Optimal Resource Control of Multi-Processor Multi-Radio Nodes Using Semi-Markov Decision Processes

Abstract

An optimal resource control mechanism for a multiprocessor, multi-radio node architecture is proposed. To achieve our objective, a constrained semi-Markov decision problem is formulated, which not only provides optimal resource control but also meets quality of service demands imposed by application. For each sensed event, the desired energy efficiency and performance tradeoff is achieved by the optimal stochastic policy, which selects an appropriate set of processors and radios. The proposed solution is assessed using the reported energy consumption measurements for the existing platforms. Performance evaluation results point towards the importance of proper pairing of processor and radio in achieving the energy efficiency and performance tradeoff. The proposed solution can also be employed for hand held devices equipped with multiple processors and radios.