Research On SPARQL Query Engine Across Different Storage Platform

As the development of LOD and widespread acceptance of some companies(e.g.,Google),RDF data in the semantic web is exploding.Efficient storage and querying of RDF data is of increasing important.And about 40 years of research on query optimization security and transaction support makes the relational database sufficient for the industrial-strength applications.While the popularity of Big Data,there are more and more NoSQL databases which break the limitation of ACID and can take advantage of more hardware easily to take care of intensive data access.This thesis construct a SPARQL query engine named SPONSON which reuse document store or relational database as the storage level.To using the document store as the storage level,we proposed the Subject-Partition method to split the RDF dataset which can result in a solution that reduce the number of join operations in the processing of SPARQL query;we also proposed a strategy specific for RDF data which is based on the statistic technology.To deal with the difference between the RDF model and relational model,we proposed an analysis framework for RDF dataset based on the work of boolean matrix factorization;this framework can unify two of the most popular approach to RDF storage(i.e.the property table and vertical partition);in the framework,we can get a better storage scheme under the guideline of optimization.Finally,we compared Jena-TDB and Virtuoso to the two sub-system of SPONSON which reuse document store and relational database separately as the storage level.Experiments under test of four authorize benchmarks can prove the efficiency and effectiveness of the methods proposed in this thesis.