This action might not be possible to undo. Are you sure you want to continue?
for publication in the IEEE Globecom 2010 proceedings.
Conflict Avoidance between Mobility Robustness Optimization and Mobility Load Balancing
Zhiqiang Liu, Peilin Hong, Kaiping Xue, Min Peng
The Information Network Lab of EEIS Department, University of Science and Technology of China (USTC), Hefei, China, 230027 firstname.lastname@example.org
Abstract—In Long Term Evolution (LTE) networks, Mobility Robustness Optimization (MRO) and Mobility Load Balancing (MLB) are two important functions to auto-optimize the network performances. There is a close correlation between them, as they both choose adjusting handover parameters as optimization actions. The conflict may occur between the two functions when they adjust the same handover parameter in opposite directions. This can not improve the performances but waste network resources. In this paper, we present a novel scheme to solve the problem. In order to prevent the occurrence of the conflict, we set an allowed range for MLB in which the handover problems can be prevented. Analyses and simulation results demonstrate that the proposed scheme can effectively solve the conflict problem and improve the performances of both functions. I. INTRODUCTION problems which can decrease user’s QoS. Furthermore, the proposal is in contradiction to the specification, as the specification declares that the load balancing shall not affect the user’s QoS negatively beyond what a user would experience at normal mobility without load-balancing . In this paper, we present a novel scheme to solve the conflict problem. We notice that the reason of the conflict is the inappropriate operation of MLB. If MLB adjusts handover parameters in an incorrect way, handover problems will occur and the conflict problem will follow on. In the proposed scheme, we set an allowed range for MLB to make sure that MLB operates correctly. Because of the restriction on MLB by the allowed range, handover problems are prevented and the conflict is avoided. The rest of this paper is organized as follows: Section II describes the operational principles of MRO and MLB, and explains the reason of the conflict. Section III presents the proposed scheme to solve the problem. Section IV evaluates the performances by system level simulation and Section V concludes the paper. II. DESCRIPTION OF THE CONFLICT
In recent years, with the rapid evolution of wireless communication networks, handover parameters have become more complex and larger, manual setting of these parameters is a time-consuming task and parameter updating after the initial deployment is also considered too costly. To reduce the operational complexity in wireless communication networks, Self-Organizing Network (SON) is seen as the promising area to simplify the management by auto-optimization . Therefore, SON is introduced to Long Term Evolution (LTE) and currently discussed in 3rd Generation Partnership Project (3GPP) standardization  . Mobility Robustness Optimization (MRO) and Mobility Load Balancing (MLB) are two important functions in SON. MRO aims to minimize handover problems   , while MLB aims to cope with the unequal traffic load between cells  . Although the two functions operate independently with each other, there is a close correlation between them, as they both choose adjusting handover parameters as optimization actions. It’s possible that the conflict occurs when they adjust the same handover parameter in opposite directions, e.g., handover parameter is adjusted by MLB to balance the load between cells and then changed back by MRO to minimize the handover problems . The conflict may lead to closed loop between the two functions, which will decrease the network performances seriously. To the best of our knowledge, the conflict problem has only been discussed in 3GPP standardization. In order to solve the problem, it is proposed that MLB should get the high priority and MRO should be suspended for a while during the operation of MLB . However, the proposal sacrifices the performance of MRO largely. This will bring about handover
In this section, at first we introduce the handover triggering procedure which is closely related with MRO and MLB, then we describe the operational principles of the two functions. At last, we explain the reason of the conflict. A. Handover triggering procedure in LTE The handover triggering procedure in LTE is illustrated in Fig.1. The x-axis represents the position of User Equipment (UE) and the y-axis represents the signal strength received by UE from E-UTRAN NodeB (eNB) in its cell. While a UE moves from cell1 to cell2, it measures the signal strengths of the two cells and checks whether the signal strengths satisfy the entering condition (1) for event A3 of cell1. If the entering condition is satisfied and lasts for duration of Time To Trigger (TTT), UE will send a measurement report to eNB1. After that, a handover procedure from eNB1 to eNB2 will be initiated on UE. M 2 > M 1 + ( H 1 − CIO1, 2 ) (1)
Where: M2: measurement result of cell2; M1: measurement result of cell1; H1: hysteresis parameter of cell1for event A3; CIO1,2: cell specific offset set by cell1 for cell2.
