Efficient processing of SPARQL joins in memory by dynamically restricting triple patterns

Abstract

Since there are a lot of similar or common properties between RDF and relational databases and between SPARQL and SQL, many efforts focus on leveraging the research results of optimizing relational query languages for optimizing SPARQL queries. However, SPARQL has its own characteristics different from SQL, which are not fully exploited by existing work. Therefore, there is still much space for research on optimizing SPARQL queries. Based on the triple nature of RDF data, we create 7 indices to retrieve RDF data quickly; based on the SPARQL-specific properties and the 7 indices, we develop a new, efficient approach to computing join by dynamically restricting triple patterns. Our experimental results show the efficiency of our approach. Copyright 2009 ACM.