==== BSBM-Jena Berlin SPARQL Benchmark with additions for testing local stores. BSBM runs all the queries over the SPARQL Protocol whereas this version calls the stores directly, so removing protocol overhead. For the smaller datasets and simpler queries, the overheads can be a significant part of the cost so evaluation of stores can be skewed. BSBM-Local adds new endpoint formats as pseudo-URI schemes: * "jena:<assembler>" where "assembler" is the file name of a Jena assembler description for the store. * "sesame:<directory>" to connect to a Sesame repository. == Get the system and setup git clone git://github.com/afs/BSBM-Local.git # Create directories. mkdir Data TDB-DB Sesame-DB mkdir src-test # The dependent jars. mkdir lib-sys mvn dependency:copy-dependencies -DoutputDirectory=lib-sys Either use Eclipse (the project setup is included) or use "mvn compile" which places the jar in target/classes == Directories Data/data-SIZE -- generated data TDB-DB/DB-SIZE/ -- TDB Sesame-DB/Repo-SIZE -- Sesame == Process - for TDB backed stores. # Choose sizes wanted genAll # Run the tdbloader, create stats files and assembler files. loadAll # Run the performance tests # Results go into "Results/" runPerf # Generate the results, naming the results directory. # NB The BSBM formatter knows the possible store sizes runResults Results == Process - Sesame loadSesame runPerfSesame == Questions? If about BSBM generaly, ask the authors, see http://www4.wiwiss.fu-berlin.de/bizer/BerlinSPARQLBenchmark/ If about the local customization, please send questions to: jena-dev@groups.yahoo.com