centralized mobility load balancing scheme in lte systems.pdf

6
Centralized Mobility Load Balancing Scheme in LTE Systems Junichi Suga, Yuji Kojima, Masato Okuda Fujitsu Laboratories Ltd. 1-1, Kamikodanaka 4, Nakahara-ku, Kawasaki, Japan  Abstract  —MLB (Mobility Load Balancing) is one of the use cases of SON (Self-Organizing Network) discussed in NGMN and 3GPP. The objective of MLB is to reduce the number of low throughput users and the call blocking rate, when cell loads such as radio resources are unbalanced among cells, by adjusting handover parameters. In this paper, we focus on adjusting the HO margin which is one of the handover parameters and propose a centralized solution for MLB where a central server decides optimal HO margins considering load distribution among cells. In addition, we also present simulations results to demonstrate that our proposed solution reduces the number of low throughput users compared with a distributed solution.  Keywords-component; mobility load balancing, handover, self- organizing network I. I  NTRODUCTION With the rapid growth of mobile data communications,  planning, optimization and maintenanc e of the cellular networks are getting more complex, higher cost and time consuming for mobile network operators. In order to reduce these tasks of mobile network operators, SON (Self-Organizin g  Network) has been discussed in NGMN (the Next Generation Management Network)[1] and 3GPP (the 3 rd  Generation Partnership Project)[2]. The main goal of SON is to mitigate operators’ network operation and management tasks by automatically adjusting parameters related to these tasks (i.e.  planning, optimization and maintenan ce) based on measurement data of radio signal quality, performance information and failure information. In cellular networks, since mobile terminals (User Equipment or UE in LTE (Long Term Evolution)) move around and start or terminate their communications randomly, cell loads such as the number of connected UEs or radio resource usage among cells can be unbalanced. While some cells might be overloaded, i.e., having many connected UEs, the other cells might be less loaded. In the overloaded cells, this situation results in some UEs not obtaining enough user throughput and also results in increased call blocking rate. MLB (Mobility Load Balancing), which is one of the SON use cases, can solve these problems. As a result of MLB, it is expected that mobile network operators can improve the user satisfaction and their revenue. The objective of MLB is to improve the system throughput, to reduce unsatisfied UEs, which receive lower user throughput than expected and to reduce the call blocking rate. This is achieved by adjusting HO (Handover) parameters between the overloaded cell and its neighboring cells. The HO parameters are related to the HO decision for the connected UEs. With these adjustments, some UEs in the overloaded cell are expected to be handed off to its neighboring cells. In this paper, we focus on PRB (Physical Resource Block) utilization as a cell load which indicates the radio resource usage in LTE and adjusting the HO margin which is one of the HO parameters. For an MLB solution, we propose a centralized solution where a central server in the cellular network determines all HO margins among cells. In addition, we also  propose an algorithm executed in the central server which decides the optimal HO margins on the basis of the load distribution predicted from the measurement reports from UEs in LTE. Furthermore we evaluate our MLB solution with our simulator in terms of the number of unsatisfied UEs with low user throughput. The rest of this paper is organized as follows: we explain the overview of MLB in section II, and provide the usage of measurement reports for our MLB solution in section III. Section IV provides the centralized solution and a HO margin adjustment algorithm executed in the central server. Section V explains simulations assumptions and scenarios for evaluation. And we show simulations results and evaluate our MLB solution in Section VI. Concluding remarks are discussed in Section VII. II. MOBILITY LOAD BALANCING Basically, each base station (enhanced Node B or eNB in LTE) measures its serving cell load such as the number of connected UEs or the PRB utilization. In MLB with a distributed solution discussed in 3GPP, an eNB cooperates with its neighboring eNBs via X2. When an eNB detects that its serving cell load is overloaded, the eNB obtain neighboring cell loads from its neighboring eNBs and adjusts the HO  parameters via X 2 to force some UEs to be handed off f rom the overloaded serving cell to its neighboring cells. In this paper, we focus on adjusting the HO margin. The followings are the explanations of HO decision process and the effect of adjusting the HO margin.  A.  Handover prameters In LTE, a handover decision can be triggered with a number of events[3]. Generally, the event A3 is used for the trigger of handover decision. The following formula is the simplified formula of event A3. 2011 8th International Symposium on Wireless Communication Systems, Aachen 978-1-61 28 4- 402-2/11 /$ 26 .00 ©2011 IEEE 306

Upload: urfriendlyjoe

Post on 02-Mar-2016

21 views

Category:

Documents


0 download

DESCRIPTION

Centralized Mobility Load Balancing Scheme in LTE Systems

TRANSCRIPT

Page 1: Centralized Mobility Load Balancing Scheme in LTE Systems.pdf

7/18/2019 Centralized Mobility Load Balancing Scheme in LTE Systems.pdf

http://slidepdf.com/reader/full/centralized-mobility-load-balancing-scheme-in-lte-systemspdf 1/5

Centralized Mobility Load Balancing Scheme

in LTE Systems

Junichi Suga, Yuji Kojima, Masato OkudaFujitsu Laboratories Ltd.

