Lcal Mbility Achrig fr Seamless Hadver i Crdiated Small Cells Ravikumar Balakrisha ad Ia F Akyildiz Bradbad Wireless Netwrkig Labratry Schl f Electrical ad Cmputer Egieerig, Gergia Istitute f Techlgy, Atlata, GA 30332, USA Email {rbalakrisha6,ia}@ecegatechedu Abstract Small cells are redefiig the traditial cellular system ccepts Crdiati amg small cells ffers several beefits i terms f utilizig existig etwrk ifrastructure t supprt advaced iterferece maagemet, mbility maagemet as well as self-rgaizig (SON) fuctis I this wrk, the prblem f hadver maagemet is studied The crdiati betwee small cells is geerally t csidered i may existig hadver maagemet slutis resultig i icreased hadver cst, sessi iterrupti time ad cre etwrk verlad T this ed, a vel lcal achr-based architecture fr crdiated small cells is prpsed i this paper based which three hadver schemes are preseted A mathematical framewrk is develped t aalyze the perfrmace f the prpsed schemes ad Markv mdels are utilized t btai clsed-frm expressis fr the key hadver parameters icludig hadver cst ad sessi iterrupti time Numerical results idicate savigs f abut 60% i the sigalig cst, abut 50% i the data frwardig cst ad mre imprtatly abut 0% i the hadver iterrupti time cmpared t existig schemes based crdiated small cells I INTRODUCTION THE rapidly grwig eed f mbile wireless data ad services has resulted i tremedus advaces i mbile etwrks bth at the radi access ad the etwrk maagemet techlgies I this regard, cellular etwrks are udergig a majr trasfrmati with the existig macrcell cverage area uderlaid with a umber f lw-pwered small cell base statis t imprve the verall system capacity The emergece f this multi-layered apprach requires a revisi f may f the majr eablig techlgies icludig iterferece maagemet ad mbility maagemet as pited ut i the Rel-2 f 3GPP LTE-Advaced systems [] I particular, the presece f a umber f small cells uderlaid withi a macrcell layer causes frequet hadvers frm e cell t ather fr mbile users with a active sessi I rder t esure seamless cectivity fr the users, it is ecessary t efficietly hadle these frequet hadvers [2] Crdiati amg a set f small cells is ideal fr small cell deplymets i scearis such as airprts, malls, auditriums ad large ffice buildigs I fact, there are several stadardizati ad idustry effrts fr eterprise small cells targeted fr the abve scearis This crdiati eables achievig imprved iterferece maagemet, mbility maagemet as well as self-rgaizig (SON) fuctis by utilizig the uderlyig etwrk ifrastructure Specifically fr mbility maagemet, this crdiati ca als play a sigificat rle i the fllwig ways First, it ca eable scalable small cell deplymets by miimizig the lad the cre etwrk durig hadvers Furthermre, sice the small cells icrprate differet backhaul techlgies icludig iteret, micrwave LOS, etc, crdiati amg small cells ca help vercme ptetial backhaul issues f lg latecy ad peratig csts ivlved fr sigalig ad data iteracti durig hadvers Hwever, it must be metied that crdiated small cells als place sme cstraits such as requirig a etwrk ifrastructure with high-speed liks Several existig wrk hadver maagemet aim t miimize the key parameters f hadver csts, iterrupti time Lcal achr-based mbility maagemet schemes were studied i [3] fr ptimizig pagig ad registrati updates I [4], ew architectures have bee prpsed t mve the mbility achr pit clser t the base statis Hwever, icrpratig these ew architectures require redefiig the security key mechaisms ad sigalig flw amg ther mdificatis I [5], the authrs prpse a fast hadver scheme Nevertheless, this scheme des t prvide sigificat savigs