Performance Study for Off-Grid Self-Backhauled Small Cells in Dense Informal Settlements

Parth Amin1, Nadew S. Kibret2, Edward Mutafungwa2, Beneyam B. Haile2, Jyri Hämäläinen2, Jukka K. Nurminen1

1Department of Computer Science & Engineering, Aalto University School of Science, Finland 2Department of Communications & Networking, Aalto University School of Electrical Engineering, Finland

{parth.amin, nadew.kibret, edward.mutafungwa, beneyam.haile, jyri.hamalainen, jukka.k.nurminen}@aalto.fi

Abstract – Urban (and suburban) informal settlements in emerging markets will be the fastest growing population hotspots in the next few decades. This presents a significant challenge in delivery of services, including affordable mobile Internet access. Operator-led upgrades through network densification (rollout of additional operator-maintained sites) are difficult to sustain in those areas due to low revenues, lack of fixed lines, energy scarcity, insecurity, and so on. This calls for alternative approaches in mobile network upgrades and operation. In this paper, we present capacity enhancement approach using user-deployed shared-access small cells in the dense informal settlements. To that end, we consider leveraging macro Long Term Evolution (LTE) networks to backhaul High Speed Packet Access (HSPA) small cells. As a case study, we present comparative network simulations based on an example informal settlement. The results of the study highlight the possibilities for cost-effective capacity upgrades for users in dense settlements for even a limited number of unplanned end-user small deployments and self-backhauling via existing macro sites. In the study, we also note possible system performance enhancements of the small cell backhaul link through improved antenna design, scaling of carrier bandwidth and introduction of traffic steering across HSPA and LTE layers.

Keywords — Small cells; self-backhauling; emerging markets; dense settlements

I. INTRODUCTION Mobile broadband technologies are increasingly the most common, and in many cases, the only economically-feasible means for providing broadband connectivity for the masses in emerging regions, such as, Africa, where the fixed-line penetration has remained virtually flat over the last decade [1]. Typically, the mobile broadband network coverage is mostly provided by 3G Wideband Code-Division Multiple Access (WCDMA) and High Speed Packet Access (HSPA) macro cellular networks [2]. The increased mobile broadband subscriptions, traffic growth and intensifying competition has prompted most operators in the region to upgrade their networks to evolved HSPA and increasingly Long Term Evolution (LTE) networks in major urban areas [2][3]. Network densification through rollout of new cell sites allows operators to increase reuse of their limited spectrum and provide needed capacity gains in urban areas, particularly in the fast expanding dense informal settlement areas [4]. However, rolling out of new sites in those settlements is complicated by lack of fixed lines for backhaul, energy scarcity, need for securing network assets at sites and limited average revenue per user (ARPU) to justify the additional investment [1]. This calls for alternative approaches for network densification and operation models suitable to this

environment.

In this paper we consider the alternative densification scenario through small cell deployment in the informal settlements. The HSPA macro site represents the legacy deployment with majority of user equipment (UE) in the settlements assumed to be HSPA-compliant. Radio Access Technology (RAT) enhancements through macro LTE upgrades are then implemented to cater for a minority but gradually expanding base of LTE UEs [2][3]. The HSPA small cells are then deployed to offload traffic from highly-loaded HSPA macro cells. Unplanned deployment of shared access small cells by end users (households, microenterprises etc.) provides a cost-effective network densification from the operators’ perspective and affordable connectivity from the end users’ perspective. Moreover, it potentially allows for novel business models that provide incentives (e.g. revenue share) for end users to deploy and maintain the shared small cells.

