Ontotext has released GraphDB 9.7, which offers a substantial improvement in SPARQL performance.
The new version of the GraphDB engine implements a major improvement in the memory management and a 30% reduction in the generated Java objects. In combination with the optimized hashing algorithms and processing of DISTINCT and GROUP BY results, the database achieves up to 7 times better throughput with such queries over large datasets like BSBM 1B. There is also substantial improvement of the utilization of multi-core CPUs for handling multi-client workloads that involve such queries. The overall performance improvement for all other operations is between 10-15% due to the lowered GC time. All changes are backward compatible and should positively affect large repositories under moderate to heavy query load.
The GraphDB Connectors enable real-time and transactionally safe replication and indexing of knowledge graphs with search systems like Lucene, Solr and Elasticsearch. In the 9.7 release, the Connectors were extended to support the indexing of all RDF literals for a given resource, inverse property paths, range searches and improved date/time support. These changes deprecated the GraphDB Lucene FTS plugin and enabled data indexing with complex filtering conditions like temporal data. All changes are backward-compatible in terms of existing connector definitions.
This release is no exception to the rule of listening to the feedback from our users. It features user experience improvements in the JDBC virtual table configuration, support for showing the owner of the running query in query/update monitoring, support for FROM and FROM NAMED over the internal federation protocol.
GraphDB comes with the latest release 3.6.1 of rdf4j – the most popular open-source RDF framework. It includes many fixes in the SHACL support for data validation with RDF shapes, contributed by Ontotext. Virtual SPARQL endpoints now use Ontop 4.1 – another open-source project with Ontotext contributions. The main new features of the later include support for more SPARQL functions, such as IF, IRI, and BNODE, new SQL generation optimizations, and the native support for Dremio.
There are important bug fixes in the FTS connectors and SPARQL evaluation (inverse property paths and BOUND in UNION clause). A deadlock related to heavy load of SHACL requests has been resolved, as well as issues related to export of virtual repositories.
Get ready to take advantage of all new GraphDB 9.7 features!
For more information, contact Doug Kimball, Chief Marketing Officer at Ontotext