CSR---AES: Programming Language and Runtime System Support for Robust Distributed Software Systems

异构的、网络连接的计算设备的集合已经成为一种新的计算范式。这种新的基于网络的范例简化了构建灵活和可扩展的硬件基础设施。不幸的是，使这样的系统对故障保持健壮是很困难的。许多这样的分布式软件系统包含大量的设备，因此，有很大的机会经历硬件故障。开发人员必须手动开发代码，以允许软件系统从此类硬件故障中恢复。因此，开发健壮的分布式软件系统通常比健壮的集中式软件系统要困难得多。这项工作建立在首席研究员之前关于开发健壮软件系统的bristlecone语言的工作基础上。Bristlecone语言背后的关键观点是，大多数错误通过软件系统传播，通过破坏数据结构或通过概念操作之间的控制流诱导耦合造成进一步的损害。Bristlecone程序被架构为一组解耦的任务，这些任务通过一组任务规范连接起来，这些任务规范描述了这些解耦的任务如何交互以及一致的数据结构是什么样子的。Bristlecone然后使用这些规范来调整程序的执行以响应失败。这个项目扩展了以前的工作，以支持分布式软件系统。该项目开发静态分析，以帮助开发人员了解底层硬件的故障将如何影响软件系统，并管理数据和任务，以便硬件故障对计算的影响最小。

Collections of heterogeneous, network-connected computational deviceshave emerged as a new computational paradigm. This new network-basedparadigm simplifies constructing flexible and scalable hardwareinfrastructure. Unfortunately, making such systems robust to failurescan be difficult. Many of these distributed software systems containa large number of devices, and therefore, have a significant chance ofexperiencing a hardware failure. Developers have to manually developcode to allow the software system to recover from such hardwarefailures. As a result, developing robust distributed software systemsis typically more difficult than robust centralized software systems.This work builds upon the Principal Investigator's prior work on theBristlecone language for developing robust software systems. The keyinsight behind the Bristlecone language is that most errors propagatethrough software systems to cause further damage either by corruptingdata structures or through the control-flow--induced coupling betweenconceptual operations. Bristlecone programs are architected as a setof decoupled tasks that are linked through a set of taskspecifications that describe how these decoupled tasks interact andwhat consistent data structures look like. Bristlecone then usesthese specifications to adapt the program's execution in response tofailures.This project extends the previous work to support distributed softwaresystems. This project develops static analyses to help developersunderstand how failures in the underlying hardware will affect thesoftware system and to manage data and tasks so that hardware failureshave a minimal affect on the computation.