Ronald G. Addie

Internet traffic modeling and future technology implications

This paper presents the Poisson Pareto burst process (PPBP) as a simple but accurate model for Internet traffic. It presents formulae relating the parameters of the PPBP to measurable traffic statistics, and describes a technique for fitting the PPBP to a given traffic stream. The PPBP is shown to accurately predict the queueing performance of a sample trace of aggregated Internet traffic. We predict that in few years, natural growth and statistical multiplexing will lead to an efficient optical Internet.

Fractal traffic: measurements, modelling and performance evaluation

Observations of both Ethernet traffic and variable bit rate (VBR) video traffic have demonstrated that these traffics exhibit self-similarity and/or infinite asymptotic index of dispersion for counts (IDC). We report here on measurements of traffic in a commercial public broadband network where similar characteristics have been observed. For the purpose of analysis and dimensioning of the central links of an ATM network we analyse in this paper the performance of a single server queue fed by Gaussian traffic with infinite IDC. The analysis lends to an approximation for the performance of a queue in which the arriving traffic is fractal Gaussian and consequently where there does not exist a dominant negative-exponential tail. The term fractal is used here in the sense that the autocovariance of the traffic exhibits self-similarity, that is to say, where the autocovariance of an aggregate of the traffic is the same, or asymptotically the same for large time lags, as the original traffic. We are not concerned with proving or exploiting this self-similarity property as such, but only with performance analysis techniques which are effective for such processes. In order to be able to test the performance analysis formulae, we show that traffic with the same autocovariance as measured in a real network over a wide range of lags (sufficiently wide a range for the traffic to be equivalent from the point of view of queueing performance) can be generated as a mixture of two Gaussian AR(1) processes. In this way we demonstrate that the analytic performance formulae are accurate.

Most probable paths and performance formulae for buffers with gaussian input traffic

[Abstract]: In this paper, performance formulae for a queue serving Gaussian traffic are presented. nThe main technique employed is motivated by a general form of Schilder’s theorem, the large ndeviation result for Gaussian processes. Most probable paths leading to a given buffer occupancy nare identified. Special attention is given to the case where the sample paths of the nGaussian process are smooth. The performance approximations are compared with known nanalytical results or by means of simulation. The approximations appear to be surprisingly naccurate.

Performance evaluation of a queue fed by a Poisson Pareto burst process

This paper provides means for performance evaluation of a queue with Poisson Pareto burst process (PPBP) input. Because of the long-range dependent nature of the PPBP, straightforward simulations are unreliable. New analytical and simulation techniques are described in this paper. Numerical comparison between the results shows consistency. Conservative dimensioning rules using zero buffer approximations are examined versus the more aggressive analytical approach based on the results of this paper to provide practical guidelines for network design.

Performance of a single server queue with self similar input

To evaluate analytically the performance of a discrete time single server queue where the input process has a general marginal distribution, its index of dispersion for counts (IDC) may be infinite, and where the tail of the unfinished work distribution is not necessarily dominant, we use the following two principles. Firstly, we use the observation that only a finite duration of time is required for consideration of the effect of correlation on queueing performance. Secondly, by a certain transformation, queueing performance results for a queue with correlated input having arbitrary marginal distribution can be evaluated using that of its noncorrelated counterpart. We confirm the results by simulation using real traffic measurements from a high speed data network.

Application of the M/Pareto process to modeling broadband traffic streams

In this paper we examine the usefulness of the M/Pareto process as a model for broadband traffic. We show that the queueing performance of the M/Pareto process depends upon the level of aggregation in the process. When the level of aggregation is high, the M/Pareto converges to a long-range-dependent Gaussian process. For lower levels of aggregation, the M/Pareto is capable of modeling the queueing performance of real broadband traffic traces.

Modeling superposition of many sources generating self similar traffic

This paper presents steps towards creating a general model for aggregated traffic streams. We show that fitting the mean, variance and Hurst parameter is insufficient to consistently model a long range dependent traffic stream. A fourth parameter, the level of aggregation, is required. We also show that with increased aggregation, the behaviour of a traffic stream tends towards that of a fractal Gaussian process.

A practical approach for multimedia traffic modeling

This paper presents the M/Pareto process as a practical model for multimedia traffic. We explain the M/Pareto model, discuss its advantages and limitations, and demonstrate its ability to accurately predict the queueing performance of multimedia traffic.

A Gaussian characterization of correlated ATM multiplexed traffic and related queueing studies

Results on performance statistics of a single server queue fed by a traffic stream which is modeled by a discrete time Gaussian process are presented. Because such a process can capture the full range of second-order statistics and is closed under superposition, it is a good model for correlated asynchronous transfer mode (ATM) statistical multiplexed traffic. The results include simple formulae for cell loss which can be applied to connection admission control for broadband integrated services digital networks (B-ISDNs).<<ETX>>

Admission control schemes for bursty multimedia traffic

We introduce a new framework for various measurement-based connection admission control (MBCAC) schemes for a multiservice network. Then, using heterogeneous bursty multimedia traffic traces, efficiency and simplicity tradeoffs are obtained by simulations. For all our MBCAC schemes, we use a procedure of available bandwidth evaluation based on online measurements and an adaptive feedback mechanism. The online measurement concept is based on the use of different traffic histograms that maintain records of the aggregated traffic in a link over a range of time-scales. The most complex MBCAC scheme involves measuring and then storing of all traffic information for every connection, the use of a warming up period technique, and updating the histograms when a connection departs. However, for the other simpler MBCAC schemes, various aspects of the most complex scheme are eliminated or simplified. We also consider two model-based CAC schemes, i.e., Gaussian and effective bandwidth, where a priori statistical knowledge of the connections are known in advance. Given that the traffic is known beforehand, the performance of such schemes will be better than if the statistics are not exactly known. A comparison between such model-based CAC schemes with our MBCAC schemes provides a benchmark that gives the best efficiency and QoS a model-based CAC may achieve. Simulation results demonstrate that while the best efficiency achievable by the model-based gives efficiency of 80%, the most complex MBCAC schemes achieved 81% efficiency, and the simplest MBCAC schemes obtained 76% value.