Get started

We use JXR and Javadoc to allow developers to easily understand our code. Start by looking at the Benchmark or Pagerank examples.

Running the code

Peregrine has two main components. The Worker node (which runs the filesystem, mapper, merger, and reducer tasks) and the controller which is used to scheduler your job across the cluster..

To startup the worker on localhost just run:

export HOSTNAME=localhost && ./bin/workerd

This will startup a new daemon on port 11112

You can then run your client code with an embedded controller against this new worker node.

To run a simple benchmark you could execute:

./bin/jexec peregrine.app.benchmark.Main

This will startup a new controller instance and execute the benchmark job.

Larger clusters

This would just run on localhost with one compute node.

To run on a larger node you will need to set a controller and edit peregrine.hosts with the servers you wish to use.

The same config and host files must be present on all hosts and the controller. Your job code should be placed in a .jar and added to the classpath.