Home » Graphalytics Benchmark

Graphalytics Benchmark

The Graphalytics benchmark is an industrial-grade benchmark for graph analysis platforms. It consists of six core algorithms, standard datasets, synthetic dataset generators, and reference outputs, enabling the objective comparison of graph analysis platforms.

The main Graphalytics components can be found at:

Documentation

The Graphalytics project leads to several academic publications, including a recent article at VLDB 2016 (available soon). The first draft of the Graphalytics specification document (https://github.com/tudelft-atlarge/graphalytics_docs) is available, which explains in details the benchmark specifications of Graphalytics.

To start using Graphalytics, a complete guide on how to install and run LDBC Graphalytics (https://github.com/tudelft-atlarge/graphalytics) can be found in our repositories, and there are also instructions for creating a new platform driver for Graphalytics (https://github.com/tudelft-atlarge/graphalytics/wiki/).

Implementation

Our repository is currently hosted at https://github.com/tudelft-atlarge, which may soon be relocated to the main LDBC repository. Graphalytics consists of a core implementation (https://github.com/tudelft-atlarge/graphalytics), which is extendable by a driver implementation for each platform. Currently, Graphalytics also already support driver implementation of several state-of-the-art graph analysis platforms (https://github.com/tudelft-atlarge/graphalytics-platforms-*), some of which are vendor-optimized.

Standard datasets and Reference outputs

Graphalytics also provides standard datasets and their reference outputs (validated outputs for each algorithm), which can be used in the benchmark process. These datasets will be made available soon.

Recommended Tools

To enhance the depth and comprehensiveness of the benchmark process, the following software tools are integrated into Graphalytics. Usage of these tools is optional, but highly recommended.

  1. Datagen (https://github.com/ldbc/ldbc_snb_datagen): Graphalytics relies not only on real-world graphs but also on synthetically generated graphs, which provide a means of testing data configurations not always available in the form of real datasets. Datagen, for example, is an advanced synthetic graph generator which not only preserves many realistic graph features, but also supports graphs with tunable degree distributions and structural characteristics.
  2. Granula (https://github.com/tudelft-atlarge/granula): To extend Graphalytics with fine-grained performance evaluation, we developed Granula, a performance evaluation system for graph analysis platform, Granula consists of four main modules: the modeller, the monitor, the archiver, and the visualizer. By using Granula, enriched performance results can be obtained for each benchmark run, which helps in facilitating in-depth performance a