Aims and Scope
High-performance computing (HPC) is at an inflection point in which the near-end of Moore’s Law, the big data explosion from AI workflows and next generation scientific instruments, the increasing operational costs beyond exascale, and the extreme heterogeneity of hardware vendors architecture and software systems have led to significant technical and economical barriers. Thus accessing cutting-edge HPC effectively has become a significant complex investment and endeavor with little reminiscence of the pre-accelerator and many-core era before the last decade.
The first “Democratizing HPC” (D-HPC) half-day workshop invites interdisciplinary communities of developers, facilities, vendors, users, researchers, educators, etc., to define, understand and quantify accessibility towards the democratization of current and future HPC technologies and ecosystems characterizing the path from idea to scientific discovery. Past examples of high-impact technologies that enabled HPC democratization are the message passing interface (MPI) and graphics processing units (GPU) providing accessible communication and processing power. We understand the technical democratization of HPC in its broadest meaning, pursuing success stories about enabling and/or improving accessibility in data- and compute-intensive applications across all domains ranging from traditional areas like simulations, data management and analysis, to more recent fields, such as AI, for a wide variety of extreme heterogeneity computing targets such as: manycore, quantum, neuromorphic, field-programmable gate arrays (FPGAs), chiplets, etc.
Call for Papers
D-HPC invites submissions of full technical papers for inclusion in the SC23 workshops proceedings. To meet our scientific goals, we require that the presented conclusions are based on rigorous analysis of empirical data, rather than anecdotal. This implies quantifying the accessibility in experiences and use-cases with a strong focus on novel or practical solutions, challenges and impact of their before/after scenarios.
Topics of interest, but are not limited to :
- The impact of programming languages and models, compiler techniques, libraries, domain-specific languages (DSLs), communication, data frameworks and ecosystems in improving access to HPC applications
- Making the convergence of HPC + AI more accessible
- HPC accessibility to big data experiments
- Accessible exascale and post-Moore heterogeneous hardware: quantum computing hardware, new architectures: many-core, accelerators, GPUs, file systems, neuromorphic systems, FPGA, chiplets, etc.
- AI techniques to improve HPC accessibility: generative AI tools, e.g. GPT, machine and deep learning models.
- Multidisciplinary end-to-end co-design practices, processes, paradigms that provide accessible HPC: e.g. fault tolerance designs, data reduction, porting to new architectures, proxy applications
- Virtualization techniques that provide performance portable cross-system access (e.g. containers)
- Proposed metrics and the theoretical framework for studies quantifying accessibility in future procurement systems based on existing empirical data
- Standards and ecosystem sustainability requirements based on empirical data for accessible future systems
Examples of metrics used for HPC accessibility are, but not limited to:
- Scalable performance improvements from providing HPC access to new problems and communities
- The number of core-hours, user or developer time, failure rates, time to overall solution, energy metrics as a result of leveraging accessibility
- The economics, benefits and impact of porting existing code to new architectures
- Technology adoption rates and communities impacted due to accessibility improvements
Out of Scope:
- Papers describing the technical merits of a particular technology or methodology without showing metrics that illustrate their impact on making HPC more accessible
- Position or review papers and papers highlighting the community or educational aspects only
- Papers describing merits on advancing a specific scientific application domain rather than HPC access
- Papers that do not describe targeted HPC systems or communities benefiting from improved accessibility
Paper Format: This year SC is implementing a workshop proceedings. Papers are encouraged to abide with the SC23 reproducibility initiative and provide an artifact description.
- Authors must use the new ACM proceedings templates
- Regular papers are between 6 to 12 pages, including the references and figures
- For Artifact Description and Evaluation appendices, include them in the regular paper page limits (i.e., up to a total of 12 pages)
- Papers submission and reviews are done through the SC23 linklings website
Best Paper Award: The best paper award will be selected by the technical committee based on impact, relevance and quality of the work.
