Web interface for Roihu

Intro

The web interface for Roihu at www.roihu.csc.fi can be used to access the supercomputer using only a web browser. A web interface for LUMI is also available at www.lumi.csc.fi, see the LUMI documentation for more details.

Please note that logging in to the Roihu web interface requires multi-factor authentication. More information on the Connecting page.

Scope

The Roihu web interface is best suited for interactive workloads that consume a modest amount of computational resources. Some examples are pre- and post-processing of data in Jupyter Notebooks using at most a few tens of CPU cores, small-scale AI/ML experiments using a single GPU, and data visualization tasks. As such, the maximum length of interactive jobs is limited to 16 hours.

Please note that the interactive applications in the web interface are not suitable for multi-node and multi-GPU jobs. Such workloads should ideally be carried out as standard batch jobs. If you are new to batch jobs and command-line usage of the HPC environment, we recommend the CSC Computing Environment self-learning materials to learn how to get the most out of CSC's supercomputers.

Available features

Features available in the Roihu web interface:
- View, download, upload and move files between Allas, the supercomputer and your local computer
- Open a shell on the login node
- Open a persistent shell on a compute node
- View running batch jobs
- View disk quotas and project status
- Launch interactive apps and connect to them directly from the browser:
  - Accelerated Visualization with applications such as Blender and ParaView
  - Desktop with applications such as Maestro and VMD
  - Julia-Jupyter
  - Jupyter
  - Jupyter for courses: An interactive Jupyter session specifically for courses
  - marimo
  - MATLAB
  - MLflow
  - R-Jupyter
  - RStudio
  - TensorBoard
  - Visual Studio Code

Shell

The Shell apps can be used to access the command-line of a supercomputer via the web interface. You can either open a connection to the login nodes, or a more persistent shell to the compute nodes. For more details, see the Shell page.

File browser

The Files app can be used to manage your files on the supercomputer and access storage services such as Allas and IDA. For more information, check the Files and storage services page.

Active jobs

Recent and running batch jobs can be viewed using the Jobs section in the top navbar and selecting Active jobs. Here you can view the current status of your jobs and what kind of resources were requested. Deleting a running job will cancel the job.

In the future, it may become possible to submit batch jobs through the web interface, but for now the recommended way to launch standard batch jobs is using sbatch from the shell.

Project view

Using the Project view under the Tools section in the top navbar, you can view current disk and project Billing Unit quotas on the supercomputers. For more information, see the Project view page.

Interactive apps

Interactive apps are programs that can be launched and run on the compute nodes and provide a graphical user interface. These are apps such as Jupyter Notebook, RStudio and Visual Studio Code. For a full list of applications and specific instructions, see the Interactive apps page.

If an interactive app does not start, or does not work as expected, you can delete the session and try to launch the app again. See also Troubleshooting.

Partitions and resources

Only a few partitions of Puhti and Mahti are available for use in the web interfaces. Some apps also have a more limited set of partitions available than others.
The maximum length of jobs in the web interface is 16 hours to keep the queueing times for interactive resources short. If you have work that requires longer jobs than that, we recommend that you run your software as batch jobs.

In the Roihu web interface, the interactive, small, test, gpuinteractive, and gputest partitions are available. In the Accelerated Visualization app, the vizinteractive partition is used, and will reserve one Nvidia L40 GPU. On the gpuinteractive and gputest partitions, one Nvidia GH200 GPU will be allocated. See the Roihu partitions page for general information about queues on Roihu.