A Middleware for Polyglot Persistence of RDF Data into NoSQL Databases

Abstract

Software engineers can consider today a multitude of storage solutions and data formats to achieve better performance, lower cost, or even explore the power expression of a data model to develop an application. We call it polyglot access. Nevertheless, the cost of developing polyglot software increases due, for instance, to the complexity of managing multiple connections to databases and the need for training people to use different tools, models and query languages. This paper presents a scalable middleware, called WA-RDF, that provides a unique gateway to multiple NoSQL databases. Different from other similar ideas, WA-RDF uses the well-known abstractions of Semantic Web to store and query RDF data into key/value, document and graph databases. Moreover, WA-RDF includes workload-awareness, fragmentation and partitioning components to meet the NoSQL high level of scalability. An experimental evaluation shows that the approach is promising. It scaled linearly to the dataset size and query frequency growth, and outperformed a multimodel database in the tested use cases.