However, small cell powering and backhauling in these dense informal settlements presents a significant challenge. Traditional approaches for small cell backhauling, such as, wireline links (fiber, DSL etc.), fixed-wireless links (digital microwave, millimeter wave radio etc.) and satellite, require sufficient cable plant (for wireline links), carefully planned deployment by the operator and are relatively expensive [7]-[8]. In this paper, we consider self-backhauling of small cells through the use of macro LTE and LTE-Advanced enhancements to provide low-cost and flexible backhauling for the unplanned HSPA small cells in the informal settlements. Extensive simulations are carried out to verify the feasibility of the considered self-backhauling approach and observe the overall performance impact on the HSPA and LTE users in the network. Furthermore, the limited and/or unreliable access to power from the main electrical grid presents significant challenges in the operation of the small cells [9]. To that end, our study also reviews various powering options for the small cells in informal settlements and considers deployments that enable off-grid operation of the small cells, for instance, using renewable energy sources, such as, solar.

Key contributions of the paper include verifying the feasibility of using LTE as a self-backhauling technique for HSPA small cell, analyzing the impact of self-backhauling on existing LTE user equipment (UE) throughput and enhancement in small cell LTE backhaul link capacity to minimize the impact on HSPA UE throughput. The rest of the paper is arranged as follows. Section II introduces small cell concept briefly and presents the motivation for study of small cell deployment in informal settlements. In Section III, we describe the deployment scenario and possible enhancements

Part of this work was performed within the ICT&E project funded by the Aalto University School of Electrical Engineering programme on energy efficiency.

978-1-4799-4912-0/14/$31.00 ©2014 IEEE

for small cell backhauling. Finally Sections IV and V, we present the simulation results and conclusions.

II. MOTIVATION FOR STUDY

A. Background on Small Cells Small cells is an umbrella term for low-powered reduced-

range (tens to several hundred meters) base stations deployed to provide improved coverage and capacity in residential areas, enterprise environments or public spaces (both indoor and outdoor)[5]. Operator deployed small cells include: extensions to macrocells like remote radio heads, metrocells, microcells, picocells and open-access femtocells. On the other hand, indoor closed-access femtocells are deployed and operated by end-users in their place of residence, analogous to traditional Wi-Fi access points. In this paper, the focus is on user-deployed but open or shared access femtocells. Henceforth the term small cell is assumed to refer to this class of femtocells, and may be used interchangeably. Radio and network architecture aspects of small cells have been under development in 3GPP (Release 8 onwards), where terms Home NodeB and Home eNodeB are used to refer WCDMA/HSPA and LTE femtocells, respectively. In this paper we focus on Home NodeB’s for which the transmission power is limited to 20dBm. Yet, we note that the discussion is also valid for local area class pico base stations (up to 24dBm transmit power).

B. Motivation for End-User Deployed Small Cells End-user deployed small cells present a number of benefits

for both the mobile operator and end-users [5]. The benefits for the operator include additional capacity for overlaying macro-cellular network after offloading user to the small cells and increased customer loyalty through retention of family members or other local resident groups in home zone pricing promotions, improved user experience and creation of value added services that use small cell contextual information. Furthermore, operator benefits from reduction in the number of required new macro-cell sites to be rolled out and associated costs for installation and site maintenance, due to the fact that the small sites are self-deployed and maintained by the end users. Although the small cells are deployed autonomously by end users, the operator is able to efficiently to manage the small cells remotely through use of Self-Organizing Networks (SON) techniques [6]. The SON techniques enable operators to implement autonomous self-configuration of end-user deployed small cells, automated discovery of neighboring cells, self-optimization of small cell transmit powers (for e.g. interference avoidance and load balancing), optimization of handovers and recovery from operation failures (self-healing).

From the end-user perspective the small cell benefits may include: improved user experience in particularly in macro coverage dead spots and highly loaded hotspots; prioritized access to radio resources for end-users belonging to a Closed Subscriber Group (CSG) (e.g. family members); operator incentives (e.g. reduced service tariffs) when user is connected to small cell rather than macrocell; access to new small cell value-added services (e.g. location-based services) and reduced end-user device power consumption (smaller uplink transmission power) compared to macro case.