hadver csts r the cre etwrk lad I [6], a 2-based data frwardig scheme aalgus t the piter frwardig techique is prpsed I this wrk, we csider the case f crdiated small cells ad prpse a vel lcal achr-based architecture fr prvidig ehaced supprt fr hadver maagemet The ctributis f the paper are summarized as fllws We prpse vel hadver schemes based the lcal achr-based architecture fr crdiated small cells We aalytically mdel a cluster f small cells t study the mbility behaviur f users We prvide clsed-frm expressis fr hadver cst, hadver iterrupti time uder the prpsed hadver schemes We preset the umerical results highlightig the perfrmace gais f ver 0% reducti i the sessi iterrupti time ad mre tha 50% reducti i the hadver csts The rest f this paper is rgaized as fllws The lcal achr-based architecture ad the prpsed hadver mechaisms are preseted i Sec II I Sec III, the aalytical mdel develped t study the mbility behavir f users is described I Sec IV, the clsed-frm expressis fr several hadver perfrmace metrics are derived Numerical results are preseted i Sec V Fially, the mai cclusis are summarized i Sec VI
Small Cell Access Pit (SC) Lcal Achr Nde (LA) Mbile Users (MU) Crdiated small cell evirmet Security GW MME MBS S-GW Cre Netwrk INTERNET Firewall Cluster Cluster 2 Fig Lcal achr based hadver architecture II LOCAL ANCHOR-BASED HANDOVER MANAGEMENT I this secti, we first mtivate the eed fr a ew hadver architecture fr crdiated small cells ad the describe ur prpsed lcal achr-based hadver architecture A Mtivati The hadver prcedure fr 3GPP LTE-A systems utilizig the direct iterface (2 iterface) betwee small cells is divided it three phases (i) hadver preparati, (ii) hadver executi, ad (iii) hadver cmpleti We refer the readers t [7] fr a detailed explaati f the hadver prcedure The key bservati frm the abve 2-based hadver prcedure is that the mbility achr fr the hadver is the cre etwrk (particularly the MME) This gives the ituiti that the backhaul fr the small cells eeds t have the key characteristics f lw-latecy ad high reliability The ttal dwlik iterrupti time fr a rmal hadver prcedure is cmputed as max(5ms, T p ) i [] where T p is the path switchig delay This cfirms that the backhaul latecy (t perfrm path switchig) shuld t exceed few tes f millisecds Hwever, may f the small cell techlgies cat meet this striget backhaul latecy requiremets Therefre, we advcate the use f lcal mbility achrig that ca satisfy the bjectives f miimizig ttal hadver csts, hadver iterrupti time ad cre etwrk lad B Lcal Achr-based Architecture With the abve cre bjectives i mid, we prpse a lcal achr-based (LA-based) architecture fr a set f crdiated small cells We divide a large array f small cells it several clusters where each cluster ctais a subset f small cells The lcal achr-based architecture is shw i Fig Oe f the small cells i a cluster is chse as the lcal achr (LA) The small cells iside a cluster are assumed t cect t a lcal etwrk The lcal achrs maitai cecti with the IP gateway f the lcal etwrk which is cected t the Mbile Cre Netwrk (CN) thrugh a firewall ad public iteret The mai advatage f the LA-based architecture is that hadver mechaisms ca be prpsed that utilize the LA as the mbility achr therefre miimizig the hadver iterrupti as well as the assciated csts E Bearer ID SC If SC Trasprt SAP ID Layer Address GTP TEID f SC CN SAP ID CN If Trasprt Layer Address GTP TEID f CN Fig 2 Lcal achr registrati table The fuctis f the LA are summarized as fllws Ccetratr f traffic betwee SCs ad CN I this sese, the LA perfrms prxy tuellig fucti fr the iterface betwee SCs i its