Efficient Location Traffic Management With Multiple Virtual Layers*

Transcription

1 JOURNAL OF INFORMATION SCIENCE AND ENGINEERING 9, (2003) Efficient Location Traffic Management With Multiple Virtual Layers* DAEWOO CHUNG, HYUNSEUNG CHOO, HEE YONG YOUN AND SEONG-MOO YOO + School of Information and Communication Engineering Sungkyunkwan University Suwon , Korea {chung, choo, youn}@ece.skku.ac.kr + Electrical and Computer Engineering Department The University of Alabama in Huntsville Huntsville, AL 35899, U.S.A. In mobile wireless networks, efficient location management for tracking and finding the mobile users is a critical issue. The traffic for location update can be excessive, especially at base stations that are near location area (LA) boundaries. In this paper we propose a new location update scheme which can significantly reduce the signaling traffic for location update. It is based on the virtual layer approach employing SubMSCs. Here the virtual layer is laid upon the original layer of LA s such that the mobile terminals moving around the boundary cells of adjacent LA s begin moving within a virtual LA. As a result, the proposed scheme significantly reduces the location update traffic compared to an overlapping scheme which is the most recent and efficient location update scheme. Keywords: location area, mobile terminals, signaling traffic, wireless network, SS7. INTRODUCTION Mobile communication has grown substantially over the last few years. Here users accustomed to the service available from wired networks tend to expect to receive the same quality of services from mobile wireless networks. One of main issues in mobile wireless networks is how to deal with moving terminals. As the movement implies a change of access point, the wireless network must be able to determine the location of moving terminals in order to set up a connection and route for incoming messages. Location management is concerned with tracking and finding the mobile terminal. Many schemes have been proposed in the literature [-3, 7, 9, 20] to keep track of the location information of mobile terminals. There exist two standards, IS-4 and GSM, for location management. Location management that keeps track of the positions of mobile terminals plays an important role in wide area roaming in personal communication systems (PCS) environment. Here two major tasks, location update and paging, are involved [8, 2-25]. A terminal performs location update whenever it enters a new location area (LA) by trans- Received September 22, 200; revised April 0 & October, 2002; accepted December 8, Communicated by Yi-Bing Lin. * Short version of this work was published in International Conference on Parallel Processing 200, Valencia, Spain. This work was supported in part by Brain Korea

2 788 DAEWOO CHUNG, HYUNSEUNG CHOO, HEE YONG YOUN AND SEONG-MOO YOO mitting its current location information to the network. This procedure is called location update. Here LA is a service area where several clustered cells are managed by a mobile switching center (MSC). When a call arrives, the MSC locates the designated terminal by sending a page message to all the cells within its territory. When a terminal responds to the page message, the network sets up a connection to that terminal. This procedure is called paging. The main problem in location update and paging is that the traffic for location update and paging can be excessive, especially at base stations that are near the LA boundaries [2-30]. We focus here on signaling for location update since its traffic is much heavier than the traffic for paging setup in the boundary cells of LA s. To cope with the problem of excessive traffic due to location update, a number of effective location update schemes have been published. For example, update upon entering another cell [4, 5]/new group []/a reporting cell, dynamic update scheme [7] based on distance/movement/time, forward pointer strategy [3, 4] and overlapping scheme [8, 9] have been proposed. Among the schemes, the recently proposed overlapping scheme is designed to handle the critical but realistic situation that mobile terminals move back-and-forth repeatedly between two adjacent LA s, which causes frequent location updates. With the overlapping scheme, the location update signaling traffic can be reduced compared to earlier non-overlapping schemes. In this paper we propose a new location update scheme which further reduces the signaling traffic for location update. It is based on the virtual layer approach employing SubMSCs. Here the virtual layer consists of virtual LA s, each of which is managed by SubMSCs. The layer of virtual LA s is laid upon the original layer of LA s such that each original LA is covered by three virtual LA s. As a result, the mobile terminals moving around the boundary cells of adjacent LA s move within a virtual LA. This can greatly save the network resources by eliminating unnecessary location updates. Moreover, the signaling traffic concentrated to some limited number of cells in earlier designs is distributed among several cells in the proposed design. The proposed scheme allows significant reduction of location update traffic compared to the overlapping scheme [8], which is the most recent and efficient scheme. The rest of the paper is organized as follows. In section 2, we describe the network architecture of PCS and existing location management techniques. We present the proposed scheme in section 3. In section 4, the proposed scheme is evaluated and compared with the overlapping scheme in terms of average location update rate per user. The conclusion and suggestions for future research are given in the last section. 2. BACKGROUND AND PREVIOUS WORKS In this section the PCS is first briefly described. Then the issues related to location update are discussed. In this paper we assume that PCS networks consist of hexagonal cells. 2. PCS Network Typically a terminal in a wired network such as a telephone network has a fixed location. Therefore, changing the location of a terminal generally involves network ad-

