exploiting multi-homing in hyper dense lte small-cells...

Exploiting Multi-homing in Hyper Dense LTESmall-Cells Deployments

Abdellaziz Walid #1, Essaid Sabir 2, Abdellatif Kobbane #3, Tarik Taleb >4, Mohammed El Koutbi #5,# ENSIAS, Mohammed V University, SIME Lab/MobiTic, Rabat, Morocco.

UBICOM Research Group, ENSEM, Hassan II University, Casablanca, Morocco>Aalto University, Finland

1 [email protected], 2 [email protected], 3 [email protected], 4 [email protected],5 [email protected]

AbstractIt is expected that in two-tier LTE heterogeneousnetworks, an extensive deployment of small cell networks (SCNs)will take place in the near future, especially in dense urbanzones; hence a hyper density of SCNs randomly distributedwithin macro cell networks (MCNs) will emerge with many over-lapping zones of neighboring SCNs. Therefore, the problems ofinterferences in co-channel deployment will be more complicatedand then the overall throughput of downlink will substantiallydecrease. In order to mitigate the effect of interferences in a hyperdensity of SCNs scenarios, a solution based on a fully distributedalgorithm for sharing time access to SCNs and multi-homingcapabilities of macro cellular users is proposed to improve theoverall data rate of downlink and at the same time to satisfy QoSthroughput requirements of macro and home cellular users. Ourtentative scheme will also reduce the signaling overhead due tothe absence of coordination among small base stations (SBSs)and macro base station (MBS). Results validate our solutionand show the improvement attained in a hyper density of SCNswithin MCNs compared to open, closed and shared time accessmechanisms based on single network selection.

Index TermsSmall Cell Network (SCN), Macro Cell Network(MCN), Base Station (BS), Co-channel, Long Term Evolution(LTE), Multi-homing.

I. INTRODUCTION

The demand for ubiquitous wireless access network andthe need for high data rates are growing significantly withthe omnipresence of demanding applications such as onlinegaming, mobile TV, Web 2.0, streaming contents, ... etc. Thesenew services require a high rate on the bit per square kilometerthat is expected to be delivered by the next generation wirelessnetworks like LTE and WIMAX.

The operators are urged to create new ways for improvingtheir coverage, enhancing their networks capacity, and reduc-ing their costs of their mobile networks. A promising way tosolving this problem is the use of SCNs that are considered asa novel networking paradigm based on the idea of deployingshort-range, low-power, and low-cost base stations operatingin conjunction with the main MCN infrastructure. SCNs areexpected to provide high data rates for the next generation 5Gnetworks, tolerate offloading traffic from the MCN, providing

This work was supported by the Mobicity Project funded by The MoroccanMinistry of Higher Education and scientific research.

high capacity to homes, enterprises, or urban hot spots. SCNsare also envisioned to pave the way for a plethora of newwireless services and cover a range of cells such as micro,pico, metro, and femto cells as well as advanced wirelessrelays, and distributed antennas that can be installed practicallyeverywhere.

SCNs have also attracted the attention of standardizationbodies, e.g., the 3rd Generation Partnership Project (3GPP)[1], Long Term Evolution (LTE) and LTE-Advanced (LTE-A). However, the utilization of SCNs in a realistic environmentfaces numerous technical challenges and issues that need tobe dealt with on various network layers.

II. MOTIVATION AND RELATED WORKThe success of using SCNs will lead to a proliferation of

SBSs within MCNs, especially in hyper dense urban zoneslike downtowns, malls, residences, metropolis, ...etc, as shownin Fig.1. Accordingly, a large number of overlapping zones

Fig. 1. Hyper-dense SCNs deployments scenario.

of SCNs shows up. The extent of the overlapping dependson the closeness of SBSs to each other (i.e the totallyoverlapping zones covered by SBSs installed in the flats thattake the same position in each floor of the building, whilepartially overlapping zones occur among neighboring SCNs).The overlapping zones of SCNs have a serious impact on thedegradation of the overall capacity of two-tier heterogeneousnetworks due to the gravity of interferences, particularly inco-channel deployments.

In two-tier heterogeneous networks, interferences are clas-sified as follows: Cross-tier interferences are caused by an element of the

small cell tier to the macro cell tier and vice versa (seeFig.2).

