Applicability of different models of burstiness to energy consumption estimation

In this paper we investigate alternative ways to model traffic burstiness with the aim of finding a model which allows mapping of traffic characteristics to energy consumption estimates. The goal is to find a model that would be simple enough to be useful for application and service developers to study how their design decisions influence mobile energy consumption. We investigate alternative traffic models, poisson model, self-similar model, and on/off model, to see how suitable they are for the characterization of WiFi traffic and for the estimation of energy consumption. We generate both bursty and smooth traffic with same average bandwidth and measure the difference in energy consumption. As expected, bursty traffic consumes less energy in the mobile device compared to the smooth data traffic because the radio interface can spend part of the time in a low-power idle state. We then fit on/off model to the measurement results and see that it can characterize the traffic very well and that it can be used to derive a good estimate for energy consumption. In addition to the extreme cases, we also apply the on/off model to YouTube download in order to see how the model works for real traffic. Our key finding is that the on/off model serves well as a simple way for traffic characterization and it can be used to estimate the energy consumption fairly well.