A Framework for Bandwidth Management and Call Admission Control in ATM Networks

One critical issue in tra c management of ATM networks is, to achieve the balance between two seemingly con icting goals: to guarantee the Quality of Service (QoS) for all service classes, while still allowing enough statistical sharing of bandwidth so that the network is e ciently utilized. Guaranteeing QoS requires tra c isolation, as well as allocation of enough network resource (e.g. bu er space and bandwidth) to each call. However, the statistical bandwidth sharing means the network resource should be occupied on-demand, leading to less tra c isolation and minimal resource allocation. Due to the complicated nature of this problem, no single measure can possible provide the proper solution. Rather, a comprehensive solution incorporating the schemes dealing with di erent aspects of the problem is necessary. In this thesis, we try to address the problem by proposing a framework for bandwidth management and connection admission control, into which further detailed control measures may be added later. In the rst part of this thesis, we propose and evaluate a networkwide bandwidth management architecture in which an appropriate compromise between the two con icting goals is achieved. Speci cally, the bandwidth management framework consists of a network model and a network-wide bandwidth allocation and sharing strategy. Implementation issues related to the framework are discussed. For real time applications, we obtain maximum queueing delay and queue length which are important in bu er design and VP (Virtual Path) routing. Furthermore, we propose a measurement-based CAC strategy based on the proposed architecture. We rst discuss how to obtain an accurate description (UPC parameters) of user tra c by using trace-based measurement in conjunction with on-line measurement and dynamic renegotiation. The results show that the proposed strategy is reliable and simple to implement. We then move on to examine the possibility and methodology for exploiting the e ect of statistical multiplexing in resource allocation to achieve higher network resource utilization. i