Digital Twins : Practices & Principles for HPC

Abstract:

Digital twins, physically accurate virtual representations of real-world systems providing beneficial information in actionable time by combining sensor data with surrogate models, have a long, successful history in industry. The recent shift in HPC combining simulation, AI and edge computing is not only an opportunity to apply digital twins in science, but also magnify their impact on global public policy and institutional decision making, including climate change, renewable energy and global healthcare. Leveraging uncertainty quantification to inform physicians, assisted by real-time AI surrogate models can save lives now and in the future. Increasingly accurate simulations become single sources of truth for virtual datasets, capable of multi-physics synchrony with the real world in time and ranging from subatomic to interstellar spaces. Evolving from crude 3D approximations to near identical digital twins is critical to enable breakthroughs in computational biomedicine, nuclear fusion, factory automation and looking forward, the role of digital twins in quantum computing. The goal of this workshop is to bring together digital twin practitioners, computational scientists, middleware developers and compute resource providers to identify challenges and opportunities in establishing digital twins as a common HPC practice and identifying key principles for their use in high performance computing. 

Workshop Format & Schedule - (link to detailed workshop agenda)

Friday, November 17th, 2023 – all times US/MT

08:30 am - 08:40 am Intros from Panel Members & Workshop Team (30m, 5 min/speaker)

08:40 am - 09:00 am Panel discussion ( 3 – 5 questions to consider/discuss)

09:00 am - 09:20 am Keynote :

Bill Tang, Princeton - Princeton Program in Plasma Physics - (pptx slides)

Digital Twins across HPC Domains (~20 min / spkr)

Digital Twins in Healthcare 

09:20 am - 09:40 am Peter Coveney – University College of London (pptx slides)

09:40 am - 10:00 am Mariano Vazquez - ELEM Biotech, Co-Founder and CTO (pdf slides)

10:00 am – 10:15 am Break

Digital Twins in Oceanography & Climate Science

10:20 am - 10:40 am Ioan Hadade European Centre for Medium-Range Weather Forecasts (ECMWF) (pptx slides)

10:40 am - 11:00 am Chris Hill, MIT EAPS - Learning Fast Climate Emulators (pptx slides)

Digital Twins in Construction / Smart Cities + Public Policy

11:00 am - 11:20 am Dimitrios Rovas, UCL (pptx slides)

11:20 am - 12:00 pm Panel Retrospective & Audience Discussion + next steps

12:00 pm  End of workshop

Keynote and Subject Matter Expert Speakers:

Keynote speaker will be Bill Tang of Princeton Program in Plasma Physics

For this ½ day workshop format, we plan to feature a panel discussion (with audience participation) for workshop topic speakers to kickoff the broader themes central to deploying digital twins in HPC among the considered domains.

Confirmed Presenters 

Peter Coveney (UCL, Professor of Physical Chemistry)

Mariano Vazquez (ELEM Biotech, Co-Founder and CTO)

Ioan Hadade European Centre for Medium-Range Weather Forecasts (ECMWF)

Chris Hill, MIT EAPS - Learning Fast Climate Emulators

Dimitrios Rovas, UCL - Smart Buildings, Simulation and optimization of energy systems, renewable energy, whole building simulation and control

Workshop Organizers

Andrea Townsend-Nicholson, Professor of Biochemistry & Molecular Biology at University College London

Bio: Andrea holds a chair in Biochemistry & Molecular Biology at University College London (UCL). She obtained a BSc degree in Molecular Genetics & Molecular Biology (University of Toronto, Canada; 1986) and a DSc degree in Cellular & Molecular Biology (Université Louis Pasteur, France; 1990), completed postdoctoral training at the Garvan Institute of Medical Research (Sydney Australia;1991-1995) and UCL (1995-1999), and held a British Heart Foundation Fellowship at UCL from 1999-2001. Her research focuses on understanding the molecular basis of cell surface receptor function in health and disease using a combination of experimental and computational methodologies. Andrea is particularly interested in facilitating the introduction of personalised medicine into clinical practice and in the development of computational methodologies that converge with experimental findings. She teaches medical and undergraduate bioscience students to use supercomputers as part of their taught university curriculum.

Peter Messmer - Director of Developer Technology - Digital Twins, HPC Viz

Bio: Peter is a senior manager in the HPC Developer Technology group at NVIDIA. He and his team work on tools and technologies to help clients use GPUs to accelerate their scientific discovery processes. Peter holds an MSc and PhD in physics from ETH Zurich, Switzerland, with a specialization in kinetic plasma physics and nonlinear optics.

Philippe Segers - PRACE Board of Directors, GENCI Head of European HPC projects

Bio: Board of Directors of PRACES, Head of European High-Performance Computing (HPC) projects for GENCI, the French HPC agency, member of PRACE Technical Board and Management Board. Co-leading the work-package in charge of the Organization of the Research Infrastructure, in charge of KPI and Stakeholder Management (including the European Open Science Cloud (EOSC)). Member of the Management Board of the Public Procurement of Innovation (PPI) PPI4HPC, former work-package leader of PRACE Pre-Commercial Procurement (PCP). He holds an Executive MBA and has an academic background in Numerical modelling of Theoretical Physics. He began his career in Canada, working in Data Assimilation for Meteorology, and at Ispra, Italy, for the Joint Research Centre of the European Commission, on Climate Sciences, moved to the scientific software industry, on pollution modelling, then to hardware industry and spent twenty years of Program Management of R&D demonstration and prototypes for various EC and NATO projects.

Barton Fiske - Senior Alliances Manager - HPC DevTools, HPC Viz

Bio: As senior alliances and product manager for math libraries and devtools  at NVIDIA, Barton has been fascinated by 3D graphics, visualization and computer gaming from a very early age and pursued his degree in Computer Science specifically to further these interests into a full blown profession. A graduate of the Rochester Institute of Technology, Barton has more than 30 years experience in a variety of prior roles, ranging from software engineer to systems engineer and demo architect to senior cloud architect and director of technical sales and evangelism. Barton is the co-author of two books on Java programming and has developed dozens of interactive 2D, 3D and VR demo experiences for a wide variety of industrial scenarios, from global product launches to end user applications. He is currently using Digital Twin technology in his own life to manage and maintain models of his seaside home in Newport, RI. 

Existing assets/artifacts

From SC22 BoF

(list of attendees)

Questions/Topics below carried forward from George Biro’s slides from SC22 DT BoF:

HPC Challenges / Opportunities

• Algorithms

  • Extremely heterogeneous, distributed computing (leadership computing to edge computing)

  • Hierarchical model management

  • Real time performance, streaming / incomplete / stochastic data

  • Federated, asynchronous, discrete events

  • Encrypted data

• Software

  • Programming APIs for such complex workflows

  • I/O, assimilation, uncertainty quantification, optimization, experimental design, physical model learning

  • Integration of simulation, AI/ML, sensor, data processing / data mining

• Systems

  • Resource allocation, runtimes, monitoring, control,

  • Interactive steering, intaroperatity with diverse systems, legacy systems, databases

  • Data movement: fault tolerant, heterogeneous / wireless network stacks

  • On demand computing

• Hardware

  • Embedded, FPGAs, ASICs, networks, sensors, …