3 EFFICIENT LOCATION UPDATE USING VIRTUAL LAYER 789 ministration. Here the incoming calls for a particular terminal are always routed to a fixed location. On the contrary, PCS networks provide wireless communication services to mobile subscribers so that PCS users carrying mobile terminals can communicate with remote terminals regardless of their current location and mobility pattern. The basic design of a PCS network consists of a wired backbone network and wireless mobile units. Current PCS networks adopt a cellular architecture as shown in Fig.. Here the entire service area is covered with cells, and several cells are grouped into an LA. A cell is serviced by a base station (BS), and several BSs are wired to a base station controller (BSC) which is connected to a mobile switching center (MSC). An MSC provides typical switching functionality, coordinated location registration, and call delivery. It is connected to the backbone wired network such as public switching telephone network (PSTN) and signaling network such as SS7 [0, ]. Fig.. The cellular architecture in PCS networks. In the fixed environment of common telephone networks, traffic is routed from a source to a destination having a static address. However, in a mobile environment, the endpoint of a connection is unknown to the source. To trace the location of mobile terminals, the network is equipped with location registers that are accessed by relevant network entities. A mobile communication network holds two types of registers. The visit location register (VLR) temporarily stores the service profile and location information of mobile terminals roaming in its area. It is associated with an MSC, which is geographically adjacent to it. The home location register (HLR) permanently stores information on the mobile terminals currently roaming. In the entire network, only one HLR can exist. Current location management schemes employing the two-type registers are mostly based on a two-level data hierarchy. The location registers are updated for tracking the mobile users when they change the LA s. Fig. 2 shows the two operations Groupfind and Groupupdate involved in location update. The movement of every mobile unit is recorded in the VLR of the corresponding LA as well as the HLR. For example, IS-4

4 790 DAEWOO CHUNG, HYUNSEUNG CHOO, HEE YONG YOUN AND SEONG-MOO YOO (AMPS cellular phone system) is a well-defined tool for location update and mobile tracking in wireless system. GroupFind( ) { Call to PCS user is detected at local switch; if the called party is the same LA then return; switch queries called party HLR; HLR queries called party current VLR, V; VLR V returns called party location to HLR; HLR returns location to the calling switch; If (the called party is found) return;else search the remaining N- cells for the called party; } GroupUpdate( ) { The mobile terminal detects that it is in a new Group; The mobile terminal sends a registration message to the new VLR; The new VLR sends a registration message to the user HLR; The HLR sends a registration cancellation message to the old VLR; The old VLR sends a cancellation confirmation message to the HLR; The HLR sends a registration confirmation message to the new VLR; } Fig. 2. The algorithms involved in location update. 2.2 Location Update In cellular systems handling a large number of subscribers, the traffic required for location update needs to be minimized. In order to solve this problem, several location update schemes have been developed. They are update upon entering another cell or reporting cell, dynamic update based on distance, movement, or time, forward pointer strategy, and overlapping scheme. However, since the location update is LA-based, the traffic due to location update is very high at the boundary cells of each LA. With the no update strategy, the location of mobile users is never updated. To detect a specific mobile user, the system must first search the entire network. This strategy involves long call setup delay. The update upon entering another cell strategy carries out the update task when a mobile user crosses a cell boundary which reduces call setup delay. However, the update cost is high because the update rate for such mobile users is pretty high. The update upon entering a reporting cell strategy designates some cells as reporting cells where the mobile users must update the location data upon entering them. Upon arrival of a call, the mobile user is paged in the vicinity of the reporting cell with which it has most recently updated the location. Choosing an optimal set of reporting cells for a general cellular network has been shown to be NP-complete. Three strategies, distance-based update, movement-based update, and time-based update, are named

5 EFFICIENT LOCATION UPDATE USING VIRTUAL LAYER 79 by the kind of threshold used to initiate an update; they are also called as dynamic update scheme. Under these schemes, a mobile user updates the location based only on its local activity in the sense that it makes the decision whether to update or not without gathering any global or design specific information about the cell. Under distance-based update strategy, the mobile unit is required to track the Euclidean distance from the location of the previous update and initiate a new update if the move distance passes a threshold, D. The distance can be specified in terms of the number of cells between the two positions. The movement-based update strategy is essentially a way of over-estimating the Euclidean distance by the traversed distance when the distance is measured by the number of cells. Under time-based update strategy, the mobile unit sends periodic updates to the system. The period, T, can easily be programmed into the mobile unit using a hardware or software timer. However, the disadvantage of these strategies is that a network-wide searching is still needed. With the forward pointer strategy, a call to a user first queries the user s HLR to determine the VLR to which the user first registered, and then follows a chain of forwarding pointers to the user s current VLR. This strategy is useful for those users who receive calls less frequently than the rate at which they change the registration area. Although this system is low-cost and does not involve searching for a mobile user network wide, redundant setting of forward pointers still occurs. In addition, the longer the pointer chain, the longer the call setup delay. The group method divides all the cells into groups, and as a result reduces the update rate to ( 4 N ) / 3 N ) from N for the basic method where N is the number of cells []. The group method performs the update task only when a mobile user enters another group. The system has to search for the mobile user only within the group when a mobile user is called. The group method is thus simpler and more effective than the other schemes mentioned above because it does not need to arrange reporting cells while restricting the search scope to a group. The overlapping scheme [8, 9] prevents a mobile user moving along the border of two LA s from causing increased location update traffic due to short term switching. Overlapping LA s can reduce the traffic as shown in Fig. 3. Observe that the two LA s do not overlap in Fig. 3(a), while they overlap in Figs. 3(b) and (c). Here, w indicates the degree of overlapping which is actually the number of rows of overlapped cells. Without overlapping as in Fig. 3(a), every time a mobile user crosses the LA boundary, the location needs to be updated. If the adjacent LA s overlap as in Fig. 3(b), only the users crossing the overlapping region cause location update. In other words, a user needs to fully cross a cell to cause location update. Note that it just needs to cross the boundary line to cause location update when the LA s do not overlap but abut each other. When the overlap is more significant as in Fig. 3(c), the users need to cross several cells (here it is 2) to cause location update. This scheme thus significantly reduces the signaling traffic due to location update compared to the non-overlapping scheme. The shortcomings of the overlapping approach are, however, that the cells in an LA do not overlap uniformly. As a result, managing the location update is complex. Also, since the LA s overlap, more MSCs (and thus VLRs) are required than with the nonoverlapping scheme. We next present the proposed scheme which can effectively reduce the location update rate without such overhead.

