Dardel Status

In the last issue of the PDC Newsletter, it was mentioned that PDC had a test rack with a different cluster manager from Hewlett Packard Enterprise (HPE) called HPCM ( www.pdc.kth.se/publications/2023-no-2/dardel-status-1.1291982 ). HPCM is a less complex and more stable cluster management system that has worked well on other cluster installations for many years. PDC considered the results of tests with the rack to be promising and decided to convert Dardel to use HPCM as this change should make it easier to keep the entire Dardel software stack up to date while maintaining stability.

The last step in the conversion was to shut down the old system and install a more modern software stack under HPCM. At the time of writing, this process was in its final steps. Almost all system software has been updated, including the Cray Programming Environment, the Slingshot software, the Linux software on the compute nodes and the AMD ROCm package. This means that all applications executing on multiple nodes must be recompiled. PDC has already installed the applications that are used most.

The package for the remote graphical desktop system ThinLinc is also being updated. PDC is installing a new version of ThinLinc, as well as fixing several of the problems with the previous installation. For users with less experience of Linux, ThinLinc is an easier alternative for logging in with SSH and using Linux shell commands. ThinLinc allows users to interactively perform data visualisation and simulation in a virtual desktop environment, thereby minimising the need to offload data from Dardel for simulations. With this upgrade, PDC will be offering support for additional applications, such as JupyterLab, Jupyter Notebook, MATLAB, Ansys Fluent, and VIAMD amongst others.