978-1-4244-5637-6/10/$26.00 ©2010 IEEE
MRO will postpone the handover triggering procedure. Ping-pong handover can also be solved by MRO. Here is an example of MLB. If RLF by too early handover is detected. As illustrated in Fig. the solution is to adjust handover parameters to satisfy (2). Operational principle of MRO Incorrect handover parameter settings will bring about handover problems. RLF affects user’s experience seriously since it causes the detachment of UE from the serving eNB. Fig. they are closely related to each other. in which the entering conditions for event A3 of the two cells are both satisfied. the MLB function of cell1 starts up and chooses cell2 to balance.1 lower.00 ©2010 IEEE . on the contrary. The two functions both optimize network performances through adjusting the handover parameters. It monitors the load periodically and adjusts handover parameter CIO when the traffic load is unbalanced . Ping-pong handover does not affect use’s QoS. This will affect user’s experiences and waste network resources. However. the value of entering condition for event A3 of cell1 is changed too low. MRO will make the handover triggering earlier. eNB1 increases CIO1. MLB increases CIO1.2 is eliminated and ping-pong handovers are avoided. and RLF by too early handover occurs after the operation of MLB. If handover is triggered too late. but it wastes the network resources. RLF will occur before the handover is initiated or during the handover procedure . the inappropriate operation of MLB may cause ping-pong handover or RLF by too late handover. RLF will occur shortly after the handover procedure. 2 ) + ( H 2 − CIO2 . The incorrect handover parameter settings cause the overlapped area. although MRO and MLB operate independently. ( H 1 − CIO1. B. If handover is triggered too early. after a while. This may lead to the long duration of heavy load conditions.1.1 ) > 0 (2) C. UE sends a measurement report to trigger handover implement at point B. The conflict occurs and leads to closed loop. Reason of the conflict between MRO and MLB In the original scheme.This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings. as the signal strength of the target cell is too low.2. eNB1 informs eNB2 to adjust CIO2. MRO can detect the handover problems in the cell through information gathering on UEs . For the purpose of load balancing in cell1. Similarly. ping-pong handovers will occur between cell1 and cell2 when a UE goes through.3. 978-1-4244-5637-6/10/$26. the overlapped area as shown in Fig. handover problems are brought during the conflict. On one hand. D. Here is an example as shown in Fig. another kind of “ping-pong”.2. In the operation procedure. The immediate cause of RLF is that the signal strength of UE’s serving cell is too low and the interference is too high .2 to make the handover triggering earlier. ping-pong handovers will occur between the two cells . The conflict problem between MRO and MLB is urgent to be solved as it decreases the performances of both functions. as the signal strength of the source cell is already too low. therefore UEs moving from cell1 to cell2 will handover out of cell1 earlier and then the load in cell1 Fig. In the overlapped area. When a UE goes through that area. therefore conflict may occur when they adjust the same parameter in opposite directions. H2 is the hysteresis parameter of cell2 and CIO2. because of the “ping-pong” effect between the two functions. TTT). On the other hand. therefore the handovers from cell2 to cell1 will be postponed and the increase of the additional load in cell1 will be slowed down. if RLF by too late handover is detected. Fig. the handover procedure accomplishes at point C. Operational principle of MLB MLB aims to cope with the unequal traffic load between cells. Meanwhile. Inappropriate handover triggering will lead to the occurrence of RLF. After that. such as Radio Link Failure (RLF) and ping-pong handover. MRO detects the RLF problem and adjusts CIO1. Unfortunately.2 to move UEs out of cell1 earlier. MLB changes it back once more. which can also lead to the occurrence of the conflict. CIO. the operation of MLB loses its efficiency.3. The entering condition for event A3 is satisfied at point A. Conflict due to inappropriate operation of MLB will be cut down.1 is the cell specific offset set by cell2 for cell1. which will bring about more blocking calls and dropping calls.2 lower to postpone the handover triggering procedure. We assume that cell1 has heavy load and its neighbour cell2 has light load. Its objective is to minimize RLFs and ping-pong handovers by auto-optimizing the handover parameters (H.