1-1, Kamikodanaka 4, Nakahara-ku, Kawasaki, Japan

 Abstract  —MLB (Mobility Load Balancing) is one of the use cases

of SON (Self-Organizing Network) discussed in NGMN and

3GPP. The objective of MLB is to reduce the number of low

throughput users and the call blocking rate, when cell loads such

as radio resources are unbalanced among cells, by adjusting

handover parameters. In this paper, we focus on adjusting the

HO margin which is one of the handover parameters and propose

a centralized solution for MLB where a central server decides

optimal HO margins considering load distribution among cells.

In addition, we also present simulations results to demonstratethat our proposed solution reduces the number of low

throughput users compared with a distributed solution.

 Keywords-component; mobility load balancing, handover, self-

organizing network

I.  I NTRODUCTION

With the rapid growth of mobile data communications, planning, optimization and maintenance of the cellularnetworks are getting more complex, higher cost and timeconsuming for mobile network operators. In order to reducethese tasks of mobile network operators, SON (Self-Organizing

 Network) has been discussed in NGMN (the Next Generation

Management Network)[1] and 3GPP (the 3rd  GenerationPartnership Project)[2]. The main goal of SON is to mitigateoperators’ network operation and management tasks byautomatically adjusting parameters related to these tasks (i.e.

 planning, optimization and maintenance) based onmeasurement data of radio signal quality, performanceinformation and failure information.

In cellular networks, since mobile terminals (UserEquipment or UE in LTE (Long Term Evolution)) movearound and start or terminate their communications randomly,cell loads such as the number of connected UEs or radioresource usage among cells can be unbalanced. While somecells might be overloaded, i.e., having many connected UEs,

the other cells might be less loaded. In the overloaded cells,this situation results in some UEs not obtaining enough userthroughput and also results in increased call blocking rate.MLB (Mobility Load Balancing), which is one of the SON usecases, can solve these problems. As a result of MLB, it isexpected that mobile network operators can improve the usersatisfaction and their revenue.

The objective of MLB is to improve the system throughput,to reduce unsatisfied UEs, which receive lower user throughputthan expected and to reduce the call blocking rate. This isachieved by adjusting HO (Handover) parameters between the

overloaded cell and its neighboring cells. The HO parametersare related to the HO decision for the connected UEs. Withthese adjustments, some UEs in the overloaded cell areexpected to be handed off to its neighboring cells.

In this paper, we focus on PRB (Physical Resource Block)utilization as a cell load which indicates the radio resourceusage in LTE and adjusting the HO margin which is one of theHO parameters. For an MLB solution, we propose a centralized

solution where a central server in the cellular networkdetermines all HO margins among cells. In addition, we also

 propose an algorithm executed in the central server whichdecides the optimal HO margins on the basis of the loaddistribution predicted from the measurement reports from UEsin LTE. Furthermore we evaluate our MLB solution with oursimulator in terms of the number of unsatisfied UEs with lowuser throughput.

The rest of this paper is organized as follows: we explainthe overview of MLB in section II, and provide the usage ofmeasurement reports for our MLB solution in section III.Section IV provides the centralized solution and a HO marginadjustment algorithm executed in the central server. Section V

explains simulations assumptions and scenarios for evaluation.And we show simulations results and evaluate our MLBsolution in Section VI. Concluding remarks are discussed inSection VII.

II.  MOBILITY LOAD BALANCING 

Basically, each base station (enhanced Node B or eNB inLTE) measures its serving cell load such as the number ofconnected UEs or the PRB utilization. In MLB with adistributed solution discussed in 3GPP, an eNB cooperateswith its neighboring eNBs via X2. When an eNB detects thatits serving cell load is overloaded, the eNB obtain neighboringcell loads from its neighboring eNBs and adjusts the HO

 parameters via X2 to force some UEs to be handed off from theoverloaded serving cell to its neighboring cells.

