III: Small: Scalable RDF Query Processing Using a Cloud Infrastructure

近年来，资源描述框架(RDF)已经成为包括万维网在内的广泛应用的日益重要的数据和知识表示形式。随着RDF数据集大小的快速增长，对用于存储、索引和查询数万亿倍大小的RDF数据集的可伸缩且高效的技术的需求日益增长。该项目由密苏里大学堪萨斯城分校的Praveen Rao博士领导，旨在通过开发：(1)一种存储、索引和查询RDF数据的新方法，将图视为一等公民，以减少使用RDF签名、RDF签名索引和折线图进行图模式匹配所需的成本和连接数量；(2)基于RDF签名的数据分发机制、跨计算节点快速查找感兴趣的RDF图的位置索引和八卦驱动的查询执行模型，提出了云平台上并行SPARQL查询处理的新方法；(3)基于RDF图模式基数估计的新八卦算法和用于有效负载平衡和提高准确率的分而治之的方法，提出了用于查询优化的RDF图模式选择性估计的新方法。该项目的更广泛影响包括涵盖RDF数据管理和云计算主题的新课程、为肿瘤学家提供的针对癌症数据的可扩展RDF推理工具、用于超大型RDF数据存储的新云服务、增加对包括女性在内的本科生和研究生进行研究型高级培训的机会。这项研究的结果，包括出版物、软件和数据集，将免费与更广泛的社区共享。有关该项目的更多信息，请访问该项目的网站：http://vortex.sce.umkc.edu/ric.html.

Resource Description Framework (RDF) has, in recent years, become an increasingly important data and knowledge representation formalism for a broad range of applications, including the World Wide Web. With rapidly growth in the size of RDF datasets, there is growing need for scalable and efficient technologies for storing, indexing, and querying RDF datasets that are trillions of triples in size. This project, led by Dr. Praveen Rao of University of Missouri-Kansas City, aims to address this need by developing: (1) A novel approach to storing, indexing, and querying of RDF data that treats graphs as first-class citizens to reduce the cost and number of joins required for graph pattern matching using RDF signatures, RDF signature indexes and line graphs; (2) A new approach for parallel SPARQL query processing on cloud platforms using data distribution schemes based on RDF signatures, location index for quickly finding RDF graphs of interest across computing nodes, and a gossip-driven query execution model and (3) A new approach for selectivity estimation of RDF graph patterns for query optimization based on new gossip algorithms for cardinality estimation of RDF graph patterns and a divide-andconquer method for effective load balancing and improved accuracy. The broader impacts of this project include new courses covering topics in RDF data management and cloud computing, a scalable RDF reasoning tool over cancer data for oncologists, new cloud services for very large RDF data stores, increased opportunities for research-based advanced training of undergraduate and graduate students, including women. The results of this research, including publications, software, and data sets will be freely shared with the broader community. Additional information about the can be accessed through the project website at http://vortex.sce.umkc.edu/ric.html.