The supercomputing infrastructure at CSCS (Swiss National Supercomputing Centre) in Ticino is sustainably operated as a scientific user laboratory. ETH Zurich is thus continuing to implement the ETH Board’s National Strategic Plan for High-Performance Computing and Networking (HPCN Strategy) at CSCS on behalf of the Swiss Confederation.

Bone implant used in case of osteoporosis. The current supercomputer of the CSCS simulates the mechanical stresses exerted by an implant on the bone tissue and provides knowledge that is used to design new models. (Illustration: Institute of Biomechanics, ETH Zurich)

High-performance computing (HPC) is a new key technology for both science and business. It complements theory and experiment and allows for new approaches to complex research questions. The ETH Board developed the national HPCN Strategy for Switzerland on behalf of the Confederation in 2007. Subsequent to resolutions adopted by the Federal Council and Parliament, the first two stages (2009–2012, 2013–2016) rested on a sound foundation totalling 172.5m CHF. The HPCN Strategy is now already engaged in its third stage, which lasts from 2017–2020.

Third stage (2017–2020): preparation for the exascale era

At the CSCS and in Switzerland, plans are already being drawn up for the coming exascale era, which is expected by 2020. In the future, exascale computers will be capable of billions by billons of calculations per second. In the third, 2017–2020 stage of the HPCN strategy, it is particularly the development and optimisation of codes that is being advanced in order to ensure that future computer architectures can be exploited better while still remaining energy-efficient. 12m CHF is available for this, with 80m CHF expected for the next hardware generation.

Successful implementation of the first two stages

During the first two stages of the HPCN Strategy the CSCS, under the aegis of ETH Zurich, developed an infrastructure which enables scientists to conduct research with the help of computer simulations at the highest level. Since computing power owing to new technologies, and the supercomputers’ energy efficiency, are continually improving on an exponential scale, existing supercomputer infrastructures had to be continually adapted and retrofitted or replaced. For the CSCS User Lab, this entails annual investment costs of approx. 20m CHF. ETH Zurich invests the same amount in operation costs on an annual basis.

In order to be able to use most cost- and energy-efficient computer architectures at any time, investments were made in the renewal of user codes from the very start. To date, an annual 3m CHF has been earmarked for this, with the same amount being contributed by the groups of researchers involved. To begin with, these projects were conducted within the scope of the Swiss Platform for High-Performance and High-Productivity Computing (HP2C). In the summer of 2013, HP2C was replaced by the Platform for Advanced Scientific Computing (PASC), in which the CSCS closely cooperates with hardware manufacturers, researchers, application software developers, mathematicians and IT experts. HP2C and PASC did not only provide the fertile soil for an annual conference which by now is well-known beyond European borders and registers approx. 400 participants; rather, they also and primarily established a network of cooperation ventures both at home and abroad.

International network

In the wake of the HPCN Strategy, Switzerland and the CSCS also established themselves in the international environment of high-performance computing. Since 2007, the CSCS has been a member of PRACE (Partnership for Advanced Computing in Europe), and since late 2016, it has also been a hosting member with its flagship supercomputer, which at present has a top performance of more than 25 petaflops (as at spring 2017). Membership of PRACE provides Swiss researchers with access to various European systems, and researchers from Europe with access to Piz Daint.