How Diverse Are Federated Query Execution Plans Really?

Anders Langballe Jakobsen, Gabriela Montoya, Katja Hose

Abstract: Federated query engines optimize and execute SPARQL queries over the data accessible via SPARQL endpoints; even in the absence of information about which sources provide relevant data. Different query optimization strategies may produce different execution plans, which in many cases explains the huge differences in performance that we encounter with state-of-the-art federated query engines. Related work has so far mainly focused on execution time and number of selected sources, overlooking the importance of the execution plans themselves. In this demonstration paper, we therefore present PIPE, a tool that allows for comparing federated query engines in terms of performance, execution plans, and query answers. Currently, PIPE supports five state-of-the-art federated query engines; we provide a Java library for straightforward integration of additional federated query engines.