A Client-based Vertical Handoff Approach for Seamless ... - IEEE Xplore

Download PDF
10 downloads 8625 Views 510KB Size Report
Comment
Email: [email protected]. Abstract—With the rapid growth of mobile Internet, offering seamless connectivity and high-speed multimedia services in different ...
A Client-based Vertical Handoff Approach for Seamless Mobility in Next Generation Wireless Networks Yong-Sung Kim

Dong-Hee Kwon

Young-Joo Suh

Dept. of Computer Science and Engineering, POSTECH Korea Email: [email protected]

POSDATA Co., LTD. Korea Email: [email protected]

Dept. of Computer Science and Engineering, POSTECH Korea Email: [email protected]

Abstract—With the rapid growth of mobile Internet, offering seamless connectivity and high-speed multimedia services in different types of wireless networks are important features in next generation wireless networks (4G networks). Many researchers envision that the 4G networks will consist of many radio access technologies which integrate into all IP-based networks while providing seamless handoffs across the heterogeneous wireless networks. Mobile IP is a prominent solution to handle mobile nodes’ movement in IP-based networks. However, Mobile IP suffers from long handoff latency since it is designed to support the macromobility of a mobile node. Therefore, it is necessary to devise a scheme that offers seamless connectivity during Mobile IP handoffs across the heterogeneous wireless networks. In this paper, we propose a client-based handoff management system for multi-networks to provide end-users with seamless connectivity across heterogeneous wireless networks. Our proposed system is designed as a common network interface at a client side and provides transparent services to IP/MIP layers regardless of used wireless technologies. With this system, we also present vertical handoff performance over a loosely coupled cdma2000/WLAN experimental test-bed. The experimental results show that our system enables a mobile node to handoff seamlessly across different types of wireless networks without any modifications to the existing IP/MIP stack and core network. Index Terms—vertical handoff; heterogeneous wireless networks; seamless handoff; 4G networks

I. I NTRODUCTION There are various wireless network technologies that offer Internet access to end-users. The technologies have been developed separately and thus offer different services, coverage areas, data rates, and so on. For example, second-generation (2G) cellular networks are designed to serve speech and lowbit-rate data services to end-users while third-generation (3G) cellular networks are developed to provide higher data-rate services. During the evolution from 2G to 3G, several wireless systems including IEEE 802.16 (WiMAX) [16], Wireless LAN (WLAN) [8], Bluetooth, and HiperLAN have been developed. WLAN offers higher data rates up to 54Mbps, but it is only suitable for small office networks and public hotspot areas such as campus and airports, whereas 2G/3G cellular networks cover a larger area. Thus, it is expected that various network architectures and technologies will coexist and interwork

978-1-4244-2413-9/08/$25.00 ©2008 IEEE

together to support high usability, seamless inter-technology handoffs, multimedia services, etc in fourth-generation (4G) networks [17]. Combining the 3G and WLAN networks is considered as one of the first steps toward achieving always-on Internet connectivity in 4G networks due to the complementary characteristics of the 3G and WLAN technologies. There are two integration architectures for providing integrated service capability across 3G and WLAN networks: tightly coupled architecture and loosely coupled architecture [3, 7]. In the tightly coupled architecture, WLANs appear just like another 3G access network and provide 3G services to WLAN users. This architecture utilizes the WLAN gateway to hide the details of WLAN from the 3G core network as shown in Figure 1(a). The WLAN gateway is connected to the PDSN and provides WLAN users with all the 3G protocols (e.g., mobility management). However, this architecture presents several disadvantages. Since the WLAN gateway is connected to the PDSN entity, a new interface between the WLAN gateway and the PDSN should be defined, which requires modifications or extensions to the 3G core network. It introduces not only the network deployment cost but also the network integration cost during initial network deployment of 3G and WLAN. Contrary to the tightly coupled architecture, the loosely coupled architecture introduces a WLAN gateway which connects WLANs to the Internet as shown in Figure 1(b). This architecture completely separates the data paths in 3G and WLAN networks and thus WLAN data traffics do not go through the 3G core network, compared to the tightly coupled architecture. With these characteristics, the loosely coupled architecture does not require to reconfigure the 3G core network. However, this architecture needs mobility management schemes to support handoffs between 3G and WLAN networks. 4G networks consist of a diverse set of wireless networks and are integrated into IP-based networks, which further require a seamless handoff support. Once the decision of vertical handoff from one type of wireless network to another has been made, the key issue for a handoff system is the

