gromacs2017

Exascale data generation and analysis for MD simulation

G├Âttingen, May 22-24

Molecular dynamics (MD) simulation performance has steadily improved over the last decade as a result of new algorithms that optimally exploit state-of-the-art computer hardware. In contrast, data formats and analysis tools for MD programs such as GROMACS, are often decades old and yield output requiring manipulation before being used by other parts of the program suite or external tools.

Definitely, improved data interfaces are needed as a prerequisite for the incorporation of new algorithms that, e.g., require external data as input, produce complex data as output, or transform data within the simulation software. Examples are simulations driven by experimental data, Monte Carlo ensemble sampling schemes, and the ╬╗-dynamics fast multipole method (FMM) developed within the GROMEX SPPEXA project.

For all these algorithms, an extensible, flexible, and scalable way of data description and storage is needed as well as software infrastructure that can efficiently operate on that data. Therefore, in this workshop, we will address how to communicate and access the data generated by massively parallel biomolecular simulations, and harmonize software interfaces for inter and intra software data communication. Beyond pure performance gains, these data interfaces will add increased flexibility and usability for the scientific community.