An Approach to an efficient Evaluation of SPARQL Queries in Databases
Christina Ehrlinger, Burkhard Freitag · 2021
Addressing the issue of efficient SPARQL 1.1 query evaluation, this paper proposes a reordering approach that can handle any SPARQL query, regardless of the version. The presented method uses the average incoming and outgoing degree of property edges in the underlying RDF graph. Elements next to basic graph patterns available in SPARQL 1.1 were analyzed regarding their impact on the performance to formulate rules for rearranging them. We achieve promising results: Using the widely known Apache Jena and Eclipse RDF4J as triple store, we reached up to 50’000 times faster execution times than the best performing optimization approach included in Apache Jena using queries from three different well-known SPARQL benchmarks: LDBC SNB, BSBM, and SP2Bench.