419

Billing Server

Billing Server

Internet

Internet Home AAA

3G Core Network

New interface for integrating 3G and WLAN

AAA: Authentication , Authorization , and Accounting AP: Access Point BSC: Base Station Controller BTS: Base Transceiver Station BS: Base Station (BSC + BTS) HA: Home Agent MN: Mobile Node PCF: Packet Control Function PDSN: Packet Data Serving Node

HA PDSN PCF Gateway

BSC cdma2000 Network

Connected via Internet

HA PDSN PCF Gateway BSC cdma2000 Network

WLAN Network BTS

Home AAA

3G Core Network

WLAN Network BTS

AP

AP

MN

MN

(a) Tightly coupled architecture Fig. 1.

(b) Loosely coupled architecture

Integration architectures for 3G and WLAN

mobility management scheme, which maintains seamless connectivity during handoffs. There are several kinds of mobility management protocols operating in different layers such as Mobile IP [5, 19] in the network layer, TCP-Migrate [20] / Stream Control Transmission Protocol (SCTP) [21] in the transport layer, and Session Initiation Protocol (SIP) [18] in the application layer. Among these protocols, Mobile IP (MIP) is the most widely studied approach to handle mobility and is a prominent solution as a mobility management protocol for heterogeneous wireless networks which are deployed on top of IP-based networks. Thus, we focus on the MIP-based mobility management protocol in this paper. However, MIP suffers from long handoff latency since it is designed to support the macromobility of the Mobile Node (MN) and is not suitable for high-speed multimedia services which require lossless handoff as well as always-on Internet connectivity. We expect the emerging 4G networks to serve continuous and high-speed data service to end-users regardless of their movements across the diverse network environments. Therefore, it is necessary to study the scheme that provides currently deployed network with backward compatibility as a cost-effective way and offers seamless connectivity during MIP handoffs across heterogeneous wireless networks. In this paper, we design and implement a handoff management system for multi-networks (HMSMN) to provide seamless connectivity to end-users across heterogeneous wireless networks without modifying or extending the network side. HMSMN is designed as a common interface that resides between the IP/MIP layer and the actual network interfaces for a MN. HMSMN provides a transparent service to the IP/MIP stack and seamless connectivity for inter-technology handoffs by simply modifying the MN side. Moreover, HMSMN can be closely coupled with the MIP stack to perform a handoff to its preferred network determined by handoff metrics such

as signal strength, QoS parameters, available services, enduser preference, etc. Thus, we address two possible designs of HMSMN based on inter-technology handoff approaches - basic integration and tight integration. To validate our approach, we have implemented HMSMN over a loosely coupled cdma2000 1xEV-DO/WLAN experimental test-bed. The experimental results show that HMSMN enables MNs to handoff seamlessly across different types of wireless networks by preserving the existing IP/MIP stack and modifying only the client side. The remainder of this paper is organized as follows. Section II describes related work regarding to the connectivity for 4G networks. In Section III, we present the proposed HMSMN system in detail to support a seamless handoff across heterogeneous wireless networks. Section IV describes the loosely-coupled based cdma2000 1xEV-DO/WLAN test-bed and examines the vertical handoff performance on realistic environments. Finally, concluding remarks are given in Section V. II. R ELATED W ORK Stemm et al. [4] proposed a vertical handoff system for infrared/WLAN networks and WLAN/Ricochet networks. The system provides an end-user with the best possible connectivity with a minimum disruption during a handoff. Siddiqui et al. [9] proposed two tight integration approaches by differentiating integration points, namely SGSN and GGSN. It integrates 3G and WLAN networks as the typical tight integration architecture, which has been proposed by 3GPP, but it still has a problem in modifying SGSN and GGSN which has been already reported in the tightly coupled architecture. Akyildiz et al. [1] studied an adaptive protocol suite for 4G wireless data networks. They have noticed the importance of adaptiveness to each layer in the protocol stack at a

