Introduction to running on AWS EC2

AWS EC2 via CloudBank¶

Amazon Web Services (AWS) EC2 provides on-demand virtual machines in the cloud. NSF-funded researchers can obtain AWS compute through CloudBank by exchanging an ACCESS allocation for cloud credits; CloudBank then provisions an AWS account. See the CloudBank documentation for the current steps.
The datasets used in these notebooks live in AWS Open Data buckets in the us-west-2 region (e.g. cmip6-pds, ncar-cesm2-lens), so launch your EC2 instance in us-west-2. Same-region access is fast, avoids egress charges, and makes the OSDF-vs-S3 comparison fair.
Launch an Ubuntu instance (m5.xlarge or m5.4xlarge suit these workflows) with an SSH key pair and inbound SSH (port 22) allowed from your IP.

: Connecting and setting up JupyterLab

Connect and forward JupyterLab to your laptop over SSH:

ssh -i your-key.pem -L 8888:localhost:8888 ubuntu@<public-dns>

Create a Python environment from this repository’s requirements.txt (a virtual environment or a conda env), then start JupyterLab:

jupyter lab --no-browser --port 8888

Open the printed http://localhost:8888/lab?token=... URL in your browser — the SSH tunnel makes it resolve to the instance.

With an EC2 instance running in us-west-2 and JupyterLab open, you are ready to run the AWS notebooks that follow: the CMIP6 GMST and ECS workflows, and the OSDF vs S3 streaming benchmark.