10. Interactive Web Servers on the LRZ AI Systems

It is possible to start interactive web servers utilizing the resources of the LRZ AI Systems. Currently supported are Jupyter Notebook, JupyterLab, RStudio Server and TensorBoard environments.

To make use of this, access to the LRZ AI Systems has to be requested (see Access and Getting Started) and a successful login at the web-based frontend https://login.ai.lrz.de is necessary.

In order to then e.g. start a RStudio Server environment, select "Interactive Apps" -> "RStudio Server" from the top panel.

For a typical use-case

  • select the type of resources (CPU only or CPU + single GPU)
  • specify your workload (a combination of CPU core and RAM requirements)
  • select the container environment you want to work with - this provides a specific version of R and RStudio (every session is containerized, i.e. the information provided here: Introduction to Enroot: The Software Stack Provider for the LRZ AI Systems does apply; additionally, see Managing R Packages in a Containerized Environment)
  • finally, specify the number of hours you plan to work with the RStudio Server environment (be aware that your session will be shut down when this time limit is reached, any unsaved work will then be lost).

When everything is set, press "Launch". You will then be re-directed to an overview of your sessions (this can also be accessed by selecting "My Interactive Sessions" from the top panel). Here you find a card representing the session you launched. Given the requested resources are available, the status of the session will change from "Queued" to "Starting" and finally "Running". This should typically take no more than 2 minutes. Once the session has started up properly, a button "Connect to RStudio Server" will appear. Clicking this button will open the RStudio Server environment in a new browser tab.