Wireless Pers Commun DOI 10.1007/s11277-017-4100-z

SON Handover Algorithm for Green LTE-A/5G HetNets Maissa Boujelben1,2 • Sonia Ben Rejeb1 • Sami Tabbane1

 Springer Science+Business Media New York 2017

Abstract Fourth generation networks and beyond (5G) have recently emerged to satisfy the increasing demand for high data bit rates. Cost-effective means such as small cells have been designed in 3GPP LTE-Advanced standard since the Release 10 to significantly enhance coverage and capacity. The deployment of small cells over Macrocells layer introduces a new type of networks called Heterogeneous Networks (HetNets). Another cost saving concept is Self-Organizing Networks (SON) which aims at minimizing human efforts in management and operating processes. The last emerging trend for cost effectiveness is to save energy especially in access networks by turning off the unused stations. In this context, we propose a new green procedure that aims at minimizing the energy consumption in LTE-Advanced/5G access networks using Handover self-optimization SON function. We also introduce a new mathematical model for energy cost calculations. Performance evaluation results show that our proposed algorithm considerably reduces the energy consumption in the network for all scenarios related to user speed. Keywords LTE-A/5G  Green HetNets  SON  Self-optimization  Handover  Energy-efficiency

& Maissa Boujelben [email protected]; [email protected] Sonia Ben Rejeb [email protected] Sami Tabbane [email protected] 1

MEDIATRON Laboratory, Higher School of Communications (Sup’Com), Ariana, Tunisia

2

Esprit School of Engineering, Tunis, Tunisia

123

M. Boujelben et al.

1 Introduction Fifth Generation (5G) networks are defined weather as an evolution or a revolution of LTE-Advanced/4G networks. Nowadays, all researchers’ efforts are directed to the definition of the appropriate technologies able to fulfill the 5G requirements by 2020. Yet, many Key Performance Indicators (KPI) must be ameliorated compared to the existing 4G networks such as capacity, latency and energy consumption which leads to sharp competition between 4G and beyond mobile operators to provide reliable and green network services. Moreover, in order to maintain their competitiveness in the market, mobile network operators should look for new strategies to keep their Total Cost of Ownership (TCO) composed of capital expenditure (CAPEX) and operational expenditure (OPEX) costs as low as possible. One promising solution for CAPEX costs minimization is the intensive introduction of small cells and especially Femtocells that provide high data bit rates for indoor and hot spots with reduced costs when compared to Macrocells. Fortunately, Femtocells can be part of the green directive of mobile operators thanks to their low transmission power and therefore their low energy consumption. However, with the densification of mobile Heterogeneous Networks (HetNets) through the introduction of Femtocells at large scale, the management of networking processes such as configuration, optimization and maintenance is becoming a real burden for mobile operators and OPEX costs will consequently increase. Self-Organizing Networks (SON) provide a complete solution for such a problem including a variety of functions with high deployment flexibility. These functions are grouped into three categories: self-configuration, self-optimization and self-healing [1]. In this paper, we propose to use SON Handover (HO) self-optimization function for energy saving which will highly reduce the OPEX costs. The remainder of this paper is outlined as follows. In the next section, previous works related to cost reduction and evaluation methods are discussed. Section 3 details the proposed green HO self-optimization procedure which addresses the issue of energy cost minimization. In Sect. 4, we report and analyze the simulation results while taking into consideration all the possible scenarios. Finally, conclusions and perspectives for this work will be given in Sect. 5.

2 State of the Art OPEX and CAPX costs optimization for different types of networks have been subject to consistent research works since several years. In [2], the authors have suggested a new cost estimation method for the network operations in order to deal with the issue of OPEX reduction. Thus, they explain how this method can estimate the operation cost more accurately and precisely than other previous methods and how this can help to analyze the cause of cost expenses and to establish the strategy and plan to improve the operational process and to reduce the operational cost. The authors of [3] have presented some approaches to model and minimize the total cost of ownership for operators which is composed of CAPEX and OPEX costs. They also highlighted that OPEX can become immense if frequent tasks are not supported by semi-/fully automated operation, administration and maintenance tools.

