Woong Shin PhD

Woong Shin, Ph.D. is a Research Scientist with over 15+ years of combined industry and research experience, spanning mission-critical enterprise software engineering and 8+ years of HPC operational data analytics, data-driven modeling, and sustainable supercomputing. After joining ORNL in 2017, his work focuses on developing AI/ML based systems for operational data analytics and improving HPC energy efficiency of supercomputers at Oak Ridge Leadership Computing Facility (OLCF). He holds a Ph.D. in Electrical Engineering and Computer Science from Seoul National University. His contributions have earned recognition, including a Best Paper Award at SC’21 and a Research Accomplishment Award from UT-Battelle.

 

Presentation Title:

Tools and methods for HPC application performance and energy analysis

Presentation Abstract:

As we enter the post-exascale era, energy has become a critical new dimension for HPC applications beyond traditional performance metrics. This creates a need for new classes of tools and methods that help application developers incorporate energy considerations into their efforts to extract maximum efficiency from underlying HPC systems.
In addressing this need, I will introduce OLCF’s ongoing efforts in enabling BYOM-style monitoring tools and EDP-based metrics. These are specifically designed to help HPC application developers balance energy and performance considerations in ways that align with system-level optimization goals such as high utilization and maximum operational efficiency. Drawing from preliminary studies across varied application scenarios, I will demonstrate how energy-aware optimizations manifest under different facility goals and show the crucial role of user-driven tools guided by these metrics in achieving system-wide efficiency improvements.