Narrow Results

Power is increasingly becoming a design constraint for embedded systems. Dynamic Power Management algorithms enable optimal utilization of hardware at runtime. The present work attempts to arrive at an optimal policy to reduce the energy consumption at system level, by selectively placing components into low power states. A new, simple algorithm for power management systems involving multiple requests and services, proposed here, has been obtained from stochastic queuing models. The proposed policy is event driven and based on a Deterministic Markov Nonstationary Policy model (DMNSP). The proposed policy has been tested using a Java-based event driven simulator. The test results show that there is about 23% minimum power saving over the existing schemes with less impact on performance or reliability.

Call centers are service networks in which agents provide telephone-based services. An important part of call center operations is represented by service durations. In recent statistical analysis of real data, it has been noted that the distribution of service times reveals a remarkable fit to the lognormal distribution. In this paper, we discuss a possible source of this behavior by resorting to classical methods of statistical mechanics of multi-agent systems. The microscopic service time variation leading to a linear kinetic equation with lognormal equilibrium density is built up introducing as main criterion for decision a suitable value function in the spirit of the prospect theory of Kahneman and Twersky.