The areas of resource allocation ancl congestion control in ATM networks have been investigated. ATM networks and the guarantees given to users have been reviewed and a new model of ATM networking has been proposed. To aid the analysis of ATM network issues, performance modelling and simulation methods have been reviewed. Typical sources have been designed : a two-state Markov model for voice ; a multi-state Markov one layer variable bit rate video source model ; an empirical file transfer data source model ; and some basic network elements. The models have been verified and validated on a discrete event simulator. It was shown that there are problems when using ATM over satellite links. A model for the noise analysed from real satellite links was developed. Based on this model a new more efficient protocol for assembling ATM cells was proposed and simulated. Again at the cell level, the traffic that can pass the standardised conformance test and still produce the worst performance in the network was investigated. Counter to the traditional wisdom it was found that the on-off source does not always produce the worst case traffic. Users have been classified with new parameters, and it has been shown that these new classes of users can still be given guarantees without giving traffic descriptors. Adaptive user classes have been modelled mathematically. A new model for efficiency has been developed, which includes both network issues and economic issues. This new model defines congestion and also describes how to allocate resources when congested. It has been shown that this economic model coupled with the adaptive user classes allow for an increase in both network and economic efficiency simultaneously for some sample cases.