A Data-Centric Perspective on Scientific Workflows in the Computing Continuum
DescriptionConverging high-performance computing with systems and methods inspired in Big Data Science and artificial intelligence is crucial to advance knowledge discovery in a data-intensive landscape. Nowadays, complex scientific workflows must be able to extract knowledge and produce insight at every stage from the instruments to the scientist. This requires the integration of heterogeneous and geographically distributed computing environments into a computing continuum to seamlessly bridge simulations, machine learning and data-driven analytics. Current approaches to large-scale computing are accommodating the need to support hyper-heterogeneous environments in which different ecosystems and devices work together. From edge devices to supercomputers, each hardware technology has unique physical properties that determine their performance for different tasks, leading to different system architectures, software stacks and data management methods that maximize their individual performance. This opens challenges on how to enable the interoperability of the different computing paradigms, programming models, and data abstractions coexisting in the continuum. This talk presents use cases and new research directions on a data-centric perspective towards scientific wokflow composition in the computing continuum.
TimeMonday, June 2615:00 - 15:30 CEST
Computer Science, Machine Learning, and Applied Mathematics