cluster ad the CN (S iterface, i case f LTE-A) The LA is als capable f prxy 2 fuctis t eable hadver t SCs belgig t ther clusters r t macrcell base stati (MBS) I rder t achieve the prxy tuellig betwee SCs ad CN, the LA maitais a lcal achr registrati table (LART) This is shw i Fig 2 The LART ctais the data ad sigalig plae ed pit addresses This is als eabled by maitaiig S security ver tw hps betwee fr CN, LA ad LA, SC liks Lcal Mbility Achr fr hadver betwee ther SCs The LA acts as a lcal mbility achr fr the users perfrmig hadver betwee SCs i its cluster The LA is able t perfrm lcal path switchig withut affectig the established prcedures ad ly frwards the path switch request t the CN ce certai cditis are met Usig the lcal achr based-architecture, the hadver mechaism are utlied i the ext subsectis C Lcal path switchig-based hadver Origially, the target SC f a MS udergig hadver seds a path switch request message t the cre etwrk t idicate the cmpleti f hadver ad t eable the CN t switch the dwlik path fr the MS twards the target SC Mre imprtatly, the path switchig als eables the CN t geerate ew keys fr securig the SC () MS iterface A detailed explaati f the key maagemet is available i [9] This ew key als helps vercme issues ccered with frward key separati, ie, ce the target SC receives the ew key (thrugh vertical key derivati) i path switch request ack frm the MME, the previus SC cat decde the keys f the target SC If the path switchig is t perfrmed, the the target SC will ctiue t use the key derived by the previus SC (hriztal key derivati) frm which the MS haded ver frm Hwever, due t lger backhaul latecy ad frequet hadver i small cells, the path switch request ack frm the MME is t always received i time t achieve the best case f 2-hp backward key separati At the same time, perfrmig the path switch with the cre etwrk seriusly degrades the hadver perfrmace I this scheme, we advcate the use f lcal path switchig fr SCs belgig t the same cluster fr upt umber f hadvers fr a give MS T achieve this, the LA maitais a cuter value fr each user util its sessi eds ad perfrms path switchig with the CN, ly whe the cuter value is equal t The the LA resets the cuter ad ctiues icremetig util the sessi fr the MS eds Therefre,
HO Preparati HO Executi HO Cmpleti MS Servig SC Target SC Lcal Achr MME S-GW Attach t Target SC Measuremet Reprt HO Cmmad Hadver Decisi Hadver Cfirm HO HO ACK SN Status Trasfer UE C Release Admissi Ctrl Data Frwardig Ed Marker Ed Marker ACK Cditi Check ACK If Cuter = aν Mdify Bearer Mdify Bearer Respse Create DL path HO Preparati HO Executi HO Cmpleti MS Servig SC Target SC Attach t Target SC Measuremet Reprt HO Cmmad Hadver Decisi Hadver Cfirm HO HO ACK SN Status Trasfer Data Frwardig UE C Release Admissi Ctrl Fig 4 DF-ehaced hadver scheme Fig 3 LP-based hadver scheme fr crdiated small cells at the ed f every hadver, the path switch message frm the target SC is t frwarded by the LA t the CN Istead, the LA creates a ew S path with the target SC ad updates its LART with the ew small cell edpit address fr the sessi As a result, the LA frwards all future dwlik data ad sigalig t the target SC Fllwig this, the LA seds a ed marker t the previus SC t idicate the switchig f the lcal path This is ctiued fr all further hadvers fr the user util the cuter value equals, beyd which the path switch request frm the latest target SC is frwarded t the CN t restre the backward key separati The hadver prcedure is illustrated i Fig 3 D Rute Optimizati-ehaced hadver The lcal path switchig-based hadver mechaism prmises sigificat gais i terms f miimizig