Exascale is heralded as a major breakthrough in computing that will pave the way for excellence in science and engineering, machine learning, and artificial intelligence. Its realization comes with a cohort of challenges, especially in processing and managing extremely large data sets. The legacy flat storage systems on classic HPC architectures are no longer capable of dealing with the emerging data-intensive computing requirements of applications. To address these concerns, hierarchical storage systems have been developed to balance computation and storage performance. However, there is still a need for adequate software interfaces which can leverage these HPC I/O architectures. It is in this area that the ADMIRE project will play a key role by creating an active I/O stack that exploits all layers of HPC I/O. This will be achieved by implementing an intelligent global coordination mechanism, malleability of computation and I/O, and the integrated scheduling of storage resources. An operational software-defined framework will be developed based on scalable monitoring and control, with aligned control and data paths as well as embedded control points which orchestrate the system components and applications. The envisioned framework will be co-designed by the project partners with several use cases from diverse scientific areas, including climate/weather, life sciences, remote sensing, software heritage, and deep learning.

JSC will contribute to the work packages “Application Co-Design” (WP7), “Intelligent Controller” (WP6), and “Dissemination and Exploitation” (WP8). JSC is the lead partner for WP7, which will produce an enhanced version of the existing applications based on the co-design methodology developed during the project lifetime.