A Novel Anchor Selection Scheme for Distributed Mobility Management

Computer Science •22(1) 2021 https://doi.org/10.7494/csci.2021.22.1.3708 Battulga Davaasambuu Tumnee Telmuun Dominik Sasko Yu Keping Shirmen Sodbileg NOVEL ANCHOR-SELECTION SCHEME FOR DISTRIBUTED MOBILITY MANAGEMENT Abstract The number of subscribers in mobile networks is growing rapidly, which challenges network management and data delivery. Efficient management and routing are key solutions. One important solution is distributed mobility management (DMM), which handles the mobility of subscribers at the edges of mobile networks and load balancing. Otherwise, mobility anchors are distributed across a network that can manage the handover procedures. In this paper, we propose a novel mobility anchor-selection scheme based on the results of a cost function with three factors to select a suitable cell as well as an anchor for moving subscribers and improving the handover performances of networks. Our results illustrate that the proposed scheme provides significantly enhanced handover performance. Keywords cost function, DMM, distributed anchor, handover, mobility management Citation Computer Science 22(1) 2021: 143–163 Copyright ©2021 Author(s). This is an open access publication, which can be used, distributed and reproduced in any medium according to the Creative Commons CC-BY 4.0 License. 143 144 Battulga Davaasambuu et al. 1. Introduction Network technologies are rapidly developing and changing to meet the requirements of customers. On the other hand, 5G network has led to challenges for mobile network operators. Also, new paradigms have been integrated into the 5G network architecture such as network function virtualization, software-defined networks, edge computing, and network slicing [11,12, 29, 36]. Following these paradigms, the 3GPP’s working groups focus on the improvement of a mobile network’s performance, developing it towards a distributed and softwarized one [1–3, 22]. As a result, the mobile network architecture is changing from homogeneous to heterogeneous and with increasing complexity. There are a huge number of mobile subscribers who are generating and transferring massive mobile data traffic, and this is rapidly growing [11]. From a Cisco report, the amount of mobile data traffic grew by 71% in 2017. In addition, this is expected to increase by nearly tenfold by 2022, reaching 77 exabytes per month. Moreover, mobile subscribers and connections grew to 8.6 billion in 2017 [12]. These growths are affecting the signaling overhead of networks. In a next-generation network, mobility management needs to efficiently manage the mobility of subscribers in heterogeneous and complex networks. However, network management is becoming more difficult due to the integration process, such as the distributed and softwarized technologies in mobile networks [36]. The following issues are challenging the performance of centralized mobility management (CMM). 1. In a dense and heterogeneous network, cells of smaller sizes (such as small cells, femtocells, and picocells) are denser and more uncontrollable [10,20]. Therefore, mobility management needs changes and features for a heterogeneous network. 2. Many factors affect network performance, including simultaneously executing handovers, mobility speed, the load of the mobility anchors (MAs), the available the radio resources (RRs) of a cell, and the number of active subscribers and their usage. The impact of these factors varies depending on the current network environment. 3. The concentration of subscribers is a cause of overload in a certain section of an access network. At a certain section of the network, the mobility of massive subscribers is a reason for the increased load of a centralized MA; this also decreases the performance of the handover procedure. In other words, a single point of failure is a significant drawback of CMM because the subscriber’s mobility is handled by a centralized MA located at the core network. This paper focuses on the second issue, where a subscriber is moving across distributed anchors. To provide efficient mobility management and solve the abovementioned issues, a network-based distributed mobility management (DMM) solution is to be developed with a load balance function and handover procedure execution at the edge of a network [9]. In DMM, MAs are distributed across a network, and their serving area is defined by geolocation or a set of cells. The subscribers’ registration Novel anchor-selection