Creating a scalable, cost-effective JupyterHub implementation on Amazon AWS

At Harvard, we deployed JupyterHub on Amazon AWS for two classes in School of Engineering. The Signal Processing class used Docker-based JupyterHub, where each user provisioned with a docker container notebook. For the Decision Theory class, we redesigned JupyterHub using a dedicated EC2 instance per user’s notebook, providing better scalability, reliability and cost efficiency.