In this paper, we focus on adjusting the HO margin. Thefollowings are the explanations of HO decision process and theeffect of adjusting the HO margin.

 A.  Handover prameters

In LTE, a handover decision can be triggered with anumber of events[3]. Generally, the event A3 is used for thetrigger of handover decision. The following formula is thesimplified formula of event A3.

2011 8th International Symposium on Wireless Communication Systems, Aachen

978-1-61284-402-2/11/$26.00 ©2011 IEEE 306

Page 2: Centralized Mobility Load Balancing Scheme in LTE Systems.pdf

7/18/2019 Centralized Mobility Load Balancing Scheme in LTE Systems.pdf

http://slidepdf.com/reader/full/centralized-mobility-load-balancing-scheme-in-lte-systemspdf 2/5

     (1) 

where M n is RSRP(Received Signal Reference Power) in dBmor RSRQ(Received Signal Reference Quality) in dB for aneighboring cell, M  s is RSRP or RSRQ for the serving cell, and HOmargin is a margin between M n  and M  s in dB. A differentvalue for  HOmargin  can be set for each neighboring cell.When the result of measurement on a UE for the serving celland its neighboring cell matches the formula (1), a handoverdecision for the UE from the serving cell to the neighboringcell as the target cell is triggered at the eNB.

 B.  Handover margin adjustment

When an eNB detects its serving cell to be overloaded, theHO margin to its neighboring cells will be adjusted to force theUEs in the serving cell to be handed off to its neighboring cells.

Figure 1 shows the relation between the HO marginadjustment and the timing of the HO for UEs. When the HOmargin is decreased, the UEs located at the cell edge tend tomeet the formula (1) and are expected to be handed off fromthe serving cell to the neighboring cell. On the other hand,when the HO margin is increased, the UEs in the serving cell

are prevented to be handed off to the neighboring cell. That is,adjusting the HO margin affects the coverage size of theserving cell and its neighboring cell virtually without changingradio environment.

III.  MEASUREMENT R EPORT UTILIZATION 

To adjust the HO margin in MLB more effectively and precisely, it is proposed that the measurement reports from UEs be used for predicting the cell loads after HO marginadjustment[4]. Since the measurement reports can contain Ms and Mn in the formula (1), the eNB collects the information onMs and Mn of each UE located at cell edge between the servingcell and its neighboring cells.

 A.  Measurement reports configuration

The trigger of measurement reports for the load predictionmight be configured as similar to the trigger of handover asfollows,

     (2)

 HOmargin  in the formula (1) of the handover trigger isreplaced to MRmargin. When a UE measures RSRP or RSRQof the serving cell and its neighboring cell and matches theresult of the measurement with the formula (2), the UE starts tosend measurement reports with M  s  and M n  to its serving eNB

 periodically. The relation between HOmargin and MRmargin isas follows,

  (3)

That is, before the handover for the UE is executed, the eNBobtains measurement reports from the UE.

 B.  Handover prediction with measurement reports

Since the eNB obtains the measurement report from theUEs located at the edge of cell, the eNB will be able to predictthe handover for some UEs located at the edge of the servingcell to its neighboring cells following the update of the HOmargin in the handover trigger. For instance, we assume that

the current HO margin is set to 2.0 dB and, M n  and M  s  in themeasurement reports from a UE satisfy the relationM n=M  s+1.5dB. If the eNB updates the HO margin as HOmargin=1.0dB, the UE is predicted to be handed off to thecorresponding neighboring cell, because M n  and M  s  wouldmatch the handover trigger with the updated HO margin.

The eNB also can measure the UE’s PRB utilization at the

serving cell and estimate the PRB utilization at the neighboringcell in accordance with its user throughput and an MCS(Modulation and Coding Scheme) on the neighboring cell

 predicted by the RSRQ of the neighboring cells in themeasurement reports.

IV.  PROPOSED SOLUTION 

In this paper, we propose a centralized solution where acentral server in the cellular network collects informationrelated to MLB and executes an MLB algorithm for all cells. Inaddition, we also propose the HO margin adjustment algorithmexecuted in the central server as an MLB algorithm using theload prediction that is based on the measurement reports from

the UEs described in Section III. A.  Centralized solution

Compared with the distributed solution proposed in[4][5][6] where an eNB itself adjusts the HO margin of theserving cell cooperating with its neighboring eNBs, thecentralized solution considering all cells is expected todistribute the unbalanced cell load, especially in the case whenseveral cells close to each other are overloaded.

