A common approach to providing persistent storage for RDF is to store statements in a three-column table in a relational database system. This is commonly referred to as a triple store. Each table row represents one RDF statement. For RDF graphs with frequent patterns, an alternative storage scheme is a property table. A property table comprises one column containing a statement subject plus one or more columns containing property values for that subject. In this approach, a single table row may store many RDF statements. This paper describes a property table design and implementation for Jena, an RDF Semantic Web toolkit. A design goal is to make Jena property tables look like normal relational database tables. This enables relational database tools such as loaders, report writers and query optimizers to work well with Jena property tables. This paper includes a basic performance analysis comparing a triple store with property tables for dataset load time and query response time. Depending on the application and data characteristics, Jena property tables may provide a performance advantage compared to a triple store for large RDF graphs with frequent patterns. The disadvantage is some loss in flexibility.
[1]
Jeff Heflin,et al.
An Evaluation of Knowledge Base Systems for Large OWL Datasets
,
2004,
SEMWEB.
[2]
Nigel Shadbolt,et al.
SPARQL Query Processing with Conventional Relational Database Systems
,
2005,
WISE Workshops.
[3]
Kevin Wilkinson,et al.
Application-Specific Schema Design for Storing Large RDF Datasets
,
2003,
PSSS.
[4]
Vassilis Christophides,et al.
Benchmarking Database Representations of RDF/S Stores
,
2005,
SEMWEB.
[5]
Frank van Harmelen,et al.
Sesame: A Generic Architecture for Storing and Querying RDF and RDF Schema
,
2002,
SEMWEB.