CSR-DMSS, SM: A Holistic Approach to Reliable Pervasive Systems

CSR-DMSS、SM：实现可靠普及系统的整体方法

• 批准号: 0834529
• 负责人: Patrick Eugster
• 金额: $ 40万
• 依托单位: Purdue University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2008
• 资助国家: 美国
• 起止时间: 2008-09-01 至 2012-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0834529&HistoricalAwards=false
• 关键词: CSR; DMSS; SM; Holistic; Approach

Holistic Approach to Reliable Pervasive SystemsThis project proposes a holistic approach to developing reliable pervasive systems through three contributions: (1) A unified view of a pervasive distributed system is proposed for debugging purposes. This view overcomes the heterogeneous requirements for various technologies. The key idea is to formulate all system artifacts, system wide or local to a given node, static or temporal, independently of their granularity, into relations. Programs are debugged by writing SQL-like queries over these relations. A query compiler is developed that translates queries into system-level or node-level replay requests, and in the meantime, instruments application code to populate relevant relations. (2) Programming language support is proposed to facilitate the tracking of dependencies among essential events in the program. To that end, a simple language built on Java is proposed, with annotations for high-level application events and their dependencies. This represents one of the cornerstones for dealing with the large number of events in dynamic pervasive applications. In addition, the integration with a programming language allows the query compiler to verify parts of queries referring to application-specific definitions. (3) Runtime and communication support through specific overlay networks is investigated to efficiently propagate events as well as causality information across nodes. At the node level, an instrumented Java virtual machine is used to log and trace local events in a runtime-adaptable manner. Such an adaptive event logging/tracing strategy significantly reduces resource demands.Involvement in computer science student organizations will be specifically leveraged to promote undergraduate student research.

可靠普适系统的整体方法本项目提出了一个整体的方法来开发可靠的普适系统，通过三个贡献：（1）提出了一个统一的普适分布式系统的视图，用于调试目的。此视图克服了各种技术的异构要求。其关键思想是制定所有的系统工件，系统范围内或本地到一个给定的节点，静态或时间，独立于他们的粒度，到关系。通过对这些关系编写类似SQL的查询来调试程序。 开发了一个查询编译器，将查询转换为系统级或节点级的重放请求，同时，仪器应用程序代码填充相关关系。(2)编程语言的支持，以促进程序中的基本事件之间的依赖关系的跟踪。为此，提出了一种基于Java的简单语言，为高级应用程序事件及其依赖关系提供注释。这是处理动态普适应用程序中大量事件的基石之一。此外，与编程语言的集成允许查询编译器验证引用特定于应用程序的定义的查询的部分。(3)研究了通过特定覆盖网络的通信和通信支持，以有效地传播事件以及跨节点的因果关系信息。在节点级别，一个插装的Java虚拟机用于以运行时自适应的方式记录和跟踪本地事件。这种自适应的事件记录/跟踪策略大大减少了资源需求。参与计算机科学学生组织将特别用于促进本科生的研究。