Skip to content. | Skip to navigation

You are here: Home » CSCI » Pressroom » Colloquium » Grad Center CS Colloquium
Document Actions

Grad Center CS Colloquium

Characterizing and optimizing workflow performance on Department of Energy (DOE) supercomputers

When Apr 16, 2021
from 10:00 am to 11:00 am
Hosting organization Graduate Center Computer Science
Speaker Devarshi Ghoshal
Speaker Information Dr. Ghoshal is a Research Scientist in the Data Science and Technology department at the Lawrence Berkeley National Lab, Department of Energy. He received his Ph.D. in Computer Science from Indiana University, and joined the Berkeley Lab as a Postdoctoral Fellow in 2014. His research focuses on efficient management and analysis of large, complex scientific data on high performance computing and distributed systems. He has an extensive publication record at high performance computing conferences as well as book chapters. His work on scientific and cloud data management has received multiple best paper awards. He is in the program committee and review board of multiple top-tier conferences and peer-reviewed journals.
Where On-line
Contact Name Saptarshi Debroy
Contact Email
Add event to calendar vCal

Scientists are increasingly processing and analyzing experimental and observational data as workflows on Department of Energy (DOE) supercomputers. There is a need to optimize the performance of these data-intensive scientific workflows on supercomputing resources. However, there are not any well-established methods for analyzing and characterizing scientific workflows on high performance computing systems. In this talk, I am going to present our methodologies towards understanding and optimizing the performance of data-intensive scientific workflows on DOE supercomputers. We present a data-driven approach to analyze and influence the design of next-generation hardware and software infrastructure for efficiently managing workflows. The methods, models, and results presented in this talk are critical for analyzing and optimizing workflows well beyond the DOE ecosystem.