节点文献

分布式文件系统高可用问题研究

Research on High Availability of Distributed File System

【作者】 史小冬

【导师】 孙凝晖;

【作者基本信息】 中国科学院研究生院(计算技术研究所) , 计算机系统结构, 2002, 博士

【摘要】 分布式文件系统高可用问题是分布式文件系统的重要研究课题。目前在分布式文件系统高可用问题的研究上缺乏一种可以进行定量或定性分析的方法,许多系统只是针对某一类的应用需求(如高可用的磁盘访问)来实现,而关于某一类应用的高可用分布式文件系统应实现到什么程度以及如何使用相应的关键技术去实现高可用的分布式文件系统等问题,则缺乏相应的定性分析理论来指导。由于软件系统不能象硬件系统那样对可用度进行定量分析,因此本文试图从定性的角度研究分析分布式文件系统的高可用问题。本文根据分布式文件系统的应用模式,将影响分布式文件系统高可用性的因素进行聚类分析,以分布式文件系统的故障因素和恢复目标因素为线索,对分布式文件系统高可用性进行了定义,并依此建立了一个分布式文件系统高可用问题定性分析模型—DFS_HAL ( Distributed File System High Availability Level )模型。在建立了DFS_HAL模型后,本文将此模型和分布式文件系统的实现技术结合,通过矩阵分析的方法,研究了在不同应用需求下,高可用分布式文件系统的关键实现技术。重点研究了DFS_HAL模型中的DSR_T ( Distributed State Recovery Technology )和CFS_T( Continuous File Service )技术,提出了DSR_T技术中的中转控制策略、满足服务连续性的充分条件和CFS_T技术中的元数据操作探寻请求算法。作为DFS_HAL模型的一个实际应用,本文给出了曙光机群文件系统DCFS高可用系统的设计与实现技术,给出了DCFS HA系统中利用可冗余的内存记录机制来保证分布式文件系统结构一致性的方法。本文的主要贡献如下:㈠根据分布式文件系统的应用模式和分布式文件系统的故障因素和恢复目标因素的聚类分析,首次提出了对分布式文件系统高可用问题进行定性分析的模型DFS_HAL模型。㈡根据DFS_HAL模型,结合分布式文件系统的实现技术,通过矩阵分析的方法,探讨了实现分布式文件系统高可用的关键技术,从而为实现高可用的分布式文件系统起到了指导性的作用。㈢提出了DSR_T技术中的中转控制策略。㈣提出了满足服务连续性的充分条件和CFS_T技术中的元数据操作探寻请求算法。㈤提出了通过可冗余的内存记录机制来保证分布式文件系统结构一致性的方法。

【Abstract】 The High Availability issue is an important research topic in distributed file systems. Till now,we are short of some qualitative and quantitative analysis method on the high availability researchesof distributed file systems, and many systems are implemented only for the requires of some certainapplication (such as the high availability of Disk Access). We are now in serious need ofcorresponding qualitative theories to instruct us to resolve the following problems: In what degree ahigh available distributed file system should be implemented for some kind of application? Howshould we use the relative key techniques to implement a high available distributed file system?Because we cannot do the quantitative analysis for the degree of availability in software systemlike hardware system, this dissertation does some researches and analysis on the distributed filesystem’s high availability issue from a point of qualitative view. Based on the apply mode ofdistributed file system, we did some clustering analysis for the factors which will influence the highavailability issue. Based on the failure and restoring factors, we defined the high availability indistributed file system, and constructed a qualitative analytical model for high availability issue indistributed file system—DFS_HAL ( Distributed File System High Availability Level ) model.After constructing the DFS_HAL model, we integrated it with the implementation techniquesof distributed file system, and by adopting the matrix analytical method, we researched the keyimplementation techniques of high available distributed file system on the conditions of differentapplication requirement. We put emphases on the research of DSR_T ( Distributed State RecoveryTechnology ) and CFS_T ( Continuous File Service ) techniques in DFS_HAL model, put forwardthe transfer-controlling policy in DSR_T, sufficient conditions to satisfy service continuity, and theheuristic requesting algorithm of meta data operation in CFS_T.As a actual application of DFS_HAL model, we present the design and implementationtechniques of the high availability system of Dawning cluster file system-DCFS, describes methodto assure the structural consistency of distributed file system by using the redundant memoryrecording mechanism.The crucial contributions are as follows:1 Based on the apply mode, the failure and restore factors of distributed file system, for the firsttime, we put forward a qualitative analytical model for high availability issue in distributed filesystem—DFS_HAL model.2 Based on the DFS_HAL model, and integrated it with the implementation techniques ofdistributed file system, by adopting the matrix analytical method, we discussed the keyimplementation techniques of high available distributed file system, which works as instructions forimplementation of high available distributed file system.3 Put forward the transfer-controlling policy in DSR_T technique.4 Put forward sufficient conditions to satisfy service continuity and the heuristic requesting

节点文献中: 

本文链接的文献网络图示:

本文的引文网络