Co-tier interferences occur between elements of the sametier, for example, between neighboring small cells. (seeFig.3).

Fig. 2. Cross-tiers interferences in LTE two-tier networks.

Fig. 3. Co-tiers interferences in LTE two-tier networks.

Furthermore, the selection of an access control mechanism toSCNs has dramatic effects on the performance of the overallnetwork, mainly due to its role in defining the degree ofinterferences. Different approaches have been proposed: Closed access: only a subset of the users (home users),

which is defined by the small cell owner, can connect tothe SCN.

Open access: all customers of the operator have the rightto make use of any SCN.

In closed access, only registered home users can communi-cate with the SBS. It seems to be efficient to the home user,however it results in severe cross-tier interference from nearbymacro users to home users in the uplink and to nearby macrousers from SBS in the downlink. To mitigate these interfer-ences in closed access, previous studies have considered powercontrol [2], fractional frequency reuse techniques [3], and aspectrum sensing approach [4]. An alternative is to providean access to the nearby macro users that cause or experiencestrong interference to the SCN. This is known as open access.

In the case of uplink, open access is generally preferred byhome users and macro users [5]. In the case of downlink, theclosed access mode is favorable for home users while macrousers prefer the open access mode. To resolve the conflict inthe case of downlink, a shared time access has been suggestedas a compromise between home users and macro users tomaximize the network throughput subject to a network-wideQoS requirement [6]. All these cited works are adapted to asingle network selection in which macro indoor cellular usershave the possibility to use the service of the MBS in closedaccess policy to SCNs or to use the service of a single SBSin open access and shared access policy. This is due to thescenarios of simulation which assume that each SBS does notoverlap with another SBS.

Nonetheless, in hyper dense urban areas, where SBSs aredeployed randomly, a large number of overlapping zones ofSCNs will appear and a lot of macro indoor cellular userswill be inside these zones. Consequently, they face difficultiesin taking their decisions, especially in open mode and sharedmode access policy in the way that each macro indoor cellularuser will be able to select a single SBS among all availableSBSs implied in the overlapping zones.

In the single network selection, a cellular user takes itsrequired bandwidth from the best wireless access networkavailable at its location. The selection decision of a cellularuser is based on a predefined criterion like RSS [7] andavailable bandwidth [8] to choose the best available wirelessaccess network. One drawback of single network selection isthat the available resources from different networks are notfully exploited. Another drawback is that an incoming call isblocked if no network in the service area can individuallysatisfy the required bandwidth by that call. In addition, acongestion network occurs if all cellular users choose the bestnetwork at the same time [9].

Oppositely, with multi-homing capabilities, a cellular usercan maintain multiple simultaneous associations with differentnetworks. Therefore, the multi-homed cellular user obtains itsrequired bandwidth from all networks BS(s)/AP(s) availableat its location using its multi-homing capabilities. This hasthe benefit of supporting applications with high required datarate through aggregating the offered resources from differentnetworks and using multiple threads at the application layer.Also, it allows for mobility support, since at least one ofthe used radio interfaces will remain active during the callduration. Moreover, the multi-homing concept can reduce thecall blocking rate and improve the system capacity. A largeset of research work has been conducted to support the multi-homing access. Convex optimization formulation in [10] forboth Constant Bit Rate (CBR) and Variable Bit Rate (VBR)services is used in multi-homing access. Multi-homing ofusers to access points in WLANs in [11] has been studiedusing a potential game model and study its equilibrium. Thework in [12] deals with a cost minimization problem fora multi-homed mobile terminal that downloads and plays aVideo-on-Demand (VoD) stream. Another work [13] modelthe communication between multi-homed terminals as a multi-

criteria non-cooperative game so as to achieve performance-cost decision frontiers.

The major contributions of this paper can be summarizedas follows: Our model takes into account the study of downlink in

a high density of SCNs with many overlapping zonesof neighboring SBSs for LTE two-tier heterogenous net-works.

A solution based on a fully distributed shared timeaccess algorithm and multi-homing access capabilitiesof macro cellular users (Multi-homed devices) has beenused to improve the overall network throughput and QoSthroughput requirements of cellular users.