scheme for distributed mobility management 145 processes and their data paths are also managed by an MA (such as a data gateway with an MA function). In cellular networks, there are two types of handover procedures: intra and inter handovers. When subscribers are moving inside the serving area of a current MA, the intra handover procedure is executed between two cells. The inter handover procedure executes between two MAs if the subscriber is leaving the serving area of a current MA. Also, the network begins the inter handover procedure between MAs when load balancing and optimal path selection are necessary. In the existing handover, a selection scheme is important. If the wrong cell or MA is selected, a handover request is dropped by the call admission control and subscriber tries again to start a new handover procedure. This is one of the reasons for high handover delay and unsuccessful handovers (including request drops). The existing selection scheme is a distance-based scheme in the current development of DMM. For example, the nearest selection scheme (NS) selects an MA that can provide the shortest path between an MA and a subscriber [9]. Also, this path can provide a low-latency connection. However, in order to provide a load balance between MAs, a network will change the MA again after a successful handover procedure. From the above example, a selection scheme with one factor cannot select an optimal connection path with low-latency and load balancing. Furthermore, optimizing the factors is necessary. In DMM implementation, we defined one possible feature individually, which is a selection scheme that selects a target cell and MA. In other words, a selection scheme selects a cell from a neighbor cell list (NCL) and an MA from the possible sets. Then, the subscriber’s data path is established over the selected cell and MA in the handover execution. For example, any cell can establish a connection with any MA to provide an optimized path over the selected cell and MA for a subscriber’s connection performance. In the current DMM, a network selects a target cell or MA because the cells have established a fixed connection to the MA. In this paper, we propose an anchor-selection scheme that is based on multiple factors for DMM called cost-based mobility anchor selection (CMAS) to select a suitable cell and MA for the provision of the QoS of a subscriber’s connection. The major contributions of our proposal are twofold: (i) we present a handover with multifactor-based MA selection that addresses the improvement of handover performance (including decreased request drops) and load balancing in handover execution; and (ii) a simulation model developed to compare our proposal with existing well-known schemes. The remainder of the paper is organized as follows: In Section 2, we review related works. The details of our proposed cost function are presented in Section 3. In Section 4, we introduce the proposed selection scheme. We describe the network model and set up to study the performance of the proposed scheme in Section 5. Finally, Section 6 concludes the paper. 146 Battulga Davaasambuu et al. 2. Related works In this section, we review some mobility solutions and classify the related works into DMM and anchor-selection schemes. The DMM paradigm has been developed by IETF [9]. DMM is an important paradigm for mobile network architecture. Recently, researchers have proposed some DMM architectures: partial, full, and hybrid solutions. The partial DMM solution is proposed in [18]. The results of this solution showed the reduced signaling messages by 30 percent. In [6], a fully distributed solution is introduced and discussed. In [8,26,28], the authors proposed hybrid mobility management that is a combination of DMM and PMIPv6. In the hybrid approach, a centralized MA can provide communications and information exchanging among regional MAs that are to control its connected macro-cells and the cell-level mobility of UEs. In [25], the authors proposed a DMM solution with Double-NAT that separated an identifier from locator and distributed anchors routing the packets in the transport network. Also, the authors of [32] proposed dynamic mobility anchoring, which can dynamically anchor a user’s traffic by routers in an access network. IPv6 and its mobility approach are presented in [13]. PMIPv6 is a network layer protocol that is meant to support mobility management by a local mobility anchor (LMA). Actually, an LMA is a centralized management