The Department of Energy’s Oak Ridge National Laboratory hosted the 17th annual Smoky Mountains Computational Sciences and Engineering Conference, or SMC, from August 26 to 28. But instead of meeting in Tennessee, attendees joined the all-virtual event from home offices, kitchen tables and other improvised workspaces.
A premier computing conference, SMC brings together scientists and engineers from government, industry and academia to discuss the current state and future directions of the computational and data science communities. With 373 people registered, SMC 2020 had the highest number of participants in the conference’s long history.
SMC 2020 included many “firsts,” including a new topic — the convergence of high-performance computing, or HPC, and artificial intelligence, or AI — to delve further into the opportunities and challenges identified during DOE’s recent Town Hall meetings. This year’s theme was also new: “Driving future science & engineering discovered through the integration of experiment, big data, and modeling and simulation.”
Conference organizers also issued SMC’s inaugural call for papers, which asked authors to identify and advocate for developments and investments in research areas in line with the conference theme. From 94 submissions, a 70-member program committee selected 37 peer-reviewed position papers from 40 international institutions, which will be published in the event’s first-ever conference proceedings.
The authors and other speakers delivered talks divided into four sessions to expand on these ideas and describe the importance of interrelated topics, including software stacks, scientific instruments, benchmarking tactics and the data life cycle. Poster presenters also contributed to these discussions in the form of brief lightning talks followed by longer Q&A sessions.
During a keynote session titled “The Future of HPC Systems in the Presence of AI,” Keynote Panel Coordinator Katie Schuman and Panel Moderator Travis Johnston led discussions with panelists Bronson Messer of ORNL, Prasanna Balaprakash of Argonne National Laboratory, Brian Van Essen of Lawrence Livermore National Laboratory and Michela Taufer of the University of Tennessee.
Primarily, the keynote speakers explained how current computing facilities and scientific problems got to where they are today, as well as where they might go in the future. In this vein, they considered what an ideal HPC system might look like, outlined opportunities to improve the performance of deep learning computation on existing resources and identified trends and gaps in computing workflows. They also listed requirements for taking the next big leap in AI-HPC convergence, including more memory and faster file systems.
“The panel highlighted key questions concerning the future of HPC, especially as workloads continue to change in the presence of heterogeneous hardware systems,” said Schuman, a computer scientist at ORNL. “Addressing those questions will require significant investments in the coming years.”
SMC’s fourth annual Data Challenge also broke previous records. Fifty-two teams participated in this data analytics competition, which provides a platform for students and experts to analyze real scientific datasets curated and provided by ORNL data sponsors. Teams from various organizations and universities competed in six categories related to active research efforts at ORNL in materials science, geophysics and other fields, as well as a seventh category linked to the COVID-19 Open Research Dataset Challenge.
Competitors analyzed these datasets and submitted papers answering a series of increasingly difficult challenge questions. Of the 23 teams that completed the challenge, the 16 selected finalists explained strategies and solutions in another virtual poster session. Finally, data sponsors and audience members voted on the winners and runners-up for “Best Advanced Paper,” “Best Student/Novice Paper,” “Best Lightning Talk” and “Most Promising Approach.”
During his closing remarks, SMC General Chair and Associate Laboratory Director for ORNL’s Computing and Computational Sciences Directorate Jeff Nichols thanked conference organizers, session chairs, speakers, presenters and other contributors for making the first virtual SMC a success. Additionally, he asked attendees to mark their calendars for next year’s conference, which is scheduled for August 24–25, 2021.
“Whether it takes place in person or online, SMC continues to be a vital place for sharing information in a rapidly changing technological landscape,” he said. “These discussions have helped experts in the fields of energy science and computing tackle some of the world’s most complex challenges for nearly two decades, and this year was no exception.”
UT-Battelle manages Oak Ridge National Laboratory for DOE’s Office of Science, the single largest supporter of basic research in the physical sciences in the United States. DOE’s Office of Science is working to address some of the most pressing challenges of our time. For more information, visit energy.gov/science.