Investigations of the performance of a measurement-based Connection Admission Control algorithm.

In this paper we present an algorithm for making connection admission decisions in ATM networks, using measurements made on existing connections and the declared parameters of the new connections. Our scheme makes use of the shape-function, a concept developed by Botvich and Duffield [28], which arises in the application of Large Deviation theory to queueing systems. By estimating the shape-function of the existing connections, we can make predictions about the effect on the network of accepting new connections. Using real traffic collected from a network, we compare the performance of this CAC scheme with that of the Mosquito [33] algorithm which is based on estimation of effective bandwidths. 1 Introduction Asynchronous Transfer Mode (ATM) allows for statistical multiplexing of traffic from many different applications, having widely differing characteristics. This is advantageous from the network's point of view as if affords the opportunity for dramatically increased utilisation of..