Figure 1 Self-backhauling of HSPA small cell deployments overlaid by HSPA/LTE macro cells

III. DESCRIPTION OF DEPLOYMENT SCENARIO

A. Macro and Self-Backhauled Small Cell Deployment Self-backhauling of small cells via the macro base stations

is an attractive solution for user-deployed small calls in dense informal settlements that are characterized by lack of fixed line infrastructure and very stringent cost constraints. The self-backhauling approach for network access nodes has previously been standardized for LTE Release 10 and IEEE 802.16m relays [10]. Furthermore, we previously investigated the possible use of relaying for cell-edge performance enhancements LTE-capable user equipment (LTE-UEs) in dense informal settlements [12]. Moreover, a recent draft proposal (IEEE 802.16r) has been initiated to specify use of WiMAX (WirelessMAN OFDMA) air interface to backhaul LTE small cells [11]. For this study, we consider a multi-RAT environment, whereby, HSPA small cells offload HSPA-UEs from HSPA macro cells, while the overlapping macro LTE coverage serves LTE-capable UEs, as well as, providing LTE connectivity for self-backhauling of HSPA small cells (seeFigure 1).

B. Small Cell Backhaul Enhancements In the considered deployment scenario of Figure 1, the macro LTE radio resources (physical resource blocks or PRBs) are shared between LTE-UEs and the HSPA small cell backhaul. This resource contention limits the capacity of the small cell backhaul which may in turn limit the achievable throughput for the HSPA-UEs served by a particular HSPA small cell. This backhaul bottleneck effect occurs when the loaded small cell capacity (including control plane traffic, IPSec overhead etc.) exceeds the provided backhaul capacity [7]. Therefore, our study considers various capacity enhancement strategies for eliminating the small cell backhaul bottlenecks in the self-backhauling configuration of Figure 1. The enhancement strategies consider the potential performance tradeoffs between provisioning of sufficient LTE resources to satisfy small cell backhaul capacity requirements and while ensuring acceptable quality of service for LTE-UEs. These enhancement strategies are outlined briefly below.

1) Multiple-Input Multiple Output (MIMO)

Utilization of higher order MIMO configurations could provide spectral efficiency gains for the small cell backhaul link (depending on channel conditions). This would enable small cell backhaul capacity needs to be potentially met with relatively lower resource allocation. In this study we consider 2x2 and 4x4 configurations for LTE Release 8 and additional gains obtained by Release 10 (LTE-Advanced) extensions to 8x8 configurations in the downlink [13].

2) Bandwidth Scaling or Carrier Aggregation This involves scaling the shared LTE macro resources by increasing the available transmission bandwidth. For LTE Release 8 transmission bandwidth scalability is possible up to 20 MHz, while Release 10 carrier aggregation enables LTE-UEs and small cell LTE backhaul connections are scheduled over multiple 20 MHz component carriers [13]. Addition of licensed spectrum bands presents a significant cost barrier, but recently proposed approaches, such as, use of unlicensed spectrum bands for LTE make this an attractive option [14].

3) Traffic Steering We define three network layers (with differing RAT, cell size and/or operating frequency) for the proposed deployment scenario of Figure 1. These are: (i) LTE macro layer, (ii) HSPA macro layer and (iii) HSPA small cell layer. Traffic steering can be utilized to optimally distribute offered load across different layers so as to balance load and/or meet particular performance requirements [15]. The traffic steering policies are operator-specific and could be enforced based on attributes, such as, service type and user class. In our considered deployment scenario of Figure 1, the LTE-UEs also possess HSPA capabilities and therefore can be steered across all three layers. Furthermore, the HSPA UEs can be steered between the HSPA macro and HSPA small cell layer (in case of overlapping coverage), while LTE backhaul for the HSPA small cells is only provided by the LTE macro layer (no steering possible). To that end, we consider two alternative traffic steering policies:

i. Small cell backhaul prioritized: This policy prioritizes the small cell backhaul by first allocating LTE macro layer resources to first satisfy capacity requirements for the HSPA small cell backhaul. The LTE-UEs are first camping in the HSPA layers (macro and small cell) and would only be steered to the LTE macro-layer if small cell backhaul capacity requirements are satisfied.

