Abstract:

      Existing predictive dynamic power management(DPM) policies based on correlations of consecutive idle times are ineffective under stochastic service requests. In view of statistics, this paper proposes an energy expectation based (EEB) predictive policy. EEB makes components into a mode of lowest energy expectation of idle times. It is shown that the method optimizes energy saving. Simulation results show that, under exponential distribution and Pareto distribution, EEB acquires lower power of component compared with traditional predictive policies.

Key words: energy expectation , predictive policy , dynamic power management , low power