ScienceDirect Performance Evaluation of Femtocells

Available online at www.sciencedirect.com

ScienceDirect Procedia Technology 17 (2014) 683 – 691

Conference on Electronics, Telecommunications and Computers – CETC 2013

Performance Evaluation of Femtocells Usage on LTE Ana Gagoa, Nuno Cotaa* a

Área Departamental de Engenharia Electrónica e Telecomunicações e de Computadores - ADEETC Instituto Superior de Engenharia de Lisboa - ISEL, Lisboa, Portugal

Abstract Nowadays femtocells are being used by the operators as a solution to increase the network capacity by use them in shadow zones and for traffic flow increasing from the cells of the higher layer. Therefore, the main goal of this article is to present a study that quantifies and qualifies the performance of this solution and presents a conclusion about the advantages introduced from a network performance and financial point of view. © © 2014 2014 The The Authors. Authors. Published Published by by Elsevier ElsevierLtd. Ltd.This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/). Selection and peer-review under responsibility of ISEL – Instituto Superior de Engenharia de Lisboa. Peer-review under responsibility of ISEL – Instituto Superior de Engenharia de Lisboa, Lisbon, PORTUGAL. Keywords: LTE, Femtocells, Performance, Simulation;

1.

Introduction

The demand for wideband services by the users has been continuously and exponentially increasing, leading the mobile operators to constantly invest in networks that support high capacities. Thus, appeared the LTE – Long Term Evolution, technology that offers larger capacities, until 300 Mbps in downlink and 75 Mbps in uplink [1] [2], greater spectral efficiency, lower latency, higher levels of service quality and infrastructures more optimized and cheaper. However, the continued growth in the number of mobile users lead to some problems related to the network coverage. Therefore, new challenges began to exist for this technology in indoor environments given that, together with the introduced penetration losses, the conventional propagation models can hardly satisfy the demand for large transmission rates.

* E-mail address: [email protected], [email protected]

2212-0173 © 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/). Peer-review under responsibility of ISEL – Instituto Superior de Engenharia de Lisboa, Lisbon, PORTUGAL. doi:10.1016/j.protcy.2014.10.210

684

Ana Gago and Nuno Cota / Procedia Technology 17 (2014) 683 – 691

This way the mobile operators have adopted the femtocells solution in order to increase the network capacity, thus fulfilling the needs of the users, and to leak the traffic of the existing cells. Yet this solution introduces some network interference issues, thus jeopardizing the global system performance and limiting the capacity of some cells because of the reuse of radio resources. In this work, a network capacity analysis will be performed considering the ICIC - Inter-Cell Interference Coordination [1], interference techniques, and an economical analysis in order to conclude whether this femtocell solution presents advantages when compared to the existing networks. 2.

Vienna System Level Simulator

In order to attain the purpose of this work, it is necessary to conduct simulation tests on the defined scenarios and to optimize the already developed within the femtocells and the LTE subject. Therefore, the Vienna System Level Simulator [3] has been chosen in order to carry out some tests, at the system level, such as to the schedulers that will be used, to the mobility of the users and to the management of interference. The reasons that lead to this choice are related to the fact that this simulator is freely-available for academic research, so there is a closer cooperation between different universities and research centers, and that it has the necessary functionalities for the present study. Another advantage of this simulator is that it offers great flexibility in terms of implementation since its construction is based on object-oriented programming. Its code is modular and it has a clear structure based on objects which results in a more organized code, perceptible and simple maintenance thus making easier the insertion of a new code. In the Figure 1 it is visible the schematic block of the Vienna System Level Simulator. So, it is possible to observe that the part corresponding to the core is divided into two main blocks: the link measurement model and the link performance model. It is important to mention that the physical layer is not included as this makes it easier in some implementations of certain models. The link measurement model abstracts the measured link quality used for link adaptation and resource allocation. On the other hand the link performance model determines the link BLER - Block Error Rate, at reduced complexity. Network layout

Mobility management

Interference structure

Macro-scale fading Antenna gain Shadow fading

Traffic model Link-measurement model

Resource scheduling strategy

Small scale fading

Precoding QoS Link adaptation strategy Power allocation strategy

Link-performance model Retransmission handling/HARQ

Throughput

Error ratios

Error distribution

Figure 1 - Schematic block of the Vienna System Level Simulator in [3].

685


At the end of the simulation graphics of the ROI – Region of Interest, are obtained, composed by the eNodeBs and its coverage area, with the network users, with SINR - Signal to Interference Noise Ratio, and pathloss and with the network throughputs and errors. 3.

