Containers for Rhinodiagnost

The Rhinodiagnost article “Web-access to HPC for Rhinodiagnost” presents reasons and efforts that led to the use of Jupyter as a multi-functional web interface for easy access to HPC resources in the Rhinodiagnost project. In order to use Jupyter in the project outside the HPC environment or without the necessity of an HPC clusters, the solution is transferred into self-sufficient containers in the second step.
Containers (e.g. with the help of the container virtualization solution Docker, allow the isolation of applications, so that several instances even on the same host can be operated completely independent from each other and independent from the host system. Even setups in which different applications have to interact with each other in a complex way in order to be perceived as a web application can be installed and operated quickly and easily as containers.
At JSC and for the Rhinodiagnost project, the Helmholtz Data Federation (HDF) Cloud ( currently offers an ideal environment for HPC-related web services in the form of container solutions. The HDF Cloud is an OpenStack cluster ( with high availability, which enables the execution of Rhinodiagnost Docker Images, is accessible worldwide, and yet is very close to the HPC resources of JSC.
This proximity makes it possible to store large amounts of simulation data on the HPC storage systems (called XCST in the figure), i.e., data can therefore be stored where it was generated. The HDF cloud is directly connected to the XCST at 40 GB/s, so that the data can be accessed from the container with high I/O performance if required. In the Rhinodiagnost setup GitLab was additionally added as a flexible method (or interface) to make simulation results worldwide accessible.
Furthermore, the container solution on the HDF cloud allows its own user administration independent from HPC operation. Thus, access to Rhinodiagnost web services can be granted without accessing the HPC systems at the same time. In addition, different HPC systems and different computing time accounts can be combined via the web services.