Model-based admission control adjustment in data networks

Abstract

Tuning for connection admission control (CAC) algorithms in broadband ATM networks is accomplished using an overbooking technique based on aggregate effective bandwidth, AEBW, as an approximation to required bandwidth for given levels and classes of network traffic. An overbooking gain factor, &agr; is developed and used for a succeeding time interval using a model-based approach illustratively using estimates of mean and variance for selected measured parameters of the aggregate cell process for the current state of connections. Well-known Kalman filtering techniques are advantageously applied to optimally estimate the mean and variance values. It also proves useful to include longer term statistics in the mean and variance calculation and to modulate the aggressiveness of overbooking based on actual system performance measures (e.g. cell losses).