摘要: Meeting power requirements of huge exascale machines the future will be a major challenge. Our focus in this paper is to minimize cooling and we propose technique that uses combination DVFS temperature aware load balancing constrain core temperatures as well save energy. scheme specifically designed suit parallel applications which are typically tightly coupled. The control, comes at cost execution time try timing penalty. We experiment with three (with different utilization profiles), run on 128-core (32-node) cluster dedicated air conditioning unit. calibrate efficacy our based metrics: ability control average thereby avoiding hot spot occurence, penalty minimization, energy savings. results show savings up 57% 19%.