Call-level VoIP traffic modelling based on data from a real-life VoIP service provider

This paper presents a call-level traffic analysis and characterization, based on real data collected from a VoIP service provider, for a time period of almost one year. This traffic includes calls within the VoIP network boundaries and others that are either initiated or terminated by other telephone networks, since most contemporary VoIP networks interconnect with other landline or cellular telephony networks. The traffic is originated by a multitude of user profiles. However on the calllevel basis, it can be described basically by two main processes, namely the Call Arrival and the Call Duration. Hence the aim of this work is the examination of the call arrival pattern and the extraction of a statistical distribution that can accurately describe this type of traffic. Also the call-duration times are analyzed to obtain the corresponding distribution. Hence, the contribution of the work is twofold. Firstly the data of a reallife VoIP network is examined for a long time-period which is independent from traffic patterns that occur within certain time periods or caused by events such as holidays and summer vacations. Secondly, the same data is differentiated based on the user behaviour in different time periods, such as working hours and weekends and the extracted results show differences between the statistical behaviour of the traffic. The presented results differ from other existing work and can be utilized to allocate resources more efficiently, to appropriately handle the network's traffic, to configure VoIP simulations in a more realistic way and to enhance the management of such networks.