123

SON Handover Algorithm for Green LTE-A/5G HetNets

The idea of radio access network virtualization was discussed in [4]. The authors stated that the virtualization of core networks is not sufficient to reach cost effective and energy efficient networks since access networks are the source of up to 40% of the total operational cost of a cellular network. In [5], the authors proposed to switch off some base stations of an UMTS (Universal Mobile Telecommunications System) network in order to achieve cost effectiveness by energy saving at access networks. In [6], the authors noticed that the utilization of base stations is not useful during out of peak hours. Thus, they proposed to dynamically switch off some base stations (BS) in order to reduce the energy according to the varying characteristic of the traffic profile over the time. Other researchers introduced the concept of cell zooming for energy savings which consists on adjusting the cell size in an adaptive way according to traffic load, user requirements and channel conditions. Centralized and distributed algorithms for cell zooming were then presented and their role in achieving green cellular networks was proven [7]. In [8], the authors proposed a dynamic energy saving algorithm which jointly optimizes some quality of service parameters and energy consumption in the whole network by studying the different possibilities of association between users and base stations. A new green base station switching off algorithm for LTE-Advanced networks was presented in [9]. The proposed idea consists of switching off the underutilized base stations during low traffic load in order to optimize the power consumption in the network without degrading the offered quality of service. This algorithm is based on the distance separating users from their associated stations. The authors of [10] proposed a dynamic base station switching on/off algorithm that can be implemented in a distributed manner and thus requires low computational complexity. The minimal impact of switching off one base station on its neighboring cells is taken into consideration for the switch off decision. In the literature, several cost models were proposed and different cost reduction methods were discussed. However, the majority of these solutions require additional capital (Hardware or Software) investments such as data centers, optical fiber connections or new software installations with the necessary supervision tools. In this paper, our major contribution is the proposal of a customized OPEX cost model coupled with the development of a new green procedure which does not require any additional capital expenditures. Our proposed solution is based on self-organizing functions.

3 Proposed Green Approach The figure below (Fig. 1) describes the Total Cost of Ownership (TCO) which is composed of OPEX and CAPEX costs. In this paper, we are exclusively interested in minimizing the OPEX costs by reducing energy consumption in the network. Indeed, we aim at developing an OPEX cost-effective procedure by HO self-optimization in an LTE-A/5G HetNet environment. The total OPEX cost can be calculated as follows [3]: OpexðTÞ ¼ CMng ðTÞ þ COpr ðTÞ þ CInf ðTÞ

ð1Þ

where T denotes the operation period, CMng ðT Þ is the cost of management during T, COpr ðT Þ is the cost of operations during T and CInf ðT Þ is the operational cost of infrastructure during T which is given by Eq. (2):

123

M. Boujelben et al. Fig. 1 Diagram of SON module for OPEX minimization

Total Cost of Ownership (TCO)

OPEX

Cost of Operations

CAPEX

Cost of Energy

Other Costs (management, space)

HO SelfOptimization

Self-Organizing Networks (SON)

CInf ðT Þ ¼ CEnergy ðT Þ þ CSpace ðT Þ

ð2Þ

where CEnergy ðT Þ is the cost of the energy consumed by all nodes of the network during T and CSspace ðT Þ is the cost of the space occupied by all nodes during T. Therefore, the OPEX cost model can be given by: OpexðT Þ ¼ COpr ðT Þ þ CEnergy ðT Þ þ Cother

ð3Þ

where Cother is the sum of the costs different from operations and energy costs which is given by: Cother ¼ CMng ðT Þ þ CSpace ðT Þ

ð4Þ

In the next sub-section, we will present our green HO self-optimization approach for energy cost minimization as well as the mathematical model used to evaluate this specific type of cost.

