In this document, we suggest workflows in the Collaboratory environment for a few tasks where Jupyter Notebooks are useful.

Documentation

When you want to document a library or service, we suggest you create an example notebook in the drive of a Collab. You can use a public Collab if you want it to be visible to everyone. Here is an example.

You can put sample data in the drive.

You can setup the library and requirements for the notebook using pip.

You can create sample output in a folder in the user's home. This will only persist so long as the user is active on the platform within the previous 24 hours.

Interactive Instruction

When you want to setup a course, you can put your notebook on the drive of a Collab and make the students viewers. You can include small datasets. Students (viewers) will have read only access to the notebook and data. They will be able to execute it, but not to save any changes. To save their version, they can save the notebook to a drive folder that they have the right to edit.

Output can be saved to a directory in the user's home, or to a folder on the user's drive that they have access to, such as "My Libraries/My Library"

Automation

To automate experiments or tasks, you can create a notebook which processes data and can launch jobs on the HPC. Save the notebook in a Collab and members of the Collab will be able to run the notebook. Use Unicore to dispatch jobs to the HPC.

Exploration

You can create notebooks to help you with your experiments. Whether it is for data visualization, analysis or simulations, you code can be shared with others. You can build rich interfaces using IPython Widgets.