TY - GEN
T1 - CPU gadients
T2 - 2010 International Conference on Green Computing, Green Comp 2010
AU - Chen, Shuyi
AU - Joshi, Kaustubh R.
AU - Hiltunen, Matti A.
AU - Schlichting, Richard D.
AU - Sanders, William H.
PY - 2010
Y1 - 2010
N2 - Dynamic voltage and frequency scaling (DVFS) and virtual machine (VM) based server consolidation are well-known CPU scaling techniques for energy conservation that can have an adverse impact on system performance. For the responsiveness-sensitive multitier applications running in today's data centers, queuing models should ideally be used to predict the impact of CPU scaling on response time, to allow appropriate runtime trade-offs between performance and energy use. In practice, however, such models are difficult to construct and thus are often abandoned for ad-hoc solutions. In this paper, an alternative measurement-based approach that predicts the impact without requiring detailed application knowledge is presented. The approach proposes a new predictive model, the CPU gradient, that can be automatically measured on a running system using lightweight and nonintrusive CPU perturbations. The practical feasibility of the approach is demonstrated using extensive experiments on multiple multitier applications, and it is shown that simple energy controllers can use gradient predictions to derive as much as 50% energy savings while still meeting response time constraints.
AB - Dynamic voltage and frequency scaling (DVFS) and virtual machine (VM) based server consolidation are well-known CPU scaling techniques for energy conservation that can have an adverse impact on system performance. For the responsiveness-sensitive multitier applications running in today's data centers, queuing models should ideally be used to predict the impact of CPU scaling on response time, to allow appropriate runtime trade-offs between performance and energy use. In practice, however, such models are difficult to construct and thus are often abandoned for ad-hoc solutions. In this paper, an alternative measurement-based approach that predicts the impact without requiring detailed application knowledge is presented. The approach proposes a new predictive model, the CPU gradient, that can be automatically measured on a running system using lightweight and nonintrusive CPU perturbations. The practical feasibility of the approach is demonstrated using extensive experiments on multiple multitier applications, and it is shown that simple energy controllers can use gradient predictions to derive as much as 50% energy savings while still meeting response time constraints.
UR - http://www.scopus.com/inward/record.url?scp=78449305816&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=78449305816&partnerID=8YFLogxK
U2 - 10.1109/GREENCOMP.2010.5598296
DO - 10.1109/GREENCOMP.2010.5598296
M3 - Conference contribution
AN - SCOPUS:78449305816
SN - 9781424476138
T3 - 2010 International Conference on Green Computing, Green Comp 2010
SP - 15
EP - 29
BT - 2010 International Conference on Green Computing, Green Comp 2010
Y2 - 15 August 2010 through 18 August 2010
ER -