2 . 2. The entering conditions for event A3should be set inside the shaded areas during the operation of MLB III. the restriction on the handover parameters during the operation of MLB can be summarized as (3). an inquiry procedure for neighbor’s handover parameters is also required.00 ©2010 IEEE .late) is the maximum value of entering condition for event A3 that cannot cause RLFs in cell1. 978-1-4244-5637-6/10/$26.1'. early < H 1 − CIO1.CIO1.2) is set outside the shaded area.1 to CIO2.1.CIO2.2) is set inside the shaded area of cell1.CIO1.CIO2. CIO2. we can get the allowed range of CIO1.1.1 ' < H 2 − CIO2 .2' and CIO2. The proposed scheme requires MRO to record the minimum value and maximum value of entering condition for event A3 than cannot cause RLFs while MRO deals with RLF problems. 2 '+CIO2.5.1 are adjusted to CIO1. The scheme we describe is based on the restriction on the operation of MLB. ⎧CIO1. Therefore the key to solving the conflict is to prevent MLB from adjusting handover parameters in an incorrect way. 1.2) and (H2 .1' respectively after the MLB procedure. the entering conditions for event A3 of the two cells should satisfy (2) to prohibit the occurrence of ping-pong handovers. CIO1.1. 2 2 2 . The allowed range is calculated by eNB1 after the inquiry for cell2’s handover parameters (H2. CIO2. therefore the handover problems are prevented and the conflict is avoided. CIO2.4. 1 ' < CIO2 .2' and CIO2.1) should be changed inside the shaded areas during the operation of MLB. THE PROPOSED SCHEME In this section. the adjustment on CIO1. the last one is to avoid ping-pong handovers between the two cells. In order to avoid the conflict. If (H1 .CIO2. 4) eNB1 requests eNB2 to adjust CIO2.CIO2.1. late ⎪( H − CIO ' ) + ( H − CIO ' ) > 0 1.1. RLFs by too early handover and too late handover will be prevented while UEs move from cell2 to cell1 Fig.2' and CIO2.This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings. ⎧ H 1 − CIO1. a novel MLB procedure is designed to bring the proposed scheme into effect.early) is the minimum value and (H1 . 3) eNB1 calculates the allowed range from the handover parameters of the two cells(H1.CIO1.early. Furthermore. 2) eNB2 responses to the inquiry. the conflict will not occur as long as the operation of MLB causes no handover problems. Similarly. a novel MLB procedure is designed as shown in Fig.2 and CIO2.1 ' < H 1 + H 2 ⎪ (4) ⎨CIO1.1'. If (H2 .1.1.2 and informs eNB2 to decrease CIO2. We assume that the MLB function of cell1 intends to balance load to cell2. We then describe the principle of the proposed scheme and the designed MLB procedure.1 < H 2 − CIO2. A.late) is the maximum value that cannot cause RLFs in cell2.2 ' < H 1 − CIO1.1) is set inside the shaded area of cell2. In addition. 1 ⎩ B.2' and CIO2. (a).1' is restricted in the allowed range to prohibit the occurrence of the conflict.CIO2. While UEs move from cell1 to cell2. CIO2. as shown in Fig. 2 < CIO1.4 (b).4 (a).2 to CIO1. RLFs by too early handover and too late handover will be prevented while UEs move from cell1 to cell2 As illustrated in Fig. In other words. H2.1 ⎩ 1 (b). As shown in Fig.1. (H1 . 5) eNB2 accepts the request and adjusts CIO2. During the operation of MLB.4 (a)-(b). we set an allowed range for MLB to prevent handover problems during its operation.2'. Besides.1.late). CIO2.2.late). handover problems will occur and the conflict problem will follow on. (H1 . if MLB operates inappropriately. Principle of the proposed scheme As described in section II.1. the operation of MLB is restricted in the allowed range. The procedure can be described as follows: 1) eNB1 inquires about the handover parameters of cell2 ( H2.late). The purpose of the first inequality in (3) is to prevent RLFs by too early handover while UEs move from cell1 to cell2. UEs moving from cell2 to cell1 will handover successfully if (H2 . In the procedure.1. 2 ' < CIO1. early ⎪CIO 2 . eNB1 increases CIO1. To assume that CIO1. late < CIO2 .2.2. RLFs by too early or too late handover will occur if (H1 . The proposed scheme is based on the restriction on the operation of MLB.CIO1. we present a simple but effective scheme to solve the conflict. the second inequality is to prevent RLFs by too late handover while UEs move from cell2 to cell1. and works out the optimum values of CIO1. CIO2. (H2 .2.1' presented as (4) which is transformed from (3). in order to prevent the RLF problems. CIO1. 6) eNB1 adjusts CIO1. 2 ⎪ (3) ⎨ H 2 − CIO2.1) is set inside the shaded area of cell2.early) is the minimum value and (H2 . in order to get the allowed range. Besides. MLB procedure in the proposed scheme In order to bring the proposed scheme into effect.CIO1.
