Degree Name

Master of Engineering (Hons.)


Department of Electrical and Computer Engineering


There are two groups of traffic models that usually used in telecommunication network design, planning and performance evaluation. They are aggregate and individual models. In case of individual models, measurement-based studies are the most common approach used to obtain an accurate traffic model. The common assumption used in this approach is that the data link layer protocol (i.e., layer 2 of OSI Reference Model) does not significantly affect traffic characteristics of its higher layer where the traffic characteristics are of interest. In other words, this hypothesis assumes that traffic characteristics at the input of the data link protocol is statistically the same as the traffic characteristics at its output, i.e., the physical layer where the measurement can be easily done. This hypothesis is addressed as the null hypothesis. This thesis is concerned with testing the validity and investigating the sensitivity of the null hypothesis against several parameters. For these purposes, an Ethernet network simulation model is developed and several statistical methods are utilised. Several simulation network configurations are considered as test configurations which represent different combination of test parameters, namely network parameters, protocol parameters and user parameters. Those configurations are used to test the null hypothesis under two different traffic source models, i.e., Poisson and Pareto traffic models. Statistical test results indicated that the validity of the null hypothesis mostly depends on traffic model under consideration and is sensitive to the activity level of monitored station. The comprehensive statistical test results concluded that the null hypothesis is valid for relatively low-level activity stations (stations having packet departure rate up to 10 packets/second) and sensitive only to the user parameter of user activity level. In addition, this thesis also presents a case study that demonstrates the liability of applying the null hypothesis in measurement-based traffic modelling without further formal validation.