6 792 DAEWOO CHUNG, HYUNSEUNG CHOO, HEE YONG YOUN AND SEONG-MOO YOO Fig. 3. The overlapping scheme. 3. THE PPOPOSED SCHEME In this section the proposed scheme is presented. First, the basic structure based on a virtual layer concept is introduced. Then the detail operation is explained. 3. The Basic Structure In the future microcellular structure is expected to be used for PCS networks in order to support high user densities. In such an environment the signaling traffic due to location update will be very high since mobile users can easily cross the LA cell boundaries. In mobile networks, the location of a user is identified by the LA in which it resides. The base stations continuously broadcast the identity of the LA they belong to. When a mobile terminal detects a change in the LA, it sends a location update message to the network. Location update messages generated by many mobile users result in a considerable amount of the signaling traffic. As shown in Fig. 3, LA s with overlapping regions could reduce the update signaling traffic. Using the virtual layer design proposed in this paper, however, the same objective can be accomplished more effectively. One of the important facts motivating the proposed design is that the cost of location update for HLR is much higher than that for VLR. This is because the HLR for a user is usually located somewhere far from the user while the VLR is close to the user. Therefore, the traffic required for updating HLR needs to be minimized, and the principle employed in the proposed scheme is to distribute the HLR bound signaling traffic to VLRs. The enhanced location management scheme proposed in this paper employs a virtual layer as shown in Fig. 4. Observe that the entire area is partitioned into seven LA s (LA_2 to LA_8) which are marked by bold lines. As mentioned, each LA has an associated MSC and VLR. This original layer of LA s is called Layer-. Also notice that there exists another partition of the area using triplicated lines, which is a virtual layer which we call Layer-2. Each LA of Layer-2 also has a SubMSC. Note that the Layer-2 partition was decided such that an LA of Layer- consists of part of three LA s of Layer-2 of almost equal size. In what follows, we denote LA_i, j the LA i of Layer-j. For example, LA_5, consists of part of LA_2,2, LA_4,2, and LA_5,2. The MSC of an LA of Layer- is connected to three SubMSCs representing the LA s of Layer-2, which partition that LA. For example, MSC_5 is connected to SubMSC_2, 4, and 5. The VLRs are connected to a HLR. The proposed structure effectively prevents the oscillation effect from occurring when a mobile user travels along the boundary of two adjacent LA s and distribute

7 EFFICIENT LOCATION UPDATE USING VIRTUAL LAYER 793 Fig. 4. The proposed virtual layer architecture. the location update signaling traffic over many cells using the virtual layer design as explained next. The proposed architecture is more understandable if Fig. 8 is referred. Observe that an LA in Layer- overlaps with three LA s in Layer-2. At the same time, an LA in Layer-2 overlapped with three LA s in Layer Operation Here we employ the same environment as in all previous schemes for PCS networks. Each terminal monitors the broadcast message from the base station. If the current LA is different from the LA registered, then the mobile terminal initiates location update to inform the system about its new LA. When an incoming call arrives for the mobile terminal, the system performs a paging operation to locate it. The proposed scheme can be implemented by assigning a unique ID [2, 3] to each LA of Layer- and 2. Note that the proposed scheme covers the service area with homogeneous LA s. The original LA s completely overlap the LA s of the virtual layer, and each cell is covered by exactly two different LA s - one of Layer- and one of Layer-2. Even though each cell belongs to two LA s, the mobile user in a cell registers with only one LA at any moment.

