K2RDF: A Distributed RDF Data Management System on Kudu and Impala

Xu Chen, Boyu Qiu, Jungang Xu, Renfeng Liu

Abstract: The Resource Description Framework (RDF) has been widely used in various applications or services as a model for displaying, sharing and connecting data. With the increase of RDF data scale, distributed RDF data management system becomes popular, but there are still many problems to be solved. To solve these problems, we proposed a distributed RDF data management system K2RDF based on the Porperty Chain model on the Kudu and Impala platforms. Kudu is a data storage engine that combines OLAP and OLTP scenario and Impala can process SQL queries in real time. The combination of these two platforms provides new options for processing RDF data, making storage more efficient and queries faster. The Property Chain model is derived from the RDF data content. The RDF data is divided into different parts stored in the corresponding attribute table by the class information extracted from RDF schema. In the attribute table, In the attribute table, each column corresponds to a property in the RDF class. This model can increase the data storage density and improve the query processing speed by reducing the number of the join operation. By comparing with the current popular distributed RDF data management systems in some experiments, our system has lower query latency and faster query speed.