420

MN to address several problems including rate adaptation, congestion control, mobility support, and coding. In particular, they present adaptiveness in layers 2 (link), 4 (transport), and 7 (application) to satisfy the demand of heterogeneity and highspeed multimedia applications in 4G wireless data networks. There are several research results to address vertical handoff issues by introducing an integration point among different heterogeneous wireless networks. Buddhikot et al. [2] introduced a new network element called integration of two access technologies (IOTA) gateway in IEEE 802.11 networks and a new service access software on client devices for an interworking solution of IEEE 802.11 and 3G networks. The IOTA gateway cooperates with the client software to offer an integrated 802.11/3G wireless data services that support seamless inter-technology handoffs, Quality of Service (QoS) requirements, and multi-provider roaming agreements. Nevertheless, they did not consider minimizing the disruption period during vertical handoffs by closely integrating their approach with Mobile IP, which is one of the issues we address in this paper. Sharma et al. [10] presented the design, implementation, and evaluation of a vertical handoff system which enables a MN to automatically switch between WLAN and GPRS networks by introducing a simple extension to existing Mobile IP implementation. However, it needs a special agent, GPRS foreign agent, for forwarding data towards a MN while it is away from the home network. Chen et al. [11] showed a vertical handoff solution by introducing an entity, handoff server, which provides IP tunneling techniques and serves as a home agent in the Mobile IP protocol. It only needs a little change at the client side, but the handoff server has to be additionally deployed in the existing network, which causes much overhead in scalability. Furthermore, this scheme has been tested for the handoff solution between 802.3 and 802.11 networks and results indicate that it is not feasible to test vertical handoff between 802.11 and 3G networks without modifying or extending the handoff server entity. III. H ANDOFF M ANAGEMENT S YSTEM M ULTI -N ETWORKS (HMSMN)

FOR

User level Handoff Management Unit Handoff execution function Handoff detection Handoff trigger function function

IP/MIP Stack Device Convergence Unit Network switching function Network detection function

RSS measurement function

Hardware Network Interface 1

Network Interface 2

…

Kernel level Fig. 2.

System architecture of HMSMN

A. Architecture of HMSMN The architecture of the proposed HMSMN is shown in Figure 2. As shown in the figure, Device Convergence Unit (DCU) is located between the MIP stack and the network interface as a virtual network interface to provide seamless handoffs to the MIP stack. This architecture offers the abstraction of a single virtual network interface to the Operating System and supports multiple network interfaces according to different types of radio technologies. HMSMN also allows the handoff management unit to gather the status of network connection, reflect several kinds of handoff metrics, and select a specific radio access technology that will be used for subsequent data communication. The functionalities of handoff management and device convergence units are as follows: •

Based on the loosely coupled architecture, we design and implement a handoff system that allows a MN to handoff across different types of wireless networks in a manner that is completely transparent to the MIP stack while providing seamless connectivity. Since our realization of the handoff system is implemented as a single virtual network interface in the protocol stack of the MN, it is apparently simple, scalable, and cost-effective compared to the modification or extension to the network side. Furthermore, the virtual network interface hides all the details about the change of network interfaces currently being used from the MIP stack. Thus, the vertical handoff is handled at the virtual network interface without the MIP stack’s knowledge. In result, the end-users can achieve seamless connectivity no matter which technologies they currently use.

421

•

•

Network detection function assists the MN to learn which available technologies can be used for MN’s later data communication. It automatically detects available network technologies when network interfaces are attached to the system and gathers network information through the attached network interfaces, such as a radio access technology (e.g., Mobile cellular system, WiMAX, WLAN, Bluetooth, etc.). RSS measurement function periodically measures the currently received signal strength of the serving BS and reports the information to the handoff trigger function which will be used for a handoff initiation trigger. Network switching function provides the end-user with intra or inter-technology handoffs according to the selected network interface. If this function is called, then the network interface is changed from the previous network interface to the new one by deactivating the currently used

•

•

•

network interface and activating the requested network interface. It also sets up proper parameters to the virtual network interface for the new network interface such as MAC address, IP address, etc. Handoff trigger function gathers information from the RSS measurement function and fires a handoff trigger to start handoff operation at an appropriate time. Handoff detection function monitors MN’s point of attachment and notifies the handoff execution function of the movement to a new network. Handoff execution function performs handoff operation by cooperating with the handoff detection function and the network switching function. When the handoff detection function notifies a new link attachment, which satisfies a specific Received Signal Strength (RSS) threshold and user preference, the handoff execution function performs the MIP handoff operation through the newly selected network interface with the aid of the network switching function.

