The World’s Fastest European Technology based Supercomputer for CINECA and Europe

The World’s Fastest European Technology based Supercomputer for CINECA and Europe

New “Leonardo” System designed by Atos integrates the Modular Supercomputing technology developed by ParTec and the services provided by E4 Computer Engineering to Power Italian’s and European’s HPC and AI Research, Cineca announced at SC20

Bologna -  Nov 18, 2020 - CINECA  the Italian inter-university consortium - one of the European and world’s most important supercomputing center - has selected Atos’ innovative BullSequana XH2000 supercomputing platform, integrated with ParTec’s ParaStation Modulo software for its new supercomputing system, named Leonardo. As part of EuroHPC, Leonardo will fully designed and manufactured in Europe to deliver world-class supercomputing capabilities to support HPC and AI Research in Italy and Europe, while E4 Computer Engineering, Italian SME will provide the services  to install, run and maintain the system.

Leonardo Hosting Entity

The Leonardo Hosting Entity consortium, composed by the Italian Research Institutions CINECA, INFN and SISSA, and participated by Austria, Greece, Hungary, Slovenia and Slovakia, is fully co-funded by the Italian Ministry of University and Research and the European Commission through the EuroHPC Joint Undertaking, a joint collaboration between national governments and the European Union to develop a supercomputing ecosystem with world-class infrastructure in Europe.

Technology and configuration

Featuring Atos’ patented DLC (Direct Liquid Cooling) solution, which minimizes global energy consumption by using warm water up to 40°C, the platform boasts a high energy-efficiency. The new supercomputer will be installed at the end of 2021 in the new data center located at the Bologna Tecnopolo designed to guarantee a PUE lower than 1.1 by combining the DLC solution with high energy efficiency of the electromechanical components of the new site's plants.

Leonardo will be configured with two partitions, the booster equipped with nearly 14,000 NVIDIA Ampere architecture-based GPUs  and the general purpose data centric partition with over 3,000 next-generation CPUs from Intel, interconnected with NVIDIA® Mellanox® HDR 200Gb/s InfiniBand networking. DDN will provide a highly scalable and performant storage solution with 5PB of flash and 100PB of capacity storage using Exascaler technology.

The modern way of scientific computing that requires the combination of high-performance simulation, data analysis, artificial intelligence, machine learning, and visualization work flows will be supported by ParTec’s ParaStation Modulo Software stack, that will combine computing power, capable of executing around 250 petaflops Linpack (HPL) and 10 exaflops FP16 AI (Artificial Intelligence), to enable advanced AI and HPC converged application use cases and push Italy towards the exascale class of high performance computing. 

Some strategic domains for the Italian research community using this approach include complex workflows in areas such as:

• Drug Repurposing and Discovery: Using genomic analysis to identify promising proteins that can be targeted with a specific drug to fight COVID-19 and other diseases;
• Resilience to natural disasters: both for the mitigation of the impact from extreme events of seismic nature as earthquakes, tsunamis, landslides, and from events of epidemiological and pandemic nature and to combat the health problems of the population;
• Green Deal:  for sustainability of the impact of human production activities, decarbonization and energy production from renewable sources and others such as nuclear fusion and hydrogen;
• Weather Modeling: Predicting extreme weather conditions with greater accuracy and speed fundamental elements underlying global challenges such as those related to the Destination Earth project.

“Cineca is proud to play a critical part in evolving both the research and industrial community in accelerated HPC application development in Europe,” said Sanzio Bassini, Director of Cineca’s HPC Department, “The Leonardo supercomputer is the result of our long-term commitment and demonstrates the European know how-and capacity to prove what an exascale supercomputer can be.”

“We are extremely proud to be contributing towards research in Europe and to be developing a strong European capability in exascale supercomputing,” said Agnes Boudot, Senior Vice President, Head of HPC & Quantum at Atos “Our BullSequana XH2000 supercomputers powered by Atos’ HPC Software Suite and Codex AI Suite provide an open and modular architecture, paving the way for exascale supercomputers. Not only do our systems provide a vast power and performance to drive HPC simulation, they also support Artificial Intelligence (AI) and Machine Learning model development, and data analysis, so that researchers can explore new applications, thereby boosting innovation and industrial competitiveness in Europe.”

“The Modular Architecture which is a development resulting from the European funded DEEP program, created mostly by a close cooperation with Forschungszentrum Jülich, Cineca and Atos, is the most advanced, flexible and capable system architecture in the world”, explains Bernhard Frohwitter, CEO of ParTec.“One can say that the innovative combination of a modular architecture and the ParaStation Modulo software allows the creation of supercomputers that greatly exploit the Petaflop / s peaks, better than traditional architectures. Our Software ParaStation dynamically reconfigures on the job hardware to suite the ever changing requirements of the algorithm to be calculated. Our modular architecture leads to most efficient machines, which are effectively faster and can handle a much broader range of different algorithms”.

###
Leonardo specifications
•    More than 120 BullSequana XH2000 Direct Liquid cooling racks
•    Booster partition with 14000 GPU nodes with 4 x NVIDIA Ampere architecture GPUs each, providing 240 PFLOPs Linpack Performance (Rmax) and 10 EFLOPS FP16 AI
•    A general purpose partition with 3000 CPU nodes using Intel Next-Generation processors, 
•    5PB of usable storage for High Performace storage from DDN
•    100PB of usable storage for Large Capacity Storage from DDN
•    Mellanox Infiniband HDR 200Gb Interconnect

About Cineca
Cineca established in 1969, is a not-for-profit consortium of 69 Italian Universities 21 national research institutions, the Italian Ministry of Education, and the Italian Ministry of University and Research. Cineca is the Italian national facility for supercomputing applications and research, one of the largest in Europe. It develops advanced Information Technology applications and services supporting the European scientific communities, the Italian academic administration offices, the Italian Ministry of Education, the Italian Ministry of Universities and Research, and the world of industry and Public Administration.
Cineca’s HPC infrastructure is equipped with cutting-edge technology managed by qualified personnel, which cooperates with researchers and customers for the most effective exploitation of the HPC systems, in both the academic and industrial fields. The mission of Cineca is to accelerate the scientific discovery by providing high performance computing resources, data management and storage systems and tools, HPC services and expertise at large. Cineca represents Italy in PRACE (the pan-European ESFRI e-infrastructure for HPC) and in the EuroHPC Joint Undertaking (a joint initiative between the EU, European countries and private partners to develop a World Class Supercomputing Ecosystem in Europe).