Memory efficient information processing for FAIR computing

The Facility for Antiproton and Ion Research (FAIR) is an international research infrastructure for multiple communities, facilitating research in areas like hadrons, heavy ions, astrophysics, nuclear physics and more. FAIR includes a new double-ring synchrotron that is able to provide ion beams of unprecedented intensity and increased energy, thus enabling scientists to perform previously impossible experiments.

These experiments yield an enormous amount of data (Petabytes) that has always been filtered by real-time trigger steps, simple filters that identify an event and eliminate the rest of the data as noise. Contrary to that, FAIR experiments will gather all information in a huge data stream and process it on-line to retrieve more information and gain more flexibility. The amount of data and the processing architecture dictate the need for data structures and algorithms that are extremely memory efficient as well as massively parallel.

Our main focus will be on the development on suitable efficient data models and structures for FAIR-specific problems as well as algorithms to process the raw data. Considering the amount of data and calculations the algorithms have to exploit the architecture of the computing backend as much as possible, making not only use of memory hierarchies but massively parallel coprocessors like GPUs.