B. Operation of HMSMN 1) Network Detection: The MN may detect multiple network technologies based on its network interfaces. With the multiple network interfaces, the MN obtains the information of neighboring wireless technologies as well as the information of currently serving technology. For example, a MN equipped with 3G and WLAN interfaces moves around two wireless technologies as shown in Figure 1(b). In the figure, the MN is aware that there are two available access technologies by receiving signals from the BS and AP. Using the network detection function, the MN knows and manages the currently available network lists, and uses the information during its handoff process in near future. 2) Handoff Decision: It is difficult to predict when the MN should handoff to a particular BS (AP) before breaking its current connection. Moreover, in heterogeneous wireless networks, there are also many handoff decision metrics such as RSS, end-user preference, QoS requirement, service type, communication cost, etc., because each wireless technology has its own characteristics and services. In this paper, we only consider the end-user preference and the RSS as our primary handoff metric since our focus is to validate our approach. It can be more effective if other handoff decision metrics listed above are used in conjunction with the end-user preference and the RSS information. The end-user preference is obtained by the manual input from the end-user at the user level (Note that we use WLAN as a preferred network in our experiment). The RSS information is also gathered from the RSS measurement function which constantly monitors network status through network interfaces equipped in the MN. 3) Handoff Initiation: Once the MN learns a new network from the handoff detection function, it estimates the right time to start the MIP handoff operation (MIP registration procedure). The handoff trigger function uses end-user preference and RSS information to make a vertical handoff decision on which network the MN should handoff to. When the measured

RSS of the currently serving BS (AP) drops below a predefined threshold, the handoff trigger function gives a handoff trigger with network information, of which the MN should initiate handoff operation, to the handoff execution function to start the MIP handoff operation. The information of the newly selected network should satisfy the end-user requirements: end-user preference and a RSS threshold. 4) Handoff Execution: When the handoff execution function receives the handoff trigger from the handoff trigger function, it starts the MIP handoff operation. When the MN starts the MIP handoff operation, there are two different approaches, basic and tight integration approaches, to integrate HMSMN and MIP on whether HMSMN closely cooperates with MIP or not. We provide a detailed discussion about the integration approaches in the following subsection. In this subsection, we describe the handoff execution process based on the basic integration approach. In the basic integration approach, the MN first disconnects its current network connection and then the MN switches to a newly selected BS (AP) using the network switching function and performs MIP handoff operation. Once the MN finishes the MIP registration operation, the MN can finally start its data communication through the newly selected network interface. The operation of HMSMN is summarized in Figure 3. First, the MN learns the available network lists from the network detection function. Then, it monitors the RSS of currently associated BS (AP) and determines whether the RSS is below the predefined threshold or not. If the RSS is lower than the threshold, then the MN selects the best BS to connect based on one of the handoff metrics (end-user preference). Based on the selected BS, the MN measures the RSS of the newly selected BS and gives the handoff trigger to the handoff execution function if the measured RSS is above the predefined threshold. Next, the handoff execution function performs MIP handoff operation based on the integration approach. When the basic integration approach is selected, the MN switches to the newly selected network interface and performs MIP handoff operation (agent discovery and registration procedures). Once the MN finishes its registration procedure, the MN starts its communication through the newly selected BS. Otherwise, in the tight integration approach, the MN starts its MIP handoff operation before its actual network switching from the previous network interface to the new network interface that would result in less handoff disruption period than the basic integration approach. Since the MN uses multiple network interfaces simultaneously in the tight integration approach; one for the data communication and one for the MIP handoff signaling, we can expect that the MN experiences less handoff latency than the basic integration approach. C. HMSMN Integration with MIP In this subsection, we describe two integration approaches between HMSMN and MIP in detail. The first integration approach, termed a basic integration approach, is to make the network interface switching of HMSMN to be transparent to the MIP operation. In this case, HMSMN initiates a vertical

422

Start

Network detection

Is there any available network ?

