Addressing Power-Performance Tradeoffs for Data-Intensive Applications on Systems with Deep Memory Hierarchies

Overview of the Project:

The goal of this project is understanding power/performance behaviors and tradeoffs associated with data placement, data motion and data processing associated with data analytics pipelines on systems with emerging architectures and deep memory hierarchies, and to develop strategies that can fundamentally enable data-intensive workflows on current and future large-scale systems. In contrast to existing work on power management at different levels we address energy efficiency from an application- and data-centric perspective, and focus on optimizing data placement/motion and computation scheduling. It also addresses energy-efficiency in a holistic manner in combination with performance, resilience, quality of solution, and other objectives.


1) Janine C. Bennett, Sandia National Laboratories

2) Hemanth Kolla, Sandia National Laboratories

3) Jacqueline Chen, Sandia National Laboratories

4) Peer-Timo Bremer, Lawrence Livermore National Laboratory

5) Aaditya G. Landge, University of Utah

6) Attila Gyulassy, University of Utah

7) Patrick McCormick, Los Alamos National Laboratory

8) Scott Pakin, Los Alamos National Laboratory

9) Valerio Pascucci, University of Utah and PNNL

10) Stephen Poole, Oak Ridge National Laboratory


1) National Science Foundation

2) Department of Energy