Jupyterhub: A barrier-free entrance to DKRZ

Jupyterhub: A barrier-free entrance to DKRZ#

With an account for levante, you can connect via ssh on a terminal interface to the system. This requires knowledge in shell programming. As a user friendly alternative, we offer you the jupyterhub web service, which, by the way, also allows you to run terminals as with ssh as well.

Features in combination with notebooks:

  • Software independent and interactive access to HPC resources

  • Tutorials prepared by experts and formatted as Jupyter-notebooks for a steep learning curve

  • Inline plot creation for a all in one solution

  • Fully supported and maintained at DKRZ

⇨ Advanced programming with a beginner friendly control

How it works:

What you have to do:

  1. You log in via: https://jupyterhub.dkrz.de/hub/login

  2. You request a notebook server

What DKRZ does in turn:

  1. A server is spawned as a slurm job on levante. It uses resources as specified by your request.

  2. Your server runs for as long as you specified in the request. In that time, you can close the tab and revisit jupyterhub without it is shut down.

Control Panel:#

  • Either work with jupyterhub or jupyterlab

  • Navigate with a file browser

  • Launch Notebooks or Consoles

Why do we promote Jupyter-Notebooks?#

  • Cells

    • Clear view: Mark Cells as Code, Markdown or Raw

    • Interactive: Run a cell individually without running the entire script

  • Code language: Python, but also Bash, R, Julia, Latex:

    • Use ! before the commands and bash runs it instead of the python interpreter

    • Define a entire cell as bash or latexwith the %% magic

[1]:

!ls

Analysis.rst                Usage.rst
CMIP6-Data-Pool.ipynb       _build
Curation.rst                _static
Dashboards.rst              backups-archive.ipynb
Demos.rst                   conf.py
FAQ.rst                     egest-service.ipynb
Find-and-Access.rst         globus-doc.ipynb
Ingestion.rst               index.rst
License-and-Citation.rst    intake-catalog-service.ipynb
Preparation.rst             intake-esgf.ipynb
Quality-Assurance.rst       intake-view.rst
Replication.rst             quality-assurance-cfchecker-ceda.ipynb
Server-side-analysis.ipynb  quality-assurance-prepare.ipynb
Server-side-resources.rst   services_overview.csv
Services.rst                statistics-cmip6.ipynb
Tools-and-platforms.rst     wdcc_dataset_tabulator.ipynb
Transfer.rst

[2]:

%%bash
ls

Analysis.rst
CMIP6-Data-Pool.ipynb
Curation.rst
Dashboards.rst
Demos.rst
FAQ.rst
Find-and-Access.rst
Ingestion.rst
License-and-Citation.rst
Preparation.rst
Quality-Assurance.rst
Replication.rst
Server-side-analysis.ipynb
Server-side-resources.rst
Services.rst
Tools-and-platforms.rst
Transfer.rst
Usage.rst
_build
_static
backups-archive.ipynb
conf.py
egest-service.ipynb
globus-doc.ipynb
index.rst
intake-catalog-service.ipynb
intake-esgf.ipynb
intake-view.rst
quality-assurance-cfchecker-ceda.ipynb
quality-assurance-prepare.ipynb
services_overview.csv
statistics-cmip6.ipynb
wdcc_dataset_tabulator.ipynb

[3]:

%%latex
\begin{align}
\frac{\partial u}{\partial t} + \nabla \cdot \left( \boldsymbol{v} u - D\nabla u \right) = f
\end{align}
\begin{align} \frac{\partial u}{\partial t} + \nabla \cdot \left( \boldsymbol{v} u - D\nabla u \right) = f \end{align}

  • Modules and extensions

    • python and jupyter benefit from a quickly evolving package universe

    • Embed html to the notebook

    • Automatic code formatting when executing cell

[4]:

from IPython.display import HTML, display, Markdown, IFrame
display(Markdown("Time series of three different data pool disk space measures. DKRZ has published about 1.5 PB, 2.5 PB are replicated data from other data nodes. An average CMIP6 dataset contains about 5 files and covers 4GB."))
IFrame(src="https://swift.dkrz.de/v1/dkrz_a44962e3ba914c309a7421573a6949a6/Pool-Statistics/pool-timeseries-hvplot.html",width="900",height="550",frameborder="0")

Time series of three different data pool disk space measures. DKRZ has published about 1.5 PB, 2.5 PB are replicated data from other data nodes. An average CMIP6 dataset contains about 5 files and covers 4GB.

[4]:

  • Kernels:

    • Predefined Software Environments can be set in the upper right by using a specific kernel

    • On Jupyterhub, you can choose between a hands full of different configurations

    • Customized kernels can be implemented with conda. See the docs for a tutorial

Conclusion:#

⇨ Jupyterhub is your laboratory in the web browser#

[ ]:
Contents