8 794 DAEWOO CHUNG, HYUNSEUNG CHOO, HEE YONG YOUN AND SEONG-MOO YOO The selection is made according to the distance from the cell to the center cells of the two LA s. Among the two, the LA whose distance is smaller is selected. When the distances are same, a random selection is made. For example, a mobile user in Cell-A belongs to both LA_5, and LA_2,2, but it registers to LA_2,2 since it is closer to the center cell of LA_2,2 than that of LA_5,. Similarly, the user in Cell-B registers to LA_5,. Location update occurs when a user leaves the LA it is currently registered with, and it always registers with the LA in the different layer from the previous one. Refer to Fig. 4. Assume that a user residing in Cell-B (who is registered with LA_5, belonging to Layer-) moves to Cell-D through Cell-C, which belongs to both LA_4, and LA_4,2. It does not register with LA_4, but with LA_4,2 belonging to Layer-2. The reason why this approach is taken is to avoid continuous location update due to the users moving around the boundary cells. Assume again that the user arriving in Cell-D registers with LA_4, and then soon comes back to Cell-C. It will then require another location update since Cell-C belongs to LA_5,. Whenever the user moves back and forth between Cell-C and D, location update is necessary. Meanwhile, if it registered with LA_4,2 as suggested in the proposed scheme, no location update is necessary since both Cell-C and D belong to LA_4,2. As we see from this example, the proposed scheme greatly reduces the frequency of location update compared to other schemes including the overlapping scheme. The functionality of a SubMSC includes switching mobile terminals in the LA of Layer-2. The SubMSCs are simple switches and do not require a connection to each base station. Each SubMSC is connected to three neighboring MSCs. As an example, SubMSC_4 in Fig. 4 is connected to MSC 4, 5, and 7. MSC and SubMSC manage the traffic in the LA s of Layer- and Layer-2, respectively. VLR communicates with the MSC connected to three SubMSCs. As long as a mobile user moves within three adjacent LA s managed by a SubMSC, no location update occurs except in a few cells. Therefore, the proposed location update with SubMSCs can significantly reduce the traffic to HLR. This is verified by performance evaluation in section 4. Also, the signaling traffic concentrated on the boundary cells of LA s as in the overlapping scheme, can be distributed to several cells, which is another important benefit of the proposed scheme. Fig. 5 shows an example of the travel path of a mobile terminal. At first the mobile terminal is located at A, and thus registers with the VLR of MSC managing LA_5, and HLR. When it moves to B, it registers with LA_4,2 managed by a SubMSC. Here only VLR is updated since the MSC is connected to the SubMSC. While it moves from B to D through C, the system does not update either HLR or VLR since the locations are all inside LA_4,2. Upon arriving at E, the VLR is updated for the change made from SubMSC to MSC. Until it reaches G, no update is necessary. When it moves from H to I, it happens to leave the group LA_5,, LA_2,2, LA_4,2, LA_5,2 which is governed by MSC5. Therefore, both VLR and HLR need to be updated. Table lists the movements along with the registered LA and updated register when a mobile user moves from location A to location O. Here 9 VLRs and 3 HLRs were needed to be updated. Note that 7 VLRs and 7 HLRs need to be updated if the proposed virtual layer scheme is not employed.

9 EFFICIENT LOCATION UPDATE USING VIRTUAL LAYER 795 Fig. 5. An example of a travel path of a mobile terminal. Table. The registered LA and updated registers for the mobile user of Fig. 5. Path Registered LA Updated register A B LA_4, 2 VLR 2 B C LA_4, 2 None 3 C D LA_4, 2 None 4 D E LA_5, VLR 5 E F LA_5, None F G LA_5, None 7 G H LA_2, 2 VLR 8 H I LA_3, HLR, VLR 9 I J LA_3, 2 VLR 0 J K LA_, HLR, VLR K L LA_, None 2 L M LA_, 2 VLR 3 M N LA_9, HLR, VLR 4 N O LA_8, 2 VLR

10 79 DAEWOO CHUNG, HYUNSEUNG CHOO, HEE YONG YOUN AND SEONG-MOO YOO 4. PERFORMANCE EVALUATION 4. Preliminaries Each cell has six neighboring cells. This model is suitable for the mobility model in which mobile users can move in any azimuthal direction. An LA denotes a set of cells located within the update boundary. We employ the concept of rings discussed in [8, 4]. The size of an LA is represented by the number of rings of cells forming the LA, d, where the center cell is ring-0 and the outermost ring is ring-(d ). The average location update rate per user adopts the concept of dwell time [5, ]. When the dwell time, T d, expires in its current cell, a mobile user moves to one of the neighboring cells with a probability of /. We evaluate the location update rate for a target cell and its six neighboring cells. Assume that the movement of a mobile user is probabilistically independent and statistical equilibrium exists. We develop analytical models of the overlapping scheme and the proposed scheme to compare them. The following notation is used in the models. Notations K: Average number of mobile users in a cell. d : Number of rings of an LA. w: Amount of overlapping (w < d)(seefig.3.) T d : Average dwell time, while the dwell time is exponentially distributed. N: Total number of mobile users in an LA. N T : Average number of mobile terminals in an LA. N C : Number of cells in an LA, which is 3d 2 3d +. R LA : Average location update rate for the given LA. R MS : Average location update rate per user. u i,j : Number of mobile users in cell-i of Layer-j (j =,2). 4.2 Conventional Non-overlapping Scheme The mobile users in a cell of an LA all register with that LA. In the nonoverlapping scheme, the average location update rate is just the average rate of departures of the users from the boundary cells. This is because only the users in the boundary cells of an LA contribute to the traffic for location update. The average location update rate for an LA, R, is thus RLA = ( 2d +) K () Td Here (2d + )K represents the number of mobile users leaving an LA, and thereby causing location update. Refer to Fig. 4 which consists of 9 cell LA s (d = 2). Among the 2 boundary cells, cells have 3 edges facing other LA s while the remaining cells have 2 such edges. Therefore, the expected number of users leaving the LA is K /2 + K /3 = 5K. In general, it is (2d +)K. The total number of mobile users in an LA is on average LA

