You are on page 1of 5

1

Joint Optimization between MLB and MRO based


on Cell Load Balance for LTE Networks
Jun Zhou, Yijun Mo, Bang Wang
Department of Electronics and Information Engineering,
Huazhong University of Science and Technology
EMail: {moyj, wangbang}@hust.edu.cn

AbstractMobility Load Balancing (MLB) and Mobility Robustness Optimization (MRO) are two important functions in
Long Term Evolved (LTE) cellular mobile networks. They are
closely related because they both adjust handover parameters to
achieve their respective objective. But some conflicts may occur
if both of the two functions work inconsistently and adjust the
same parameter in opposite directions. This will result in huge
resource waste, and both of them will not work efficiently. In this
paper, we present a scheme embedded in MLB based on the cell
load to solve the conflict problem. Analysis and simulation results
show that the proposed scheme works efficiently and improves
the performance of the network.

I. I NTRODUCTION
In order to meet the ever-growing needs, the 3rd Generation
Partnership Project (3GPP) makes a comprehensive improvement in Long Time Evolution (LTE) system relative to the
3G network system. However, along with the performance improvement, the system becomes more complicated and the network optimization becomes time-consuming, capital-costing
and error probabilities-raising for the traditional manual way.
Therefore, the Next Generation Mobile Networks (NGMN)
organization and 3GPP propose Self-organization Networks
(SON) technology aiming at solving those contradictions.
The SON technology brings in some automation mechanisms
instead of manual work during the network deployment and
operation stages [1][3].
Mobility Robustness Optimization (MRO) and Mobility
Load Balancing (MLB) are two important functions in SON.
MRO adjusts handover parameters to minimize handover problems such as too-late handover, too-early handover, Ping-Pong
handover and so on, while MLB adjusts handover parameters
to balance the unequal traffic load among cells. Both of MRO
and MLB adjust the parameters to achieve their objectives
independently. So there may be a conflict when they adjust
a same parameter in opposite directions according to their
respective requirement at the same time. For example, MLB
needs to adjust the value of some handover parameter smaller
to balance the load, and meanwhile MRO needs to adjust the
value of the same handover parameter larger to reduce the
handover problems. This conflict will lead to a vicious cycle
between these two functions, which will waste the resources
and decrease the performance of the network severely.
This conflict problem has only been discussed in 3GPP
standardization. In order to solve the problem, it is proposed
that MLB should get higher priority and MRO should be

suspended while MLB is working. However, this proposal


just ensures that MLB and MRO will not work at the same
time but it does not solve the problem essentially. Some other
proposals set an allowed range for MLB or MRO to make them
operate correctly to avoid the conflict problem [4][5]. But these
proposals do not consider the users QoS requirements.
In this paper, we present a novel scheme to solve the
conflict problem. We notice that users QoS will reduce if its
serving cell is overloaded. The proposed scheme is triggered
when some cell is overloaded. Its objective is to reduce the
overloaded cells load level rapidly through adjusting the
handover parameter to a reasonable value without causing
the handover problems. The proposed scheme can improve
users QoS while solving the conflict problem of MLB and
MRO. Simulation results show that the system too-early, toolate, ping-pong handovers and Radio Link Failure rate can be
reduced compared with the original one and an existing improved proposal. Furthermore, the overall system throughput
can also be increased.
The rest of this paper is organized as follows: Section
II describes the operational principles of MRO and MLB.
Section III explains the conflict and resents the proposed
scheme. Section IV evaluates its performances via system level
simulations. The paper is concluded in Section V.
II. D ESCRIPTION OF H ANDOVER IN LTE
As both of MLB and MRO adjust handover parameters to
realize their respective goals, we first discuss the handover
triggering procedure, and then describe the operational principles of the two functions. At last, we explain the reason of
the conflict.
A. Handover triggering procedure in LTE
In the LTE system, a UEs handover from one cell to another
is triggered when the received signal strength from the two
cells meet the following A3 event during a certain time period
[2][7]. We suppose that UEs handover from source cell1 to
target cell2. The A3 event inequality is given by:
M2 > M1 + (H1 CIO1,2 ),

