Query Optimization And Performance Evaluation Of RDFBase

As a new generation of Internet, more and more researchers pay attention to the semantic web. RDF (Resource Description Framework), recommendation of W3C organization to depict web resources, is the data model for describing metadata. As the fundamental of RDF, the RDF graph is a directed hypergraph. Meanwhile, in order to query RDF data, W3C provide semi-structured SPARQL query language, as the future development trend for semantic web.Relational database management system can't meet the requirements of the RDF data model effectively, so it needs to develop new system. This paper is deeply engaged in the research of RDFBase, a store and query of RDF data. In storage, RDFBase uses triple table to store data, builds multi-index to improve the query plan, and applies algorithm to reduce the cost of space. In query, RDFBase adopts cost estimation method, heuristic rules and sideways information passing mechanism to optimize the SPARQL query language. Rather than linearly joining triples in a basic graph pattern, it may be more efficient to evaluate each of the groups independently, and then merge the resulting sets. Based on this idea, we introduce star-shaped groups technique for RDFBase, and define its cost estimation optimization strategy.From the experiment for performance evaluation which is made in the last section, we can see compression algorithm has a good compression ratio, saving the space effictively; start-shaped groups technique allows for quick query processing, shortening the query time; RDFBase has a good query performance and scalability, and it can load data in the limited time, better than other systems for large data sets.