國立臺灣大學電機工程學系暨研究所郭斯彥2006-07-252018-07-062006-07-252018-07-061999-07-31http://ntur.lib.ntu.edu.tw//handle/246246/7690行動計算(Mobile Computing) 是 目前一個熱門研究領域。它是由一群 固定處理機(fixed hosts) 和行動處理 機(mobile hosts) 所形成的一個分散 式處理系統。在這種系統架構下,固 定處理機(fixed hosts) 彼此間是由一個 靜態網路(static network) 連接起來。 而行動處理機和其它行動處理機間之 通訊必需經由固定處理機。一個固定 處理機連接多個行動處理機,形成一 個無線通訊網路(wireless network)。 行動計算系統主要的特性是: 使用者 可在任何地點、任何時間存取所需資 訊。 先前已有許多文獻在探討分散式系 統容錯問題(fault-tolerant problem),在 他們所提出的容錯方法,並未對錯誤如 何被偵測加以說明,也不適用於行動區 網路系統。 行動區網路系統具有下列三項特性: (一) 行動處理機具有隨意行動的特 性。 (二) 行動處理機的電能是由一個電池 組(battery pack) 供應,電池組 所能供應的電能時間有一定的限 度。 (三) 行動處理機和固定處理機間無線 網路頻寬,與一般有線網路頻寬比較起 來,有一大段差距。無線通訊也極易受 到干擾。 本整合型計畫之最終目標是在研發一 個行動區域網際網路,並在這種系統架 構之下,提供立即反應資訊服務。但隨 系統裏處理機數目以及在此系統上應用 程式執行時間的增加,這時行動計算系 統裏會發生錯誤(failures) 的機率也隨 之提高。這裏所指錯誤包括固定處理機 和行動處理機內部錯誤(processor failures),以及行動處理機和固定處理 機間無線通訊網路裏通訊錯誤 (communication failures)。錯誤發生後, 系統裏所執行的程式必需重新開始執 行。故一旦錯誤發生,系統無法對使用 者需求(requests) 立即回應。這種情況 違反本整合型計畫目標。在第一年計 畫,我們將一個用於分散式系統的容錯 方法log-based rollback recovery加以修 改後,設計出一個適用於行動資訊系統 的容錯方法。這個容錯方法能容忍固定 處理機和行動處理機的內部錯誤,也考 慮到行動處理機的行動性(mobility)、 電能消耗量(energy consumption)、無線 鏈結低頻寬(low bandwidth of wireless links)。在第二年計畫,我們發展一個新 的錯誤偵測方法,它能偵測出較多錯 誤,並區分錯誤是發生在通訊線路或是 處理機、和降低因錯誤偵測對系統所引 起overhead。我們也利用replication方法 重建因處理機發生錯誤所造成位址資訊 遺失並且探討如何改進行動處理機之間 無線通訊品質。Mobile computing is a hot research area in recent years. It consists of a set of fixed hosts and mobile hosts to form a distributed processing system. In this system architecture, the fixed hosts are connected to each other by a static network. A mobile host communicates with other nodes in the system through a fixed host. The communication between the mobile host and the fixed host is via a wireless network. The main characteristic of the mobile computing system is as follows: The mobile computing system can provide users to access information at anywhere and anytime. A large body of literature concerning the fault tolerance problem exists, but most of them deal with distributed systems rather mobile computing systems. The fault-tolerant approaches for distributed systems can not be directly applied to solve the fault-tolerant problem in mobile computing systems. The reasons are as follows. (1) In mobile computing systems, the location of a mobile host changes with time. (2) A mobile host is usually powered by a stand alone energy source, like a battery pace, that has to be replenished after a certain period of time. (3)Wireless communication is lower bandwidth and higher error rate than wired communication The goal of this integrated project is to design a responsive information service system for mobile environment. However, as the number of hosts in the system and running time of distributed applications increase, the probability of one or more failures also increase. Here, the failures can occur in the fixed nodes and mobile hosts, or on the wireless links between the mobile host and the fixed host. The failures in the mobile computing system will result in the system to restart from beginning. Therefore, if an error occurs, the system can not respond the user request with a predetermined interval. This behavior contradicts the goal of our integrated project. In the first year, we utilized the log-based approach to implement a mobile information system with the capability for tolerating fixed hosts and mobile hosts' failures. In the second year, we proposed a new error detection method to detect more errors and identify the sources of errors correctly. In addition, we used the replication technique to reconstruct the lost location information, and used the dynamic channel allocation technique to improve the quality of wireless communication.application/pdf55909 bytesapplication/pdfzh-TW國立臺灣大學電機工程學系暨研究所行動計算固 定處理機行動處理機無線通訊網路容錯問題Mobile computingfixed nodesmobile nodeswireless networkfault-tolerant problem高可靠度行動計算系統(II)Highly Reliable Mobile Computing Systems(II)reporthttp://ntur.lib.ntu.edu.tw/bitstream/246246/7690/1/882213E002040.pdf