(1)

where M1 and M2 , respectively, is the received signal strength


UE from cell1 and from cell2, H1 the hysteresis parameter of
cell1 for A3 event, and CIO1,2 the cell individual offset set
by cell1 for cell2. The time to trigger (TTT) is just a time

978-1-4799-0308-5/13/$31.00 2013 IEEE

0 +&,2

0

0

0

0

0 +&,2

0 +&,2

&HOO

&HOO

3RVLWLRQ

8(
$ % &

Fig. 1: A3 Event in LTE

&HOO

&HOO

3RVLWLRQ

8(

; <
Fig. 2: Ping-pong Handover

period. If M1 and M2 satisfy the A3 event by Eq.(1) during


one TTT, then the handover from cell1 to cell2 is triggered.
Fig. 1 illustrates the handover process: The point A means
that the entering condition for A3 event has just been satisfied,
the point B means that the UE starts to handover from cell1 to
cell2, and the point C means the that handover is just finished.
The time period while the UE moves from point A to point B
equals to TTT.
B. Mobility Load Balance
MLB aims to balance some load from the overloaded cell
to its light-loaded neighbor cell[6]. MLB works in three ways:
1) Relax the handover conditions from overloaded cell to
its neighbor cell to make UEs handover out of the overloaded
cell in advance.
2) Tighten the handover conditions from neighbor cell to
the overloaded cell to make UEs postpone the handover into
the overloaded cell.
3) Do the above two at the same time.
As can be seen from the A3 event, we assume that cell1
has heavy load and its neighbor cell2 has light load. The MLB
function in cell1 will choose cell2 to balance the load. The
MLB function in cell1 will increase the parameter CIO1,2
to make UEs handover from cell1 to cell2 earlier. Meanwhile
cell1 informs cell2 to decrease CIO2,1 to make UEs handover
from cell2 to cell1 postponed. So the increase of the additional
load in cell1 will be slowed down on the whole.
C. Mobility Robustness Optimization
In the 3GPP standard, the primary target of MRO is to
reduce radio link failure (RLF) caused by handover and reduce
unnecessary handover [8]. Specifically, RLF will be caused by
too-early handover and too-late handover and the unnecessary
handover means Ping-Pong handover in general.
1) Too-early handover: When UE moves from cell1 to cell2,
the entering condition for A3 event is just satisfied at point A
(as shown in Fig. 1) and the handover is triggered at point B.
But because of unreasonable values of handover parameters,
the received signal strength of UE from cell2 is below Mth ,
which is required to establish a radio link with a cell. So the
RLF between the UE and cell2 occurs during the handover or

in a short time after the UE successfully handover to cell2. In


this case, UE will reconnect to cell1.
2) Too-late handover: When UE moves from cell1 to cell2,
the handover is triggered at point B (as shown in Fig. 1).
Because of unreasonable values of handover parameters, the
received signal strength of UE from cell1 is below Mth before
the UE gets to point C so that the RLF between UE and cell1
occurs. In this case, the UE will connect to cell2.
3) Ping-Pong handover: As it is shown in Fig. 2, when
UEs move between point X and point Y, the A3 events which
trigger handover from cell1 to cell2 and handover from cell2
to cell1 are all satisfied. So after UEs handover from cell1 to
cell2, they will handover back to cell1 immediately. The PingPong handover occurs when UE moves between point X and
point Y. Although Ping-Pong handover does not affect users
QoS too much, it wastes the network resources severely.
III. T HE P ROPOSED S CHEME
A. Conflict between MLB and MRO
We next introduce how conflict may occur between MLB
and MRO. As shown in Fig. 3, we assume that cell1 is
overloaded and it chooses its neighbor cell2 to balance the
load. The MLB in cell1 will adjust the value of the handover
parameter CIO1,2 larger to CIO1,2M LB to relax the handover
conditions such that UEs in cell1 will handover to cell2 in
advance. But due to the inappropriate operation of MLB, the
CIO1,2 is adjusted to a value which unfortunately causes
many too-early handovers of UEs in cell1. Then the MRO
function detects many RLFs, so it will adjust the value of
CIO1,2 smaller to CIO1,2M RO to prevent too-early handover.
However, the MLB changes it back once again after a while,
and MRO will also change it in the opposite direction immediately. This will lead to an endless loop, unless one of these
two functions stops working. This conflict is a serious problem
because it will significantly reduce the performance of the two
functions and waste network resources severely.
B. The Proposed Scheme
In this section, we present an effective scheme to avoid the
conflict between MLB and MRO. The scheme is to adjust

