As part of our “Inside Quantum & HPC Minds” series, we have interviewed Hans-Christian Hoppe, Senior Technical Project Manager at ParTec. Hans-Christian discusses his pathway to HPC and his motivations for joining ParTec and Jülich Supercomputing Centre (JSC) in 2022. He also shares the most interesting results of the DEEP-SEA project, the most exciting initiatives ParTec is currently involved in, and the future of HPC. Hans-Christian also offers valuable advice for those considering how to start their career in HPC.
Hans-Christian Hoppe
Senior Technical Project Manager
As a young student of Computer Science at the University of Bonn, Prof. Trottenberg, one of my mathematics professors stepped up to lead a large government funded project to build a German supercomputer (SUPRENUM) and hired me as a student assistant. This was in the very early days of highly parallel, clustered systems, and the opportunity to get involved in a rapidly progressing technology filed got me “hooked”.
Allow me to share a bit more about my professional background – after working for a German government lab on HPC communication libraries, I spent a decade with a small independent software vendor (ISV). There I developed key system software and performance analysis tools for nearly all HPC hardware players active at the time. I led European R&D projects on scalable HPC libraries, tools and applications as well as federation middleware (at that time called “Grid computing”). Due to our work with the US-side DoE labs, Intel became interested and acquired the ISV. I spent the next two decades at Intel driving key HPC SW development activities and branching out into system architecture and computer graphics plus early AI projects.
At Intel, my team was instrumental in the DEEP projects, fostering close collaboration with Jülich Supercomputing Centre and ParTec and pitching advanced ideas for future Intel HPC systems.
Around 2020, as Intel’s focus shifted away from HPC as a leading edge CPU and system market, I sought new opportunities closer to actual scientific research and HPC use. Thus, when the opportunity arose to co-lead the DEEP-SEA project and contribute to ParTec’s success, it was an easy decision to make.
Since I was deeply (no pun intended) involved in all the DEEP projects, I can confidently say that the key outcome of this project series is the design and validation of a European architecture for current and future heterogeneous systems in the HPC, AI, and Quantum computing domains—the “Modular Supercomputing Architecture.”
With the DEEP-EST and DEEP-SEA project results, we now possess an integrated SW stack capable of operating heterogeneous HPC systems with highest efficiency, supporting dynamic resource management, orchestration and scheduling.
DEEP-SEA has been instrumental in providing excellent programming model and tools support, and is leading the way for a successful installation and operation of the first European Exascale system JUPITER. And the story does not end there – the MSA facilitates the integration of Quantum computing resources with HPC infrastructures and usage models. It is a key component for developing and deploy systems that combine leading-edge HPC and AI capabilities with maximum capital and energy efficiency.
The rapid advancements in AI are fundamentally transforming the computing landscape, from personal and edge devices to the largest computer installations. HPC is in a very interesting place here – its numerical methods and computational capabilities drive large-scale AI, while AI methods promise to augment or even replace traditional HPC-centric, simulation-based approaches for solving key scientific, industrial, and societal challenges.
This opens up immense opportunities in system architecture and design, system software and programming models, and of course applications. Research and development in Quantum Computing is making impressive, steady progress, and will in due time lead to leaps-and-bounds progress for important use cases, such as optimization problems. One of the key challenges today is to prepare integration approaches and SW stacks/applications to intercept the progress in Quantum Computer systems and be poised to support impactful and exciting use cases as performance and reliability requirements are reached.
Disruptive approaches like AI or Quantum Computing quickly open up avenues for rapid progress, which are just not there in the constant, incremental improvement of established HW or SW technologies.
HPC will continue to be a cornerstone of scientific progress, the development of superior technical products, understanding of the severe environmental and societal challenges we face today, as well as identifying effective solutions. Its value will increasingly lie in its ability to generate insights into and propose solutions for important problems with a minimal carbon footprint, rather than merely achieving higher aggregated performance for rarified benchmarks. Clearly, it will bring maximum benefits in combination with AI approaches now and Quantum Computing in the near future.
If at all feasible, study AI and Quantum Computing methods alongside “classic” HPC techniques, and seek out beneficial combinations of these approaches.
If you’re interested in Mathematics, consider exploring “old-fashioned” topics like numerical stability and error propagation, since use of reduced size datatypes in numerical algorithms is a key vector to reducing energy use. If your interest lies in device and system architecture/design, be sure to always consider how to effectively use exciting new architectures or devices to solve real-world problems. Without matching algorithms and software, even the best hardware ideas are likely to fail.