Methodology and Assumptions

3.1. Evaluation Scenarios In order to make some conclusions about the advantages of the femtocells usage, it was made a study of the network performance in three different scenarios. So, it was considered an only macrocellular deployment, an only femtocellular deployment and a joint macro and femtocellular deployment. For each one of this scenario it was assumed the use of 20MHz of spectrum; 2.6 GHz; 4x4 transmission mode; 20dB of wall losses for indoor solutions; tests were made for 2000, 4000 and 6000 users, with and without mobility; suburban and urban environments; and two interference rings. For suburban environments it was considered a range of 750m, in an urban macro environment a range of 650m and, for an urban micro environment a range of 500m; and for femtocells a coverage of 20m. 3.2. Economic Structure After a network performance analysis, it was made an analysis of the operator costs for each one of the three network solutions through the capital expenditure – CAPEX, and the operational expenditure – OPEX. To know the viability of the solution and the recovery period of the investment, it was necessary to study the NPV – Net Present Value, and Payback values respectively. All the financial analysis was made based on the values represented in Table 1 that are presented in [4], [5] e [6]. The assumptions for the analysis realized were that the investment was done with an interest rate of 7%, a loan for 10 years and an OPEX growth rate of 3%. Table 1. Network Elements Cost. Description

Cost

Units

Fast Ethernet lines 100Mbps (with enodeB and S-GW co-located)

2.40

k€/year

Fast Ethernet lines 100Mbps

3.60

k€/year

Fast Ethernet lines 1Gbps (with enodeB and S-GW co-located)

3.60

k€/year

Fast Ethernet lines 1Gbps

4.80

k€/year

Backhaul equipment

146.63

k€/year

Core network

85.63

k€/year

EnodeB, site acquisition and construction

108.00

k€/year

Access point femtocell

0.15

k€

Maintenance cost as a fraction of CAPEX

0.12xCAPEX

%/year

Site lease

14.40

k€/year

 Macro Deployment Costs , and the To calculate the CAPEX of a macrocellular solution it was considered the cost of the lines, � backhaul equipment, � ; the cost of all the necessary equipment in the network, � ; and the cost of the website acquisition and construction, � . In terms of OPEX, it was considered the annual cost of the website maintenance and the annual cost of managing it, � . Another parcel to be accounted is the cost of administrative, selling and general, � & , but since it was not possible to obtain its value, its existence was excluded.

686


Thus, the total amount of the capital invested in the macrocell solution is given by the equation (1) where n represents the number of enodeBs used. � = � +� +� +� +� ×� +� (1) +� � +� &  Femto Deployment Costs

Regarding the calculation of the total capital invested, this is also achieved using the sum of CAPEX and OPEX, but some of the parcels mentioned above will cease to exist. In this solution, the agreement made with the operator contributes to the change of the network costs. According to [4], in the initial agreement the femtocells are subsidized by the operator while in the second agreement, the initial investment is made only by the user. In the present study, was considered the second position thereby causing that the cost of the home cell is not included in the CAPEX value. Another difference in CAPEX is that it will be added the value of the core equipment needed to support the femtocells - security and home enodeB gateway and, removed the parcel of the costs of the acquisition and construction of the site. Since the femtocell becomes the responsability of the user, the site maintenance costs and the SG&A costs will not be considered in OPEX and will only be supported by the user. Thus, the total and final amount of the capital invested, is given by the equation (2), where n represents the total number of femtocells in the network. �

= �

 Hybrid Deployment Costs

+�

+�

+�

+�

×� +�×�

&

(2)

In terms of the final cost of a hybrid network, this is calculated using the following equations (1) and (2), representatives of the capital invested in a network macro and femtocell respectively. Thus, in equation (1) it is possible to observe the final expression which gives the value of the capital invested in a hybrid solution where n represents the number of enodeBs and m the number of enodeBs and femtocells. � =� +� +� +� +� ×�+� +� (3) × � +� +� & ×�

687


4.

Results and Discussion

4.1. Network Performance In this section is going to be presented and explained the results obtained by the system level simulator. In order to perform the analysis of the network performance, it is necessary to verify which are the interference values; the SINR values and the CQI - Channel Quality Indicator, values accomplished, considering the fading of the channel; which are the capacities acquired considering those variables and which are the values of the network spectral efficiency. The results shown below are for a macrocellular network; with 6000 users, with mobility, since they represent a bigger demand to the level of resources; and with the traffic model fullbuffer where they are seen to be transferred as much data as possible during the simulation time. Thus in the left figure of the Figure 2, it is possible to observe the values of SINR achieved in each eNodeB throughout the network, taking into account the fading channel, and in the figure to the right the corresponding CQI values. The level of SINR achieved with the eNodeB are better than levels at the border of the same, since this is where there are larger and smaller power values and minor incidence of interference caused by terrain obstacles. The CQI mapping is performed depending on the value of the SINR for each user hence the higher values of CQI next to the eNodeB. 20 8