sigalig cst, hadver iterrupti time as well as the cre etwrk verladig Hwever, the data frwardig cst ca be further ptimized Durig hadver frm e SC t ather, the i trasit dwlik data frwarded t the target SC fllws the path LA ) Servig SC ) Target SC This, althugh ly fr the hadver durati, is aalgus t the triagular rutig that takes place i mbile IP etwrks ad results i icreased data frwardig csts Hwever, this is vercme by sedig the path switch request message frm the target SC befre it has established a radi lik with the MS By dig s, the LA ca already perfrm lcal path switchig ad establish a ew S path with the target SC I rder t avid ay lss f i-trasit packets, the target SC ca idicate, i the path switch request message, the sequece umber f the data (SN) r ther higher layer ifrmati (delivered ver the SN status trasfer message) f the DL packets Nw, the dwlik i-trasit data is directly set frm the LA t the target SC Hwever, the target SC ctiues receivig uplik data frm the servig SC s buffer We term this mechaism as rute ptimizatiehaced (RO-ehaced) hadver scheme Similar t the case f LP-based hadver, the RO-ehaced scheme ivlves sedig the path switch request message t the cre etwrk every time the MS cmpletes a hadver fr a give sessi E Data Frwardig-ehaced hadver While the LA achrs mbility fr ther SCs, it is als pssible that the MS is attached t the LA ad eeds a hadver t a eighburig SC I this special case, it is pssible t etirely elimiate the path switchig perati by the target SC Istead, the servig SC, which is the LA, perfrms data frwardig t the target SC usig the 2 lik created betwee LA ad the target SC Whe the user mves ut f the target SC t ather SC, the RO-ehaced hadver prcedure ca be applied t btai maximum savigs i terms f sigalig cst The DF-ehaced hadver scheme is illustrated i Fig 4 III ANALYTICAL MODEL I rder t evaluate the perfrmace f the hadver schemes, we eed t study the evluti f the user s behavir i the crdiated small cell etwrk T this ed, we utilize discrete-time Markv mdel t determie the statiary prbabilities f a user preset i each f the small cells I this wrk, we mdel a sigle cluster f small cells that als icludes a lcal achr small cell Based the btaied results, we prvide clsed-frm expressis fr differet hadver perfrmace parameters A Mdel Descripti We utilize a grid tplgy t mdel a cluster f small cells as prpsed i [0], [] The tw-dimesial grid mdel adpted i this paper is shw i Fig 5 Each blck f the grid is a SC ad is represeted usig the state variable S j i f the discrete-time Markv mdel The LA (represeted by S 0 ) is lcated cetrally i the grid ad the ther small cells are deplyed surrudig the lcal achr i tiers The tplgy csists f up t K tiers It is cstructed such that each SC has fur eighbrig SCs except fr the SCs i the Kth tier The umber f SCs i tier i equals 4i I the Markv mdel, we have a additial state S idle idicatig that there is active sessi fr the MS idepedet f which cell the user is i The MS chages state ly at the ed f a discrete-time slt t If the user has a active sessi, it ca be i ay f the states S j i where i represets the tier ad j represets the idex f the cell i its tier The sessi arrival parameter, sessi durati parameter, ad cell residece parameter are give by, µ ad r The
