Application of stochastic hybrid systems in power management of streaming data

In this paper, we study the optimal power management problem for a pipeline of streaming data consisting of several components and buffers in between. The production rate of the source component is assumed to be random. We aim to find the optimal switching strategy and the optimal size of the buffers so that the expected average power consumption of the pipeline system is minimized. For the case of two components with one buffer in between, we model the system by a stochastic hybrid system, and derive analytically the solution to the optimal power management problem.