node that manages the mobility of subscribers across mobile access gateways (MAGs). The researchers have proposed the LMA discovery and selection schemes in [36, 22]. These schemes provide two main improvements: (1) the load balancing of LMAs and MAGs; and (2) a reduced number of dropped handovers at LMA/MAG. In order to meet 5G’s mobility requirements, PIMPv6 replaces a centralized LMA with a control mobility database (CMD) that includes the rules between the distributed anchors and routers [6,8,38]. The integration of SDN and DMM has recently been presented. Publications such as [23,35] introduced SDN-based DMM solutions. In addition, [23] introduced an SDN controller, which included two functions: location, and handover management. In [31], a DMM solution with two controllers is presented. These controllers are a local controller for cell-level mobility and a regional controller for handover from a serving cell controlled by an MA to a target cell controlled by another MA. Also, in [31], the authors presented the characteristics of inter and intra handovers (the same as LTE mobility management). In [27], the author introduced a method of using SDN-based DMM in 5G networks and compared the existing DMM proposals with the proposed SDN-based DMM. The authors placed their proposed DMM on top of an SDN controller as an application server. The result of this solution showed that the complexity of the control plane decreased and became more scalable in terms of handover procedure delay and transmission delay. The integration of SDN and DMM can manage mobile data traffic by MAs and gateways that are located at the edges of access networks. Two main advantages can improve the handover performance in an SDN-based DMM [19]; these are a fast handover procedure and an SDN controller’s optimal data path selection. Novel anchor-selection scheme for distributed mobility management 147 In [34], the author introduced and evaluated a novel SDN/OpenFlow-based DMM approach that could be applied to virtualized LTE systems. In the proposed approach, an X2 interface was used for handover procedures between P-GWs, and network traffic could be seamlessly propagated by a target P-GW. Simulation results showed that the handover time was less than 150 ms, which meets the requirements of LTE and LTE-A networks. In [4], the researchers introduced the possible use-cases of an MA-selection approach based on three different contexts (mobile, application, and network). However, they have not introduced the use-case of the hybrid approach that is the multiple factor-based approach. A dynamically MA selection approach is presented in [7]. This approach can provide minimized packet delay and a balanced load of switches. However, the author’s proposed mobility management is only for SDN-enabled networks. In [21], anchor selection is presented to provide load balancing. The main proposal is a middle point, which is an optimal location of an anchor between the corresponding and mobile nodes. The blocking and dropping probability of new call and handover requests are presented with reduced values in the numerical results. This approach can only provide load balance. However, the other performance factors of the network are uncertain. In [14], the introduced context-aware anchor selection includes handover procedures and twoqueued anchoring. The main feature is re-selection after the handover procedures. All of the works discussed above introduce MA selection. Furthermore, distance and load-based MA selections for multicast and distributed anchors are presented in [17]. 3. Cost function In this section, we describe the cost function with three normalized functions that calculate the cost of all of the possible combinations of MAs and cells. In the other words, our proposal is focused on achieving efficient mobility management with MA selection, taking advantage of improved handover performance and load balancing. In [24], the authors proposed an adaptive handover hysteresis based on a cost function with the following factors: 1) a difference in the load information between the serving and target cells; 2) the moving speed of subscribers; and 3) the subscribers’ active services. Based on these factors, the handover triggering is optimized to decrease unsuccessful handovers (including radio link failures). In our previous work [15,16], we proposed an adaptive hysteresis that was based on multiple factors. Also, we defined the cost function to provide the relationship between these