Our solution requires no coordination among SBSs andMBS and therefore reduces the huge signaling overheadof the whole system in hyper dense zones.

III. SYSTEM MODEL ANALYSIS

A. Two-tier Cellular Network Model

In our model, we focus the study on the downlink ofan LTE system. We assume that orthogonal multiple accessuses TDMA scheduling or OFDMA, consequently no intra-cell interference is raised. We consider a single LTE macrocell M with radius RM centred at an MBS (Macro BaseStation) and a variable number of LTE SBSs (called FemtoBSs) Ns distributed uniformly within the area covered bymacro cell M. The MBS and SBS use fixed transmit powersof PM and Ps, respectively. Nm macro cellular users arerandomly distributed within the area of macro cell M andns,i users are randomly distributed within the serving SBSi.We consider a co-channel deployment when all SCNs and theMCN use the same band frequency. In our study, we take intoconsideration the cross-tier interferences between each SCNand MCN, co-tiers interferences among SCNs and a path lossand propagation models in order to estimate the SINR in thedownlink for each user. For more details and to study differentaccess mechanisms, we divide all the users in three categories,home users or SBS owners, macro indoor cellular users andmacro outdoor cellular users.

Nm = Nm(out) +Nm(in) (1)

Nm: represents the total number of macro cellular users.Nm(in) : represents the number of macro indoor cellular userswho are under SBS and MBS coverage.Nm(out) : represents the number of macro outdoor cellularusers who are under MBS coverage only.

Nm(in) =

Nsi=1

ni (2)

ni : represents the number of macro indoor users in the areaof a SBSi.Ns : represents the total number of SBSs within the MCN.ns,i : represents the number of home users in the area of aSBSi.

B. SINR analysis in closed and open access policy

In our analysis, the estimation of the received SINR moutof a macro outdoor user mout in the downlink, when the macrouser is interfered from all the adjacent SCNs, is expressed bythe following equation:

mout =PM,kGm,M,k

N04f +Nsj=1

Psj ,kGm,sj ,k

(3)

where PM,k is the transmit power of serving macro cell Mon sub-carrier k. Gm,M,k is the channel gain between macrouser m and serving macro cell M on sub-carrier k. Similarly,Psj ,k is the transmit power of neighboring SBSj on sub-carrier k. Gm,sj ,k is the channel gain between macro user mand neighboring SBSj . N0 is a white noise power spectraldensity, and 4f is sub-carrier spacing.The estimation of the received SINR in the downlink iminof a macro indoor user min, located inside the coverage of aSBSi is given by :

imin =

PM,kGm,M,k

N04f+

Nsj=1

Psj ,kGm,sj ,k

Closed access

Psi,kGh,si,k

N04f+PM,kGh,M,k+

Nsj 6=i

Psj ,kGh,sj ,k

Open access(4)

Psi,k is the transmit power of SBSi. Gh,si,k is the channelgain between home user h and the SBSi.In the case of a home user h of a SBSi interfered the MCNand adjacent SCNs, the received SINR in the downlink ih canbe given by :

ih =Psi,kGh,si,k

N04f + PM,kGh,M,k +Nsj 6=i

Psj ,kGh,sj ,k

(5)

C. Data rate analysis in closed and open access

Having estimated the SINR, we can now proceed with thedata rate calculation. The practical data rate rmout of macrooutdoor user mout can be given by the following equation :

rmout =

4f.Bdl. log2(1+mout )

NmClosed access

4f.Bdl. log2(1+mout )NmNm(in)

Open access

(6)

Bdl : is the bandwidth occupied by the data sub-carriers inthe downlink. : is a constant for target Bit Error Rate(BER),and defined by = 1.5ln(5BER) .The practical data rate rimin of macro indoor user min, locatedinside the coverage of a SBSi can be given by the followingequations :

rimin =

4f.Bdl. log2(1+

imin

)

NmClosed access

4f.Bdl. log2(1+imin

)

ns,i+niOpen access

(7)

The practical data rate rih of a home user h in SBSi can begiven by the following equations:

rih =

4f.Bdl. log2(1+

ih)

ns,iClosed access

4f.Bdl. log2(1+ih)

ns,i+niOpen access

(8)

