Characteristics, impact, and tolerance of partial disk failures
Characteristics, impact, and tolerance of partial disk failures
复制标题
部分磁盘故障的特征、影响和容忍度
DOI：
--
复制
发表时间：
2008
期刊：
影响因子：
--
通讯作者：
Lakshmi N. Bairavasundaram
中科院分区：
文献类型：
--
作者：
Andrea C. Arpaci;Remzi H. Arpaci;Lakshmi N. Bairavasundaram
硬盘故障是数据丢失的主要原因之一。大多数磁盘故障是部分故障，其中只有一些部门是不可用的，由于潜在的扇区错误或一些块是无声损坏。本论文的重点是这样的部分磁盘故障的各个方面-他们的特点，他们对不同的系统的影响，以及可以用来容忍他们的技术。
我们进行了第一次大规模的部分磁盘故障的研究，涉及153万个磁盘。我们发现，部分磁盘故障会影响很大比例的磁盘。我们还发现，（i）SATA驱动器有更高的概率开发部分磁盘故障和（ii）故障是不独立的，故障在同一个磁盘具有较高的空间和时间的局部性。
我们研究了部分磁盘故障对各种系统的影响。我们使用模型检查来检查RAID系统中的数据保护，发现大多数方案不能防止一个或多个故障，导致数据丢失。我们应用类型感知的故障注入来检查部分磁盘故障对Linux、FreeBSD和Windows XP的虚拟内存系统的影响。我们发现，这些系统使用简单或不一致的故障处理政策。我们分析了损坏的磁盘指针对文件系统NTFS和ext 3的影响。我们发现，这些系统不使用容错技术有效地，导致数据丢失。
我们已经构建了一个N版本的文件系统来容忍部分磁盘故障。该系统将数据存储在N个不同的文件系统中，从而消除了对单个复杂文件系统的依赖。我们的系统使用现有的文件系统，如ext 3和JFS，从而避免了构建不同版本的开发成本。我们的实验表明，N版本文件系统显着降低了数据丢失的可能性。
Hard-disk failures are one of the primary causes of data loss. Most disk failures are partial failures, where only some sectors are unavailable due to a latent sector error or some blocks are silently corrupted. This dissertation focuses on all aspects of such partial disk failures—their characteristics, their impact on different systems, and techniques that can be used tolerate them.
We perform the first large-scale study of partial disk failures, involving 1.53 million disks. We find that partial disk failures affect a large percentage of disks. We also find that (i) SATA drives have a higher probability of developing partial disk failures and (ii) failures are not independent; failures within the same disk have high spatial and temporal locality.
We examine the impact of partial disk failures on a variety of systems. We use model checking to examine data protection in RAID systems, and find that most schemes do not protect against one or more failures, leading to data loss. We apply type-aware fault injection to examine the impact of partial disk failures on the virtual-memory systems of Linux, FreeBSD, and Windows XP. We find that these systems use simplistic or inconsistent failure-handling policies. We analyze the impact of corrupt on-disk pointers on file systems NTFS and ext3. We find that these systems do not use fault-tolerance techniques effectively, resulting in data loss.
We have built an N-version file system to tolerate partial disk failures. This system stores data in N different file systems, thereby eliminating the reliance on a single complex file system. Our system uses existing file systems, such as ext3 and JFS, thus avoiding the development costs of building different versions. Our experiments show that an N-version file system significantly reduces the probability of data loss.