One distinctive feature of ATM-based Broadband Integrated Service Digital Networks is to guarantee Quality of Service (QoS). [italic]Call [italic]admission [italic]control (CAC) is imposed to limit the number of accepted connections under certain QoS requirements. There are two major functions of CAC mechanism: bandwidth allocation which determines the amount of bandwidth required by a new connection; and performance monitoring to ensure that the QoS required by the existing connections are not affected when multiplexed together with this new connection. The major difficulty in CAC is either to estimate the required bandwidth to support a target QoS or to monitor the actual QoS experienced by connections within a short time period. If the system size (e.g., buffer sizes, number of conne...[ Read more ]

One distinctive feature of ATM-based Broadband Integrated Service Digital Networks is to guarantee Quality of Service (QoS). [italic]Call [italic]admission [italic]control (CAC) is imposed to limit the number of accepted connections under certain QoS requirements. There are two major functions of CAC mechanism: bandwidth allocation which determines the amount of bandwidth required by a new connection; and performance monitoring to ensure that the QoS required by the existing connections are not affected when multiplexed together with this new connection. The major difficulty in CAC is either to estimate the required bandwidth to support a target QoS or to monitor the actual QoS experienced by connections within a short time period. If the system size (e.g., buffer sizes, number of connections, etc.) is large or the cell loss ratio (CLR) is small (e.g., l0^-9), the time required to compute or measure the CLR can be prohibitively long.

In this research, down-sized systems are used to obtain quickly the corresponding CLRs by computation or measurement. Different methods are proposed to relate the statistics of cell losses of the down-sized systems to those of the original system. Based on this idea, several problems in CAC are addressed. First, a new performance monitoring mechanism to monitor stringent CLR value within a reasonable measurement period is introduced. This mechanism is used to verify the QoS conformance of existing connections in the network. Second, an estimation method to predict the CLR of a large buffer system using information from smaller-buffer systems and the asymptotic behaviour is described. Based on this method, a simple scheme to compute the required bandwidth to guarantee a specified QoS requirement is proposed. Third, due to the unpredictable nature of real-time VBR video traffic, the usual approach is to reserve resources for the worst case traffic scenario to guarantee the QoS in the CAC phase. This approach results in low utilization of network resources if the worst case scenario rarely happens. To improve the utilization, a measurement-based dynamic bandwidth allocation scheme for aggregate traffic is proposed. The performance of this scheme is discussed. Finally, the [italic]routing of calls in virtual path (VP) based ATM networks is addressed. An efficient dynamic routing algorithm based on least-loaded routing (LLR) with packing for VP-based ATM networks is proposed. The performance of the proposed algorithm is compared with other algorithms.