Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
FSD MODELS FOR OPEN-LOOP GENERATION OF INTERNET PACKET TRAFFIC 226 ABSTRACT OF PRESENTATION FSD Models for Open Loop Generation of Internet Packet Traffic William S.Cleveland, Bell Laboratories (with Jin Cao and Don X.Sun) Abstract: The packet traffic on an Internet link is a marked point process. The packet arrival times are the point process, and the packet sizes are the marks. The traffic arises from connections between pairs of computers; for each pair, the link is part of a path of links over which files are transferred between the computers; each file is broken up into packets on one computer, which are then sent to the other computer where they are reassembled to form the file. Packets arriving for transmissions on the link enter a queue. Many issues of Internet engineering depend heavily on the queue-length distribution, which in turn depends on the statistical properties of the packet process, so understanding and modeling the process are important for engineering. These statistical properties change as the mean connection load changes; consequently, the queuing characteristics change with the load. While much important analysis of Internet packet traffic has been carried out, comprehensive statistical models for the packet marked point process that reflect the changes in statistical properties with the connection load have not previously been developed. We introduce a new class of parametric statistical models fraction sum-different (FSD) models for the packet marked point process and describe the process we have used to identify the models and to then validate them. The models account for the changes in the statistical properties through different values of the parameters, and the parameters are modeled as a function of the mean load, so the modeling is hierarchical. The models are simple, and the simplicity enhances the basic understanding of traffic characteristics that arise from them. The models can be used to generate synthetic packet traffic for engineering studies; only the traffic load and certain parameters of the size marginal distribution that do not change with the load need to be specified. The mean load can be held fixed for the generation or can be varied. FSD models provides good fits to the arrivals and sizes provided the mean connection loadâthe mean number of simultaneous active connections using the linkâis above about 100. The models apply directly only to traffic where packets on the link input router delay only a small fraction of the packets, about 15 or less; but if delayed traffic is needed, it can be very simply generated by putting the synthetic model traffic through a queue. C code is available for generation as well as an implementation in the widely used NS-2 simulation system.