Getting started on gadi

The National Computing Infrastructure (NCI) hosts the supercomputer gadi where you will be conducting most of your research. Through NCI, you will have access to high-performance computing (HPC) resources and large-scale data storage, including some of the most used datasets like ERA5 and CMIP6 data.

In this section, we will guide you through the steps to get started on gadi using some handy tutorials developed by ACCESS-NRI.

1. Setting up an account on gadi

To access gadi, you will need to set up an account through NCI. Follow the instructions here to create your account. You will need a institutional email and the name of the projects you need to join. Projects gives you access to computational and disk storage resources on gadi. 21st Century Weather has 5 projects you can join:

NCI Project	Theme	Centre Projectcs	NCI Project Lead	SU/quarter + storage
gb02	Centre-wide Strategic Projects	All	gb02 Panel*	2.25 MSUs + 80 TB Storage
fy29	High-Resolution Modelling	Modelling	Bethan White	1.25 MSUs + 20TB Storage
if69	Circulation Change	Weather System Dynamics, Variability & Warmer World,	Chenhui Jin	1.25 MSUs + 20TB Storage
ng72	Weather Change	Weather Resources & High Impact Weather	Andrew Brown	1.25 MSUs + 20TB Storage
su28	Datasets for the Centre	All	Sam Green	20 TB Storage

*gb02 panel: Navid Constantinou, Paul Gregory, Bethan White, Chenhui Jin, Andrew Brown, and Sam Green

If you are unsure which projects to join, please ask on Cumulus.

Accessing gadi via a terminal.

To connect to gadi via the terminal we use ‘Secure Shell Protocol’ or ssh. Even if you are not planning to use the terminal often, it is useful to know how to connect this way. The steps bellow will also help you if you want to use VSCode with gadi.

Follow the instructions in the section Logging into Gadi. You will need your password.

Logging to Gadi without a password

Typing your password every time you want to log into gadi can be tedious. To avoid this, you can set up ssh keys to log in automatically without a password. ssh keys are a pair of cryptographic keys that allow secure access to remote systems. These keys are stored in your local machine and the remote server (gadi in this case) in a folder called .ssh/ (it is a hidden folder!).

Follow the instructions in the section Automatic login to create your ssh keys and copy the public key to gadi. Once you have done this, you should be able to log into gadi without a password.

Using the Australian Research Environment (ARE)

NCI have created a set of web-based graphical tools to help perform your computational research. These tools (including a Virtual Desktop and the JupyterLab environment) are made available via the ‘Australian Research Environment’ (ARE) on gadi.

To access ARE, you will need to use your NCI credentials (the ones you used to create your gadi account).

Follow the instructions in the ACCESS-NRI documentation to start an ARE session. The tutorial show how to create a:

• Virtual Desktop Interface (VDI)

• JupyterLab session

but there are other option available too.

Note

Any ARE session will use service units or SUs, which are the currency used on gadi to allocate computational resources. Each project you join will have a certain number of SUs allocated to it per quarter. Small JupyterLab sessions will use very few SUs, but it can add up over time during the quarter. Please always check your SU usage via the Computational Resource Dashboard and stop any ARE sessions you are not using.