The global race towards the establishment of exascale computing will not be decided by the machine showing the highest peak performance. Rather, research activities which aim at a complete supercomputing ecosystem will succeed.

Thisi is a guest blog post writen by Prof. Dr.-Ing. Wolfgang Marquardt, Chairman of the Board of Directors of Forschungszentrum Jülich, Germany

Supercomputing in Europe

Europe has always been a leader in many aspects of supercomputing, in particular in computational science and engineering with an emphasis on software, algorithms, models, and applications. This has been possible due to the successful collaboration between different scientific communities including biomedicine, brain research, fundamental physics, astrophysics, geosciences, materials science, economics and finance, or engineering. A major enabler of this success story is PRACE, the Partnership for Advanced Computing in Europe, whose members invest in supercomputing facilities and support user groups from very different scientific disciplines. The outstanding reputation of supercomputing in Europe comes from an excellence-driven peer review system as well as simulation and data labs. Many examples of ground-breaking scientific results facilitated by supercomputing in the basic and applied sciences could be quoted.

Information infrastructures on the rise

However, a different quality is currently evolving: the development and dissemination of information infrastructures. They are enabled by frontier supercomputing hosting diverse simulation and data management and analytics capabilities. A prominent example is the Human Brain Project. It uses infrastructure platforms to accelerate and facilitate scientific discovery to decode the function of the human brain. This basic understanding can then be translated into clinical and technological applications.

European processor technology missing so far

However, the strong European supercomputing community also faces a weak spot in the suite of technologies that make up an HPC ecosystem: the development and commercialization of domestic computer chips, processor technologies and computer systems. A competitive European processor technology could be built in an ambitious pan-European research effort to catch up with technology leaders outside of Europe. Indeed, there seems to be a fair chance, in particular if we joined forces with European technology providers for mass markets driven by autonomous mobility or industry 4.0. Furthermore, emerging disruptive basic technologies such as low-temperature CMOS, optical interlinks, memresistors, neuromorphic chips, or even quantum processors, offer new opportunities for industrial development.

Use the best available technology

The roadmap towards a European HPC ecosystem with exascale capabilities should, however, not become dependent on the development of domestic technology. Rather, a world-class European HPC ecosystem in the spirit of Open Innovation has to be built on the world’s best supercomputer technology. It may not always be available from European providers. A delicate balance between capacity and capability building has to be established – most likely on different time scales – in all relevant areas of supercomputing, from basic technologies to products and services, to achieve a global leadership position.

The importance of building a European supercomputing ecosystem

Why is it so important for Europe to take up the challenge and enter the demanding competition in building up a powerful supercomputing ecosystem? It is widely agreed that scientific progress as well as economic growth will increasingly rely on top level HPC-enabled methods and tools, services and products. Digitalization of science itself, of commercial services and products, of the health system of society as a whole, will at least draw on or even be driven by HPC-empowered simulation and data analytics. These enabling key technologies will accelerate scientific discovery and will also significantly contribute to value creation in the European economy. In 2015, a DOE-funded study meticulously analyzed several hundred concrete projects in the United States. The study revealed an approximate profit of $43 per dollar that was invested in HPC. This impressive result was found before big data analytics and artificial intelligence started to be rolled-out broadly in industry in recent years. There are enormous expectations regarding the pace of growth in revenues, profits and jobs driven by HPC-enabled technologies. We already see disruptive changes in business models: Traditional businesses vanish; new and unexpected ones suddenly appear. Hence, Europe needs to get into the driver’s seat on a global scale by combining resources – human and financial – to overcome its fragmentation and the dilution of efforts. The Jülich Research Center with its Supercomputing Centre as well as its information technology and neuroscience branches is willing to contribute significantly to the generation of a globally competitive, leading European supercomputing ecosystem in the EuroHPC effort.