|
This research project focuses on developing new formulations and analysis strategies to support the increasing volumes and rates at which scientific simulations running at extreme scale generate data, which needs to be transported and analyzed before scientific discovery can be realized.
The overarching goal is exploring data-related energy/performance trade-offs at extreme scales. Specifically, it aims at analyzing the behavior of large-scale simulation workflows with an in-situ and other data analytics pipelines, running on a current high-end computing platform and beyond to develop performance and power models, which can be validated using an instrumented platform. Models can be used then to explore energy/performance tradeoffs on current systems, to help answer system design questions, and to analyze the power requirements and usage modes for emerging architectures.
To enable this research, a key element of this project is the development of a prototype meta-skeleton simulation platform to support experimental data analytics research and explore different types of data analytic workflows at scale.This simulation platform will consider different parameters such as: 1) algorithm designs, 2) deployment/runtime options; 3) data exchange patterns, and 4) system architecture. It will enable the research of energy-performance behaviors and tradeoffs and studying system/software co-design of future architectures.
|
Sign in
to view more information about this project.
