You are here: Home / News & Events / DKRZ Tech Talks / Tech Talk: Jupyterhub

Tech Talk: Jupyterhub

Jupyter Notebooks have become increasingly popular among different fields of science. At the DKRZ, we provide different solutions to launch notebooks on our Supercomputer Mistral. In this talk, Sofiane Bendoukha covered access and log in to the service, launching Jupyter notebooks, switching/enabling kernels and interfaces and more.
  • Tech Talk: Jupyterhub
  • 2020-09-01T15:15:00+02:00
  • 2020-09-01T16:15:00+02:00
  • Jupyter Notebooks have become increasingly popular among different fields of science. At the DKRZ, we provide different solutions to launch notebooks on our Supercomputer Mistral. In this talk, Sofiane Bendoukha covered access and log in to the service, launching Jupyter notebooks, switching/enabling kernels and interfaces and more.
When
Sep 01, 2020 from 03:15 PM to 04:15 PM (Europe/Berlin / UTC200)
Add event to calendar
iCal

JupyterhubJupyter Notebooks have become increasingly popular among different fields of science.  A Jupyter Notebook provides a flexible and user-friendly interface to perform interactive (super) computing and also enables reproducible science.  This includes live code, equations and visualizations. It supports several languages like Python (IPython), Julia, R etc.

At the DKRZ, we provide different solutions to launch notebooks on our Supercomputer Mistral: ssh scripts or centralized services. In this first talk we are going to introduce the Jupyterhub service at the DKRZ, more precisely the newly deployed release. Jupyterhub is a multi-user server that allows spawning single-user notebook servers via a web browser. In the talk, we covered the following items:JupyterHub

- access and log in to the service
- launch Jupyter notebooks
- switch/enable kernels and interfaces
- and more

More on the new server at https://jupyterhub.gitlab-pages.dkrz.de/jupyterhub-docs/

Find the full talk here: https://youtu.be/f0wZX9i0uWQ

 

The notes of the Q&A Session are:

  • Does the same update routine as is used for python applied to other systems like R or Julia? Meaning latest R distribution and half yearly versions. 

    • No, this currently only holds for python.

    • They do get regular updates, tough.

  • It is really great to have your jupyterhub service! What NB extensions do you recommend most for science workflows?

  • Is there an easy way to just use the existing conda environment as a kernel? Without creating a new one as in the example. Say, I have one already and want to use it. 

    • If you already have a python environment, you need to install the ipy kernel, and then you can create the kernel.

  • I did not understand what the “Binder” is needed for?

  • Is it planned to allow for multi-node workflows? Say dask that uses several nodes. But not from the interface now?

  • Is it possible to jointly work together on a notebook? Ok, thank you!

There is currently no native jupyterhub feature to allow this. There are some external services that claim collaboration on jupyter notebook, but unfortunately not in real-time. One of them is https://cocalc.com/policies/pricing.html. We will take a look at it and see if we can deploy it at DKRZ. A solution a la Google doc is not yet available.

 

Document Actions