3.1 Green Handover Self-optimization Approach In our work, we will bring few changes to the HO procedure [11] in order to save the energy consumed by unused Macrocells or Femtocells. The added steps are highlighted by the yellow color in Fig. 2. At the end of each measurement gap, the source eNB asks the UE to send it a measurement report. Traditionally, this report contains some parameters such as the received signal level and quality. In our modified HO procedure, the UE will also report its speed to its serving eNB. The latter continuously checks whether the received power of the signal, known as RSRP (Reference Signal Received Power), is less than a prefixed threshold (3GPP condition for the A2 event). If the A2 event is triggered, the UE begins looking for a potential target cell by calculating the RSRP level relative to the neighboring cells and includes this

123

SON Handover Algorithm for Green LTE-A/5G HetNets

UE

Serving eNB

MME/S-GW

Target eNB/ HeNB

TargetGW

Measurment Control Measurment Report

Candidate List Selecon Target Cell selecon HO Request

HO Request HO Request Admission Control HO Response

HO Command

HO Response

HO Response

HO Confirm Path switch Request Path switch Request Path switch ACK Path switch ACK Release Resources Release Resources Release Resources

Fig. 2 Self-optimized handover procedure

information in its measurement report. Thus, the source eNB will be able to set a list for HO candidate cells. Our proposed HO procedure is based on optimizing the target cell selection from the candidate list based on two criteria: the candidate cell load and the user speed (see subsection IV-B-a). Once the best target cell is selected, the source eNB will send a HO request to the target cell. The latter will check if the available resources are sufficient to accept the new user according to the admission control condition. In the favorable case, the response of HO will be relayed to the source eNB by the different gateways (target gateway and serving gateway). The serving eNB will then send the HO command to the user which confirms the HO execution to the target cell. Finally, the path of data packets will be switched to the new cell and the old resources at the source eNB side will be released.

3.1.1 Optimized Target Cell Selection In this paper, we propose to optimize the choice of the target cell during the HO execution according to energy saving perspective. Let us assume that the UE is initially attached to its serving cell (that can be either a Macrocell or a Femtocell). If the user is moving far from its serving station, the reference received signal power (RSRP) will decrease until reaching a predefined threshold. Thus, the UE will perform HO to one of the neighboring cells. In the traditional approach, the candidate cells for HO are cells that provide sufficient RSRP level to the moving user. The target cell is in general, the cell that offers the best RSRP level (see left side in Fig. 3). According to this strategy, an edge user may be handed

123

M. Boujelben et al.

Better RSRP Level

High Loaded Cell + Sufficient capacity

Traditional Approach

Proposed Approach

Low Loaded Cell

Fig. 3 Traditional versus proposed target cell selection approach

over to the best transmitting Femtocell in its neighborhood. This HO decision is certainly beneficial in terms of energy consumption minimization. Nevertheless, if the user is moving with high speed, the quality of its link to the Femtocell will rapidly be deteriorated since the Femtocell coverage is too limited. At that time, the user will be obliged to switch to another cell and may suffer from the problem of continuous HO. Our proposed target cell selection will avoid this issue by adapting the candidate cell type to the speed of the user as follows: Users moving with a high speed (greater than V1_threshold) must only be handed over to Macrocells, whereas low speed users (lower or equal to V1_threshold) should be handed over in priority to Femtocells and then to Macrocells if there is no available resources at the neighbor HeNBs. In addition to the type of the cell, the load is also an important parameter that we will use in order to save more energy. Indeed, in the traditional approach, a user in HO can be directed to a very low loaded cell because it has the best offered RSRP level. As a consequence, the node of this cell will consume energy to serve few users. Therefore, in our proposes approach, The user will be handed over to the highest loaded cell from the candidate list that still has sufficient resources. Using this new strategy, low loaded cells will be gradually until completely offloaded. Finally, empty cells will be switched in order to ensure less energy consumption in the whole network. Our proposed rules for target cell selection optimization are gathered in the table below: (Table 1)

3.1.2 Proposed Green Handover Self-optimization Algorithm Our proposed HO optimization algorithm detailed in Fig. 4 allows to minimize both the unnecessary Handover number as well as the energy consumption at operator level.