Handoff trigger

No

Yes Is RSS below the threshold ?

Basic integration approach

Tight integration approach

Network interface switching

MIP handoff operation through new network interface

MIP handoff operation

Network interface switching

No

Yes Select next BS based on user preference RSS measurement for selected BS

Communication through new BS Is RSS above the threshold?

No End

Yes

Fig. 3.

Flow diagram of HMSMN operation

handoff by simply switching the network access technology (via. network switching function) between multiple network interfaces. HMSMN activates the newly selected network interface once it deactivates the previously enabled network interface to avoid unnecessary oscillations between multiple networks. If the handoff decision policy of the MN is eager cell switching, i.e. the MN performs MIP handoff as soon as it receives an Agent Advertisement (AA) message from the new agents, then it might be under a puzzled state because two different AA messages arrive at the MN by enabling multiple network interfaces at the same time. Thus, the basic integration approach provides a simple method of inter-technology handoff to MIP by enabling a single network interface at a time. Thus, it does not offer an effective handoff mechanism since it simply uses a network interface at a time and the MN may experience the same handoff disruption period (agent discovery and MIP registration period) as it does in horizontal MIP handoffs. If some mechanism notifies the MIP stack of the current network status (whether a MN will handoff soon or not) and the MIP stack prepares its handoff in advance before its actual handoff, then the vertical handoff latency can be further reduced than the horizontal one. For this, we propose another integration approach, termed a tight integration approach, where the operations of HMSMN and MIP are closely coupled. In this approach, HMSMN interworks with MIP to maintain seamless connectivity during the vertical handoff by triggering the MIP handoff operation for the new network while maintaining the connection to the previous network. HMSMN informs its network switching event to the MIP stack to prepare the impending handoff event.

The MIP stack performs its handoff operation in advance through the newly selected network interface while the MN maintains its connectivity for the previous network. Compared to the basic integration approach, this approach enables multiple network interfaces at the same time. Since HMSMN filters AA messages from newly selected network interfaces until the MIP handoff operation finishes, the tight integration approach does not introduce unnecessary ping-pong movement between multiple networks compared to the basic integration approach. Note that the MIP stack only sees a virtual network interface of HMSMN due to its transparency to the upper IP/MIP layer, which implies that the MIP stack is not aware of the fact that multiple network interfaces are currently activated by HMSMN. Therefore, HMSMN should decide which network interface to use depending on the types of outgoing packets. If the MN is a receiver, then only the outgoing packet related to handoff operations is the MIP registration message, which should be sent through the new network interface during an impending handoff. The data packets destined to the MN are received by the previous network interface until the binding cache of the Home Agent (HA) is updated. Once the binding cache is updated, the data packets for the MN are tunneled to the MN’s newly visiting network. When the MN receives tunneled packets via the new network interface, HMSMN deactivates the previous network interface. If the MN is a sender, then the MN should use the previous network interface for its outgoing packets except the one for the MIP registration until the MIP registration finishes. After finishing the MIP registration, HMSMN deactivates the previous network interface. In

423

MN obtains Co-located Care-of Address (CCoA) when the MN moves to the WLAN network. Since we cannot modify some entities to support MIP in the cdma2000 network, we used CCoA mode operation of the MIP during MN’s handoff. Figures 5 and 6 show the performance of the vertical MIP handoff for two integration approaches of HMSMN when the MN changes its connectivity from the cdma2000 to WLAN network and vice versa as shown in Figure 4. In both cases, we measured end-to-end delay of each UDP packet and the number of lost packets during the handoffs. The UDP packet size is 256 bytes and each packet is generated at every 50ms interval by the Correspondent Node (CN) to emulate the properties of VoIP traffic. We measured the handoff performance at least 10 times for each handoff case and they showed similar results. Therefore, we only show a snap shot of the experimental results in this paper.

HA

Public Internet IP Network Router 1

PDSN

AR (DHCP Server)

Router 2

BSC CN WLAN Network

cdma2000 1xEV-DO Network BTS

AP

MN Vertical Handoff

Fig. 4.

