Passion for Training and Education
From early on in his education, Volker Weinberg had eclectic learning interests—he began learning Japanese with eight years old, developed a passion for training and education working for the Munich Volkshochschule (adult education centre) during his year of civil service between high school and university, got inspired to study physics by similarities between the Asian worldview and modern quantum field theory, and found an interest in scientific computing shortly after beginning his university studies.
Ultimately his passion for educating others has proven to play the major part in his role at the Leibniz Supercomput-ing Centre (LRZ) in Garching near Munich. As the HPC Training and Education Coordinator at LRZ, Weinberg uses not only his background in elementary particle physics re-search, but also his deep interest in training and education to help coordinate and bolster the LRZ and Gauss Centre for Supercomputing (GCS) training programs, ensuring users are making the best use possible of their allocations on GCS supercomputing resources.
“From early on, I was interested in education and training in general,” Weinberg said. “Even during my civil service year at the Munich Volkshochschule, I was involved in the course program organization for seniors and did some teaching myself.” As a physics student at the Ludwig Maximilians University of Munich he worked for the faculty’s computer cluster, leading to many interactions with IT experts at LRZ.
Weinberg’s interests in physics, computing, and education merged during his PhD studies at the Free University of Berlin (FU) and the research center DESY. In his research Weinberg explored the vacuum structure of lattice quan-tum chromodynamics (LQCD) a theory focused on using advanced numerical simulations to understand how quarks and gluons, as building blocks of matter, behave and inter-act. His research team had large allocations on the super-computers at the Jülich Supercomputing Centre (JSC) and LRZ, and a lot of expertise in high-performance computing (HPC). Not only did LQCD help make Weinberg aware of the world of HPC early in his career, but it also informs how he approaches the GCS training program. “I think it is re-ally important for someone responsible for coordination of the training program to go through all those difficulties of highly demanding and challenging real-world simulations on different platforms in order to understand the needs of our users and offer courses on relevant topics,” he said.
Between his experience with cutting-edge LQCD simula-tions at DESY, teaching undergraduate students in com-puting at the FU, and his time organizing adult education, Weinberg sees value in having education programs touch on a variety of knowledge levels and skill sets, and he brought that view to the GCS and LRZ training programmes. He pointed out that the three GCS centres have a wide variety of educational offerings, and that there is strength in that diversity.
As a hosting member of the Partnership for Advanced Com-puting in Europe (PRACE), GCS offers advanced training courses for European researchers looking to further refine and improve their HPC knowledge and code performance. On the other hand, the individual centres partner with local universities and industries to regularly host workshops aimed at training working professionals how to incorporate simulation into their workflows.
Weinberg also noted that while staff members at the three centres excel at providing training for their own unique architectures on their respective leading (Tier 0) HPC sys-tems, a lot of training focuses on parallelizing and optimiz-ing codes for HPC environments generally. This encourages the GCS centres’ training staffs to collaborate and coordi-nate in their training activities. Weinberg especially enjoys the fruitful collaboration in training within GCS, but also with other European supercomputing centres, especially in Austria, Czech Republic and Finland.
There is more to a successful HPC training program than just making sure users are prepared to use a centre’s current supercomputer, though—HPC centres want ensure that users can navigate and take advantage of new and emerging technologies as well. The GCS centres are at the frontline of emerging technologies’ developments, including the rise of artificial intelligence (AI), machine learning, and data ana-lytics applications as well as relatively new HPC architec-tures, such as those employing GPUs or other accelerators.
“In the last two years, the rise of AI, machine learning, and data analytics has been one of the biggest changes in how we think about HPC training,” he said. “There is an immense interest from our users in these fields. Four LRZ colleagues, including myself, were recently certified by NVIDIA as University Ambassadors to teach GPU programming and deep learning. When we offer courses on these topics, we usually get more than 100 responses. These topics have gotten so hot that we are regularly being invited to HPC centres and universities internationally.”
While it is important to grow and respond to user needs related to training for new and emerging technologies, Weinberg also emphasized that one of the most important aspects of continued success for the GCS training program revolved around sticking to parallel computing funda-mentals. “Emerging technologies, they are always coming and going,” he said. “Having well-established courses on programming standards like MPI and OpenMP, while also following current trends, is really important. Even with new and emerging technologies, it is important to adapt these programming models for these new architectures.” Recently he became LRZ representative in the OpenMP architecture review board (ARB) and language committee to actively influence the future of the OpenMP standard in this direction.