Distributed Systems Development with Multitier Reactive Programming

传统上，分布式应用程序是使用分布式系统中每个组件的单独模块来开发的，通常使用不同的编程语言。这些模块对诸如用户输入或来自网络的消息之类的事件作出反应，并可能进而产生由其他模块处理的新事件。分成几个层通常是有问题的：组合模块是出了名的困难，并且需要广泛且耗时的集成测试。手动实现通信也是一个问题：开发人员被迫在主机之间编写复杂的基于事件的通信方案-这是一项通常级别较低且容易出错的活动。这两者的结合使情况变得更糟：基于事件的分布式数据流分散在多个模块中，这使得很难对整个系统进行推理。由于这些原因，尽管当今大多数软件是分布式的，但分布式系统的设计和开发仍然具有惊人的挑战性。在这个项目中，我们提出了通过专门的编程语言支持来解决开发分布式应用程序的复杂性的路线图。我们的愿景是，基于多层方法的方法可以提供一个连贯的模型来推理分散在多个模块中的数据流，通过专用的语言功能支持多个软件体系结构，并对主机间通信和数据转换等低级别细节进行抽象。预期的影响是，我们的解决方案可以显著减少开发和维护分布式应用程序的工作，自动生成本来需要手动编写的容易出错的代码，增强对错误的早期检测，并引入在不同平台和通信机制之间轻松迁移应用程序的方法。

Distributed applications are traditionally developed using separate modules for each component in the distributed system, often using different programming languages. These modules react to events such as user input or messages from the network and may in turn produce new events handled by other modules. Separation into tiers is often problematic: combining modules is notoriously hard and requires extensive and time consuming integration tests. Manual implementation of communication is also an issue: developers are forced to program complex event-based communication schemes among hosts - an activity which is often low level and error prone. The combination of the two makes things even worse: distributed event-based data flow is scattered among multiple modules, which makes it hard to reason about the system as a whole. For these reasons, despite most software is today distributed, the design and development of distributed systems remains surprisingly challenging.In this project, we present our roadmap for attacking the complexity of developing distributed applications via specialized programming language support. Our vision is that methods based on a multitier approach can provide a coherent model to reason about data flows otherwise scattered across multiple modules, support multiple software architectures through dedicated language features, and abstract over low level details such as inter-host communication and data conversions. The expected impact is that our solution can significantly reduce the effort of developing and maintaining distributed applications, automat- ically generating error prone code that would be otherwise manually written, enhancing early detection of bugs, and introducing means to easily migrate applications among different platforms and communication mechanisms.