Traffic modeling based on FARIMA models

We provide a procedure to fit a FARIMA(p,d,q) (fractional autoregressive integrated moving average) model to the actual traffic trace, as well as a method to generate a FARIMA process with given parameters. We show how to model the traffic by fitting FARIMA models to four measured traces. Our experiments illustrate that the FARIMA model is a good traffic model and is capable of capturing the property of real traffic with long-range and short-range dependent behavior. Unlike previous work on FARIMA models, we deduce some guidelines to reduce the complexity of fitting the FARIMA model which would allow us to reduce the computational time of fitting.