Parametric analysis of statistical communication nets

Abstract

The existing traffic within the branches of a communication net can often be assumed to be normally distributed random variables. A natural problem is to determine the probability that a particular flow rate between a pair of stations can be attained. If this probability is too small, it is necessary to improve the net with minimum cost. In this paper, analysis techniques on which effective synthesis procedures can be based are developed. An exact method for evaluating the flow rate probability is obtained as well as upper and lower bounds. Monte Carlo techniques are applied and the flow rate is seen to be approximately normally distributed. A method of finding the approximate mean and variance of the flow rate is given, as well as a Uniformly Most Powerful Invariant Statistical test.