ii. LTE-UEs prioritized: In this case, the small cell backhaul links are allocated LTE macro layer resources only after the LTE-UEs camping in the LTE macro layer have sufficient resources to achieve a specified quality of service (QoS) objective (e.g. guaranteed bit rate). The steering of LTE-UEs from LTE macro to the HSPA layers is implemented in case better throughput is achieved in the HSPA layers due to better channel conditions and/or reduced cell load in HSPA layers.

C. Powering Options The powering of small cells is another key operational

consideration. Small cells deployed indoors may possibly be powered from the main grid. However, this option is limited in low-income settlements due to frequent outages in the power supply from the main grid and in many cases access to main

grid distribution lines may be unavailable at small cell site [9]. Therefore, indoor small cells require standby power sources or access to local microgrid source to avoid downtime when no power is available from the main grid. On the other hand, small cells deployed outdoors (e.g. rooftop) may conveniently harvest energy from renewable sources (solar, wind, etc.) and as a result also reduce operating costs due to powering. However, upfront cost and form factor aspects have to be considered for dedicated renewable energy systems for standalone loads, such as, small cells. For instance, for solar energy systems, the cost and size of the solar panel and energy storage has to be dimensioned to maximize availability for given small cell power consumption profiles and typical variation in daily insolation for a given area [16].

IV. PERFOMANCE STUDY METHODOLOGY

A. Simulation Scenario To exemplify a high-density urban informal settlement we

have used Hanna Nassif ward in Dar es Salaam, Tanzania, as a simulation study area. Hanna Nassif has an estimated population of 40000 people, living in a 1 km2 land area (see Figure 2). The area includes around 3000 (mostly single story) buildings and is located on a terrain with a topographical difference of 19 m. The small cells are deployed at random buildings by end-users in the service area. We consider two possible deployment scenarios: (a) indoor deployment; (b) rooftop deployment (akin to a television antenna). Indoor deployment enables small cells to provide indoor coverage and indoor-to-outdoor coverage for other UE in close proximity of the building. Rooftop deployed small cells provide increased range for outdoor coverage, but at the expense of reduced signal strength for indoor users (outdoor-to-indoor coverage) due to building penetration losses. Moreover, as noted in Section 0, rooftop deployment of the small cell is convenient for off-grid operation by harvesting energy from renewable sources.

B. System Simulation Parameters and Assumptions Our simulation study focuses on downlink performance. The radio coverage estimations are based on realistic three-dimensional building vectors and topographical data for the Hanna Nassif area and are evaluated using the dominant path model implemented in the WinProp propagation modeling tool [17] (see example of Figure 3). Static system-level simulations are then performed to investigate network performance, whereby, small cells are dropped at random building locations for each snapshot, while half of the UEs (HSPA- and LTE-UEs) are dropped in clusters around small cells and the rest of the UEs are dropped randomly over whole area. Throughput results are found through mapping the SINRs results using a modified Shannon formula [18]-[19]:

min

minmax 2

0

,2 min , log 1PRB PRB eff

eff

SINR SINR

TP SINR SINRSINRN BW S BWSINR

τ

<⎧⎪ ⎛ ⎞⎛ ⎞= ≥⎨ +⎜ ⎟⎜ ⎟⎪ ⎜ ⎟⎜ ⎟⎝ ⎠⎝ ⎠⎩

(1)

where τ reflects the amount of time resource allocated for the transmission, PRBN is the number of PRBs, and PRBBW is the

bandwidth per PRB. Note also that Smax is the maximum spectral efficiency, minSINR is the minimum required SINR,

effBW adjusts bandwidth to fit with LTE the system bandwidth efficiency and

effSINR adjusts for the SINR implementation efficiency. The macro and small cell parameter and system simulation assumptions follow commonly used 3GPP guidelines [20]-[21], and are listed in Table I. The SINR-to-throughput mapping parameters for different MIMO configurations are based on previous link-level simulations campaigns [22].

