Multiuser Cluster Material

Overview

It has been about 350 years that separate the original 1673 Leibniz mechanical calculator from today's Leibniz Supercomputing Centre's (LRZ) facilities located on the Campus Garching. And yet, the spirit has not changed. To quote the German mathematician:

"It is beneath the dignity of excellent men to waste their time in calculation when any peasant could do the work just as accurately with the aid of a machine."

This course module will allow participants to live up to their dignity by providing a comprehensive walkthrough and usage guide to such contemporary types of these machines that potentially fill whole buildings.

In a general overview, historical and current developments and trends in the space of scientific computing and cluster systems will be presented. This will, amongst others, address the following questions: How do modern cluster systems work and how are they architected? How did we come to High Performance Computing, High Performance Data Analytics and High Performance AI? What makes a system adequate to a specific workload? How are these systems operated and how are they made available to their users?
In addition, typical interaction methods and usage patterns will be covered, including various possibilites of setting up user environments (e.g. environment variables and modules, user space package managers, containers) as well as tools for resource allocation (i.e. Slurm Workload Manager) and efficient parallelization (MPI, OpenMP, ...). Finally, an overview of different compute clusters as well as their background storage systems operated by LRZ will be provided. The requirements for acquiring access to these systems will be covered as well.

Participants will gain a good understanding of the characteristics of multiuser cluster systems in general and will practise basic methods of typical interaction. They will familiarize themselves with the landscape of cluster systems available at LRZ and this will allow them to choose the right system for their own compute projects.

Available Material

Corresponding person

Questions and comments can be addressed at Florent Dufour: florent@lrz.de