Engineering fault tolerance

基于计算机的系统和控制系统的范围和复杂性继续急剧增加。现代软件必须满足越来越多的需求，提供实时功能，并且必须在异类和不断变化的环境中与其他软件系统交互。越来越多的提供关键功能的责任由软件承担，因此，这种系统以不可预测的方式运行的后果可能是相当严重的。为了成功地构建这种可靠的“系统系统”，应用程序必须以这样一种方式设计，即它们可以在出现故障的情况下提供服务。他们必须能够容忍环境中外部实体的意外行为、不可靠的底层组件、通信故障、错误数据和错误的算法。这个持续的发现授权项目的目标是通过遵循模型驱动的方法将容错问题集成到整个软件开发周期来解决这种情况。容错感知过程的总体好处是可观的。从一开始就牢记容错使软件开发人员能够“设计”对容错的支持。它可以及早发现容错需求，尽早从客户那里获得有价值的反馈，做出明智的决策或权衡，并在可靠性和安全性之间设置优先级，准确地说明系统不同部分所需的容错级别，然后选择适当的模型、设计和基础设施来实现它。良好定义和指导的过程有助于降低容错的复杂性，在需要的地方更有效地使用最合适的容错模型，因此所产生的软件体系结构、设计和代码清晰可读，易于维护并提供足够的性能。

The scope and complexity of computer-based and controlled systems continue to increase dramatically. Modern software must satisfy an increasing number of requirements, provide real-time functionality, and has to interact with other software systems in a heterogeneous and changing environment. More and more responsibility of providing critical functionality is assumed by software, and hence the consequences of such systems behaving in an unpredictable way can be considerable. To successfully build such dependable "systems of systems", applications have to be engineered in such a way that they can provide service in spite of faults. They must be able to tolerate unexpected behavior of external entities in their environment, unreliable underlying components, communication failures, erroneous data and faulty algorithms.The goal of this continuing discovery grant project is to address this situation by integrating the concern of fault tolerance into the entire software development cycle following a model-driven approach. The overall benefits of a fault tolerance-aware process are considerable. Having fault tolerance in mind from the beginning allows software developers to "engineer" support for fault tolerance. It makes it possible to discover the need for fault tolerance early on, get valuable feedback from the client as early as possible, make informed decisions or tradeoffs and set priorities between reliability and safety, state the required levels of fault tolerance for different parts of the system precisely, and then choose the appropriate models, design and infrastructure to achieve it. A well-defined and guided process helps reducing the complexity of fault tolerance, results in more efficient use of the most appropriate fault tolerance models where they are needed, and hence the resulting software architecture, design and code are clear and readable, easy to maintain and deliver adequate performance.