Ensemble Toolkit (EnTK)


The Ensemble Toolkit (EnTK) is a Python library developed by the RADICAL Research Group at Rutgers University for developing and executing large-scale ensemble-based workflows. This tutorial shows how to get up and running with EnTK on Summit specifically. For in-depth information about EnTK itself, please refer to its documentation.


Before using EnTK itself, you will need MongoDB and RabbitMQ services running on Slate. The tutorials for this are forthcoming.

You will need to know the connection information for both MongoDB and RabbitMQ so that EnTK can be configured to connect to the services.

Hello world!

To run EnTK on Summit, you will create two files and then execute two commands from a Summit login node. Currently, EnTK must be run from a Summit login node, rather than within a batch job.

First, create a setup file setup.bash that will load the correct modules and define environment variables.

module load python/3.7.0
module load py-virtualenv/16.0.0
module load py-setuptools/40.4.3-py3

python3 -m venv ./
source ./bin/activate
pip install --upgrade pip
pip install radical.entk

export RADICAL_PILOT_DBURL="mongodb://admin:password@apps.marble.ccs.ornl.gov:32767/test"
export RMQ_HOSTNAME="apps.marble.ccs.ornl.gov"
export RMQ_PORT="30256"
export RMQ_USERNAME="admin"
export RMQ_PASSWORD="password"

Replace, without renaming, the RADICAL_PILOT_DBURL environment variable in setup.bash with the MongoDB connection string that corresponds to your own service running on Slate. EnTK uses this environment variable directly.

Then, replace the RMQ_ variables in setup.bash with the corresponding values for your RabbitMQ service on Slate. These variables can be renamed, because their only use is in the following example Python program.

Now, create a demo.py3 file with the following lines:

import os
from radical.entk import AppManager, Pipeline, Stage, Task

# Create objects.

appman = AppManager(

p = Pipeline()

s = Stage()

t = Task()

# Use the objects to model the workflow.

appman.resource_desc = {
    'resource': 'ornl.summit',
    'walltime': 10,
    'cpus':     1,
    'project':  'stf011' # replace with your own project identifier

t.name = 'my-first-task'
t.executable = '/bin/echo'
t.arguments = ['Hello world!']



appman.workflow = set([p])

# Execute the workflow.


In demo.py3, only one line needs to be changed, so that EnTK knows which project identifier to use when submitting batch jobs to Summit.

Finally, run the demo program by executing the following commands from a Summit login node:

$ source setup.bash
$ python3 demo.py3

Congratulations! You should now see interactive output from EnTK while it launches and monitors your job on Summit.