Table 1 Optimized target cell selection based on user velocity and cell load Low loaded Macrocell V [ V1_threshold V B V1_threshold

123

High loaded Macrocell

Low loaded Femtocell

High loaded Femtocell

X X

SON Handover Algorithm for Green LTE-A/5G HetNets

RLF_Number ++

RLF_Number ++

Start

Calculate RSRP for Neighbor List No RSRP_t satisfies A3 condition Yes Add Cell_t to Macro/Femto Candidate_List

UE_speed < V1

No

Target_Cell=cell from Macro Candidate_List With highest load

Target_Cell=cell from Femto Candidate_List With highest load

No

yes

No

Macro Candidate_List = {} Yes

yes

Femto Candidate_List = {}

Remove cell from Candidate_List

No

C’ + Creq < = Th_HO * C

Yes

RSRP_s < RSRP_th

RSRP_s < RSRP_th Remove cell from Candidate_List

No No

No

C’ + Creq < = Th_HO * C

yes

yes

HO command to target Macrocell

HO command to target Femtocell

No

RSRP_t < RSRP_th

Yes

Fig. 4 Proposed green handover Self-optimization algorithm

Our algorithm is compliant with the 3GPP standard condition for Handover defined as the A3 event [11]. The UE keeps measuring the RSRP level received from its serving cell and all its neighbor ones. If the signal level received from the source cell becomes worse than a certain threshhold, and one of the neighboring cells offers a signal level to the user that is offset better than the one received from the first cell during a time period equal to TTT (Time To Trigger), then a HO will be executed. RSRPt þ Oft þ Oct  Hys [ RSRPs þ Ofs þ Ocs þ Off

ð5Þ

Where RSRPt and RSRPs are the received signal powers respectively received from target and source cell, Of t and Of s are the frequency specific offsets respectively of target and source cell, Oct and Ocs are the cell specific offsets respectively of target and source cell, Off is the offset parameter of A3 event and Hys is the Hysteresis margin for Handover.

123

M. Boujelben et al.

After that, neighbor cells that verify the A3 condition will be grouped into two different candidate lists for each user according to the cell type: Macro Candidate List and Femto Candidate List. According to the user speed related rules already discussed, the source cell will choose the appropriate target cell for HO. In the specific case where user is moving with low speed, it is possible that no Femtocell could take the user in Handover in charge. Therefore, a cell from the Macro candidate list will be selected in order to avoid a radio link failure. The target cell will be selected according to the highest load strategy. We compute the cell load using the following formula: Cellload ¼

N X

Consumed PRB i=Total PRB

ð6Þ

i¼1

Where i = 1…N is the total number of active users in the cell, PRB is the Physical Resource Block, Consumed_PRB_i is the number of PRBs allocated to user i and Total_PRB is the total number of available PRBs in the cell according to the used frequency bandwidth. After the selection of the appropriate target cell, the Admission Control (AC) condition [12] should be checked so that we ensure that the target cell will be able to provide the user with its required data bit rate. 0

C þ Creq  Th HO  C

ð7Þ

Where C is the cell total capacity, C’ is the already consumed capacity by the cell active users, Creq is the capacity required by the user in HO and TH_HO is the amount of cell capacity reserved for HO calls. If the AC is verified, a HO command will be sent from source to target cell. In the opposite case, this cell will be deleted from the candidate list and the next one will go through the same verification steps. When applying our new HO procedure, cells that serve initially few users will be gradually offloaded and the other ones will use their available capacity in an efficient way. If one cell is completely offloaded, we will turn it off in order to save energy. These cells can be turned on again if the users’ demands exceed the available capacity. Consequently, Femtocells will provide low speed users with their required QoS and empty cells will be switched off so that the total energy consumption of the networkwould be minimized.

3.2 Cost of Energy Model The total cost of energy in the network can be calculated as: CEnergy ðT Þ ¼

N X

CEnergyn ðT Þ

ð8Þ

n¼1

Where N is the total number of nodes in the network and CEnergyn ðT Þ is the elementary cost of energy consumed by one node during T. As mentioned before in this paper, we will consider two types of nodes which are eNB and Home eNB (HeNB). The energy consumption models of these two types of stations are so different. Therefore, the total amount of energy cost in the network can be further detailed as follows:

123

SON Handover Algorithm for Green LTE-A/5G HetNets

CEnergy ðT Þ ¼

M X

CEnergym ðT Þ þ

m¼1

F X

CEnergyf ðT Þ

ð9Þ

f ¼1

Where M is the number of Macrocells and F the number of Femtocells in the network. The models of power consumption for both base stations were studied in previous works [13–15]. Based on these models, the simplified energy equations can be given by: CEnergym ðT Þ ¼ ðTx Powerm þ Site Powerm Þ  T  CPower   CEnergyf ðT Þ ¼ Tx Powerf þ Site Powerf  T  CPower

unit unit

ð10Þ ð11Þ

Where Tx_Power is the transmission power of the node and Site_Power is the power consumed by the node for cooling, backhaul, lighting and monitoring, while CPower unit is the cost of one power unit. Some recent researches have been carried in order to manage power and improve the energy efficiency in each node [15, 16]. However, these approaches have limited and local impact on energy consumption. Therefore, our proposed model aims at acting at global scale by reducing the total number of active nodes with the switching off technique.

