Flink Gelly Algorithms

Repository containing various graph algorithm implementations using Flink's Gelly Graph API

Installation

All jobs were run on Flink 1.8.0.

Information about setting up Flink can be found in Flink's Official Documentation.

Building JARs

1. cd into an algorithm implementation directory (e.g. cd PageRank)
2. Run mvn clean package in order to build the project
3. The JAR is located within the target folder

Configuring the cluster

To configure the system edit the conf/flink-conf.yaml file

• jobmanager.heap.size: Nm (the heap size for the JobManager JVM - e.g. 4096m)

• taskmanager.heap.size: Nm (the heap size for the TaskManager JVM - e.g. 4096m)

• taskmanager.numberOfTaskSlots: N (the number of parallel operator or user function instances that a single TaskManager can run (DEFAULT: 1). If this value is larger than 1, a single TaskManager takes multiple instances of a function or operator. That way, the TaskManager can utilize multiple CPU cores, but at the same time, the available memory is divided between the different operator or function instances. This value is typically proportional to the number of physical CPU cores that the TaskManager’s machine has - e.g. equal to the number of cores, or half the number of cores)

Starting/Monitoring/Stopping the cluster

• To start the (local) Flink cluster run:

./Flink-1.8.0/bin/start-cluster.sh

• You can check jobs through the web UI (default: http://localhost:8081/#/overview)

• To stop the cluster run:

./Flink-1.8.0/bin/stop-cluster.sh

Running a job

To run a Flink job via the Flink CLI execute the following command:

./Flink-1.8.0/bin/flink run path/to/JAR --links path/to/edgelist/csv/file

SW Versions

• Scala 2.12.8
• Java 1.8
• Maven 3.6.0

License

As stated in LICENSE.