Fault-Tolerant Programming

容错的分布式程序是在处理器和网络出现故障时仍能继续执行的程序。这项研究的总体目标是开发技术和支持系统，以促进这类软件的设计、构建和验证。为了实现这一目标，将研究从通信协议到正式验证等领域的许多主题。特别是：操作系统支持容错、分布式编程；便于构建容错分布式系统的编程语言抽象；扩展现有的软件工程技术和编程环境，以包含容错编程；并对容错程序的验证进行了研究。第一个、第二个和第四个主题代表了目前正在进行的调查的继续和这些领域的自然延伸，而第三个主题是一个新的倡议。由于容错很难单独研究，这项工作主要基于现有的系统和语言，包括SR分布式编程语言、Psync进程间通信机制和软件设计师的合作项目。这种方法还通过消除建设必要基础设施的需要，促进了研究问题的直接进展。

A fault-tolerant, distributed program is a program that can continue executing despite processor and network failures. The general goal of this research is to develop techniques and support systems that facilitate the design, construction, and verification of this type of software. To accomplish this goal, a number of topics will be investigated in areas ranging from communication protocols to formal verification. In particular: operating system support for fault-tolerant, distributed programming; programming language abstractions that facilitate construction of fault-tolerant, distributed systems; extension of existing software engineering techniques and programming environments to encompass fault-tolerant programming; and verification of fault-tolerant programs will be studied. The first, second, and fourth topics represent the continuation of investigations currently underway and natural extensions in those areas, while the third is a new initiative. Since fault-tolerance is difficult to study in isolation, this work is based largely on existing systems and languages, including the SR distributed programming language, the Psync interprocess communication mechanism, and the Software Designer's Associate project. This approach also facilitates immediate progress on research issues by eliminating the need to construct necessary infrastructure.