factors, such as the speed of a subscriber, the loads of the cells, and the service type of the subscriber’s active sessions. The simulation results of our previous work show a reduction in the radio link failures that affected the handover performance. Also, the cost function and these factors are related to special environments such as high-speed mobility and centralized management. In this paper, we redefined the cost function for DMM. Our proposed solution is based on several factors: the load information of candidate MAs, the available 148 Battulga Davaasambuu et al. RRs of candidate cells, and handover performance. The first two factors mean those indicators that are the number of dropped handover requests at the MA or cell, and handover performance is an indicator of other handover failures. An MA’s load is a new influential factor for performing a successful handover in DMM. An MA’s total capacity and the size of all active traffic can predict the quality of service (QoS) after a successful handover. If an MA with a high load is selected, the network may begin the handover procedure because of the load-balancing function. Also, a handover request drops by an MA with a full load. The available RR at the target cell affects the number of dropped handovers at the beginning of the handover procedure. For example, a target cell rejects the handover request because the number of available RRs is fewer than the necessary RRs for the active traffic of a subscriber. If a handover request is dropped at a target cell, the number of wrong cells (handover failure) increases, and the handover procedure restarts anew. The handover performance of an MA is the ratio of unsuccessful to successful handovers. This determines the behavior and the performance history of the MA based on the attempted handovers that were directed to the MA. This factor contributes to a suitable MA selection. For example, we will increase the weight of the handover performance because it is more important than others are when all MA loads are light or ample for all of the connected subscribers. Our CMAS is proposed to calculate the cost of candidate MAs based on cost function f(CMAS), which includes three normalized functions: (i) f(rr) is a function of the available RRs of a target cell; (ii) f(l) is a function of MA load; and (iii) f(hp) is a function of the handover performance of an MA. Weights (w)s are used by the normalized functions to define the priority of the factors. Here, proposed function f(CMAS) can be simplified by the following equation: fn=wlf(l) + wrrf(rr) + whpf(hp)if ls≥lthr f(rr)if ls< lthr(1) where wl,wrr, and whp represent the weights of functions f(l), f(rr), and f(hp), respectively. The sum of all weights is equal to one (wl+wrr +whp = 1); lsand lthr are the current loads of the serving MA and the threshold for MA selection, respectively; lthr is a key parameter for handover-type selection: intra and inter handovers. The normalized functions are explained in detail in the following three subsections. 3.1. Function based on load The function f(l) represents the differences between a target MA’s availability and the requirement of the subscriber’s sessions. A target MA’s availability is defined as the ratio of the sum of the subscribers’ traffic flows to the capacity of the MA. Here, the capacity of the MA is defined by the number of supported traffic flows and subscribers. From the viewpoint of hardware, the capacity is the maximum size of Novel anchor-selection scheme for distributed mobility management 149 the data receive/send sessions over a network adapter. Therefore, a function of MA i load can be simplified by the following equation: fi(l) = Pm n=1(Pt u=1 fu)−Pt hft h Ci (2) where Ciis the capacity of MA i,mthe number of current subscribers in MA i,ft u the total flow of subscriber u, and his an incoming subscriber. Note that an MA with a high load is not suitable for a subscriber. A light-load MA is more suitable because it can provide high-quality services for a subscriber’s active sessions after a successful handover. 