Capturing the spatio-temporal behavior of real traffic data

Traffic data, such as disk and memory accesses, typically exhibits burstiness, temporal locality, and spatial locality. However, except for qualitative speculations, it is not even known how to measure the spatio-temporal correlation, let alone how to re-produce it realistically. In this paper, we propose the “entropy plots” to quantify the correlation and develop a new statistical model, the “PQRS” model, to capture the burstiness and correlation of the real spatio-temporal traffic. Moreover, the model requires very few parameters and offers linear scalability. Experiments with multiple real data sets show that our model can mimic real traces very well.