主從式容錯架構之檔案系統備份設計

Design of File Backup for Client-Server Fault Tolerant System

鄭武昇(Wu-Sheng Jeng)

指導教授 ： 丁鏞

中原大學/工學院/機械工程研究所/碩士(2002年)

https://doi.org/10.6840/cycu200200616

摘要

摘 要
本研究提出在主從式架構下利用檔案備份來提高系統可靠度的容錯備份模式，以期解決在主從式檔案系統中伺服端因故障或當機造成服務的中斷以及資料的流失之問題。然後建立系統之馬可夫模型進行可靠度分析。該容錯模式包含了模組重覆(modular redundancy)、主/備(primary-backup)模式、及雙網路卡三種設計以滿足容錯的需求。為了達到以檔案目錄階層為導向的複製，設計判斷式的同步資料備份結合虛擬目錄與應用於代理伺服器的mirror技術進行主/備伺服端之間檔案樹的複製。最後於實際所架構的容錯備份系統中進行模擬，結果利用判斷式的同步備份可以有效的減少伺服端因備份工作所佔用的系統資源，提高備份的效率。另外可以預防因主伺服端發生錯誤，備份伺服端取代主伺服端繼續提供服務給工作端，並且保護檔案系統資料的完整性。

關鍵字

備份 ； 馬可夫 ； 虛擬目錄 ； 容錯

並列摘要

Abstract
The research makes propose a fault tolerant model for backup in client-server system, to solve the problem that the server fault cause file and data lost of file system in client-server system. Then the reliability analysis of fault tolerant system for backup is established with the Markov Chain method. To fulfill the fault tolerant requirements, the fault tolerant model includes modular redundancy, primary-backup, and double ethernet adapters. In order to reach the idea of directory-oriented backup, combine judicially synchronous backup, virtual directory, and the technique of mirror that always used in proxy server to backup the file tree between primary server and backup server. Finally, the ststem reliability and performance are analyzed and evaluated via simulations to verify that judicially synchronous backup can effectively reduce the use of system source and improve the efficiency of backup. Besides, backup for fault tolerant system can prevent as the primary server fault that the backup server can displace the service to clients and preserve the integrity of file system.

並列關鍵字

fault tolerance ； backup ； Markov Chain ； virtual directory

參考文獻

[1] G. Popek and B. Walker. “The LOCUS Distributed System Architecture” MIT Press, Cambridge Mass. 1985.

[2] M. Satyanarayanan , J.J. Kistler, P. Kumar, M.E. Okasaki, E.H. Siegel, and D.C. Steere. Coda “A Highly Available File System for a Distributed Workstation Environment” IEEE Transaction on computer, Vol. 39 No.4, April 1990.

[3] B. Liskov, R. S. Gruber, P. Johnson, L. Shrira and M.Williams. “Replicated Unix File System” ACM Operating System Review, Vol. 25, No. 1, Jan. 1991, pp. 60-64.

[5]A, Bhide, E.N. Elnozahy, S.P. Morgan and A. Siegel. A. “A Comparison of Two Approaches to Build Reliable Distributed File Servers” Proc. 11th International Conference on Distributed Computing System IEEE 1991, pp. 616-623.

[11]Peterson, J.L.(1981). Petri Net theory and the modeling of systems. Englewood Cliffs:NJ: Pretice-Hall.