Scientific user facilities (SUFs) at the U.S. Department of Energy (DOE) drive scientific discovery and innovation by delivering world-class experimental capabilities that expand the frontiers of biology, chemistry, physics, and materials science. Over the next 5 years, upgrades at SUFs will generate over an order of magnitude more data, promising to accelerate the pace of scientific innovation if correctly harnessed. However, this flood of data poses challenges for the scientific community, despite continued growth in HPC hardware performance. The current state of the practice and tools optimized for HPC are insufficiently flexible and productive to address the high-stakes, short timelines, and rapidly evolving requirements of highly dynamic scientific user experiments. Additionally, traditional HPC software tools demand expertise that most users of SUFs cannot realistically apply within the pace and pressures of modern experiments, underscoring the need for more accessible, high-productivity approaches. Emerging AI/ML technologies, though promising, do not address these needs, and will not lead to a productive, high-performance software ecosystem without decisive action.
This workshop will explore the research challenges and opportunities in building a highly productive, high-performance software ecosystem for large scale scientific data analysis for users at the SUFs. The goal of the workshop is to identify key research directions that, if addressed, would substantially change the status quo and deliver an order of magnitude increase in productivity and performance for users of SUFs across the DOE complex.
To address these goals, short (8 minute) talks are solicited in the following areas.
Talks should be purposefully forward-looking, identifying key research directions for the next 5–10 years while moving beyond refining current practice toward transformative solutions that could increase productivity and performance by an order of magnitude.
Each talk will be accompanied by an abstract (up to 2 pages maximum). Multiple abstracts may be submitted to address different topics, but please include a different lead author for each submission. Accepted abstracts will be published on the workshop website prior to the beginning of the workshop.
All participants are strongly encouraged to attend in person to maximize the productivity of the event, though a virtual option will be available if necessary. Participants in the workshop will also be invited to help write a report that summarizes the findings of the workshop.
Presentation of a lightning talk is not a requirement for participation in the workshop, but note that advance registration is required.
Talks are 8 minutes with separate Q&A. All participants are strongly encouraged to attend in person to maximize the productivity of the event. A virtual presentation option will be available if necessary.
Abstracts are up to 2 pages maximum (including all text, figures, footnotes, citations, etc.), in 12-point font, on letter paper, with 1-inch margins. Abstracts should be submitted in PDF format only.
By submitting an abstract, the authors consent to publishing it publicly.
The workshop will be held at SLAC National Accelerator Laboratory in Menlo Park, CA. Parking is available for free on site. Note that registration is required for all attendees and that all visitors must present valid identification when entering SLAC.
The workshop will be held in the Redwood conference room (A/B/C/D), located in building 048. A map of SLAC is available at: https://vue.slac.stanford.edu/meeting-rooms
Attendees may book at the Stanford Guest House (subject to availability). Additional hotels may be available in the area.