Figure 2 Aerial view of the Hanna Nassif area

Figure 3 Example comparative pathloss maps for a small cell deployed indoors (left) and on rooftop (right) in the Hanna Nassif area (WinProp simulation result)

V. SIMULATION RESULTS AND DISCUSSIONS A total of seven deployment scenarios are initially considered in the simulation study. This includes the three scenarios where three macro sites plus 10, 30 and 60 small cells are deployed indoors in randomly selected buildings in the service area. Another three scenarios are created by deploying the same number of small cells on rooftops of the same buildings.

The conventional macro-only deployment, that is, with no small cells, is used as the baseline scenario. Figure 4 shows the HSPA-UE throughput gains (relative to macro-only case) after deployment of rooftop and indoor small cells, respectively. Performance is considered on 10th, 50th and 90th percentile of the throughput CDF representing the cell edge, median and cell center performances. In all cases, throughput gains increase with dense small cell deployment particularly for rooftop deployments, which enable relative higher macro-to-small cell offloading.

TABLE I. SIMULATION PARAMETERS AND ASSUMPTIONS

Parameter Values/Assumptions Air Interface HSPA FDD, LTE FDD Carrier Freq./ Bandwidths

LTE: 2600 MHz / 10 MHz and 2 x 10 MHz HSPA: 2112.5 MHz / 5 MHz

Simulation Radio propagation modeling (WinProp) [17], Static system level simultion (Matlab), 2.5 m resolution

SINR-throughput mapping [22]

2×2 MIMO 4×4 MIMO 8×8 MIMO SINRmin (dB)

-10 -10 -10

BWeff 0.42 0.40 0.33 SINReff 1.1 1.5 2.1 Smax (b/s/Hz)

7.67 14.5 25.6

Macro Paramaeters Macro Sites Site1 (3 cells) Site2 (3 cells) Site3 (1 cells) Transmit Power LTE: 46 dBm, HSPA: 37.8 dBm (10% for CPICH) Antenna Height 10 m 15 m 10 m Antenna Patterns Kathrein 741984 Sector Azimuths 20°,140°,260° 0°, 120°, 240° 250° Intersite distance Site 1-2: 955 m, Site 2-3: 880 m, Site 3-1: 585 m

Small Cells Parameters SC number Three cases considered: 10, 30 and 60 small cells Location/Height Randomly deployed, Indoor:1.5 m, Rooftop: 4 m or 7 m LTE backhaul Antenna Gain: 0 dBi, Noise Figure: 9 dB, Antenna

config: 2x2, 4x4, 8x8 MIMO HSPA access Tx power 20 dBm (10% for common pilot channel),

Omni, 0 dBi antenna gain, 3 dB cell selection bias UE Parameters

UE height/location

1.5 UE height, 50% HSPA-UEs/LTE-UEs dropped randomly in whole area (both indoor and outdoor), 50% HSPA-UEs/LTE-UEs cluster-dropped within 40m radius of small cells (both indoor and outdoor)

UE number 45 HSPA-UEs or LTE-UEs in service area LTE-UEs Noise Figure: 9 dB, 2 × 2 MIMO, Ant. Gain: 0 dBi HSPA-UEs Omni.; 0 dBi gain, -99dBm noise; 15Code HSDPA

Buildings and Fading Characteristics Shadow Fading Shadow fading: WinProp ray tracing Fast fading Rician (for small cell-eNB, K =2), Rayleigh (for UEs) Buildings Variable dimensions, heights 3-6 m, penet. loss: 10 dB

Furthermore, we compared the fairness of achievable throughput for HSPA-UEs by using Jain’s fairness index (J) obtained by [23],

( ) ( )∑∑

=

=

⋅= N

i i

N

i iN

tN

ttttJ

12

2

121 ,,, … (2)

where N is the number of users and ti is the throughput achieved by the ith user. The value J = 1 (most fair case) is occurs when all N users experience the same throughput and it decreases to 1/N in the least fair case. As depicted in Table II,

throughput gains are more equitably distributed when the small cells are deployed on rooftop with relatively high offloading from macro cells achieved compared to indoor small cell case. This is attributed to the fact that for indoor small cells, the highest throughputs are only achievable for HSPA-UEs located in houses with deployed small cells.

