There is a performance-at-all-costs mentality at most of the nation's
supercomputing centers that has resulted in significant and growing
energy use.  This unchecked consumption costs the government a
considerable amount of money and wastes natural resources.  Moreover,
energy consumption and the resultant heat dissipation are becoming
important performance-limiting factors that we believe will eventually
come to bear on high-performance computing users.  The goal of our
research is to consume less energy (generating less heat) with no more
than a modest performance penalty and to do so without burdening
computational scientists.

This talk first discusses the energy consumption and execution time of
applications from a standard benchmark suite (NAS) on a power-scalable
cluster.  Our results show that many standard scientific applications
executed on a such cluster can save energy by reducing the processor
frequency and voltage without a significant increase in time.

Additionally, this talk presents two techniques for dynamically
controlling power consumption and increasing the energy efficiency of
application without undue performance penalties.  One technique
reduces CPU performance in phases that are not CPU-bound.  The other
technique reduces the CPU performance on nodes that are on the
critical path.  In both these situations, the application performance
is not heavily dependent on the CPU performance, thus the resultant
energy savings comes with little time increase.