gMark is a domain- and query language-independent framework targeting highly tunable generation of both graph instances and graph query workloads based on user-defined schemas.
For more details about gMark, please refer to our technical report: http://arxiv.org/abs/1511.08386
gMark was demonstrated in VLDB 2016. The gMark research paper has been accepted in the TKDE journal.
If you use gMark, please cite:
@article{BBCFLA17,
author = {Bagan, G. and Bonifati, A. and Ciucanu, R. and Fletcher, G. H. L. and Lemay, A. and Advokaat, N.},
title = {{gMark}: Schema-Driven Generation of Graphs and Queries},
journal = {IEEE Transactions on Knowledge and Data Engineering},
year = {2017 (to appear)}
}
and/or
@article{BBCFLA16,
author = {Bagan, G. and Bonifati, A. and Ciucanu, R. and Fletcher, G. H. L. and Lemay, A. and Advokaat, N.},
title = {Generating Flexible Workloads for Graph Databases},
journal = {PVLDB},
volume = {9},
number = {13},
pages = {1457--1460},
year = {2016}
}
To compile the code:
cd demo/scripts
./compile-all.sh
In the rest of the readme, suppose that we are in demo/scripts.
To generate an entire workflow, use the prepared script play.sh:
./play.sh
This executes the following three steps:
1. Generation of the graph and query workload in internal format, and html reports for both
cd ../../src
./test -c ../use-cases/test.xml -g ../demo/play/play-graph.txt -w ../demo/play/play-workload.xml -r ../demo/play/
where the parameters are:
- -c : the configuration file
- -g : the output file for the graph instance
- -w : the output file for the query workload generated on this instance (in internal format)
- -r : the output directory for the html reports
and optionally
- -a : to use aliases for the predicates in the generated graph and queries
- -n : to specify the number of nodes in the graph (it overrides the parameter from the config file)
The provided configuration files in the directory use-cases are:
- test.xml : schema of a bibliographical graph database
- shop.xml : schema of an online shop (our gMark encoding of the default schema from WatDiv)
- social-network.xml : schema of a social network (our gMark encoding of the schema from LDBC SNB)
- uniprot.xml : schema of a protein network (our gMark encoding of the schema extracted from UniProt)
2. Translation of the queries into the four concrete syntaxes
cd querytranslate
./test -w ../../demo/play/play-workload.xml -o ../../demo/play/play-translated
where the parameters are:
- -w : the query workload in internal format generated at step 1.
- -o : the output directory for the translations of the queries
3. Generation of the query workload interface
cd ../queryinterface
./test -w ../../demo/play/play-workload.xml -t ../../demo/play/play-translated -o ../../demo/play/play-interface
where the parameters are:
- -w : the query workload in internal format generated at step 1.
- -t : the translations of the queries generated at step 2.
- -o : the output directory for the query workload interface
Provided examples
We provide several examples of generated graphs, query workloads in internal format, html reports, translated queries, and query workload interfaces.
You can find them in the directory demo, subdirectories test, test-a, shop, shop-a, social, social-a, uniprot, uniprot-a.
These scenarios basically correspond to the aforementioned four configuration files from use-cases, and for each of them we generated versions without or with aliases (i.e., using integers as predicates or using real-world predicates specified in the configuration file, respectively).