3.2. Function based on available radio resources of target cell The function f(rr) represents the differences between the ratios of the available RRs before and after a handover. This factor defines a target cell that is better than the others. In a baseline handover of LTE [7], the serving cell makes an NCL based on the signal-to-noise-plus-interference ratio (SNIR) of the candidate cells that are listed in a subscriber’s measurement report. However, our proposal uses the RR calculation instead of SNIR-based sorting, thus reducing the handover-dropping probability. When a handover begins, a serving cell exchanges the context information with its neighbors and re-sorts the NCL based on the available RRs. Based on the context information, f(rr) determines whether target cell jcan provide the required RRs for a subscriber’s sessions according to the following equation: fj(rr) = RRh RRmax −RRj RRmax =RRh−RRj RRmax (3) where RRh,RRj, and RRmax are the available RRs after and before a handover and the maximum number of RRs of cell j, respectively. 3.3. Function based on handover performance of anchor The function f(hp) represents the ratio of successful to unsuccessful handovers that are destined to an MA. Now, we define the handover performance of MA ias follows: fi(hp) = HFi+P Pi+RLFi HOsucc (4) where HOsucc is the number of successful handovers. HFi,P Pi, and RLFiare the handover failures (including handover failures at the execution and completion steps), ping-pong handovers, and radio-link failures (too-late handovers), respectively. All of these parameters (defined in [10,11]) are the results of the handover procedures that are executed from any MA to MA i. 150 Battulga Davaasambuu et al. 4. Anchor and cell selection In DMM, one of the most significant issues is the mechanism that is needed to select a suitable candidate from the distributed MAs to provide improved handover performance and good QoS for subscriber sessions. In this section, we present the proposed MA and cell-selection strategy. Depending on the type of handover, our proposed algorithm selects the target cell or both an MA and a cell. For example, if the current MA’s load is high, our algorithm selects both an MA and a cell. On the other hand, the target cell is selected. In the following subsections, we introduce our proposed algorithm and two versions of the handover procedure. 4.1. Proposed mobility anchor selection In our CMAS, the MA selection depends on the cost of the combination, which is defined as the possible collaboration of MAs and cells. If the serving MA’s load is above a certain threshold, the decision is made according to the type of handover procedure shown in Algorithm 1. Hence, the algorithm calculates the cost of all possible combinations if an inter handover is necessary. Then, the combination with the lowest cost is selected. The algorithm then sends the handover requests to the selected MA and cell. Algorithm 1 Proposed CMAS Input : M⇐List of MAs and S⇐Cells in NCL Output : Selected a combination fij or only target cell j 1: if ls≥lthr then ▷If inter handover is necessary 2: for MAi ∈Mdo 3: for j∈Sdo ▷Calculate a cost of combinations of MA iand all cells in NCL 4: fij =wlfi(l) + wrrfj(rr) + whpfi(hp) 5: end for 6: end for 7: MAselection =min {fij } 8: return : MAselection 9: else if ls< lthr then 10: T ime ⇐an average value of cell residence time 11: for j∈Sdo ▷If intra handover is necessary 12: fj(rr) = RRh−RRj RRmax ▷RR’s availability of cell j 13: end for 14: Tcell =max {SNIR ×fj(rr)×T ime} 15: return : Tcell 16: end if Novel anchor-selection scheme for distributed mobility management 151 4.2. Proposed cell selection Our Algorithm 1 selects an only cell based on factors that are related to the neighbor cells, while the current MA can continue a subscriber’s connection with high QoS. We assume that the cell selection knows two facts regarding the cells in the NCL: the available RRs, and the SNIR. Thus, the output of the cell selection can be expressed as follows: Tcell =max {SNIR ×f(rr)}(5) . Then, a handover request is sent to the target cell. Note that SNIR is an additional factor only if a cell selection is needed. 4.3. Handover procedures Now, we describe handover preparation with the proposed CMAS. We defined two versions of the handover procedure with an additional option to control and decide about handovers for subscribers. It is possible to move the handover decision from a serving cell to the MA. Hence, the handover procedure using the different methods and the flowchart of handover messages related to the MA or serving cell. We define our proposed handover procedure, which adds and modifies messages on the baseline handover of LTE networks. 