An Efficient Approach to Triple Search and Join of HDT Processing Using GPU

Abstract

Resource Description Framework (RDF) is originally designed as a metadata data model. It has an advantage of efficient exchange between different metadata by supporting a set of common rules. To process RDF data efficiently, SPARQL (SPARQL Protocol and RDF Query Language) was introduced. In this era of emerging Web of Data, the amount and size of published RDF data have dramatically increased. Most studies so far have focused on the compression of RDF data and fast SPARQL query processing using single core CPUs. Thus, they do not utilize well current multicore environment. We in this study propose SPARQL query processing using a GPU multicore system. We focus on a search and join method using Header, Dictionary, Triples (HDT) data format and present its experimental results.