What have in common tsunami waves, earthquakes, nuclear fusion, meteorology, gamma rays explosions and combustion process? These themes which attract the interest of scientists from all over the world have a universal mathematical formulation (under the conservation laws, for example the conservation of mass and energy) and are application fields for supercomputers.
Machines with extraordinary computational capabilities compared to the normal computers; the supercomputers with hundred of thousands of processors are able to execute simulations useful to research because they verify theories by putting into relation data gathered in the experimental phase and they produce more and more complex and precise models.
Since many years they make possible big forward steps in the progress of knowledge in various areas and it is foreseen in about a decade supercomputers will be able to manage computations a thousand time faster than today.
The next challenge for scientists is to create the Exascale computers, machines able to run billion of billions (1018) of computational operations per second.
The research group ExaHype (An Exascale Hyperbolic PDE Engine) is working on this project, financed by the European Commission within the Horizon 2020 framework with 2,8 million euros. The project ExaHype aims at developing a new simulation software for the conservation laws, able to exploit the power of supercomputers of new generation to be ready in 2020.
At this project is working an interdisciplinary and international team consisting of seven institutions from Germany, Italy, Great Britain and Russia, which sees among the leaders the University of Trento - sole Italian research institution - with the Laboratory of Applied Mathematics (Department of Civil, environmental and mechanical engineering), headed by professor Michael Dumbser.
The making of software applications for this kind of machines is an extremely complex operation compared to what happens for conventional computational computers. But to what purpose could this computational power be needed for?
«The likely applications to which we are working on, in the ExaHype project - explains Michael Dumbser - regards two different scenarios: geophysics, with the risk calculation associated to earthquakes, and astrophysics with the simulation of the gravitational waves and the explosions of the gamma rays. Earthquakes cannot be foreseen. Nevertheless the exascale computer simulations may help to assess the risk of the aftershocks. The regional simulations seams to acknowledge a better understanding of what happens during large scale earthquakes and in the following aftershock phase».
«In the field of astrophysics the new mathematical instruments developed in ExaHype could simulate the fusion of the orbiting neutrons stars. It is believed this phenomena not only could be at the root of the creation of the gravitational waves, but that could be the cause of the gamma rays explosions. The simulation on the exa scale could allow to study in a new perspective these mysteries, investigated since a long time in the field of astrophysics».
Beyond these two specific applications, the researchers are trying to keep the algorithm as general and open as possible to future applications in very different fields. It could be useful, for example, in the meteorological field, to simulate the trend of climatic and meteorological events, or in engineering to better understand the complex processes of circulation and combustion, or in the forecasting of natural disasters as the tsunami or the tides.
«The objective - clarifies Dumbser - is to make this technology adaptable and easily usable for the most different needs of research, in the shortest time possible. This is why the results from the project will be accessible to everyone in open source».
The supercomputers: a challenge for research
One of the biggest hindrance to overcome in order to win this scientific challenge is about the energy consumption. At present, the most powerful built computers - the Chinese Tianhe-2, the American Titan and Sequoia and the Japanese KComputer - are able to carry out almost 34 million of billions operations per second (1015) and use between 8 to 18 megawatts of energy, for an average yearly cost of a milion dollars per megawatt. «Considering today’s technologies - explains the scientific coordinator of the project, professor Michael Bader of the Technische Universität di München - to built an exascale computer which requires 70 megawatts to function is seriously a considerable commitment, both on the financial level, as well as at the infrastructures level. This is why the simulation software which is part of the ExaHype project will be designed taking into accounts the needs of an energy saving hardware which will have to be able to withstand this huge performance’s efficiency improvement».
The rise of computational speed (a thousand times higher than today) will have to come with an improvement of the data transfer and data storage processes, which will have to be the more and more swift and efficient to enable the supercomputer to process the information correctly without wasting energy. Another important challenge is the internal hardware communication. In this direction, the project forecasts to develop a scalable algorithm which increases in a dynamic and self-adaptive way the simulations’ resolution only when needed. This way the scientists can achieve very accurate simulations, by reducing to the minimum the necessary computer operations and therefore saving energy and memory.
An international and interdisciplinary team
Driving force of the project ExaHype is without any doubt the international and interdisciplinary intense collaboration. In addition to the University of Trento (Department of Civil, environmental and mechanical engineering), the consortium includes from the German side the Technische Universität di München, the Frankfurt Institute for Advanced Studies, the Ludwig-Maximilians-Universität München and the Bavarian Research Alliance. Present as well Great Britain with the Durham University. The consortium is integrated, from an industrial point of view, with the participation of the supplier of the Russian supercomputer ZAO RSC Technologies.