Anda di halaman 1dari 5

Mobility Management Schemes at Radio Network Layer for LTE Femtocells

Lan Wang, Yongsheng Zhang, Zhenrong Wei DOCOMO Beijing Communications Laboratories Co. Ltd., Beijing, China
{wangl;zhangy;wei}@docomolabs-beijing.com.cn
AbstractFemtocell, a small cellular base station in home and small business environment, is an attractive solution for operators to improve indoor coverage and network capacity in 3G networks. However, there are technical problems due to its mass deployment. The paper presents a femtocell architecture for LTE and investigates different handover scenarios. Two mobility management schemes at Radio Network Layer (RNL) are proposed and their signaling cost, complexity, standard impact and application scenarios are also discussed. Keywords-handover; femtocells; LTE; network architecture

I.

INTRODUCTION

Femtocell is a small cellular base station designed for use in subscribers home and small business environments. It enables Fixed Mobile Convergence (FMC) service by connecting to the cellular network via broadband communications (such as DSL or cable). Femtocell radiates very low power (<10mW) and can typically support 2-6 simultaneous mobile users. It has attracted strong interests within telecommunication industry recently due to its unique benefits: for the operators, the small, low-cost and low power home base station improves the indoor converge and network capacity, increases Average Revenue Per User (ARPU), and enhances customer loyalty; for the end users, the femtocell solution provides better in-building call quality and reduced calling cost at home. No expensive dualmode handset with both Wi-Fi and cellular radio access is needed. The battery life is also improved because of the low power radiation. A recent research report visions huge market potential for femtocells and estimates that by 2011 there will be 102 million users of femtocell products on 32 million access points worldwide [1]. Before the mass commercial deployment of femtocells, operators need to ensure that new femtocell deployments should meet regulatory requirements and enable smooth integration with the macro networks. The femtocell concept can be applied on different radio access technologies and related discussions have been actively conducted in 3GPP standardization meetings for both UMTS Terrestrial Radio Access Network (UTRAN) and Evolved UTRAN (EUTRAN). Market and technical data indicates wireless technologies, such as Long Term Evolution (LTE), will be of most use in metropolitan and enterprise hotspots, delivering improved coverage and enhanced capacity for operators [2]. Thus, the best way forward LTE may be to roll them out as femtocells first. Several technical challenges due to the mass deployment of femtocells have been discussed in recent literatures: (i) RF

interference. Unlike WiFi, femtocells operate in licensed spectrum owned by mobile operators and they may share the same spectrum with macrocell network. RF interference could happen between neighboring femtocells, femtocells to macrocells and vice versa. The spectrum has to be efficiently allocated in the femtocell network to mitigate the interference problems. In [3], interference avoidance strategies were developed in a coexisting environment of macrocells and femtocells. (ii) Self-optimizing and auto-configuration. As a Consumer Premise Equipment (CPE), femtocell is expected to operate in a plug and play fashion to ease installation, configuration and management. Methods for self-optimization and auto-configuration have been investigated in [4][5] to optimize the coverage of femtocells and minimize the impact on macrocell network. (iii) Integration and interoperability with core network. Femtocells extend the operators cellular network into home providing high data-rate services. Thus, integration and inter-operability with operators existing network and services are the first concern for the operators. UMTS-based, GAN-based and IMS-based femtocell solutions have been discussed in Femto Forum [6]. Our discussions on femtocells focus on the effects of the mass deployment on the flat architecture of LTE: a femtocell architecture for LTE is presented in the paper. Two mobility management schemes at Radio Network Layer (RNL) are proposed and their signaling cost, complexity, standard impact and application scenarios are also discussed. The rest of the paper is organized as follows. Section II presents a femtocell architecture for LTE and discusses the solution for the scalability and security problems. Section III investigates different handover scenarios and proposes two mobility management schemes. Section IV introduces an analytical model and the performance of the two schemes are evaluated in Section V. Section VI summarizes the paper. II. FEMTOCELL ARCHITECTURE IN E-UTRAN

Two types of nodes consist of the UTRAN: the NodeBs handle radio access in physical and network layers; the RNCs take care of radio related functions and it is the termination point for radio protocols. For the evolution in the radio access network, one of the important changes in E-UTRAN for LTE is the flat architecture where more functions are added to the base stations, i.e. eNodeB. As shown in Fig. 1a, each eNodeB is connected to the Evolved Packet Core (EPC) by the S1 interface. On the user plane the S1 interface terminates the Serving Gateway (S-GW); on the control plane the S1 interface terminates the Mobility Management Entity (MME) [7]. The MME handles control plane signaling, especially for mobility