11 EFFICIENT LOCATION UPDATE USING VIRTUAL LAYER 797 N = N K ( 3 2 N = d + 3 d + ) C (2) C The average location update rate per user is then R MS R = N LA (2d + ) = (3d + 3d + ) 2 T d (3) For details, refer to [8]. 4.3 The Model for the Overlapping Scheme Compared to the conventional scheme discussed above, this scheme basically reduces the location update signaling traffic due to the reduced number of mobile users registered in the boundary cells of the LA s. Overlapping is shown in more detail in Fig.. LA overlaps are six neighboring LA s even though the overlap with only LA2 and LA3 is shown here. The mobile users in any overlapping region may register with different LA s. Each LA consists of six symmetric sectors and a center cell. In Fig. one of the sectors is identified. Each cell in the LA is associated with a unique coordinate (i, j). The number of users in cell-( i, j ) is denoted x i,. In order to find the average number of mobile users in LA, the number of mobile users in the cells of each sector needs to be j known. The entire area of a sector is divided into four regions as shown in Fig.. Fig..OverlapinanLAwithd=9andw=5. Case of nonoverlapping region: This region consists of the cells belonging only to LA and not overlapped by other LA s. Entire mobile users in this region register with LA. To calculate the number of mobile users in this case, we use the following equation. The nonoverlapping region has only one center cell as shown in Fig..

12 798 DAEWOO CHUNG, HYUNSEUNG CHOO, HEE YONG YOUN AND SEONG-MOO YOO x i,j = K (i =,2,,d w,andj =,2,,d w i) (4) Case of overlapping region: This region consists of the cells belonging to two or three LA s depending on the location of cells in the LA s. In Fig., the three-overlapping regions are represented. The mobile terminals in this region remain registered to their previous LA s upon entering the cell. The average rate of mobile terminals departing from cell-(i, j) is equal to the average rate of mobile terminals arriving at cell-(i, j). Therefore is as follows. ( x + x + x + x + x x ) (5), x i j = i, j i, j+ i, j i+, j i, j+ + i+, j x i, j where, i = d w, d w +,,d, and j =0,,,d i. Eqs. (4) and (5) above constitute a set of equations on xi, j ' s for the given sector of LA. The shaded cells in Fig. 7 should be considered for solving Eq. (5). x 0,0 must be K because cell-(0, 0) is in the non-overlapping region as discussed above. Fig. 7. The boundary condition for equations (4) and (5). x 0,0 = K (Center cell) (5-a) We evaluate just one sector for obtaining the number of mobile users in LA since the overall structure is symmetric. The portion of boundary cells such as (b) and (c) in Fig. 7 are considered similarly. x 0, j = j, 0 x i, = i x ( j = d w, d w +,..., d ) x, ( i = d w, d w +,..., d ) (5-b) (5-c)

13 EFFICIENT LOCATION UPDATE USING VIRTUAL LAYER 799 Mobile users in cell-(d) and (e) register with different LA s. Hence, even though mobile users enter LA through the cells, they do not register with LA. Mobile users in cell-(d) and (e) of Fig. 7 do not cause location update, and thus they are all 0 as follows: x d, = 0 (5-d) x, = 0 ( i =, 2,... d ) i d i (5-e) The cells requiring update are xi, d i (i =,2, d 2). When i =,2, d 2, mobile users move to one of the neighboring cells with a probability of /. However, they leave LA with a probability of 2/. When i = d, mobile users leave LA with a probability of 3/. This is because the cell-(d, 0) has three neighboring cells not belonging to LA. The average location update rate for a given LA is as follows. The average location update rate for LA is obtained in the same way as explained for Eq. (). R LA d 3 2 = xd + x d i j T d 2, 0, () T i= d d = 3 x + 2 x 2 d, 0 d i, i i= Td The total average number of mobile terminals for the given LA is d d i N = x0 0 + x i, j i= j=0, (7) Finally, the average location update rate per user is R MS R = N LA = 3 x x d, 0 0, d 2 x d i, j i= d d i i= j=0 x i, j T d (8) 4.4 The Model for the Proposed Scheme As we can see from Fig. 4, MSC 5 and its neighboring SubMSCs_2, 4, and 5 are connected to VLR5. Therefore, the location update for the four LA s, LA_5,, LA_2,2, LA_4,2, and LA_5,2 is handled by VLR5. Since an LA in the proposed scheme has three partitions that are not of equal size, each cell is assigned a unique number instead of coordinates. Cells in an LA of the proposed scheme are numbered from the top row and left column as shown in Fig. 8. Here the upper and lower numbers at each cell represent the cell numbers for Layer- and 2, respectively.

