CAREER: Novel Query Processing Techniques for Distributed Probabilistic Data

职业：分布式概率数据的新颖查询处理技术

• 批准号: 1200792
• 负责人: Feifei Li
• 金额: $ 49.74万
• 依托单位: University of Utah
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2011
• 资助国家: 美国
• 起止时间: 2011-09-29 至 2017-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1200792&HistoricalAwards=false
• 关键词: CAREER; Novel; Query; Processing; Techniques

Data are increasingly generated, stored, and processed distributively. Meanwhile, when large amounts of data are generated, ambiguity, uncertainty, and errors are inherently introduced, especially in a distributed setup. It is best to represent such data in a distributed probabilistic database. In distributed data management, summary queries are useful tools for obtaining the most important answers from massive quantities of data effectively and efficiently, e.g., top-k queries, heavy hitters (aka frequent items), histograms and wavelets, threshold monitoring queries, etc. This project investigates novel query processing techniques for various, important summary queries in distributed probabilistic data.Broadly classified, this project examines both snapshot summary queries in static (i.e., no updates) distributed probabilistic databases, and continuous summary queries in dynamic (i.e., with updates) distributed probabilistic databases. A number of techniques are explored to design novel, communication and computation efficient algorithms for processing these queries. A distributed probabilistic data management system (DPDMS) prototype is implemented based on the query processing techniques developed in this project. This DPDMS is released to and used in practice by scientists and engineers from other science disciplines as well as industry.Graduate and undergraduate students, including those from minority groups, are actively involved in this project. Findings from the project have been integrated into different courses, demos, and educational projects. For further information, such as publications, data sets, source code, and education initiatives, please visit the project website at http://www.cs.fsu.edu/~lifeifei/dpdm.

数据越来越多地以分布式方式生成、存储和处理。同时，当产生大量数据时，必然会引入歧义、不确定性和错误，特别是在分布式设置中。最好在分布式概率数据库中表示此类数据。在分布式数据管理中，摘要查询是有效且高效地从大量数据中获取最重要答案的有用工具，例如top-k查询、重磅查询（又称频繁项）、直方图和小波查询、阈值监控查询等。本项目研究分布式概率数据中各种重要摘要查询的新颖查询处理技术。从广义上讲，这个项目既检查静态（即没有更新）分布式概率数据库中的快照摘要查询，也检查动态（即有更新）分布式概率数据库中的连续摘要查询。为了处理这些查询，我们探索了许多技术来设计新颖、通信和计算效率高的算法。基于本项目开发的查询处理技术，实现了分布式概率数据管理系统（DPDMS）原型。该DPDMS发布给来自其他科学学科以及行业的科学家和工程师并在实践中使用。研究生和本科生，包括少数民族学生，都积极参与到这个项目中来。该项目的研究成果已被整合到不同的课程、演示和教育项目中。欲了解更多信息，如出版物、数据集、源代码和教育计划，请访问项目网站http://www.cs.fsu.edu/~lifeifei/dpdm。