Workload-Awareness in a NoSQL-Based Triplestore

RDF and SPARQL are increasingly used in a broad range of information management scenarios. Scalable processing of SPARQL queries has been the main goal for virtually all the recently proposed RDF triplestores. Workload-awareness is considered an important feature for the current generation of triplestores. This paper presents WA-RDF, a middleware that addresses workload-adaptive management of large RDF graphs. These graphs are stored into NoSQL databases, which provide high availability and scalability. The focus of this paper is on the Workload-Aware component (WAc) of WA-RDF. WAc was developed to avoid data fragmentation, improve data placement and reduce the intermediate results. Our experimental evaluation shows that the solution is promising, outperforming a recent baseline.