Figure 4 HSPA-UE throughput gains (multiple) over the baseline macro-only scenario TABLE II. AVERAGE JAIN’S THROUGHPUT GAIN FAIRNESS INDEX

Macro only # Indoor Small Cells # Rooftop Small Cells

0.29 10 30 60 10 30 60

0.29 0.26 0.25 0.40 0.38 0.40

TABLE III. SIMULATED CASES OF LTE BACKHAUL FOR SMALL CELLS

Case 1 (Baseline)

Case 2 Case 3 Case 4

MIMO 2x2 4x4 8x8 2x2 Carrier (MHz) 10 10 10 2x10

TABLE IV. AVERAGE PERCENTAGE OF SMALL CELLS (FOR 60 SMALL CELLS) WITH BACKHAUL CAPACITY FULLY SATISFIED BY LTE CONNECTION

Small cell Backhaul Prioritized LTE-UEs Prioritized

LTE-UEs: HSPA-UEs

0:45 (Initially) 5:40 22:23 40:5

Case 1 87.4% 96.1% 91.3% 78.4% Case 2 99.1% 99.5% 96.4% 87.8% Case 3 99.9% 100% 98.4% 90.8% Case 4 99.9% 100% 99.9% 99.8%

For the rest of the paper we focus our study on the rooftop deployment due to aforementioned performance gains and possibility of off-grid operation (noted in Section 0). To that end, the performance of LTE-UE and LTE backhaul for small cells is simulated for four different combinations (cases 1-4 of Table III) of MIMO configurations for small cell backhaul (2×2, 4×4, 8×8) and carrier bandwidths (10 MHz, 2×10 MHz). Case 1 with 2×2 MIMO and 10 MHz carrier bandwidth for small cell backhaul connection is considered the baseline case (no enhancements).

The 4 cases of Table III are studied under two different traffic steering policies (described in Section III.B.3) that

prioritize either small cell backhaul or LTE-UEs. The objective of the simulation study was twofold: first to understand how different LTE link capacity enhancements and traffic steering policies can enable the HSPA small cell backhaul needs to be met (Table IV); secondly to observe how these enhancement and LTE resource sharing (with small cell backhaul) impacts normal LTE UEs (Table V). In this study the HSPA small cell backhaul capacity is served cell capacity of the small cell scaled by a factor 1.26 to account for overheads of the Iu-h interface between HSPA small cell and core network [7]. From the LTE macro layer perspective, it is obvious that denser HSPA small cells deployments creates greater demand for LTE connections to cater for small cell backhauling. Therefore, for sake of brevity the simulations only consider the 60 small cells deployment scenario.

Table IV shows the percentage of HSPA small cells whose backhaul capacity needs are fully satisfied by the LTE macro networks. When the small cell backhaul is prioritized the backhaul capacity is initially evaluated assuming all 45 UEs are camping in HSPA layers (macro or small cell) and LTE capable UEs are steered to LTE macro layer in case of spare capacity after small cell backhaul is served. It is noted that the higher order MIMO configurations for small cell backhaul (improved spectral efficiency) allows for backhaul capacity requirements to be fully met without the need for extra LTE spectral resources.

