一個適用於行動計算之高效率行動代理者系統

Abstract

行動計算的目標是無論在何時或何地都能提供使用者無所不在的資訊服務。行動代理者是一種新的分散式計算的方法，它具有一些適合行動計算環境的特性。然而它們最初是針對傳統的有線網路所設計。本論文探討行動計算環境和有線計算環境之差異，以及前者所面臨的困難，並且提出一個新的且適用於行動環境下的行動代理者系統架構。無線網路的傳輸代價相較於有線網路的傳輸是較昂貴的，因此我們設計的主要構想是，在一個行動代理者由一個基地台進入一個行動主機前，在基地台上先複製一份此行動代理者。藉由基地台的幫助，我們可以利用有線網路的傳輸來減少無線網路所需的傳輸。由於不僅行動代理者，行動主機本身也會改變位置，所以我們提出一個有效率的行動代理者遷徙機制，來解決行動計算環境下的換手問題。除此之外，藉由基地台上的行動代理者複本，我們提出一個有效的回復機制來增加行動代理者系統的可靠性。實驗結果顯示，我們的設計顯著地降低行動代理者系統的遷徙執行時間及無線網路負載。對為了減少能源消耗而需經常中斷的行動主機而言，我們的系統會更為適用。在大部分的情況下，我們的方法可以比其他方法減少18.3% - 39.8%的遷徙執行時間和14.7% - 34.8%的無線網路負載。雖然我們的方法需要額外的儲存空間，藉有效的回復機制，我們可以讓系統更為可靠。

The goal of mobile computing is to provide ubiquitous information services to users at anytime and anywhere. Mobile agents are a new paradigm for distributed computing and have some features which are suitable for a mobile computing environment. However, they are originally designed for conventional wired networks. This thesis discusses the differences and difficulties in a mobile computing environment by comparing it with a wired computing environment and proposes a new mobile agents system architecture which is feasible in a mobile computing environment. Note that the cost of wireless transmission is more expensive than that of wired transmission. The key idea of our design is to duplicate an agent in a base station before the agent jumps from the base station into a mobile host. With the aid of the base station, wired transmission is used to reduce wireless transmission. Since not only mobile agents but also mobile hosts will change their locations, we present an efficient migration mechanism for mobile agents to resolve the handoff problem in a mobile computing environment. In addition, with the help of the duplicated agent in the base station, an efficient agents recovery mechanism is proposed to make our agents system more reliable. Experimental results show that our design significantly reduces the migration execution time and wireless network load in a mobile agents system and is especially suitable for a mobile host that is frequently disconnected for saving power. In comparison with other approaches, in most cases, our approach can reduce 18.3% - 39.8% and 14.7% - 34.8% migration execution time and wireless network load, respectively. Although our scheme needs additional storage, it makes a mobile agents system more reliable through an efficient agent recovery.