4 Energy Cost Evaluation Results In this work, simulations were conducted using a Matlab simulator for an LTE-A HetNet as presented in Fig. 3 of Sect. 3. Simulation parameters and results will be detailed and analyzed in the following subsections.

4.1 Simulation Parameters Table 2 summarizes the different parameters used for our simulations [11, 17]. The simulation parameters relative to energy calculations [12] are grouped in Table 3. Since our proposed HO optimization algorithm is based on user speed, we will test and analyze the obtained results for three different scenarios related to the UE velocity so that the results can be significant: • Scenario 1: users with low speed In this scenario, we expect to use all the resources of the available Femtocells and to offload some or many Macrocells. • Scenario 2: users with high speed In this scenario, we expect to offload all Femtocells and probably some Macrocells. • Scenario 3: users with random speed In this scenario, we cannot predict the obtained results since the speed is random.

4.2 Simulation Results This part is dedicated to the presentation and the analysis of simulation results according to the three different scenarios and their comparison to the reference HO procedure where all the cells in the network are normally turned on. Therefore, the energy consumption in

123

M. Boujelben et al. Table 2 Simulation parameters

Simulation parameter

Value

Number of Macrocells

7

Number of Femtocells per Macrocell

10

eNB transmission power

46 dBm

HeNB transmission power

20 dBm

Total number of users

420

Femtocell maximum users number

8

Users distribution model

Uniform

Users mobility model

RWP

Users traffic model

Non real time

Low mobility user speed

B5 km/h

High mobility user speed

[50 km/h

Random mobility user speed

[3 km/h and B120 km/h

RSRP_th (dBm) [23]

-101.5

Hysteresis (dB)

6

Th_HO

0.7

V1_threshold

5 m/s

Table 3 Power consumption parameters in HetNets Power component

Average power consumption per eNB

Average power consumption per HeNB

Tx_power

46 dBm

20 dBm

P_cooling

2000 W

0W

P_backhaul

200 W

100 W

P_lighting

50 W

50 W

P_monitoring

50 W

50 W

Total consumption

2340 W

200.1 W

the traditional case reaches its maximum and is assumed to be constant during the simulation time (see Table 3). We recall that the cell load is given by Eq. (6).

4.2.1 Scenario 1 Since all users move with a relatively low velocity, it is better,from energy saving point of view, to direct them towards Femtocells and to switch off the completely offloaded Macrocells. Our HO self-optimization procedure is run at t = 0. As shown by Fig. 5, all Macrocells were completely offloaded after 220 s. This could be explained by the fact that the overlaying Femtocells have sufficient resources to serve all the users. Figure 6 shows that Macrocell 2 is completely offloaded (from 72% to 0% after 50 s) while the overlaying Femtocells are gradually admitting new users in HO from Macrocell 2 and their loads increased until stabilization between 82% and 100%. As a result, the total energy consumption of the network is gradually decreased as the Macrocells are completely offloaded and then switched off. For this particular scenario, the

123

SON Handover Algorithm for Green LTE-A/5G HetNets 1

Fig. 5 Macrocells load evolution _Scenario1

Macro cell 1

0.9

Macro cell 2

Cell Load

0.8

Macro cell 3

0.7

Macro cell 4

0.6

Macro cell 5

0.5

Macro cell 6 Macro cell 7

0.4 0.3 0.2 0.1 0

0

50

100 150 200 250 300 350 400 450 500 550

Time in sec

simulations have shown stable energy consumption after 220 s with a total energy gain of 58.5% (see Fig. 7).