upt K tiers Fig 5 Tw-dimesial grid deplymet mdel upt K tiers Fig 6 Grid mdel after state aggregati crrespdig prbabilities are P, P µ, ad P r I this wrk, we csider radm-walk mbility where users ca mve frm a SC t ay f its eighbrs with equal prbability I the abve mdel, each cell is represeted by a state variable Hwever, this may result i state space explsi, ad therefre, we apply state aggregati fr the Markv mdel makig use f lcati symmetry ad the adpted mbility mdel By perfrmig state aggregati, we have K tiers ad M states i each tier where M = d(k + )/2e The tplgy after perfrmig state aggregati is shw i Fig 6 The discrete-time Markv mdel fr the aggregated states is represeted i Fig 7 Here, N represets the ttal umber f small cells i the mdel, P sijs ī j represets the prbability f the user mvig frm cell S j i t cell which is btaied S jī based the user mbility characteristics I this wrk, we csider that the sessi arrivals fllw Piss distributi, while the cell residece time ad sessi durati fllw expetial distributi As a result we have P = t, P r = r t, ad P µ = µ t Based the abve discrete-time Markv mdel, we btai S 0 P λ/n -P λ S idle 4P λ/n P r(-p μ)p ss0 P r(-p μ)p s0s S P λ/n 4KP λ/n P r(-p μ)p s2s P r(-p μ)p ss2 S 2 P μ (-P μ)(-p r) S K M Fig 7 Discrete-time Markv mdel the statiary prbability distributi f a MS i state S j i Befre prvidig the balace ad rmalizati equatis, we defie the fllwig parameters Let a = P /N, b =( P µ )( P r ), ad d = P r ( P µ ) I additi, we defie the fllwig parameters < { i, i} = { 2, 4 } if i<k, {, } if i = K, {0, 0} if i = K The balace equatis are give as fllws idle =( Na) idle + P µ K i=0 d i+ j= () j i, (2) 0 = a idle + b 0 + d 0, (3) =4a idle + b + d 0 + 2 + 2 2, (4) i =4a idle + b i + d 4 i + i i+ + 2 i i+ 2 ; i >, (5) 2 2 =4a idle + b 2 2 + d 2 + 2 2 3 2, (6) 3 2 =a idle + b 3 2 + d 2 2 + 2 2 2 + 2 3 4 2 + 3 4 3, (7) i 2 =a idle + b i 2 + d 2 i + 4 2 i + 2 i i+ 2 + 2 i i+ 3 ; i >3, () j 2j 2 =4a idle + b j 2j 2 + d 4 j 2j 3 + 2 2j 2 j 2j ; j >2, (9) j 2j =a idle + b j 2j + d 4 j 2j 2 + 2 j 2j 2 + 2 2j j 2j + 2j j+ 2j ; j >2, (0) j i =a idle + b j i + d 4 j i + 4 j i + 2 i j i+ + 2 i j+ i+ ; j >2,i>2j, () where j i is f the frm j i = xj i idle + y j i j i + zj i d ; i, j Fr simplicity, i the future sectis, we use the tati j j i = i + j i i 6=,j 6=, j i + j i + (2) j i i =,j =, where j i = xj i idle + y j i j i, = d 0, j i = zj i d ad = z j i d d 0 The rmalizati equati is give by idle + i=0 d i+ j i j= = (3) Usig equatis (2) - () ad (3), the statiary prbability distributi f the Markv mdel ca be derived We utilize the statiary prbability distributi j i ad idle t derive the perfrmace metrics
IV HANDOVER PERFORMANCE METRICS The clsed-frm expressis fr differet perfrmace characteristics are derived i this secti A Hadver Cst The hadver cst based the abve Markv mdel is btaied as where C HO = j kċ0 0 + Ċ + i= j= + j k C0 0 + C + i= d K+ j= C ij j i d K+ Ċ ij j i, (4) Maximum umber f hadvers per MS per sessi Number f hadvers befre full path switchig Ċ ij, C ij Hadver csts frm SC t SC i state S j i with full path switchig ad lcal path switchig respectively Equati (4) is used t cmpute bth the sigalig cst Cij s ad the data frwardig cst Cij D icurred durig a hadver B Hadver Iterrupti Time The hadver iterrupti time is ather key perfrmace metric fr hadver schemes The clsed-frm expressi fr hadver iterrupti time is btaied as give as where HO = j k 0 0 + + + + i= d K+ j= i= d K+ j= j k 0 0 + ij j i ij j i, (5) ij, ij Iterrupti times fr hadver frm SC (r LA) t SC (r LA) i state S j i with full path switchig ad lcal path switchig respectively C Cst Cmputati The cst fuctis fr the prpsed hadver schemes are prvided here I ur tplgy, the SC i a cluster ca be cected t the LA thrugh multi-hp usig the lcal etwrk Hwever, sice the itermediary SCs ly act as IP ruters fr the SC, LA cmmuicati, the prcessig cst at the itermediate SCs are ly accuted frm the ruter prcessig cst This cst