calls are accepted whenever there are free channels . It is obvious to view that the proportions of handover problems in the proposed scheme are lower than that in the original scheme. 61cell sites 1732m 2GHz 5MHz 43dBm 128. There are two reasons for this. in the original scheme. and it only choose the neighbor cell whose resource usage is lower than 80% to balance the load. CIO2.8 shows the ping-pong rate varying the number of UEs in the simulation topology. Meanwhile. CIO2.1. Fig. The traffic load in each cell is measured once 5s.2 Adjust CIO2. therefore the probability of the conflict increases and more handover problems are brought. it will reject other neighbors’ requests. R in km 10dB 10m 100ms 3dB 0dB 700~1300 144kbps. However. MRO begins operating.6log10(R).10. we evaluate the proposed scheme by dynamic system level simulation including MRO. Especially in 3kmph case. inappropriate operation of MLB brings a lot of ping-pong handovers which waste the network resources of the adjacent cells seriously. the operation of MLB is restricted in the allowed range to avoid the conflict problem. A. the rate in the original scheme is above 15%.9 and Fig. RLF rate and ping-pong rate are used to evaluate the performance of MRO. Call admission control adopts Queueing Priority (QP) scheme. the heavy load conditions last longer and more blocking calls 978-1-4244-5637-6/10/$26.6. real-time 3kmph. conflict occurs frequently and unnecessary handover problems are brought. the operation of MLB loses its efficiency. in the proposed scheme. MLB operates without restriction. This is due to the fact that in the proposed scheme. In the simulation topology. In the original scheme.1 Fig.1.6. the proportions of handover problems rise in both schemes. especially in the large number of UEs conditions. eNB1 eNB2 Inquire the handover parameters Response(H2.7 shows the RLF rate and Fig. it can be noted that with the growth of the number of UEs in the simulation topology.This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings. Besides. One reason is that in the original scheme. MRO and MLB both choose adjusting CIO as optimization actions. the conflict is avoided and the handover problems Parameter Cell layout Inter-site distance Carrier frequency Bandwidth eNB power SIMULATION PARAMETERS Assumption Hexagonal grid.5. UEs will connect to the network once more. call blocking rate and call dropping rate have higher percentages in the original scheme than that in the proposed scheme. Simulation topology TABLE I IV. Therefore. For another reason. 30kmph 2000s Distance-dependent path loss Shadowing standard deviation Correlation distance of Shadowing Time to Triger (TTT) Hysteresis parameter (H) Initial cell specific offset (CIO) configuration Number of UEs Traffic on UE UE speeds Simulation time are prevented during the operation of MLB. call blocking rate and call dropping rate are used to evaluate the performance of MLB.late) Calculate the allowed range and work out the optimum new settings Request eNB2 to adjust (CIO2. Other simulation parameters can be found in table I. The reason is that larger number of UEs causes more heavy load conditions and more operations related to load balancing in cells. PERFORMANCE EVALUATION In this section. i. the ping-pong rate is much reduced in the proposed scheme. and after the idle state of 20s’ duration. with the restriction on MLB.1 + 37. MLB procedure in the proposed scheme Fig. MLB operates without restriction. As illustrated in Fig.1') Accept the request Adjust CIO1. The simulation is driven in the Manhattan scenario . B. due to the “ping-pong” effect between MRO and MLB. MLB starts up when the resource usage of its cell reaches 90%. UEs move in the street represented by the bold lines as shown in Fig. which is kept below 1%. or 2 RLFs by too early / too late handover within 200s in its cell. If a cell accepts one neighbor’s request of MLB. MLB and handover functions. Operations of MRO and MLB In the simulation. there are 61 cells and wrap-around technique is used to avoid the edge effect. Simulation results This section presents a comparison between the original scheme and the proposed scheme.00 ©2010 IEEE . UEs’ probability of turning in the crossings is 0. The average call length of UEs is 20s.5. The method to detect RLFs and ping-pong handovers refers to  and . After detecting 3 ping-pong handovers within 50s.e.
