Description
The grand challenge of Exascale computing, a critical pillar for global scientific progress, requires co-designed architectures, system software and applications. Massive worldwide collaboration of leading centres, already underway, is crucial to achieve pragmatic, effective solutions. Existing funding programs do not support this complex multidisciplinary effort. Severo Ochoa will cover this critical gap, strengthening cohesive internal synergies and external world-class collaborations to pursue new high risk strategic research and consolidating BSC's international leadership and influence of Exascale designs. The program will also attract talented researchers and provide unique training and mobility opportunities for career development.Next generation Exascale supercomputers are crucial for raising numerical simulation to the capacity levels required by different scientific fields to tackle major societal problems in human health and the environment. BSC-CNS, the Spanish national supercomputing centre, proposes a bold multidisciplinary research program to address the complex challenges in the path towards Exascale, co-developing energy efficient and high performance hardware and software platforms with applications in personalised medicine, modelling of human organs, and global models for climate change and air quality prediction. The program will leverage existing national and international collaborations, consolidating BSC-CNS as a world leader in the field.Key issues at the forefront of scientific endeavour require both immense computing power and sophisticated data handling capabilities, way beyond current state of the art. BSC envisions a fundamental break with past advances based on faster hardware, instead achieving Exascale by dramatically increasing the number and heterogeneity of computational devices. The proposed research program builds on BSC's in-depth expertise developed during 25 years driving the design and programming of world-class supercomputers.The intellectual merits of the BSC proposal lie in:1. Addressing the extremely complex issues of connectivity, reliability and power efficiency presented by Exascale systems with over 100 million processors. BSC proposes to make extensive use of low-power components to achieve a 25x improvement in energy efficiency.2. Managing to express and exploit parallelism for these systems. BSC will introduce asynchrony and dataflow concepts to new programming models to overcome limitations in delivering billions of threads synchronising globally, and more intelligent performance tools to understand and optimise applications.3. Handling unprecedented amounts of heterogeneous data in highly reliable, scalable storage architectures. BSC will propose new systems, tightly coupling asynchronous computation with data management and optimising for minimal energy usage, targeting both Exascale for research and Data Centres for Internet/Cloud computing services.4. Co-developing these Exascale technologies with specific applications in Life, Earth and Engineering sciences addressing urgent challenges in these fields.The impact of success will be momentous for global science and business computing, making Exascale systems and enterprise scale Data Centres economically feasible to construct and operate, and provide user-friendly programming platforms, optimisation tools and data handling systems. It presents a unique opportunity to leverage European hardware and BSC software, strategically positioning Europe in the future supercomputing landscape with key applications in health and climate modeling.The relevance and international context lies in developing new strategies to address power efficiency and performance, the key challenges facing computer system design and use at all levels. BSC will be positioned as a global leader in this critical technology that will impact all aspects of human society, from telephony to frontier research.