4.3.1. MA and cell selection developed at cells We introduce the cell and MA selections that are placed on the serving cell in Figure 1. The neighbor cells exchange the messages about the RR’s information (including the RR availability of the cell). Also, the messages of context information exchanged between MAs include the load and other information. The serving cell controls the radio signal strength of the subscribers as a baseline handover. If the Event A3 condition is satisfied, the serving cell exchanges the context information with the serving MA (subscriber anchored) and begins the MA and cellselection step. Then, the serving cell sends a handover request to the target cell. Also, it sends a handover request to the target MA via the serving MA if inter handover is necessary (see the solid blue box in Figure 1). Finally, the serving cell sends a handover command to the subscriber with the target’s information for establishing the connection. 4.3.2. MA and cell selection developed at gateways When the Event A3 condition is satisfied, the serving cell sends a handover necessary message (which includes the availability of the neighbor cells). Based on the result of Algorithm 1, the serving MA sends a handover request to a target cell if an intra handover is necessary. However, the serving MA sends a handover request to a target MA only if an inter handover is necessary (see the solid blue box in Figure 2). Finally, 158 Battulga Davaasambuu et al. 5.2.3. Handover failures Figures 8 show the ratio of handover failures for a number of attempted handovers when the MA selection is placed at the gateways. When the number of attempted handovers increased, our proposed CMAS demonstrated fewer handover failures than the other three cases did. 250 1,000 1,750 2,500 3,250 4,000 4,750 Attempted Handovers 0.175 0.200 0.225 0.250 0.275 0.300 0.325 0.350 0.375 Ratio of Handover Failures NS LS HS CMAS Figure 8. Ratio of handover failures for number of attempted handovers This reduction can be explained by the result of two normalized functions: (i) the function based on MA load affected the cost of the MA as follows: an MA with an optimal load and good handover performance is selected with a high probability; and (ii) the function of handover performance must be recalculated after a handover procedure has finished (using the next handover procedure). Then, the proposed CMAS scheme selects a different MA and cell when subscribers simultaneously begin a handover procedure as well as in the case where the candidates listed in the NCL are the same. 5.2.4. Analysis of DMM solutions In this subsection, we compared our solution with other related solutions for DMM that are PMIPv6-based DMM solution (PMIPv6-DMM) [6, 33], and SDN- -based DMM solution (SDN-DMM) [19]. In PMIPv6-DMM, the LMA/MAG selection scheme can provide a load balance between anchors but cannot select an optimal path. Also, the controller selects a path with low-latency in SDN-DMM. However, the load-balancing function works after the handover procedures. Novel anchor-selection scheme for distributed mobility management 159 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 5,000 Attempted Handovers 1 2 3 4 5 6 Handover Request Dropping ( %) PMIPv6-DMM SDN-DMM CMAS Figure 9. Ratio of handover request dropping versus number of attempted handovers Figure 9 depicts the effect that increasing the attempted handovers has on the request dropping rate. The results show that the percentage of the handover request dropping of SDM-DMM is significantly increased. PMIPv6-DMM and CMAS show a low percentage of handover request dropping when the number of attempted handovers is low. When the number of handovers increases, CMAS keeps a lower percentage of handover request dropping than PMIPv6-DMM. This is due to a load of an MA that is defined as a factor in the handover decision phase of CMAS. Also, the result of PMIPv6-DMM can be explained by the load balancing function and LMA/MAG selection function. In addition, SDN-DMM cannot provide load balancing in the handover procedures. 5.3. Impact of weight configuration In this subsection, we study the impact of weight configuration. Figure 10 shows the percentage of handover request dropping for the number of attempted handovers. We reconfigured the weights as shown in Table 2. Table 2 Weight configuration Cases Values 1wl= 0.2, wrr = 0.2 and whp = 0.6 2wl= 0.4, wrr = 0.2 and whp = 0.4 3wl= 0.6, wrr = 0.2 and whp = 0.2 160 Battulga Davaasambuu et al. 250 1,000 1,750 2,500 3,250 4,000 4,750 Attempted Handovers 2 3 4 5 6 7 8 Handover Request Dropping (%) case 1 case 2 case 3 Figure 10. Handover request dropping versus number of attempted handovers From Figure 10, the handover signaling cost of the three configurations of weights is almost identical (at about 3%) when the number of attempted handovers is between 250 and 2,500. However, when the number of attempted handovers is greater than 3,000, the handover request dropping increases significantly. Then, when the number of attempted handovers is greater than 3,000, the handover request dropping of Case 2 is around 4%. In addition, the handover request dropping of Cases 1 and 3 is increased by up to 7%. 