Europe Builds World-Leading Supercomputer

The European High-Performance Computing Joint Undertaking (EuroHPC JU) has invested 144.5 million euros to build a European supercomputer called LUMI (Large Unified Modern Infrastructure). With more than 550 peak petaflops, LUMI is faster than the current number 1 in the world, the Japanese Fugaku supercomputer, and will take European competitiveness and research to a new level. The LUMI system is supplied by Hewlett Packard Enterprise.

Supercomputers need a lot of energy. LUMI is hosted at CSC’s data center in Kajaani, Finland, and will use only hydroelectric power. It will also provide 20 % of the district heating needs of the area.

Together with the other EuroHPC pre-exascale and petascale supercomputers, which will become operational in 2021, the LUMI supercomputer will help Europe’s public and private users to address research and innovation challenges in a variety of areas, from weather and climate change to cybersecurity, drug discovery, and personalized medicine. A fifth of the supercomputer’s capacity will be made available to industry and SMEs. A mission-critical fast-track capacity will also be integrated into LUMI’s operations. The awareness of this need has been further increased by the COVID19 pandemic.

The participants of the LUMI consortium are Finland, Belgium, the Czech Republic, Denmark, Estonia, Iceland, Norway, Poland, Sweden, and Switzerland. The word “Lumi” means “snow” in Finnish.

LUMI will be based on Hewlett Packard’s HPE Cray EX supercomputer. Its capability of 552 x 10¹⁵ floating-point operations per second (flops) will represent a significant step towards exascale computing, which refers to the capacity to compute 10¹⁸ flops and is the estimated processing power of the human brain. LUMI will pack a storage of 117 petabytes and be configured with a bandwidth of 160 terabytes per second. This is enough to accommodate the entire global internet traffic twice. LUMI will be capable of executing 375 petaflops, or more than 375 million billion calculations per second, with a theoretical maximum performance of more than 550 petaflops per second, equaling that of 1.5 million laptops.