Skip to content

Getting started on galileo

Jump to bottom
Przemysław Juda edited this page Jun 21, 2021 · 3 revisions

If your galileo account has just been created, you will find here some information that will get you started.

IMPORTANT: Please change your password upon your first login using the command passwd

Connection

To connect to galileo, use ssh:

  • on Linux or MacOS, type in your command line: ssh <your_username>@galileo.unine.ch (substitute with your login)
  • on Windows, you need an application which allows creating ssh connections.

TIP for first-time Linux users: when typing your password on the cluster, you will not see any characters appear (or placeholders, such as *). It is expected and common for all Linux systems! Just type your password as usual and press Enter.

File transfer and quota

You have a quota of 200 GB in your home directory (/home/). Home is on a fast SSD drive. If you need to store big files, you can use the directory /server/, but do not run jobs from this directory.

To transfer files between the cluster and your local machine, we suggest:

  • sshfs for Linux/MacOS,
  • an application such as FileZilla for Windows.

CCFS architecture

The cluster is currently composed of the following machines:

  • master node, which runs the queue system (SLURM) and home server,
  • 10 dell nodes node01 - node10, 2x32 CPU each (2 of them 01 and 02 have 1024 GB RAM, the rest 256 GB RAM each), which are managed by SLURM (you should not connect to these nodes with ssh!),
  • 1 node with 4 GPU node-gpu and 2x22 CPU,
  • 1 node with 2x64 CPU node-epyc for interactive computing. The /home directory is shared among all these nodes.

To launch tasks (jobs) on dell nodes (node01 - node10), you have to use SLURM. On contrary, node-gpu and node-epyc are independent of SLURM. You can connect to node-epyc directly with ssh from the master nodes. Then, you can launch tasks from the command line. node-epyc serves principally for testing codes or interactive computing (such as notebooks), but you should not consume too many CPUs for longer periods of time, as the resource is shared among many users. As for GPU node, it is reserved for users who need to use GPU for their computations and have a suitable code. Should you require access to the GPU node, please contact the administrator.

There is a link to a simple schema of the cluster's architecture: schema

R

If you use R, consider using Rstudio which is available under the address: http://galileo.unine.ch:8787/. Use your CCFS (galileo) credentials to sign in. The backend is running on the epyc-node, and, consequenntly, Rstudio has access to your galileo home. Rstudio serves for interactive computing. Intensive jobs (multiple CPU over multiple hours) should be submitted via SLURM instead.

Happy computing, Your galileo admin

Clone this wiki locally