UC San Diego
MAC layer power management schemes for efficient energy- delay tradeoffs in wireless local area networks
- Author(s): Sarkar, Mahasweta
- et al.
In order to minimize power consumption and thereby prolong the system lifetime of battery powered wireless devices, it makes sense for such devices to transit to a very low power "Sleep" state when they are not communicating with their peers. However the main challenge of the sleep mechanism lies in the wireless nodes inability to "wake up" as soon as a packet arrives for it during its sleep state. This leads to an obvious tradeoff between power saving and packet delay. In this dissertation we are interested in the problem of optimizing the timing and duration of sleep states of wireless nodes in an infrastructure WLAN scenario with the objective of minimizing average overall system power consumption with respect to a QoS constraint. The QoS parameter we have focused on is average packet delay. We first considered a simple model comprising of a single transmitter and a single receiver. We formulated this as an optimization problem and solved it numerically using dynamic programming. We were able to derive closed form expressions for the optimal sleep duration for a given packet delay, as well as the associated minimal rate of power consumption. We extended the model to a multi-user system. Results indicated that the optimal policy for a specific node was a function of its buffer length as well as the sleep states of the other nodes in the system. We next considered the problem of scheduling multiple streams with either same or different packet delay constraints. We proposed an adaptive sleep-scheduling algorithm based on a heuristic derived from the results observed in the dynamic programming formulation. Our algorithm has three different sleep scheduling schemes - Round Robin scheme, Shortest Sleep First scheme and the Steep Descent method. We compare and contrast each of these schemes to the theoretical lower bound that we have previously computed by the method of dynamic programming. Results show that our algorithm performs favorably when compared to the lower bound. Further, in order to evaluate the performance of the Steep Descent method we construct a simplified simulation of the 802.11 sleep scheduling algorithm. Our simulation results show a substantial improvement in the performance of our new sleep scheduling policy when compared to the static sleep schedule of 802.11 especially in scenarios where traffic conditions change slowly over time which we model as a three state Markov process