Figure 2 shows the sequential flow of the proposedcentralized solution. The central server in the figure collects theinformation on the PRB utilization of all cells and the results ofthe measurement reports via S1 from eNBs. With thisinformation, the central server executes an MLB algorithm andadjusts all HO margins between two neighboring cells.

 B.  Overall procedure in the central sever

At the beginning of the procedure in the central server, itdetects the overloaded cells in the cellular network inaccordance with comparing their PRB utilization with a

 predetermined threshold. The central server applies the HOmargin adjustment algorithm described later to eachoverloaded cell in the descending order of the PRB utilizationand adjusts the HO margin between the overloaded cell and itsneighboring cells. With the HO margin adjustment in the

Figure 1. HO margin adjustment

 

307

Page 3: Centralized Mobility Load Balancing Scheme in LTE Systems.pdf

7/18/2019 Centralized Mobility Load Balancing Scheme in LTE Systems.pdf

http://slidepdf.com/reader/full/centralized-mobility-load-balancing-scheme-in-lte-systemspdf 3/5

 procedure, the central server predicts the handover for UEs andthe transition of the cell load from the overloaded cell to itsneighboring cells. When a neighboring cell is predicted to

 become overloaded because of the cell load from anoverloaded cell, the central server also applies the HO marginadjustment algorithm to the neighboring cell. The followingshows the overall procedure in the central server.

C.  HO margin adjustment algorithm

The HO margin adjustment algorithm is performed for eachoverloaded cell. The objective of this algorithm is to move thecell load from the overloaded cell to its neighboring cellsconsidering the efficiency of the total PRB utilization among cells by adjusting the HO margin between the overloaded celland its neighboring cells.

When the HO margin is adjusted to move the load from the

overloaded cell to its neighboring cell, it is also required tominimize the increase in the total PRB utilization among cells.To realize both requirements, an objective function for the HOmargin adjustment algorithm with a penalty value is defined asfollows,

          (4)

where L s is the PRB utilization of the overloaded cell,  P  s is the penalty value of the overloaded cell.  Li  and  P i  are the PRButilization and the penalty value for a neighboring cell i,respectively. The penalty value P  in the objective function, thatindicates a degree of overload, is calculated for thecorresponding PRB utilization L as follows,

      (5)

where  P   is the penalty value, Th is a predetermined thresholdthat indicates whether a cell is overloaded or not, and   is aweighted value affecting the penalty value itself. That is, thehigher the PRB utilization becomes, the higher the penaltyvalue is to the objective function. In addition, the PRButilization L in the objective function is calculated as follows,

 

    (6)

where  N  prb is the number of PRBs in one subframe,  N  sub is thenumber of subframes in a second(i.e. 1,000 in LTE), TR is theminimum user throughput guaranteed to all UEs in bit/s andMCS(RSRQk  ) is the bit rate per PRB calculated given RSRQ ofUE k in the measurement reports. The serving cell of each UEis predicted by the set HO margin between the overloaded celland its neighboring cells as described in chapter III.

In order to find the minimized value of the objectivefunction effectively, dynamic programming[7] is used bywhich the number of the HO margin patterns to search will bereduced. The following is the process of the proposed HOadjustment algorithm using dynamic programming.

1.  Define list A as all neighboring cells of the overloadedcell.

2.  Initialize the value of the object function (=eval_value)

as INIT_MAX as a huge value.

3.  Take one neighboring cell i  from  A  and set the

temporal HO margin (=temp_margin[i]) for this

neighboring cell to a minimum value.

4.  Add the selected neighboring cell on the list B 

5.  num_cell  is set with the number of cells in B 

6.   Flag =1;

7.  while( Flag ==1) {

8.   Flag =0;9.  for(i=0; i<num_cell ; i++) {

10.  for( j=temp_margin[i]; j<Max_margin; j++) {11.  Calculate eval_value with j;12.  if (eval_value is smaller) {

13.  Update eval_value;

14.  Update temp_margin[i];15.   Flag =1; }}}}

16. Continue the process 2-15, until no cell remains in A.

17. Set the HO margin for each neighboring cell i  with

temp_margin[i].

1.  Collect information on the PRB utilization of all cells,the result of the measurement reports from UEs and thecurrent HO margin setting.

2.  Detect the overloaded cells by comparing the PRB

utilization of each cell with the predetermined

threshold.

