Capturing overload probability in modeling network traffic for discrete-time loss analysis

Design of network devices requires accurate models for network traffic. The main difficulty lies in determination of the traffic statistics which effectively influence the network performance, and in construction of a tractable and easy analytical model capturing these statistics. This paper demonstrates that, in order to evaluate performance of a multiplexer in a discrete-time environment, it is necessary to define a model which is able to capture the probability density function, autocovariance function, and the overload probability of the input traffic, simultaneously. The paper also provide an algorithm for the construction of an SBBP model matching these statistics