European Commission wants prototype exascale supercomputer systems based on European hardware technology by 2018

This is a bit less than the estimates of what is needed in public and private investments. For Europe to achieve HPC leadership estimates are: by 2020 3,3 billion euro in 5 years (2016 to 2020) or by 2022 5,3 billion euro in 7 years (2016 to 2022). These estimates are in order to match the developments of Europe's main competitors for HPC leadership in competitive time frames. The effort would be similar to the U.S.A. plans in terms of number of systems: two pre-exascale systems around 2019-2020 and two exascale systems around 2022.

According to the Commission, there is a surging demand in Europe for a world-class High Performance Computing (HPC) infrastructure to process data in science and engineering. The simulation of a complete next-generation airplane; climate modelling; linking genome to health; understanding the human brain; in silico testing of cosmetics to reduce animal testing - all need exascale computing capabilities. The European Union competitiveness also depends on the support of HPC for pan-European data infrastructures.

The Commission sees that worldwide, the USA, China, Japan, Russia and India are advancing swiftly. They have declared HPC a strategic priority, they fund programmes to develop national HPC ecosystems (hardware, software, applications, skills, services and interconnections) and work on the deployment of exascale supercomputers. Europe is not participating in the HPC race in line with its economic and knowledge potential. The Commission even thinks it is falling behind other regions as it fails to invest in its HPC ecosystem and to reap the benefits of intellectual property in this field.

On the supply side, the industry in the European Union provides about 5% of HPC resources worldwide, while it consumes one third of it. As Europe depends ever more on other regions for critical technology, the risk exists that it gets technologically locked, delayed or deprived of strategic know-how. Europe also lags in terms of sheer total computing power: only one out of ten supercomputers in the TOP500 is in the European Union, Germany's Hoechstleistungsrechenzentrum Stuttgart ranking 8th. Of course there is also a second European machine in the top 10, the Piz Dant at CSCS in Switzerland at position 7.

According to the Commission a single Member State alone has not the financial resources to develop the necessary HPC ecosystem, in competitive time frames with respect to the US, Japan or China. However to date no common action is taken to bridge the gap between internal demand and EU supply. The European Union set up a contractual Public-Private Partnership on HPC to develop exascale technology, but there is no European framework to integrate it in large scale computing systems.

The Commission does not answer the question why a single small European country, Switzerland, has a system in the top 10, and a dozen or so similar or larger countries in the European Union, are not able to buy a similar supercomputer on their own.

According to the Commission Europe needs integrated world-class HPC capability, high-speed connectivity and leading-edge data and software services for its scientists and for other lead users from industry (including SMEs) and the public sector. The European Data Infrastructure will also support the EU to rank among the world's top supercomputing powers by realising exascale supercomputers around 2022, based on EU technology, which would rank in the first 3 places of the world. Europe should aim to have at least two sources of this technology, says the Commission.

The supercomputing plans themselves are not new: the European Union already supports a supercomputing infrastructure - PRACE - for several years and also supports projects to develop new exascale hardware and software. However the existing plans did not foresee realising an exascale supercomputer. The European Data Infrastructure will gather the necessary resources and capabilities, to close the chain from research and development to the delivery and operation of the exascale HPC systems co-designed between users and suppliers.

Building on the Pan-European High Performance Computing infrastructure and services (PRACE), the trans-European high speed network (GEANT), the contractual Public-Private Partnership on HPC, the ECSEL Joint Undertaking, and the IPCEI on HPC and Big Data, the Commission and participating Member States will:

• foster an HPC ecosystem capable of developing new European technology such as low power HPC chips;
• integrate technologies into system prototypes, co-designing 50 solutions and procuring HPC systems; the resulting HPC infrastructure will focus on supercomputers of top-range capabilities connected to mid-range EU national computing centres and to pan-European data and software infrastructure to offer supercomputing as a service;
• provide seamless, high-speed, reliable and secure connectivity to make HPC accessible across the EU; the trans-European high speed network (GEANT) and National Research and Education Networks (NREN) already connect 50 million researchers and students; these infrastructures will be upgraded to match the increase of data volumes to be transferred and the extension of the user base.

The Commission expects that the European Data Infrastructure will contribute to the digitisation of industry, to develop European platforms for new, strategic applications (e.g. medical research, aerospace, energy) and to foster industrial innovation. It will widen the user base of HPC, providing easier access via the Cloud both to researchers in key scientific disciplines and to the long tail of science.

The Commission expects that industry, particularly SMEs without in-house capabilities and public authorities with smart cities and transport will benefit from Cloud-based and easy-to-use HPC resources, applications and analytics tools. In this context, the Commission will foster the deployment of processing and exploitation capacities for Sentinel satellites' data, Copernicus services information and other Earth Observation data, so as to enable the cross-fertilisation of different data sets, encourage the development of innovative products and services and maximise the socio-economic benefits of Earth Observation data in Europe.

In the 2016-2020 time frame, the Commission and participating Member States should develop and deploy a large-scale European HPC, data and network infrastructure, including the acquisition of two co-designed, prototype exascale supercomputers and two operational systems which will rank in the top three of the world as of 2018.

If Europe would succeed in building exascale supercomputers right down from the processor up to the machine integration and get it in the TOP500 of supercomputers, it would be for the first time since about 20 years when systems like the 1024 processor German Parsytec GC system based on transputer chips from the UK entered the TOP500. However these systems never made it in the TOP100. So getting in the TOP3 with European hardware is really a challenge.