarios for European 3G Markets", Chalmers University of Technology, Göteborg, Sweden, http://www.idate.fr/fic/revue_telech/24/bjorkdahl.pdf

[31] http://www.wirelessbroadbandsystem.fi/bbstats.htm

[32] UMTS forum white paper, “3G/UMTS Evolution: towards a new generation of broadband mobile services”, December 2006.

[33] A. Kivi, "Mobile Data Service Usage Measurements", Helsinki University of Technology Networking Laboratory, 2.4.2007

[34] http://europa.eu/rapid/pressReleasesAction.do?reference=IP/07/1118&format=HTML&aged=0&language=EN&guiLanguage=en

[35] IPDC Forum/HPI Research – Sept 2003

[36] Milan Polytechnic – School of Management – Proceedings of conference on Mobile & Wireless Business – July, 2007

[37] iMOBICOM 2007 – Proceedings of the Conference

[38] Joakim Bjorkdahl & Erik Bohlin, IMIT & Chalmers - 3G Network Investments in Sweden – Study for Post & Telestyren – February 2003

[39] Viviane Reding - Television is going Mobile – and needs a pan European policy approach - International CeBIT Summit - Hannover, Germany, 8 March 2006

[40] Commission of the European Communities – Strengthening the Internal Market for Mobile TV – Impact Assessment – July 2007

[41] J. WANG, "Will WiMAX+WLAN Constitute a Substitute to 3G? – A Techno-Economic Case Study", Master’s Degree Project Stockholm, Sweden March 2004



ACRONYMS CAPEX CAPital Expenditures CRRM Common Radio Resource Management DCH Data CHannel DVB-H Digital Video Broadcasting Handheld EDGE Enhanced Data rates for GSM Evolution GERAN GSM EDGE Radio Access Network HSDPA High Speed Downlink Packet Access KPI Key Performance Indicators MBMS Multimedia Broadcast Multicast Service MPLS MultiProtocol Label Switching NPV Net Present Value OPEX Operation EXpenditures RRM Radio Resource Management S-MDB Satellite Digital Multimedia Broadcasting T-DBM Terrestrial Digital Multimedia Broadcasting UMTS Universal Mobile Radio Access Network UTRAN UMTS Terrestrial Radio Access Network VAS Value Added Service

aroma ist-4-027567 d17 economic evaluation of novel aroma ... · in mbms case study, the target of...

Documents