A Big Data Benchmark For Graph-Processing Platforms

Graph processing is of increasing interest for many scientific areas and revenue-generating applications, such as social networking, bioinformatics, online retail, and online gaming. To address the growing diversity of graph datasets and graph-processing algorithms, developers and system integrators have created a large variety of graph-processing platforms, which we define as the combined hardware, software, and programming system that is being used to complete a graph processing task. LDBC Graphalytics, an industrial-grade benchmark under LDBC, is developed to enable objective comparisons between graph processing platforms by using six representative graph algorithms, and a large variety of real-world and synthetic datasets. Visit our website for the most recent updates of the Graphalytics project.

Want to know more about Graphalytics? Read our VLDB paper and the specification.

The Graphalytics provides platform drivers for the state-of-the-arts graph processing platforms. To start your first benchmark with Graphalytics, we recommend using our reference implementations: GraphBLAS and Umbra. Our datasets are hosted publicly – see the Graphalytics website for download instructions.

LDBC Graphalytics hosts competitions for graph processing platforms. Are you interested in the state-of-the-art performance? To participate, reach out to Gabor Szarnyas and David Puroja. Our email addresses are under [email protected].

The project uses the Build Number Maven plug-in to ensure reproducibility. Hence, builds fail if the local Git repository contains uncommitted changes.

To build & install locally regardless (for testing), run:

scripts/install-local.sh

We use a manual process for deploying Maven artifacts for the Graphalytics framework.

1. Clone the graphalytics-mvn repository next to the driver repository's directory.

2. In the driver repository, run:

   scripts/package-mvn-artifacts.sh

3. Go to the graphalytics-mvn directory, check whether the JAR files are correct.

4. Add the newly created and updates files using git, then commit and push.

5. Wait for approx. 5 minutes for the deployment process to finish.