Gerenciamento de energia em sistemas embarcados.

Abstract

The complexity of battery-powered embedded system such as mobile phones, and personal digital assistants (PDAs) is growing quite fast. Devices based on high speed and multicore cpus having multiple cameras, display devices and several network interfaces increase the demand for more power. However, battery capacity does not grow at the same rate as it does the complexity of mobile embedded devices. Therefore, extending battery lifetime using power management strategies has became one of the key challenges in the design of complex mobile embedded systems. The dynamic power management techniques, named (DPM), allow power reduction at runtime by shutting down or reducing frequency or voltage of idle system components. An strategy of power management must consider the workload of the system. At a general purpose processor e.g. the combination of applications running on such system may vary strongly, depending on what is being executed. Moreover, the workload may vary drastically during the day, or over the days of the week, or when the system is operated by different users. It happens because of the nonstationarity of the workload. Besides, traditional power optimization strategies may not be optimal for battery-powered devices if the characteristics of the battery are not properly modeled and exploited. In order to optimize the battery lifetime, all of these must be taken into account: the model of the workload of the system, the electric parameters (e.g. values of currents) of the electronic system and the electrochemical features of the battery. A battery-aware DPM technique that exploits an acurate analytical battery model to increase battery lifetime in a non-stationary environment is proposed in this work. The system is modeled by discrete-time Markov chains coupled to the battery model. Such model allows a rigorous mathematical formulation of the problem and a trade-off between performance and battery lifetime. The proposed DPM technique has been simulated at Matlab and implemented using the Texas Instruments OMAP 1611 platform running Linux. Simulation and experimental results have shown that the technique introduced here results in longer battery lifetimes compared to previous DPM techniques.