0/%

0 +&,2PLQ

052

0 +&,2

0

0

0

0 +&,2

0PD[
0 +&,20/%

0

0 +&,2052
0 +&,2

&HOO

07+
&HOO

&HOO

&HOO

3RVLWLRQ

8(

Fig. 3: Reason of Conflict between MLB and MRO

$ $

Fig. 4: Adjust point A to A1


the handover parameters based on the load distribution of the
serving cell and the target cell.
We assume that the serving cell named Cell1 is over loaded.
It will choose the light loaded target cell named Cell2 to
balance the load. To achieve this purpose, Cell1 should adjust
the handover parameters to make the UEs handover from Cell1
to Cell2 more easily and make the UEs handover from Cell2 to
Cell1 more difficultly. This is equivalent to that the handover
position should be closer to Cell1 whether UEs handover from
Cell1 to Cell2 or handover from Cell2 to Cell1.
To make the handover position of UEs handover from Cell2
to Cell1 move from point A to point A1 (Fig. 4), the handover
parameters are adjusted as follows:
CIO2,1min = MT H + H2 M1max ,
Load2
1 = (CIO2,1 CIO2,1min ) (1
),
Load1

0 +&,2

0 +&,2

0

0

0 +&,2
0 +&,2

07+

&HOO

&HOO
3RVLWLRQ

8(
$ $ % %

(2)

Fig. 5: Adjust point B to B1


(3)

and
CIO2,1 = CIO2,1 1 ,

(4)

where MT H is the threshold of the received signal strength


required by handover, M1max the received signal strength of
the UE from Cell1 when the received signal strength the UE
receives from Cell2 is equal to MT H , CIO2,1min a critical
value that makes the A3 event just to be satisfied. To prevent
the too late handover, when UE moves from Cell2 to Cell1,

CIO2,1 should be adjusted to CIO2,1 which is between the


value of CIO2,1 and CIO2,1min . The step 1 is decided
by the load of Cell1 and Cell2. The smaller the value of
(Load2 /Load1 ), the larger the value of 1 and the closer the
point A1 to Cell1.
To make the UEs in Cell1 handover to Cell2 more difficultly,
we will make the UEs handover positions from Cell1 to Cell2
be closer to Cell1. This means point B will be changed to point
B1 as illustrated in Fig. 5.
In Fig. 5, the point A1 is the new handover position when
UEs move from Cell2 to Cell1 and point B1 is the new
handover position when UEs move from Cell1 to Cell2. In
consideration of Ping-Pong handover, the point B1 should be
further from Cell1 than the point A1 , but closer to Cell1 than
the point B. As we know from Fig. 2, if the point A1 and

point B1 are overlapped, the parameters H1 , H2 , CIO1,2 and

CIO2,1 should be satisfied the following equation:

(H1 CIO1,2 ) + (H2 CIO2,1 ) = 0.


So, if the position of the point B1 is just as shown in Fig. 5,
two inequalities should be satisfied:
{

(H1 CIO1,2 ) + (H2 CIO2,1 ) > 0,


(5)

CIO1,2 > CIO1,2 .

Then we can get the adjustable range of CIO1,2 as follows:


{

CIO1,2 + CIO2,1 < H1 + H2 ,


(6)

CIO1,2 > CIO1,2 .


This proposed scheme is triggered by MLB, so the MLB
procedure should work as follows:
(1) The MLB function in Cell1 collects the load statistics
of neighbor cells;
(2) The Cell1 detects an eligible neighbor (Cell2) and
requests its handover parameters CIO2,1 and H2 ;
(3) The Cell1 adjusts the parameter CIO1,2 and CIO2,1 to

CIO1,2 and CIO2,1 according to equality (4) and inequality


(6);

(4) The Cell1 transfers the parameter CIO2,1 to Cell2, and


both of the Cell1 and Cell2 synchronize the new parameters.