978-1-4244-2517-4/09/$20.00 2009 IEEE

management and idle mode signaling. The S-GW processes the user plane data and is the local mobility anchor point. The eNodeBs are also interconnected with each other by the X2 interface, which is mainly used for inter-eNodeB handover purpose. As hundreds of thousand of femtocells are deployed, the scalability issue imposes costly re-configuration and operation in MME/S-GW. Because femtocells use residential broadband as the backhaul to connect to the mobile core network (CN), security issue needs be considered in order to protect the integrity of the network from malicious operations. Therefore, the femtocell network needs to consider both of the problems. Fig. 1b shows a femtocell E-UTRAN architecture where femtocell is referred to as Home eNB (HeNB). An intermediate node called HeNB Gateway (HeNB GW) is proposed to be located between HeNBs and the mobile CN. It acts as a virtual macro eNodeB towards CN and as a virtual CN node towards the HeNBs. The interface between HeNB HeNB GW, and the interface between HeNB GW MME/S-GW are S1 interface, no X2 interface exists between neighboring HeNBs.

management schemes applied in the HeNB GW at RNL and discuss their effects and applications in the following sections. III. HANDOVER PERCEDURE

In LTE, the network-controlled UE-assisted handovers are performed. For the handover between two eNodeBs, the handover decisions are made by eNodeBs without consulting with the MME: the source eNodeB decides to move the UE to the target eNodeB based on UE measurement report; the target eNodeB prepares radio resources before accepting the handover; after a successful handover, the target eNodeB indicates the source eNodeB to release the resources and sends a path switch message to the MME. The MME then sends a User Plane Update Request to the S-GW (the anchor for the user plane) about to which eNodeB the packets for the user shall route. The control messages are exchanged via the X2 interface between the two eNodeBs. The downlink packet data is also forwarded from the source to the target eNodeB via the same X2 interface. More details regarding the handover procedure can be found in [7][8]. However, in the femtocell architecture shown in Fig. 1b, no X2 interface exists between neighboring HeNBs. An intermediate node HeNB GW is also introduced in the EUTRAN between HeNBs and the CN. The difference of the two architectures will affect the mobility management procedures. Actually, we believe that the HeNB GW can play an important role in mobility management at RNL, especially during handover, instead of just working as a security gateway and traffic CNT/DST at TNL. There are three types of handover in femtocell deployment scenarios: (a) macrocell -> femtocell (b) femtocell -> macrocell (c) femtocell -> femtocell. The HeNB GW is assigned with a normal eNodeB ID (a unique identity for an eNodeB among nodes in E-UTRAN) so that it is seen by the MME as any other eNodeBs. HeNB GW can allocate private identities for each HeNB within its range and maintain a list of these HeNB IDs. A unique Tracking Area Code (TAC) is assigned to HeNB GW and HeNBs in the list. The MME associates the TAC and the eNodeB ID for the HeNB GW as a tuple (TAC, eHodeB ID) and notify the surrounding eNodeBs. For handover type (a), the source eNode B knows that the potential target cell is a HeNB cell from the TAC and the HeNB ID reporting from the UE. It identifies its HeNB GW and sends handover required message to MME. MME then routes the message to the HeNB GW, the latter delivers the message to the right HeNB. For handover type (b), HeNB GW gets the handover required message from source HeNB and forwards the message to MME. It is obvious to see that the MME is involved in handover type (a) and type (b) because it is the anchor point to route handover required message either inbound or outbound the HeNB GW. For handover type (c), it is of interest to investigate who will make the final handover decisions, or what is the mobility management strategy. Unlike the inter-eNodeB handover for LTE, where control plane messages and user plane packets are relocated from source to target eNodeB via X2 interface without the involvement of the EPC, the inter-HeNB handover has to be supported by upper nodes due to the lack of X2 interface. Therefore, we propose two mobility management methods for handover type (c).

