Fitting mixtures of exponentials to long-tail distributions to analyze network performance models

@article{Feldmann1997FittingMO,
  title={Fitting mixtures of exponentials to long-tail distributions to analyze network performance models},
  author={Anja Feldmann and Ward Whitt},
  journal={Proceedings of INFOCOM '97},
  year={1997},
  volume={3},
  pages={1096-1104 vol.3}
}
Traffic measurements from communication networks have shown that many quantities characterizing network performance have long-tail probability distributions, i.e., with tails that decay more slowly than exponentially. Long-tail distributions can have a dramatic effect upon performance, but it is often difficult to describe this effect in detail, because performance models with component long-tail distributions tend to be difficult to analyze. We address this problem by developing an algorithm… 

Figures from this paper

Accurate long-tailed network traffic approximation and its queueing analysis by hyper-Erlang distributions
TLDR
A hyper-Erlang model (mixed Erlang distribution) is proposed for such a long-tailed network traffic approximation that is more accurate in fitting the tail behavior and also computationally efficient.
Fitting heavy-tailed HTTP traces with the new stratified EM-algorithm
  • R. Sadre, B. Haverkort
  • Computer Science
    2008 4th International Telecommunication Networking Workshop on QoS in Multiservice IP Networks
  • 2008
TLDR
This paper presents a data aggregation algorithm which accelerates the fitting of heavy- tailed distributed data to hyper-exponential distributions and illustrates the performance of the algorithm by applying it to empirical and artificial data traces.
Modeling and Analysis of Heavy-Tailed Distributions viaClassical Teletra
TLDR
This derivation shows that classical teletraac methods can be employed for the analysis of heavy-tailed distributions and allows to derive both quantitative numerical results and asymptotic closed-form results.
1 Numerical Methods for Analyzing Queues with Heavy-Tailed Distributions
TLDR
Numerical issues of implementing the Transform Approximation Method for simple queueing systems are investigated and it is argued that TAM can be used in conjunction with the Fourier-series method for inverting Laplace transforms, even though TAM is a discrete approximation and the Fouriers method requires a continuous distribution.
Efficient fitting of long-tailed data sets into phase-type distributions
TLDR
The new method is accurate, efficient, and allows one to apply existing analytic tools to analyze the behavior of queueing systems that operate under workloads that exhibit long-tail behavior, such as queues in Internet-related systems.
Recursive EM algorithm for finite mixture models with application to Internet traffic modeling
TLDR
This paper uses mixture distributions to model Internet traffic and applies the EM algorithm to fit the models, and the Bayesian information criterion is applied to select the best model.
How heavy-tailed distributions affect simulation-generated time averages
TLDR
This study compares the accuracy of point estimators based on the Lomax distribution with those for lognormal and Weibull file-size distributions whose parameters are determined by matching their means and a selected extreme quantile with those of the LOMax.
Modeling service-time distributions with non-exponential tails : Beta mixtures of exponentials
Motivated by interest in probability density functions (pdf's) with nonexponential tails in queueing and related areas, we introduce and investigate two classes of beta mixtures of exponential pdf's.
Performance Evaluation with Heavy Tailed Distributions
Over the last decade an important new direction has developed in the performance evaluation of computer systems: the study of heavy-tailed distributions. Loosely speaking, these are distributions
...
1
2
3
4
5
...

References

SHOWING 1-10 OF 102 REFERENCES
Waiting-time tail probabilities in queues with long-tail service-time distributions
TLDR
Algorithms for computing the waiting-time distribution by Laplace transform inversion when the Laplace transforms of the interarrival-time and service-time distributions are known are developed and a convenient two-parameter family of long-tail distributions on the positive half line with explicit Laplace transformations is introduced.
On approximations for queues, III: Mixtures of exponential distributions
  • W. Whitt
  • Mathematics
    AT&T Bell Laboratories Technical Journal
  • 1984
TLDR
It is shown that additional information about the distributions is more important for the interarrival-time than for the service time, and one kind of shape constraint — assuming that the distribution is a mixture of exponential distributions — can be examined analytically.
Wide-area traffic: the failure of Poisson modeling
TLDR
It is found that user-initiated TCP session arrivals, such as remote-login and file-transfer, are well-modeled as Poisson processes with fixed hourly rates, but that other connection arrivals deviate considerably from Poisson.
Planning queueing simulations
TLDR
Simple heuristic formulas are developed to estimate the simulation run lengths required to achieve desired statistical precision in queueing simulations and apply to stochastic processes that can be approximated by reflected Brownian motion, such as the queue-length process in the standard GI/G/1 model.
Modeling call holding time distributions for CCS network design and performance analysis
  • V. Bolotin
  • Business, Computer Science
    IEEE J. Sel. Areas Commun.
  • 1994
TLDR
Analysis of actual holding time distributions in conversations, facsimile and voice mail connections revealed that these distributions radically differ from the exponential distribution, which can have a significant impact on the traffic stream presented to the CCS network.
Can Self-Similar Traffic Be Modeled by Markovian Processes?
  • S. Robert, J. Boudec
  • Computer Science
    International Zurich Seminar on Digital Communications
  • 1996
TLDR
It is shown that a simple 5-state Markov process integrating different time scales can reasonably model the behavior of measured traffic up to a certain time interval.
The Effect of Multiple Time Scales and Subexponentiality in MPEG Video Streams on Queueing Behavior
TLDR
A video model is constructed that captures both multiple time scale and subexponential characteristics of real-time MPEG video streams, and a fluid model, whose arrival process is obtained from the video data by replacing scene statistics with their means, is shown to asymptotically converge to the exact queue distribution.
Self-similarity through high-variability: statistical analysis of ethernet LAN traffic at the source level
TLDR
This paper provides a plausible physical explanation for the occurrence of self-similarity in high-speed network traffic based on convergence results for processes that exhibit high variability and is supported by detailed statistical analyses of real-time traffic measurements from Ethernet LAN's at the level of individual sources.
Economies of scale in queues with sources having power-law large deviation scalings
We analyse the queue Q L at a multiplexer with L sources which may display long-range dependence. This includes, for example, sources modelled by fractional Brownian motion (FBM). The workload
...
1
2
3
4
5
...