Simulation and Evaluation of Decentralized SPARQL Query Processing

Previously, we proposed efficient, scalable decentralized processing of SPARQL queries for an ad-hoc Semantic Web data sharing system and explored optimization techniques. However, it has proven to be difficult to measure the performance of the proposed query processing in a decentralized settings with existing tools. This is because assessments on SPARQL query performance were typically targeted at a centralized or single machine settings, and node-to-node communication costs occurring when (sub-)queries were forwarded among multiple nodes have rarely been taken into consideration. We hereby developed a simulator, SShare, that bridges Jena and ns-3 (Network Simulator 3). With SShare, one can submit any proper SPARQL query that involves RDF data of interest scattered around distributed hosts (the details of which are unknown to the query initiator), evaluate important performance metrics (e.g. the inter-site data transmission volume and communication delay) obtained at the network level, and finally get visualized results. We anticipated that SShare would be beneficial to others who are keen on better capturing and analyzing the inherent feature of various distributed and decentralized SPARQL processing mechanisms over a large scale network.