On the other hand, when the LTE-UEs are prioritized, there is a notable dependency on the fraction of 45 UEs that are LTE-capable (see Table IV). It is clear that in the case that a significant majority of the UEs are LTE-capable the approach of scaling carrier bandwidths or aggregating carriers becomes necessary to meet small backhaul capacity needs and serve LTE-UEs with an acceptable quality of service (in this case set a minimum bit rate target of 512 kbps). However, when the fraction of HSPA-only UEs is higher there is an improvement in the percentage of satisfied small backhaul capacity demands as the small backhaul capacity requirement is reduced from the previous case of prioritizing small cell backhaul (whereby all 45 UEs assumed to be camping initially in the HSPA layers). This implies that some optimum steering of LTE-capable UEs from the HSPA small cell layer to LTE macro layer would enable small cell backhaul capacity demand to be satisfied when the aforementioned LTE-UEs are prioritized.

We can also observe the performance from the perspective of LTE-UEs. The throughput CDFs for UEs (HSPA- and LTE-UEs) assuming baseline implementation of small cell backhaul (Case 1 of Table III) is shown in Figure 5. As expected the HSPA-UE throughputs improve as the fraction of LTE-capable UEs increases, and vice-versa. This is due to the strong dependency in resource utilization between the HSPA small layer and LTE macro layers, as the higher load on HSPA small cells would create demand for more LTE macro resources to satisfy the small backhaul capacity needs.

Using Case 1 as the baseline we evaluate the LTE-UE throughput gains when MIMO and carrier bandwidth enhancements of (Cases 2-4 of Table III) are applied (see Table V). It is noted that that the most significant gains are when the additional 10 MHz carrier is provided. The MIMO

enhancements to small cell backhaul become less significant when LTE capable UEs are prioritized and make up a larger fraction of the served UEs.

10-3

10-2

10-1

100

101

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1UE Throughput CDF

Throughput[Mbps]

CD

F

HSPA-UEs (SC backhaul prioritized)

HSPA-UEs (5 LTE-UEs prioritized)HSPA-UEs (22 LTE-UEs prioritized)

HSPA-UEs (40 LTE UEs prioritized)

LTE-UEs (SC backhaul prioritized)

LTE-UEs (5 LTE UEs prioritized)LTE-UEs (22 LTE UEs prioritized)

LTE-UEs, (40 LTE UEs prioritized)

Figure 5 CDFs of HSPA-UE and LTE-UE throughput for baseline Case 1 of Table III (Total UE number = 45; SC: Small Cell). TABLE V. LTE-UES THROUGHPUT GAIN COMPARED TO BASELINE (CASE 1)

LTE-UEs: HSPA-UEs

Percentile Case 2

Case 3

Case 4

Small cell Backhaul

Prioritized

0:45 (Initially)

10th 1.25 1.53 2.58 50th 1.16 1.36 2.52 95th 1.02 1.09 2.28

LTE-UEs Prioritized

5:40

10th 1.39 1.64 3.14 50th 1.42 1.64 3.20 95th 1.20 1.28 2.47

22:23

10th 1.05 1.13 2.04 50th 1.11 1.29 2.87 95th 1.09 1.12 2.41

40:5

10th 1.01 1.02 1.83 50th 1.04 1.07 2.68 95th 1.02 1.08 2.24

VI. CONCLUSIONS AND FUTURE WORK This paper demonstrates through extensive simulation

study that deployment of shared access small cells in densely populated informal settlements can provide notable throughput gains even for limited levels of small cell penetration. The simulation study also considered performance issues for flexible and cost-effective self-backhauling of HSPA small cells via LTE macro sites. These observations highlight the potential of small cells as a viable and cost effective tool for increased access to affordable broadband connectivity in dense informal settlements with challenging conditions for supporting increased site rollout by operators. This self-backhauling approach can be extended to consider other RAT combinations in the backhaul and access segments. For instance, 5G RATs currently under standard could be used to provide even higher capacity self-backhauling for HSPA, LTE or multi-radio small cells. Future work is required to investigate joint radio resource management schemes across different layers and link segments (access and backhaul) for the deployment scenario considered in this study. Furthermore, the authors are currently conducting on SON algorithms for optimum load balancing across different layers and energy-sustainable operation of off-grid

small cells in this context. This ongoing work considers the impact of intermittent and limited battery capacity of renewable energy sources (e.g. solar) for powering self-backhauled small cells.

