Evaluation of Main Memory Data Structures for Link Traversal Based Query Execution

Olaf Hartig
Database and Information System Research Group
Department of Computer Science
Humboldt-Universität zu Berlin

On this page we provide everything that is necessary to reproduce the caching experiments of our LDOW 2011 submission; this includes the query templates and the software used to generate the queried datasets and to run our experiments. Furthermore, we provide all our measurements. The content provided on this page is licensed under the Creative Commons Attribution-Share Alike 3.0 License.

Software

We implemented the data structures and the corresponding operations described in the paper as part of our link traversal based query execution system SQUIN. SQUIN is available as Free Software. The JAR file of the SQUIN version used for the experiments is this: squin-20110124.jar. The source code is the SVN snapshot from Jan.24, 2011: SQUIN-20110124.zip. The relevant package is org.squin.dataset.

The SQUIN code depends on the latest CVS version of NG4J and on ARQ v.2.8.7 (which is packaged with NG4J). The CVS snapshot of NG4J, including all required libraries, is here: ng4j-0.9.4-SNAPSHOT-20111024.zip.

The tool to execute the experiments, including the source code and all shell scripts, is available here: BenchmarkQueriedDatasetRepresentation-20110124.zip. Notice, this software is Free Software, licensed under the GNU General Public License (GPL), version 3. In addition to SQUIN, NG4J, ARQ, and Jena, this tool depends on the tools for the Berlin SPARQL Benchmark, V2.0 and on the Java object profiler described in the article "Sizeof for Java".

Measurements

The raw measurements recorded during our experiments can be found in the following files:

OpenOffice spreadsheet documents that contain these measurement and the charts we generated from them are here: