String-Based Semantic Web Data Management Using Ternary B-Trees

The Resource Description Framework (RDF) stems from the Semantic Web but can also be regarded simply as a data model, independent of its origins. Its simple structure is ideal for describing and merging heterogeneous data from different sources quickly, without having to design a complex schema first. The different nature of RDF requires new approaches for data management and query processing and entails new problems when it comes to efficiency and scalability. Current solutions rely on string mapping and extensive indexing, but concentrate on a subset of the RDF query language SPARQL. They also ignore inherent properties residing in RDF itself, particularly its string characteristics. In this work we propose the ternary B-tree as a new data structure for storing and accessing RDF. It is string-based, making use of the intrinsic features of RDF. Strings are decomposed into fixed-size elements and organized as tree nodes. This approach permits new ways of filtering RDF, joining query results, partitioning for different orderings, and for parallelization.