3.  Arrange overloaded cells in the descending order of the

PRB utilization and put the overloaded cells on a list.4.  Take the first entry from the list, and apply the HO

margin adjustment algorithm.

A)  Once, a HO margin is set between the cell and

its neighboring cell, it will not be adjusted again

B)  If the neighboring cell is predicted to become

overloaded with the HO margin adjustment, theneighboring cell is also put on the list as an

overloaded cell in the descending order.

5.  While a cell remains in the list, rocess 4 continues.

  

  

 

Figure 2. Centralized solution

308

Page 4: Centralized Mobility Load Balancing Scheme in LTE Systems.pdf

7/18/2019 Centralized Mobility Load Balancing Scheme in LTE Systems.pdf

http://slidepdf.com/reader/full/centralized-mobility-load-balancing-scheme-in-lte-systemspdf 4/5

V.  MLB SIMULATION 

To evaluate the performance of the proposed centralizedsolution described in chapter IV, we perform a systemsimulation employing a distributed solution as a conventionalmethod and the proposed centralized solution in the downlinkof an LTE network. TABLE I gives the main simulation

 parameters.

TABLE I. SIMULATIONS PARAMETERS 

Parameter Value

Cell Layout  7 hexagonal sites, 3 cells per site

Site Distance 500 m

Pathloss 113.7+37.6*log(D) D in km

Shadowing

Lognormal shadowing SD: 8 dB

Intra-site correlation: 1.0

Inter-site correlation: 0.5

Antenna pattern70°(-3dB) with 20 dB front-to-back

ratio

System Bandwidth 10 MHz

Scheduler Max C/I

Call Admission Control(CAC) Not taken into account

Simulation duration 30 min

For setting cell loads for the simulations system, a background load with a low number of UEs for all cells and 10hotspots figured by small circles with a high number of UEsare allocated. Figure 3 and Figure 4 show two patterns ofhotspot distributions. In order to generate overloaded cellswhich are close to each other, 10 hotspots are generated atrandom within a rectangle as shown in Figure 3 and Figure 4.A UE in the background or a hotspot is generated on the basisof a Poison process and its connection time for which the UEconnects to the eNB is set on the basis of an exponentialdistribution. When the connection time expires, the UEdisappears from the simulation. While a UE connects to aneNB, constant data traffic for the UE is generated and sent

from the eNB to the UE. We assume a video streaming trafficwith a constant data rate of 64kbps per a UE and we regard thedata rate as a minimum guaranteed throughput for the UE. TheSINR value measured by a UE is mapped to a MCS based onlook up tables obtained from a link level simulation. Table II shows the values of the simulations parameters related to cellloads.

TABLE II. LOAD PARAMETERS 

Parameter Value

 Number of hotspots within a rectangle 10

Size of rectangle 500 m 500 m

Radius of hotspots 50 m

Averaged connection time  5 min

Averraged number of UEs in the

 background200

Averaged number of UEs in a hotspot 35 – 55

Generated user throughput (=guranteed

throughput)64 kbps

MCSsDecided base on the result of

the link level simulation

MLB is performed every 1 minute. In the centralizedsolution, the central server collects the information from eNBsand updates all HO margins in one turn as described chapter IV.In the distributed solution, when the serving cell of the eNB

exceeds a predetermined threshold which indicates whetheroverloaded or not, the eNB adjusts a HO margin between itsserving cell and its neighboring cell. If the neighboring cellload would exceed the threshold, the HO margin between theserving cell and the neighboring cell is not updated. Table III

shows parameters values of the proposed centralized solutionand the distributed solution.

TABLE III. ALGORITHM PARAMETERS 

Parameter Value

Initial HO margin 1 dB

MLB peforming interval 1 min

Overloaded threshold  90 %

  in the penalty function 3.5

Th in th penalty function 80 %

Threshold under which the HO marginis updated in the distributed solution

80 %

Each eNB in the simulations performs data scheduling witha simple Max C/I algorithm where a UE which has high SINRis prioritized. CAC (Call Admission Control) in an eNB is nottaken into account in these simulations. That is some UEslocated at the cell edge with a low MCS are assumed not toreceive the generated data traffic without MLB.

VI.  SIMULATIONS R ESULT AND EVALUATION 

 A.  Unsatisfied UE

We evaluated the performance of our proposed solution interms of the number of unsatisfied UEs against the distributedsolution and no MLB case. Since data traffic for each UE isgenerated as much as a minimum guaranteed throughput inthese simulations, a UE that does not receive a guaranteedthroughput is regarded as an unsatisfied UE.