IV. OUR SOLUTION : SHARED TIME AND MULTI-HOMINGACCESS POLICY

A. Shared time access policy analysis

We consider the shared time access where a SBSi allocatesi fraction of time-slots to home users and the remaining(1 i) fraction of time-slots to macro indoor users in thedownlink. Due to the absence of the coordination among SBSsand MBS, each SBS tries to maximize selfishly its through-put capacity of downlink and satisfies the QoS throughputrequirements of home users. The time-slot allocation problemto maximize the overall network throughput of each SBSi isformulated as :

CtotalSBSi = i.

ns,ih=1

rih + (1 i).nim=1

rimin

= iTh,itotal + (1 i)T

min,itotal (9)

i [0, 1[In the shared time access, we have :

rih =4f.Bdl. log2(1 + ih)

ns,i(10)

rimin =4f.Bdl. log2(1 + imin)

ni(11)

Maximize iTh,itotal + (1 i)Tmin,itotal subject to

i.Th,itotalns,i

ih and (1 i).T

min,i

total

ni im

-ih : represents the average throughput requirement of ahome user h in SBSi.-im : represents the average throughput requirement of amacro user m in SBSi.

Theorem 1The optimal value i of the time allocation of a SBSiis given as :- if Th,itotal T

min,itotal

S1 = i = 1

im.ni

Tmin,itotal(12)

this solution S1 is feasible if i ih.ns,i

Th,itotal

- if Th,itotal < Tmin,itotal

S2 = i =

ih.ns,i

Th,itotal(13)

-

If i = 1, it will act as a closed access SBS.The algorithm proposed is based on Theorem 1 and

executed by each SBS to search the optimal valuesof sharing time access i for each SBSi in order tomaximize the total throughput of users inside SBS, andto satisfy the QoS throughput requirements of home usersand existing macro users inside the coverage of SCNs.

Algorithm 1 : Fully Distributed Algorithm for SharingTime Accessi =1 (closed mode is activated by default in each SBS)- each SBS sensing periodically for macro users in thearea of its coverageif (Exists)

Input parameters : ni, ns,i, ih, im ;

get Th,itotalget Tmin,itotalif (Th,itotal T

min,itotal ) Then

i = 1 im.ni

Tmin,i

total

if (i ih.ns,i

Th,itotal) Then

i = iend if

elsei =

ih.ns,i

Th,itotal

end if

end if

B. Multi-homing access analysis of macro cellular users

In the previous section, the shared time access policywas analyzed. In this section, we add to this analysis themulti-homing capabilities of macro cellular users. With thesecapabilities, macro cellular users located in overlapping zonesof SCNs can maintain multiple simultaneous associationswith different SCNs due to their multiple homogeneous radiointerfaces. Hence, in multi-homing radio resource allocation,the macro cellular user obtains its required traffic from allavailable SBSs at its location using its multiple homogeneousinterfaces. This has the advantage of supporting applicationswith high required data rate through aggregating the offeredresources from different networks. The total downlink through-put capacity of a macro indoor cellular user Tmin with multi-homing capabilities located in an overlapping zone covered bya set of SBSs noted Nover is formulated as:

Tmin =

iNover(1 i ).

4f.Bdl. log2(1 + imin)ni

=

iNover(1 i ).rimin (14)

V. RESULTS AND SIMULATIONS

The simulations are event-based and developed accordingto 3GPP standards. The simulations scenario is given inFig.4. The plotted values are an average of 10000 independentsimulations using Matlab. The assumed system parameter forthe simulations is given in Table 1.

1000 800 600 400 200 0 200 400 600 800 10001000

800

600

400

200

0

200

400

600

800

1000

X [m]

Y [m

]

Macro BS

Area Macro

Macro user

Area Small cell

Small BS

Home user

Fig. 4. Simulation of a hyper density deployments of SCNs scenario.

