Computing and Data Requirements in the Exascale Age
The DOE Office of Science brought together key computational domain scientists, and DOE planners and administrators to determine the requirements for an exascale ecosystem that includes computation, data analysis, software, workflows, HPC services, and the full-scale range of computer requirements needed to support forefront scientific research through 2025.
This effort helps ASCR Facilities Division develop a strategic roadmap to inform future ASCR compute and data facility investments, ASCR Facility Division partnerships with Office of Science stakeholders (ASCR R&D, BES, BER, FES, HEP, and NP) and inform the Office of Science research agenda.
Meetings were held for each of the DOE’s six Office of Science (SC) program offices, as follows:
- High-Energy Physics (HEP) — June 2015.
- Basic Energy Sciences (BES) — November 2015.
- Fusion Energy Sciences (FES) — January 2016.
- Biological and Environmental Science (BER) — March 2016
- Nuclear Physics (NP) — June 2016
- Advanced Scientific Computing Research (ASCR) — September 2016
The overarching goal of the reviews was to determine the requirements for an exascale ecosystem that includes computation, data analysis, software, workflows, high-performance computing (HPC) services, and other programmatic or technological elements that may be needed to support forefront scientific research.
Participants in the Exascale Requirements Reviews addressed the following:
- Identify forefront scientific challenges and opportunities in that could benefit from exascale computing over the next decade.
- Establish the specifics of how and why new HPC capability will address issues at various scientific frontiers.
- Promote the exchange of ideas among application scientists, computer scientists, and applied mathematicians to maximize the potential for use of exascale computing to advance scientific discovery.
The requirements collection process was defined and implemented by a Steering Committee* consisting of these representatives from the ASCR HPC Facilities.
- Katie Antypas, NERSC**
- Richard Coffey, ALCF
- Eli Dart, ESnet
- Sudip Dosanjh, NERSC
- Richard Gerber, NERSC
- James Hack, OLCF
- Inder Monga, ESnet
- Michael E. Papka, ALCF
- Katherine Riley, ALCF
- Lauren Rotman, ESnet
- Tjerk Staatsma, OLCF
- Jack Wells, OLCF
* We would like to acknowledge the early contributions of Paul Messina and Tim Williams of ALCF.
** Deborah Bard of NERSC is serving on the Steering Committee while Katie Antypas is on leave.
An Organizing Committee was selected for each review. The Committee included a chair and co-chair who are leaders in the scientific community associated with the given SC program office. The chair and the Organizing Committee selected the themes and objectives for the breakout sessions to be held at the review and identified the breakout leads. Breakout session leads were selected on the basis of their technical expertise and vision for the computational needs of the scientific field, their ability to focus on the key technical issues, and their skills and experience in writing a reporting document that will serve the intended audience.
Attendees were selected in consultation with the Organizing Committee chair, co-chair, and DOE SC program managers. Each review engaged a limited set of participants with expertise in the domain science, experimental facilities, applied mathematics and computer science, and software development. Pre-registration was open to invited guests. The Organizing Committee was also free to invite observers from other divisions in ACSR, in the program offices, and from other agencies.
Outlines and input from white papers and case studies authored by the participants and submitted to the Organizing Committee chairs in advance of the reviews guided the discussions in general sessions and topical breakouts. Committee members and review participants collaborated at the meeting to identify the grand challenges, priority research directions, and computing requirements for their fields of research — communicating these requirements to the DOE SC offices and ASCR facilities. The reports therefore reflect extensive and varied forms of input from many voices in the each scientific community regarding HPC requirements for SC’s world-class initiatives.
The reviews afforded a rare opportunity for the participants to interact and learn about each other’s areas of expertise, challenges faced, and the exciting opportunities made possible by the exascale computing environment.
The final deliverable will be a report for each Office detailing the review findings. This report will help ASCR Facilities Division develop a strategic roadmap to inform future ASCR compute and data facility investments, ASCR Facility Division partnerships with Office of Science stakeholders (ASCR R&D, BES, BER, FES, HEP, and NP) and inform the Office of Science research agenda.
Carolyn Lauzon, [email protected], U.S. Department of Energy