 B.  Simulation Result and Evaluation

Figure 5 and Figure 7 show the average number ofunsatisfied UEs for pattern 1 and pattern 2 respectively with noMLB, with the distributed solution and with the proposedcentralized solution. The simulations are repeated varying theaverage number of UEs in each hotspot from 35 to 55. AsFigure 5 and Figure 7 show, the average number of unsatisfiedUEs of the proposed centralized solution is the lowestthroughout all simulations.

Figure 3. Hotspots distribution

 pattern 1

Figure 4. Hotspots distribution

 pattern 2

309

Page 5: Centralized Mobility Load Balancing Scheme in LTE Systems.pdf

7/18/2019 Centralized Mobility Load Balancing Scheme in LTE Systems.pdf

http://slidepdf.com/reader/full/centralized-mobility-load-balancing-scheme-in-lte-systemspdf 5/5

Figure 6 and Figure 8 show the transitiounsatisfied UEs through the simulations

 pattern 2 respectively where the average nuhotspot is 40. Since the number of UEs in thotspots vary because UEs are generated ba

 process and their connection time are sexponential distribution, the number of un

varies throughout the simulations. As Figushow, the proposed centralized solution canof unsatisfied UEs compared with the dithroughout the simulations.

In the distributed solution, when theexceeds the threshold 80%, the correspondallow to update the HO margin between itoverloaded cells. While the PRB utilizationunder the threshold, the overloaded cell isunsatisfied UEs to that cell. As a resulunsatisfied UEs on the overloaded cells in threduced.

On the other hand, the proposed ceupdates the HO margin in the descendingutilization of cells considering the total PR the penalty value among the overloadneighboring cells. Consequently, unsaoverloaded cells are handed off to their nereduce the total number of unsatisfied UEs a

 

Figure 5. Average number of unsatisfied UEs

Figure 6. Number of unsatisfied UEs in pattern 1 with

 of the number ofin pattern 1 and ber of UEs in onee background andsed on the Poisonet based on theatisfied UEs also

e 6 and Figure 8educe the numberstributed solution

PRB utilizationing eNB does nots serving cell andof that cell is keptnot able to move, the number ofe simulation is not

ntralized solutionorder of the PRBB utilization withed cell and itsisfied UEs onighboring cells toong all cells.

VII.  C

In this paper, we explaineadjustment using measurementintroduced the centralized soluthe cellar network adjusts

 predicting the transition of ce

margin adjustment algorithm esimulations results showed, thcan reduce the number ofcompared with the distributehotspots were close to each oth

R EFE

[1]   NGMN, “Use Cases related tDescription”, ver. 2.02.

[2]  3GPP TR 36.902 9.2.0, “Self-c(SON) use cases and solutions”.

[3]  3GPP TS 36.300 9.2.0, “EU protorol specification”

[4]  A. Lobinger, S. Stefanski, T.

downlink LTE self-optimizing nWorkshop COST 2100 SWGGreece, February 5, 2010.

[5]  R. Nasri and Z. Altman, “HaBalancing in 3GPP Long TInternational Conference onMultimedia (MoMM), Dec. 2007

[6]  I. Viering, M. D¨ottling, A. Loself-optimizing wireless networ Communications 2009 (ICC), Dr 

[7]  R. Bellman, “Dynamic Program1957.

in pattern 1

40 UEs in a hotspot

Figure 7. Average number o

Figure 8. Number of unsatisfied UE

 NCLUTION 

the aspects of the HO marginreports from UEs for MLB. Wetion where the central server inall HO margin among cellsll loads, and proposed the HO

ecuted in this central server. Ase proposed centralized solutionnsatisfied UEs in the system

solution in case that severaler among cells.

ENCES 

o Self Organising Network, Overall

nfiguring and self-optimizing network

TRA Radio Resource Control(RRC)

Jansen, I. Balan, “Load balancing in

tworks”, COST 2100 TD(10)071, Joint3.1 & FP7-ICT-SOCRATES, Athens,

dover Adaptation for Dynamic Loaderm Evolution Systems”, Proc. OfAdvances in Mobile Computing &

inger, “A mathematical perspective ofs”, IEEE International Conference on

esden, Germany, May 2009

ming”, Princeton University, Princeton

f unsatisfied UEs in pattern 2

s in pattern 2 with 40 UEs in a hotspot

310