S1

Figure 1. (a) E-UTRAN architecture (b) Femtocell architecture in EUTRAN

Two main functions are located in the HeNB GW: Concentration/Distribution (CNT/DST) and Security Gateway (SeGW). CNT/DST is a concentrator at Transport Network Layer (TNL), i.e., the HeNB GW transports many S1AP connections generated by a large number of HeNBs in single SCTP association between HeNB GW and MME (SCTP is the protocol used for signaling transport). As a distributor, HeNB GW distributes messages and traffic to different HeNBs with its range. SeGW performs HeNB and UE authentication for the access to the mobile CN. It is the end point for an IP-Sec tunnel to protect traffic in the public transport network between HeNB and HeNB GW. HeNB GW also have interface to operators O&M system for configuration and control. Since the UMTS RNC was removed in LTE, most of the RNC functionalities are shifted to eNodeB. The introduction of HeNB GW in E-UTRAN will affect the functional split. Whether the HeNB GW has the mobility management function is an interesting issue. At least, the HeNB GW needs to analyze dedicated S1AP messages for routing purpose at the Radio Network Layer (RNL). In this paper, we propose two mobility

IP -S ec

X2

tu nn e
X2

S1

Method 1 is to move the mobility anchor for user plane from the S-GW to the HeNB GW and let the HeNB GW make the handover decisions. This method implies a micro-mobility anchor at the HeNB GW for both control plane and user plane because the MME and the S-GW are not involved during interHeNB handover. It also implies that the S-GW does not need to be updated for the path switch after handover. When the HeNB GW receives a HO Required Message from the source HeNB, it will check the Target ID IE in the message: if the target cell is under its control, HeNB GW handles the handover. The signaling procedure is shown in Fig. 2.

(10-14) After receiving HO Command Message from the source HeNB(10), the UE detaches from the source HeNB and synchronizes the target HeNB (11); With L1/L2 processing, the UE accesses the target HeNB (12) and sends HO Confirm Message (13) to the target HeNB. The target HeNB notifies HeNB GW the success of handover by HO Notify Message (14). Both downlink and uplink data are then transferred through target HeNB. (15-17) The HeNB GW indicates the source HeNB to release the resources (15-16). The handover is completed after the HeNB GW receives the Release Complete Message (17). Method 2 is similar to the S1-based handover in LTE: the MME confirms the handover request and makes sources release decision (more details in [8]). The S-GW still remains as the anchor for the data. Therefore, HeNB GW simply forwards handover messages between HeNB and MME, working more like a transparent node or relay in terms of mobility management at RNL. The handover procedure for method 2 below the HeNB GW is the same as the left part shown in Fig.2. Since the HeNB GW acts as a relay, it forwards handover messages either from the source HeNB or from the MME instead of local processing. Thus, more control signaling messages are exchanged between radio access network and core network. The S-GW also has to be notified with the change of end point of GTP tunnel after a successful handover. The benefit for method 2 is that it has low impact on LTE standards. IV. ANALYTICAL MODEL

Figure 2. Method 1: Inter-HeNB handover with HeNB GW as the mobility anchor

(1-3) Once the source HeNB receives the UE measurement report (1), it makes the handover decisions (2) based on the measurement report and sends HO Required Message to the HeNB GW (3). The target ID IE in the HO Required Message is set to be the identity of the target HeNB. (4-5) The HeNB GW analyzes the HO Required Message and finds that the target ID is under its control, it then performs the access control to check that whether the UE has right to access the target HeNB (4) and sends HO Request Message to the target HeNB (5). However, if the target ID is not among the list, the HeNB GW forwards the message to MME. In this case, it is handover type (b). (6-7) The target HeNB performs the admission control based on the availability of the required resources (6), prepares HO over L1/L2 layers and then sends the HO Request Ack. Message to the HeNB GW (7). (8) As soon as receiving the HO Request Ack. Message, the HeNB GW switches the downlink path from source HeNB to target HeNB on the user plane. (9) The HeNB GW sends HO Command Message to source HeNB to indicate that the handover have been prepared at the target side.

We develop a simple analytical model to investigate the signaling cost based on the work in [9][10]. In the analytical model, if UE moves across the border of two HeNBs in active state, it requires handover from one HeNB to another and generates handover signaling messages.

Figure 3. Timing diagram for the analyticla model

The timing diagram for the analytical model is shown in Fig. 3. We assume that the call may occur at any moment. The call arrival rate follows the Poisson process. t1 is the moment when UE enters the range of a HeNB; t2 is the moment when UE leaves the range of the HeNB and enters the range of another HeNB. Two cases generate handover messages at the moment t2: (1) UE has a call at moment P1 before UE enters the range of the HeNB and remains connected after UE leaves the range of the HeNB. (2) UE starts at moment P2 in the range of the HeNB and remains connected after UE leaves. The call ends at the moment P3. The probability of the handover that happens at moment t2 is defined as the sum of the probability of case (1) Pr1 and the probability of case (2) Pr 2 :

Pr = Pr1 + Pr 2 .

