Open OnDemand provides a browser-based interface into several graphical packages on the FRCE cluster. Applications such as Jupyter, RStudio, and VS Code can by run interactively without logging directly into FRCE. A full-fledged Linux desktop using the Xfce window manager can also be used. The applications will run on one of the FRCE CPU or GPU servers and it is possible to submit further slurm jobs from within the application.

Documentation

Open OnDemand home
Resouces and Videos from sites running Open OnDemand
Video on using OnDemand on Biowulf (mostly applicable)

Available Applications

Jupyter
Shiny
R Studio
MATLAB
VS Code

Other applications may be requested by sending an email to the FRCE administrators. While being able to launch an app directly from OnDemand is convenient, it may be just as easy to open the Linux desktop and run the app from a terminal window.

FAQ & Tips

  • Having the OnDemand web page displayed does not consume any resources from the FRCE cluster; this occurs only when you start a session. However, as with running interactive sessions from the FRCE command line, we ask that you stop and delete sessions that are either completed or expected to be idle for some time so that the resources are freed for other users.
  • Multiple OnDemand sessions may be run simultaneously. Resource limitations are the same as for batch scripts submitted through the command line.
  • When GPUs are allocated for the session, the applications are already configured to take advantage of the GPU's capabilities. The exception to this is when running the Linux desktop. When running a GPU-enabled program from the terminal command line, the command must be preceded by vglrun. Without this command, the program should run with no warnings or errors but it will execute on the CPU only.

Usage

The OnDemand application is hosted at https://ondemand.ncifcrf.gov and can be accessed using any web browser while on site or connected through VPN. NIH Login is used to authenticate to the service. The site's home page presents a menu bar allowing one to select specific apps to run and check on the status of running apps. A few popular apps are also pinned to the home page for quick access.

Image
On Demand home page

A pinned application can be chosen directly from the list of pinned apps or selected from the Interactive Apps dropdown. Once the app is chosen, you will be presented with several run-time options. Many of these are common to all applications, such as the number of core and the allocated time. Others are application-specific, such as choosing the version of R or Python to use. After the task is launched, the screen will refresh with a message that the session is being launched. Once the process is scheduled with slurm, some status information is presented and a button to connect to the session is available. It is possible to disconnect from a running session by closing the browser tab and re-connect using this screen.

Image
R Studio parameter selection
Image
R Studio start process

After connecting to the session, the new tab will appear very similar to what you would see if the application was running natively in a GUI window. Applications that may need to open new windows will run under a limited Linux desktop, others will run as a stand-alone package.

Image
R Studio initial run screen
Image
R Studio executed script