ACKNOWLEDGMENT Hanna Nassif GIS data was kindly provided by Prof. R.

Sliuzas of ITC-Faculty of Geo-Information Science & Earth Observation, University of Twente.

REFERENCES [1] GSMA, “Sub-Saharan Africa Mobile Observatory 2012” Nov. 2012. [2] Ericsson, “Mobility Report,” June 2014. [3] GSA, “Evolution to LTE Report,” GSA Market Update, Dec. 2013. [4] UN-HABITAT, “Slum Dwellers to double by 2030,” April 2007. [5] Small Cell Forum, “Small Cells – What is the Big Idea?,” SCF

Release 1, Doc. 030.01.01, February 2012. [6] 3GPP, “Self-configuring and self-optimizing network use cases and

solutions (Release 9)”, TR 36.902 v9.3.1, April 2011. [7] NGMN, “Small Cell Backhaul Requirements,” NGMN Alliance

White Paper, June 2012. [8] D. Bojic et al, “Advanced wireless and optical technologies for small

cell mobile backhaul with dynamic software defined management,” IEEE Comm. Mag., vol. 51, no. 9, pp. 86-93, September 2013.

[9] A. Eberhard et al, “Africa’s Power Infrastructure: Investment, Integration, Efficiency” World Bank Report (with IBRD), 2011.

[10] C. Hoymann, et al, “Relaying operation in 3GPP LTE: challenges and solutions,” IEEE Com. Mag., vol. 50, pp. 156-162, February 2012.

[11] IEEE 802.16r, “Architecture and Requirements for Small Cell Backhaul,” 2013.

[12] B. Haile, E. Mutafungwa and J. Hämäläinen, “Use of Coordinated multipoint transmission for relaxation of relay link bottlenecks,” VTC2014-Spring 18–21 May 2014, pp. 1-5.

[13] E. Dahlman, S. Parkvall and J. Sköld, “4G LTE/LTE-Advanced for Mobile Broadband,” 1st ed., Academic Press, Oxford, 2011.

[14] Huawei, “U-LTE: Unlicensed spectrum utilization of LTE,” 2014. [15] P. M. Luengo, D. Laselva, R. Barco and P. E. Mogensen, “Adjustment

of mobility parameters for traffic steering in multi-RAT multi-layer wireless networks,” EURASIP J. Wireless Comm. and Networking 2013: 133, 14 p., 2013.

[16] J. W. Kimball, B. T. Kuhn, and R. S. Balog, “A system design apporach for unattended solar energy harvesting supply,” IEEE Trans. Power Electron., vol. 24, no. 4, pp. 952–962, April 2009.

[17] http://www.awe-communications.com/ [18] F. Haider, E. Hepsaydir and N. Binucci, "Performance analysis of a

live mobile broadband-HSDPA network," in IEEE VTC Spring 2011, May 2011, 5 p.

[19] P. Mogensen et al, "LTE system capacity compared to Shannon bounds," in IEEE VTC Spring 2007, May 2007, 5 p.

[20] 3GPP TR 25.814, “Physical layer aspect for evolved Universal Terrestrial Radio Access (UTRA).”

[21] 3GPP, TR 36.814, “Evolved Universal Terrestrial Radio Access (E-UTRA); Further advancements for E-UTRA Physical layer aspects.”

[22] E. Lahetkangas, et al, “On the performance of LTE-Advanced MIMO: How to set and reach beyond 4G targets,” 18th European Wireless Conference, 18-20 April 2012, 6p.

[23] R. Jain, D. Chiu, and W. Hawe, “A Quantitative Measure Of Fairness And Discrimination For Resource Allocation In Shared Computer Systems”, DEC Research Report TR-301, September 1984.