STATISTICAL ANALYSIS AND MODELING OF INTERNET TRAFFIC IP-BASED NETWORK FOR TELE-TRAFFIC ENGINEERING
This paper presents a statistical analysis and best fitted distribution model of internet traffic IP-based Network for tele-traffic engineering. One IP-based campus network architecture is studied which support of 16Mbps Committed Access Rate (CAR) speed line to Wide Area Network (WAN). Solarwinds network monitoring traffic toolbox is setup at the gateway router from inside campus network to the WAN in collecting real live throughput internet traffics. Daily throughput flows in Mbyte are
... ed in every ten minutes inter-arrival time. Statistical method on fitted Cumulative Distribution Function (CDF) is evaluated on collected throughput with Matlab software. Maximum Likelihood Estimator (MLE) technique is used to identify the maximum MLE log-likelihood which characterized as best fitted CDF distribution. Normal, Lognormal, Exponential and Weibull CDF fitted on throughput are presented. Among the four distributions, CDF Weibull is identified as the best traffic characteristic based on MLE maximum log-likelihood. Day7 is identified as best fitted that presents fitted 2-parameter Weibull which is Scale α =641.04 and Shape β =1.36 and fitted 3-parameter Weibull which is Scale α =551.76, Shape β =1.15 and threshold θ=63. Detail characteristics on day 7 and day 1 are presented and taken as benchmark model for future traffic algorithm.These results are valuable on modeling future tele-traffic engineering algorithm like policing, shaping, scheduling or queue which is based in real IP-based campus network environment. It is also useful for future prediction of tele-traffic models.