SC21 Proceedings

The International Conference for High Performance Computing, Networking, Storage, and Analysis

Two Worlds Collide: Traditional HPC Simulation and the Modern AI/Machine Learning Stack


Authors: Sam Partee (Hewlett Packard Enterprise), Venkatram Vishwanath (Argonne National Laboratory (ANL)), Scott Bachman (National Center for Atmospheric Research (NCAR)), Austin Ellis (Oak Ridge National Laboratory (ORNL)), Riccardo Balin (Argonne National Laboratory (ANL))

Abstract: The Birds of a Feather (BoF) Two Worlds Collide: Traditional HPC Simulation and the Modern AI/Machine Learning Stack will highlight the current experiences, challenges, and future opportunities of several laboratories as they grapple with the need to leverage AI and machine learning to advance state-of-the-art research. The goal of this BoF is to present current thinking, execution and future needs between traditional HPC simulation and the modern AI/machine learning stack. This BoF will also wrestle with the topic; given current capability, what does the community need to prioritize to achieve the desired outcomes of combining HPC Simulation and AI/ML.

Long Description: This is the first of a desired on-going series Birds of a Feather (BoF) Two Worlds Collide: Traditional HPC Simulation and the Modern AI/Machine Learning Stack. This BoF seeks to highlight the current experiences, challenges, and future opportunities of several Laboratories as they grapple with the need to leverage AI and Machine Learning to advance state-of-the-art research. The goal of this BoF is to present current thinking, execution, and future needs between traditional HPC simulation and the modern AI/Machine Learning stack. This BoF will also wrestle with the question: given current capability, what does the community need to prioritize to achieve the desired outcomes of combining HPC Simulation and AI/ML?

The language gap between tradition HPC simulation, typically written in Fortran, C, or C++, with the modern AI/Machine Learning stack, usually written in Python, present significant programmatic, data, and parallelism challenges in being able to integrate efficiently into uniform solutions and capability. The BoF Two Worlds Collide: Traditional HPC Simulation and the Modern AI/Machine Learning Stack will discuss these practical challenges as well as present the experiences between supercomputing centers, industry, and the open-source community highlighting the need for collaboration between these three entities in order to achieve Machine Learning capability at scale with HPC Simulations.

The supercomputing centers represented in this BoF include, Argonne National Laboratory, the National Center for Atmospheric Research (NCAR), and Oak Ridge National Laboratory. Also represented from industry will be Hewlett Packard Enterprise (HPE) and the research, started at Cray around Simulation and AI, that has carried over to today. Participants will demonstrate what work is currently being conducted at their centers to set the context for further discussion.

The outcome of this BoF will be as follows. First, develop a current understanding of capability that researchers can leverage today when it comes to injecting Machine Learning capability into traditional HPC simulations. This will allow participants with little to no understanding of current capabilities to see what is possible given today’s technology. Second, we’ll enumerate the desired outcomes of bringing together HPC Simulation and AI/Machine Learning that participants would feel would most advance current state-of-the-art research if certain capability were available. Lastly, we’ll capture the delta between these first two exercises to use as a baseline to gauge progress over the next several years’ time. This BoF seeks to further develop a burgeoning community interested in the intersection of HPC Simulation and the AI/ML stack.


URL:


Back to Birds of a Feather Archive Listing