GRAFS (GRaph Analytics with Apache Flink's Stream-API) is a prototype for scalable graph stream analytics using Apache Flink's DataStream-API, which is part of my Master Thesis. It's graph model is adapted from the Property Graph Model and the Extended Property Graph Model introduced by University of Leipzig. An element in the stream, called Triplet, contains exactly one edge of the graph and its source and target vertex. Each Graph element can hold a label and multiple properties, as well as its graph memberships.

This module contains the implementation of GRAFS. It features 4 operators, namly Subgraphing, Transformation, Windowed Grouping and Windowed Dual Simulation, as well as the auxilary operator Reduce. An introduction to each operator can be found in the Gradoop wiki. For detailed explanation of those read the master thesis.

GraphStream and GCStream are the main classes. Simply convert your graph stream source to a Flink DataStream using Triplets and use that as parameter for the stream. An example for an analytics pipeline can be found below

// define config before with the streaming environment used for the DataStream
GraphStream gs = new GraphStream(stream, config);
// Let's assume this is a stream of persons (vertices) and its relationsships with each other.
// Mark everyone as adult or child and find only relationships between two adults or two children
GCStream gcs = gs
    .transformVertices(v -> {
      v.getPropertyValue("age").getInt() > 18
        ? v.setProperty("adult", true)
        : v.setProperty("adult", false);
        return v;
        })
    .dualSimulation("MATCH (a)->(b)" + 
                    "WHERE a.adult = b.adult");

This module contains the benchmark test for GRAFS.

This module contains the data setup to upload the csv-files as triplets to a Apache Kafka cluster.