Accelerating the adoption of cloud HPC for materials R&D

Details

Presenters

Dr. Timur Bazhirov, Founder and CEO of Mat3ra.com

Email: timur@mat3ra.com 

Bio: Timur obtained a Ph.D. in physics from UC Berkeley. There, he worked with Marvin Cohen and Steven Louie on first-principles studies of materials, publishing over a dozen peer-reviewed articles and one day realizing first-hand that we do research in materials and chemistry needs to be modernized. In 2015, he founded Exabyte Inc./Mat3ra.com, which currently has over 15,000 users from academia and industry, with dozens of enterprise customers in the electronics, energy, and manufacturing sectors, including Merck KGaA Darmstadt Germany, Air Force Research Laboratory (USA), and Tokyo University of Science.

Dr. Pranab Das, Senior Research Fellow, National University of Singapore • Scientific Software Engineer (Consultant), Mat3ra.com

Email: das@nus.edu.sg • pranab.das@mat3ra.com 

Bio: Pranab received PhD in condensed matter physics from Tata Institute of Fundamental Research, Mumbai. Then, he worked at the Eletra Synchrotron Trieste as a postdoctoral fellow from 2014-17. Since then, he has joined as a research fellow at Singapore Synchrotron Light Source, National University of Singapore.

Target audience

Scientists and Engineers working in condensed matter physics, materials science, chemistry, and related fields specializing in Density Functional Theory (DFT), Molecular Dynamics (MD), and related approaches. Our introductory session will also interest HPC architects and planners.

Pre-requisites: some knowledge of basic physics concepts in materials science, such as crystal structures and physical properties of materials.

Recommended prerequisites: familiarity with Linux/UNIX command line,  basic knowledge of DFT, familiarity with simulation codes such as Quantum ESPRESSO, VASP, or similar.

The workshop is intended to be interactive with involved hands-on sessions, and a target audience size of 15-25.

Workshop structure

This workshop aims to provide participants with an in-depth understanding of the high-performance computing (HPC) capabilities offered by Mat3ra.com, specifically tailored for materials science research. Through interactive sessions, hands-on demonstrations, and Q&A discussions, attendees will gain insights into optimizing HPC resources for computational modeling, data management, and automation workflows within the Mat3ra platform.

Agenda

As below:

1. Welcome and Workshop Overview 9:00 AM – 9:15 AM
   • Presenter: Workshop Facilitator/Mat3ra Representative
   • ‍Description: Introduction to the workshop goals, agenda, and overview of Mat3ra’s HPC capabilities
     ‍
2. Introduction to Mat3ra HPC Platform 9:15 AM – 9:45 AM
   • Topics Covered:
     • Overview of Mat3ra.com and its cloud-based HPC infrastructure.
     • Explanation of key computational resources and parallelization capabilities.
     • Overview of available simulation tools for materials science applications.
   • Objectives: Provide foundational knowledge of Mat3ra’s HPC environment and its relevance to computational research in materials science
   ‍
3. ‍ Setting Up and Optimizing HPC Workflows in Mat3ra 9:45 AM – 10:30 AM
   • Format: Live Demonstration and Hands-On Session
   • Topics Covered:
     • Step-by-step guide to setting up computational workflows on Mat3ra.
     • Resource allocation strategies: CPU vs. GPU usage, parallel processing, and optimization tips.
     • Integrating multi-step workflows for materials simulations, including electronic structure calculations and molecular dynamics.
     • benchmarking per https://github.com/Exabyte-io/exabyte-benchmarks-suite and https://ui.adsabs.harvard.edu/abs/2018arXiv181205257M/abstract
   • Objectives: Equip participants with the skills to set up and optimize their own HPC workflows on the platform.
     ‍
4. Break 10:30 AM – 10:45 AM
5. Advanced Capabilities: High-Throughput and AI-Driven Workflows 10:45 AM – 11:15 AM
   • Topics Covered:
     • Automating high-throughput computations and data handling.
     • Leveraging HPC for AI/ML-enhanced material discovery workflows.
     • Examples of high-throughput applications include screening large material datasets for specific properties.
   • Objectives: Demonstrate advanced automation and machine learning integration within HPC workflows to accelerate materials discovery.
     ‍
6. Case Studies: HPC-Enabled Research on Mat3ra 11:15 AM – 12:00 PM
   • Presenter: Guest Researcher
   • Format: Presentation with Q&A
   • Topics Covered:
     • Real-world examples of materials research projects executed on Mat3ra’s HPC platform: compute-demanding cases for advanced approaches (DFT HSE, GW), including large cell sizes, magnetism, relativistic effects, etc.
     • Discuss applications in electronics, energy storage, and catalysis.
     • Lessons learned and best practices for utilizing Mat3ra HPC capabilities.
   • Objectives: Highlight successful applications of Mat3ra in materials research, emphasizing best practices and innovative uses of HPC resources.
     ‍
7. Interactive Q&A and Troubleshooting Session 12:00 PM – 12:30 PM
   • Format: Open Q&A with Mat3ra’s technical team
   • Topics Covered:some text
     • Addressing specific questions from participants regarding HPC setups, optimization, and technical issues.
     • Troubleshooting common challenges when using HPC resources in computational materials science.
   • Objectives: Provide personalized support and guidance for participants’ ongoing or future projects on Mat3ra’s platform‍
   
   
8. Closing Remarks and Next Steps 12:30 PM – 12:45 PM
   • Description: Recap of key takeaways, additional resources for HPC use on Mat3ra, and information on follow-up support
   • Objectives: Encourage continued use and exploration of Mat3ra’s HPC capabilities with guidance on accessing further assistance and resources.

Total duration: 4-6 hours.

References

1. Timur Bazhirov, Mohammad Mohammadi, Sergey Barabash, Kevin Zhang, “Large-scale high-throughput computer-aided discovery of advanced materials using cloud computing”, Proceedings of the American Physical Society meeting 2017, http://meetings.aps.org/link/BAPS.2017.MAR.C1.7 
2. Timur Bazhirov, “Data-centric online ecosystem for digital materials science”, manuscript, 2019, arxiv.org preprint: https://arxiv.org/abs/1902.10838 
3. Protik Das, Mohammad Mohammadi, and Timur Bazhirov, “Accessible computational materials design with high fidelity and high throughput”, manuscript, 2018, arxiv.org preprint: https://arxiv.org/abs/1807.05623 
4. Protik Das and Timur Bazhirov, “Electronic properties of binary compounds with high fidelity and high throughput”, Journal of Physics: Conference Series, volume 1290, pages 012011, 2019 
5. Timur Bazhirov and E. X. Abot, “Fast and accessible first-principles calculations of vibrational properties of materials”, manuscript, 2018, arxiv.org preprint: https://arxiv.org/abs/1808.10011 
6. Mohammad Mohammadi, Timur Bazhirov, “Comparative benchmarking of cloud computing vendors with high-performance Linpack”, Proceedings of the 2nd International Conference on High-Performance Compilation, Computing and Communications, 2018, https://dl.acm.org/citation.cfm?doid=3195612.3195613