The PRACE Scientific Steering Committee Erik Lindahl!1
European Computing Solves Societal Challenges PRACE s goal is to help solve these challenges. The days when scientists did not have to care about the hardware are over, and so are the days when compute centers did not have to worry about the scientific application!!2
Computing drives Science - and Science Computing Remarkable Repeated Success Stories: Recurring core part of Nobel Prizes in Physics & Chemistry Saving billions with better weather forecasting Improving human health with genomics, personalized medicine 3-4% better fuel efficiency of aircraft & wind turbines every year Disrupting communication, transportation and manufactoring Design of future materials from scratch based on desired properties Batteries & supercapacitors Artificial intelligence, machine learning, sensors, open data Scientific computing keeps delivering impact, but European impact is limited by resources. The Scientific Case for Computing in Europe showcases the achievements we predict will come true within the next 5-10 years with a factor 50-100x more computing power in Europe!3
Expanding the Frontiers of Fundamental Sciences Simulation of traces producing Higgs Boson (Nobel Prize 2013) LIGO gravitational wave interference (Nobel Prize 2017)!4
PRACE supports outstanding European research Space weather models supported by ERC & PRACE predict how solar eruptions will influence electric grids & satellites (Palmroth, Helsinki) Simulations of neutron starts merging supported by ERC & PRACE unravel the character of matter (Rezzola, Frankfurt)!5
Climate, Weather & Earth Ecosystems Irma & José!6
Europe leads international code development The best forecast codes are European Arguably, the US would have done better by investing less in machines and more in software Irma & José Predicting evacuation needs is a life/death matter but avoiding it saves 250M Maintaining our European lead on software should be one of our investment priorities!7
Improving Human Health Cryo-electron microscopy (Nobel 2017): Better computers & algorithms make it possible to understand how molecules move, and design new types of drugs!8
Human Health Computing is Translational It will become possible to design microfluidics & nanobiotechnology devices to target e.g. specific cancer cells by modeling flow/interactions Lattice-Boltzmann enable simultations of blood flow, e.g. effects of using a stent to treat brain aneurysm. On-demand computing! Tumor classification with deep learning neural networks now beats the best pathologists!9
Energy applications need Exascale Oil & gas is one of the world s largest users of HPC because computing saves time & money Fluid dynamics is used to understand heat & flow in nuclear reactors!10
Computing is Driving Renewable Energy Optimisation of wing shapes placement of wind farms HPC is creating new generations of insulators that enable higher voltage cables, which reduces losses New generations of solar cells (materials) Advanced MHD simulations are critical to make the plasma stable enough to make Tokamaks like ITER useful for energy production through fusion!11
Infrastructure & Manufacturing for Humankind Optimized wing Aage et al., Nature 550, 84 (2017) PRACE industrial project Reach design goals such as: Stronger, Lighter, Faster Use less raw material Cheaper, Safer Reduce time-to-market Direct numerical simulation is a revolution for fluid dynamics!12
Key Infrastructure Requirements Urgent need for more compute cycles - but also high demands for memory bandwidth & I/O Scientific Applications are not only ready to use Exascale, but have specific high-impact research goals that can only be realized with access to next-generation resources However: Exascale will require new approaches to scaling by using ensembles, deep learning, and statistical models. High-end computing will dominate, but tomorrow s solutions will be different than yesterday s Exascale systems will need to be able to handle tens of thousands of active jobs and large I/O requests Software & algorithms take longer to change than hardware - first-generation systems must be based on present concrete needs, rather than hopes about being able to co-design Linux rules the modern computing world: All systems will need to support a full Linux stack of development tools Centers & their staff need to engage directly in software development together with scientific communities to build on Europe s lead in these areas!13