Space Weather: Improving the parallel performance of codes
Space weather is the study of the processes originating in the sun and propagating in the space of the solar system and affecting people and technology in space and on the Earth. These events can can range from the harmless occurrence of the auroras in the polar regions to radiation threats to astronauts in space and to electromagnetic disturbances causing disruptive currents in ground infrastructures such as power and communication lines. KTH has implemented the massively parallel Particle-in-Cell code, iPIC3D. The code simulates the interaction between the solar wind and Earth magnetic field. Plasma particles from the solar wind are mimicked by computational particles. At each computational cycle, the velocity and location of each particle are updated, the current and charge density are interpolated to the mesh grid and Maxwell equations are solved. Earth magnetosphere is a large system with many complex physical processes, which requires realistically domain size and billions of computational particles.
The Space Weather application iPIC3D produces 10-100 TB of data for each simulation. This data includes the information of the trajectories of billions of particles. The main interest of iPIC3D users is to identify particles subject to strong acceleration in space and to understand the physical “engine” for accelerating particles to very high energy. The generation of such large amount of data in iPIC3D affects the performance of the code, as large part of the execution time is spent in I/O operations and for the analysis to detect particles subject to strong acceleration. SAGE Compute Storage and Offload modes in the iPIC3D application will enable an improvement of parallel scalability. We expect to increase the parallel efficiency by 20% in a typical production run with 10,000 MPI processes. Using post-processing analysis of particle trajectories in the full data centric SAGE mode, we expect to halve the time that is required now to detect particles subject to strong accelerations.
The example picture below shows the simulation of solar wind interaction with a planetary magnetosphere ( The tubes represent the magnetic field lines while the contour-plot represent the electron density).