Does the TES stitching function merely stich?

TES (transform-expand-sample) is among the younger traffic modeling techniques. Its basic scheme consists of creating a randomly governed (background) sequence, smoothing it by a so-called stitching function and adapting it to a given marginal distribution. It is demonstrated that the stitching function in fact is much more powerful as its shape has crucial influence on the visual appearance of the resulting (foreground) sequence and the autocorrelation function. Using suitable generalized stitching functions is a very simple way to receive simulated sample paths being visually indistinguishable from the original time series. This method is demonstrated especially for the case of periodic time series (e.g. network utilization and MPEG frame sizes). Finally, it turns out that it could also be useful for simulating self-similar traffic. Hence, this approach allows one to model a couple of characteristic features of given traffic in great detail while still supporting the major advantages of TES