Important Dates
- Submission opens: June 1 link
- Paper Submission deadline:
August 4August 16 (final extension) - Author Notification: September 8
- Camera-Ready Submission: September 21
- SC23 conference: November 12-17
- D-HPC Workshop Date: Sunday, November 12, 1:30-5pm MST
Agenda
| Time | Title | Presenter |
|---|---|---|
| First part | ||
| 2:00pm-2:10pm | Opening: “LLMs and Democratizing HPC” | William F Godoy and Pedro Valero-Lara, Oak Ridge National Laboratory |
| 2:10pm-2:45pm | Invited Talk: “The History and Future of Making HPC Technologies Accessible to the Wider Community” | Al Geist, Corporate Fellow, Oak Ridge National Laboratory |
| 2:45pm-3:05pm | Presentation: “Democratizing HPC by Building a Diverse and Inclusive Workforce” | Mary Ann Leung, Director, Sustainable Horizons Institute |
| 3:05pm-3:35pm | Afternoon Break | |
| Second part | ||
| 3:35pm-4:00pm | Paper: “Democratizing HPC Access and Use with Knowledge Graphs” | Pouya Kousha, Vivekananda Sathu, Matthew Lieber, Hari Subramoni, DK. Panda |
| 4:00pm-4:40pm | Invited Talk: “Democratizing Science Through Equitable Access to Computing and Data” | Manish Parashar, Director, Scientific Computing and Imaging Institute, University of Utah |
| 4:40pm-5:25pm | Panel: “S4PST: Stewardship of Programming Systems and Tools” | Moderator: Jeffrey Vetter, Corporate Fellow, Oak Ridge National Laboratory |
| Panelists: | ||
| Sunita Chandrasekaran, University of Delaware | ||
| Damian Rouson, Lawrence Berkeley National Laboratory | ||
| Johannes Doerfert, Lawrence Livermore National Laboratory | ||
| Johannes Blaschke, Lawrence Berkeley National Laboratory | ||
| Patrick Diehl, Louisiana State University | ||
| 5:25pm-5:30pm | Closing Remarks and Adjourn. | William F Godoy and Pedro Valero-Lara, Oak Ridge National Laboratory |
Organization
Please contact the organizers for questions and if would like to participate in future committees
Program Organizers
- William F Godoy, Oak Ridge National Laboratory, USA
- Pedro Valero-Lara, Oak Ridge National Laboratory, USA
Steering Committee
- Sunita Chandrasekaran, University of Delaware, USA
- Barbara Chapman, Hewlett Packard Enterprises, USA
- Jack Dongarra, University of Tennessee/Oak Ridge National Laboratory, USA
- Hartmut Kaiser, Louisiana State University, USA
- Mary Ann Leung, Sustainable Horizons Institute, USA
- Manish Parashar, University of Utah, USA
- Antonio Pena, Barcelona Supercomputing Center, Spain
- Christian Trefftz, Grand Valley State University, USA
- Jeffrey S Vetter, Oak Ridge National Laboratory, USA
- Michael Wong, Codeplay Software, Canada
Technical Committee
- Olivier Aumage, National Institute for Research in Digital Science and Technology (Inria), France
- Carlos Barrios Hernandez, Universidad Industrial de Santander, Colombia
- George Bosilca, University of Tennessee, USA
- Silvina Caino-Lores, University of Tennessee, USA
- Rocio Carratala-Saez, University of Valladolid, Spain
- Erin Carrier, Grand Valley State University, USA
- Jan Ciesko, Sandia National Laboratories, USA
- Patrick Diehl, Louisiana State University, USA
- Johannes Doerfert, Lawrence Livermore National Laboratory, USA
- Todd Gamblin, Lawrence Livermore National Laboratory, USA
- Simon Garcia de Gonzalo, Sandia National Laboratories, USA
- Patricia Grubel, Los Alamos National Laboratory, USA
- Gokcen Kestor, Pacific Northwest National Laboratory, USA
- Nicholson Koukpaizan, Oak Ridge National Laboratory, USA
- Het Mankad, Carnegie Mellon University, USA
- Jose Manuel Monsalve, Argonne National Laboratory, USA
- Suzanne Parete-Koon, Oak Ridge National Laboratory, USA
- Swaroop Pophale, Oak Ridge National Laboratory, USA
- Raul Sirvent, Barcelona Supercomputing Center, Spain
- Miwako Tsuji, RIKEN Center for Computational Science, Japan
- Veronica Vergara-Melese, Oak Ridge National Laboratory, USA