(1)

TD is denoted as the call duration and is exponentially distributed with mean 1 / . Its probability density function, denoted as fTD (t ) , is given by: f T (t ) = e t . TR is UE residence
D

time in the range of a HeNB and is exponentially distributed with mean 1 / . Its probability density function, denoted as fTR (t ) , is given by: f T (t ) = e t . TD and TR are independent
R

mode is the sum of the transmission cost ij and processing cost i , where i and j indicate nodes in the network, and pair ( i , j ) indicates the two nodes exchanging the signaling messages between each other, such as (HeNB, HeNB GW) or (HeNB GW, CN). Transmission cost is the cost of delivering handover message between two nodes and processing cost is the cost of processing messages at each node in the network. Notations for different costs are given as the following:
H UE GW H CN GW UE H GW CN

random variables.
TDr , denoted as the residual time for call duration, has the same distribution as TD since moment P1 can be infinitely near moment t1. Similarly, TRr , denoted as the residual time for UE residence time, has the same distribution as TR since moment P2 can be infinitely near moment t2.

Transmission cost between UE and HeNB Transmission cost between HeNB and HeNB GW Transmission cost between HeNB GW and CN Processing cost at UE Processing cost at HeNB Processing cost at HeNB GW Processing cost at CN

The probability for case (1) becomes: Pr1 = P( P1 < t1 < P1 + TD ) P(TDr > TR ) (2)

For a UE roaming between two neighboring HeNBs, the total handover signaling cost per UE is related to the probability of handover and is defined as:
= Pr ( ij + i ) (6)

The first term in the Eq. (2) ensures that the call should start before the moment t1 under a constraint, i.e., the amount of time ahead of t1 cannot be longer than the call duration TD . The second term in the Eq. (2) guarantees that the call should end after the moment t2, i.e., TDr > TR . Please note that the condition TD > TR can not guarantee the end of call after the moment t2. Taking Laplace transform of fTD (t ) and fTR (t ) , Eq.(2) can be expressed as:

where the term in the brackets is the signaling cost for one handover. V. PRERFORMANCE EVALUATION

Pr1 =

0 t

t x te fTD ( y)dydt (1 e f Dr (t )dxdt ) 0 t

= ( + )2 ( + )
The probability for case (2) is given by: Pr 2 = P (t1 P 2 t1 + TRr ) P(TD > TRr )

(3)

The transmission cost and processing cost can be expressed in terms of time needed to deliver and process signaling messages as in [10], i.e., the transmission cost is one-way transfer delay between a pair of nodes; the processing cost is the delay to process one message in one node. Although the lengths of handover messages are varied, we assume that transmitting time for different messages between the same pair of nodes are approximately the same and the processing time for different message at the same node are also approximately the same for simplification. The cost parameters for performance evaluation are shown in Table I. Their scales (no units) are derived from [11][12] for the delay on IP-based network.
TABLE I.
Transmission Cost Value

(4)

COST PARAMETERS
Processing Cost Value

The first term in Eq. (4) ensures that the call should start in the range of the HeNB. The second term in Eq. (4) ensures that the call should end after the moment t2. Similarly, Eq (4) can be expressed as: Pr 2 = te t f Rr (t )dt e y f Rr (t )dydt
0 0 t

H UE

2 2 4

CN GW H UE *

4 3 2 40

GW H CN GW

2 ( + )( + ) 2

(5)

* It is the interruption time from the point when UE detaches from the source cell to the point when UE accesses to the target cell.

The signaling cost associated with each handover that is occurred when UE crosses border of two HeNBs in active

When the average call arrival rate is set as 0.1call/minute and the average residence time is set as 1 and 10 minutes, the total signaling cost versus average call duration time is shown in Fig. 4. When the call lasts longer, the probability of handover occurred across the border is bigger. So the total signaling cost increases when average call duration time

increases. It is also observed that the signaling cost is increasing with UE residence time increasing. Since the HeNB GW handles the local handover between two HeNBs in method 1 instead of forwarding messages to MME as in method 2, the signaling cost for method 1 is around 30% lower than that for method 2. Fig. 5 shows the handover signaling cost versus the average call arrival rate (call/minute) when the residence time is 1 / = 10 minutes and call duration is 1 / = 3 minutes. The probability of handover increases when more calls arrive during UE residence time in the analytical model. The signaling cost is then increasing.

