Adaptive Policing Algorithms on inbound internet traffic using Generalized Pareto model

Abstract

This paper present an analysis of live internet traffic and development of an Adaptive Policing Algorithms to control burst traffic based on fitted traffic model. Objectives of this research is to characterize inbound IP-based campus internet traffic, then traffic is fitted to 2-parameters Cumulative Distribution Function (CDF) traffic model. A Percentage level Policing and algorithm is developed to control the bandwidth used. Open Distribution Fitting application is used to fit to the collected data. Maximum Log likelihood estimation technique is used to fit the best 2-parameter CDF which are Generalized Pareto, Weibull, Normal and Rician distribution model. Results presents best CDF fitted model is Generalized Pareto which present highest maximum likelihood value for this case. Thus, a percentage level of 5% under original bandwidth used is developed on policing algorithms to control internet bandwidth using Pareto traffic model. Result present performances upgraded around 3% to 5% of time processing and approximately 74% of bandwidth saved with Gen Pareto model. This result help to expand the view of new idea in modelling the tele-traffic algorithm based on bandwidth management and time processing improvement. Control algorithms on bandwidth can be developed especially on new Software Defined Network with this algorithms.