is egligible Similarly, the margi f lik cst betwee sigle hp ad multi-hp is egligible sice the cluster spas ly a few hudred metres Hece, i the fllwig cst fuctis, we csider the cst icurred by tw small cells lcated at differet distaces frm the LA icur the same cst as lg as they are belg t the same cluster TABLE I System Parameters Parameter Value Parameter Value 000/s C sc 5ms µ 00/s C la 0ms r 0/s C 2 5ms K, 2, 3, 4 C s 5ms N 5, 3, 25, 4 C s 50ms t 00s C scgw 0ms ) LP-based scheme The sigalig cst is give as < 4C sc +2C la +4C 2 + C s ; if i =0,j =, C ij s = 5C sc +2C la +4C 2 +2C s ; if i =,j =, 5C sc +2C la +4C 2 +3C s ; therwise (6) The data frwardig cst is give as < C sc + C la + C 2 + C s ; if i =0,j =, C ij D = C sc + C 2 ; if i =,j =, (7) C sc + C la + C 2 + C s ; therwise 2) RO-ehaced scheme The sigalig cst is the same as i equati (6) The data frward cst is give as < q(c sc + C 2 )+C la ; if i =0,j =, C ij D = C la + C s ; if i =,j =, q(c sc + C 2 )+( q)(c la + C s ); therwise, () where q ad q are fractis f uplik ad dwlik data respectively 3) DF-ehaced scheme The sigalig cst ad data frwardig cst uder DF-ehaced hadver scheme differs frm the RO-ehaced scheme ly fr the case whe a MS hads ver frm LA t SC These are give as C s = 4C sc +4C 2 ad C D = C sc + C 2 Whe path switchig is applied, the data frwardig cst will still be Ċs ij = C ij s Hwever, the sigalig cst is the same fr all the schemes ad is give by Ċ s ij =5C sc +4C 2 +3C s +2C scgw + C ps (9) As we see abve, this is als the cst icurred i the absece f ur hadver schemes Hece, we cmpare ur sluti fr the case whe lcal path switchig is utilized I this case, we apply C ij s = Ċs ij t determie the sigalig cst Similarly the data frwardig cst is give by ĊD ij = C ij D = C sc + C 2 + C la + C s V NUMERICAL RESULTS I this secti, we umerically evaluate the prpsed schemes based the perfrmace metrics i Secti IV The system parameter values used i this secti are prvided i Table I as recmmeded i [] We prvide results as a rati with respect t the curret best scheme defied i the previus secti This meas that a smaller rati crrespds t imprved hadver perfrmace I Fig (a), the sigalig cst rati uder the prpsed hadver schemes is pltted as a fucti f the umber f tiers (K) ad the umber f hadvers util path switchig ( ) We bserve that all three prpsed schemes shw similarity i their behavir i varyig with K ad Fr small values f, the sigalig cst rati is large This is because f the frequet path switchig perfrmed fr each hadver Hwever, as
(a) (b) (c) Fig Hadver perfrmace vs umber f tiers (K) ad umber f hadvers util path switchig ( ) icreases, we bserve that the sigalig cst rati decreases expetially ad reaches t arud 40% f the maximum rati This validates ur claims that miimizig the umber f full path switchig with the etwrk results i sigalig cst savigs The DF-ehaced scheme ffers the best perfrmace f achievig abut 70% reducti i the sigalig cst fr K apple 2 ad I geeral, all the three schemes are able t ffer almst 60% sigalig cst savigs fr =6 It is als iterestig t te that icrease i tier size (K) des t sigificatly affect the sigalig cst perfrmace This allws fr scalable cluster sizes as lg as the etwrk ifrastructure is able t supprt crdiati amg the member SCs I Fig (b), the ttal data frwardig cst rati as a fucti f ad K is pltted It is see that the LP-based scheme des t ffer sigificat perfrmace gai whe K 3 This is expected as there is a triagular rutig f the i-trasit data frm the servig SC t the target SC durig a hadver Hwever, bth the RO-ehaced ad DF-ehaced schemes are able t achieve abut 50% gai i the data