Analysis and Modeling of VoIP Conversation Traffic in the Real Network

This paper presents the characteristic and traffic modeling of VoIP conversation. The traffic data are measured from the operating IP network of Telephone Organization of Thailand (TOT) Corporation. The observed distributions of talkspurt and silent durations considerably differ from the standard ON-OFF model (ITU-T P.59). It is proposed that a new state "long burst" representing the background noise at the talker\´s place, is added into the model based on continuous-time Markov process. The other two states, "normal speech" and "long silence", represent the normal behavior of the VoIP user. If background noise occurs during the speech, the model is classified as "noisy speech" otherwise as "noiseless speech". The simulation of traffic aggregation shows that the background noise significantly increases the data rate mean but reduces the variance. The measurement results support that both directions of the conversation can be modeled by the same model