skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Analysis of Application Power and Schedule Composition in a High Performance Computing Environment

Abstract

As the capacity of high performance computing (HPC) systems continues to grow, small changes in energy management have the potential to produce significant energy savings. In this paper, we employ an extensive informatics system for aggregating and analyzing real-time performance and power use data to evaluate energy footprints of jobs running in an HPC data center. We look at the effects of algorithmic choices for a given job on the resulting energy footprints, and analyze application-specific power consumption, and summarize average power use in the aggregate. All of these views reveal meaningful power variance between classes of applications as well as chosen methods for a given job. Using these data, we discuss energy-aware cost-saving strategies based on reordering the HPC job schedule. Using historical job and power data, we present a hypothetical job schedule reordering that: (1) reduces the facility's peak power draw and (2) manages power in conjunction with a large-scale photovoltaic array. Lastly, we leverage this data to understand the practical limits on predicting key power use metrics at the time of submission.

Authors:
 [1];  [1];  [1];  [1];  [1]
  1. National Renewable Energy Lab. (NREL), Golden, CO (United States)
Publication Date:
Research Org.:
National Renewable Energy Lab. (NREL), Golden, CO (United States)
Sponsoring Org.:
USDOE
OSTI Identifier:
1235236
Report Number(s):
NREL/TP-2C00-65392
DOE Contract Number:  
AC36-08GO28308
Resource Type:
Technical Report
Country of Publication:
United States
Language:
English
Subject:
97 MATHEMATICS AND COMPUTING; supercomputer; data center; power scheduling; campus integration

Citation Formats

Elmore, Ryan, Gruchalla, Kenny, Phillips, Caleb, Purkayastha, Avi, and Wunder, Nick. Analysis of Application Power and Schedule Composition in a High Performance Computing Environment. United States: N. p., 2016. Web. doi:10.2172/1235236.
Elmore, Ryan, Gruchalla, Kenny, Phillips, Caleb, Purkayastha, Avi, & Wunder, Nick. Analysis of Application Power and Schedule Composition in a High Performance Computing Environment. United States. https://doi.org/10.2172/1235236
Elmore, Ryan, Gruchalla, Kenny, Phillips, Caleb, Purkayastha, Avi, and Wunder, Nick. 2016. "Analysis of Application Power and Schedule Composition in a High Performance Computing Environment". United States. https://doi.org/10.2172/1235236. https://www.osti.gov/servlets/purl/1235236.
@article{osti_1235236,
title = {Analysis of Application Power and Schedule Composition in a High Performance Computing Environment},
author = {Elmore, Ryan and Gruchalla, Kenny and Phillips, Caleb and Purkayastha, Avi and Wunder, Nick},
abstractNote = {As the capacity of high performance computing (HPC) systems continues to grow, small changes in energy management have the potential to produce significant energy savings. In this paper, we employ an extensive informatics system for aggregating and analyzing real-time performance and power use data to evaluate energy footprints of jobs running in an HPC data center. We look at the effects of algorithmic choices for a given job on the resulting energy footprints, and analyze application-specific power consumption, and summarize average power use in the aggregate. All of these views reveal meaningful power variance between classes of applications as well as chosen methods for a given job. Using these data, we discuss energy-aware cost-saving strategies based on reordering the HPC job schedule. Using historical job and power data, we present a hypothetical job schedule reordering that: (1) reduces the facility's peak power draw and (2) manages power in conjunction with a large-scale photovoltaic array. Lastly, we leverage this data to understand the practical limits on predicting key power use metrics at the time of submission.},
doi = {10.2172/1235236},
url = {https://www.osti.gov/biblio/1235236}, journal = {},
number = ,
volume = ,
place = {United States},
year = {Tue Jan 05 00:00:00 EST 2016},
month = {Tue Jan 05 00:00:00 EST 2016}
}