14 800 DAEWOO CHUNG, HYUNSEUNG CHOO, HEE YONG YOUN AND SEONG-MOO YOO Figure 8. Cell numbering in the proposed scheme. Fig. 8. Cell numbering in the proposed scheme. Note that u i, is the number of mobile users in cell-i of Layer-j (j =,2).Thenumber of cells in an LA of d = 2 is 9, and thus 9 equations need to be manipulated. The K j denotes the number of mobile users in a cell. In Fig. 8, adding the number of mobile users in cell-7 of LA_7, and that of cell-4 of LA_4,2 results in K. u 7, + u 4, 2 = K, u 8, + u, 2 = K, u, + u 4, 2 = K (9) We can apply the same rule for other cells; refer to Appendix I. The expected number of mobile users in cell-, u, is obtained by adding the influx from the adjacent cells u 3,, u 7,, u 9,, u 2,2, u 4,2, u 5,2. Also notice that a cell has edges, and thus the users crossing one of the edges are one sixth of the users in the cell. Finally, u is obtained as follows. u 2 u 9 are listed in Appendix II. (0) u = ( u 3. + u7. + u9. + u2,2 + u4,2 + u5,2 ) The average number of mobile terminals in an LA is 2 3d N= + 3d + 4 u () i= i

15 EFFICIENT LOCATION UPDATE USING VIRTUAL LAYER 80 α β γ Fig. 9. The group of cells identified as α, β, γ. The average location update rate for a given LA is obtained using the cells grouped as shown in Fig. 9. R LA = ( α ) + ( β ) + ( γ ) 2 3 T d (2) Where, α = u + 2 u + 2 u + u + u + 2 ) (, 3, 8, 2, 7, u9, β = u + 2 u + 2 u + 2 u + u + 2 u + 2 u + 2 u + 2 ) (, 2, 4, 7, 2, 3,, 7, u8, γ = u + u + ) (, 2, u7, The average location update rate per user is R MS = R LA (3) N 4.5 Numerical Results As done in other papers, the average number of mobile users in a cell is assumed to be 00. To consider various speeds of mobile users in a cell, we assume that the dwell times, T d, are, 2, 4, and 8 minutes. We compare the proposed scheme with the overlapping scheme in terms of average location update rate per user. Figs. 0,, and 2 show that the proposed scheme significantly outperforms the overlapping scheme for all the cases studied. Also notice that the rate decreases as the size of LA increases. This is an important fact since the size of LA in typical PCS network is expected to grow as communication techniques and equipment improve.

16 802 DAEWOO CHUNG, HYUNSEUNG CHOO, HEE YONG YOUN AND SEONG-MOO YOO 7EB>MBI>DB F LGA>KBI>KB GBILJBI "!$! $ $! 2MBI >GJ@EB B 3IFGFJBA J@EB B * FL KFCFMBI >GN FMBI >GJ@EB B Fig. 0. Average location update rate per user when d = 2. 7EB>MBI>DB F@>K F LGA>KBI>KBGBI LJBI! ' % #! ' % #!! " 2MBI >G J@EB B 3IFGFJBA J@EB B * FL KFCFMBI >GNFMBI >G J@EB B Fig.. Average location update rate per user when d = 3. In the overlapping scheme, the update rate decreases as overlapping increases. However, the number of MSCs and VLRs also increase. For example, refer to Fig. 0 where d = 2 and w = 2. In this case, the overlapping scheme needs almost 38 MSCs and VLRs (double the original number) since the overlapping is quite deep. For this condition, the proposed scheme still needs 9 MSCs, VLRs and SubMSCs. Note that SubMSC is a much simpler switch than regular MSCs, and the VLRs and corresponding connections also require some significant resources. Therefore the overhead of the proposed scheme is much smaller than that for the overlapping scheme. Moreover, the proposed scheme reduces the traffic at HLR which is very important. Another advantage is that the signaling traffic concentrated in a limited number of boundary cells in the overlapping scheme is distributed over many cells in our scheme.

17 EFFICIENT LOCATION UPDATE USING VIRTUAL LAYER 803 KEB>MBI>DB F LGA>KBI>KBGBI LJBI % #! ' % #!! " # * FL KFCFMBI >GNFMBI >GJ@EB B 2MBI >GJ@EB B 3IFGFJBA J@EB B Fig. 2. Average location update rate per user when d = CONCLUSIONS In this paper we have proposed an efficient location update scheme employing SubMSCs to reduce the update signaling traffic in cellular systems. The system has a two-layer architecture and is configured by homogenous LA s. Conceptually, the proposed scheme is a combination of grouping, overlapping, and local updating in VLR. This scheme yields a significant performance improvement over the overlapping scheme in terms of the average location update rate per user. Moreover, the new method offers considerable enhancement in utilizing the network resources which otherwise will be wasted by mobile users causing frequent update in the conventional scheme. The signaling traffic concentrated in boundary cells in the conventional scheme is also distributed over many cells. In addition to the mobile users at the boundary cells, location update needs to consider other factors such as mobility pattern, dwell time, call to movement ratio, etc. We will investigate the relationship between these factors, and include them in the model of the update rate. This will provide us with a good measure by which efficient location management policy can be derived. REFERENCES. U. Madhow, M. L. Honig, and K. Steiglitz, Optimization of wireless resources for personal communications mobility tracking, IEEE/ACM Transactions on Networking, Vol. 3, 995, pp A. Bar-Noy and I. Kessler, Tracking mobile users in wireless communication networks, IEEE Transactions on Information Theory, Vol. 39, 993, pp

