J. Gregory Pauloski and Rohan Basu Roy Named Recipients of 2023 ACM/IEEE CS George Michael Memorial HPC Fellowships
Share this on:
Los Alamitos, Calif., October 26, 2023 – The Association for Computing Machinery (ACM) and the IEEE Computer Society (IEEE CS) announced today that J. Gregory Pauloski of the University of Chicago and Rohan Basu Roy of Northeastern University are recipients of the 2023 ACM/IEEE CS George Michael Memorial HPC Fellowships. Hua Huang of the Georgia Institute of Technology received an Honorable Mention.
ACM/IEEE CS George Michael Memorial HPC Fellowship Recipients
J. Gregory Pauloski (University of Chicago)
Pauloski is cited “for optimizing HPC resource usage via scalable optimization methods for deep learning training and improving the efficiency of data fabrics in applications spanning heterogeneous resources.” His research is grounded in developing tools and systems that enable efficient use of high-performance computing (HPC) resources. His research approaches HPC from multiple aspects: large-scale machine learning (ML) training and data fabrics that support distributed and federated scientific applications.
The rapid increase in demand for AI tools has promoted scalable ML training to a core challenge for HPC, so he developed KAISA, an optimization framework that adapts the memory footprint, communication frequency, and computation requirements for ML training on HPC clusters. While individual clusters become more powerful, advancements in specialized accelerators, edge devices, and cloud computing have encouraged the development of federated scientific applications that span heterogeneous systems. He leads the ProxyStore project which abstracts the routing of data between processes in distributed and federated applications, enabling dynamic selection of different data movement methods depending on what, where, and when data are moved.
Rohan Basu Roy (Northeastern University)
Roy is cited “for methods and tools that exploit cloud computing and on-premise HPC resources to enhance the productivity of computational scientists and the sustainability of HPC.” A Ph.D. student advised by Professor Devesh Tiwari in the Goodwill Computing Lab at Northeastern University, Roy designs new tools and methods for enhancing programmer productivity making large-scale computing systems more cost-effective and environmentally sustainable.
Roy’s HPC performance auto-tuner tools significantly improve the program productivity. While there has been plenty of work on HPC performance auto-tuners, Roy’s work bridges two important gaps: portability across different architectures to improve productivity, especially in the context of increasing hardware heterogeneity and the interpretability of different tuning configurations for programmers, leveraging game theory. He observed that despite highly-tuned codes, domain scientists are bottlenecked by long queue wait times.
Roy’s research contributions include the first demonstration of the effective use of the serverless computing model for complex scientific workflows — he demonstrates that, although not originally designed for HPC workloads, the serverless paradigm has significant productivity and performance advantages for HPC including quick elasticity in the cloud (eliminating the long queue wait time), and ease of use due to raised abstraction.
Unfortunately, raising the abstraction level via utilizing cloud computing resources can significantly increase the operational cost for computing service providers. Roy designs serverless-specific function prediction and resource allocation methods to reduce cost and improve resource utilization in large-scale computing systems – via injecting controlled hardware heterogeneity for serverless functions and opportunistic co-location of workloads. Opportunistic co-location of workloads in the cloud reduces the carbon footprint of on-premise HPC cluster/supercomputer, especially during periods of high carbon intensity in the geographical region of the supercomputer — Roy aims to improve the environmental sustainability of HPC systems as their carbon footprint expands rapidly due to ease of use, increasing scale of the system, and emerging compute-intensive applications.
ACM/IEEE CS George Michael Memorial HPC Fellowship Honorable Mention
Hua Huang (Georgia Institute of Technology)
Huang is cited “for contributions to high-performance parallel matrix algorithms and implementations and their application to quantum chemistry calculations.” His work focuses on improving the performance of matrix operations, a fundamental part in scientific computing and data analysis, with innovative algorithms. Huang’s main contributions include developing a high-performance, multi-purpose, rank-structured matrix library for multiple scientific computing tasks, designing innovative parallel algorithms for large-scale matrix operations, and new optimization strategies for quantum chemistry calculations. Hua’s work has been integrated into widely used codes such as Psi4, NWChem, and SPARC.
About the ACM/IEEE CS George Michael Memorial Fellowship
The ACM/IEEE CS George Michael Memorial HPC Fellowship is endowed in memory of George Michael, one of the founders of the SC Conference series. The fellowship honors exceptional PhD students throughout the world whose research focus is on high performance computing applications, networking, storage, or large-scale data analytics using the most powerful computers that are currently available. The Fellowship includes a $5,000 honorarium and travel expenses to attend the SC conference, where the Fellowships are formally presented.
About the IEEE Computer Society
Engaging computer engineers, scientists, academia, and industry professionals from all areas of computing, the IEEE Computer Society (IEEE CS) sets the standard for the education and engagement that fuels continued global technological advancement. Through conferences, publications, and programs, and by bringing together computer science and engineering leaders at every phase of their career for dialogue, debate, and collaboration, IEEE CS empowers, shapes, and guides the future of not only its members but the greater industry, enabling new opportunities to better serve our world. Visit www.computer.org for more information.
SC, the International Conference for High Performance Computing, sponsored by ACM and IEEE CS, offers a complete technical education program and exhibition to showcase the many ways high performance computing, networking, storage, and analysis lead to advances in scientific discovery, research, education, and commerce. This premier international conference includes a globally attended technical program, workshops, tutorials, a world-class exhibit area, demonstrations, and opportunities for hands-on learning.