TABLE I: Parameters&Values

0/%WULJJHUHG

/RDG
VWDWLVWLFV
1R

052,QGH[

,VWKHUHVRPHHOLJLEOH
QHLJKERU&HOO"

Value

Frequency

2GHz

Bandwidth

5Mth

Inter-site Distance

500m

eNodeB TX Power

43dBm

Microscopic Pathloss

<HV
$GMXVW&,2 WR&,2

Parameter

5HGXFH

7RRODWH+2

40ms

Hysteresis Parameter

3dB

Initial Cell Individual Offset


Number of UEs
UE Speed

$GMXVW&,2 WR&,2

5HGXFH

7RRHDUO\+2
3LQJ3RQJ+2

128.1+37.6lg(R)

Time to Thigger

Simulation Time
TTI length

0dB
42,63,84,105,126,147,168
50km/h
10000TTI
1ms

3DUDPHWHUV
V\QFKURQL]DWLRQ

Fig. 6: Improved MLB Procedure

The improved MLB procedure not only modifies the handover parameters of the native cell but also the handover parameters of neighbor cells. And it will not work independently,
because it ensures the MRO indexes when it is working to
balance the load.

IV. P ERFORMANCE E VALUATION


In this section, we evaluate the proposed scheme by a LTE
system level simulator. This simulator is developed by Josep
Colom Ikuno et. al., from Vienna University of Technology in
Austria[10]. In our simulations, there are 7 eNBs (one is in
center and other six around it) and 21 cells. At the beginning
of the simulation, UEs are generated at some random positions
in the network. During the simulation time, UEs always
connect to the network and keep calling. When the simulator
starts running, every UE chooses random direction and moves
straight with a special speed. And if some UE moves out of the
region of interest(ROI), this UE will be randomly regenerated
to appear again in the ROI and reallocated some recourse.
Other parameters can be found in Table I. We collect the
simulation statistics to compute the following performance
metrics: Too-early handover rate, too-late handover rate, pingpong handover rate and the RLF rate are used to evaluate the
performance of MRO, and the number of overloaded cell per
TTI and the average throughput of all cells are used to evaluate
the performance of MLB.
At the beginning of every TTI, the system will detect the
load level of every cell. If the resource usage is above 70
percentage, the MLB will be triggered in this cell. It will
choose its neighbor cell whose resource usage is below 50
percentage to balance the load. In consideration of MRO, the
parameter CIO of this overload cell and its neighbor cell will
be adjusted to reasonable values according to our proposed
scheme.

A. Simulation Results
In this section, we present the simulation results for performance comparison among the original scheme[1][9], the
improved scheme[4] and the proposed scheme. In the original
scheme, the MLB and MRO operate independently and they
modify the handover parameters according to their respective
demand. In the improved scheme, the MLB modifies the
parameter CIO to a reasonable range to prevent the conflict
between MRO and MLB. In the proposed scheme, the MLB
will modify the parameters according to the MRO indexes and
the load level of neighbor cells to avoid the conflict.
Figure 7(a),7(b),7(c) and 7(d) show that the MRO indexes in
the proposed scheme including the rate of too-early handover,
too-late handover, ping-pong handover and RLF are significantly better than the original scheme, and also outperform
the improved scheme. This is because the MLB function in the
proposed scheme adjusts the parameter CIO to a reasonable
value according to the load distribution of neighboring cells
and MRO indexes, while the MLB in the improved scheme just
obtains the reasonable range according to the MRO indexes
and so the parameter values may not be appropriately adjusted.
Figure 7(e) and Figure 7(f) show that the system average
throughput increases and the average number of overload
cells per TTI decreases in the proposed scheme. This is
because after the MLB operation, the handover positions of
UEs moving from overloaded cell to its light load neighbor
cell and the handover position of UEs moving from the light
load neighbor cell to the overload cell all get closer to the
overloaded cell. Therefore, the UEs can handover from the
overloaded cell to its neighbor cell more easily, and the UEs
handover from the neighbor cell to the overload cell more
difficultly. This makes the UE growth rate of the overloaded
cell is lower than the UE decrement rate. So the load level of
the overload cell decreases. Meanwhile all of the UEs can get
sufficient resource from the cells they connect to so that the
system average throughput increases as we expected. While the
improved scheme only gives a reasonable parameter change
range without considering the load distribution particularly, so
the overall handover performance of the proposed scheme is

speed=50km/h

3
2.5
2
1.5
1
0.5
40

4
3.5
3
2.5
2
1.5

60

80

100
120
number of UEs

140

160

1
40

180

60

80

average throughput of all cells(Mbps)

rate of RLF(%)

