节点文献

基于存储区域网络的数据可靠性技术研究

【作者】 张晓

【导师】 李战怀;

【作者基本信息】 西北工业大学 , 计算机软件与理论, 2006, 博士

【摘要】 全球经济正经历着一场深刻的大变革,这场变革的关键就在于信息的应用。信息正成为最具有活力、最重要的生产要素之一。信息量的激增同时也增加了信息资源管理的难度,给信息的存储和处理带来了巨大的挑战,。存储管理技术的研究得到了普遍的关注,并逐渐成为研究热点。存储硬件技术的发展,使海量数据存储成为可能,也为存储管理技术的发展提供了机遇。借助于存储区域网络(Storage Area Network,SAN),人均可管理数据量大大超过直接相连存储系统(Dircct Attach Storage,DAS)。存储区域网络的出现,改变了存储系统的管理方式,数据由分散存储到集中存储。在存储能力大幅度提高的同时,数据管理的工作量和复杂度也不断增加,而原有存储管理技术制约着人均可管理数据量的进一步提高。多项调查表明,存储管理技术是限制存储容量进一步发展的瓶颈。存储管理一方面要提高存储资源的利用效率,另一方面要提高人的管理效率,需要将管理员从繁杂的重复劳动中解放出来,提高存储管理软件的智能程度,使其成为存储管理工作的主体。存储系统的自动化管理的研究目标就是提高存储管理系统的智能程度,减少人为干预,降低存储管理工作的难度和强度。 另一方面,数据激增带给企业巨大财富的同时,数据的价值已经远远超过了计算机系统本身的价值,意外的停机或数据损坏会给企业带来巨大的损失。数据灾难造成的经济损失远远大于计算机系统本身故障造成的损失。从911事件之类的恐怖袭击,到停电、硬盘故障等意外事件都会导致数据丢失。为了避免数据损失,已经实现了各种不同的高可靠性技术,包括冗余磁盘阵列(Redundent Array of Incxpcnsivc Disks,RAID),复制技术等。用户在选择存储系统时,可靠性也正日益成为一个重要的指标。但是目前的数据高可靠性研究还存在以下几个问题:(1)缺乏数据可靠性的准确定义。目前研究存储系统可靠性主要着眼于存储系统中的磁盘阵列或存储子系统的可靠性,而不是其上保存数据的可靠性。海量数据存储管理中,同一磁盘阵列上的数据可能具有不同的可靠性要求,其次,使用不同的备份周期和参数,不同的数据也确实具有不同的可靠性。(2)缺乏统一的可靠性度量标准,尤其是衡量组合容灾方案的方法。各个厂商为用户提供了多种多样的备份技术,并提供了这些方案的设置和使用方法,但是如何评价组合方案的效果仍存在巨大的空白。不同方案涉及到的技术和参数各不相同,也和用户的需求表示方式存在有巨大的差异,急需适用于用户和业界的统一的可靠性度量标准。(3)缺乏智能化的管理系统。备份和恢复是数据管理员最繁重的工作,而且单调乏味,适合计算机来进行管理。定义好恢复规则后,在监测到故障发生后,自治系统自动恢复数据可以加快数据恢复速度。

【Abstract】 A revolution is changing the economic of the earth. The key factor of revolution is knowledge. Information is one of the most important factors of producers’ goods. Rapid increasing information amount bring huge challenge to information handler and storage. The research on storage management gains a lot of attention and becomes a hotspot in storage research field. The development of storage hardware brings great opportunity to storage management technology. People can manage more data by storage area network than before. SAN centralize the management of multiple devices to one console, but it will never exceed the limit of one’s ability. Several surveys indicate that the storage management is the bottleneck of storage system development. The storage management should not only promote the value of storage device, but also promote the value of storage manager. The storage manage software should do most regular and reputably work to decrease the difficulty of storage management.With the popularity of data intensive applications, the value of data is greater than the value of computer system itself. Unexpected halt down or data lost will bring huge damage to enterprise. Terrorism attack as 911, power failure and hard disk failure will cause data lost. To avoid data lost, many technologies were developed, such as tape backup, remote mirror and snapshot. Dependability is coming to a important factor when user choose storage system. But the research of storage dependability remains following problems: (1) the dependability request of different data in one disk array is different, but people always concern the dependability of disk array, which should be dependability of data on it. Though different backup method, we can provide different dependability of data, but we didn’t compute the dependability of data before. (2) Over the last decade, the set of data protection techniques has increased significantly. In addition to traditional tape-based backup, online techniques using high-density disks and incremental snapshots are becoming attractive. F.ach technique provides some portion of the protection that is needed; combined, they can cover a much broader range. The key question, then, is how to determine the appropriate combination of different techniques to provide the data protection desired by the user. The complex technologies and the parameter make it different to define a proper solution suit the user’s request. It needs a standard can be use to define user request and compare different solution. (3) Backup is not the

节点文献中: