This project focuses on providing the necessary R&D to support the I/O performance and portability requirements of the ATDM applications and system software. Without this R&D, ATDM applications will not be able to effectively manage and store data in the more complex ATS platforms. The work includes support and maintenance of production-level I/O libraries such as IOSS, netCDF, HDF5, as well as the research-grade Flexible Asynchronous Object Data-Exchange Libraries (FAODEL)—I/O middleware that provides efficient data movement and management of transient data within an HPC system. This data-management layer leverages available system memory, non-volatile memory (e.g., burst buffers), and parallel file systems and is available as both a user-level library and a persistent service to HPC applications, application workflows, and various runtime systems (e.g., AMT systems). The R&D performed will be evaluated and deployed on existing ASC platforms (primarily ATS systems), Sandia’s advanced architecture testbed, and ASC-developed simulation capabilities (where necessary) to demonstrate proof-of-concept software and develop production-capable packages for use by the ATDM program with eventual deployment to the IC program as well as the broader needs of the Exascale Computing Project.
The Scalable Visualization and Analysis work provides enabling capabilities for the visualization and analysis needs of the ASC/ATDM applications on next-generation, many core platforms. The primary emphasis is on development, integration, and deployment of in situ analysis and visualization capabilities as well as the testing, tuning, and deployment of many-core algorithms for visualization to support the direct needs of ATDM applications on the ASC/ATS systems. Without these advances, we will be unable to capture and analyze enough data from large-scale ATDM applications to understand the science.