Application of Erlang B Model for Estimation of Performance Parameters in Modern IP Network Design

Abstract

One of the most commonly used techniques for communications network design and evaluation is the
analytic one. Unfortunately, it becomes inapplicable, due to the size and complexity of models. Analytical network design
techniques are usually specific to circuit-switched networks. Network design techniques specific to packet-switched
networks usually rely on network simulations. One class of problems for which simulation is rather unsuitable is those
involving the estimation of probabilities of rare events. Such events as packet loss probabilities are typically designed to
have very low values to guarantee a good quality of service. This implies simulating the system for a long time without
the event occurring even once. Fortunately, recently, the Erlang B formula has been modified for IP networks. Therefore,
in this paper, the Erlang B model is used for estimation of packet loss probability and bandwidth utilization for the core
transmission links in the process of a large-scale IP backbone network design. Specifically, the paper describes the
performance analysis of the IP backbone network topology designed for Benue State in Nigeria within the framework of
a doctoral thesis. The entire scope of the paper includes performance analysis and comparison of the results between the
Erlang B and simulation models of the IP backbone network. The results of the analysis of the two models compare
favourably and further confirms the issue of rare packet loss probability associated with simulation among other
interesting outcomes.