PACKET LOSS PROBABILITY ESTIMATION USING ERLANG B AND M/G/1/K MODELS IN MODERN VOIP NETWORKS

Abstract

The paper presents an approach to packet loss probability estimation in modern VoIP networks. Our proposal is based on statistical characteristics of telecommunication networks and VoIp traffic packet flows. We adapted simple Erlang B model that is widely used for classic telecommunication networks dimensioning and analyze its potential applicability to modern convergent IP networks. Furthermore we derived M/G/1/K model of aggregated data flow to study influence of queue size on packet loss probability. At last we verified the model outcome and applicability using extensive simulations using NS2 software.