frwardig cst sice bth the schemes eable rute-ptimizati fr the i-trasit data packets durig hadver It is iterestig t te that bth these schemes d t vary sigificatly with as the rute-ptimizati des t deped hw frequetly the full path switchig is perfrmed frm the target SC It is wrthy f bservig that the RO-ehaced scheme has a particular behavir differet frm the ther schemes as the tier size icreases This is due t the larger csts ivlved whe a MS hads ver frm LA t ather SC, ad with a large K, the prbability f such a hadver becmes very lw I Fig (c), the hadver iterrupti time rati is pltted agaist ad K The parameter affectig the hadver iterrupti time is maily the frequecy f path switchig As icreases, the path switchig takes place less frequetly ad hece we ca bserve that the hadver iterrupti time decreases ad reaches a miimum f up t 20% f the maximum rati This crrespds t abut 0% reducti i the iterrupti time This is e f the key perfrmace gais f ur prpsed schemes as lwer iterrupti time reduces hadver failures ad eable seamless mbility fr users VI CONCLUSION Crdiati amg smalls cells helps realize imprved mbility maagemet ad iterferece maagemet fuctis amg thers I this wrk, we utilized this crdiati betwee small cells t prpse a lcal achr-based hadver architecture ad usig this, prpsed vel hadver schemes Fr this, we develped a mathematical framewrk t aalyze the hadver schemes usig Markv mdels Based this mathematical framewrk, we derived clsed-frm expressis fr the key hadver perfrmace parameters The umerical results idicate savigs f abut 60% i the sigalig cst, abut 50% i the data frwardig cst ad up t 0% reducti i the hadver iterrupti ver existig schemes The prpsed framewrk ca be utilized t aalyze the perfrmace f ew hadver schemes Fr future wrk, we ited t study ad prpse ew hadver mechaisms supprtig hadvers amg differet clusters, ad t develp aalytical mdels fr such a tplgy We als ited t use differet prbability distributis f cell residece time, sessi durati ad sessi arrival time it the develped aalytical mdel t prvide a mre thrugh framewrk fr studyig hadver mechaisms REFERENCES [] 3GPP, Evlved Uiversal Terrestrial Radi Access (E-UTRA); Further advacemets fr E-UTRA physical layer aspects, TR 364, Mar 200 [2], Evlved Uiversal Terrestrial Radi Access (E-UTRA); Mbility ehacemets i hetergeeus etwrks, TR 3639, Dec 202 [3] J H ad I Akyildiz, Lcal achr scheme fr reducig sigalig csts i persal cmmuicatis etwrks, Netwrkig, IEEE/ACM Trasactis, vl 4, 5, pp 709 725, ct 996 [4] F Zdarsky, A Maeder, S Al-Sabea, ad S Schmid, Lcalizati f Data ad Ctrl Plae Traffic i Eterprise Femtcell Netwrks, i Prc IEEE 73rd Vehicular Techlgy Cferece (VTC Sprig), May 20, pp 5 [5] A Rath ad S Pawar, Fast Hadver i Cellular Netwrks with Femtcells, i Prc IEEE Iteratial Cferece Cmmuicatis (ICC), Ju 202, pp 2752 2757 [6] T Gu, A ul Quddus, N Wag, ad R Tafazlli, Lcal Mbility Maagemet fr Netwrked Femtcells Based 2 Traffic Frwardig, IEEE Trasactis Vehicular Techlgy, vl pp, 99, p, 202 [7] 3GPP, Evlved Uiversal Terrestrial Radi Access (E-UTRA) ad Evlved Uiversal Terrestrial Radi Access Netwrk (E-UTRAN); Overall descripti; Stage 2, 36300, Mar 203 [], Feasibility study fr evlved Uiversal Terrestrial Radi Access (UTRA) ad Uiversal Terrestrial Radi Access Netwrk (UTRAN), TR 2592, Sep 202 [9], 3GPP System Architecture Evluti (SAE); Security architecture, TR 3340, Mar 203 [0], Evlved Uiversal Terrestrial Radi Access (E-UTRA); TDD Hme ende B (HeNB) Radi Frequecy (RF) requiremets aalysis, TR 36922, Sep 202 [] S C Frum, A Small Cell Frum Whitepaper, Eterprise Femtcell Deplymet Guidelies, Tech Rep, Feb 202