CSR---PDOS: Support for Atomic Sequences of File System Operations

存储在计算机上的数据是最宝贵、不可替代的组成部分。文件系统和数据库是提供受监管的数据访问的两种最常见方式。 数据库提供事务，允许原子地应用任意操作序列;但是数据库使用各种不兼容的API。 然而，文件系统为数据提供了标准的POSIX API，但由于它们不提供原子地应用一系列操作的设施，因此不太可靠。该项目将完整的数据库（ACID）属性集成到商品操作系统中，同时为用户级应用程序提供具有事务支持的POSIX兼容API。 我们正在将Berkeley数据库（BDB）移植到商品内核--一个高度可移植的，高效的，通用的嵌入式数据库。 除了集成的内核支持，我们还开发了两个ACID语义的文件系统。我们正在开发几个高度实用的应用程序：一个溯源感知存储系统（PASS），跟踪文件所有权的起源和历史;一个事务感知包装器共享库，为未修改的遗留应用程序提供事务功能;和一个邮件传递服务器，原子地更新邮箱，访问控制列表，和更多。这个项目的最终目标是开发和评估一个新的操作系统体系结构，提供事务所有的方式给用户应用程序，即使是未经修改的应用程序;由此产生的应用程序可靠性和安全性的增加将有助于社会对软件的不断增长的使用。 这将使开发人员能够更快地编写更安全的软件。长期的影响将是整个社会将受益于更可靠、更安全和更强大的软件。

Data stored on computers is the most precious, irreplaceable component.File systems and databases are the two most common ways to provide regulated access to the data. Databases provide transactions, which allow arbitrary sequences of operations to be applied atomically; but databases use a wide variety of incompatible APIs. File systems, however, provide a standard POSIX API to data, but are less reliable because they do not provide facilities to apply a sequence of operations atomically.This project integrates full database (ACID) properties into commodityoperating systems, while offering user-level applications a POSIX-compliant API with transactions support. We are porting the Berkeley Database (BDB) to commodity kernels---a highly portable, efficient, and versatile embeddable database. Aside from integrated kernel support, we are also developing two file systems with ACID semantics.We are developing several highly practical applications: a Provenance-Aware Storage System (PASS) that tracks the origin and history of file ownership; a transaction-aware wrapper shared library that provides transaction capabilities to unmodified legacy applications; and a mail-delivery server that atomically updates mailboxes, access control lists, and more.The end goal of this project is to develop and evaluate a new operatingsystem architecture that offers transactions all the way to userapplications, even unmodified applications; the resulting increase inapplication reliability and security will help society's ever-growing use of software. This will allow developers to write safer software more rapidly. The long term impact will be that society as a whole will benefit from more reliable, secure, and robust software.