An Efficient Approach for Adaptive Online Power Management in Mixed-Criticality Systems

In this paper, we present the first results known to date on applying the online dynamic power management to the mixed-criticality system, where tasks of varying criticality levels run in the same platform. We formulate the power management problem, and provide an efficient policy to adaptively manage the power mode based on the actual demand, in order to reduce the static power waste as much as possible. We explore three strategies to do the decision making of mode switching, and evaluate their effects on the online energy management. Experiments are conducted to show the effects of parameters on the energy savings.