4.2.2 Scenario 2 In this scenario, users move with such a high speed that they can no more be served by Femtocells. On the one hand, UEs who were initially attached to Femtocells will be directed to the appropriate Macrocells (sufficient RSRP and highest load). Therefore, Femtocells will be completely offloaded and then switched off. This will result on 100% of Femtocells energy saving. Figure 8 illustrates the example of Macrocell 4 and some of its overlaying Femtocells which are completely offloaded since 90 s.

1

Fig. 6 Offloading of macrocell 2 by some overlaying Femtocells_Scenario1

Cell Load

0.9

Macro cell 2

0.8

Femto cell 21

0.7

Femto cell 22

0.6

Femto cell 23 Femto cell 24

0.5

Femto cell 25

0.4

Femto cell 26

0.3

Femto cell 27

0.2 0.1 0

0

50

100 150 200 250 300 350 400 450 500 550

Time in sec

123

M. Boujelben et al. Fig. 7 Energy consumption evolution_Scenario1

x 10

3.2

4

Energy consumption (Watt)

3

New Energy consumption

2.8

Traditional Energy consumption

2.6 2.4 2.2 2 1.8 1.6 1.4 1.2

0

50

100 150 200 250 300 350 400 450 500 550

Time (sec)

On the other hand, users who were first affected to Macrocells will be directed to neighbor Macrocells having a higher load. Then, if a Macrocell is totally offloaded, it will consequently be turned off to save more energy. Simulation results reported by Fig. 9 show that Macrocell 3 was also offloaded. This will result on additional energy savings (14.3% of Macrocells energy saving) at the end of the simulation. When considering the network as a whole (Macrocells and Femtocells), the energy consumption of the network implementing our HO algorithm has clearly decreased compared to the traditional one. Energy savings have reached 53.8% as shown by Fig. 10.

1

Fig. 8 Offloading of femtocells by overlaying Macrocell 4_Scenario2

Cell Load

0.9

Macro cell 4

0.8

Femto cell 41

0.7

Femto cell 42

0.6

Femto cell 43 Femto cell 44

0.5

Femto cell 45

0.4

Femto cell 46

0.3 0.2 0.1 0

0

50

100

150 200

250 300

Time in sec

123

350 400

450 500

550

SON Handover Algorithm for Green LTE-A/5G HetNets 1

Fig. 9 Macrocells load evolution _Scenario2

Cell Load

0.9 0.8

Macro cell 1

0.7

Macro cell 2

0.6

Macro cell 3 Macro cell 4

0.5

Macro cell 5

0.4

Macro cell 6

0.3

Macro cell 7

0.2 0.1 0

0

100

200

300

400

500

600

700

800

Time in sec

4.2.3 Scenario 3 This scenario is based on the random speed of users. That is why it is almost not possible to predict the results obtained after each iteration. As shown by Fig. 11, the Macrocells loads vary randomly as the users speed is also randomly variable. In this iteration, three Macrocells were completely offloaded after t = 850 s (Macrocell 2, 4 and 6). Thus, 42.8% of Macrocells energy consumption can be saved. The users normally served by these Macrocells will be served by either the overlaying Femtocells (10 Femtocells per Macrocell which are almost all activated) or by the neighboring Macrocells which can admit new users. The choice of the target cell type is based on the user speed as already explained. Figure 12 shows the load evolution of Macrocell 6 and some of its overlaying Femtocells. The results show that Macrocell 6 is completely offloaded while Femtocells have admitted new users performing HO. Fig. 10 Energy consumption evolution_Scenario2

3.2

x 10

4

Energy consumption (Watt)

3

New Energy consumption

2.8

Traditional Energy consumption 2.6 2.4 2.2 2 1.8 1.6 1.4

0

100

200

300

400

500

600

700

800

Time (sec)

