CAREER: Diagnosing Distributed Systems with Provenance

职业：诊断分布式系统的来源

• 批准号: 1453392
• 负责人: Wenchao Zhou
• 金额: $ 49.84万
• 依托单位: Georgetown University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2015
• 资助国家: 美国
• 起止时间: 2015-06-01 至 2021-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1453392&HistoricalAwards=false
• 关键词: CAREER; Diagnosing; Distributed; Systems; Provenance

Despite the ubiquitous adoption of distributed systems, designing and deploying distributed systems remain challenging because of the ever-increasing scale, complexity, and unpredictability of system executions. In occurrences of unanticipated faults, system operators often find themselves needing to diagnose the systems to discover software bugs, misconfiguration or malicious intrusions. Emerging provenance techniques can construct an explanation of an observed system state, in the form of a directed dependency graph. However, the use of provenance is limited; they lack query interfaces that allow users to easily extract information from provenance data, as well as, techniques to fully leverage provenance data. System operators have to further analyze provenance data, and manually perform management actions such as system recovery and reconfiguration.This project introduces a framework that efficiently maintains and queries provenance data, and leverages the exposed dependency information for system diagnosis. The research will i) develop layered privacy-aware provenance maintenance that provides availability and security guarantees in untrusted environments, ii) introduce a SQL-like declarative query language for users to annotate and abridge provenance data, and iii) investigate various methods of leveraging provenance data in order to identify and control suspicious misbehavior. The work provides a holistic view of provenance support for practical applications that impact performance, privacy, availability and security.More broadly, this project will provide a novel approach towards enhancing the reliability and security of distributed systems in an automated manner. Tools developed in the project will be freely distributed, and research findings will be incorporated into networking and databases courses.

尽管分布式系统被普遍采用，但由于系统执行的规模、复杂性和不可预测性不断增加，设计和部署分布式系统仍然具有挑战性。在发生意外故障时，系统操作员经常发现自己需要诊断系统以发现软件错误、错误配置或恶意入侵。新兴的起源技术可以构建一个解释所观察到的系统状态，在一个有向依赖图的形式。然而，出处的使用是有限的;它们缺乏允许用户轻松从出处数据中提取信息的查询接口，以及充分利用出处数据的技术。系统操作员需要进一步分析起源数据，并手动执行管理操作，如系统恢复和重新配置。本项目介绍了一个框架，有效地维护和查询起源数据，并利用暴露的依赖信息进行系统诊断。该研究将i）开发分层的隐私感知出处维护，在不受信任的环境中提供可用性和安全性保证，ii）引入类似SQL的声明性查询语言，供用户注释和删节出处数据，以及iii）研究利用出处数据的各种方法，以识别和控制可疑的不当行为。这项工作为影响性能、隐私、可用性和安全性的实际应用程序提供了来源支持的整体视图。更广泛地说，该项目将为以自动化方式增强分布式系统的可靠性和安全性提供一种新方法。该项目开发的工具将免费分发，研究结果将纳入网络和数据库课程。