18 804 DAEWOO CHUNG, HYUNSEUNG CHOO, HEE YONG YOUN AND SEONG-MOO YOO 3. Y. B. Lin and W. N. Tsai, Location tracking with distributed HLRs and pointer forwarding, IEEE Transactions on Vehicular Technology, Vol. 47, 998, pp R. Jain, et al., A forwarding strategy to reduce network impacts of PCS, IEEE INFOCOM, 995, pp R. Jain, et al., A caching strategy to reduce network impacts of PCS, IEEE Journal on Select Areas in Communications, Vol. 2, 994, pp C. M. Weng and P. W. Huang, Modified group method for mobility management, Computer Communications, Vol. 23, 2000, pp A. Bar-Noy, I. Kessler, and M. Sidi, Mobile users: To update or not to update? Wireless Networks, Vol., 995, pp D. Gu and S. S. Rappaport, Mobile user registration in cellular systems with overlapping location areas, IEEE VTC, 999, pp T. P. Chu and S. S. Rapparot, Overlapping coverage with reuse partitioning in cellular communication systems, IEEE Transactions on Vehicular Technology, 997, Vol Y. B. Lin and S. K. DeVries, PCS network signaling using SS7, IEEE Personal Communications Magazine, 995, pp K. S. Meier-Hellstem and E. Alonso, The use of SS7 and GSM to support high density personal communications, IEEE International Conference on Communications, Vol. 3, 992, pp S. Mohan and R. Jain, Two user location strategies for personal communications services, IEEE Personal Communications, Vol., 994, pp L. R. Hu and S. S. Rappaport, An adaptive location management scheme for global personal communications, IEE Proceedings on Communication, Vol. 44, 997, pp J. S. M. Ho and I. F. Akyildiz, Mobile user location update and paging under delay constraints, Wireless Networks, Vol., 995, pp P. V. Orlik and S. S. Rappaport, Traffic performance and mobility modeling of cellular communications with mixed platforms and highly variable mobility, in Proceedings of the IEEE, Vol. 8, 998, pp D. Hong and S. S. Rappaport, Traffic model and performance analysis for cellular mobile radio telephone systems with prioritized and nonprioritized handoff procedures, IEEE Transactions on Vehicular Technology, Vol. VT-35, 98, pp A. Bhattacharya and S. K. Das, LeZi-Update: An information-theoretic approach to track mobile users in PCS networks, in Proceedings of ACM/IEEE International Conference on Mobile Computing and Networking, 999, pp Y. B. Lin and W. N. Tsai, Location tracking with distributed HLR s and pointer forwarding, IEEE Transactions on Vehicular Technology, Vol. 47, 998, pp D. Plassmann, Location management strategies for mobile cellular networks of 3rd generation, IEEE Vehicular Technology Conference, 994, pp C. Rose, Minimizing the average cost of paging and registration: A timer-based method, Wireless Networks, Vol. 2, 99, pp S. J. Kim and C. Y. Lee, Modeling and analysis of the dynamic location registration and paging in microcellular systems, IEEE Transactions on Vehicular Technology, Vol. 45, 99, pp I. F. Akyildiz and J. S. M. Ho, Dynamic mobile user location update for wireless

19 EFFICIENT LOCATION UPDATE USING VIRTUAL LAYER 805 PCS networks, Wireless Networks, Vol., 995, pp I. F. Akyildiz and J. S. M. Ho, Movement-based location update and selective paging for PCS networks, IEEE/ACM Transactions on Networking, Vol. 4, 995, pp T. X. Brown and S. Mohan, Mobility management for personal communication systems, IEEE Transactions on Vehicular Technology, Vol. 4, 997, pp S. K. Das and S. K. Sen, A new location update strategy for cellular networks and its implementation using a genetic algorithm, in Proceedings of ACM/IEEE International Conference on Mobile Computing and Networking, 997, pp Y. B. Lin, Determining the user locations for personal communications networks, IEEE Transactions on Vehicular Technology, Vol. 43, 994, pp Y. B. Lin, Reducing location update cost in a PCS network, IEEE/ACM Transactions Network, Vol. 5, 997, pp M. Zaid, Personal mobility in PCS, IEEE Personal Communication Magazine, 994, pp M. Mouly and M. B. Pautet, Current evaluation of the GSM systems, IEEE Personal Communication Magazine, 995, pp S. Tabbane, ESPIT: location management methods for third-generation mobile systems, IEEE Communication Magazine, 997, pp Daewoo Chung received the B.S. degree in Computer Communication from Hongik University, Korea in 2000, and the M.S. degree in the School of Information and Communication Engineering from the Sungkyunkwan University in 2002, Mr. Chung is currently an Associate Research Engineer of the LG Electronics Inc. His research interests include call processing, mobile computing, and system performance evaluation. He has published seven papers in international journals and refereed conferences. Hyunseung Choo received the B.S. degree in mathematics from Sungkyunkwan University, Korea in 988, the M.S. degree in computer science from the University of Texas at Dallas, Richardson, TX in 990, and the Ph.D. degree in computer science from the University of Texas at Arlington (UTA), TX in 99. From 997 to 998, he was a Patent Examiner at Korean Industrial Property Office, Seoul, Korea. Dr. Choo is currently an Assistant Professor of the School of Information and Communication Engineering at Sungkyunkwan University where he has finished his undergraduate program. His research interests include switching networks, mobile computing, parallel and distributed computing, system performance evaluation, and design of algorithms. He has published over fifty papers in international journals and refereed conferences. Dr. Choo is a member of IEEE.