TABLE ISIMULATION PARAMETERS

Parameter ValueMacro cell Radius (Rm) 1000 mSmall cell Radius (Rs) 75 mNumber of macro cellular users (Nm) 200Number of home users 1-2Frequency 2 GhzMBS Power PM 25 WSBS Power Ps 200 mW

Channel gain G 10PL10

PL (Path loss) See formulas in [1]Indoor Walls Loss (Liw) 5 dBOutdoor Walls Loss (Low) 20 dBBandwidth of downlink Bdl 20 MHzModulation Scheme 64 QAMSub-carrier Spacing (f) 15 KHzBit Error Rate (BER) 106

White noise power density (N0) -174 dBm/Hzhome user QoS throughput requirements ih 1 Mbpsmacro user QoS throughput requirements im 0.1 MbpsNumber of homogenous interfaces of macrocellular users

4

0 40 80 120 160 200 240 2800

1

2

3

4

5

6x 10

9

Number of small cells

Tota

l D

ow

nlin

k T

hro

ughput C

apacity (

bps)

Shared Access with multihoming

Shared Access with single network selection

Open Access with single network selection

Closed Access

Fig. 5. Total throughput capacity of downlink.

As depicted in Fig.5, the overall throughput capacity of

downlink reaches its high level when the shared access policyis used with multi-homing capabilities of macro cellular users.Shared access with multi-homing mitigates the cross-tiersinterferences caused by SBS to macro indoor users in closedmode and offloads the MCN. Furthermore, it keeps the QoSthroughput parameters of home users degraded in the openmode. Moreover, multi-homing improves the QoS throughputparameter of macro indoor users in overlapping zones withthe increasing of SCNs compared to the shared access.

40 80 120 160 200 240 2800

2

4

6

8

10

12

14

16x 10

8


Tota

l T

hro

ughput C

apacity o

f M

acro

Indoor

Users

(bps)




Closed Access

Fig. 6. Total throughput of indoor macro cellular users.

Fig.6 shows the total throughput of macro indoor users inthe downlink. Shared mode with multi-homing is the suitablemode for downlink because it mitigates the interferencescaused by SBSs to macro indoor users in the closed mode andthe degradation of throughput capacity of macro indoor usersin the open mode. What is more, multi-homing capabilitiesof macro indoor users in overlapping zones of SCNs increasethe throughput of macro indoor with the increasing of SCNscompared to the shared access. For the reason that the greaterthe number of SCNs, the greater the number of macro indoorusers will be inside overlapping zones.

40 80 120 160 200 240 2800

1

2

3

4

5

6

7

8

9

10

x 107


Tota

l T

hro

ughput C

apacity o

f M

acro

Outd

oor

Users

(bps)




Closed Access

Fig. 7. Total throughput of outdoor macro cellular users.

As depicted in Fig.7, the open access is the appropriatemode for macro outdoor users in downlink because it offloads

the MCN by decreasing the number of macro users usingthe service of MCN and increasing the throughput capacityof macro outdoor users. But, this will be at the expenseof QoS throughput requirement of home cellular users thatwill be degraded in this access mode. In shared and closedmode, the total throughput capacity decreases due to theincrease of macro indoor users at the expense of macrooutdoor users by increasing the number of SCNs within MCN.Nonetheless, the shared access with multi-homing mode givesbest results compared to the closed and shared access due tothe simultaneous connections of macro cellular users insideoverlapping zones of SCNs.

40 80 120 160 200 240 2800

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5x 10

9


Tota

l T

hro

ughput C

apacity o

f H

om

e U

sers

(bps)




Closed Access

Fig. 8. Total throughput of home users.

As shown in Fig.8, the preferred access policy for homeusers in the downlink is closed access. Shared access policyand shared access with multi-homing are also suitable modesfor home users since they take into consideration the QoSthroughput requirements of home users.

VI. CONCLUSION

In this paper, we have studied the problem of co-channelinterferences in hyper dense SCNs deployments within MCNs.Macro cellular users inside overlapping area coverage of SCNssuffer from high degradation of downlink throughput due tosevere interferences from SBSs. Our proposed solution whichcombines shared time access policy and multi-homing capa-bilities of macro cellular users to SCNs is really promising.In fact, it improves the overall capacity network and QoSthroughput requirements of both home and macro cellularusers. Multi-homing capabilities of macro cellular users havethe substantial advantage of supporting applications with highrequired data rate through aggregating the offered resourcesfrom different SCNs. The results corroborate our proposaland display the improvement reached compared to the open,closed, and shared time access policy based on single networkselection. It is noteworthy that future studies can combineour proposal with power control methods to achieve far betterresults.

