FFT-FGN is an S program for synthesizing a type of self-similar process known as fractional Gaussian noise. The program is fast but approximate.
The program requires the S statistical environment. I imagine it will also work with S-Plus, but haven't tried it.
Fractional Gaussian noise is only one type of self-similar process. When using this program for synthesizing network traffic, you must keep in mind that it may be that the traffic you seek is better modeled using one of the other processes.
The output of the program (perhaps transformed - see the documentation) describes an arrival process, not an interarrival process. That is, the output can be interpreted as a count of network arrivals during a particular time interval; but if your application requires exact arrival times, a separate mechanism is needed to transform the count into individual arrivals (doing this is a research area).
Written by Vern Paxson. If a publication studies traffic synthesized by this program, please include a corresponding citation to the paper mentioned below. A citation to the Internet Traffic Archive would be nice, too, but is not required.
Report bugs to vern@ee.lbl.gov.
Fast Approximation of Self-Similar Network Traffic, V. Paxson, technical report LBL-36750/UC-405, April 1995.
I am working with Murad Taqqu on a journal version of this paper.
The current release is 1.0. Updates will appear directly in the Internet Traffic Archive.
The software is copyrighted by the U.C. Regents (a "BSD-style" copyright). See the file COPYING either here or in the distribution for details. This copyright means you can redistribute the software freely, provided you keep the authorship information intact.
A short text file and the technical report mentioned above.
Available from the Archive in compressed tar format (7 KB).

Up to Software In The Internet Traffic Archive.