20 80 DAEWOO CHUNG, HYUNSEUNG CHOO, HEE YONG YOUN AND SEONG-MOO YOO Hee Yong Youn received the BS and MS degree in electrical engineering from Seoul National University, Seoul, Korea, in 977 and 979, respectively, and the PhD degree in computer engineering from the University of Massachusetts at Amherst, in 988. From 979 to 984, he was on the research staff of Gold Star Precision Central Research Laboratories, Korea. He had been Associate Professor of Department of Computer Science and Engineering, The University of Texas at Arlington until 999. He was also a faculty of School of Engineering, Information and Communications University, Taejon, Korea from 999 to He is presently Endowed Professor of School of Information and Communication Engineering, Sungkyunkan University, Suwon, Korea. His research interests include storage system, distributed computing and networking, Internet and mobile computing, and fault-tolerant computing. He has published more than 30 papers in international journals and conference proceedings, and received Outstanding Paper Award from the 988 International Conference on Distributed Computing Systems, 992 Supercomputing, and 200 Korean Society Internet Information Spring Symposium, respectively. He also served as a lecturer of the ACM Lectureship Series from 993 to 997. Dr. Youn is a senior member of the IEEE Computer Society. Seong-Moo Yoo received the B.S. degree in economics from Seoul National University, Seoul, Korea in 970, the M.S. and Ph.D. degree in computer science from the University of Texas at Arlington in 989 and 995, respectively. Since September 200, he is an associate professor in Electrical and Computer Engineering Department of the University of Alabama in Huntsville, Huntsville, Alabama, U.S.A. From September 99 to August 200, he was an assistant professor in Computer Science Department of Columbus State University in Columbus, Georgia, U.S.A. Dr. Yoo is the co-program chair of ISCA th International Conference on Parallel and Distributed Computing Systems (PDCS-2003), August 2003, Reno, Nevada, U.S.A. Dr. Yoo s research interests include wireless networks, parallel computer architecture, computer network security, and design of algorithms. Dr. Yoo is a member of IEEE Computer Society and Communication Society.

21 EFFICIENT LOCATION UPDATE USING VIRTUAL LAYER 807 u 8, + u, 2 = K, u 9, + u2, 2 = K, u, + u3, 2 = K APPENDIX I 0, u 7, + u4, 2 = K, u 3, + u5, 2 = K, u, + u, 2 = K 4, u 5, + u7, 2 = K, u, + u8, 2 = K, u, + u9, 2 = K 2, u 7, + u0, 2 = K, u 8, + u, 2 = K, u, + u2, 2 = K 9, u, + u3, 2 = K, u, + u4, 2 = K, u u = K 2, + 5, 2, 5, + u8, 2, u 3, + u, 2 = K, u 4, + u7, 2 = K, u = K u u = K, + 9, 2 u = ( u + u + u + u + u + u ) APPENDIX II ,2 4,2 5,2,2 + u3,2 + u5,2 + u,2 2,2 + u,2 + u7,2,2 + u5,2 + u8,2 + u9,2 + u8, + u3,,2 + u2,2 + u4,2 + u,2 + u9,2 + u0,2 + u7, + u2, 2,2 + u3,2 + u5,2 + u7,2 + u0,2 + u,2 3,2 + u,2 + u,2 + u2,2 4,2 + u9,2 + u3,2 4,2 + u5,2 + u8,2 + u0,2 + u3,2 + u4,2 + u, + u3, + u7, 5,2 + u,2 + u9,2 + u,2 + u4,2 + u5,2 + u, + u2, + u2,,2 + u7,2 + u0,2 + u2,2 + u5,2 + u,2 + u2, + u3, 7,2 + u,2 + u,2 + u3, + u4, + u8, 8,2 + u9,2 + u4,2 + u7,2 + u, + u4, 9,2 + u0,2 + u3,2 + u5,2 + u7,2 + u8,2 + u2, + u, + u u 2 = ( ) u 3 = ( ) u 4 = ( ) u 5 = ( ) u = ( ) u 7 = ( ) u 8 = ( ) u 9 = ( ) u 0 = ( ) u = ( ) u 2 = ( ) u 3 = ( ) u 4 = ( ) 7,

22 808 DAEWOO CHUNG, HYUNSEUNG CHOO, HEE YONG YOUN AND SEONG-MOO YOO 0,2,2 4,2,2 8,2 9,2 7,,2 + u2,2 + u5,2 + u5,2 + u9,2 + u8, + u8, + u 3,2 + u4,2 + u8,2 + u, + u9, 4,2 + u5,2 + u7,2 + u9,2 5,2 + u,2 + u8,2 u 5 = ( u + u + u + u + u + u + u + u ) u = ( ) u 7 = ( ) u 8 = ( ) u 9 = ( ) 8, 9,