About Nexus

For over a decade, Argonne National Laboratory researchers have been developing tools and methods to blend its world-class computing resources with data-intensive experiments. This connected approach, called Nexus, encompasses:

• Powerful facilities including the Argonne Leadership Computing Facility (ALCF) and Argonne’s Advanced Photon Source (APS), and an exploration of near-real-time analysis of experimental data from remote DOE user facilities. The ALCF and APS are DOE Office of Science user facilities open to the world’s scientific community.
• Artificial intelligence/machine learning and edge computing to conduct real-time analysis of experimental data on leadership-class computing resources, with AI eventually suggesting future experiment directions.
• Autonomous Discovery and self-driving labs that streamline processes, save resources, and accelerate the pace of discovery.

Nexus builds on the Globus research automation fabric for data management and its cloud-hosted federating services. Using Globus, experimental facilities can integrate large-scale computing into experiments, providing near-real-time analysis capabilities at scales attainable only at remote computing facilities. At the APS, when data is collected and catalogued in the APS Data Management system a process is automatically triggered that transfers the data to ALCF computers, submits related analysis jobs, and makes the results available to APS users at the beamline. This workflow pattern demonstrates the essential functions required by experimental facilities: ensuring security, transferring data, submitting and monitoring jobs, and visualizing results. It has already proven to be adaptable to other facilities.

Argonne researchers are creating new techniques for automating inter-facility workflows. Workflows that support experimental research activities largely feature the same general patterns and have the same general steps, such as data collection, reduction, inversion, storage and publication, machine-learning model training, experiment steering, and coupled simulation. By identifying how to express the steps in general terms, researchers aim to create workflows that are reusable and adaptable to different experiments.