123

M. Boujelben et al. 1

Cell Load

Fig. 11 Macrocells load evolution _Scenario3

0.9

Macro cell 1

0.8

Macro cell 2

0.7

Macro cell 3 Macro cell 4

0.6

Macro cell 5

0.5

Macro cell 6

0.4

Macro cell 7

0.3 0.2 0.1 0

0

200

400

600

800 1000 1200 1400 1600 1800 2000

Time in sec

Next, we report the obtained results relative to the total energy consumption of two different iterations (see Fig. 13 and Fig. 14). Both of figures show that our green selfoptimization HO procedure clearly reduces the total energy consumption in the network. However, the obtained energy gain was different for both iterations. For the first iteration, the total gain of energy was oscillating with a decreasing tendency as shown by Fig. 13. The average gain obtained at the end of simulations is about 32.7% with a variance of ± 4%. For the second iteration, the total gain of energy was also oscillating with a decreasing tendency as shown by Fig. 14. The average gain obtained at the end of simulations is about 36.8% with a variance of ± 3.8%. In the table below, we provide a summary of the energy consumption of Macrocells, Femtocells and the whole network for the different tested scenarios. Simulation results confirm that our proposed HO procedure is energy efficient independently from the user speed. Besides, the maximum of energy gain was reached in the first scenario by 100% 1

Fig. 12 Femtocells load evolution_Scenario3

Macro cell 6

0.9

Femto cell 51

Cell Load

0.8

Femto cell 52

0.7

Femto cell 53

0.6

Femto cell 54

0.5

Femto cell 55 Femto cell 56

0.4 0.3 0.2 0.1 0

0

200

400

600

800

1000 1200 1400 1600 1800 2000

Time in sec

123

SON Handover Algorithm for Green LTE-A/5G HetNets Fig. 13 Energy consumption evolution_Scenario3_iteration 1

3.2

x 10

4

Energy consumption (Watt)

3

New Energy consumption 2.8

Traditional Energy consumption

2.6 2.4 2.2 2 1.8

0

200

400

600

800 1000 1200 1400 1600 1800 2000

Time (sec)

Fig. 14 Energy consumption evolution_Scenario3_iteration 2

3.2

x 10

4

Energy consumption (Watt)

3

New Energy consumption 2.8

Traditional Energy consumption

2.6 2.4 2.2

X: 4464 Y: 2.037e+004

2 1.8

0

500 1000 1500 2000 2500 3000 3500 4000 4500 5000

Time (sec) Table 4 Energy gain per scenario Scenario1 (%)

Scenario2 (%)

Scenario3_ iteration1 (%)

Scenario3_ iteration2 (%)

Macrocell energy gain

100

14.3

42.8

42.853

Femtocell energy gain

10

100

21.4

30

Total energy gain

58.5

53.8

32.7

36.8

saving for Macrocells and 10% of energy saving for Femtocells. Moreover, our method bypasses other previous researches [4], [10] dealing only with energy issue and saving up to 80% of the energy consumption (Table 4).

123

M. Boujelben et al.

5 Conclusion In this work, we first introduce an OPEX cost model for mobile network operators. Then, we propose a new OPEX minimization method for LTE-A HetNets based on green HO self-optimization module that aims at minimizing the cost of energy consumption in the network by switching off the completely offloaded cells. Simulations have shown that our new SON function for Handover minimizes the energy consumption of the network in all possible scenarios relative to users speed. More specifically, the obtained results show that our HO self-optimization module gives the best results when the mobile users move with low speed (scenario 1) such as in dense urban areas. In the next works, we intend to address the inter cell interference issue in LTE-A/5G HetNets by exploiting the COMP feature as well as the Carrier aggregation option. Besides, we will study the tradeoff between energy saving and quality of service (QoS).