5
4
3
2
80

100
120
number of UEs

30
25
20
15
10

140

160

0
40

180

60

80

140

160

180

(d) rate of RLF

100
120
number of UEs

140

160

180

160

180

(c) rate of pingpong HO

speed=50km/h

60

100
120
number of UEs

speed=50km/h

1240

original
proposed
improved

1
40

35

(b) rate of toolate HO

speed=50km/h

original
proposed
improved

(a) rate of tooearly HO


9

40
rate of pingpong hanover(%)

4.5

speed=50km/h
45

original
proposed
improved

1220
1200

original
proposed
improved

1180
1160
1140

number of overloadcell per tti

rate of tooearly hanover(%)

3.5

speed=50km/h
5

original
proposed
improved
rate of toolate hanover(%)

2.5

original
proposed
improved

1.5

1120
1100
40

60

80

100
120
number of UEs

140

160

180

(e) average sum of throughput of all cells

0.5
40

60

80

100
120
number of UEs

140

(f) average number of overload cells per tti

Fig. 7: Simulation results

better than the improved scheme.


V. C ONCLUSION
In this paper, we have proposed a new scheme to solve the
conflict problem between MLB and MRO for LTE networks.
We found that when a cell is overloaded or blocked, the RLF
and the call dropping occur frequently, which affects users
QoS and system throughput severely. So the proposed scheme
is based on the load distribution of neighboring cells to adjust
the handover parameters. System level simulations have shown
that the proposed scheme can greatly reduce the too-early, toolate, ping-pong handover rate and the RLF rate. Furthermore,
it can reduce the number of overloaded cell and improve the
overall system throughput.
Acknowledgement: This paper is supported in part by Doctoral
Fund of Ministry of Education of China (20110142120078)
and Fundamental Research Funds for the Central Universities
(No. 2013QN139, 2013QN140).
R EFERENCES
[1] 3GPP standardization, Self-configuration and self-optimization network
use cases and solution (Release 9), TR 36.902 v1.2.0, May 2009
[2] Alcatel-Lucent, Exchange of handover parameters directly between eNBs,
San Francisco, US, 3GPP TSG RAN WG3 Meeting, R3-091294, May
2009

[3] Sujuan Feng, Eiko Seidel, Self-Optimization Networks(SON) in 3GPP


Long Term Evolution, Nomor Research GmbH, Munich, Germany, May
2008
[4] Zhiqiang Liu, Peilin Hong, Conflict Avoidance between Mobility Robustness Optimization and Mobility Load Balancing, IEEE Globecom
2010 proceedings, University of Science and Technology of China, Hefei,
China, 2010
[5] Jiantao Yu, Honglin Hu, Conflict Coordination Between Mobility Load
Balancing and Mobility Robustness Optimization, Computer Engineering
January 2012, Shanghai Research Center of Wireless Communication,
Shanghai, China, 2012
[6] Andreas Lobinger, Szymon Stefanski, Coordinating Handover Parameter
Optimization and Load balancing in LTE Self-Optimization Networks,
Vehicular Technology Conference (VTC Spring), Nokia Siemens Network, 2011
[7] Jihai Han, Bingyang Wu, Handover in the 3GPP Long Term Evolution
(LTE) Systems,
Mobile Congress (GMC), 2010 Global, Southeast
university, Nanjing, China, 2010
[8] Koichiro Kitagawa, Toshihiko Komine, A Handover Optimization Algorithm with Mobility Robustness for LTE systems, Personal Indoor and
Mobile Radio Communications (PIMRC), Fujimino-shi, Saitama, Japan,
2011
[9] CMCC, Bingyang Wu, Dependencies among SON use cases and CCO
priority,
San Francisco, US, 3GPP TSG RAN WG3 Meeting, R3091294, May 2009
[10] Jcolom, Martin.taranetz, Vienna LTE Simulators System Level Simulator
Documentation, Institute of Telecommunications, Vennia University of
Technology, Austria, 2010

You might also like