Application Power Signature Analysis

Conference · Wed Jan 01 00:00:00 EST 2014

OSTI ID:1132366

Hsu, Chung-Hsing ^[1]; Combs, Jacob ^[2]; Nazor, Jolie ^[2]; Santiago, Fabian ^[2]; Thysell, Rachelle ^[2]; Rivoire, Suzanne ^[2]; Poole, Stephen W ^[1]

---

1. ORNL
2. Sonoma State University

The high-performance computing (HPC) community has been greatly concerned about energy efficiency. To address this concern, it is essential to understand and characterize the electrical loads of HPC applications. In this work, we study whether HPC applications can be distinguished by their power-consumption patterns using quantitative measures in an automatic manner. Using a collection of 88 power traces from 4 different systems, we find that basic statistical measures do a surprisingly good job of summarizing applications' distinctive power behavior. Moreover, this study opens up a new area of research in power-aware HPC that has a multitude of potential applications.

• Research Organization: Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States)
• Sponsoring Organization: Work for Others (WFO)
• DOE Contract Number: DE-AC05-00OR22725
• OSTI ID: 1132366
• Resource Relation: Conference: The 10th Workshop on High-Performance, Power-Aware Computing (HPPAC 2014), Phoenix, AZ, USA, 20140519, 20140519
• Country of Publication: United States
• Language: English