MCMC simulation of GARCH model to forecast network traffic load
The performance of a computer network can be enhanced
by increasing number of servers, upgrading the hardware,
and gaining additional bandwidth but this solution require
the huge amount to invest. In contrast to increasing the
bandwidth and hardware resources, network traffic
modeling play a significant role in enhancing the network
performance. As the emphasis of telecommunication
service providers shifted towards the high-speed networks
providing integrated services at a prescribed Quality of
Service (QoS), the role of accurate traffic models in
network design and network simulation becomes ever
more crucial. We analyze a traffic volume time series of
internet requests made to a workstation. This series
exhibits a long-range dependence and self-similarity in
large time scale and exhibits multifractal in small time
scale. In this paper, for this time series, we proposed
Generalized Autoregressive Conditional Heteroscedastic,
(GARCH) model, and practical techniques for model
fitting, Markov Chain Monte Carlo simulation and
forecasting issues are demonstrated. The proposed model
provides us simple and accurate approach for simulating
internet data traffic patterns.
Keywords: GARCH, Simulation, forecasting, MCMC, network traffic, load
Download Full-Text
ABOUT THE AUTHORS
Akhter Raza Syed
Department of Computer Science, University of Karachi, Karachi, 75270, Pakistan
Hussain Saleem
Department of Computer Science, University of Karachi, Karachi, 75270, Pakistan
Habib-Ur-Rehman Syed
Department of Mathematics, Hamdard Institute of Management Sciences, Karachi, Pakistan
Akhter Raza Syed
Department of Computer Science, University of Karachi, Karachi, 75270, Pakistan
Hussain Saleem
Department of Computer Science, University of Karachi, Karachi, 75270, Pakistan
Habib-Ur-Rehman Syed
Department of Mathematics, Hamdard Institute of Management Sciences, Karachi, Pakistan