Title :
CPU MISER: A Performance-Directed, Run-Time System for Power-Aware Clusters
Author :
Ge, Rong ; Feng, Xizhou ; Feng, Wu-chun ; Cameron, Kirk W.
Author_Institution :
Dept. of Comput. Sci., Center for High-End Comput. Syst., Blacksburg, VA
Abstract :
Performance and power are critical design constraints in today´s high-end computing systems. Reducing power consumption without impacting system performance is a challenge for the HPC community. We present a runtime system (CPU MISER) and an integrated performance model for performance-directed, power-aware cluster computing. CPU MISER supports system-wide, application-independent, fine-grain, dynamic voltage and frequency scaling (DVFS) based power management for a generic power-aware cluster. Experimental results show that CPU MISER can achieve as much as 20% energy savings for the NAS parallel benchmarks. In addition to energy savings, CPU MISER is able to constrain performance loss for most applications within user-specified limits. These constraints are achieved through accurate performance modeling and prediction, coupled with advanced control techniques.
Keywords :
power aware computing; workstation clusters; CPU MISER; cluster computing; dynamic voltage frequency scaling; high-end computing systems; performance-directed run-time system; power aware cluster; power management; runtime system; Dynamic voltage scaling; Energy consumption; Energy management; Frequency; High performance computing; Performance loss; Power system management; Power system modeling; Predictive models; System performance;
Conference_Titel :
Parallel Processing, 2007. ICPP 2007. International Conference on
Conference_Location :
Xi´an
Print_ISBN :
978-0-7695-2933-2
DOI :
10.1109/ICPP.2007.29
