近年來隨著行動網路的蓬勃發展，行動IP已成為在全球網路提供終端裝置移動能力的重要協定之一。在行動IP協定中，使用本籍代理人來為漫遊中的行動裝置負責封包的轉送與位置資訊的記錄，然而當有愈來愈多的行動裝置在網路上，這些本籍代理人可能會成為網路的瓶頸，造成封包傳送的延遲。因此，一個以多個本籍代理人之間的動態負載平衡來降低封包傳送延遲的設計被提出。
　　本論文將多個本籍代理人的負載平衡考量分為兩階段，對初始註冊階段，我們考慮有仲裁伺服器於本地網域的多個本籍代理人架構，透過仲裁伺服器，我們可以依本籍代理人負載平衡的考量，來動態選擇一個行動IP的本籍代理人給行動主機的使用者，使得每個行動裝置可以適當地被分配給本籍代理人，
　　對註冊後階段，我們以雙重閥值動態負載平衡策略為出發點，衍伸並提出新的動態負載平衡策略考量，以增加其彈性和效能。首先，我們用單一閥值取代雙閥值,然後再將佇列分割為多個區域，每個佇列分割區代表不同的負載狀態,當本籍代理人的負載狀態改變時，它會廣播更新訊息通知其他本籍代理人，而佇列分割越多則訊息的精確度越高，我們想瞭解不同佇列分割數對整體系統效能的影響。第二，當本籍代理人負載轉移發生時，我們把行動裝置的選擇納入考量，在此我們提出四種行動裝置選擇方法---隨機選擇、最大傳輸量優先、最多佇列封包優先及最多佇列封包加上較長停駐時間考量，藉以找出最佳的方案以降低在本籍代理人下封包傳送的延遲。
　　最後我們根據所述模型發展了一個模擬程式，藉以比較不同策略與設定下的效能，首先第一個實驗發現即使忽略處理表格更新封包的負載,佇列分割數目增進效能的效果並不明顯，對行動裝置選擇方法而言，實驗結果呈現最多佇列封包加上較長停駐時間考量策略獲得最小的封包延遲。最後一個模擬實驗結果則顯示加入仲裁伺服器可以幫助負載平衡，且可以降低轉換的次數。

　　In recent year, the mobile and wireless network is expanding continuously, Mobile IP has become an important protocol for providing Internet connectivity to roaming mobile hosts. In Mobile IP standard, location and packet forwarding functions are provided by servers referred to as home agents. When there are more and more mobile hosts in the network, these home agents may become the bottleneck and cause the packet delivery latency. Therefore, a multiple home agents extension of the Mobile IP had been proposed to reduce the delay across the home agents by dynamic load balancing policy.
　　In this thesis, we separate the load balancing considerations of multiple home agents into two stages, the initial registration and after the initial registration. For the initial registration stage, we consider the multiple home agents environment with an arbiter server. Under this environment, the mobile nodes can dynamically select a home agent by load balancing reasons. We utilize a server as arbiter and propose the mechanisms to fairly distribute the mobile nodes to each home agent.
　　For the second stage, we use the idea of the double threshold dynamic load balancing scheme as a starting point, and then we extend and adapt its concept to make it more flexible and efficient. In the first place, we use single threshold instead of double thresholds, then we divide the area below the threshold into a number of queue partitions. Each partition stands for a queue state. A home agent advertises the update information when its queue state changes. The more queue partitions, the information will be more accurate. We would like to study the impact when using different number of queue partitions. In the second place, when a transfer takes place, we take account of the mobile node selection. Here we propose four mobile node selection policies, namely, random selection, most packets received first, most packets in queue first and most packets in queue with longer resident time to find the best policy for reducing the delay across the home agents.
　　In the end, we develop a simulator by which we studied the performance under various policies and configurations. For queue partitions, even though we have ignored the overhead of table update message, the benefit is not clear when we use the larger number of queue partitions. For mobile node selection policy, the result shows that the most packets in queue with longer resident time first is the best policy to reduce the packet delivery latency. In our final experiment, we found that the environment adopting our mechanism can balance the load in advance and reduce the transfer times.

[1] J. Jue and D. Ghosal. “Design and Analysis of a Replicated Server Architecture for Supporting IP-Host Mobility”. ACM Mobile Computing and Communications Revue,2(3):16-23, 1998
[2] A. Vasilache, J. Li and H. Kameda. “Load balancing Policies for Multiple Home Agents Mobile IP Networks”. IEEE Proceedings of the Second International Conference on Web Information Systems Engineering, Volume: 2 , 3-6 Dec. 2001.
[3] C. Perkins. “IP Mobility Support for IPv4”. RFC3220, January 2002
[4] C. Perkins and D. Johnson. “Route Optimization in Mobile IP”. Internet draft, draft-ietf-mobileip-optim-08.txt, Feburary 1999.
[5] C. Perkins. “Mobile IP – Design Principles and Practices”. Addison-Wesley, 1998.
[6] P. Calhoun and C. Perkins, “Mobile IP Network Access Identifier Extension for IPv4”, RFC 2794, March 2000
[7] M. Kulkarni, Alpesh Patel, and Kent Leung. “Mobile IPv4 Dynamic Home Agent Assignment”. Internet draft, draft-ietf-mip4-dynamic-assignment-01.txt, April 2004
[8] H. Deng, Rong Zhang, and K.Zhang. “Load Balance for Distributed Home Agents in Mobile IPv6”. Internet draft, draft-deng-mip6-ha-loadbalance-00.txt, November 2003.
[9] C. Perkins. “Mobile IP Joins Forces with AAA”. IEEE Personal Communication, August 2000.
[10] Jay L. Devore. “Probability and Statistics for Engineering and the Sciences”, Fourth edition, 1995.
[11] R. Ghosh and G. Varghese. “Fault Tolerant Mobile IP”. Technical report, 1998.
[12] Christian Huitema. “Routing in the Internet”, Second edition, 1999.
[13] W. Stevens. “TCP/IP Illustrated, Volume 1: The Protocols”, Addison-Wesley, Reading, Massachussetts, 1994
[14] W. Fisher and K. Meier-Hellstern. “The Markov-modulated Poisson process (MMPP) cookbook”. Performace Evaluation 18, 1992.
[15] D. McDysan. “QoS & Traffic Management in IP & ATM Networks”, McGraw-Hill, 2000
[16] A. Campbell, J. Gomez, S. Kim and Chieh-Yih Wan, “Comparison of IP Micromobility Protocols”, IEEE Wireless Communications, Feburary, 2002
[17] A. Valko, “Cellular IP: A New Approach to Internet Host Mobility”, ACM SIGCOMM Comp.Commun. Rev., vol. 29, no.1, January 1999, pp.50-65
[18] R. Remjee et al., “HAWAII: A Domain-Based Approach for Supporting Mobility in Wide-area Wireless Network”, Proc. IEEE Int’l. Conf. Network Protocols, 1999