VII. APPENDIX : PROOF OF THEOREM 1

We define a function g of i as :g(i) = iT

h,itotal + (1 i)T

min,itotal

maximize g(i) subject to

i.Th,itotal

ns,i ih i

ih.ns,i

Th,itotal(15)

and

(1 i).Tmin,itotalni

im i 1im.ni

Tmin,itotal(16)

Using (15) and (16)

ih.ns,i

Th,itotal i 1

im.ni

Tmin,itotal(17)

g(i) = i(Th,itotal T

min,itotal ) + T

min,itotal

dg(i)di

= Th,itotal Tmin,itotal

- if Th,itotal Tmin,itotal , i.e

dg(i)di

0 then g(i) monotonicallyincreases with increasing i, using (17) the solution i =1

im.ni

Tmin,i

total

this solution is feasible for i ih.ns,i

Th,itotal.

- if Th,itotal < Tmin,itotal , i.e

dg(i)di

< 0 then g(i) monotonicallydecreases with decreasing i, then using (17) the solution isi =

ih.ns,i

Th,itotal

REFERENCES[1] 3GPP TR 36.814 V9.0.0. Evolved Universal Terrestrial Radio Access

(E-UTRA); Further advancements for E-UTRA physical layer aspects(Release 9). Technical report, 3rd Generation Partnership Project, (2010).

[2] P. Mach, Z. Becvar, QoS-guaranteed power control mechanism based onthe frame utilization for femtocells. EURASIP J. Wirel. Commun. Netw(Article ID 802548), (2011)

[3] A. Mahmud and K. A. Hamdi, A unified framework for the analysis offractional frequency reuse techniques, IEEE Trans. Commun., vol. 62,no. 10, pp. 3692-3705, Oct. (2014).

[4] S. Al-Rubaye, A. Al-Dulaimi, J. Cosmas, Cognitive femtocell. IEEE Veh.Technol. Mag 6, 44 - 51 (2011)

[5] P. Xia, V. Chandrasekhar, J. Andrews, Open vs. closed access femtocellsin the uplink. IEEE Trans. Wirel. Commun 9(12), 3798 - 3809 (2010).

[6] HS. Jo, P. Xia, JG. Andrews, Open, closed, and shared access femtocellsin the downlink. EURASIP J. Wireless Comm. and Networking (2012).

[7] S. Mohanty and I. F. Akylidiz, A cross-layer (layer 2 + 3) handoffmanagement protocol for next generation wireless systems, IEEE Trans.mobile computing, vol. 5, no. 10, pp. 1347-1360, (2006).

[8] W. Shen and Q. Zeng, Resource management schemes for multiple trafficin integrated heterogeneous wireless and mobile networks, In Proceedingsof ICCCN, August , pp. 105-110 (2008).

[9] A. Walid, M. El Kamili, A. Kobbane, A. Mabrouk, S. Essaid, M. ElKoutbi. A Decentralized Network Selection Algorithm for Group VerticalHandover in Heterogeneous Networks. In proceeding of IEEE WCNCConference,Istanbul, pp. 2817-2821 (2014).

[10] M. Ismail and W. Zhuang, A distributed multi-service resource allocationalgorithm in heterogeneous wireless access medium, IEEE J. SelectedAreas Communications, vol. 30, no. 2, pp. 425-432, (2012).

[11] S. Shakkottai, E. Altman, A. Kumar, Multihoming of Users to AccessPoints in WLANs: A Population Game Perspective, IEEE Journal onSelected Areas in Communications 25(6): 1207-1215 (2007).

[12] L. Jongwook and B. Saewoong, On the MDP-Based cost cinimization forvideo-on-cemand services in a heterogeneous wireless Network with mul-tihomed terminals, IEEE Trans. Mob. Comput. 12(9):1737-1749 (2013).

[13] S. Nguyen, T. Trang Nguyen, G. Pujolle, S. Secci, Strategic evaluationof performance-cost trade-offs in a multipath TCP multihoming context,In Proceedings of IEEE ICC Canada, pp. 14431447 (2012).

exploiting multi-homing in hyper dense lte small-cells...

Documents