Nokia-Siemens Networks.  In this paper. 2009AA012002 and the National S&T Major Project of China under Grant No.0 1. “Analysis of a hierarchical cellular system with reneging and dropping for waiting new calls and handoff calls. Besides. 3GPP TSG RAN WG1 Meeting #56bis. VI.902 v1.00 ©2010 IEEE . pp. http://www. 2008ZX03003-005..281-286.9. Nomor Research GmbH. In sum.5 4. Munich. “LTE Rel-8 Handover Performance in Manhattan Environment”.11. National High-Tech Key program of China under Grant No. no. http://www. http://www. pp. Ping-pong rate Original scheme (3kmph) Proposed scheme(3kmph) Original scheme (30kmph) Proposed scheme(30kmph) Fig.1080–1091.3gpp. Tiwana.0 3. “Exchange of handover parameters directly between eNBs”. Seoul. 2008. http://www. Eiko Seidel.0 2. and K. R1-091578.3gpp. 3GPP TSG RAN WG3 Meeting #64. J. R3-091032. CONCLUSION REFERENCES Sujuan Feng. Veh. R3-091294.5 1.org/  Nokia. May 2009. B.3gpp. “Distributed Self-Optimization of Handover for the Long Term Evolution”.3gpp. and Z. Chang. Call dropping rate Fig. Number of ping-pongs / handovers (%) 4. Indonesia. IEEE ICC 2009.10. therefore we set an allowed range for MLB to prevent the handover problems. 3GPP TSG RAN WG1 Meeting #56bis. Sayrac. In order to bring the proposed scheme into effect. Henrik Lundqvist. Call blocking rate Number of dropping calls / calls (%) 12 10 8 6 4 2 0 Number of operations of MLB 6000 5000 4000 3000 2000 1000 0 700 800 900 1000 1100 1200 1300 Number of UEs in simulation topology 700 800 900 1000 1100 1200 1300 Number of UEs in simulation topology Fig. R. Dresden.7. it can be deduced that more operations of MLB are needed in the original scheme. pp. Zwi Altman. The simulation results have demonstrated that the conflict can be avoided effectively in the proposed scheme. May 2009. Lo. 3GPP TSG RAN WG3 Meeting #64. we present the number of operations of MLB in both schemes as shown in Fig. 2009. pp. Jakarta.11. vol. It is obvious to view that with the growth of number of UEs in the simulation topology.0 0. R1-091559. more operations of MLB are needed. Austria.  M.5. Simulation has been done to evaluate the performances.  3GPP standardization.0 70 60 50 40 30 20 10 0 Number of blocking calls / calls (%) Original scheme (3kmph) Proposed scheme(3kmph) Original scheme (30kmph) Proposed scheme(30kmph) Original scheme (3kmph) Proposed scheme(3kmph) Original scheme (30kmph) Proposed scheme(30kmph) Number of RLFs / calls (%) 20 16 12 8 4 0 Original scheme (3kmph) Proposed scheme(3kmph) Original scheme (30kmph) Proposed scheme(30kmph) 700 800 900 1000 1100 1200 1300 Number of UEs in simulation topology 700 800 900 1000 1100 1200 1300 Number of UEs in simulation topology 700 800 900 1000 1100 1200 1300 Number of UEs in simulation topology Fig.  Ridha Nasri. Besides.org/  NTT DOCOMO.2. it can be noted that there are more operations of MLB in the original scheme than that in proposed the scheme. ACKNOWLEDGEMENT This work is supported by the National Natural Science Foundation of China under Grant No. we have designed a novel MLB procedure. “Statistical Learning for Automated RRM: Application to eUTRAN Mobility”.5 0.48. Germany. San Francisco. in contrast with the original scheme. These phenomena coincide with our analyses above. March 2009. March 2009. “Evaluation model for Rel-8 mobility performance”. we have proposed a novel scheme to solve the conflict problem between MRO and MLB. At last. I.145-154. “Dependencies among SON use cases and CCO priority”. Korea. http://www.  CMCC. “Handover Adaptation for Dynamic Load Balancing in 3GPP Long Term Evolution Systems”. INC. 2007. Vienna. May 2008. and Giorgio Nunzi.. 1999. Germany.org/  Alcatel-Lucent. TR 36.org/  André Schröder. 978-1-4244-5637-6/10/$26. Technol. on Advanced in Mobile Computing Multimedia (MoMM2007). Proceedings of the 3rd International Workshop on Self-Organizing Systems. 60772033. Conf.org/  C. V. US. Seoul.This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.” IEEE Trans. Altman.5 3. Korea. in order to prove our analyses. Number of operations of MLB and dropping calls are brought. “Self-Organizing Networks (SON) in 3GPP Long Term Evolution”.0. “Self-configuring and self-optimizing network use cases and solutions (Release 9)”.3gpp.8. US. Chang. We noticed that reason of the conflict is the inappropriate operation of MLB. 60903216. May 2009. C. San Francisco.5 2.4.1 . the proposed scheme can achieve better performances of both functions. RLF rate Original scheme (3kmph) Proposed scheme(3kmph) Original scheme (30kmph) Proposed scheme(30kmph) Fig.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.