6. Conclusions In this paper, we propose and analyze a cost function-based MA selection for DMM architecture. The proposed CMAS can select a suitable MA and cell in the handoverpreparation step. First, we compared our proposed scheme with three simple solutions – distance-based, load-based, and hybrid solutions. Simulation results such as signaling cost, handover request dropping, and handover failures show the feasibility of CMAS. The proposed CMAS was analyzed and shown to have better handover performance than that of the other three solutions. In addition, our proposal reduces the ratio of handover request dropping, as CMAS gives the subscriber a better chance at succeeding with the first attempted handover. Second, we compared major solutions such as PMIPv6-based DMM, SDN-based DMM, and CMAS. Our CMAS maintains lower handover request dropping when the number of attempted handovers increases. Novel anchor-selection scheme for distributed mobility management 161 Acknowledgements The research presented in this paper was partially supported by Project P2016-1249 supported by the Asia Research Center, Mongolia and Korea Foundation for Advanced Studies, Korea. References [1] 23.799 G.T.: Study on Architecture for Next Generation System, Technical report, Internet Engineering Task Force (IETF), 2016. [2] 36.300 G.T.: Technical Specification Group Radio Access Network Evolved Universal Terrestrial Radio Access (E-UTRA) and Evolved Universal Terrestrial Radio Access Network (E-UTRAN), Overall description; (Release 13), 2015. [3] 38.913 G.T.: Technical Specification Group Radio Access Network; Study on Scenarios and Requirements for Next Generation Access Technologies, 2017. [4] Ali-Ahmad H., Moses D., Moustafa H., Seite P., Condexia T.: Mobility anchor selection in dmm: Use-case scenarios, IETF-Draft (work-in-progress), 2013. [5] Baldo N.: The ns-3 LTE module by the LENA project, Center Tecnologic de Telecomunicacions de Catalunya, 2011. [6] Bernardos C.J., de la Oliva A., Giust F.: A PMIPv6-based solution for Distributed Mobility Management, 2013. [7] Bradai A., Benslimane A., Singh K.D.: Dynamic anchor points selection for mobility management in Software Defined Networks, Journal of Network and Computer Applications, vol. 57, pp. 1–11, 2015. [8] Carmona-Murillo J., Friderikos V., Gonz´alez-S´anchez J.: A hybrid DMM solution and trade-off analysis for future wireless networks, Computer Networks, vol. 133, pp. 17–32, 2018. [9] Chan H., Liu D., Seite P., Yokota H., Korhonen J.: Requirements for distributed mobility management, Technical report, Internet Engineering Task Force (IETF), 2014. [10] Chen Z., Bj¨ornson E.: Channel Hardening and Favorable Propagation in Cell-Free Massive MIMO with Stochastic Geometry, IEEE Transactions on Communications, vol. 66(11), pp. 5205–5219, 2018. [11] Cisco Visual Networking Index: Forecast and Trends, 2017–2022 White paper, Cisco Systems, pp. 1–7, 2019. [12] Cisco Visual Networking Index: Global Mobile Data Traffic Forecast Update, 2017–2022 White paper, Cisco Systems, pp. 1–7, 2019. [13] Cominardi L., Giust F., Bernardos C.J., De La Oliva A.: Distributed mobility management solutions for next mobile network architectures, Computer Networks, vol. 121, pp. 124–136, 2017. [14] Condeixa T., Sargento S.: Context-aware adaptive IP mobility anchoring, Computer Networks, vol. 71, pp. 84–99, 2014. 