Industrial application recommendations
Large petaflop computers are operational now in several companies in Europe.
Big data management is already a reality. Nevertheless, the European ecosystem is outpaced by other regions.
Indeed, users of such infrastructures are mainly large companies or academia, and most of computing resources and applications are developed outside Europe. Actions need to be undertaken so as to keep Europe competitiveness.
For the industrial applications, the EESI1 roadmap is followed by most companies with a slight delay. In aeronautics, there is the need to couple LES simulations and to develop both simulation codes as well as flexible coupler and optimised coupling techniques.
For the Weather/Climate and Solid earth Sciences, the climate community is now preparing all the workflows and the simulations codes toward the next IPCC campaign. This will need to address data management issues and improving scalability of the applications, developing dynamical cores, coupling multiple codes, assessing uncertainties of the models
In fundamental Sciences a lot of massive simulations in cosmology are posing crucial issues for handling very large amount of data to be treated on the fly during the computation and made available during years to worldwide communities.
In Life Sciences the European Commission announced in early 2013 the launch of 2 major Flagships : Human Brain Project and Graphene with a funding of more than 1B€ during 10 years for each project. These two initiatives will lead to the creation in Europe of strong structured communities highly connected with existing research infrastructures like PRACE or GEANT.
Disruptive technologies in the following fields :
- Simulation and Optimization with uncertainties
- Simulation and Optimization in Complex Networks
- Data Driven Simulations
- Bridging scales by ultra-large simulations
- Advanced Adaptive Resolution
- Learning-on-the-fly
- Multiscale: hybrid and beyond
Additional recommendations are the following:
- To reinforce the training program of PRACE across Europe and the support of HPC and numerical simulation trainings from education (under graduate programs) to permanent training. MOOC tools need to be developed in complement of physical trainings in order to train more people from new disciplines or coming from remote locations.
- To fund co-educational programs on the master level which provides a broad knowledge of hardware technology and system- and application software not only for computer scientists but also for students in scientific computing.
- Foster educational programs for training of postdoctoral and senior scientists, who have to follow the trends and developments of hardware architectures, software and optimization and tuning of codes.
- With the rise of Open Innovation between academia and industry, it become mandatory to industrialize the different codes developed by academia in order to make them usable by industry with higher TRL. For the European Commission as well as the others funding agencies there is a need to invest on structures/teams which are able to provide software engineering methodologies for developing new standard components