HOPOVER: A New Handoff Protocol for Overlay Networks Fan Du, Lionel M. Ni and Abdol-Hossein Esfahanian Department of Computer Science and Engineering Michigan State University East Lansing, MI 48824 Abstract-This paper presents a new handoff protocol, Handoff Protocol for Overlay Networks (HOPOVER). This protocol is compatible with Mobile IP and is designed specifically for overlay networks where handoffs happen both horizontally and vertically. Handoff performance is enhanced by a number of measurements including pre-reserving resources, packet buffering in the new network and packet forwarding from the old network to the new network. Our simulation proved the effectiveness of these measurements. 1 INTRODUCTION Handoff is the process in which a wireless device moves from current cell to a new one. The performance of handoff scheme directly affects the overall performance of mobile applications. The importance of handoff performance keeps going up as today’ s wireless networks have increasingly large user population and decreasing cell size. Traditionally, the term handoff is used to refer to the case in which the two cells involved belong to the same wireless network, as shown in the left side of Figure 1. Such handoffs are referred as horizontal. As number of wireless networks exponentially increases, they are overlaid with each other to form hierarchies. In such structures, mobile devices can perform handoffs among different layers of networks. Such handoffs are referred as vertical, as shown in the right side of Figure 1. Figure 1. Horizontal and Vertical Handoff A lot of research work has been done to improve the performance of horizontal handoffs. Examples include [4][6][7][8][9][10][11][12]. However, much less effort has been spent on vertical handoff research [2][5][11][15]. Up to now, most networks still support vertical handoff through Mobil IP [11], which was designed to provide macro-level and slow moving mobility. The scheme suffers from high overhead when high frequency handoffs are needed. More importantly, handoffs cause big gaps in packet flows. Besides Mobile IP, there are only few other approaches and none of them got extensive use due to their limitations and shortcomings. To address this problem, we designed a new overlay handoff protocol named HOPOVER (HandOff Protocol for OVERlay networks), which supports both horizontal and vertical handoffs and is compatible with Mobile IP. The main approaches used in this protocol include: facilitating a special wireless resource reservation protocol to pre-reserve resources in the new cell and along the path from the new cell to the flow transmission parties; buffering in the new network for the MH and forwarding packets from the old network to the new network. With these measurements, HOPOVER significantly enhances handoff performance. The remaining part of this paper is organized as follows. First, related work is reviewed in Section 2. Then we present the design of HOPOVER in Section 3. Section 4 presents the simulation results. Section 5 discusses issues and attributes of HOPOVER. Finally, conclusion is given in Section 6. 2 RELATED WORK In this section, we go through some of the most representative research works in the area of handoff. 2.1 Mobile IP Traditionally, handoff schemes across networks are constructed based on Mobile IP [11]. Mobile IP provides a simple and efficient solution to maintain IP connectivity for mobile hosts. It is a proposed standard protocol by IETF, and it is widely used and supported by current systems. Mobile IP handles mobility as follows: • Every Mobile Host (MH) has a Home Agent (HA), which knows the MH’ s permanent IP address. Every site that wants to allow visitors creates a Foreign Agent (FA). • When a MN moves to a foreign network, it registers with the FA and gets assigned a Care-Of-Address (COA). • The FA notifies the HA of the MH’s COA. The FA also starts forwarding packets for the MH. • When packets are sent to the MH’ s home network, the HA forwards them to the current COA. At the same time, the HA notifies the sender about the new location. • The sender then sends following packets directly to the new location. Mobile IP was designed to support macro level mobility and slow moving hosts. It requires that the HA be notified each time the MH receives a new COA, which happens when the MH moves into a foreign network or performs a handoff between cells. Such requirement is OK when MHs move at low speed and handoffs only happen at low frequency. 32340-7803-7400-2/02/$17.00 © 2002 IEEEHowever, as wireless communication evolves, the number of mobile devices increases rapidly, and the use of high-speed moving mobile devices becomes more and more popular. At the same time, to accommodate the increasingly large number of devices, cells have to be designed smaller, thus causing even more frequent handoffs for mobile devices. All of these made the design principle behind Mobile IP no longer suitable and the overhead associated with handoff too expensive. 2.2 Cellular IP To provide a low-overhead handoff scheme, Cellular IP is proposed. [5] Cellular IP adopts a hierarchical approach to manage mobility. At the higher level, Mobile IP will still be used, and Cellular IP is used to handle lower level mobility. When a MH arrives a Cellular IP network network the first time, traditional Mobile IP operations are performed to inform the HA. Now all the packets for the MH will be routed to the gateway router then to the MH. Later, when the MH moves within the wireless network and performs handoffs, it only causes the gateway router to change the base station to forward to. No operation with HA is necessary. Cellular IP reduced the overhead associated with handoffs and made handoff process smoother. However, it is only applicable for horizontals handoffs where the MH moves within the same network. For handoffs between two networks, MHs still go back to Mobile IP. 2.3 BARWAN The Bay Area Research Wireless Access Network (BARWAN) [3] also proposed a vertical handoff scheme. This scheme is based on Mobile IP, but with modifications to make it fast. Basically, BARWAN improves handoff performance by converting packet delivery mode from 1-1 to 1-n. For each MH, packets are multicast to multiple BSs. BARWAN requires one (local) multicast address for each MH. This requirement makes it unrealistic in real life. This approach is possible only if all the
View Full Document
Unlocking...