Synthetic Traffic Generation Techniques For ATM Network Simulations

Abstract

This paper presents a synthetic traffic modeling approach to generate bursty, long-range dependent (LRD) traffic flows for ATM network simulations. The approach uses Fractional-ARIMA processes and a set of mathematical transformations to generate traffic streams with a wide range of user-specified traffic characteristics. With this modeling approach, the user can control the short-range and longrange correlation structure in the generated traffic, as well as the marginal distribution. A description of the techniques is provided, along with a case study illustrating the use of these traffic generation techniques in a capacity planning study for ATM networks.

The synTraff suite of traffic modeling toolkits

“…This paper has presented a set of traffic modeling toolkits capable of generating and analyzing monofractal and multifractal traffic. The traffic modeling toolkits are based on traffic modeling techniques from the published literature [4,6,9,12,13], and have an easy-to-use graphical interface written in Tcl/Tk. We have found the toolkits useful for generating independent monofractal and multifractal streams for network performance studies [1,2].…”

“…Scale widgets constrain users to provide valid specifications. The LRD time series is generated using an F-ARIMA process (Fractional Auto-Regressive Integrated Moving Average) [13]. This process has…”

“…The toolkits can be used to model monofractal and multifractal traffic [1]. The synTraff toolkit uses a three-step modeling approach [13] based on F-ARIMA processes to generate monofractal traffic, which can be described using the three Norros traffic parameters ( , ¡ , and ¢ ) [10]. The other toolkits are for modeling multifractal traffic: Wsyn-Traff implements the Wavelet-domain Independent Gaussian (WIG) model [9], and MsynTraff implements the Multifractal Wavelet Model (MWM) [12].…”

A comparison of Markovian Arrival and ARMA/ARTA processes for the modeling of correlated input processes

“…Since MAPs only generate valid interarrival times we also considered this aspect and compared the characteristics of the original traces with those of the traces generated from the fitted processes where we took only feasible values into account by ignoring negative values or by replacing them (by 0). Another possibility to avoid infeasible values is to transform the generated traces of the fitted processes as for example described in (Williamson 1999). We do not consider this approach here since the determination of the parameters for the transformation requires additional expert knowledge and is not straightforward.…”

Simulating Stochastic Processes with OMNeT++

“…This issue can be addressed by transforming the Zt, e.g. [31] discusses several linear and non-linear transformations and its properties.…”