Hands-on Tutorial Deploying Portable Research and Learning Environments on Commercial Clouds and XSEDE
TimeMonday, July 238:30am - 5pm
DescriptionJupyterHub connects custom computational environments to multiple users via the cloud. It is a flexible, rapidly-growing technology that has proven useful in both pedagogy and research. For example, it has been used for single- and multi-day workshops as well as a 1000 student class at UC Berkeley. Building upon JupyterHub, BinderHub enables users to easily share reproducible computational environments with others. For an example of a public BinderHub service, see mybinder.org.
This day-long workshop will cover setting up BinderHub and JupyterHub in the cloud. By the end, you will have your own public-facing BinderHub deployment that allows users at your own institution to share computational environments with a public link.
We will begin by deploying a bare JupyterHub on a commercial cloud platform using Kubernetes, a popular technology for robustly managing cloud resources. We'll cover best-practices in debugging, managing, and customizing your Kubernetes deployment, complete with hands-on exercises. We’ll show how to customize the environment your JupyterHub provides to users by using a Docker image of your choice. Next, we’ll cover how to set up a BinderHub service that allows users to automatically create their own computational environments. . We will also deploy the same infrastructure on XSEDE resources to demonstrate how easily this can be generalized to other cloud platforms.
We’ll close the workshop with a discussion about future directions of the technology and the community, as well as an open question: How can we federate BinderHub instances with support at national and institutional levels?