References 1. http://www.3gpp.org/technologies/keywords-acronyms/105-son. 2. Kwak, E. J., Kim G. E., & Yoo J. H. (2011). Network operation cost model to achieve efficient operation and improving cost competitiveness, ICACT. 3. Gruber, C. G. (2009). CAPEX and OPEX in aggregation and core networks, IEEE NFOEC. 4. Rahman, M. M., Despins, C., & Affes, S. (2013). Analysis of CAPEX and OPEX benefits of wireless access virtualization, IEEE ICC. 5. Chiaraviglio, L., Ciullo, D., Meo, M., Marsan, M. A., & Torino, I. (2008). Energy-aware UMTS access networks WPMC. 6. Oh, E., & Krishnamachari, B. (2010). Energy savings through dynamic base station switching in cellular wireless access networks. GLOBECOM. 7. Niu, Z., Wu, Y., Gong, L., & Yang, Z. (2010). Cell zooming for cost-efficient green cellular networks. IEEE Communications Magazine, 48(11). 8. Kang, T., Sun, X., & Zhang T. (2012). Base station switching based dynamic energy saving algorithm for cellular networks, Proceedings of IC-NIDC. 9. Bousia1, A., Antonopoulos, A., Alonso1, L., & Verikoukis, C. (2012, June). ‘‘Green’’ distance-aware base station sleeping algorithm in LTE-Advanced. IEEE International Conference on Communications (ICC). 10. Oh, E., Son, K., & Krishnamachari, B. (2013, May). Dynamic base station switching-on/off strategies for green cellular networks, IEEE Transactions On Wireless Communications. 11. 3GPP TS 36.331 V12.3.0: ‘‘Evolved Universal Terrestrial Radio Access (E-UTRA); Radio Resource Control (RRC); Protocol specification’’, September 2014. 12. Sas, B., Spaey, K., Balan, I., Litjens. R., & Zetterberg, K. (2011, May). Self-optimisation of admission control and handover parameters in LTE, IEEE. 13. Arnold, O., Richter, F., Fettweis, G., & Blume, O. (2010). Power consumption modeling of different base station types in heterogeneous cellular networks. In Future Network and MobileSummit 2010 Conference Proceedings. 14. Keating, W. (2011, August). Reducing Energy consumption in access networks, master of engineering. In Telecommunications Engineering, Dublin City University. 15. Laguidi, A., Hayar, A., & Wetterwald, M. (2013, March 24–26) Improving energy efficiency of Femtocell network, ICSPT 2013, International Conference on Signal Processing and Telecommunications, Sousse, Tunisia. 16. Xu, X., Kutrolli, G., & Mathar, R. Energy efficient power management for 4G heterogeneous cellular networks, 1st International Workshop on Green Optimized Wireless Networks (GROWN’13). 17. 3GPP TS 36.104 V12.4.0: ‘‘Evolved Universal Terrestrial Radio Access (E-UTRA); Base Station (BS) radio transmission and reception’’, June 2014.

123

SON Handover Algorithm for Green LTE-A/5G HetNets

Maissa Boujelben is an Assistant Professor at Esprit School of Engineering of Tunisia an a researcher at MEDIATRON Laboratory of the Higher School of Communications of Tunis (Sup’Com). She graduated from Sup’Com in 2012 as a Network Engineer and began her professional carrier as a CS Core Planning and Optimization Engineer at OoredooTunisia in the same year. She obtained the Doctorate diploma from Sup’Com in 2016 with Honors. Her research interests are Self-Organizing Networks (SON), Load Balancing in LTE-Advanced Heterogeneous Networks (HetNets), Handover optimization, Energy saving (Green networks), Inter Cell Interference management and Inter Operator Cooperation.

Sonia Ben Rejeb is M. Assistante at ISIE. She has graduated from Ecole Nationale d’Inge´nieur de Tunis (ENIT) in 1999 and got a PhD of Ecole Supe´rieure des Communications de Tunis (Sup’- Com) and Universite´ Occidentale de Bretagne (UBO) in 28 Februrary 2006. She is currently a member in the research unit MEDIATRON_SUPCOM. Her research interests include wireless mobile networks: 3G and 4G, QoS provisioning, resource allocation management, mobility management, policy-based networking architecture.

Sami Tabbane is a professor at the School of Communications of Tunis (Sup’Com). He is a specialist in mobile radio communication systems. He graduated from the Ecole des Mines de Paris (France) in 1998 and obtained a doctorate from the Ecole Nationale Superieure des Telecommunications de Paris (ENST) in 1991. He began his career at France Telecom from 1992 to 1994 and was recruited by Sup’Com in 1994. He has conducted numerous missions for the ITU in the field of mobile network planning, management and training, and spectrum management in organizations regulations. He’s co-authors of several IEEE conference and journal papers. He is the co-author of ‘GSM Networks’ (Hermes, 1995) and ‘Engineering Services Telecommunications’ (Hermes-Lavoisier, 2005). He is the author of ‘Mobile Networks’ (Hermes, 1997),’ Manuel mobile networks (Artech House, 2000), Engineering of Cellular Networks’ (Hermes-Lavoisier, 2002). He served as a co-chair/session chair for several IEEE conferences.

123