162 Battulga Davaasambuu et al. [15] Davaasambuu B.: Study on Mobility Management Scheme of Moving Relay Nodes in LTE Network, Ph.D. thesis, Waseda University, 2016. [16] Davaasambuu B., Sato T.: A Cost Based Handoff Hysteresis Scheme in Wireless Mobile Relay Node. In: 2014 IEEE 80th Vehicular Technology Conference (VTC2014-Fall), pp. 1–5, 2014. doi: 10.1109/VTCFall.2014.6965808. [17] Figueiredo S., Jeon S., Gomes D., Aguiar R.L.: D3M: Multicast listener mobility support mechanisms over distributed mobility anchoring architectures, Journal of Network and Computer Applications, vol. 53, pp. 24–38, 2015. [18] Huang C.M., Dao D.T., Chiang M.S.: A Bursty Multi-node Handover scheme for mobile internet using the partially Distributed Mobility Management (BMH– DMM) architecture, Telecommunication Systems, vol. 69, pp. 113–130, 2018. [19] Huang C.M., Dao D.T., Chiang M.S.: SDN-FHOR-DMM: a software defined network (SDN)-based fast handover with the optimal routing control method for distributed mobility management (DMM), Telecommunication Systems, vol. 72, pp. 1–21, 2019. [20] Interdonato G., Ngo H.Q., Frenger P., Larsson E.G.: Downlink Training in Cell-Free Massive MIMO: A Blessing in Disguise, arXiv preprint arXiv:190310046, 2019. [21] Jeon S., Aguiar R.L., Kang N.: Load-balancing proxy mobile IPv6 Networks with Mobility Session Redirection, IEEE Communications Letters, vol. 17(4), pp. 808–811, 2013. [22] Kim J., Kim D., Choi S.: 3GPP SA2 architecture and functions for 5G mobile communication system, ICT Express, vol. 3(1), pp. 1–8, 2017. [23] Ko H., Jang I., Lee J., Pack S., Lee G.: SDN-based distributed mobility management for 5G. In: Consumer Electronics (ICCE), 2017 IEEE International Conference on, pp. 116–117, IEEE, 2017. [24] Lee D.W., Gil G.T., Kim D.H.: A Cost-Based Adaptive Handover Hysteresis Scheme to Minimize the Handover Failure Rate in 3GPP LTE System, EURASIP Journal on Wireless Communications and Networking, vol. 2010(1), p. 750173, 2010. [25] Liebsch M., Seite P., Karagiannis G.: Distributed mobility management: Framework and analysis, Internet Engineering Task Force (IETF), 2013. [26] Nguyen T.T., Bonnet C.: A hybrid centralized-distributed mobility management architecture for Network Mobility. In: 2015 IEEE 16th International Symposium on A World of Wireless, Mobile and Multimedia Networks (WoWMoM), pp. 1–9, IEEE, 2015. [27] Nguyen T.T., Bonnet C., Harri J.: SDN-based distributed mobility management for 5G networks. In: Wireless Communications and Networking Conference (WCNC), 2016 IEEE, pp. 1–7, IEEE, 2016. [28] Nguyen T.T., Bonnet C.: A hybrid centralized-distributed mobility management for supporting highly mobile users. In: IEEE-ICC, pp. 3945–3951, 2015. Novel anchor-selection scheme for distributed mobility management 163 [29] Ordonez-Lucena J., Ameigeiras P., Lopez D., Ramos-Munoz J.J., Lorca J., Folgueira J.: Network slicing for 5G with SDN/NFV: Concepts, architectures, and challenges, IEEE Communications Magazine, vol. 55(5), pp. 80–87, 2017. [30] Roy R.R.: Handbook of Mobile Ad Hoc Networks for Mobility Models, vol. 170, Springer, 2011. [31] Sanchez M.I., De la Oliva A., Mancuso V.: Experimental evaluation of an SDN-based distributed mobility management solution. In: Proceedings of the Workshop on Mobility in the Evolving Internet Architecture, pp. 31–36, ACM, 2016. [32] Seite P., Bertin P., Lee J.H.: Distributed Mobility Anchoring, InternetDraft draft-seite-dmm-dma-07, Internet Engineering Task Force, 2014. https: //datatracker.ietf.org/doc/html/draft-seite-dmm-dma-07. Work in Progress. [33] Sun K., Kim Y.: Flow Mobility Management in PMIPv6-based DMM (Distributed Mobility Management) Networks, Journal of Wireless Mobobile Networks Ubiquitous Computing and Dependable Applications, vol. 5(4), pp. 120–127, 2014. [34] Valtulina L., Karimzadeh M., Karagiannis G., Heijenk G., Pras A.: Performance evaluation of a SDN/OpenFlow-based Distributed Mobility Management (DMM) approach in virtualized LTE systems. In: Globecom Workshops (GC Wkshps), 2014, pp. 18–23, IEEE, 2014. [35] Yang H., Kim Y.: Routing Optimization with SDN, Internet-Draft draft-yangdmm-sdn-dmm-01, Internet Engineering Task Force, 2014. [36] Yousaf F.Z., Bredel M., Schaller S., Schneider F.: NFV and SDN-Key Technology Enablers for 5G Networks, IEEE Journal on Selected Areas in Communications, vol. 35(11), pp. 2468–2478, 2017. doi: 10.1109/JSAC.2017.2760418. Affiliations Battulga Davaasambuu National University of Mongolia, Department of Electronics and Communication Engineering, School of Engineering and Applied Sciences, [email protected] Tumnee Telmuun National University of Mongolia, Department of Electronics and Communication Engineering, School of Engineering and Applied Sciences, telm[email protected]um.edu.mn Dominik Sasko Brno University of Technology, sasko[email protected] Yu Keping Waseda University, Global Information and Telecommunication Institute, yu.k[email protected] Shirmen Sodbileg National University of Mongolia, Department of Electronics and Communication Engineering, School of Engineering and Applied Sciences, [email protected] Received: 05.03.2020 Revised: 29.11.2020 Accepted: 30.12.2020