LUBMex20

Here we provide the detailed information for the LUBMex20 experimentation (Sections 3.1 and 3.2). Note that in this case there is no given SQL schema; and hence, no mappings. Ontop creates a relational schema together with mappings itself using the "Semantic Index" technique described in the paper.

Ontology and Queries

The ontology (the original can be found at http://informatik.uni-bremen.de/~clu/combined/)

LUBM-ex-20.owl

Queries (R1-R5 from the Requiem evaluation, Q1-Q6 from the combined approach evaluation and Q7-Q9 from the Clipper evaluation)

LUBM-ex-20-SPARQL.txt (SPARQL format used by Ontop, Clipper, OWLIM and Stardog)
LUBM-ex-20-datalog.txt (datalog format used by Requiem, IQAROS, Rapid)

The data for these experiments (with 50, 200 and 1000 universities) was generated using the data generator available at http://informatik.uni-bremen.de/~clu/combined/.

Loading the data into Stardog and OWLIM

To load the data into Stardog or OWLIM, do the following steps:

generate the desired number of Universities using EUGen.zip;
merge each set of University0_x.owl files into a single .ttl file using Jena's rdfcat (you can use our bash script for this);
load each uni file from the .ttl file using the usual commands for Stardog/OWLIM.

Loading the data into MySQL and DB2

The data for the semantic index schema can only be generated using a version of Ontop which is not publicly available at the moment. However, we have created dumps of the DBs generated by Ontop's Semantic Index for LUBMex20 with 50, 200 and 1000 universities that can be used with MySQL, DB2 or any other DB that supports loading data from CSV files. Follow the instructions below to prepare the databases:

create the schema in MySQL/DB2 using the lubm-create-schema-for-csv.sql script;
download the gzipped CSV files with the data for the tables and load them into the appropriate tables using the databases bulk data load command (e.g., LOAD in MySQL/DB2)
- 50 universities: CA, RA, URI
- 200 universities: CA, RA, URI
- 1000 universities: CA, RA, URI
create the indexes using the lubm-create-indexes.sql script and collect index statistics using "ANALYZE TABLE x" (MySQL) or "RUNSTATS ON x" (for DB2) for each table x;
use the SQL queries generated by Ontop.

Summary

Query execution time (in seconds) for each query in the LUBMex20 scenario.

These are the averages of 5 warm executions.

https://docs.google.com/spreadsheet/ccc?key=0AkzBFnHPAuKedGNVczdiX21TRXZwTC04VUI3R1M3Q1E&usp=sharing

Detailed results

Detailed results for each system, including the loading time (SQL or triple) and the forward chaining time (for OWLIM):