After the official contract signing yesterday, configuration details were made public for the new BullSequana system that the Finnish IT Center for Science (CSC) is procuring from Atos in two phases over the next year-and-a-half. CSC is the inaugural customer for Atos’ BullSequana XH2000 supercomputer, announced at SC18 and the first Atos system to support AMD processors. Finland is investing 37 million euros in new high-performance computing infrastructure and another 2 million euros for skills development.
“The acquisition increases CSC’s customer base, as the new hardware will be used not only for universities and polytechnics, but also for academic use by research institutes,” said CSC on its website. “The acquisition ensures the international competitiveness of the Finnish research community in data and computational areas of research.”
Once completed in 2020 at CSC’s datacenter in Kajaani, the combined resources will comprise 11 petaflops of theoretical performance and will provide five times more computing capacity for Finnish scientists from what was previously available, accelerating climate research, quantum mechanics, life sciences, fusion energy and many other domains. AI-relevant computing is an expanding focus.
Expected to be made available in the summer of 2019, the phase one air-cooled BullSequana cluster features Mellanox HDR InfiniBand and Intel Cascade Lake Xeon processors that altogether provide 2 petaflops of theoretical performance. Node memory ranges between 96 GB and 1.5 TB. DDN has been tapped to provide a 4.9 PB Lustre parallel file system. A portion of the system (both compute and storage) is allocated to the Finnish Meteorological Institute.
CSC will also field a 12 PB data management solution, based on CEPH object storage.
Under the second phase, scheduled to be ready in the spring of 2020, Atos’ liquid-cooled, HDR-connected BullSequana XH2000 supercomputer will extend aggregate computing capacity by 6.4 (peak) petaflops, owing to approximately 200,000 AMD Epyc “Rome” cores. The system offers 256 GB memory per node. DDN will supply an 8PB Lustre parallel file system.
Researchers see the new resources and capabilities as critical for pushing the bounds of science in ways that expand our knowledge and better our lives.
“We have together with my research group developed Vlasitor, the world’s most detailed model for simulating conditions in space,” said Professor Minna Palmroth, University of Helsinki. “The new supercomputer enables six-dimensional modelling and with that, we will be able to shed light on the most complex problems of astrophysics that have remained elusive for decades.”
“In the field of nano science, one of the key challenges in the near future will be understanding the boundary between nano materials and biological materials in terms of structure, dynamics and properties. The increased computing capability provided by the system will enable us to find more in-depth answers to these questions,” stated Hannu Häkkinen, Academy Professor, University of Jyväskylä, Häkkinen.
“CSC’s new supercomputer will enable us to take a new approach to cancer and antibiotic drug development. We can now look at drugs and their target proteins as a dynamic system. This will enable us to consider completely new targets and ultimately develop more efficient and better medicine,” said Professor Antti Poso, University of Eastern Finland and University of Tübinge in Germany.
“The whole Finnish society, including transport, authorities, energy, business, citizens and media, is utilizing the services of the Finnish Meteorological Institute. With the new supercomputer resource, which we have acquired in cooperation with CSC, we will be able to develop new methods for modelling weather, sea conditions, air quality and climate. The increased capacity enables Finland to participate in international climate scenario calculations, which will provide research data for country level decision making and mitigation planning in addition to IPCC reports,” said Sami Niemelä, director of the Finnish Meteorological Institute.