A. Handoff Performance from cdma2000 to WLAN network

Experimental test-bed

this way, the MIP stack proceeds with its handoff operation by using the new network interface in advance while the MN maintains its network connectivity through the previous network interface. As explained above, the tight integration approach does not interrupt MN’s current communication and the MN rarely experiences packet loss due to the MIP handoff. It is obtained by the best utilizing multiple network interfaces and the close interaction between HMSMN and MIP at the MN. By closely coupling HMSMN with MIP, the handoff performance can be improved and seamless connectivity across the different types of networks can be provided. IV. P ERFORMANCE E VALUATION In this section, we present the handoff performance of HMSMN which supports two wireless technologies: cdma2000 1xEV-DO and WLAN. Using two network interfaces, we measured handoff performance of the basic integration and tight integration approaches of HMSMN for VoIP traffic. Our experimental setup consists of a looselycoupled architecture for integrating cdma2000 1xEV-DO and WLAN networks as shown in Figure 4. The cellular cdma2000 1xEV-DO network infrastructure currently in use is the SK Telecom production cdma2000 network in Korea. To access the cdma2000 network, the MN uses the USB type of the cdma2000 modem, which supports maximum 2.4Mbps for the downlink and 153kbps for the uplink. The WLAN network access is offered by IEEE 802.11b and the MN is equipped with an 802.11b PCMCIA-based card. Several other entities are used for handoff experiments as shown in the figure. Especially, to support MN’s mobility between cdma2000 and WLAN networks, the MIP agent, HA, is implemented in the experiment network. The HA and the MN in the test-bed network operate on top of Linux operating system and use Dynamics Mobile IP which was developed by Helsinki University of Technology (HUT) [6]. The AR also operates as a DHCP Server for the MN which implies that the

Figure 5 shows the snapshot of the measured end-to-end packet delivery (transmission time) for each integration case when the CN sends UDP packets to the MN. In Figure 5(a), the MN performs a handoff from cdma2000 to WLAN during the experiment by using the basic integration approach. We observed that the MN lost 25 packets (from the 92th to 116th packets) during the MIP vertical handoff for the basic integration approach. It implies that the vertical handoff latency from cdma2000 to WLAN takes about 1.25 seconds since the packet generation interval is 50ms in this experiment. The long handoff latency of the MIP vertical handoff is due to the fact that the MN must establish the WLAN connection for L2 handoff and then, performs the MIP handoff operation (agent discovery and registration procedures) for L3 handoff. Since in the basic integration approach, HMSMN does not simultaneously utilize the multiple network interfaces as a make-before-break model, but uses only one interface at a time, it suffers from many packet losses during the vertical handoff as the conventional MIP does. The average end-to-end delay in the cdma2000 network is around 70ms and the packets whose delay ranges between 70ms and 250ms are mainly due to buffering [12] offered by current cdma2000 networks. Most cdma2000 networks provide a substantial amount of buffering (for every cdma2000 mobile device) in their Packet Control Function (PCF) nodes due to their low bandwidth. Thus, packets destined for the downlink become queued up at the PCF node and the end-to-end delay is varied during the vertical handoff. Contrary to the basic integration approach, the tight integration approach presents better handoff performance as shown in Figure 5(b). In this case, there are almost no packet losses during the vertical handoff from cdma2000 to WLAN since HMSMN performs switching to the new network while maintaining the connection to the previous network (the cdma2000 network in this case). Thus, the MN can seamlessly communicate with the CN using the previous network while it completes the MIP handoff operation using the new network. Until the handoff finishes, HMSMN maintains the previous

424

300

250

250

End-to-end delay (ms)

End-to-end delay (ms)

300

cdma2000 to WLAN Handoff Period

200 150 100 50

cdma2000 to WLAN Handoff Period

200 150 100 50

0

0

0

50

100

150

200

0

50

(a) Basic integration approach

200

150

200

cdma2000 → WLAN handoff performance

300

300

250

250

End-to-end delay (ms)

End-to-end delay (ms)

150

(b) Tight integration approach

Fig. 5.

WLAN to cdma2000 Handoff Period

200

100 Packet number

Packet number

150 100 50

WLAN to cdma2000 Handoff Period

200 150 100 50

0

0 0

50

100

150

200

0

50

Packet number (a) Basic integration approach Fig. 6.