2000

13

15

18

8

13

18

2000

15 7

12

17

y pos [m]

0

2

6

11

16

19 5

-1000

1

5

10

3

1000

10

y pos [m]

3

1000

12

17 10

0

2

6

-1000

15

7

1

11

5

16

10

19 5

15

0 -2000

4

9

-2000

14

-3000 -2000 -1000 0 1000 x pos [m]

4

2000

-5

3000

9

14

-3000 -2000 -1000 0 1000 x pos [m]

2000

0

3000

Figure 2 – left: ROI max SINR; right: CQI mapping

In the Figure 3 there is shown the same that in figures described above, with the same network configuration, but with the difference that in these there is no fading in the channel. So it is apparent that the values of SINR and CQI are quite better than in the border of the cell than those obtained in Figure 2. 15

20 8

2000

13

8

18

13

18

2000 15

y pos [m]

7

12

17

0

2

6

11

16

19 5

-1000

1

5

10

3

1000 10

y pos [m]

3

1000

0

-1000

15

2

7

6

1

12

11

5

17

16

10

10 19 5

15

0

-2000

-2000 4

9

4

14

-3000 -2000 -1000 0 1000 x pos [m]

2000

3000

-5

9

-3000 -2000 -1000 0 1000 x pos [m]

14 2000

3000

0

Figure 3 – left: ROI max SINR; right: CQI mapping

Knowing the network fading conditions it is easier to make the interpretation of the capacity values obtained in network and per user.

688


Average Throughput per UE 10

Mbps

8 6 Urban Macro

3,97 4

2,2

2,1

Urban Micro

1,98 1,07

2

1,03

0,69

1,29

0,68

Suburban Macro

0 2000 UEs

4000 UEs

6000 UEs

Number of UEs

Figure 4 – Average throughput in non femtocellular network.

Average Spectral Efficiency per cell 4,00 3,86

3,90

3,74

bit/s/Hz

3,80 3,68

3,70 3,58

3,60

3,58

3,62

3,61

3,56

Urban Macro

3,49

3,50

Urban Micro

3,40

Suburban Macro

3,30 3,20 2000 UEs

4000 UEs

6000 UEs

Number of UEs

Figure 5 – Spectral efficiency in non femtocellular network.

Looking at Figure 4, it can be seen the average throughput achieved in non femtocellular configurations for the three types of studied environments, as well as in a study for the three densities users. So it is concluded that for a smaller number of network users are achieved higher throughput because there are more resources in the network to assign. Regarding the change of the throughput values obtained depending on the propagation environment, it is possible to see that the difference is practically reduced but, in a suburban environment, throughputs are higher. In Figure 5, we can see the efficiency values per cell where, for the same number of users, higher values are achieved in an urban macrocellular environment. The most efficient average is for a smaller number of users. Average Throughput per UE 9,19

10

8,46

7,9 Mbps

8

6,44

6

7,9

6,01 4,65

5,6 4,45

4

Urban Macro Urban Micro

2

Suburban Macro

0 2000 UEs

4000 UEs

6000 UEs

Numver of UEs

Figure 6 - Average throughput in femtocellular network.


Average Spectral Efficiency per cell 4,00 2,76

bit/s/Hz

3,00 2,00

2,02 1,49

2,69 2,24 1,87

2,83 2,50 2,13 Urban Macro Urban Micro

1,00

Suburban Macro 0,00 2000 UEs

4000 UEs

6000 UEs

Number of UEs

Figure 7 - Spectral efficiency in femtocellular network.

In Figure 6, for a femtocell network, it is also achieved higher throughputs to a smaller number of network users because the interference caused by users is smaller and the environment that allows higher values is the suburban macro environment since there are more resources to less users. In terms of Figure 7, there is a greater efficiency for the suburban environment and the average efficiency per cell is greater for 4000 users. Average Throughput per UE 10 7,64

Mbps

8

7,08 5,04 4,68

6 4

3,72

2,89

4,3 4,25 2,55

Urban Macro Urban Micro

2

Suburban Macro

0 2000 UEs

4000 UEs

6000 UEs

Number of UEs

Figure 8 - Average throughput in hybrid network.

