The ExaScience Lab is organized around a linchpin application: space weather simulation and prediction. This application is chosen as the driver for the lab,

• to better understand space weather and to enable better predictions of its effects on Earth
• because space weather simulations need exascale computing to achieve the required accuracy and performance to be economically relevant.
• because the algorithms and parallelization strategies needed for such simulations are similar in nature to the ones that are expected to drive a host of applications on future exascale systems

An architectural simulation framework will be developed to allow us to study the behavior of this exascale workload on existing computing systems.

This framework will enable the study of the detailed behavior of a very reduced (in space and in time) but representative workload on a software model of the exascale hardware.

These experiments will lead to recommendations for the actual hardware design and to performance, power and reliability feedback to the other components of the project.

The application will be linked with visualization software to allow for the interpretation of the simulation results. This visualization software will operate in-situ: on data in memory.

It will create virtual telescopes which will allow space weather forecasters to make better forecasts of the impact of space weather on earth by creating virtual observation points in space.

In order to shield the application developer from the technical details of the hardware platform, the space weather simulation and prediction will be developed based on a toolkit.

This toolkit will also have to cope with the effects of failing hardware components.