PROPOSAL OF A BAYESIAN PERFORMANCE ESTIMATION METHOD AND ITS APPLICATION

Abstract

We have developed an adaptive performance estimation method called Bayesian performance estimation. This method combines conventional performance evaluation with Bayesian analysis. It, is to be used for estimating the performance of networks under operation. It uses a priori information such as the performance derived by conventional methods. It does not need detailed traffic measurement, which is not, usually performed in networks, but uses the data, from a performance monitor, which is generally used. After learning the a priori estimate, the estimator updates the performance estimate through Bayesian estimation using the monitored performance. The proposed Bayesian performance estimation method is applied to VP bandwidth control in ATM networks. We show to apply this method, that is, how to determine the functions and variances in the method. Numerical examples of VP bandwidth control demonstrate the effectiveness of the method.