CAREER: Heterogeneity-Enriched Communication for Advancing HPC Systems and Applications

In an era where High-Performance Computing (HPC) cyberinfrastructure systems are undergoing rapid and transformative evolution, marked by a surge in heterogeneity and scale, this project stands at the forefront of transformative innovation in HPC cyberinfrastructure, where rapid development in heterogeneity and scale presents both opportunities and challenges. Its significance lies in its unwavering commitment to advancing the usability, efficiency, and scalability of HPC systems and applications. By pioneering Heterogeneity-Enriched Communication designs and software, this project not only propels the field forward but also resonates with the broader mission of the National Science Foundation (NSF), which aspires to promote the progress of science, advance national welfare, and secure the national defense. Beyond its scientific impact, this endeavor prioritizes education and diversity, fostering a learning culture and addressing workforce shortages in critical STEM (Science, Technology, Engineering, and Mathematics) fields. It seeks to democratize knowledge and empower the scientific community through open-source principles, amplifying its potential benefits for data-intensive scientific research. The primary objective of this project is to pioneer innovative heterogeneous-enriched communication designs for modern and next-generation HPC systems and applications, significantly enhancing their usability, efficiency, and scalability in the face of rapid evolution characterized by increased heterogeneity and scale. This will be achieved through a multi-pronged approach involving analytical modeling and architectural performance optimization. This research project targets three key challenges: (a) achieving high usability by developing precise cost modeling, prediction, and simulation for heterogeneity-enriched communication; (b) addressing efficiency through the composition of innovative communication schemes and adaptive orchestration; and (c) ensuring scalability through optimization techniques that reduce message and connection overhead. These designs, released as open-source tools and integrated into existing HPC libraries, can revolutionize HPC and Machine Learning (ML) applications across domains such as agriculture, industrial workloads, biostatistics, and geospatial information science. Furthermore, this project aligns with the NSF's core mission to advance the national welfare by actively fostering inclusivity, education, and underrepresented groups in STEM. It aims to drive innovation and leave a lasting impact on the scientific community. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2623548 | Program: 01002425DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Xiaoyi Lu | Institution: University of Florida, GAINESVILLE, FL | Award Amount: $482,193 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2623548 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2623548.html