HeNB GW can be removed with low impact on the functionalities in EPC and the E-UTRAN architecture. Considering that the initial LTE femtocell deployment may start in offices or campus hot spots, method 1 can be used in corporate environment where inter-HeNB handover happens quite frequently. The HeNB GW can effectively remove the backhaul signaling traffic and reduce the handover processing time. For the home environment where users have different access rights, most of mobility scenarios are between femtocells and macrocells. In this case, the HeNB GW is used for security purpose and traffic concentrator at TNL. Therefore, a less complex HeNB GW as in method 2 can be applied. VI. CONCLUSION

Figure 4. Totoal signaling cost versus average call duration time

This paper presents a femtocell architecture for LTE. An intermediate node called HeNB GW is introduced to solve the scalability and security problems raised by mass deployment of femtocells. Two mobility management schemes are proposed to be applied in the HeNB GW at RNL for the purpose of integrating the LTE architecture. In method 1, HeNB GW acts as a mobility anchor to control handovers among femtocells. While in method 2, HeNB GW is more like a relay between HeNB and EPC. The two methods are compared with handover signaling cost using an analytical model. In reality, they can be implemented in different application scenarios. Despite the complexity in HeNB GW, method 1 reduces the signaling traffic in EPC and is more suitable for enterprise and campus use. Method 2 is based on S1 handover in LTE. It has small impact on E-UTRAN standards and can be applied for the home use. REFERENCES
ABI research, Femtocell Access Points, Fixed-Mobile Convergence for Residential, SMB, and Enterprise Markets, research report, 2007. [2] D. Pulley, The macrocell is dead, long live the network, Sept. 2008, http://www.electronicsweekly.com/Articles/2008/09/15/ [3] V. Chandrasekhar and J. G. Andrews, Uplink Capacity and Interference Avoidance for Two-Tier Cellular Networks, Proc. IEEE GLOBECOM 2007. [4] L. T. W. Ho and H. Claussen, Effects of User-deployed, Co-channel Femtocells on the Call Drop Probability in a Residential Scenario, Proc. IEEE Indoor and Mobile Radio Communications (PIMRC) , 2007. [5] H. Claussen, Performance of Macro- and Co-channel Femtocells, Proc. IEEE Indoor and Mobile Radio Communications (PIMRC) , 2007. [6] Femto Forum, http://www.femtoforum.org. [7] 3GPP TS 36.300, Evolved Universal Terrestrial Radio Access (E-TRA) and Evolved Universal Terrestrial Radio Access Network (E-UTRAN) Overall description; Stage 2, Rel. 8, v8.5.0, May 2008. [8] 3GPP TS 23.401, General Packet Radio Service (GPRS) enhancements for Evolved Universal Terrestrial Radio Access Network (E-UTRAN) access, Rel. 8, v8.3.0, Sept. 2008. [9] U. Narayanan and J. Xie, Signaling Cost Analysis of Handoffs in a Mixed IPv4/IPv6 Mobile Environment, Proc. IEEE GLOBECOM, 2007. [10] J.S.M. Ho and I. F. Akyildiz, Local Anchor Scheme for Reducing Signaling Costs in Personal Communications Networks, IEEE/ACM Transactions on networking, Vol 4(5), 1996 [11] ITU-T Recommendation G.114, One-way transmisson time, 2003. [12] E.J.Hernandez-Valencia and M.C. Chuah, Transport Delays for UMTS VoIP, Proc. IEEE WCNC 2000. [1]

Figure 5. Totoal signaling cost versus call arrival rate

For handovers between femtocell macrocell and vice versa, MME/S-GW is always involved. For handover between two femtocells, method 1 reduces signaling cost and processing load in the CN because the HeNB GW blocks local roaming among HeNBs. It also reduces the time for full handover procedure. However, it requires the HeNB GW to understand and responds S1AP messages in stead of simply forwarding messages during inter-HeNB handover, the HeNB GW is therefore more complex. On the other hand, in method 2, the HeNB GW acts more like a relay between HeNB and MME/SGW. If the S-GW is deployed in a separate physical unit, the

Anda mungkin juga menyukai