Advancing HPC and ML Systems via Efficient Data Management

Dingwen Tao, Associate Professor, Indiana University

Wegmans Hall 1400

Abstract: The new generation of supercomputers is exascale (10¹⁸ floating-point operations per second) computer systems. These systems can help scientists and engineers tackle extremely complex high-performance computing (HPC) and machine learning (ML) problems for critical societal challenges, such as climate change, water management, advanced manufacturing, and vaccine and drug design. However, due to the gap between ever-increasing compute power and limited memory/storage capacity and I/O bandwidth, scientists and engineers must create intelligent and effective methods to efficiently manage massive amounts of data generated by HPC and ML applications for fast storage and transmission. This talk will introduce our promising solution – error-bounded lossy compression – that significantly reduces the data sizes while maintaining high data fidelity, which can benefit data management (including I/O, memory, and storage) in many in HPC and ML applications. The talk will cover the design, optimization, and use of our error-bounded lossy compression to advance HPC and ML systems for large-scale data processing applications such as HPC simulations, ML model training, large graph analytics.

Bio: Dingwen Tao is an associate professor in Intelligent Systems Engineering at Indiana University’s Luddy School of Informatics, Computing, and Engineering. At Luddy, he leads the High-Performance Data Analytics and Computing (HiPDAC) research group. Before joining IU, he worked as an assistant professor at Washington State University and University of Alabama in between 2018 and 2022. Prior to that, he worked in multiple DOE national laboratories including Brookhaven National Laboratory, Argonne National Laboratory, and Pacific Northwest National Laboratory. He has won multiple awards, including Meta Research Award (2022), R&D100 Awards Winner (2021), IEEE Computer Society TCHPC Early Career Researchers Award for Excellence in High Performance Computing (2020), and NSF CISE Research Initiation Initiative (CRII) Award (2020). He is serving on the Technical Review Board (TRB) of IEEE Transactions on Parallel and Distributed Systems (TPDS). He served as the Program Co-chair of 2021 IEEE International Conference on Scalable Computing and Communications and International Workshops on Big Data Reduction. He has published in the top-tier HPC and big data conferences and journals, including SC, ICS, HPDC, PPoPP, DAC, PACT, IPDPS, CLUSTER, ICPP, BigData, IEEE TC, IEEE TPDS, etc. His research has been supported by NSF, DOE, NOAA, AMD, Meta, and Xilinx.

Refreshments will be provided.