Average Spectral Efficiency per cell 4,00 3,00 z H /s /t 2,00 i b 1,00

2,84

3,17

2,19

3,10 2,86 2,35

3,18 2,98 2,45 Urban Macro Urban Micro Suburban Macro

0,00 2000 UEs

40 00 UEs

6000 UEs

Number of UEs

Figure 9 - Spectral efficiency in hybrid network. Finally, in Figure 8 are shown the values obtained for a hybrid solution where it can be seen that the higher values of throughputs are for 2000 users and for an urban micro environment. That is related to the fact that the propagation model used was COST 231 which means that the results obtained for urban micro environments are a little bit different from reality. Regarding the

Figure 9 the values of spectral efficiency per cell are very close to macro urban and suburban environments being

689

690


that the least efficient is micro urban environment. In all the spectral efficiency graphics it is possible to see that the average values are good which means that the scheduler used is a good choice in what concerns to fairness. 4.2. Financial Analysis In order to obtain the following values, were considered the parcels and rates above and the (1), (2) and Error! Reference source not found. equations. In the macrocellular solution it was assumed 19 enodeBs; in femtocellular solution it was assumed 50 femtocells and in the hybrid solution, it was assumed 19 enodeBs and 50 femtocells. In Table 2 it is presented the CAPEX, NPV and Payback values for each one of the three network solutions for low level demand and high level demand where low level demand represents 100Mbps lines and the high level demand represents 1Gbps lines. Table 2 – Network Solutions Investiments Network Solution

CAPEX [€]

NPV [€]

PayBack [time]

Low level demand

2 652 938

1 682 742

5 years and 6 months

High level demand

2 675 738

1 690 082


Low level demand

417 941

2 259 685

Investment year

High level demand

477 941

2 274 930

Investment year

Low level demand

2 840 638

1 608 841


High level demand

2 923 438

1 635 497


MacroCellular Deployment

FemtoCellular Deployment

Hybrid Deployment

The NPV value is one of the most used indicators in assessing the viability of a project. If the value obtained is positive the project is profitable but if otherwise, the project does not demonstrate the ability to generate sufficient profits therefore does not have conditions to advance and to be feasible. Present in the table above, the NPV values are bigger for high level demand and the higher values are achieved for the femtocell solution. The NPV values are all positive which indicates that all the solutions are viable and according to the values represented the best solution of the three is the femtocellular since the NPV values are bigger. Relative to the payback the bigger values are in the hybrid network as expected since the CAPEX of the same is bigger than the CAPEX of the other solutions so, because of that, the recovery period of the investment is significantly higher. The best solution is the femtocellular since the investment is going to be all paid at the beginning. Comparing the CAPEX for low level demand solutions with the CAPEX for high level demand solutions, it is possible to see that the difference between the two is small which means that the most viable solution is going to be with 1Gbps lines. 5.

Conclusions

We conclude in general that, at the network performance level the solution with better efficiency values is the macrocellular network and that the average throughput values per user are bigger in femtocells networks and urban macro environment. So, when it is desired to provide services with high flow requirements, the femtocells are a better solution despite the less efficient values. For low level demand networks, with a big number of users, the most viable solution is the macrocellular.


In terms of monetary quantification of the various solutions through the payback values is possible to see that the most viable solution is a femtocell network. For macrocellular networks it is clear that the costs are recovered only five years after the initial investment as opposed to femtocells and as opposed to hybrid networks that are recovered after six years. Comparing the CAPEX for low level demand solutions with the CAPEX for high level demand solutions, it is possible to see that the difference between the two is small which means that the most viable solution is going to be with 1Gbps lines. So after this, it is possible to conclude that the more viable solutions are for high level demand. The greenfield deployment of a femtocellular network introduces high values of CAPEX for the capacity gain that presents so the best solution is a hybrid network. If there is already exists a macrocellular network and there is the need to increase the network capacity because of shadow zones; or traffic flow increasing from the cells of the higher layer flow some of the traffic to another cell; or give coverage for users that needs big values of QoS, the best solution is a femtocellular network. This is due to the fact that, each femtocell covers few users and because, for the same number of users the number of femtocells is going to be bigger than the number of macrocells so, the CAPEX is going to increase. References [1] H. H. a. A. Toskala, LTE for UMTS OFDMA and SC-FDMA Based Radio Access, WILEY. [2] “Overview of 3GPP Release 8 V0.2.11,” 2013-06. [3] M. W. M. R. Josep Colom Ikuno, “System level simulation of LTE networks,” 2010. [4] L. T. W. H. L. G. S. Holger Claussen, “Financial Analysis of a Pico-Cellular Home Network Deployment,” 2007 IEEE. [5] J. P. Zoraida Frias, “Techno-economic analysis of femtocell deployment in long-term evolution networks,” 2012. [6] P. Wholesale, “ORCE – Oferta de Referência de Circuitos Ethernet,Anexos 2, Preços” 2013.

691