3rd iHURT Meeting Spotlights HPC Usage in Industry

Keyvisual image main
Discussion session at the 3rd Industrial HPC User Roundtable at HLRS.

The Industrial HPC-User Roundtable focused on the computational challenges industry faces, with a special focus on the growing interest in solutions involving artificial intelligence.

In addition to enabling academic research, the High-Performance Computing Center Stuttgart (HLRS) supports industry by making its supercomputing resources for simulation, machine learning, and visualization available for research and development. To better understand and address the specific needs of industrial HPC users, SICOS BW and HLRS hosted the third annual Industrial HPC User Roundtable (iHURT) on December 3, 2019. The event offered a forum for HLRS and its industrial users to exchange perspectives on the state of the art in high-performance computing, specific computing challenges industrial users face, and how HLRS could help address them.


The program opened with presentations by industrial users of HLRS's HPC systems. Andreas Link of automotive supplier Mann + Hummel described his team's computational fluid dynamics (CFD) simulations, which they run at HLRS to optimize car component geometries and reduce noise caused by air flow around them. When combined with rapid prototyping technologies, the approach makes it possible to test new part geometries quickly. René Thümmler of CFD Consultants GmbH, a small consulting firm that assists companies in running CFD simulations, talked about his experience using HLRS computing resources, pointing to challenges the company faces with respect to data transfer and software licensing.

This year's iHURT meeting paid particular attention to growing interest in the integration of artificial intelligence applications and high-performance computing. Iris Pantle of a startup called Falquez, Pantle und Printz GbR described her company's goal of integrating machine learning and classical CFD simulation. The company is in the early stages of developing a cloud-based platform that would use machine learning to identify key parameters in CFD data sets, analyze simulation results, and optimize product development in an interactive, automated way. Alexander Thieß of Spicetech GmbH described how his company provides AI as a service for other companies interested in using machine learning to extract insights from the data they collect.

Representatives of HLRS and SICOS BW described recent developments at HLRS that are relevant to industrial HPC users. SICOS BW Executive Director Andreas Wierse provided an update on EXCELLERAT, a Europe-wide project led by HLRS that brings together academic researchers and industry to develop new and more complex software tools that industry will need as supercomputers approach exascale. Thomas Bönisch, who leads the HLRS user support team, introduced the HPC center's new flagship supercomputer, called Hawk, which is due to go online early in 2020 and will offer more than triple the computing power of its outgoing system, Hazel Hen. SICOS BW's Nicole Dobner provided an update on the Supercomputing-Akademie, a continuing professional education program involving HLRS, Ulm University, and University of Freiburg that is designed to address the specific needs of industrial HPC users.

Capping the meeting was an open discussion focusing on practical challenges that industrial HPC users face. Topics discussed included training, data management, workflow development and data transfer, access to software licenses, and the integration of HPC and artificial intelligence.

Christopher Williams