100 Packet number

(b) Tight integration approach WLAN → cdma2000 handoff performance

network (the cdma2000 network) as well as the new network (the WLAN network). In the tight integration approach, we expect no packet loss during the vertical handoff. However, there are some packet losses during the handoff because the MN loses the buffered packets in the previous cdma2000 interface when it switches to the WLAN interface once it finishes the MIP handoff operation. Since HMSMN manages multiple network interfaces transparently, the buffered packets in each network interface are not managed and handled in HMSMN for low layer transparency. However, it still achieves a good handoff performance compared to the basic integration approach where the handoff latency of the tight integration approach is around 250ms in this experiment. B. Handoff Performance from WLAN to cdma2000 network In the handoff case from WLAN to cdma2000, as shown in Figure 6, we can see that the MIP vertical handoff performance

is almost the same as that from the cdma2000 to WLAN. The basic integration approach still shows lots of packet losses during the handoff as shown in Figure 6(a). There are 18 packet losses (from 80th to 97th) and the handoff latency is around 900ms in this experiment. Since the MN does not simultaneously use multiple network interfaces for the vertical handoff as discussed earlier, the MN experiences lots of packet losses during the handoff in the basic integration approach. Generally, cdma2000 ’ WLAN handoff performance is better than WLAN ’ cdma2000 due to the make-before-break handoff, but the basic integration approach cannot use makebefore-break handoff due to the its characteristics. On the other hand, the tight integration approach shows better handoff performance compared to the basic integration approach. Since the tight integration approach supports multiple network interfaces simultaneously with the aid of HMSMN, the MN uses

425

its previous network interface (the WLAN network interface in this case) for data communication while the MN performs L2 and L3 handoff operation using the cdma2000 network interface. Although the 6 lost packets (from 63th to 68th) are presented as shown in Figure 6(b) due to the dropped packets in the previous network interface as explain before, the MN experiences less handoff latency (300ms) than the case of the basic integration approach. It implies that HMSMN intelligently manipulates the multiple network interfaces and coordinates them for heterogeneous wireless networks. V. C ONCLUSION In this paper, we presented client software for supporting integration of multiple wireless technologies such as cdma2000, WiMAX, WLAN, Bluetooth, etc. We proposed a Handoff Management System for Multi-Networks (HMSMN) that offers seamless vertical handoffs to end-users across heterogeneous wireless networks without modifying or extending the network side. With the benefit of only modifying the client side, HMSMN is more simple, scalable, and cost-effective than modifying or extending the network side. We described the design and implementation of HMSMN and measured its handoff performance. HMSMN is designed to provide transparent services to its upper layers, IP/MIP layers and manipulates different types of wireless network interfaces in a single virtual network interface. Besides the basic integration approach, we proposed the tight integration approach where the operations of HMSMN and MIP are closely coupled. To analyze the HMSMN handoff performance, we presented an experimental inter-network mobility between cdma2000 1xEV-DO and WLAN hot-spots. The experimental results show how HMSMN provides seamless mobility during vertical handoffs. In our experiences, HMSMN can mitigate MN’s traffic flow with least disruptions during vertical handoffs across heterogeneous wireless networks. ACKNOWLEDGMENTS This research was supported by the MIC(Ministry of Information and Communication), Korea, under the ITRC(Information Technology Research Center) support program supervised by the IITA(Institute for Information Technology Advancement)” (IITA-2008-C1090-0801-0045) and supported by the Korea Science and Engineering Foundation(KOSEF) grant funded by the Korea government(MOST) (No. R01-2007-000-20154-0).

