II-NEW: A Platform to Advance Research in Energy-Efficient Computing

Brown University, Providence RI

Investigators

Abstract

Computing power consumption is the most difficult and pervasive challenge towards improving the performance of computing systems. This project seeks the acquisition of a new class of cluster computing systems based on low-power System-on-a-Chip (SoC) processor nodes. Compared to traditional servers, this new class of servers promises large improvements in throughput per Watt. Using the acquired infrastructure, the project explores the ramifications of this new class of servers on the performance, energy consumption and thermal characteristics of application workloads. The results from different types of parallel workloads are analyzed to understand the interactions between application characteristics and system architecture and the impact of these interactions on performance and power consumption. The project also investigates the impact of this new class of servers on the effectiveness of power and thermal management techniques. The proposed infrastructure enables the pursuit of new research directions in energy-efficiency of computing systems. The results of this project will be disseminated in premier research venues, enabling the computing community to better understand the impact of emerging low-power SoC-based systems on the energy efficiency of computing clusters. The broader impacts of this project include outreach to local high school students in Providence, RI, and research experiences for undergraduate students at Brown University.