[4] Mark Stemm and Randy H. Katz, ”Vertical Handoff in Wireless Overlay Networks,” ACM Mobile Networking and Application (MONET), vol. 3, no. 4, pp. 335-350, 1998. [5] C. Perkins, ”IP Mobility Support,” RFC 2002, IETF, October 1996. [6] Bjorn Andersoon, Jari Malinen, Dan Forsberg, Hannu Kari, and Jari Hautio, ”Dynamics Mobile IP,” available from http://dynamics.sourceforge.net/?paget=main. [7] Milind M. Buddhikot, Girish Chandranmenon, Seungjae Han, Yui-Wah Lee, Scott Miller, and Luca Salgarelli, ”Design and Implementation of a WLAN/CDMA2000 Interworking Architecture,” IEEE Wireless Communications, vol. 41, no. 11, pp. 90-100, November 2003. [8] IEEE Std. 802.11-1999, ”IEEE Standard for Local and Metropolitan Area Networks, Part 11,” 1999. [9] F. Siddiqui, S. Zeadally, and E. Yaprak, ”Design Architectures for 3G and IEEE 802.11 WLAN Integration,” Lecture Notes in Computer Science (LNCS), vol. 3421, pp. 1047-1054, Springer, April 2005. [10] Srikant Sharma, Inho Baek, Yuvrajsinh Dodia, and Tzi-cker Chiueh, ”OmniCon: A Mobile IP-based Vertical Handoff System forWireless LAN and GPRS Links,” in Proceedings of IEEE International Conference on Parallel Processing Workshops (ICPPW), August 2004. [11] Ling-Jyh Chen, Tony Sun, Guang Yang, and Mario Gerla, ”USHA: a simple and practical seamless vertical handoff solution,” in Proceedings of IEEE Consumer Communications and Networking Conference (CCNC), January 2006. [12] ”TIA/EIA/IS-835B cdma2000 Wireless IP Network Standard,” 3GPP2, 2000. [13] Rajiv Chakravorty, Pablo Vidales, Kavitha Subramanian, Ian Pratt, and Jon Crowcroft, ”Performance Issues with Vertical Handovers Experiences from GPRS Cellular and WLAN Hot-spots Integration,” in Proceedings of IEEE Conference on Pervasive Computing and Communications (PERCOM), March 2004. [14] M. Ylianttila, M. Pande, J. Makela, and P. Mahonen, ”Optimization Scheme for Mobile Users Performing Vertical Handoffs between IEEE 802.11 and GPRS/EDGE networks,” in Proceedings of IEEE Global Telecommunications Conference (GLOBECOM), November 2001. [15] R. Caceres and L. Iftode, ”Improving the Performance of Reliable Transport Protocols in Mobile Computing Environments,” IEEE Journal on Selected Areas in Communications, vol. 13, no. 5, June 1995. [16] IEEE Std. 802.16-2001, ”IEEE Standard for Local and Metropolitan Area Networks, Part 16,” 2001. [17] Suk Yu Hui and Kai Hau Yeung, ”Challenges in the Migration to 4G Mobile Systems,” IEEE Communications Magazine, vol. 41, no. 12, pp. 54-59, December 2003. [18] J. Rosenberg, H. Schulzrinne, G. Camarillo, A. Johnston, J. Peterson, R. Sparks, M. Handley, and E. Schooler, ”SIP: Session Initiation Protocol,” RFC 3261, IETF, June 2002. [19] D. Johnson, C. Perkins, and J. Arkko, ”Mobility Support in IPv6,” RFC 3775, IETF, June 2004. [20] A. C. Snoren and H. Balakrishnan, ”An end-to-end approach to host mobility,” in Proceedings of ACM MobiCom, August 2000. [21] L. Ma, F. Yu, and V. C. M. Leung, ”A New Method to Support UMTS/WLAN Vertical Handover Using SCTP”, IEEE Wireless Communications, vol. 11, no. 4, pp. 44-51, August 2004.

R EFERENCES [1] Ian Akyildiz, Yucel Altunbasak, Faramarz Fekri, and Raghupathy Sivakumar, ”AdaptNet: An Adaptive Protocol Suite for the NextGeneration Wireless Internet,” IEEE Communications Magazine, vol. 42, no. 3, pp. 128-136, March 2004. [2] M. Buddhikot, G. Chandranmenon, S. Han, Y. W. Lee, S. Miller, and L. Salgarelli, ”Integration of 802.11 and Third-Generation Wireless Data Networks,” in Proceedings of IEEE Conference on Computer Communications (INFOCOM), April 2003. [3] A. Salkintzis, C. Fors, and R. Pazhyannur, ”WLAN-GPRS Integration for Next-generation Mobile Data Networks,” IEEE Wireless Communications, vol. 9, no. 5, pp. 112-124, October 2002.

426